poly_int: expand_ifn_atomic_compare_exchange_into_call
[official-gcc.git] / gcc / ChangeLog
blob2eea29ba5e0a349f7e665b52a6d64e0022afbd23
1 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
2             Alan Hayward  <alan.hayward@arm.com>
3             David Sherwood  <david.sherwood@arm.com>
5         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
6         that the mode size is in the set {1, 2, 4, 8, 16}.
8 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
9             Alan Hayward  <alan.hayward@arm.com>
10             David Sherwood  <david.sherwood@arm.com>
12         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
13         Use plus_constant instead of gen_rtx_PLUS.
15 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
16             Alan Hayward  <alan.hayward@arm.com>
17             David Sherwood  <david.sherwood@arm.com>
19         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
20         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
21         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
22         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
23         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
24         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
25         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
26         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
27         * config/i386/i386.c (ix86_push_rounding): ...this new function.
28         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
29         a poly_int64.
30         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
31         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
32         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
33         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
34         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
35         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
36         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
37         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
38         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
39         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
40         function.
41         * expr.c (emit_move_resolve_push): Treat the input and result
42         of PUSH_ROUNDING as a poly_int64.
43         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
44         (emit_push_insn): Likewise.
45         * lra-eliminations.c (mark_not_eliminable): Likewise.
46         * recog.c (push_operand): Likewise.
47         * reload1.c (elimination_effects): Likewise.
48         * rtlanal.c (nonzero_bits1): Likewise.
49         * calls.c (store_one_arg): Likewise.  Require the padding to be
50         known at compile time.
52 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
53             Alan Hayward  <alan.hayward@arm.com>
54             David Sherwood  <david.sherwood@arm.com>
56         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
57         Use plus_constant instead of gen_rtx_PLUS.
59 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
60             Alan Hayward  <alan.hayward@arm.com>
61             David Sherwood  <david.sherwood@arm.com>
63         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
64         rather than an int.
66 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
67             Alan Hayward  <alan.hayward@arm.com>
68             David Sherwood  <david.sherwood@arm.com>
70         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
71         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
72         via stack temporaries.  Treat the mode size as polynomial too.
74 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
75             Alan Hayward  <alan.hayward@arm.com>
76             David Sherwood  <david.sherwood@arm.com>
78         * expr.c (expand_expr_real_2): When handling conversions involving
79         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
80         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
81         as a poly_uint64 too.
83 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
84             Alan Hayward  <alan.hayward@arm.com>
85             David Sherwood  <david.sherwood@arm.com>
87         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
89 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
90             Alan Hayward  <alan.hayward@arm.com>
91             David Sherwood  <david.sherwood@arm.com>
93         * combine.c (can_change_dest_mode): Handle polynomial
94         REGMODE_NATURAL_SIZE.
95         * expmed.c (store_bit_field_1): Likewise.
96         * expr.c (store_constructor): Likewise.
97         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
98         and polynomial REGMODE_NATURAL_SIZE.
99         (gen_lowpart_common): Likewise.
100         * reginfo.c (record_subregs_of_mode): Likewise.
101         * rtlanal.c (read_modify_subreg_p): Likewise.
103 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
104             Alan Hayward  <alan.hayward@arm.com>
105             David Sherwood  <david.sherwood@arm.com>
107         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
108         numbers of elements.
110 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
111             Alan Hayward  <alan.hayward@arm.com>
112             David Sherwood  <david.sherwood@arm.com>
114         * match.pd: Cope with polynomial numbers of vector elements.
116 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
117             Alan Hayward  <alan.hayward@arm.com>
118             David Sherwood  <david.sherwood@arm.com>
120         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
121         in a POINTER_PLUS_EXPR.
123 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
124             Alan Hayward  <alan.hayward@arm.com>
125             David Sherwood  <david.sherwood@arm.com>
127         * omp-simd-clone.c (simd_clone_subparts): New function.
128         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
129         (ipa_simd_modify_function_body): Likewise.
131 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
132             Alan Hayward  <alan.hayward@arm.com>
133             David Sherwood  <david.sherwood@arm.com>
135         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
136         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
137         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
138         (expand_vector_condition, vector_element): Likewise.
139         (subparts_gt): New function.
140         (get_compute_type): Use subparts_gt.
141         (count_type_subparts): Delete.
142         (expand_vector_operations_1): Use subparts_gt instead of
143         count_type_subparts.
145 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
146             Alan Hayward  <alan.hayward@arm.com>
147             David Sherwood  <david.sherwood@arm.com>
149         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
150         (vect_compile_time_alias): ...this new function.  Do the calculation
151         on poly_ints rather than trees.
152         (vect_prune_runtime_alias_test_list): Update call accordingly.
154 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
155             Alan Hayward  <alan.hayward@arm.com>
156             David Sherwood  <david.sherwood@arm.com>
158         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
159         numbers of units.
160         (vect_schedule_slp_instance): Likewise.
162 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
163             Alan Hayward  <alan.hayward@arm.com>
164             David Sherwood  <david.sherwood@arm.com>
166         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
167         constant and extern definitions for variable-length vectors.
168         (vect_get_constant_vectors): Note that the number of units
169         is known to be constant.
171 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
172             Alan Hayward  <alan.hayward@arm.com>
173             David Sherwood  <david.sherwood@arm.com>
175         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
176         of units as polynomial.  Choose between WIDE and NARROW based
177         on multiple_p.
179 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
180             Alan Hayward  <alan.hayward@arm.com>
181             David Sherwood  <david.sherwood@arm.com>
183         * tree-vect-stmts.c (simd_clone_subparts): New function.
184         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
186 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
187             Alan Hayward  <alan.hayward@arm.com>
188             David Sherwood  <david.sherwood@arm.com>
190         * tree-vect-stmts.c (vectorizable_call): Treat the number of
191         vectors as polynomial.  Use build_index_vector for
192         IFN_GOMP_SIMD_LANE.
194 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
195             Alan Hayward  <alan.hayward@arm.com>
196             David Sherwood  <david.sherwood@arm.com>
198         * tree-vect-stmts.c (get_load_store_type): Treat the number of
199         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
200         for variable-length vectors.
201         (vectorizable_mask_load_store): Treat the number of units as
202         polynomial, asserting that it is constant if the condition has
203         already been enforced.
204         (vectorizable_store, vectorizable_load): Likewise.
206 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
207             Alan Hayward  <alan.hayward@arm.com>
208             David Sherwood  <david.sherwood@arm.com>
210         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
211         of units as polynomial.  Punt if we can't tell at compile time
212         which vector contains the final result.
214 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
215             Alan Hayward  <alan.hayward@arm.com>
216             David Sherwood  <david.sherwood@arm.com>
218         * tree-vect-loop.c (vectorizable_induction): Treat the number
219         of units as polynomial.  Punt on SLP inductions.  Use an integer
220         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
221         cast of such a series for variable-length floating-point
222         reductions.
224 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
225             Alan Hayward  <alan.hayward@arm.com>
226             David Sherwood  <david.sherwood@arm.com>
228         * tree.h (build_index_vector): Declare.
229         * tree.c (build_index_vector): New function.
230         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
231         of units as polynomial, forcibly converting it to a constant if
232         vectorizable_reduction has already enforced the condition.
233         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
234         to create a {1,2,3,...} vector.
235         (vectorizable_reduction): Treat the number of units as polynomial.
236         Choose vectype_in based on the largest scalar element size rather
237         than the smallest number of units.  Enforce the restrictions
238         relied on above.
240 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
241             Alan Hayward  <alan.hayward@arm.com>
242             David Sherwood  <david.sherwood@arm.com>
244         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
245         number of units as polynomial.
247 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
248             Alan Hayward  <alan.hayward@arm.com>
249             David Sherwood  <david.sherwood@arm.com>
251         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
252         * target.def (autovectorize_vector_sizes): Return the vector sizes
253         by pointer, using vector_sizes rather than a bitmask.
254         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
255         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
256         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
257         Likewise.
258         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
259         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
260         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
261         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
262         * omp-general.c (omp_max_vf): Likewise.
263         * omp-low.c (omp_clause_aligned_alignment): Likewise.
264         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
265         * tree-vect-loop.c (vect_analyze_loop): Likewise.
266         * tree-vect-slp.c (vect_slp_bb): Likewise.
267         * doc/tm.texi: Regenerate.
268         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
269         to a poly_uint64.
270         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
271         the vector size as a poly_uint64 rather than an unsigned int.
272         (current_vector_size): Change from an unsigned int to a poly_uint64.
273         (get_vectype_for_scalar_type): Update accordingly.
274         * tree.h (build_truth_vector_type): Take the size and number of
275         units as a poly_uint64 rather than an unsigned int.
276         (build_vector_type): Add a temporary overload that takes
277         the number of units as a poly_uint64 rather than an unsigned int.
278         * tree.c (make_vector_type): Likewise.
279         (build_truth_vector_type): Take the number of units as a poly_uint64
280         rather than an unsigned int.
282 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
283             Alan Hayward  <alan.hayward@arm.com>
284             David Sherwood  <david.sherwood@arm.com>
286         * target.def (get_mask_mode): Take the number of units and length
287         as poly_uint64s rather than unsigned ints.
288         * targhooks.h (default_get_mask_mode): Update accordingly.
289         * targhooks.c (default_get_mask_mode): Likewise.
290         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
291         * doc/tm.texi: Regenerate.
293 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
294             Alan Hayward  <alan.hayward@arm.com>
295             David Sherwood  <david.sherwood@arm.com>
297         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
298         * omp-general.c (omp_max_vf): Likewise.
299         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
300         (expand_omp_simd): Handle polynomial safelen.
301         * omp-low.c (omplow_simd_context): Add a default constructor.
302         (omplow_simd_context::max_vf): Change from int to poly_uint64.
303         (lower_rec_simd_input_clauses): Update accordingly.
304         (lower_rec_input_clauses): Likewise.
306 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
307             Alan Hayward  <alan.hayward@arm.com>
308             David Sherwood  <david.sherwood@arm.com>
310         * tree-vectorizer.h (vect_nunits_for_cost): New function.
311         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
312         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
313         (vect_analyze_slp_cost): Likewise.
314         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
315         (vect_model_load_cost): Likewise.
317 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
318             Alan Hayward  <alan.hayward@arm.com>
319             David Sherwood  <david.sherwood@arm.com>
321         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
322         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
323         from an unsigned int * to a poly_uint64_pod *.
324         (calculate_unrolling_factor): New function.
325         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
327 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
328             Alan Hayward  <alan.hayward@arm.com>
329             David Sherwood  <david.sherwood@arm.com>
331         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
332         from an unsigned int to a poly_uint64.
333         (_loop_vec_info::slp_unrolling_factor): Likewise.
334         (_loop_vec_info::vectorization_factor): Change from an int
335         to a poly_uint64.
336         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
337         (vect_get_num_vectors): New function.
338         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
339         (vect_get_num_copies): Use vect_get_num_vectors.
340         (vect_analyze_data_ref_dependences): Change max_vf from an int *
341         to an unsigned int *.
342         (vect_analyze_data_refs): Change min_vf from an int * to a
343         poly_uint64 *.
344         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
345         than an unsigned HOST_WIDE_INT.
346         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
347         (vect_analyze_data_ref_dependence): Change max_vf from an int *
348         to an unsigned int *.
349         (vect_analyze_data_ref_dependences): Likewise.
350         (vect_compute_data_ref_alignment): Handle polynomial vf.
351         (vect_enhance_data_refs_alignment): Likewise.
352         (vect_prune_runtime_alias_test_list): Likewise.
353         (vect_shift_permute_load_chain): Likewise.
354         (vect_supportable_dr_alignment): Likewise.
355         (dependence_distance_ge_vf): Take the vectorization factor as a
356         poly_uint64 rather than an unsigned HOST_WIDE_INT.
357         (vect_analyze_data_refs): Change min_vf from an int * to a
358         poly_uint64 *.
359         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
360         vfm1 as a poly_uint64 rather than an int.  Make the same change
361         for the returned bound_scalar.
362         (vect_gen_vector_loop_niters): Handle polynomial vf.
363         (vect_do_peeling): Likewise.  Update call to
364         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
365         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
366         be constant.
367         * tree-vect-loop.c (vect_determine_vectorization_factor)
368         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
369         (vect_get_known_peeling_cost): Likewise.
370         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
371         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
372         (vect_transform_loop): Likewise.  Use the lowest possible VF when
373         updating the upper bounds of the loop.
374         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
375         rather than an int.
376         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
377         polynomial unroll factors.
378         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
379         (vect_make_slp_decision): Likewise.
380         (vect_supported_load_permutation_p): Likewise, and polynomial
381         vf too.
382         (vect_analyze_slp_cost): Handle polynomial vf.
383         (vect_slp_analyze_node_operations): Likewise.
384         (vect_slp_analyze_bb_1): Likewise.
385         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
386         than an unsigned HOST_WIDE_INT.
387         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
388         (vectorizable_load): Handle polynomial vf.
389         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
390         a poly_uint64.
391         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
393 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
394             Alan Hayward  <alan.hayward@arm.com>
395             David Sherwood  <david.sherwood@arm.com>
397         * match.pd: Handle bit operations involving three constants
398         and try to fold one pair.
400 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
402         * tree-vect-loop-manip.c: Include gimple-fold.h.
403         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
404         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
405         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
406         Add a path that uses a step of VF instead of 1, but disable it
407         for now.
408         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
409         and niters_no_overflow parameters.  Update calls to
410         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
411         Create a new SSA name if the latter choses to use a ste other
412         than zero, and return it via niters_vector_mult_vf_var.
413         * tree-vect-loop.c (vect_transform_loop): Update calls to
414         vect_do_peeling, vect_gen_vector_loop_niters and
415         slpeel_make_loop_iterate_ntimes.
416         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
417         (vect_gen_vector_loop_niters): Update declarations after above changes.
419 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
421         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
422         128-bit round to integer instructions.
423         (ceil<mode>2): Likewise.
424         (btrunc<mode>2): Likewise.
425         (round<mode>2): Likewise.
427 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
429         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
430         unaligned VSX load/store on P8/P9.
431         (expand_block_clear): Allow the use of unaligned VSX
432         load/store on P8/P9.
434 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
436         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
437         New function.
438         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
439         swap associated with both a load and a store.
441 2018-01-02  Andrew Waterman  <andrew@sifive.com>
443         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
444         * config/riscv/riscv.md (clear_cache): Use it.
446 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
448         * web.c: Remove out-of-date comment.
450 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
452         * expr.c (fixup_args_size_notes): Check that any existing
453         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
454         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
455         (emit_single_push_insn): ...here.
457 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
459         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
460         (const_vector_encoded_nelts): New function.
461         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
462         (const_vector_int_elt, const_vector_elt): Declare.
463         * emit-rtl.c (const_vector_int_elt_1): New function.
464         (const_vector_elt): Likewise.
465         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
466         of CONST_VECTOR_ELT.
468 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
470         * expr.c: Include rtx-vector-builder.h.
471         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
472         directly on the tree encoding.
473         (const_vector_from_tree): Likewise.
474         * optabs.c: Include rtx-vector-builder.h.
475         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
476         sequence of "u" values.
477         * vec-perm-indices.c: Include rtx-vector-builder.h.
478         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
479         directly on the vec_perm_indices encoding.
481 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
483         * doc/rtl.texi (const_vector): Describe new encoding scheme.
484         * Makefile.in (OBJS): Add rtx-vector-builder.o.
485         * rtx-vector-builder.h: New file.
486         * rtx-vector-builder.c: Likewise.
487         * rtl.h (rtx_def::u2): Add a const_vector field.
488         (CONST_VECTOR_NPATTERNS): New macro.
489         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
490         (CONST_VECTOR_DUPLICATE_P): Likewise.
491         (CONST_VECTOR_STEPPED_P): Likewise.
492         (CONST_VECTOR_ENCODED_ELT): Likewise.
493         (const_vec_duplicate_p): Check for a duplicated vector encoding.
494         (unwrap_const_vec_duplicate): Likewise.
495         (const_vec_series_p): Check for a non-duplicated vector encoding.
496         Say that the function only returns true for integer vectors.
497         * emit-rtl.c: Include rtx-vector-builder.h.
498         (gen_const_vec_duplicate_1): Delete.
499         (gen_const_vector): Call gen_const_vec_duplicate instead of
500         gen_const_vec_duplicate_1.
501         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
502         (gen_const_vec_duplicate): Use rtx_vector_builder.
503         (gen_const_vec_series): Likewise.
504         (gen_rtx_CONST_VECTOR): Likewise.
505         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
506         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
507         Build a new vector rather than modifying a CONST_VECTOR in-place.
508         (handle_special_swappables): Update call accordingly.
509         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
510         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
511         Build a new vector rather than modifying a CONST_VECTOR in-place.
512         (handle_special_swappables): Update call accordingly.
514 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
516         * simplify-rtx.c (simplify_const_binary_operation): Use
517         CONST_VECTOR_ELT instead of XVECEXP.
519 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
521         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
522         the selector elements to be different from the data elements
523         if the selector is a VECTOR_CST.
524         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
525         ssizetype for the selector.
527 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
529         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
530         before testing each element individually.
531         * tree-vect-generic.c (lower_vec_perm): Likewise.
533 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
535         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
536         * selftest-run-tests.c (selftest::run_tests): Call it.
537         * vector-builder.h (vector_builder::operator ==): New function.
538         (vector_builder::operator !=): Likewise.
539         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
540         (vec_perm_indices::all_from_input_p): New function.
541         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
542         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
543         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
544         instead of reading the VECTOR_CST directly.  Detect whether both
545         vector inputs are the same before constructing the vec_perm_indices,
546         and update the number of inputs argument accordingly.  Use the
547         utility functions added above.  Only construct sel2 if we need to.
549 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
551         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
552         the broadcast of the low byte.
553         (expand_mult_highpart): Use an explicit encoding for the permutes.
554         * optabs-query.c (can_mult_highpart_p): Likewise.
555         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
556         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
557         (vectorizable_bswap): Likewise.
558         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
559         explicit encoding for the power-of-2 permutes.
560         (vect_permute_store_chain): Likewise.
561         (vect_grouped_load_supported): Likewise.
562         (vect_permute_load_chain): Likewise.
564 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
566         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
567         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
568         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
569         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
570         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
571         (vect_gen_perm_mask_any): Likewise.
573 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
575         * int-vector-builder.h: New file.
576         * vec-perm-indices.h: Include int-vector-builder.h.
577         (vec_perm_indices): Redefine as an int_vector_builder.
578         (auto_vec_perm_indices): Delete.
579         (vec_perm_builder): Redefine as a stand-alone class.
580         (vec_perm_indices::vec_perm_indices): New function.
581         (vec_perm_indices::clamp): Likewise.
582         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
583         (vec_perm_indices::new_vector): New function.
584         (vec_perm_indices::new_expanded_vector): Update for new
585         vec_perm_indices class.
586         (vec_perm_indices::rotate_inputs): New function.
587         (vec_perm_indices::all_in_range_p): Operate directly on the
588         encoded form, without computing elided elements.
589         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
590         encoding.  Update for new vec_perm_indices class.
591         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
592         the given vec_perm_builder.
593         (expand_vec_perm_var): Update vec_perm_builder constructor.
594         (expand_mult_highpart): Use vec_perm_builder instead of
595         auto_vec_perm_indices.
596         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
597         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
598         or double series encoding as appropriate.
599         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
600         vec_perm_indices instead of auto_vec_perm_indices.
601         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
602         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
603         (vect_permute_store_chain): Likewise.
604         (vect_grouped_load_supported): Likewise.
605         (vect_permute_load_chain): Likewise.
606         (vect_shift_permute_load_chain): Likewise.
607         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
608         (vect_transform_slp_perm_load): Likewise.
609         (vect_schedule_slp_instance): Likewise.
610         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
611         (vectorizable_mask_load_store): Likewise.
612         (vectorizable_bswap): Likewise.
613         (vectorizable_store): Likewise.
614         (vectorizable_load): Likewise.
615         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
616         vec_perm_indices instead of auto_vec_perm_indices.  Use
617         tree_to_vec_perm_builder to read the vector from a tree.
618         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
619         vec_perm_builder instead of a vec_perm_indices.
620         (have_whole_vector_shift): Use vec_perm_builder and
621         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
622         truncation to calc_vec_perm_mask_for_shift.
623         (vect_create_epilog_for_reduction): Likewise.
624         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
625         from auto_vec_perm_indices to vec_perm_indices.
626         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
627         instead of changing individual elements.
628         (aarch64_vectorize_vec_perm_const): Use new_vector to install
629         the vector in d.perm.
630         * config/arm/arm.c (expand_vec_perm_d::perm): Change
631         from auto_vec_perm_indices to vec_perm_indices.
632         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
633         instead of changing individual elements.
634         (arm_vectorize_vec_perm_const): Use new_vector to install
635         the vector in d.perm.
636         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
637         Update vec_perm_builder constructor.
638         (rs6000_expand_interleave): Likewise.
639         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
640         (rs6000_expand_interleave): Likewise.
642 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
644         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
645         to qimode could truncate the indices.
646         * optabs.c (expand_vec_perm_var): Likewise.
648 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
650         * Makefile.in (OBJS): Add vec-perm-indices.o.
651         * vec-perm-indices.h: New file.
652         * vec-perm-indices.c: Likewise.
653         * target.h (vec_perm_indices): Replace with a forward class
654         declaration.
655         (auto_vec_perm_indices): Move to vec-perm-indices.h.
656         * optabs.h: Include vec-perm-indices.h.
657         (expand_vec_perm): Delete.
658         (selector_fits_mode_p, expand_vec_perm_var): Declare.
659         (expand_vec_perm_const): Declare.
660         * target.def (vec_perm_const_ok): Replace with...
661         (vec_perm_const): ...this new hook.
662         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
663         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
664         * doc/tm.texi: Regenerate.
665         * optabs.def (vec_perm_const): Delete.
666         * doc/md.texi (vec_perm_const): Likewise.
667         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
668         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
669         expand_vec_perm for constant permutation vectors.  Assert that
670         the mode of variable permutation vectors is the integer equivalent
671         of the mode that is being permuted.
672         * optabs-query.h (selector_fits_mode_p): Declare.
673         * optabs-query.c: Include vec-perm-indices.h.
674         (selector_fits_mode_p): New function.
675         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
676         is defined, instead of checking whether the vec_perm_const_optab
677         exists.  Use targetm.vectorize.vec_perm_const instead of
678         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
679         fit in the vector mode before using a variable permute.
680         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
681         vec_perm_indices instead of an rtx.
682         (expand_vec_perm): Replace with...
683         (expand_vec_perm_const): ...this new function.  Take the selector
684         as a vec_perm_indices rather than an rtx.  Also take the mode of
685         the selector.  Update call to shift_amt_for_vec_perm_mask.
686         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
687         Use vec_perm_indices::new_expanded_vector to expand the original
688         selector into bytes.  Check whether the indices fit in the vector
689         mode before using a variable permute.
690         (expand_vec_perm_var): Make global.
691         (expand_mult_highpart): Use expand_vec_perm_const.
692         * fold-const.c: Includes vec-perm-indices.h.
693         * tree-ssa-forwprop.c: Likewise.
694         * tree-vect-data-refs.c: Likewise.
695         * tree-vect-generic.c: Likewise.
696         * tree-vect-loop.c: Likewise.
697         * tree-vect-slp.c: Likewise.
698         * tree-vect-stmts.c: Likewise.
699         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
700         Delete.
701         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
702         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
703         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
704         (aarch64_vectorize_vec_perm_const): ...this new function.
705         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
706         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
707         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
708         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
709         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
710         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
711         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
712         into...
713         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
714         check for NEON modes.
715         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
716         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
717         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
718         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
719         into...
720         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
721         the old VEC_PERM_CONST conditions.
722         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
723         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
724         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
725         (ia64_vectorize_vec_perm_const_ok): Merge into...
726         (ia64_vectorize_vec_perm_const): ...this new function.
727         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
728         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
729         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
730         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
731         * config/mips/mips.c (mips_expand_vec_perm_const)
732         (mips_vectorize_vec_perm_const_ok): Merge into...
733         (mips_vectorize_vec_perm_const): ...this new function.
734         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
735         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
736         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
737         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
738         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
739         (rs6000_expand_vec_perm_const): Delete.
740         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
741         Delete.
742         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
743         (altivec_expand_vec_perm_const_le): Take each operand individually.
744         Operate on constant selectors rather than rtxes.
745         (altivec_expand_vec_perm_const): Likewise.  Update call to
746         altivec_expand_vec_perm_const_le.
747         (rs6000_expand_vec_perm_const): Delete.
748         (rs6000_vectorize_vec_perm_const_ok): Delete.
749         (rs6000_vectorize_vec_perm_const): New function.
750         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
751         an element count and rtx array.
752         (rs6000_expand_extract_even): Update call accordingly.
753         (rs6000_expand_interleave): Likewise.
754         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
755         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
756         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
757         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
758         (rs6000_expand_vec_perm_const): Delete.
759         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
760         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
761         (altivec_expand_vec_perm_const_le): Take each operand individually.
762         Operate on constant selectors rather than rtxes.
763         (altivec_expand_vec_perm_const): Likewise.  Update call to
764         altivec_expand_vec_perm_const_le.
765         (rs6000_expand_vec_perm_const): Delete.
766         (rs6000_vectorize_vec_perm_const_ok): Delete.
767         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
768         reference to the SPE evmerge intructions.
769         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
770         an element count and rtx array.
771         (rs6000_expand_extract_even): Update call accordingly.
772         (rs6000_expand_interleave): Likewise.
773         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
774         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
775         new function.
776         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
778 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
780         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
781         vector mode and that that mode matches the mode of the data
782         being permuted.
783         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
784         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
785         directly using expand_vec_perm_1 when forcing selectors into
786         registers.
787         (expand_vec_perm_var): New function, split out from expand_vec_perm.
789 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
791         * optabs-query.h (can_vec_perm_p): Delete.
792         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
793         * optabs-query.c (can_vec_perm_p): Split into...
794         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
795         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
796         particular selector is valid.
797         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
798         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
799         (vect_grouped_load_supported): Likewise.
800         (vect_shift_permute_load_chain): Likewise.
801         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
802         (vect_transform_slp_perm_load): Likewise.
803         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
804         (vectorizable_bswap): Likewise.
805         (vect_gen_perm_mask_checked): Likewise.
806         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
807         implementations of variable permutation vectors into account
808         when deciding which selector to use.
809         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
810         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
811         with a false third argument.
812         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
813         to test whether the constant selector is valid and can_vec_perm_var_p
814         to test whether a variable selector is valid.
816 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
818         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
819         * optabs-query.c (can_vec_perm_p): Likewise.
820         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
821         instead of vec_perm_indices.
822         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
823         (vect_gen_perm_mask_checked): Likewise,
824         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
825         (vect_gen_perm_mask_checked): Likewise,
827 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
829         * optabs-query.h (qimode_for_vec_perm): Declare.
830         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
831         (qimode_for_vec_perm): ...this new function.
832         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
834 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
836         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
837         does not have a conditional at the top.
839 2018-01-02  Richard Biener  <rguenther@suse.de>
841         * ipa-inline.c (big_speedup_p): Fix expression.
843 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
845         PR target/81616
846         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
847         for generic 4->6.
849 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
851         PR target/81616
852         Generic tuning.
853         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
854         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
855         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
856         cond_taken_branch_cost 3->4.
858 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
860         PR tree-optimization/83581
861         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
862         TODO_cleanup_cfg if any changes have been made.
864         PR middle-end/83608
865         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
866         convert_modes if target mode has the right side, but different mode
867         class.
869         PR middle-end/83609
870         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
871         last argument when extracting from CONCAT.  If either from_real or
872         from_imag is NULL, use expansion through memory.  If result is not
873         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
874         the parts directly to inner mode, if even that fails, use expansion
875         through memory.
877         PR middle-end/83623
878         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
879         check for bswap in mode rather than HImode and use that in expand_unop
880         too.
882 2017-12-31  Jakub Jelinek  <jakub@redhat.com>
884         PR target/83536
885         * config/i386/i386.c (ix86_attribute_table): Remove excess
886         initializer for "shared" attribute.
888 2017-12-30  Tom de Vries  <tom@codesourcery.com>
890         PR libgomp/83046
891         * omp-expand.c (expand_omp_target): If in_lto_p, mark offload_funcs with
892         DECL_PRESERVE_P.
893         * lto-streamer-out.c (prune_offload_funcs): New function.  Remove
894         offload_funcs entries that no longer have a corresponding cgraph_node.
895         Mark the remaining ones as DECL_PRESERVE_P.
896         (output_lto): Call prune_offload_funcs.
898 2017-12-30  Jakub Jelinek  <jakub@redhat.com>
900         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
901         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>,
902         vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
903         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
904         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
905         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1, vpdpbusd_<mode>,
906         vpdpbusd_<mode>_mask, vpdpbusd_<mode>_maskz, vpdpbusd_<mode>_maskz_1,
907         vpdpbusds_<mode>, vpdpbusds_<mode>_mask, vpdpbusds_<mode>_maskz,
908         vpdpbusds_<mode>_maskz_1, vpdpwssd_<mode>, vpdpwssd_<mode>_mask,
909         vpdpwssd_<mode>_maskz, vpdpwssd_<mode>_maskz_1, vpdpwssds_<mode>,
910         vpdpwssds_<mode>_mask, vpdpwssds_<mode>_maskz,
911         vpdpwssds_<mode>_maskz_1, vaesdec_<mode>, vaesdeclast_<mode>,
912         vaesenc_<mode>, vpclmulqdq_<mode>,
913         avx512vl_vpshufbitqmb<mode><mask_scalar_merge_name>): Formatting fixes.
915 2017-12-28  Michael Meissner  <meissner@linux.vnet.ibm.com>
917         * builtins.def: (_Float<N> and _Float<N>X BUILT_IN_CEIL): Add
918         _Float<N> and _Float<N>X variants for rounding built-in
919         functions.
920         (_Float<N> and _Float<N>X BUILT_IN_FLOOR): Likewise.
921         (_Float<N> and _Float<N>X BUILT_IN_NEARBYINT): Likewise.
922         (_Float<N> and _Float<N>X BUILT_IN_RINT): Likewise.
923         (_Float<N> and _Float<N>X BUILT_IN_ROUND): Likewise.
924         (_Float<N> and _Float<N>X BUILT_IN_TRUNC): Likewise.
925         * builtins.c (mathfn_built_in_2): Likewise.
926         * internal-fn.def (CEIL): Likewise.
927         (FLOOR): Likewise.
928         (NEARBYINT): Likewise.
929         (RINT): Likewise.
930         (ROUND): Likewise.
931         (TRUNC): Likewise.
932         * convert.c (convert_to_integer_1): Likewise.
933         * fold-const.c (tree_call_nonnegative_warnv_p): Likewise.
934         (integer_valued_real_call_p): Likewise.
935         * fold-const-call.c (fold_const_call_ss): Likewise.
936         * gencfn-macros.c (print_case_cfn): Change CFN and operator
937         printers to take a const char * suffix instead of a bool.
938         (print_define_operator_list): Likewise.
939         (fltall_suffixes): New list of suffixes, that include the
940         traditional suffixes as well as all of the _Float<N> and
941         _Float<N>X suffixes.
942         (main): For _Float<N> and _Float<N>X functions, emit both
943         <name>_FN and <name>_ALL variants.  The <macro>_FN variant only
944         has the _Float<N> and _Float<N>X case names or operators.  The
945         <name>_ALL variant has both the traditional and the
946         _Float<N>/_Float<N>X case names or operators.
947         * match.pd (COPYSIGN optimizations): Provide optimizations for
948         _Float<N> and _Float<N>X types where possible.
949         (MIN/MAX optimizations): Likewise.
950         (sqrt optimizations): Likewise.
951         (rounding optimizations): Likewise.
953 2017-12-28  Richard Sandiford  <richard.sandiford@linaro.org>
955         * config/powerpcspe/powerpcspe.md (ieee_128bit_negative_zero): Use
956         gen_int_mode rather than GEN_INT.
957         * config/rs6000/rs6000.md (ieee_128bit_negative_zero): Likewise.
959 2017-12-28  Richard Sandiford  <richard.sandiford@linaro.org>
961         * emit-rtl.c (gen_const_vec_series): Use valid_for_const_vector_p
962         instead of CONSTANT_P.
963         (gen_vec_series): Likewise.
964         * simplify-rtx.c (simplify_binary_operation_1): Likewise.
966 2017-12-28  Andreas Schwab  <schwab@linux-m68k.org>
968         * config/m68k/m68k.md (ashrdi3_const1, lshrdi3_const1): Add
969         CC_STATUS_INIT.
971 2017-12-27  Kugan Vivekanandarajah  <kuganv@linaro.org>
973         * config/aarch64/aarch64-simd.md (aarch64_ld1x2<VQ:mode>): New.
974         (aarch64_ld1x2<VDC:mode>): Likewise.
975         (aarch64_simd_ld1<mode>_x2): Likewise.
976         (aarch64_simd_ld1<mode>_x2): Likewise.
977         * config/aarch64/arm_neon.h (vld1_u8_x2): New.
978         (vld1_s8_x2): Likewise.
979         (vld1_u16_x2): Likewise.
980         (vld1_s16_x2): Likewise.
981         (vld1_u32_x2): Likewise.
982         (vld1_s32_x2): Likewise.
983         (vld1_u64_x2): Likewise.
984         (vld1_s64_x2): Likewise.
985         (vld1_f16_x2): Likewise.
986         (vld1_f32_x2): Likewise.
987         (vld1_f64_x2): Likewise.
988         (vld1_p8_x2): Likewise.
989         (vld1_p16_x2): Likewise.
990         (vld1_p64_x2): Likewise.
991         (vld1q_u8_x2): Likewise.
992         (vld1q_s8_x2): Likewise.
993         (vld1q_u16_x2): Likewise.
994         (vld1q_s16_x2): Likewise.
995         (vld1q_u32_x2): Likewise.
996         (vld1q_s32_x2): Likewise.
997         (vld1q_u64_x2): Likewise.
998         (vld1q_s64_x2): Likewise.
999         (vld1q_f16_x2): Likewise.
1000         (vld1q_f32_x2): Likewise.
1001         (vld1q_f64_x2): Likewise.
1002         (vld1q_p8_x2): Likewise.
1003         (vld1q_p16_x2): Likewise.
1004         (vld1q_p64_x2): Likewise.
1006 2017-12-27  Martin Liska  <mliska@suse.cz>
1008         PR tree-optimization/83552
1009         * tree-ssa-strlen.c (fold_strstr_to_strncmp): Assign result
1010         of get_string_lenth to a SSA_NAME if not a GIMPLE value.
1012 2017-12-27  Tom de Vries  <tom@codesourcery.com>
1014         * config/nvptx/nvptx.c (nvptx_option_override): Disable
1015         -gstatement-frontiers.
1017 2017-12-26  Alexander Monakov  <amonakov@ispras.ru>
1019         PR rtl-optimization/83513
1020         * sel-sched.c (sel_rank_for_schedule): Order by non-zero usefulness
1021         before priority comparison.
1023 2017-12-25  Jakub Jelinek  <jakub@redhat.com>
1025         PR target/83488
1026         * config/i386/i386.opt (-mavx512vpopcntdq, -mmavx512bitalg): Move from
1027         ix86_isa_flags2 to ix86_isa_flags.
1028         * config/i386/i386-c.c (ix86_target_macros_internal): Test
1029         OPTION_MASK_ISA_AVX512BITALG and OPTION_MASK_ISA_AVX512VPOPCNTDQ in
1030         isa_flags rather than isa_flags2.
1031         * config/i386/i386.c (ix86_target_string): Move -mavx512vpopcntdq
1032         and -mavx512bitalg from isa2_opts to isa_opts.
1033         (ix86_option_override_internal): Test OPTION_MASK_ISA_AVX512VPOPCNTDQ
1034         in x_ix86_isa_flags_explicit rather than x_ix86_isa_flags2_explicit
1035         and set it in x_ix86_isa_flags rather than x_ix86_isa_flags2.
1036         Formatting fixes.
1037         (def_builtin): Treat OPTION_MASK_ISA_AVX512BW or
1038         OPTION_MASK_ISA_AVX512F ored with another option similarly to
1039         OPTION_MASK_ISA_AVX512VL.  Even for OPTION_MASK_ISA_AVX512VL don't
1040         clear it if mask is just OPTION_MASK_ISA_AVX512VL itself.
1041         (ix86_expand_builtin): Don't handle OPTION_MASK_ISA_GFNI and
1042         OPTION_MASK_ISA_VPCLMULQDQ specially, instead handle
1043         OPTION_MASK_ISA_AVX512BW and OPTION_MASK_ISA_AVX512F that way.
1044         * config/i386/i386-builtin.def: Move AVX512VPOPCNTDQ and AVX512BITALG
1045         builtins from bdesc_args2 to bdesc_args section.
1046         (__builtin_ia32_compressstoreuqi512_mask,
1047         __builtin_ia32_compressstoreuhi512_mask,
1048         __builtin_ia32_compressstoreuqi256_mask,
1049         __builtin_ia32_expandloadqi512_mask,
1050         __builtin_ia32_expandloadqi512_maskz,
1051         __builtin_ia32_expandloadhi512_mask,
1052         __builtin_ia32_expandloadhi512_maskz,
1053         __builtin_ia32_compressqi512_mask, __builtin_ia32_compresshi512_mask,
1054         __builtin_ia32_compressqi256_mask, __builtin_ia32_expandqi512_mask,
1055         __builtin_ia32_expandqi512_maskz, __builtin_ia32_expandhi512_mask,
1056         __builtin_ia32_expandhi512_maskz, __builtin_ia32_expandqi256_mask,
1057         __builtin_ia32_expandqi256_maskz, __builtin_ia32_vpshrd_v32hi_mask,
1058         __builtin_ia32_vpshld_v32hi_mask, __builtin_ia32_vpshrdv_v32hi_mask,
1059         __builtin_ia32_vpshrdv_v32hi_maskz, __builtin_ia32_vpshldv_v32hi_mask,
1060         __builtin_ia32_vpshldv_v32hi_maskz,
1061         __builtin_ia32_vpopcountb_v64qi_mask,
1062         __builtin_ia32_vpopcountw_v32hi_mask,
1063         __builtin_ia32_vpshufbitqmb512_mask,
1064         __builtin_ia32_vpshufbitqmb256_mask): Add
1065         " | OPTION_MASK_ISA_AVX512BW".
1066         (__builtin_ia32_expandloadqi256_mask,
1067         __builtin_ia32_expandloadqi256_maskz,
1068         __builtin_ia32_vpopcountb_v32qi_mask): Add
1069         " | OPTION_MASK_ISA_AVX512VL | OPTION_MASK_ISA_AVX512BW".
1070         (__builtin_ia32_expandloadhi256_mask,
1071         __builtin_ia32_expandloadhi256_maskz,
1072         __builtin_ia32_expandloadqi128_mask,
1073         __builtin_ia32_expandloadqi128_maskz,
1074         __builtin_ia32_expandloadhi128_mask,
1075         __builtin_ia32_expandloadhi128_maskz,
1076         __builtin_ia32_vpshrd_v16hi, __builtin_ia32_vpshrd_v16hi_mask,
1077         __builtin_ia32_vpshrd_v8hi, __builtin_ia32_vpshrd_v8hi_mask,
1078         __builtin_ia32_vpshrd_v8si, __builtin_ia32_vpshrd_v8si_mask,
1079         __builtin_ia32_vpshrd_v4si, __builtin_ia32_vpshrd_v4si_mask,
1080         __builtin_ia32_vpshrd_v4di, __builtin_ia32_vpshrd_v4di_mask,
1081         __builtin_ia32_vpshrd_v2di, __builtin_ia32_vpshrd_v2di_mask,
1082         __builtin_ia32_vpshld_v16hi, __builtin_ia32_vpshld_v16hi_mask,
1083         __builtin_ia32_vpshld_v8hi, __builtin_ia32_vpshld_v8hi_mask,
1084         __builtin_ia32_vpshld_v8si, __builtin_ia32_vpshld_v8si_mask,
1085         __builtin_ia32_vpshld_v4si, __builtin_ia32_vpshld_v4si_mask,
1086         __builtin_ia32_vpshld_v4di, __builtin_ia32_vpshld_v4di_mask,
1087         __builtin_ia32_vpshld_v2di, __builtin_ia32_vpshld_v2di_mask,
1088         __builtin_ia32_vpshrdv_v16hi, __builtin_ia32_vpshrdv_v16hi_mask,
1089         __builtin_ia32_vpshrdv_v16hi_maskz, __builtin_ia32_vpshrdv_v8hi,
1090         __builtin_ia32_vpshrdv_v8hi_mask, __builtin_ia32_vpshrdv_v8hi_maskz,
1091         __builtin_ia32_vpshrdv_v8si, __builtin_ia32_vpshrdv_v8si_mask,
1092         __builtin_ia32_vpshrdv_v8si_maskz, __builtin_ia32_vpshrdv_v4si,
1093         __builtin_ia32_vpshrdv_v4si_mask, __builtin_ia32_vpshrdv_v4si_maskz,
1094         __builtin_ia32_vpshrdv_v4di, __builtin_ia32_vpshrdv_v4di_mask,
1095         __builtin_ia32_vpshrdv_v4di_maskz, __builtin_ia32_vpshrdv_v2di,
1096         __builtin_ia32_vpshrdv_v2di_mask, __builtin_ia32_vpshrdv_v2di_maskz,
1097         __builtin_ia32_vpshldv_v16hi, __builtin_ia32_vpshldv_v16hi_mask,
1098         __builtin_ia32_vpshldv_v16hi_maskz, __builtin_ia32_vpshldv_v8hi,
1099         __builtin_ia32_vpshldv_v8hi_mask, __builtin_ia32_vpshldv_v8hi_maskz,
1100         __builtin_ia32_vpshldv_v8si, __builtin_ia32_vpshldv_v8si_mask,
1101         __builtin_ia32_vpshldv_v8si_maskz, __builtin_ia32_vpshldv_v4si,
1102         __builtin_ia32_vpshldv_v4si_mask, __builtin_ia32_vpshldv_v4si_maskz,
1103         __builtin_ia32_vpshldv_v4di, __builtin_ia32_vpshldv_v4di_mask,
1104         __builtin_ia32_vpshldv_v4di_maskz, __builtin_ia32_vpshldv_v2di,
1105         __builtin_ia32_vpshldv_v2di_mask, __builtin_ia32_vpshldv_v2di_maskz,
1106         __builtin_ia32_vpopcountb_v32qi, __builtin_ia32_vpopcountb_v16qi,
1107         __builtin_ia32_vpopcountb_v16qi_mask, __builtin_ia32_vpopcountw_v16hi,
1108         __builtin_ia32_vpopcountw_v16hi_mask, __builtin_ia32_vpopcountw_v8hi,
1109         __builtin_ia32_vpopcountw_v8hi_mask): Add
1110         " | OPTION_MASK_ISA_AVX512VL".
1111         * config/i386/avx512vbmi2intrin.h (_mm512_shrdi_epi16,
1112         _mm512_shrdi_epi32, _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
1113         _mm512_shrdi_epi64, _mm512_mask_shrdi_epi64, _mm512_maskz_shrdi_epi64,
1114         _mm512_shldi_epi16, _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
1115         _mm512_maskz_shldi_epi32, _mm512_shldi_epi64, _mm512_mask_shldi_epi64,
1116         _mm512_maskz_shldi_epi64, _mm512_shrdv_epi16, _mm512_shrdv_epi32,
1117         _mm512_mask_shrdv_epi32, _mm512_maskz_shrdv_epi32, _mm512_shrdv_epi64,
1118         _mm512_mask_shrdv_epi64, _mm512_maskz_shrdv_epi64, _mm512_shldv_epi16,
1119         _mm512_shldv_epi32, _mm512_mask_shldv_epi32, _mm512_maskz_shldv_epi32,
1120         _mm512_shldv_epi64, _mm512_mask_shldv_epi64,
1121         _mm512_maskz_shldv_epi64): Don't require avx512bw for these intrinsics.
1122         * config/i386/avx512bitalgintrin.h (_mm_bitshuffle_epi64_mask,
1123         _mm_mask_bitshuffle_epi64_mask): Likewise.
1124         * common/config/i386/i386-common.c
1125         (OPTION_MASK_ISA_AVX512VPOPCNTDQ_SET,
1126         OPTION_MASK_ISA_AVX512BITALG_SET): Or in OPTION_MASK_ISA_AVX512F_SET.
1127         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
1128         OPTION_MASK_ISA_AVX512VPOPCNTDQ_UNSET and
1129         OPTION_MASK_ISA_AVX512BITALG_UNSET.
1130         (OPTION_MASK_ISA2_AVX512F_UNSET,
1131         OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Define.
1132         (ix86_handle_option): For -mno-general-regs-only, clear from
1133         ix86_isa_flags2 OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET rather than
1134         just OPTION_MASK_ISA_MPX.  For -mno-sse{,2,3,4,4.1,4.2,avx,avx2} and
1135         -mno-ssse3 clear OPTION_MASK_ISA2_AVX512F_UNSET bits from
1136         ix86_isa_flags2.  For -mno-avx512f likewise, instead of masking
1137         individually listed ISAs.  For -m{,no-}avx512{vpopcntdq,bitalg} adjust
1138         for moving from ix86_isa_flags2 to ix86_isa_flags.
1140 2017-12-23  Jakub Jelinek  <jakub@redhat.com>
1142         PR c++/83553
1143         * fold-const.c (struct contains_label_data): New type.
1144         (contains_label_1): Return non-NULL even for CASE_LABEL_EXPR, unless
1145         inside of a SWITCH_BODY seen during the walk.
1146         (contains_label_p): Use walk_tree instead of
1147         walk_tree_without_duplicates, prepare data for contains_label_1 and
1148         provide own pset.
1150 2017-12-22  Jakub Jelinek  <jakub@redhat.com>
1152         PR debug/83547
1153         * tree-iterator.c (alloc_stmt_list): Start with cleared
1154         TREE_SIDE_EFFECTS regardless whether a new STATEMENT_LIST is allocated
1155         or old one reused.
1157         PR target/83488
1158         * config/i386/avx512vnniintrin.h: Don't check for __AVX512F__ nor
1159         enable avx512f explicitly in #pragma GCC target.
1160         * config/i386/i386-builtin.def (__builtin_ia32_vpdpbusd_v8si,
1161         __builtin_ia32_vpdpbusd_v8si_mask, __builtin_ia32_vpdpbusd_v8si_maskz,
1162         __builtin_ia32_vpdpbusd_v4si, __builtin_ia32_vpdpbusd_v4si_mask,
1163         __builtin_ia32_vpdpbusd_v4si_maskz, __builtin_ia32_vpdpbusds_v8si,
1164         __builtin_ia32_vpdpbusds_v8si_mask,
1165         __builtin_ia32_vpdpbusds_v8si_maskz, __builtin_ia32_vpdpbusds_v4si,
1166         __builtin_ia32_vpdpbusds_v4si_mask,
1167         __builtin_ia32_vpdpbusds_v4si_maskz, __builtin_ia32_vpdpwssd_v8si,
1168         __builtin_ia32_vpdpwssd_v8si_mask, __builtin_ia32_vpdpwssd_v8si_maskz,
1169         __builtin_ia32_vpdpwssd_v4si, __builtin_ia32_vpdpwssd_v4si_mask,
1170         __builtin_ia32_vpdpwssd_v4si_maskz, __builtin_ia32_vpdpwssds_v8si,
1171         __builtin_ia32_vpdpwssds_v8si_mask,
1172         __builtin_ia32_vpdpwssds_v8si_maskz, __builtin_ia32_vpdpwssds_v4si,
1173         __builtin_ia32_vpdpwssds_v4si_mask,
1174         __builtin_ia32_vpdpwssds_v4si_maskz): Use
1175         OPTION_MASK_ISA_AVX512VNNI | OPTION_MASK_ISA_AVX512VL instead of
1176         just OPTION_MASK_ISA_AVX512VNNI.
1178 2017-12-22  Martin Jambor  <mjambor@suse.cz>
1180         PR lto/82027
1181         * lto-cgraph.c (output_cgraph_opt_summary_p): Also check former
1182         clones.
1184 2017-12-22  Julia Koval  <julia.koval@intel.com>
1185             Sebastian Peryt  <sebastian.peryt@intel.com>
1187         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512BITALG_SET,
1188         OPTION_MASK_ISA_AVX512BITALG_UNSET): New.
1189         (ix86_handle_option): Handle -mavx512bitalg, fix 4VNNIW formatting.
1190         * config.gcc: Add avx512vpopcntdqvlintrin.h and avx512bitalgintrin.h.
1191         * config/i386/avx512bitalgintrin.h (_mm512_popcnt_epi8, _mm512_popcnt_epi16,
1192         _mm512_mask_popcnt_epi8, _mm512_maskz_popcnt_epi8, _mm512_mask_popcnt_epi16,
1193         _mm512_maskz_popcnt_epi16, _mm512_bitshuffle_epi64_mask, _mm256_popcnt_epi8,
1194         _mm512_mask_bitshuffle_epi64_mask, _mm256_mask_popcnt_epi8, _mm_popcnt_epi8,
1195         _mm256_maskz_popcnt_epi8, _mm_bitshuffle_epi64_mask, _mm256_popcnt_epi16,
1196         _mm_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
1197         _mm256_mask_bitshuffle_epi64_mask, _mm_popcnt_epi16, _mm_maskz_popcnt_epi8,
1198         _mm256_mask_popcnt_epi16, _mm256_maskz_popcnt_epi16, _mm_mask_popcnt_epi8,
1199         _mm_mask_popcnt_epi16, _mm_maskz_popcnt_epi16): New intrinsics.
1200         * config/i386/avx512vpopcntdqvlintrin.h (_mm_popcnt_epi32, _mm_popcnt_epi64,
1201         _mm_mask_popcnt_epi32, _mm_maskz_popcnt_epi32, _mm256_popcnt_epi32,
1202         _mm256_mask_popcnt_epi32, _mm256_maskz_popcnt_epi32, _mm_mask_popcnt_epi64,
1203         _mm_maskz_popcnt_epi64, _mm256_popcnt_epi64, _mm256_mask_popcnt_epi64,
1204         _mm256_maskz_popcnt_epi64): New intrinsics.
1205         * config/i386/cpuid.h (bit_AVX512BITALG): New bit.
1206         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mavx512bitalg.
1207         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI, V64QI_FTYPE_V64QI,
1208         V4DI_FTYPE_V4DI, UHI_FTYPE_V2DI_V2DI_UHI, USI_FTYPE_V4DI_V4DI_USI,
1209         V4SI_FTYPE_V4SI_V4SI_UHI, V8SI_FTYPE_V8SI_V8SI_UHI): New types.
1210         * config/i386/i386-builtin.def (__builtin_ia32_vpopcountq_v4di,
1211         __builtin_ia32_vpopcountq_v4di_mask, __builtin_ia32_vpopcountq_v2di,
1212         __builtin_ia32_vpopcountq_v2di_mask, __builtin_ia32_vpopcountd_v4si,
1213         __builtin_ia32_vpopcountd_v4si_mask, __builtin_ia32_vpopcountd_v8si,
1214         __builtin_ia32_vpopcountd_v8si_mask, __builtin_ia32_vpopcountb_v64qi,
1215         __builtin_ia32_vpopcountb_v64qi_mask, __builtin_ia32_vpopcountb_v32qi,
1216         __builtin_ia32_vpopcountb_v32qi_mask, __builtin_ia32_vpopcountb_v16qi,
1217         __builtin_ia32_vpopcountb_v16qi_mask, __builtin_ia32_vpopcountw_v32hi,
1218         __builtin_ia32_vpopcountw_v32hi_mask, __builtin_ia32_vpopcountw_v16hi,
1219         __builtin_ia32_vpopcountw_v16hi_mask, __builtin_ia32_vpopcountw_v8hi,
1220         __builtin_ia32_vpopcountw_v8hi_mask, __builtin_ia32_vpshufbitqmb128_mask,
1221         __builtin_ia32_vpshufbitqmb256_mask,
1222         __builtin_ia32_vpshufbitqmb512_mask): New builtins.
1223         * config/i386/i386-c.c (__AVX512BITALG__): New.
1224         * config/i386/i386.c (isa2_opts): Add -mavx512bitalg.
1225         (ix86_valid_target_attribute_inner_p): Ditto.
1226         (ix86_expand_args_builtin): Handle new types.
1227         * config/i386/i386.h (TARGET_AVX512BITALG, TARGET_AVX512BITALG_P): New.
1228         * config/i386/i386.opt: Add -mavx512bitalg.
1229         * config/i386/immintrin.h: Add avx512vpopcntdqvlintrin.h and
1230         avx512bitalgintrin.h.
1231         * config/i386/sse.md (VI48_AVX512VLBW): New iterator.
1232         (vpopcount<mode><mask_name>): Add more types.
1233         (avx512vl_vpshufbitqmb<mode><mask_scalar_merge_name>): New.
1234         * doc/invoke.texi: Add -mavx512bitalg and -mavx512vpopcntdq.
1236 2017-12-22  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
1238         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VNNI_SET):
1239         Or in OPTION_MASK_ISA_AVX512F_SET.
1240         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
1241         OPTION_MASK_ISA_AVX512VNNI_UNSET.
1242         (ix86_handle_option): Adjust for
1243         OPTION_MASK_ISA_AVX512VNNI_*SET being in ix86_isa_flags.
1244         * config/i386/i386-builtin.def: Move VNNI builtins from ARGS2
1245         section to ARGS.
1246         * config/i386/i386-c.c: Check for OPTION_MASK_ISA_AVX512VNNI in
1247         isa_flag instead of isa_flag2.
1248         * config/i386/i386.c (ix86_target_string): Move -mavx512vnni from
1249         isa_opts2 to isa_opts.
1250         * config/i386/i386.opt (mavx512vnni): Move from ix86_isa_flags2
1251         to ix86_isa_flags.
1253 2017-12-22  Mike Stump  <mikestump@comcast.net>
1254             Eric Botcazou  <ebotcazou@adacore.com>
1256         * doc/extend.texi (Loop-Specific Pragmas): Document pragma GCC unroll.
1258 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1260         PR middle-end/83487
1261         * config/i386/i386.c (ix86_function_arg_boundary): Return
1262         PARM_BOUNDARY for TYPE_EMPTY_P types.
1264         PR c/83448
1265         * gimple-ssa-sprintf.c (maybe_warn): Don't call set_caret_index
1266         if navail is >= dir.len.
1268 2017-12-21  Steve Ellcey  <sellcey@cavium.com>
1270         * config/aarch64/t-aarch64-linux (MULTILIB_OSDIRNAMES): Fix
1271         triplet for ilp32.
1273 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1275         PR rtl-optimization/80747
1276         PR rtl-optimization/83512
1277         * cfgrtl.c (force_nonfallthru_and_redirect): When splitting
1278         succ edge from ENTRY, copy partition from e->dest to the newly
1279         created bb.
1280         * bb-reorder.c (reorder_basic_blocks_simple): If last_tail is
1281         ENTRY, use BB_PARTITION of its successor block as current_partition.
1282         Don't copy partition when splitting succ edge from ENTRY.
1284         PR tree-optimization/83523
1285         * tree-ssa-math-opts.c (is_widening_mult_p): Return false if
1286         for INTEGER_TYPE TYPE_OVERFLOW_TRAPS.
1287         (convert_mult_to_fma): Likewise.
1289         PR tree-optimization/83521
1290         * tree-ssa-phiopt.c (factor_out_conditional_conversion): Use
1291         gimple_build_assign without code on result of
1292         fold_build1 (VIEW_CONVERT_EXPR, ...), as it might not create
1293         a VIEW_CONVERT_EXPR.
1295 2017-12-21  Andrew Pinski  <apinski@cavium.com>
1296             Steve Ellcey  <sellcey@cavium.com>
1298         * config/aarch64/t-aarch64-linux (MULTILIB_OSDIRNAMES): Handle
1299         multi-arch for ilp32.
1301 2017-12-21  Uros Bizjak  <ubizjak@gmail.com>
1303         PR target/83467
1304         * config/i386/i386.md (*ashl<mode>3_mask): Add operand
1305         constraints to operand 2.
1306         (*ashl<mode>3_mask_1): Ditto.
1307         (*<shift_insn><mode>3_mask): Ditto.
1308         (*<shift_insn><mode>3_mask_1): Ditto.
1309         (*<rotate_insn><mode>3_mask): Ditto.
1310         (*<rotate_insn><mode>3_mask_1): Ditto.
1312 2017-12-21  Alexandre Oliva  <aoliva@redhat.com>
1314         * reorg.c (make_return_insns): Reemit each insn with its own location.
1316 2017-12-21  James Greenhalgh  <james.greenhalgh@arm.com>
1318         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
1319         generation for cases where splatting a value is not useful.
1320         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
1321         across a vec_duplicate and a paradoxical subreg forming a vector
1322         mode to a vec_concat.
1324 2017-12-21  James Greenhalgh  <james.greenhalgh@arm.com>
1326         * combine.c (simplify_set): Do not transform subregs to zero_extends
1327         if the destination is not a scalar int mode.
1329 2017-12-21  Eric Botcazou  <ebotcazou@adacore.com>
1331         PR c++/82872
1332         * convert.c (convert_to_integer_1) <POINTER_TYPE>: Do not return the
1333         shared zero if the input has overflowed.
1335 2017-12-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1337         * config/arm/driver-arm.c (arm_cpu_table): Specify dotprod
1338         support for Cortex-A55 and Cortex-A75.
1340 2017-12-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1342         * common/config/arm/arm-common.c (compare_opt_names): Add function
1343         comment.  Use strcmp instead of manual loop.
1345 2017-12-21  Martin Liska  <mliska@suse.cz>
1347         PR gcov-profile/83509
1348         * gcov-dump.c (dump_gcov_file): Do not read info about
1349         support_unexecuted_blocks for gcda files.
1351 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1353         PR rtl-optimization/82973
1354         * emit-rtl.h (valid_for_const_vec_duplicate_p): Rename to ...
1355         (valid_for_const_vector_p): ... this.
1356         * emit-rtl.c (valid_for_const_vec_duplicate_p): Rename to ...
1357         (valid_for_const_vector_p): ... this.  Adjust function comment.
1358         (gen_vec_duplicate): Adjust caller.
1359         * optabs.c (expand_vector_broadcast): Likewise.
1360         * simplify-rtx.c (simplify_const_unary_operation): Don't optimize into
1361         CONST_VECTOR if some element isn't simplified valid_for_const_vector_p
1362         constant.
1363         (simplify_const_binary_operation): Likewise.  Use CONST_FIXED_P macro
1364         instead of GET_CODE == CONST_FIXED.
1365         (simplify_subreg): Use CONST_FIXED_P macro instead of
1366         GET_CODE == CONST_FIXED.
1368         PR target/83488
1369         * config/i386/i386.c (ix86_target_string): Move -mavx512vbmi2 and
1370         -mshstk entries from isa_opts2 to isa_opts and -mhle, -mmovbe,
1371         -mclzero and -mmwaitx entries from isa_opts to isa_opts2.
1372         (ix86_option_override_internal): Adjust for
1373         OPTION_MASK_ISA_{HLE,MOVBE,CLZERO,MWAITX} moving to ix86_isa_flags2
1374         and OPTION_MASK_ISA_SHSTK moving to ix86_isa_flags.
1375         (BDESC_VERIFYS): Remove SPECIAL_ARGS2 related checks.
1376         (ix86_init_mmx_sse_builtins): Remove bdesc_special_args2 handling.
1377         Use def_builtin2 instead of def_builtin for OPTION_MASK_ISA_MWAITX
1378         and OPTION_MASK_ISA_CLZERO builtins.  Use def_builtin instead of
1379         def_builtin2 for CET builtins.
1380         (ix86_expand_builtin): Remove bdesc_special_args2 handling.  Fix
1381         up formatting in IX86_BUILTIN_RDPID code.
1382         * config/i386/i386-builtin.def: Move VBMI2 builtins from SPECIAL_ARGS2
1383         section to SPECIAL_ARGS and from ARGS2 section to ARGS.
1384         * config/i386/i386.opt (mavx512vbmi2, mshstk): Move from
1385         ix86_isa_flags2 to ix86_isa_flags.
1386         (mhle, mmovbe, mclzero, mmwaitx): Move from ix86_isa_flags to
1387         ix86_isa_flags2.
1388         * config/i386/i386-c.c (ix86_target_macros_internal): Check for
1389         OPTION_MASK_ISA_{CLZERO,MWAITX} in isa_flag2 instead of isa_flag.
1390         Check for OPTION_MASK_ISA_{SHSTK,AVX512VBMI2} in isa_flag instead
1391         of isa_flag2.
1392         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VBMI2_SET):
1393         Or in OPTION_MASK_ISA_AVX512F_SET.
1394         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
1395         OPTION_MASK_ISA_AVX512VBMI2_UNSET.
1396         (ix86_handle_option): Adjust for
1397         OPTION_MASK_ISA_{SHSTK,AVX512VBMI2}_*SET being in ix86_isa_flags
1398         and OPTION_MASK_ISA_{MOVBE,MWAITX,CLZERO}_*SET in ix86_isa_flags2.
1400 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1401             Alan Hayward  <alan.hayward@arm.com>
1402             David Sherwood  <david.sherwood@arm.com>
1404         * tree-data-ref.h (prune_runtime_alias_test_list): Take the
1405         factor as a poly_uint64 rather than an unsigned HOST_WIDE_INT.
1406         * tree-data-ref.c (prune_runtime_alias_test_list): Likewise.
1407         Track polynomial offsets.
1409 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1410             Alan Hayward  <alan.hayward@arm.com>
1411             David Sherwood  <david.sherwood@arm.com>
1413         * tree-vect-data-refs.c (vect_compute_data_ref_alignment):
1414         Treat drb->init as a poly_int.  Fail if its misalignment wrt
1415         vector_alignment isn't known.
1417 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1418             Alan Hayward  <alan.hayward@arm.com>
1419             David Sherwood  <david.sherwood@arm.com>
1421         * tree-vectorizer.h (_loop_vec_info): Add a versioning_threshold
1422         field.
1423         (LOOP_VINFO_VERSIONING_THRESHOLD): New macro
1424         (vect_loop_versioning): Take the loop versioning threshold as a
1425         separate parameter.
1426         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
1427         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
1428         versioning_threshold.
1429         (vect_analyze_loop_2): Compute the loop versioning threshold
1430         whenever loop versioning is needed, and store it in the new
1431         field rather than combining it with the cost model threshold.
1432         (vect_transform_loop): Update call to vect_loop_versioning.
1433         Try to combine the loop versioning and cost thresholds here.
1435 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1436             Alan Hayward  <alan.hayward@arm.com>
1437             David Sherwood  <david.sherwood@arm.com>
1439         * tree-ssa-loop-ivopts.h (strip_offset): Return the offset as
1440         poly_uint64_pod rather than an unsigned HOST_WIDE_INT.
1441         * tree-loop-distribution.c (classify_builtin_st): Update accordingly.
1442         * tree-ssa-loop-ivopts.c (iv_use::addr_offset): Change from
1443         an unsigned HOST_WIDE_INT to a poly_uint64_pod.
1444         (group_compare_offset): Update accordingly.
1445         (split_small_address_groups_p): Likewise.
1446         (record_use): Take addr_offset as a poly_uint64 rather than
1447         an unsigned HOST_WIDE_INT.
1448         (strip_offset): Return the offset as a poly_uint64 rather than
1449         an unsigned HOST_WIDE_INT.
1450         (record_group_use, split_address_groups): Track polynomial offsets.
1451         (add_iv_candidate_for_use): Likewise.
1452         (addr_offset_valid_p): Take the offset as a poly_int64 rather
1453         than a HOST_WIDE_INT.
1454         (strip_offset_1): Return the offset as a poly_int64 rather than
1455         a HOST_WIDE_INT.
1457 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1458             Alan Hayward  <alan.hayward@arm.com>
1459             David Sherwood  <david.sherwood@arm.com>
1461         * tree.h (get_binfo_at_offset): Take the offset as a poly_int64
1462         rather than a HOST_WIDE_INT.
1463         * tree.c (get_binfo_at_offset): Likewise.
1465 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1466             Alan Hayward  <alan.hayward@arm.com>
1467             David Sherwood  <david.sherwood@arm.com>
1469         * ipa-prop.h (build_ref_for_offset): Take the offset as a poly_int64
1470         rather than a HOST_WIDE_INT.
1471         * tree-sra.c (build_ref_for_offset): Likewise.
1473 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1474             Alan Hayward  <alan.hayward@arm.com>
1475             David Sherwood  <david.sherwood@arm.com>
1477         * fold-const.h (mem_ref_offset): Return a poly_offset_int rather
1478         than an offset_int.
1479         * tree.c (mem_ref_offset): Likewise.
1480         (build_simple_mem_ref_loc): Treat MEM_REF offsets as poly_ints.
1481         * builtins.c (get_object_alignment_2): Likewise.
1482         * expr.c (get_inner_reference, expand_expr_real_1): Likewise.
1483         * gimple-fold.c (get_base_constructor): Likewise.
1484         * gimple-ssa-strength-reduction.c (restructure_reference): Likewise.
1485         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
1486         Likewise.
1487         * ipa-polymorphic-call.c
1488         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Likewise.
1489         * ipa-prop.c (compute_complex_assign_jump_func): Likewise.
1490         (get_ancestor_addr_info): Likewise.
1491         * ipa-param-manipulation.c (ipa_get_adjustment_candidate): Likewise.
1492         * match.pd: Likewise.
1493         * tree-data-ref.c (dr_analyze_innermost): Likewise.
1494         * tree-dfa.c (get_addr_base_and_unit_offset_1): Likewise.
1495         * tree-eh.c (tree_could_trap_p): Likewise.
1496         * tree-object-size.c (addr_object_size): Likewise.
1497         * tree-ssa-address.c (copy_ref_info): Likewise.
1498         * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Likewise.
1499         (indirect_refs_may_alias_p): Likewise.
1500         * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Likewise.
1501         * tree-ssa.c (maybe_rewrite_mem_ref_base): Likewise.
1502         (non_rewritable_mem_ref_base): Likewise.
1503         * tree-vect-data-refs.c (vect_check_gather_scatter): Likewise.
1504         * tree-vrp.c (vrp_prop::check_array_ref): Likewise.
1505         * varasm.c (decode_addr_const): Likewise.
1507 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1508             Alan Hayward  <alan.hayward@arm.com>
1509             David Sherwood  <david.sherwood@arm.com>
1511         * gimple-ssa-stor-merging.c (find_bswap_or_nop_load): Track polynomial
1512         offsets for MEM_REFs.
1514 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1515             Alan Hayward  <alan.hayward@arm.com>
1516             David Sherwood  <david.sherwood@arm.com>
1518         * tree-ssanames.h (adjust_ptr_info_misalignment): Take the increment
1519         as a poly_uint64 rather than an unsigned int.
1520         * tree-ssanames.c (adjust_ptr_info_misalignment): Likewise.
1522 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1523             Alan Hayward  <alan.hayward@arm.com>
1524             David Sherwood  <david.sherwood@arm.com>
1526         * varasm.c (addr_const::offset): Change from HOST_WIDE_INT
1527         to poly_int64.
1528         (decode_addr_const): Update accordingly.
1530 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1531             Alan Hayward  <alan.hayward@arm.com>
1532             David Sherwood  <david.sherwood@arm.com>
1534         * tree.h (bit_field_size, bit_field_offset): New functions.
1535         * hsa-gen.c (gen_hsa_addr): Use them.
1536         * tree-ssa-forwprop.c (simplify_bitfield_ref): Likewise.
1537         (simplify_vector_constructor): Likewise.
1538         * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Likewise.
1539         * tree-cfg.c (verify_expr): Require the sizes and offsets of a
1540         BIT_FIELD_REF to be poly_uint64s rather than uhwis.
1541         * fold-const.c (fold_ternary_loc): Protect tree_to_uhwi with
1542         tree_fits_uhwi_p.
1544 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1545             Alan Hayward  <alan.hayward@arm.com>
1546             David Sherwood  <david.sherwood@arm.com>
1548         * expr.h (emit_group_load, emit_group_load_into_temps)
1549         (emit_group_store): Take the size as a poly_int64 rather than an int.
1550         * expr.c (emit_group_load_1, emit_group_load): Likewise.
1551         (emit_group_load_into_temp, emit_group_store): Likewise.
1553 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1554             Alan Hayward  <alan.hayward@arm.com>
1555             David Sherwood  <david.sherwood@arm.com>
1557         * ira-int.h (ira_spilled_reg_stack_slot::width): Change from
1558         an unsigned int to a poly_uint64.
1559         * ira.h (ira_reuse_stack_slot, ira_mark_new_stack_slot): Take the
1560         sizes as poly_uint64s rather than unsigned ints.
1561         * ira-color.c (ira_reuse_stack_slot, ira_mark_new_stack_slot):
1562         Likewise.
1564 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1565             Alan Hayward  <alan.hayward@arm.com>
1566             David Sherwood  <david.sherwood@arm.com>
1568         * lra-constraints.c (emit_inc): Change inc_amount from an int
1569         to a poly_int64.
1571 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1572             Alan Hayward  <alan.hayward@arm.com>
1573             David Sherwood  <david.sherwood@arm.com>
1575         * cfgexpand.c (stack_var::size): Change from a HOST_WIDE_INT
1576         to a poly_uint64.
1577         (add_stack_var, stack_var_cmp, partition_stack_vars)
1578         (dump_stack_var_partition): Update accordingly.
1579         (alloc_stack_frame_space): Take the size as a poly_int64 rather
1580         than a HOST_WIDE_INT.
1581         (expand_stack_vars, expand_one_stack_var_1): Handle polynomial sizes.
1582         (defer_stack_allocation, estimated_stack_frame_size): Likewise.
1583         (account_stack_vars, expand_one_var): Likewise.  Return a poly_uint64
1584         rather than a HOST_WIDE_INT.
1586 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1587             Alan Hayward  <alan.hayward@arm.com>
1588             David Sherwood  <david.sherwood@arm.com>
1590         * target.def (return_pops_args): Treat both the input and output
1591         sizes as poly_int64s rather than HOST_WIDE_INTS.
1592         * targhooks.h (default_return_pops_args): Update accordingly.
1593         * targhooks.c (default_return_pops_args): Likewise.
1594         * doc/tm.texi: Regenerate.
1595         * emit-rtl.h (incoming_args): Change pops_args, size and
1596         outgoing_args_size from int to poly_int64_pod.
1597         * function.h (expr_status): Change x_pending_stack_adjust and
1598         x_stack_pointer_delta from int to poly_int64.
1599         (args_size::constant): Change from HOST_WIDE_INT to poly_int64.
1600         (ARGS_SIZE_RTX): Update accordingly.
1601         * calls.c (highest_outgoing_arg_in_use): Change from int to
1602         unsigned int.
1603         (stack_usage_watermark, stored_args_watermark): New variables.
1604         (stack_region_maybe_used_p, mark_stack_region_used): New functions.
1605         (emit_call_1): Change the stack_size and rounded_stack_size
1606         parameters from HOST_WIDE_INT to poly_int64.  Track n_popped
1607         as a poly_int64.
1608         (save_fixed_argument_area): Check stack_usage_watermark.
1609         (initialize_argument_information): Change old_pending_adj from
1610         a HOST_WIDE_INT * to a poly_int64_pod *.
1611         (compute_argument_block_size): Return the size as a poly_int64
1612         rather than an int.
1613         (finalize_must_preallocate): Track polynomial argument sizes.
1614         (compute_argument_addresses): Likewise.
1615         (internal_arg_pointer_based_exp): Track polynomial offsets.
1616         (mem_overlaps_already_clobbered_arg_p): Rename to...
1617         (mem_might_overlap_already_clobbered_arg_p): ...this and take the
1618         size as a poly_uint64 rather than an unsigned HOST_WIDE_INT.
1619         Check stored_args_used_watermark.
1620         (load_register_parameters): Update accordingly.
1621         (check_sibcall_argument_overlap_1): Likewise.
1622         (combine_pending_stack_adjustment_and_call): Take the unadjusted
1623         args size as a poly_int64 rather than an int.  Return a bool
1624         indicating whether the optimization was possible and return
1625         the new adjustment by reference.
1626         (check_sibcall_argument_overlap): Track polynomail argument sizes.
1627         Update stored_args_watermark.
1628         (can_implement_as_sibling_call_p): Handle polynomial argument sizes.
1629         (expand_call): Likewise.  Maintain stack_usage_watermark and
1630         stored_args_watermark.  Update calls to
1631         combine_pending_stack_adjustment_and_call.
1632         (emit_library_call_value_1): Handle polynomial argument sizes.
1633         Call stack_region_maybe_used_p and mark_stack_region_used.
1634         Maintain stack_usage_watermark.
1635         (store_one_arg): Likewise.  Update call to
1636         mem_overlaps_already_clobbered_arg_p.
1637         * config/arm/arm.c (arm_output_function_prologue): Add a cast to
1638         HOST_WIDE_INT.
1639         * config/avr/avr.c (avr_outgoing_args_size): Likewise.
1640         * config/microblaze/microblaze.c (microblaze_function_prologue):
1641         Likewise.
1642         * config/cr16/cr16.c (cr16_return_pops_args): Update for new
1643         TARGET_RETURN_POPS_ARGS interface.
1644         (cr16_compute_frame, cr16_initial_elimination_offset): Add casts
1645         to HOST_WIDE_INT.
1646         * config/ft32/ft32.c (ft32_compute_frame): Likewise.
1647         * config/i386/i386.c (ix86_return_pops_args): Update for new
1648         TARGET_RETURN_POPS_ARGS interface.
1649         (ix86_expand_split_stack_prologue): Add a cast to HOST_WIDE_INT.
1650         * config/moxie/moxie.c (moxie_compute_frame): Likewise.
1651         * config/m68k/m68k.c (m68k_return_pops_args): Update for new
1652         TARGET_RETURN_POPS_ARGS interface.
1653         * config/vax/vax.c (vax_return_pops_args): Likewise.
1654         * config/pa/pa.h (STACK_POINTER_OFFSET): Add a cast to poly_int64.
1655         (EXIT_IGNORE_STACK): Update reference to crtl->outgoing_args_size.
1656         * config/arm/arm.h (CALLER_INTERWORKING_SLOT_SIZE): Likewise.
1657         * config/powerpcspe/aix.h (STACK_DYNAMIC_OFFSET): Likewise.
1658         * config/powerpcspe/darwin.h (STACK_DYNAMIC_OFFSET): Likewise.
1659         * config/powerpcspe/powerpcspe.h (STACK_DYNAMIC_OFFSET): Likewise.
1660         * config/rs6000/aix.h (STACK_DYNAMIC_OFFSET): Likewise.
1661         * config/rs6000/darwin.h (STACK_DYNAMIC_OFFSET): Likewise.
1662         * config/rs6000/rs6000.h (STACK_DYNAMIC_OFFSET): Likewise.
1663         * dojump.h (saved_pending_stack_adjust): Change x_pending_stack_adjust
1664         and x_stack_pointer_delta from int to poly_int64.
1665         * dojump.c (do_pending_stack_adjust): Update accordingly.
1666         * explow.c (allocate_dynamic_stack_space): Handle polynomial
1667         stack_pointer_deltas.
1668         * function.c (STACK_DYNAMIC_OFFSET): Add a cast to poly_int64.
1669         (pad_to_arg_alignment): Track polynomial offsets.
1670         (assign_parm_find_stack_rtl): Likewise.
1671         (assign_parms, locate_and_pad_parm): Handle polynomial argument sizes.
1672         * toplev.c (output_stack_usage): Update reference to
1673         current_function_pushed_stack_size.
1675 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1676             Alan Hayward  <alan.hayward@arm.com>
1677             David Sherwood  <david.sherwood@arm.com>
1679         * function.c (in_arg_offset, var_offset, dynamic_offset)
1680         (out_arg_offset, cfa_offset): Change from int to poly_int64.
1681         (instantiate_new_reg): Return the new offset as a poly_int64_pod
1682         rather than a HOST_WIDE_INT.
1683         (instantiate_virtual_regs_in_rtx): Track polynomial offsets.
1684         (instantiate_virtual_regs_in_insn): Likewise.
1686 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1687             Alan Hayward  <alan.hayward@arm.com>
1688             David Sherwood  <david.sherwood@arm.com>
1690         * rtl.h (get_args_size, add_args_size_note): New functions.
1691         (find_args_size_adjust): Return a poly_int64 rather than a
1692         HOST_WIDE_INT.
1693         (fixup_args_size_notes): Likewise.  Make the same change to the
1694         end_args_size parameter.
1695         * rtlanal.c (get_args_size, add_args_size_note): New functions.
1696         * builtins.c (expand_builtin_trap): Use add_args_size_note.
1697         * calls.c (emit_call_1): Likewise.
1698         * explow.c (adjust_stack_1): Likewise.
1699         * cfgcleanup.c (old_insns_match_p): Update use of
1700         find_args_size_adjust.
1701         * combine.c (distribute_notes): Track polynomial arg sizes.
1702         * dwarf2cfi.c (dw_trace_info): Change beg_true_args_size,
1703         end_true_args_size, beg_delay_args_size and end_delay_args_size
1704         from HOST_WIDE_INT to poly_int64.
1705         (add_cfi_args_size): Take the args_size as a poly_int64 rather
1706         than a HOST_WIDE_INT.
1707         (notice_args_size, notice_eh_throw, maybe_record_trace_start)
1708         (maybe_record_trace_start_abnormal, scan_trace, connect_traces): Track
1709         polynomial arg sizes.
1710         * emit-rtl.c (try_split): Use get_args_size.
1711         * recog.c (peep2_attempt): Likewise.
1712         * reload1.c (reload_as_needed): Likewise.
1713         * expr.c (find_args_size_adjust): Return the adjustment as a
1714         poly_int64 rather than a HOST_WIDE_INT.
1715         (fixup_args_size_notes): Change end_args_size from a HOST_WIDE_INT
1716         to a poly_int64 and change the return type in the same way.
1717         (emit_single_push_insn): Track polynomial arg sizes.
1719 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1720             Alan Hayward  <alan.hayward@arm.com>
1721             David Sherwood  <david.sherwood@arm.com>
1723         * expr.h (push_block, emit_push_insn): Change the "extra" parameter
1724         from HOST_WIDE_INT to poly_int64.
1725         * expr.c (push_block, emit_push_insn): Likewise.
1727 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1728             Alan Hayward  <alan.hayward@arm.com>
1729             David Sherwood  <david.sherwood@arm.com>
1731         * function.h (frame_space): Change start and length from HOST_WIDE_INT
1732         to poly_int64.
1733         (get_frame_size): Return the size as a poly_int64 rather than a
1734         HOST_WIDE_INT.
1735         (frame_offset_overflow): Take the offset as a poly_int64 rather
1736         than a HOST_WIDE_INT.
1737         (assign_stack_local_1, assign_stack_local, assign_stack_temp_for_type)
1738         (assign_stack_temp): Likewise for the size.
1739         * function.c (get_frame_size): Return a poly_int64 rather than
1740         a HOST_WIDE_INT.
1741         (frame_offset_overflow): Take the offset as a poly_int64 rather
1742         than a HOST_WIDE_INT.
1743         (try_fit_stack_local): Take the start, length and size as poly_int64s
1744         rather than HOST_WIDE_INTs.  Return the offset as a poly_int64_pod
1745         rather than a HOST_WIDE_INT.
1746         (add_frame_space): Take the start and end as poly_int64s rather than
1747         HOST_WIDE_INTs.
1748         (assign_stack_local_1, assign_stack_local, assign_stack_temp_for_type)
1749         (assign_stack_temp): Likewise for the size.
1750         (temp_slot): Change size, base_offset and full_size from HOST_WIDE_INT
1751         to poly_int64.
1752         (find_temp_slot_from_address): Handle polynomial offsets.
1753         (combine_temp_slots): Likewise.
1754         * emit-rtl.h (rtl_data::x_frame_offset): Change from HOST_WIDE_INT
1755         to poly_int64.
1756         * cfgexpand.c (alloc_stack_frame_space): Return the offset as a
1757         poly_int64 rather than a HOST_WIDE_INT.
1758         (expand_one_stack_var_at): Take the offset as a poly_int64 rather
1759         than a HOST_WIDE_INT.
1760         (expand_stack_vars, expand_one_stack_var_1, expand_used_vars): Handle
1761         polynomial frame offsets.
1762         * config/m32r/m32r-protos.h (m32r_compute_frame_size): Take the size
1763         as a poly_int64 rather than an int.
1764         * config/m32r/m32r.c (m32r_compute_frame_size): Likewise.
1765         * config/v850/v850-protos.h (compute_frame_size): Likewise.
1766         * config/v850/v850.c (compute_frame_size): Likewise.
1767         * config/xtensa/xtensa-protos.h (compute_frame_size): Likewise.
1768         * config/xtensa/xtensa.c (compute_frame_size): Likewise.
1769         * config/pa/pa-protos.h (pa_compute_frame_size): Likewise.
1770         * config/pa/pa.c (pa_compute_frame_size): Likewise.
1771         * explow.h (get_dynamic_stack_base): Take the offset as a poly_int64
1772         rather than a HOST_WIDE_INT.
1773         * explow.c (get_dynamic_stack_base): Likewise.
1774         * final.c (final_start_function): Use the constant lower bound
1775         of the frame size for -Wframe-larger-than.
1776         * ira.c (do_reload): Adjust for new get_frame_size return type.
1777         * lra.c (lra): Likewise.
1778         * reload1.c (reload): Likewise.
1779         * config/avr/avr.c (avr_asm_function_end_prologue): Likewise.
1780         * config/pa/pa.h (EXIT_IGNORE_STACK): Likewise.
1781         * rtlanal.c (get_initial_register_offset): Return the offset as
1782         a poly_int64 rather than a HOST_WIDE_INT.
1784 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1785             Alan Hayward  <alan.hayward@arm.com>
1786             David Sherwood  <david.sherwood@arm.com>
1788         * reload1.c (elim_table): Change initial_offset, offset and
1789         previous_offset from HOST_WIDE_INT to poly_int64_pod.
1790         (offsets_at): Change the target array's element type from
1791         HOST_WIDE_INT to poly_int64_pod.
1792         (set_label_offsets, eliminate_regs_1, eliminate_regs_in_insn)
1793         (elimination_costs_in_insn, update_eliminable_offsets)
1794         (verify_initial_elim_offsets, set_offsets_for_label)
1795         (init_eliminable_invariants): Update after above changes.
1797 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1798             Alan Hayward  <alan.hayward@arm.com>
1799             David Sherwood  <david.sherwood@arm.com>
1801         * reload.h (reload::inc): Change from an int to a poly_int64_pod.
1802         * reload.c (combine_reloads, debug_reload_to_stream): Likewise.
1803         (decomposition): Change start and end from HOST_WIDE_INT
1804         to poly_int64_pod.
1805         (decompose, immune_p): Update accordingly.
1806         (find_inc_amount): Return a poly_int64 rather than an int.
1807         * reload1.c (inc_for_reload): Take the inc_amount as a poly_int64
1808         rather than an int.
1810 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1811             Alan Hayward  <alan.hayward@arm.com>
1812             David Sherwood  <david.sherwood@arm.com>
1814         * tree.h (get_inner_reference): Return the bitsize and bitpos
1815         as poly_int64_pods rather than HOST_WIDE_INT.
1816         * fold-const.h (ptr_difference_const): Return the pointer difference
1817         as a poly_int64_pod rather than a HOST_WIDE_INT.
1818         * expr.c (get_inner_reference): Return the bitsize and bitpos
1819         as poly_int64_pods rather than HOST_WIDE_INT.
1820         (expand_expr_addr_expr_1, expand_expr_real_1): Track polynomial
1821         offsets and sizes.
1822         * fold-const.c (make_bit_field_ref): Take the bitpos as a poly_int64
1823         rather than a HOST_WIDE_INT.  Update call to get_inner_reference.
1824         (optimize_bit_field_compare): Update call to get_inner_reference.
1825         (decode_field_reference): Likewise.
1826         (fold_unary_loc): Track polynomial offsets and sizes.
1827         (split_address_to_core_and_offset): Return the bitpos as a
1828         poly_int64_pod rather than a HOST_WIDE_INT.
1829         (ptr_difference_const): Likewise for the pointer difference.
1830         * asan.c (instrument_derefs): Track polynomial offsets and sizes.
1831         * config/mips/mips.c (r10k_safe_mem_expr_p): Likewise.
1832         * dbxout.c (dbxout_expand_expr): Likewise.
1833         * dwarf2out.c (loc_list_for_address_of_addr_expr_of_indirect_ref)
1834         (loc_list_from_tree_1, fortran_common): Likewise.
1835         * gimple-laddress.c (pass_laddress::execute): Likewise.
1836         * gimple-ssa-store-merging.c (find_bswap_or_nop_load): Likewise.
1837         * gimplify.c (gimplify_scan_omp_clauses): Likewise.
1838         * simplify-rtx.c (delegitimize_mem_from_attrs): Likewise.
1839         * tree-affine.c (tree_to_aff_combination): Likewise.
1840         (get_inner_reference_aff): Likewise.
1841         * tree-data-ref.c (split_constant_offset_1): Likewise.
1842         (dr_analyze_innermost): Likewise.
1843         * tree-scalar-evolution.c (interpret_rhs_expr): Likewise.
1844         * tree-sra.c (ipa_sra_check_caller): Likewise.
1845         * tree-vect-data-refs.c (vect_check_gather_scatter): Likewise.
1846         * ubsan.c (maybe_instrument_pointer_overflow): Likewise.
1847         (instrument_bool_enum_load, instrument_object_size): Likewise.
1848         * gimple-ssa-strength-reduction.c (slsr_process_ref): Update call
1849         to get_inner_reference.
1850         * hsa-gen.c (gen_hsa_addr): Likewise.
1851         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Likewise.
1852         * tsan.c (instrument_expr): Likewise.
1853         * match.pd: Update call to ptr_difference_const.
1855 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1856             Alan Hayward  <alan.hayward@arm.com>
1857             David Sherwood  <david.sherwood@arm.com>
1859         * fold-const.c (fold_comparison): Track sizes and offsets as
1860         poly_int64s rather than HOST_WIDE_INTs when folding address
1861         comparisons.
1863 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1864             Alan Hayward  <alan.hayward@arm.com>
1865             David Sherwood  <david.sherwood@arm.com>
1867         * expr.h (get_bit_range): Return the bitstart and bitend as
1868         poly_uint64s rather than unsigned HOST_WIDE_INTs.  Return the bitpos
1869         as a poly_int64 rather than a HOST_WIDE_INT.
1870         * expr.c (get_bit_range): Likewise.
1871         (expand_assignment): Update call accordingly.
1872         * fold-const.c (optimize_bit_field_compare): Likewise.
1874 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1875             Alan Hayward  <alan.hayward@arm.com>
1876             David Sherwood  <david.sherwood@arm.com>
1878         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): New enum.
1879         (aarch64_legitimate_address_p): Use it instead of an rtx code,
1880         as an optional final parameter.
1881         * config/aarch64/aarch64.c (aarch64_classify_address): Likewise.
1882         (aarch64_legitimate_address_p): Likewise.
1883         (aarch64_print_address_internal): Take an aarch64_addr_query_type
1884         instead of an rtx code.
1885         (aarch64_address_valid_for_prefetch_p): Update calls accordingly.
1886         (aarch64_legitimate_address_hook_p): Likewise.
1887         (aarch64_print_ldpstp_address): Likewise.
1888         (aarch64_print_operand_address): Likewise.
1889         (aarch64_address_cost): Likewise.
1890         * config/aarch64/constraints.md (Uml, Umq, Ump, Utq): Likewise.
1891         * config/aarch64/predicates.md (aarch64_mem_pair_operand): Likewise.
1892         (aarch64_mem_pair_lanes_operand): Likewise.
1894 2017-12-20  Richard Biener  <rguenther@suse.de>
1896         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Call
1897         update_stmt_if_modified.
1899 2017-12-20  Wilco Dijkstra  <wdijkstr@arm.com>
1901         PR tree-optimization/83491
1902         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Check for SSA_NAME
1903         before walking uses.  Improve coding style and comments.
1905 2017-12-20  Tom de Vries  <tom@codesourcery.com>
1907         * gimple-fold.c (fold_internal_goacc_dim): Simplify.
1909 2017-12-20  Jakub Jelinek  <jakub@redhat.com>
1911         PR ipa/83506
1912         * ipa-fnsummary.c (pass_data_ipa_free_fn_summary): Use 0 for
1913         todo_flags_finish.
1914         (pass_ipa_free_fn_summary): Add small_p private data member,
1915         initialize to false in the ctor.
1916         (pass_ipa_free_fn_summary::clone,
1917         pass_ipa_free_fn_summary::set_pass_param,
1918         pass_ipa_free_fn_summary::gate): New methods.
1919         (pass_ipa_free_fn_summary::execute): Return TODO_remove_functions
1920         | TODO_dump_symtab if small_p.
1921         * passes.def: Add true parm for the existing pass_ipa_free_fn_summary
1922         entry and add another instance of the pass with false parm after
1923         ipa-pure-const.
1924         * ipa-pure-const.c (pass_ipa_pure_const): Don't call
1925         ipa_free_fn_summary here.
1927 2017-12-20  Paolo Carlini  <paolo.carlini@oracle.com>
1929         * gimplify.c (gimplify_return_expr): Remove dead error_mark_node check.
1931 2017-12-20  Martin Sebor  <msebor@redhat.com>
1933         PR testsuite/83131
1934         * builtins.c (expand_builtin_strlen): Use get_callee_fndecl.
1935         (expand_builtin_strcmp): Call maybe_warn_nonstring_arg.
1936         (expand_builtin_strncmp): Same.
1938 2017-12-20  Alexandre Oliva  <aoliva@redhat.com>
1940         PR bootstrap/83396
1941         * cfgexpand.c (label_rtx_for_bb): Revert SFN changes that
1942         allowed debug stmts before labels.
1943         (expand_gimple_basic_block): Likewise.
1944         * gimple-iterator.c (gimple_find_edge_insert_loc): Likewise.
1945         * gimple-iterator.h (gsi_after_labels): Likewise.
1946         * tree-cfgcleanup (remove_forwarder_block): Likewise, but
1947         rename reused variable, and simplify using gsi_move_before.
1948         * tree-ssa-tail-merge.c (find_duplicate): Likewise.
1949         * tree-cfg.c (make_edges, cleanup_dead_labels): Likewise.
1950         (gimple_can_merge_blocks_p, verify_gimple_in_cfg): Likewise.
1951         (gimple_verify_flow_info, gimple_block_label): Likewise.
1952         (make_blocks): Move debug markers after adjacent labels.
1953         * cfgrtl.c (skip_insns_after_block): Revert SFN changes that
1954         allowed debug insns outside blocks.
1955         * df-scan.c (df_insn_delete): Likewise.
1956         * lra-constraints.c (update_ebb_live_info): Likewise.
1957         * var-tracking.c (get_first_insn, vt_emit_notes): Likewise.
1958         (vt_initialize, delete_vta_debug_insns): Likewise.
1959         (reemit_marker_as_note): Drop BB parm.  Adjust callers.
1961 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1962             Alan Hayward  <alan.hayward@arm.com>
1963             David Sherwood  <david.sherwood@arm.com>
1965         * poly-int-types.h (round_down_to_byte_boundary): New macro.
1966         (round_up_to_byte_boundary): Likewise.
1967         * expr.h (get_bit_range): Add temporary shim.
1968         * gimple-ssa-store-merging.c (store_operand_info): Change the
1969         bitsize, bitpos, bitregion_start and bitregion_end fields from
1970         unsigned HOST_WIDE_INT to poly_uint64.
1971         (merged_store_group): Likewise load_align_base.
1972         (compatible_load_p, compatible_load_p): Update accordingly.
1973         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
1974         (split_group, imm_store_chain_info::output_merged_store): Likewise.
1975         (mem_valid_for_store_merging): Return the bitsize, bitpos,
1976         bitregion_start and bitregion_end as poly_uint64s rather than
1977         unsigned HOST_WIDE_INTs.  Track polynomial offsets internally.
1978         (handled_load): Take the bitsize, bitpos,
1979         bitregion_start and bitregion_end as poly_uint64s rather than
1980         unsigned HOST_WIDE_INTs.
1981         (pass_store_merging::process_store): Update call to
1982         mem_valid_for_store_merging.
1984 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1985             Alan Hayward  <alan.hayward@arm.com>
1986             David Sherwood  <david.sherwood@arm.com>
1988         * builtins.c (get_object_alignment_2): Track polynomial offsets
1989         and sizes.  Update the alignment handling.
1991 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1992             Alan Hayward  <alan.hayward@arm.com>
1993             David Sherwood  <david.sherwood@arm.com>
1995         * tree.h (get_inner_reference): Add a version that returns the
1996         offset and size as poly_int64_pods rather than HOST_WIDE_INTs.
1997         * cfgexpand.c (expand_debug_expr): Track polynomial offsets.  Simply
1998         the case in which bitpos is not associated with the first byte.
2000 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2001             Alan Hayward  <alan.hayward@arm.com>
2002             David Sherwood  <david.sherwood@arm.com>
2004         * tree-affine.h (get_inner_reference_aff): Return the size as a
2005         poly_widest_int.
2006         * tree-affine.c (get_inner_reference_aff): Likewise.
2007         * tree-data-ref.c (dr_may_alias_p): Update accordingly.
2008         * tree-ssa-loop-im.c (mem_refs_may_alias_p): Likewise.
2010 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2011             Alan Hayward  <alan.hayward@arm.com>
2012             David Sherwood  <david.sherwood@arm.com>
2014         * fold-const.c (pointer_may_wrap_p): Take the offset as a
2015         HOST_WIDE_INT rather than a poly_int64.
2017 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2018             Alan Hayward  <alan.hayward@arm.com>
2019             David Sherwood  <david.sherwood@arm.com>
2021         * gimple-ssa-store-merging.c (symbolic_number::bytepos): Change from
2022         HOST_WIDE_INT to poly_int64_pod.
2023         (perform_symbolic_merge): Update accordingly.
2024         (bswap_replace): Likewise.
2026 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2027             Alan Hayward  <alan.hayward@arm.com>
2028             David Sherwood  <david.sherwood@arm.com>
2030         * tree-affine.h (aff_tree::offset): Change from widest_int
2031         to poly_widest_int.
2032         (wide_int_ext_for_comb): Delete.
2033         (aff_combination_const, aff_comb_cannot_overlap_p): Take the
2034         constants as poly_widest_int rather than widest_int.
2035         (aff_combination_constant_multiple_p): Return the multiplier
2036         as a poly_widest_int.
2037         (aff_combination_zero_p, aff_combination_singleton_var_p): Handle
2038         polynomial offsets.
2039         * tree-affine.c (wide_int_ext_for_comb): Make original widest_int
2040         version static and add an overload for poly_widest_int.
2041         (aff_combination_const, aff_combination_add_cst)
2042         (wide_int_constant_multiple_p, aff_comb_cannot_overlap_p): Take
2043         the constants as poly_widest_int rather than widest_int.
2044         (tree_to_aff_combination): Generalize INTEGER_CST case to
2045         poly_int_tree_p.
2046         (aff_combination_to_tree): Track offsets as poly_widest_ints.
2047         (aff_combination_add_product, aff_combination_mult): Handle
2048         polynomial offsets.
2049         (aff_combination_constant_multiple_p): Return the multiplier
2050         as a poly_widest_int.
2051         * tree-predcom.c (determine_offset): Return the offset as a
2052         poly_widest_int.
2053         (split_data_refs_to_components, suitable_component_p): Update
2054         accordingly.
2055         (valid_initializer_p): Update call to
2056         aff_combination_constant_multiple_p.
2057         * tree-ssa-address.c (addr_to_parts): Handle polynomial offsets.
2058         * tree-ssa-loop-ivopts.c (get_address_cost_ainc): Take the step
2059         as a poly_int64 rather than a HOST_WIDE_INT.
2060         (get_address_cost): Handle polynomial offsets.
2061         (iv_elimination_compare_lt): Likewise.
2062         (rewrite_use_nonlinear_expr): Likewise.
2064 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2065             Alan Hayward  <alan.hayward@arm.com>
2066             David Sherwood  <david.sherwood@arm.com>
2068         * tree-dfa.h (get_addr_base_and_unit_offset_1): Return the offset
2069         as a poly_int64_pod rather than a HOST_WIDE_INT.
2070         (get_addr_base_and_unit_offset): Likewise.
2071         * tree-dfa.c (get_addr_base_and_unit_offset_1): Likewise.
2072         (get_addr_base_and_unit_offset): Likewise.
2073         * doc/match-and-simplify.texi: Change off from HOST_WIDE_INT
2074         to poly_int64 in example.
2075         * fold-const.c (fold_binary_loc): Update call to
2076         get_addr_base_and_unit_offset.
2077         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
2078         (maybe_canonicalize_mem_ref_addr): Likewise.
2079         (gimple_fold_stmt_to_constant_1): Likewise.
2080         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
2081         Likewise.
2082         * ipa-param-manipulation.c (ipa_modify_call_arguments): Likewise.
2083         * match.pd: Likewise.
2084         * omp-low.c (lower_omp_target): Likewise.
2085         * tree-sra.c (build_ref_for_offset): Likewise.
2086         (build_debug_ref_for_model): Likewise.
2087         * tree-ssa-address.c (maybe_fold_tmr): Likewise.
2088         * tree-ssa-alias.c (ao_ref_init_from_ptr_and_size): Likewise.
2089         * tree-ssa-ccp.c (optimize_memcpy): Likewise.
2090         * tree-ssa-forwprop.c (forward_propagate_addr_expr_1): Likewise.
2091         (constant_pointer_difference): Likewise.
2092         * tree-ssa-loop-niter.c (expand_simple_operations): Likewise.
2093         * tree-ssa-phiopt.c (jump_function_from_stmt): Likewise.
2094         * tree-ssa-pre.c (create_component_ref_by_pieces_1): Likewise.
2095         * tree-ssa-sccvn.c (vn_reference_fold_indirect): Likewise.
2096         (vn_reference_maybe_forwprop_address, vn_reference_lookup_3): Likewise.
2097         (set_ssa_val_to): Likewise.
2098         * tree-ssa-strlen.c (get_addr_stridx, addr_stridxptr)
2099         (maybe_diag_stxncpy_trunc): Likewise.
2100         * tree-vrp.c (vrp_prop::check_array_ref): Likewise.
2101         * tree.c (build_simple_mem_ref_loc): Likewise.
2102         (array_at_struct_end_p): Likewise.
2104 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2105             Alan Hayward  <alan.hayward@arm.com>
2106             David Sherwood  <david.sherwood@arm.com>
2108         * tree-dfa.h (get_ref_base_and_extent): Return the base, size and
2109         max_size as poly_int64_pods rather than HOST_WIDE_INTs.
2110         (get_ref_base_and_extent_hwi): Declare.
2111         * tree-dfa.c (get_ref_base_and_extent): Return the base, size and
2112         max_size as poly_int64_pods rather than HOST_WIDE_INTs.
2113         (get_ref_base_and_extent_hwi): New function.
2114         * cfgexpand.c (expand_debug_expr): Update call to
2115         get_ref_base_and_extent.
2116         * dwarf2out.c (add_var_loc_to_decl): Likewise.
2117         * gimple-fold.c (get_base_constructor): Return the offset as a
2118         poly_int64_pod rather than a HOST_WIDE_INT.
2119         (fold_const_aggregate_ref_1): Track polynomial sizes and offsets.
2120         * ipa-polymorphic-call.c
2121         (ipa_polymorphic_call_context::set_by_invariant)
2122         (extr_type_from_vtbl_ptr_store): Track polynomial offsets.
2123         (ipa_polymorphic_call_context::ipa_polymorphic_call_context)
2124         (check_stmt_for_type_change): Use get_ref_base_and_extent_hwi
2125         rather than get_ref_base_and_extent.
2126         (ipa_polymorphic_call_context::get_dynamic_type): Likewise.
2127         * ipa-prop.c (ipa_load_from_parm_agg, compute_complex_assign_jump_func)
2128         (get_ancestor_addr_info, determine_locally_known_aggregate_parts):
2129         Likewise.
2130         * ipa-param-manipulation.c (ipa_get_adjustment_candidate): Update
2131         call to get_ref_base_and_extent.
2132         * tree-sra.c (create_access, get_access_for_expr): Likewise.
2133         * tree-ssa-alias.c (ao_ref_base, aliasing_component_refs_p)
2134         (stmt_kills_ref_p): Likewise.
2135         * tree-ssa-dce.c (mark_aliased_reaching_defs_necessary_1): Likewise.
2136         * tree-ssa-scopedtables.c (avail_expr_hash, equal_mem_array_ref_p):
2137         Likewise.
2138         * tree-ssa-sccvn.c (vn_reference_lookup_3): Likewise.
2139         Use get_ref_base_and_extent_hwi rather than get_ref_base_and_extent
2140         when calling native_encode_expr.
2141         * tree-ssa-structalias.c (get_constraint_for_component_ref): Update
2142         call to get_ref_base_and_extent.
2143         (do_structure_copy): Use get_ref_base_and_extent_hwi rather than
2144         get_ref_base_and_extent.
2145         * var-tracking.c (track_expr_p): Likewise.
2147 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2148             Alan Hayward  <alan.hayward@arm.com>
2149             David Sherwood  <david.sherwood@arm.com>
2151         * ipa-param-manipulation.h (ipa_parm_adjustment::offset): Change from
2152         HOST_WIDE_INT to poly_int64_pod.
2153         * ipa-param-manipulation.c (ipa_modify_call_arguments): Track
2154         polynomail parameter offsets.
2156 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2157             Alan Hayward  <alan.hayward@arm.com>
2158             David Sherwood  <david.sherwood@arm.com>
2160         * gengtype.c (main): Handle poly_int64_pod.
2161         * dwarf2out.h (dw_cfi_oprnd_cfa_loc): New dw_cfi_oprnd_type.
2162         (dw_cfi_oprnd::dw_cfi_cfa_loc): New field.
2163         (dw_cfa_location::offset, dw_cfa_location::base_offset): Change
2164         from HOST_WIDE_INT to poly_int64_pod.
2165         * dwarf2cfi.c (queued_reg_save::cfa_offset): Likewise.
2166         (copy_cfa): New function.
2167         (lookup_cfa_1): Use the cached dw_cfi_cfa_loc, if it exists.
2168         (cfi_oprnd_equal_p): Handle dw_cfi_oprnd_cfa_loc.
2169         (cfa_equal_p, dwarf2out_frame_debug_adjust_cfa)
2170         (dwarf2out_frame_debug_cfa_offset, dwarf2out_frame_debug_expr)
2171         (initial_return_save): Treat offsets as poly_ints.
2172         (def_cfa_0): Likewise.  Cache the CFA in dw_cfi_cfa_loc if either
2173         offset is nonconstant.
2174         (reg_save): Take the offset as a poly_int64.  Fall back to
2175         DW_CFA_expression for nonconstant offsets.
2176         (queue_reg_save): Take the offset as a poly_int64.
2177         * dwarf2out.c (dw_cfi_oprnd2_desc): Handle DW_CFA_def_cfa_expression.
2179 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2180             Alan Hayward  <alan.hayward@arm.com>
2181             David Sherwood  <david.sherwood@arm.com>
2183         * rtl.h (operand_subword, operand_subword_force): Take the offset
2184         as a poly_uint64 an unsigned int.
2185         * emit-rtl.c (operand_subword, operand_subword_force): Likewise.
2187 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2188             Alan Hayward  <alan.hayward@arm.com>
2189             David Sherwood  <david.sherwood@arm.com>
2191         * doc/rtl.texi: Update documentation of SUBREG_BYTE.  Document the
2192         'p' format code.  Use INT_LIST rather than SUBREG as the example of
2193         a code with an XINT and an XEXP.  Remove the implication that
2194         accessing an rtx field using XINT is expected to work.
2195         * rtl.def (SUBREG): Change format from "ei" to "ep".
2196         * rtl.h (rtunion::rt_subreg): New field.
2197         (XCSUBREG): New macro.
2198         (SUBREG_BYTE): Use it.
2199         (subreg_shape): Change offset from an unsigned int to a poly_uint16.
2200         Update constructor accordingly.
2201         (subreg_shape::operator ==): Update accordingly.
2202         (subreg_shape::unique_id): Return an unsigned HOST_WIDE_INT rather
2203         than an unsigned int.
2204         (subreg_lsb, subreg_lowpart_offset, subreg_highpart_offset): Return
2205         a poly_uint64 rather than an unsigned int.
2206         (subreg_lsb_1): Likewise.  Take the offset as a poly_uint64 rather
2207         than an unsigned int.
2208         (subreg_size_offset_from_lsb, subreg_size_lowpart_offset)
2209         (subreg_size_highpart_offset): Return a poly_uint64 rather than
2210         an unsigned int.  Take the sizes as poly_uint64s.
2211         (subreg_offset_from_lsb): Return a poly_uint64 rather than
2212         an unsigned int.  Take the shift as a poly_uint64 rather than
2213         an unsigned int.
2214         (subreg_regno_offset, subreg_offset_representable_p): Take the offset
2215         as a poly_uint64 rather than an unsigned int.
2216         (simplify_subreg_regno): Likewise.
2217         (byte_lowpart_offset): Return the memory offset as a poly_int64
2218         rather than an int.
2219         (subreg_memory_offset): Likewise.  Take the subreg offset as a
2220         poly_uint64 rather than an unsigned int.
2221         (simplify_subreg, simplify_gen_subreg, subreg_get_info)
2222         (gen_rtx_SUBREG, validate_subreg): Take the subreg offset as a
2223         poly_uint64 rather than an unsigned int.
2224         * rtl.c (rtx_format): Describe 'p' in comment.
2225         (copy_rtx, rtx_equal_p_cb, rtx_equal_p): Handle 'p'.
2226         * emit-rtl.c (validate_subreg, gen_rtx_SUBREG): Take the subreg
2227         offset as a poly_uint64 rather than an unsigned int.
2228         (byte_lowpart_offset): Return the memory offset as a poly_int64
2229         rather than an int.
2230         (subreg_memory_offset): Likewise.  Take the subreg offset as a
2231         poly_uint64 rather than an unsigned int.
2232         (subreg_size_lowpart_offset, subreg_size_highpart_offset): Take the
2233         mode sizes as poly_uint64s rather than unsigned ints.  Return a
2234         poly_uint64 rather than an unsigned int.
2235         (subreg_lowpart_p): Treat subreg offsets as poly_ints.
2236         (copy_insn_1): Handle 'p'.
2237         * rtlanal.c (set_noop_p): Treat subregs offsets as poly_uint64s.
2238         (subreg_lsb_1): Take the subreg offset as a poly_uint64 rather than
2239         an unsigned int.  Return the shift in the same way.
2240         (subreg_lsb): Return the shift as a poly_uint64 rather than an
2241         unsigned int.
2242         (subreg_size_offset_from_lsb): Take the sizes and shift as
2243         poly_uint64s rather than unsigned ints.  Return the offset as
2244         a poly_uint64.
2245         (subreg_get_info, subreg_regno_offset, subreg_offset_representable_p)
2246         (simplify_subreg_regno): Take the offset as a poly_uint64 rather than
2247         an unsigned int.
2248         * rtlhash.c (add_rtx): Handle 'p'.
2249         * genemit.c (gen_exp): Likewise.
2250         * gengenrtl.c (type_from_format, gendef): Likewise.
2251         * gensupport.c (subst_pattern_match, get_alternatives_number)
2252         (collect_insn_data, alter_predicate_for_insn, alter_constraints)
2253         (subst_dup): Likewise.
2254         * gengtype.c (adjust_field_rtx_def): Likewise.
2255         * genrecog.c (find_operand, find_matching_operand, validate_pattern)
2256         (match_pattern_2): Likewise.
2257         (rtx_test::SUBREG_FIELD): New rtx_test::kind_enum.
2258         (rtx_test::subreg_field): New function.
2259         (operator ==, safe_to_hoist_p, transition_parameter_type)
2260         (print_nonbool_test, print_test): Handle SUBREG_FIELD.
2261         * genattrtab.c (attr_rtx_1): Say that 'p' is deliberately not handled.
2262         * genpeep.c (match_rtx): Likewise.
2263         * print-rtl.c (print_poly_int): Include if GENERATOR_FILE too.
2264         (rtx_writer::print_rtx_operand): Handle 'p'.
2265         (print_value): Handle SUBREG.
2266         * read-rtl.c (apply_int_iterator): Likewise.
2267         (rtx_reader::read_rtx_operand): Handle 'p'.
2268         * alias.c (rtx_equal_for_memref_p): Likewise.
2269         * cselib.c (rtx_equal_for_cselib_1, cselib_hash_rtx): Likewise.
2270         * caller-save.c (replace_reg_with_saved_mem): Treat subreg offsets
2271         as poly_ints.
2272         * calls.c (expand_call): Likewise.
2273         * combine.c (combine_simplify_rtx, expand_field_assignment): Likewise.
2274         (make_extraction, gen_lowpart_for_combine): Likewise.
2275         * loop-invariant.c (hash_invariant_expr_1, invariant_expr_equal_p):
2276         Likewise.
2277         * cse.c (remove_invalid_subreg_refs): Take the offset as a poly_uint64
2278         rather than an unsigned int.  Treat subreg offsets as poly_ints.
2279         (exp_equiv_p): Handle 'p'.
2280         (hash_rtx_cb): Likewise.  Treat subreg offsets as poly_ints.
2281         (equiv_constant, cse_insn): Treat subreg offsets as poly_ints.
2282         * dse.c (find_shift_sequence): Likewise.
2283         * dwarf2out.c (rtl_for_decl_location): Likewise.
2284         * expmed.c (extract_low_bits): Likewise.
2285         * expr.c (emit_group_store, undefined_operand_subword_p): Likewise.
2286         (expand_expr_real_2): Likewise.
2287         * final.c (alter_subreg): Likewise.
2288         (leaf_renumber_regs_insn): Handle 'p'.
2289         * function.c (assign_parm_find_stack_rtl, assign_parm_setup_stack):
2290         Treat subreg offsets as poly_ints.
2291         * fwprop.c (forward_propagate_and_simplify): Likewise.
2292         * ifcvt.c (noce_emit_move_insn, noce_emit_cmove): Likewise.
2293         * ira.c (get_subreg_tracking_sizes): Likewise.
2294         * ira-conflicts.c (go_through_subreg): Likewise.
2295         * ira-lives.c (process_single_reg_class_operands): Likewise.
2296         * jump.c (rtx_renumbered_equal_p): Likewise.  Handle 'p'.
2297         * lower-subreg.c (simplify_subreg_concatn): Take the subreg offset
2298         as a poly_uint64 rather than an unsigned int.
2299         (simplify_gen_subreg_concatn, resolve_simple_move): Treat
2300         subreg offsets as poly_ints.
2301         * lra-constraints.c (operands_match_p): Handle 'p'.
2302         (match_reload, curr_insn_transform): Treat subreg offsets as poly_ints.
2303         * lra-spills.c (assign_mem_slot): Likewise.
2304         * postreload.c (move2add_valid_value_p): Likewise.
2305         * recog.c (general_operand, indirect_operand): Likewise.
2306         * regcprop.c (copy_value, maybe_mode_change): Likewise.
2307         (copyprop_hardreg_forward_1): Likewise.
2308         * reginfo.c (simplifiable_subregs_hasher::hash, simplifiable_subregs)
2309         (record_subregs_of_mode): Likewise.
2310         * rtlhooks.c (gen_lowpart_general, gen_lowpart_if_possible): Likewise.
2311         * reload.c (operands_match_p): Handle 'p'.
2312         (find_reloads_subreg_address): Treat subreg offsets as poly_ints.
2313         * reload1.c (alter_reg, choose_reload_regs): Likewise.
2314         (compute_reload_subreg_offset): Likewise, and return an poly_int64.
2315         * simplify-rtx.c (simplify_truncation, simplify_binary_operation_1):
2316         (test_vector_ops_duplicate): Treat subreg offsets as poly_ints.
2317         (simplify_const_poly_int_tests<N>::run): Likewise.
2318         (simplify_subreg, simplify_gen_subreg): Take the subreg offset as
2319         a poly_uint64 rather than an unsigned int.
2320         * valtrack.c (debug_lowpart_subreg): Likewise.
2321         * var-tracking.c (var_lowpart): Likewise.
2322         (loc_cmp): Handle 'p'.
2324 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2325             Alan Hayward  <alan.hayward@arm.com>
2326             David Sherwood  <david.sherwood@arm.com>
2328         * ira.c (get_subreg_tracking_sizes): New function.
2329         (init_live_subregs): Take an integer size rather than a register.
2330         (build_insn_chain): Use get_subreg_tracking_sizes.  Update calls
2331         to init_live_subregs.
2333 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2334             Alan Hayward  <alan.hayward@arm.com>
2335             David Sherwood  <david.sherwood@arm.com>
2337         * expr.c (store_constructor_field): Change bitsize from a
2338         unsigned HOST_WIDE_INT to a poly_uint64 and bitpos from a
2339         HOST_WIDE_INT to a poly_int64.
2340         (store_constructor): Change size from a HOST_WIDE_INT to
2341         a poly_int64.
2342         (store_field): Likewise bitsize and bitpos.
2344 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2345             Alan Hayward  <alan.hayward@arm.com>
2346             David Sherwood  <david.sherwood@arm.com>
2348         * expmed.h (store_bit_field): Change bitregion_start and
2349         bitregion_end from unsigned HOST_WIDE_INT to poly_uint64.
2350         * expmed.c (adjust_bit_field_mem_for_reg, strict_volatile_bitfield_p)
2351         (store_bit_field_1, store_integral_bit_field, store_bit_field)
2352         (store_fixed_bit_field, store_split_bit_field): Likewise.
2353         * expr.c (store_constructor_field, store_field): Likewise.
2354         (optimize_bitfield_assignment_op): Likewise.  Make the same change
2355         to bitsize and bitpos.
2356         * machmode.h (bit_field_mode_iterator): Change m_bitregion_start
2357         and m_bitregion_end from HOST_WIDE_INT to poly_int64.  Make the
2358         same change in the constructor arguments.
2359         (get_best_mode): Change bitregion_start and bitregion_end from
2360         unsigned HOST_WIDE_INT to poly_uint64.
2361         * stor-layout.c (bit_field_mode_iterator::bit_field_mode_iterator):
2362         Change bitregion_start and bitregion_end from HOST_WIDE_INT to
2363         poly_int64.
2364         (bit_field_mode_iterator::next_mode): Update for new types
2365         of m_bitregion_start and m_bitregion_end.
2366         (get_best_mode): Change bitregion_start and bitregion_end from
2367         unsigned HOST_WIDE_INT to poly_uint64.
2369 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2370             Alan Hayward  <alan.hayward@arm.com>
2371             David Sherwood  <david.sherwood@arm.com>
2373         * rtl.h (simplify_gen_subreg): Add a temporary overload that
2374         accepts poly_uint64 offsets.
2375         * expmed.h (extract_bit_field): Take bitsize and bitnum as
2376         poly_uint64s rather than unsigned HOST_WIDE_INTs.
2377         * expmed.c (lowpart_bit_field_p): Likewise.
2378         (extract_bit_field_as_subreg): New function, split out from...
2379         (extract_bit_field_1): ...here.  Take bitsize and bitnum as
2380         poly_uint64s rather than unsigned HOST_WIDE_INTs.  For vector
2381         extractions, check that BITSIZE matches the size of the extracted
2382         value and that BITNUM is an exact multiple of that size.
2383         If all else fails, try forcing the value into memory if
2384         BITNUM is variable, and adjusting the address so that the
2385         offset is constant.  Split the part that can only handle constant
2386         bitsize and bitnum out into...
2387         (extract_integral_bit_field): ...this new function.
2388         (extract_bit_field): Take bitsize and bitnum as poly_uint64s
2389         rather than unsigned HOST_WIDE_INTs.
2391 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2392             Alan Hayward  <alan.hayward@arm.com>
2393             David Sherwood  <david.sherwood@arm.com>
2395         * expmed.h (store_bit_field): Take bitsize and bitnum as
2396         poly_uint64s rather than unsigned HOST_WIDE_INTs.
2397         * expmed.c (simple_mem_bitfield_p): Likewise.  Add a parameter
2398         that returns the byte size.
2399         (store_bit_field_1): Take bitsize and bitnum as
2400         poly_uint64s rather than unsigned HOST_WIDE_INTs.  Update call
2401         to simple_mem_bitfield_p.  Split the part that can only handle
2402         constant bitsize and bitnum out into...
2403         (store_integral_bit_field): ...this new function.
2404         (store_bit_field): Take bitsize and bitnum as poly_uint64s rather
2405         than unsigned HOST_WIDE_INTs.
2406         (extract_bit_field_1): Update call to simple_mem_bitfield_p.
2408 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2409             Alan Hayward  <alan.hayward@arm.com>
2410             David Sherwood  <david.sherwood@arm.com>
2412         * lra-int.h (lra_reg): Change offset from int to poly_int64.
2413         (lra_insn_recog_data): Change sp_offset from HOST_WIDE_INT
2414         to poly_int64.
2415         (lra_eliminate_regs_1, eliminate_regs_in_insn): Change
2416         update_sp_offset from a HOST_WIDE_INT to a poly_int64.
2417         (lra_update_reg_val_offset, lra_reg_val_equal_p): Take the
2418         offset as a poly_int64 rather than an int.
2419         * lra-assigns.c (find_hard_regno_for_1): Handle poly_int64 offsets.
2420         (setup_live_pseudos_and_spill_after_risky_transforms): Likewise.
2421         * lra-constraints.c (equiv_address_substitution): Track offsets
2422         as poly_int64s.
2423         (emit_inc): Check poly_int_rtx_p instead of CONST_INT_P.
2424         (curr_insn_transform): Handle the new form of sp_offset.
2425         * lra-eliminations.c (lra_elim_table): Change previous_offset
2426         and offset from HOST_WIDE_INT to poly_int64.
2427         (print_elim_table, update_reg_eliminate): Update accordingly.
2428         (self_elim_offsets): Change from HOST_WIDE_INT to poly_int64_pod.
2429         (get_elimination): Update accordingly.
2430         (form_sum): Check poly_int_rtx_p instead of CONST_INT_P.
2431         (lra_eliminate_regs_1, eliminate_regs_in_insn): Change
2432         update_sp_offset from a HOST_WIDE_INT to a poly_int64.  Handle
2433         poly_int64 offsets generally.
2434         (curr_sp_change): Change from HOST_WIDE_INT to poly_int64.
2435         (mark_not_eliminable, init_elimination): Update accordingly.
2436         (remove_reg_equal_offset_note): Return a bool and pass the new
2437         offset back by pointer as a poly_int64.
2438         * lra-remat.c (change_sp_offset): Take sp_offset as a poly_int64
2439         rather than a HOST_WIDE_INT.
2440         (do_remat): Track offsets poly_int64s.
2441         * lra.c (lra_update_insn_recog_data, setup_sp_offset): Likewise.
2443 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2444             Alan Hayward  <alan.hayward@arm.com>
2445             David Sherwood  <david.sherwood@arm.com>
2447         * rtl.h (mem_attrs): Add a default constructor.  Change size and
2448         offset from HOST_WIDE_INT to poly_int64.
2449         * emit-rtl.h (set_mem_offset, set_mem_size, adjust_address_1)
2450         (adjust_automodify_address_1, set_mem_attributes_minus_bitpos)
2451         (widen_memory_access): Take the sizes and offsets as poly_int64s
2452         rather than HOST_WIDE_INTs.
2453         * alias.c (ao_ref_from_mem): Handle the new form of MEM_OFFSET.
2454         (offset_overlap_p): Take poly_int64s rather than HOST_WIDE_INTs
2455         and ints.
2456         (adjust_offset_for_component_ref): Change the offset from a
2457         HOST_WIDE_INT to a poly_int64.
2458         (nonoverlapping_memrefs_p): Track polynomial offsets and sizes.
2459         * cfgcleanup.c (merge_memattrs): Update after mem_attrs changes.
2460         * dce.c (find_call_stack_args): Likewise.
2461         * dse.c (record_store): Likewise.
2462         * dwarf2out.c (tls_mem_loc_descriptor, dw_sra_loc_expr): Likewise.
2463         * print-rtl.c (rtx_writer::print_rtx): Likewise.
2464         * read-rtl-function.c (test_loading_mem): Likewise.
2465         * rtlanal.c (may_trap_p_1): Likewise.
2466         * simplify-rtx.c (delegitimize_mem_from_attrs): Likewise.
2467         * var-tracking.c (int_mem_offset, track_expr_p): Likewise.
2468         * emit-rtl.c (mem_attrs_eq_p, get_mem_align_offset): Likewise.
2469         (mem_attrs::mem_attrs): New function.
2470         (set_mem_attributes_minus_bitpos): Change bitpos from a
2471         HOST_WIDE_INT to poly_int64.
2472         (set_mem_alias_set, set_mem_addr_space, set_mem_align, set_mem_expr)
2473         (clear_mem_offset, clear_mem_size, change_address)
2474         (get_spill_slot_decl, set_mem_attrs_for_spill): Directly
2475         initialize mem_attrs.
2476         (set_mem_offset, set_mem_size, adjust_address_1)
2477         (adjust_automodify_address_1, offset_address, widen_memory_access):
2478         Likewise.  Take poly_int64s rather than HOST_WIDE_INT.
2480 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2481             Alan Hayward  <alan.hayward@arm.com>
2482             David Sherwood  <david.sherwood@arm.com>
2484         * rtlanal.c (rtx_addr_can_trap_p_1): Take the offset and size
2485         as poly_int64s rather than HOST_WIDE_INTs.  Use a size of -1
2486         rather than 0 to represent an unknown size.  Assert that the size
2487         is known when the mode isn't BLKmode.
2488         (may_trap_p_1): Use -1 for unknown sizes.
2489         (rtx_addr_can_trap_p): Likewise.  Pass BLKmode rather than VOIDmode.
2491 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2492             Alan Hayward  <alan.hayward@arm.com>
2493             David Sherwood  <david.sherwood@arm.com>
2495         * dse.c (store_info): Change offset and width from HOST_WIDE_INT
2496         to poly_int64.  Update commentary for positions_needed.large.
2497         (read_info_type): Change offset and width from HOST_WIDE_INT
2498         to poly_int64.
2499         (set_usage_bits): Likewise.
2500         (canon_address): Return the offset as a poly_int64 rather than
2501         a HOST_WIDE_INT.  Use strip_offset_and_add.
2502         (set_all_positions_unneeded, any_positions_needed_p): Use
2503         positions_needed.large to track stores with non-constant widths.
2504         (all_positions_needed_p): Likewise.  Take the offset and width
2505         as poly_int64s rather than ints.  Assert that rhs is nonnull.
2506         (record_store): Cope with non-constant offsets and widths.
2507         Nullify the rhs of an earlier store if we can't tell which bytes
2508         of it are needed.
2509         (find_shift_sequence): Take the access_size and shift as poly_int64s
2510         rather than ints.
2511         (get_stored_val): Take the read_offset and read_width as poly_int64s
2512         rather than HOST_WIDE_INTs.
2513         (check_mem_read_rtx, scan_stores, scan_reads, dse_step5): Handle
2514         non-constant offsets and widths.
2516 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2517             Alan Hayward  <alan.hayward@arm.com>
2518             David Sherwood  <david.sherwood@arm.com>
2520         * inchash.h (inchash::hash::add_poly_int): New function.
2521         * tree-ssa-alias.h (ao_ref::offset, ao_ref::size, ao_ref::max_size):
2522         Use poly_int64 rather than HOST_WIDE_INT.
2523         (ao_ref::max_size_known_p): New function.
2524         * tree-ssa-sccvn.h (vn_reference_op_struct::off): Use poly_int64_pod
2525         rather than HOST_WIDE_INT.
2526         * tree-ssa-alias.c (ao_ref_base): Apply get_ref_base_and_extent
2527         to temporaries until its interface is adjusted to match.
2528         (ao_ref_init_from_ptr_and_size): Handle polynomial offsets and sizes.
2529         (aliasing_component_refs_p, decl_refs_may_alias_p)
2530         (indirect_ref_may_alias_decl_p, indirect_refs_may_alias_p): Take
2531         the offsets and max_sizes as poly_int64s instead of HOST_WIDE_INTs.
2532         (refs_may_alias_p_1, stmt_kills_ref_p): Adjust for changes to
2533         ao_ref fields.
2534         * alias.c (ao_ref_from_mem): Likewise.
2535         * tree-ssa-dce.c (mark_aliased_reaching_defs_necessary_1): Likewise.
2536         * tree-ssa-dse.c (valid_ao_ref_for_dse, normalize_ref)
2537         (clear_bytes_written_by, setup_live_bytes_from_ref, compute_trims)
2538         (maybe_trim_complex_store, maybe_trim_constructor_store)
2539         (live_bytes_read, dse_classify_store): Likewise.
2540         * tree-ssa-sccvn.c (vn_reference_compute_hash, vn_reference_eq):
2541         (copy_reference_ops_from_ref, ao_ref_init_from_vn_reference)
2542         (fully_constant_vn_reference_p, valueize_refs_1): Likewise.
2543         (vn_reference_lookup_3): Likewise.
2544         * tree-ssa-uninit.c (warn_uninitialized_vars): Likewise.
2546 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2547             Alan Hayward  <alan.hayward@arm.com>
2548             David Sherwood  <david.sherwood@arm.com>
2550         * tree-ssa-alias.c (indirect_ref_may_alias_decl_p)
2551         (indirect_refs_may_alias_p): Use ranges_may_overlap_p
2552         instead of ranges_overlap_p.
2554 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2555             Alan Hayward  <alan.hayward@arm.com>
2556             David Sherwood  <david.sherwood@arm.com>
2558         * tree-ssa-alias.c (same_addr_size_stores_p): Take the offsets and
2559         sizes as poly_int64s rather than HOST_WIDE_INTs.
2561 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2562             Alan Hayward  <alan.hayward@arm.com>
2563             David Sherwood  <david.sherwood@arm.com>
2565         * gimple-fold.h (fold_ctor_reference): Take the offset and size
2566         as poly_uint64 rather than unsigned HOST_WIDE_INT.
2567         * gimple-fold.c (fold_ctor_reference): Likewise.
2569 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2570             Alan Hayward  <alan.hayward@arm.com>
2571             David Sherwood  <david.sherwood@arm.com>
2573         * target.def (dwarf_poly_indeterminate_value): New hook.
2574         * targhooks.h (default_dwarf_poly_indeterminate_value): Declare.
2575         * targhooks.c (default_dwarf_poly_indeterminate_value): New function.
2576         * doc/tm.texi.in (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Document.
2577         * doc/tm.texi: Regenerate.
2578         * dwarf2out.h (build_cfa_loc, build_cfa_aligned_loc): Take the
2579         offset as a poly_int64.
2580         * dwarf2out.c (new_reg_loc_descr): Move later in file.  Take the
2581         offset as a poly_int64.
2582         (loc_descr_plus_const, loc_list_plus_const, build_cfa_aligned_loc):
2583         Take the offset as a poly_int64.
2584         (build_cfa_loc): Likewise.  Use loc_descr_plus_const.
2585         (frame_pointer_fb_offset): Change to a poly_int64.
2586         (int_loc_descriptor): Take the offset as a poly_int64.  Use
2587         targetm.dwarf_poly_indeterminate_value for polynomial offsets.
2588         (based_loc_descr): Take the offset as a poly_int64.
2589         Use strip_offset_and_add to handle (plus X (const)).
2590         Use new_reg_loc_descr instead of an open-coded version of the
2591         previous implementation.
2592         (mem_loc_descriptor): Handle CONST_POLY_INT.
2593         (compute_frame_pointer_to_fb_displacement): Take the offset as a
2594         poly_int64.  Use strip_offset_and_add to handle (plus X (const)).
2596 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2597             Alan Hayward  <alan.hayward@arm.com>
2598             David Sherwood  <david.sherwood@arm.com>
2600         * rtl.h (reg_attrs::offset): Change from HOST_WIDE_INT to poly_int64.
2601         (gen_rtx_REG_offset): Take the offset as a poly_int64.
2602         * inchash.h (inchash::hash::add_poly_hwi): New function.
2603         * gengtype.c (main): Register poly_int64.
2604         * emit-rtl.c (reg_attr_hasher::hash): Use inchash.  Treat the
2605         offset as a poly_int.
2606         (reg_attr_hasher::equal): Use must_eq to compare offsets.
2607         (get_reg_attrs, update_reg_offset, gen_rtx_REG_offset): Take the
2608         offset as a poly_int64.
2609         (set_reg_attrs_from_value): Treat the offset as a poly_int64.
2610         * print-rtl.c (print_poly_int): New function.
2611         (rtx_writer::print_rtx_operand_code_r): Treat REG_OFFSET as
2612         a poly_int.
2613         * var-tracking.c (track_offset_p, get_tracked_reg_offset): New
2614         functions.
2615         (var_reg_set, var_reg_delete_and_set, var_reg_delete): Use them.
2616         (same_variable_part_p, track_loc_p): Take the offset as a poly_int64.
2617         (vt_get_decl_and_offset): Return the offset as a poly_int64.
2618         Enforce track_offset_p for parts of a PARALLEL.
2619         (vt_add_function_parameter): Use const_offset for the final
2620         offset to track.  Use get_tracked_reg_offset for the parts
2621         of a PARALLEL.
2623 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2624             Alan Hayward  <alan.hayward@arm.com>
2625             David Sherwood  <david.sherwood@arm.com>
2627         * target.def (truly_noop_truncation): Take poly_uint64s instead of
2628         unsigned ints.  Change default to hook_bool_puint64_puint64_true.
2629         * doc/tm.texi: Regenerate.
2630         * hooks.h (hook_bool_uint_uint_true): Delete.
2631         (hook_bool_puint64_puint64_true): Declare.
2632         * hooks.c (hook_bool_uint_uint_true): Delete.
2633         (hook_bool_puint64_puint64_true): New function.
2634         * config/mips/mips.c (mips_truly_noop_truncation): Take poly_uint64s
2635         instead of unsigned ints.
2636         * config/spu/spu.c (spu_truly_noop_truncation): Likewise.
2637         * config/tilegx/tilegx.c (tilegx_truly_noop_truncation): Likewise.
2639 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2640             Alan Hayward  <alan.hayward@arm.com>
2641             David Sherwood  <david.sherwood@arm.com>
2643         * optabs.h (expand_operand): Add an int_value field.
2644         (create_expand_operand): Add an int_value parameter and use it
2645         to initialize the new expand_operand field.
2646         (create_integer_operand): Replace with a declaration of a function
2647         that accepts poly_int64s.  Move the implementation to...
2648         * optabs.c (create_integer_operand): ...here.
2649         (maybe_legitimize_operand): For EXPAND_INTEGER, check whether
2650         the mode preserves the value of int_value, instead of calling
2651         const_int_operand on the rtx.  Use gen_int_mode to generate
2652         the new rtx.
2654 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2655             Alan Hayward  <alan.hayward@arm.com>
2656             David Sherwood  <david.sherwood@arm.com>
2658         * dumpfile.h (dump_dec): Declare.
2659         * dumpfile.c (dump_dec): New function.
2660         * pretty-print.h (pp_wide_integer): Turn into a function and
2661         declare a poly_int version.
2662         * pretty-print.c (pp_wide_integer): New function for poly_ints.
2664 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2665             Alan Hayward  <alan.hayward@arm.com>
2666             David Sherwood  <david.sherwood@arm.com>
2668         * doc/generic.texi (POLY_INT_CST): Document.
2669         * tree.def (POLY_INT_CST): New tree code.
2670         * treestruct.def (TS_POLY_INT_CST): New tree layout.
2671         * tree-core.h (tree_poly_int_cst): New struct.
2672         (tree_node): Add a poly_int_cst field.
2673         * tree.h (POLY_INT_CST_P, POLY_INT_CST_COEFF): New macros.
2674         (wide_int_to_tree, force_fit_type): Take a poly_wide_int_ref
2675         instead of a wide_int_ref.
2676         (build_int_cst, build_int_cst_type): Take a poly_int64 instead
2677         of a HOST_WIDE_INT.
2678         (build_int_cstu, build_array_type_nelts): Take a poly_uint64
2679         instead of an unsigned HOST_WIDE_INT.
2680         (build_poly_int_cst, tree_fits_poly_int64_p, tree_fits_poly_uint64_p)
2681         (ptrdiff_tree_p): Declare.
2682         (tree_to_poly_int64, tree_to_poly_uint64): Likewise.  Provide
2683         extern inline implementations if the target doesn't use POLY_INT_CST.
2684         (poly_int_tree_p): New function.
2685         (wi::unextended_tree): New class.
2686         (wi::int_traits <unextended_tree>): New override.
2687         (wi::extended_tree): Add a default constructor.
2688         (wi::extended_tree::get_tree): New function.
2689         (wi::widest_extended_tree, wi::offset_extended_tree): New typedefs.
2690         (wi::tree_to_widest_ref, wi::tree_to_offset_ref): Use them.
2691         (wi::tree_to_poly_widest_ref, wi::tree_to_poly_offset_ref)
2692         (wi::tree_to_poly_wide_ref): New typedefs.
2693         (wi::ints_for): Provide overloads for extended_tree and
2694         unextended_tree.
2695         (poly_int_cst_value, wi::to_poly_widest, wi::to_poly_offset)
2696         (wi::to_wide): New functions.
2697         (wi::fits_to_boolean_p, wi::fits_to_tree_p): Handle poly_ints.
2698         * tree.c (poly_int_cst_hasher): New struct.
2699         (poly_int_cst_hash_table): New variable.
2700         (tree_node_structure_for_code, tree_code_size, simple_cst_equal)
2701         (valid_constant_size_p, add_expr, drop_tree_overflow): Handle
2702         POLY_INT_CST.
2703         (initialize_tree_contains_struct): Handle TS_POLY_INT_CST.
2704         (init_ttree): Initialize poly_int_cst_hash_table.
2705         (build_int_cst, build_int_cst_type, build_invariant_address): Take
2706         a poly_int64 instead of a HOST_WIDE_INT.
2707         (build_int_cstu, build_array_type_nelts): Take a poly_uint64
2708         instead of an unsigned HOST_WIDE_INT.
2709         (wide_int_to_tree): Rename to...
2710         (wide_int_to_tree_1): ...this.
2711         (build_new_poly_int_cst, build_poly_int_cst): New functions.
2712         (force_fit_type): Take a poly_wide_int_ref instead of a wide_int_ref.
2713         (wide_int_to_tree): New function that takes a poly_wide_int_ref.
2714         (ptrdiff_tree_p, tree_to_poly_int64, tree_to_poly_uint64)
2715         (tree_fits_poly_int64_p, tree_fits_poly_uint64_p): New functions.
2716         * lto-streamer-out.c (DFS::DFS_write_tree_body, hash_tree): Handle
2717         TS_POLY_INT_CST.
2718         * tree-streamer-in.c (lto_input_ts_poly_tree_pointers): Likewise.
2719         (streamer_read_tree_body): Likewise.
2720         * tree-streamer-out.c (write_ts_poly_tree_pointers): Likewise.
2721         (streamer_write_tree_body): Likewise.
2722         * tree-streamer.c (streamer_check_handled_ts_structures): Likewise.
2723         * asan.c (asan_protect_global): Require the size to be an INTEGER_CST.
2724         * cfgexpand.c (expand_debug_expr): Handle POLY_INT_CST.
2725         * expr.c (expand_expr_real_1, const_vector_from_tree): Likewise.
2726         * gimple-expr.h (is_gimple_constant): Likewise.
2727         * gimplify.c (maybe_with_size_expr): Likewise.
2728         * print-tree.c (print_node): Likewise.
2729         * tree-data-ref.c (data_ref_compare_tree): Likewise.
2730         * tree-pretty-print.c (dump_generic_node): Likewise.
2731         * tree-ssa-address.c (addr_for_mem_ref): Likewise.
2732         * tree-vect-data-refs.c (dr_group_sort_cmp): Likewise.
2733         * tree-vrp.c (compare_values_warnv): Likewise.
2734         * tree-ssa-loop-ivopts.c (determine_base_object, constant_multiple_of)
2735         (get_loop_invariant_expr, add_candidate_1, get_computation_aff_1)
2736         (force_expr_to_var_cost): Likewise.
2737         * tree-ssa-loop.c (for_each_index): Likewise.
2738         * fold-const.h (build_invariant_address, size_int_kind): Take a
2739         poly_int64 instead of a HOST_WIDE_INT.
2740         * fold-const.c (fold_negate_expr_1, const_binop, const_unop)
2741         (fold_convert_const, multiple_of_p, fold_negate_const): Handle
2742         POLY_INT_CST.
2743         (size_binop_loc): Likewise.  Allow int_const_binop_1 to fail.
2744         (int_const_binop_2): New function, split out from...
2745         (int_const_binop_1): ...here.  Handle POLY_INT_CST.
2746         (size_int_kind): Take a poly_int64 instead of a HOST_WIDE_INT.
2747         * expmed.c (make_tree): Handle CONST_POLY_INT_P.
2748         * gimple-ssa-strength-reduction.c (slsr_process_add)
2749         (slsr_process_mul): Check for INTEGER_CSTs before using them
2750         as candidates.
2751         * stor-layout.c (bits_from_bytes): New function.
2752         (bit_from_pos): Use it.
2753         (layout_type): Likewise.  For vectors, multiply the TYPE_SIZE_UNIT
2754         by BITS_PER_UNIT to get the TYPE_SIZE.
2755         * tree-cfg.c (verify_expr, verify_types_in_gimple_reference): Allow
2756         MEM_REF and TARGET_MEM_REF offsets to be a POLY_INT_CST.
2758 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2759             Alan Hayward  <alan.hayward@arm.com>
2760             David Sherwood  <david.sherwood@arm.com>
2762         * doc/rtl.texi (const_poly_int): Document.  Also document the
2763         rtl sharing behavior.
2764         * gengenrtl.c (excluded_rtx): Return true for CONST_POLY_INT.
2765         * rtl.h (const_poly_int_def): New struct.
2766         (rtx_def::u): Add a cpi field.
2767         (CASE_CONST_UNIQUE, CASE_CONST_ANY): Add CONST_POLY_INT.
2768         (CONST_POLY_INT_P, CONST_POLY_INT_COEFFS): New macros.
2769         (wi::rtx_to_poly_wide_ref): New typedef
2770         (const_poly_int_value, wi::to_poly_wide, rtx_to_poly_int64)
2771         (poly_int_rtx_p): New functions.
2772         (trunc_int_for_mode): Declare a poly_int64 version.
2773         (plus_constant): Take a poly_int64 instead of a HOST_WIDE_INT.
2774         (immed_wide_int_const): Take a poly_wide_int_ref rather than
2775         a wide_int_ref.
2776         (strip_offset): Declare.
2777         (strip_offset_and_add): New function.
2778         * rtl.def (CONST_POLY_INT): New rtx code.
2779         * rtl.c (rtx_size): Handle CONST_POLY_INT.
2780         (shared_const_p): Use poly_int_rtx_p.
2781         * emit-rtl.h (gen_int_mode): Take a poly_int64 instead of a
2782         HOST_WIDE_INT.
2783         (gen_int_shift_amount): Likewise.
2784         * emit-rtl.c (const_poly_int_hasher): New class.
2785         (const_poly_int_htab): New variable.
2786         (init_emit_once): Initialize it when NUM_POLY_INT_COEFFS > 1.
2787         (const_poly_int_hasher::hash): New function.
2788         (const_poly_int_hasher::equal): Likewise.
2789         (gen_int_mode): Take a poly_int64 instead of a HOST_WIDE_INT.
2790         (immed_wide_int_const): Rename to...
2791         (immed_wide_int_const_1): ...this and make static.
2792         (immed_wide_int_const): New function, taking a poly_wide_int_ref
2793         instead of a wide_int_ref.
2794         (gen_int_shift_amount): Take a poly_int64 instead of a HOST_WIDE_INT.
2795         (gen_lowpart_common): Handle CONST_POLY_INT.
2796         * cse.c (hash_rtx_cb, equiv_constant): Likewise.
2797         * cselib.c (cselib_hash_rtx): Likewise.
2798         * dwarf2out.c (const_ok_for_output_1): Likewise.
2799         * expr.c (convert_modes): Likewise.
2800         * print-rtl.c (rtx_writer::print_rtx, print_value): Likewise.
2801         * rtlhash.c (add_rtx): Likewise.
2802         * explow.c (trunc_int_for_mode): Add a poly_int64 version.
2803         (plus_constant): Take a poly_int64 instead of a HOST_WIDE_INT.
2804         Handle existing CONST_POLY_INT rtxes.
2805         * expmed.h (expand_shift): Take a poly_int64 instead of a
2806         HOST_WIDE_INT.
2807         * expmed.c (expand_shift): Likewise.
2808         * rtlanal.c (strip_offset): New function.
2809         (commutative_operand_precedence): Give CONST_POLY_INT the same
2810         precedence as CONST_DOUBLE and put CONST_WIDE_INT between that
2811         and CONST_INT.
2812         * rtl-tests.c (const_poly_int_tests): New struct.
2813         (rtl_tests_c_tests): Use it.
2814         * simplify-rtx.c (simplify_const_unary_operation): Handle
2815         CONST_POLY_INT.
2816         (simplify_const_binary_operation): Likewise.
2817         (simplify_binary_operation_1): Fold additions of symbolic constants
2818         and CONST_POLY_INTs.
2819         (simplify_subreg): Handle extensions and truncations of
2820         CONST_POLY_INTs.
2821         (simplify_const_poly_int_tests): New struct.
2822         (simplify_rtx_c_tests): Use it.
2823         * wide-int.h (storage_ref): Add default constructor.
2824         (wide_int_ref_storage): Likewise.
2825         (trailing_wide_ints): Use GTY((user)).
2826         (trailing_wide_ints::operator[]): Add a const version.
2827         (trailing_wide_ints::get_precision): New function.
2828         (trailing_wide_ints::extra_size): Likewise.
2830 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2831             Alan Hayward  <alan.hayward@arm.com>
2832             David Sherwood  <david.sherwood@arm.com>
2834         * emit-rtl.h (gen_int_shift_amount): Declare.
2835         * emit-rtl.c (gen_int_shift_amount): New function.
2836         * asan.c (asan_emit_stack_protection): Use gen_int_shift_amount
2837         instead of GEN_INT.
2838         * calls.c (shift_return_value): Likewise.
2839         * cse.c (fold_rtx): Likewise.
2840         * dse.c (find_shift_sequence): Likewise.
2841         * expmed.c (init_expmed_one_mode, store_bit_field_1, expand_shift_1)
2842         (expand_shift, expand_smod_pow2): Likewise.
2843         * lower-subreg.c (shift_cost): Likewise.
2844         * optabs.c (expand_superword_shift, expand_doubleword_mult)
2845         (expand_unop, expand_binop, shift_amt_for_vec_perm_mask)
2846         (expand_vec_perm_var): Likewise.
2847         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
2848         (simplify_binary_operation_1): Likewise.
2849         * combine.c (try_combine, find_split_point, force_int_to_mode)
2850         (simplify_shift_const_1, simplify_shift_const): Likewise.
2851         (change_zero_ext): Likewise.  Use simplify_gen_binary.
2853 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2855         * poly-int.h (multiple_p): Fix handling of two non-poly_ints.
2857 2017-12-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2859         * doc/invoke.texi (ARM Options): Document accepted extension options
2860         for -march=armv8.3-a.
2862 2017-12-20  Richard Earnshaw  <rearnsha@arm.com>
2864         PR target/83105
2865         * config.gcc (arm*-*-linux*): When configured with --with-float=hard
2866         or --with-float=softfp, set the default CPU to arm10e.
2868 2017-12-20  Eric Botcazou  <ebotcazou@adacore.com>
2870         * config/visium/constraints.md (J, K, L): Use IN_RANGE macro.
2871         * config/visium/predicates.md (const_shift_operand): Likewise.
2872         * config/visium/visium.c (visium_legitimize_address): Fix oversight.
2873         (visium_legitimize_reload_address): Likewise.
2875 2017-12-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2877         PR target/82975
2878         * config/arm/arm.h (TEST_REGNO): Adjust comment as expected in r255830.
2880 2017-12-20  Jakub Jelinek  <jakub@redhat.com>
2882         PR c++/83490
2883         * calls.c (compute_argument_addresses): Ignore TYPE_EMPTY_P arguments.
2885 2017-12-20  Julia Koval  <julia.koval@intel.com>
2887         * common/config/i386/i386-common.c (OPTION_MASK_ISA_VPCLMULQDQ_SET,
2888         OPTION_MASK_ISA_VPCLMULQDQ_UNSET): New.
2889         (ix86_handle_option): Handle -mvpclmulqdq, move cx6 to flags2.
2890         * config.gcc: Include vpclmulqdqintrin.h.
2891         * config/i386/cpuid.h: Handle bit_VPCLMULQDQ.
2892         * config/i386/driver-i386.c (host_detect_local_cpu): Handle
2893         -mvpclmulqdq.
2894         * config/i386/i386-builtin.def (__builtin_ia32_vpclmulqdq_v2di,
2895         __builtin_ia32_vpclmulqdq_v4di, __builtin_ia32_vpclmulqdq_v8di): New.
2896         * config/i386/i386-c.c (__VPCLMULQDQ__): New.
2897         * config/i386/i386.c (isa2_opts): Add -mcx16.
2898         (isa_opts): Add -mpclmulqdq, remove -mcx16.
2899         (ix86_option_override_internal): Move mcx16 to flags2.
2900         (ix86_valid_target_attribute_inner_p): Add vpclmulqdq.
2901         (ix86_expand_builtin): Handle OPTION_MASK_ISA_VPCLMULQDQ.
2902         * config/i386/i386.h (TARGET_VPCLMULQDQ, TARGET_VPCLMULQDQ_P): New.
2903         * config/i386/i386.opt: Add mvpclmulqdq, move mcx16 to flags2.
2904         * config/i386/immintrin.h: Include vpclmulqdqintrin.h.
2905         * config/i386/sse.md (vpclmulqdq_<mode>): New pattern.
2906         * config/i386/vpclmulqdqintrin.h (_mm512_clmulepi64_epi128,
2907         _mm_clmulepi64_epi128, _mm256_clmulepi64_epi128): New intrinsics.
2908         * doc/invoke.texi: Add -mvpclmulqdq.
2910 2017-12-20  Tom de Vries  <tom@codesourcery.com>
2912         PR middle-end/83423
2913         * config/i386/i386.c (ix86_static_chain): Move
2914         DECL_STATIC_CHAIN test ...
2915         * calls.c (rtx_for_static_chain): ... here.  New function.
2916         * calls.h (rtx_for_static_chain): Declare.
2917         * builtins.c (expand_builtin_setjmp_receiver): Use rtx_for_static_chain
2918         instead of targetm.calls.static_chain.
2919         * df-scan.c (df_get_entry_block_def_set): Same.
2921 2017-12-19  Tom de Vries  <tom@codesourcery.com>
2923         PR tree-optimization/83493
2924         * graphite-isl-ast-to-gimple.c (translate_isl_ast_node_for): Unshare ub
2925         and lb.
2927 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2929         * gimple-ssa-sprintf.c (format_directive): Use inform_n instead of
2930         inform with hardcoded english plural handling.
2932 2017-12-18  Jeff Law  <law@redhat.com>
2934         PR tree-optimization/83477
2935         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis): For
2936         a non-virtual PHI, always push a new range.
2938 2017-12-19  Martin Sebor  <msebor@redhat.com>
2940         PR middle-end/77608
2941         * builtins.c (compute_objsize): Handle non-constant offsets.
2943 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2945         PR tree-optimization/83444
2946         * tree-ssa-strlen.c (strlen_check_and_optimize_stmt): For the
2947         character load case, if get_stridx on MEM_REF's operand doesn't
2948         look usable, retry with get_addr_stridx.
2950 2017-12-19  Alexandre Oliva  <aoliva@redhat.com>
2952         PR debug/83422
2953         * var-tracking.c (vt_debug_insns_local): Do not drop markers.
2954         (variable_tracking_main_1): Keep markers even when VTA fails.
2956         PR bootstrap/83396
2957         * cfgexpand.c (expand_gimple_basic_block): Expand label first,
2958         even if there are markers before it.
2959         * cfgrtl.c (rtl_verify_bb_layout): Reject DEBUG_INSNs outside BBs.
2961 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2963         * read-rtl.c (parse_reg_note_name): Replace Yoda conditions with
2964         typical order conditions.
2965         * sel-sched.c (extract_new_fences_from): Likewise.
2966         * config/visium/constraints.md (J, K, L): Likewise.
2967         * config/visium/predicates.md (const_shift_operand): Likewise.
2968         * config/visium/visium.c (visium_legitimize_address,
2969         visium_legitimize_reload_address): Likewise.
2970         * config/m68k/m68k.c (output_reg_adjust, emit_reg_adjust): Likewise.
2971         * config/arm/arm.c (arm_block_move_unaligned_straight): Likewise.
2972         * config/avr/constraints.md (Y01, Ym1, Y02, Ym2): Likewise.
2973         * config/avr/avr-log.c (avr_vdump, avr_log_set_avr_log,
2974         SET_DUMP_DETAIL): Likewise.
2975         * config/avr/predicates.md (const_8_16_24_operand): Likewise.
2976         * config/avr/avr.c (STR_PREFIX_P, avr_popcount_each_byte,
2977         avr_is_casesi_sequence, avr_casei_sequence_check_operands,
2978         avr_set_core_architecture, avr_set_current_function,
2979         avr_legitimize_reload_address, avr_asm_len, avr_print_operand,
2980         output_movqi, output_movsisf, avr_out_plus, avr_out_bitop,
2981         avr_out_fract, avr_adjust_insn_length, avr_encode_section_info,
2982         avr_2word_insn_p, output_reload_in_const, avr_has_nibble_0xf,
2983         avr_map_decompose, avr_fold_builtin): Likewise.
2984         * config/avr/driver-avr.c (avr_devicespecs_file): Likewise.
2985         * config/avr/gen-avr-mmcu-specs.c (str_prefix_p, print_mcu): Likewise.
2986         * config/i386/i386.c (ix86_parse_stringop_strategy_string): Likewise.
2987         * config/m32c/m32c-pragma.c (m32c_pragma_memregs): Likewise.
2988         * config/m32c/m32c.c (m32c_conditional_register_usage,
2989         m32c_address_cost): Likewise.
2990         * config/m32c/predicates.md (shiftcount_operand,
2991         longshiftcount_operand): Likewise.
2992         * config/iq2000/iq2000.c (iq2000_expand_prologue): Likewise.
2993         * config/nios2/nios2.c (nios2_handle_custom_fpu_insn_option,
2994         can_use_cdx_ldstw): Likewise.
2995         * config/nios2/nios2.h (CDX_REG_P): Likewise.
2996         * config/cr16/cr16.h (RETURN_ADDR_RTX, REGNO_MODE_OK_FOR_BASE_P):
2997         Likewise.
2998         * config/cr16/cr16.md (*mov<mode>_double): Likewise.
2999         * config/cr16/cr16.c (cr16_create_dwarf_for_multi_push): Likewise.
3000         * config/h8300/h8300.c (h8300_rtx_costs, get_shift_alg): Likewise.
3001         * config/vax/constraints.md (U06, U08, U16, CN6, S08, S16): Likewise.
3002         * config/vax/vax.c (adjacent_operands_p): Likewise.
3003         * config/ft32/constraints.md (L, b, KA): Likewise.
3004         * config/ft32/ft32.c (ft32_load_immediate, ft32_expand_prologue):
3005         Likewise.
3006         * cfgexpand.c (expand_stack_alignment): Likewise.
3007         * gcse.c (insert_expr_in_table): Likewise.
3008         * print-rtl.c (rtx_writer::print_rtx_operand_codes_E_and_V): Likewise.
3009         * cgraphunit.c (cgraph_node::expand): Likewise.
3010         * ira-build.c (setup_min_max_allocno_live_range_point): Likewise.
3011         * emit-rtl.c (add_insn): Likewise.
3012         * input.c (dump_location_info): Likewise.
3013         * passes.c (NEXT_PASS): Likewise.
3014         * read-rtl-function.c (parse_note_insn_name,
3015         function_reader::read_rtx_operand_r, function_reader::parse_mem_expr):
3016         Likewise.
3017         * sched-rgn.c (sched_rgn_init): Likewise.
3018         * diagnostic-show-locus.c (layout::show_ruler): Likewise.
3019         * combine.c (find_split_point, simplify_if_then_else, force_to_mode,
3020         if_then_else_cond, simplify_shift_const_1, simplify_comparison):
3021         Likewise.
3022         * explow.c (eliminate_constant_term): Likewise.
3023         * final.c (leaf_renumber_regs_insn): Likewise.
3024         * cfgrtl.c (print_rtl_with_bb): Likewise.
3025         * genhooks.c (emit_init_macros): Likewise.
3026         * poly-int.h (maybe_ne, maybe_le, maybe_lt): Likewise.
3027         * tree-data-ref.c (conflict_fn): Likewise.
3028         * selftest.c (assert_streq): Likewise.
3029         * expr.c (store_constructor_field, expand_expr_real_1): Likewise.
3030         * fold-const.c (fold_range_test, extract_muldiv_1, fold_truth_andor,
3031         fold_binary_loc, multiple_of_p): Likewise.
3032         * reload.c (push_reload, find_equiv_reg): Likewise.
3033         * et-forest.c (et_nca, et_below): Likewise.
3034         * dbxout.c (dbxout_symbol_location): Likewise.
3035         * reorg.c (relax_delay_slots): Likewise.
3036         * dojump.c (do_compare_rtx_and_jump): Likewise.
3037         * gengtype-parse.c (type): Likewise.
3038         * simplify-rtx.c (simplify_gen_ternary, simplify_gen_relational,
3039         simplify_const_relational_operation): Likewise.
3040         * reload1.c (do_output_reload): Likewise.
3041         * dumpfile.c (get_dump_file_info_by_switch): Likewise.
3042         * gengtype.c (type_for_name): Likewise.
3043         * gimple-ssa-sprintf.c (format_directive): Likewise.
3045 2017-12-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3047         PR target/82975
3048         * config/arm/arm.h (TEST_REGNO): Check reg_renumber is set before
3049         accessing it.  Adjust comment.
3051 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
3053         PR middle-end/81914
3054         * predict.c (zero_one_minusone): New function.
3055         (apply_return_prediction): Avoid return prediction for functions
3056         returning only -1, 0 and 1 values, unless they only return -1 and 0
3057         or 0 and 1.
3059 2017-12-19  Claudiu Zissulescu  <claziss@synopsys.com>
3061         * config/arc/arc.c (legitimate_scaled_address_p): Clean
3062         fall-through warning.
3063         (arc_compute_frame_size): Remove unused variables.
3064         (arc_print_operand): Fix fprintif format.
3065         (arc_can_follow_jump): Clean fall-through warning.
3067 2017-12-19  Marek Polacek  <polacek@redhat.com>
3069         PR c++/83489
3070         * config/i386/i386.c (init_cumulative_args): Don't check TYPE_EMPTY_P
3071         on an error node.
3073 2017-12-19  Claudiu Zissulescu  <claziss@synopsys.com>
3075         * config/arc/arc.c (overriderregs): New variable.
3076         (arc_override_options): Track fixed/call saved/call options.
3077         (arc_conditional_register_usage): Check against overrideregs
3078         variable whenever we change register properties.
3080 2017-12-19  Nathan Sidwell  <nathan@acm.org>
3082         * opts.c (finish_options): Don't prefix dump_base_name if it
3083         already contains directories.
3085 2017-12-19  Martin Liska  <mliska@suse.cz>
3087         PR rtl-optimization/82675
3088         * loop-unroll.c (unroll_loop_constant_iterations): Allocate one
3089         more element in sbitmap.
3091 2017-12-19  Paolo Carlini  <paolo.carlini@oracle.com>
3093         * gimplify.c (gimplify_expr): Use error_operand_p.
3095 2017-12-19  Sebastian Huber  <sebastian.huber@embedded-brains.de>
3097         PR target/83387
3098         * config/rs6000/t-rtems (MULTILIB_REQUIRED): Remove 64-bit soft-float
3099         multilib.
3101 2017-12-19  Daniel Cederman  <cederman@gaisler.com>
3103         * config/sparc/sparc.c (sparc_do_work_around_errata): Make sure
3104         the jump is to a label.
3106 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
3108         PR tree-optimization/83444
3109         * tree-ssa-strlen.c (strlen_check_and_optimize_stmt): Optimize
3110         character loads.
3112         PR ipa/82801
3113         PR ipa/83346
3114         * ipa-inline.c (flatten_remove_node_hook): New function.
3115         (ipa_inline): Keep only nodes with flatten attribute at the end of
3116         the array in the order from ipa_reverse_postorder, only walk that
3117         portion of array for flattening, if there is more than one such
3118         node, temporarily register a removal hook and ignore removed nodes.
3120         PR tree-optimization/80631
3121         * tree-vect-loop.c (vect_create_epilog_for_reduction): Compare
3122         induc_code against MAX_EXPR or MIN_EXPR instead of reduc_fn against
3123         IFN_REDUC_MAX or IFN_REDUC_MIN.
3125 2017-12-18  Jeff Law  <law@redhat.com>
3127         * tree-ssa-dom.c (record_equivalences_from_phis): Do not
3128         record symbolic equivalences from backedges in the CFG.
3130         Revert
3131         2017-11-19  Jeff Law  <law@redhat.com>
3133         * tree-ssa-dom.c (record_equivalences_from_phis): Fix handling
3134         of degenerates resulting from ignoring an edge.
3136 2017-12-18  Martin Sebor  <msebor@redhat.com>
3138         PR middle-end/83373
3139         PR tree-optimization/78450
3140         * tree-ssa-strlen.c (maybe_set_strlen_range): New function.
3141         (handle_builtin_strlen): Call it.
3143 2017-12-18  Segher Boessenkool  <segher@kernel.crashing.org>
3145         PR rtl-optimization/83424
3146         * rtlanal.c (dead_or_set_regno_p): Handle CLOBBER just like SET.
3148 2017-12-18  Marek Polacek  <polacek@redhat.com>
3150         PR middle-end/83463
3151         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
3152         Check if TYPE is INTEGRAL_TYPE_P before accessing its min/max
3153         values.
3155 2017-12-18  Claudiu Zissulescu  <claziss@synopsys.com>
3157         * config/arc/arc.md (maddsidi4, maddsidi4_split): Update pattern.
3158         (umaddsidi4, umaddsidi_split): Likewise.
3160 2017-12-18  Claudiu Zissulescu  <claziss@synopsys.com>
3162         * config/arc/arc.c (arc_legitimate_constant_p): Always check all
3163         constants.
3165 2017-12-18  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
3167         PR target/83420
3168         * config/s390/s390.c (s390_option_override): Avoid strncpy.
3170 2017-12-18  Richard Biener  <rguenther@suse.de>
3172         PR tree-optimization/81877
3173         * tree-ssa-loop-im.c (ref_indep_loop_p): Remove safelen parameters.
3174         (outermost_indep_loop): Adjust.
3175         (ref_indep_loop_p_1): Likewise.  Remove safelen handling again.
3176         (can_sm_ref_p): Adjust.
3178 2017-12-18  Richard Biener  <rguenther@suse.de>
3180         PR middle-end/77291
3181         * tree.c (array_at_struct_end_p): Return true if the underlying
3182         object has space for at least one element in excess of what
3183         the array domain specifies.
3185 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
3187         * doc/extend.texi (x86 Function Attributes): Reformat nocf_check
3188         example to avoid overfull hbox.
3189         * doc/invoke.texi (Option Summary): Add missing @gol.
3190         (C++ Dialect Options): Reformat -Wnoexcept-type example to avoid
3191         overfull hbox.
3193 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
3194             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3196         * doc/invoke.texi (Option Summary): Add -mverbose-cost-dump
3197         to AArch64 and ARM lists, plus missing -mflip-thumb for ARM.
3198         (AArch64 Options): Document -mverbose-cost-dump.
3199         (ARM Options): Likewise, plus -mflip-thumb.
3201 2017-12-17  Martin Sebor  <msebor@redhat.com>
3203         PR bootstrap/83446
3204         * gimple-ssa-warn-restrict.c
3205         (builtin_memref::offset_out_of_bounds): Correct the handling of
3206         anti-ranges.
3208 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
3210         * doc/invoke.texi (ARC Options): Add missing -mlra entry.
3212 2017-12-17  John David Anglin  <danglin@gcc.gnu.org>
3214         * config/pa/pa.c (pa_som_asm_init_sections): Fix comment.
3216 2017-12-17  Markus Trippelsdorf  <markus@trippelsdorf.de>
3218         * x86-tune-costs.h (skylake_cost, core_cost): Decrease r64 multiply
3219         latencies.
3221 2017-12-16  Sandra Loosemore  <sandra@codesourcery.com>
3223         * doc/invoke.texi: Fix some typos.
3225 2017-12-16  Martin Sebor  <msebor@redhat.com>
3227         PR tree-optimization/78918
3228         * Makefile.in (OBJS): Add gimple-ssa-warn-restrict.o.
3229         * builtins.c (check_sizes): Rename...
3230         (check_access): ...to this.  Rename function arguments for clarity.
3231         (check_memop_sizes): Adjust names.
3232         (expand_builtin_memchr, expand_builtin_memcpy): Same.
3233         (expand_builtin_memmove, expand_builtin_mempcpy): Same.
3234         (expand_builtin_strcat, expand_builtin_stpncpy): Same.
3235         (check_strncat_sizes, expand_builtin_strncat): Same.
3236         (expand_builtin_strncpy, expand_builtin_memset): Same.
3237         (expand_builtin_bzero, expand_builtin_memcmp): Same.
3238         (expand_builtin_memory_chk, maybe_emit_chk_warning): Same.
3239         (maybe_emit_sprintf_chk_warning): Same.
3240         (expand_builtin_strcpy): Adjust.
3241         (expand_builtin_stpcpy): Same.
3242         (expand_builtin_with_bounds): Detect out-of-bounds accesses
3243         in pointer-checking forms of memcpy, memmove, and mempcpy.
3244         (gcall_to_tree_minimal, max_object_size): Define new functions.
3245         * builtins.h (max_object_size): Declare.
3246         * calls.c (alloc_max_size): Call max_object_size instead of
3247         hardcoding ssizetype limit.
3248         (get_size_range): Handle new argument.
3249         * calls.h (get_size_range): Add a new argument.
3250         * cfgexpand.c (expand_call_stmt): Propagate no-warning bit.
3251         * doc/invoke.texi (-Wrestrict): Adjust, add example.
3252         * gimple-fold.c (gimple_fold_builtin_memory_op): Detect overlapping
3253         operations.
3254         (gimple_fold_builtin_memory_chk): Same.
3255         (gimple_fold_builtin_stxcpy_chk): New function.
3256         * gimple-ssa-warn-restrict.c: New source.
3257         * gimple-ssa-warn-restrict.h: New header.
3258         * gimple.c (gimple_build_call_from_tree): Propagate location.
3259         * passes.def (pass_warn_restrict): Add new pass.
3260         * tree-pass.h (make_pass_warn_restrict): Declare.
3261         * tree-ssa-strlen.c (handle_builtin_strcpy): Detect overlapping
3262         operations.
3263         (handle_builtin_strcat): Same.
3264         (strlen_optimize_stmt): Rename...
3265         (strlen_check_and_optimize_stmt): ...to this.  Handle strncat,
3266         stpncpy, strncpy, and their checking forms.
3268 2017-12-16  Jan Hubicka  <hubicka@ucw.cz>
3270         PR rtl-optimization/82849
3271         * modulo-sched.c (sms_schedule): Use get_estimated_loop_iterations_int
3272         and get_max_loop_iterations_int.
3274 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3275             Alan Hayward  <alan.hayward@arm.com>
3276             David Sherwood  <david.sherwood@arm.com>
3278         * machmode.h (mode_for_size, int_mode_for_size, float_mode_for_size)
3279         (smallest_mode_for_size, smallest_int_mode_for_size): Take the mode
3280         size as a poly_uint64.
3281         (mode_for_vector, mode_for_int_vector): Take the number of vector
3282         elements as a poly_uint64.
3283         * stor-layout.c (mode_for_size, smallest_mode_for_size): Take the mode
3284         size as a poly_uint64.
3285         (mode_for_vector, mode_for_int_vector): Take the number of vector
3286         elements as a poly_uint64.
3288 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3289             Alan Hayward  <alan.hayward@arm.com>
3290             David Sherwood  <david.sherwood@arm.com>
3292         * machmode.h (MACRO_MODE): New macro.
3293         * addresses.h (base_reg_class, ok_for_base_p_1): Use it.
3294         * targhooks.c (default_libcall_value, default_secondary_reload)
3295         (default_memory_move_cost, default_register_move_cost)
3296         (default_class_max_nregs): Likewise.
3298 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3299             Alan Hayward  <alan.hayward@arm.com>
3300             David Sherwood  <david.sherwood@arm.com>
3302         * doc/sourcebuild.texi: Document IN_TARGET_CODE.
3303         * genattrtab.c (write_header): Define IN_TARGET_CODE to 1 in the
3304         target C file.
3305         * genautomata.c (main): Likewise.
3306         * genconditions.c (write_header): Likewise.
3307         * genemit.c (main): Likewise.
3308         * genextract.c (print_header): Likewise.
3309         * genopinit.c (main): Likewise.
3310         * genoutput.c (output_prologue): Likewise.
3311         * genpeep.c (main): Likewise.
3312         * genpreds.c (write_insn_preds_c): Likewise.
3313         * genrecog.c (writer_header): Likewise.
3314         * config/aarch64/aarch64-builtins.c (IN_TARGET_CODE): Define.
3315         * config/aarch64/aarch64-c.c (IN_TARGET_CODE): Likewise.
3316         * config/aarch64/aarch64.c (IN_TARGET_CODE): Likewise.
3317         * config/aarch64/cortex-a57-fma-steering.c (IN_TARGET_CODE): Likewise.
3318         * config/aarch64/driver-aarch64.c (IN_TARGET_CODE): Likewise.
3319         * config/alpha/alpha.c (IN_TARGET_CODE): Likewise.
3320         * config/alpha/driver-alpha.c (IN_TARGET_CODE): Likewise.
3321         * config/arc/arc-c.c (IN_TARGET_CODE): Likewise.
3322         * config/arc/arc.c (IN_TARGET_CODE): Likewise.
3323         * config/arc/driver-arc.c (IN_TARGET_CODE): Likewise.
3324         * config/arm/aarch-common.c (IN_TARGET_CODE): Likewise.
3325         * config/arm/arm-builtins.c (IN_TARGET_CODE): Likewise.
3326         * config/arm/arm-c.c (IN_TARGET_CODE): Likewise.
3327         * config/arm/arm.c (IN_TARGET_CODE): Likewise.
3328         * config/arm/driver-arm.c (IN_TARGET_CODE): Likewise.
3329         * config/avr/avr-c.c (IN_TARGET_CODE): Likewise.
3330         * config/avr/avr-devices.c (IN_TARGET_CODE): Likewise.
3331         * config/avr/avr-log.c (IN_TARGET_CODE): Likewise.
3332         * config/avr/avr.c (IN_TARGET_CODE): Likewise.
3333         * config/avr/driver-avr.c (IN_TARGET_CODE): Likewise.
3334         * config/avr/gen-avr-mmcu-specs.c (IN_TARGET_CODE): Likewise.
3335         * config/bfin/bfin.c (IN_TARGET_CODE): Likewise.
3336         * config/c6x/c6x.c (IN_TARGET_CODE): Likewise.
3337         * config/cr16/cr16.c (IN_TARGET_CODE): Likewise.
3338         * config/cris/cris.c (IN_TARGET_CODE): Likewise.
3339         * config/darwin.c (IN_TARGET_CODE): Likewise.
3340         * config/epiphany/epiphany.c (IN_TARGET_CODE): Likewise.
3341         * config/epiphany/mode-switch-use.c (IN_TARGET_CODE): Likewise.
3342         * config/epiphany/resolve-sw-modes.c (IN_TARGET_CODE): Likewise.
3343         * config/fr30/fr30.c (IN_TARGET_CODE): Likewise.
3344         * config/frv/frv.c (IN_TARGET_CODE): Likewise.
3345         * config/ft32/ft32.c (IN_TARGET_CODE): Likewise.
3346         * config/h8300/h8300.c (IN_TARGET_CODE): Likewise.
3347         * config/i386/djgpp.c (IN_TARGET_CODE): Likewise.
3348         * config/i386/driver-i386.c (IN_TARGET_CODE): Likewise.
3349         * config/i386/driver-mingw32.c (IN_TARGET_CODE): Likewise.
3350         * config/i386/host-cygwin.c (IN_TARGET_CODE): Likewise.
3351         * config/i386/host-i386-darwin.c (IN_TARGET_CODE): Likewise.
3352         * config/i386/host-mingw32.c (IN_TARGET_CODE): Likewise.
3353         * config/i386/i386-c.c (IN_TARGET_CODE): Likewise.
3354         * config/i386/i386.c (IN_TARGET_CODE): Likewise.
3355         * config/i386/intelmic-mkoffload.c (IN_TARGET_CODE): Likewise.
3356         * config/i386/msformat-c.c (IN_TARGET_CODE): Likewise.
3357         * config/i386/winnt-cxx.c (IN_TARGET_CODE): Likewise.
3358         * config/i386/winnt-stubs.c (IN_TARGET_CODE): Likewise.
3359         * config/i386/winnt.c (IN_TARGET_CODE): Likewise.
3360         * config/i386/x86-tune-sched-atom.c (IN_TARGET_CODE): Likewise.
3361         * config/i386/x86-tune-sched-bd.c (IN_TARGET_CODE): Likewise.
3362         * config/i386/x86-tune-sched-core.c (IN_TARGET_CODE): Likewise.
3363         * config/i386/x86-tune-sched.c (IN_TARGET_CODE): Likewise.
3364         * config/ia64/ia64-c.c (IN_TARGET_CODE): Likewise.
3365         * config/ia64/ia64.c (IN_TARGET_CODE): Likewise.
3366         * config/iq2000/iq2000.c (IN_TARGET_CODE): Likewise.
3367         * config/lm32/lm32.c (IN_TARGET_CODE): Likewise.
3368         * config/m32c/m32c-pragma.c (IN_TARGET_CODE): Likewise.
3369         * config/m32c/m32c.c (IN_TARGET_CODE): Likewise.
3370         * config/m32r/m32r.c (IN_TARGET_CODE): Likewise.
3371         * config/m68k/m68k.c (IN_TARGET_CODE): Likewise.
3372         * config/mcore/mcore.c (IN_TARGET_CODE): Likewise.
3373         * config/microblaze/microblaze-c.c (IN_TARGET_CODE): Likewise.
3374         * config/microblaze/microblaze.c (IN_TARGET_CODE): Likewise.
3375         * config/mips/driver-native.c (IN_TARGET_CODE): Likewise.
3376         * config/mips/frame-header-opt.c (IN_TARGET_CODE): Likewise.
3377         * config/mips/mips.c (IN_TARGET_CODE): Likewise.
3378         * config/mmix/mmix.c (IN_TARGET_CODE): Likewise.
3379         * config/mn10300/mn10300.c (IN_TARGET_CODE): Likewise.
3380         * config/moxie/moxie.c (IN_TARGET_CODE): Likewise.
3381         * config/msp430/driver-msp430.c (IN_TARGET_CODE): Likewise.
3382         * config/msp430/msp430-c.c (IN_TARGET_CODE): Likewise.
3383         * config/msp430/msp430.c (IN_TARGET_CODE): Likewise.
3384         * config/nds32/nds32-cost.c (IN_TARGET_CODE): Likewise.
3385         * config/nds32/nds32-fp-as-gp.c (IN_TARGET_CODE): Likewise.
3386         * config/nds32/nds32-intrinsic.c (IN_TARGET_CODE): Likewise.
3387         * config/nds32/nds32-isr.c (IN_TARGET_CODE): Likewise.
3388         * config/nds32/nds32-md-auxiliary.c (IN_TARGET_CODE): Likewise.
3389         * config/nds32/nds32-memory-manipulation.c (IN_TARGET_CODE): Likewise.
3390         * config/nds32/nds32-pipelines-auxiliary.c (IN_TARGET_CODE): Likewise.
3391         * config/nds32/nds32-predicates.c (IN_TARGET_CODE): Likewise.
3392         * config/nds32/nds32.c (IN_TARGET_CODE): Likewise.
3393         * config/nios2/nios2.c (IN_TARGET_CODE): Likewise.
3394         * config/nvptx/mkoffload.c (IN_TARGET_CODE): Likewise.
3395         * config/nvptx/nvptx.c (IN_TARGET_CODE): Likewise.
3396         * config/pa/pa.c (IN_TARGET_CODE): Likewise.
3397         * config/pdp11/pdp11.c (IN_TARGET_CODE): Likewise.
3398         * config/powerpcspe/driver-powerpcspe.c (IN_TARGET_CODE): Likewise.
3399         * config/powerpcspe/host-darwin.c (IN_TARGET_CODE): Likewise.
3400         * config/powerpcspe/host-ppc64-darwin.c (IN_TARGET_CODE): Likewise.
3401         * config/powerpcspe/powerpcspe-c.c (IN_TARGET_CODE): Likewise.
3402         * config/powerpcspe/powerpcspe-linux.c (IN_TARGET_CODE): Likewise.
3403         * config/powerpcspe/powerpcspe.c (IN_TARGET_CODE): Likewise.
3404         * config/riscv/riscv-builtins.c (IN_TARGET_CODE): Likewise.
3405         * config/riscv/riscv-c.c (IN_TARGET_CODE): Likewise.
3406         * config/riscv/riscv.c (IN_TARGET_CODE): Likewise.
3407         * config/rl78/rl78-c.c (IN_TARGET_CODE): Likewise.
3408         * config/rl78/rl78.c (IN_TARGET_CODE): Likewise.
3409         * config/rs6000/driver-rs6000.c (IN_TARGET_CODE): Likewise.
3410         * config/rs6000/host-darwin.c (IN_TARGET_CODE): Likewise.
3411         * config/rs6000/host-ppc64-darwin.c (IN_TARGET_CODE): Likewise.
3412         * config/rs6000/rs6000-c.c (IN_TARGET_CODE): Likewise.
3413         * config/rs6000/rs6000-linux.c (IN_TARGET_CODE): Likewise.
3414         * config/rs6000/rs6000-p8swap.c (IN_TARGET_CODE): Likewise.
3415         * config/rs6000/rs6000-string.c (IN_TARGET_CODE): Likewise.
3416         * config/rs6000/rs6000.c (IN_TARGET_CODE): Likewise.
3417         * config/rx/rx.c (IN_TARGET_CODE): Likewise.
3418         * config/s390/driver-native.c (IN_TARGET_CODE): Likewise.
3419         * config/s390/s390-c.c (IN_TARGET_CODE): Likewise.
3420         * config/s390/s390.c (IN_TARGET_CODE): Likewise.
3421         * config/sh/sh-c.c (IN_TARGET_CODE): Likewise.
3422         * config/sh/sh-mem.cc (IN_TARGET_CODE): Likewise.
3423         * config/sh/sh.c (IN_TARGET_CODE): Likewise.
3424         * config/sh/sh_optimize_sett_clrt.cc (IN_TARGET_CODE): Likewise.
3425         * config/sh/sh_treg_combine.cc (IN_TARGET_CODE): Likewise.
3426         * config/sparc/driver-sparc.c (IN_TARGET_CODE): Likewise.
3427         * config/sparc/sparc-c.c (IN_TARGET_CODE): Likewise.
3428         * config/sparc/sparc.c (IN_TARGET_CODE): Likewise.
3429         * config/spu/spu-c.c (IN_TARGET_CODE): Likewise.
3430         * config/spu/spu.c (IN_TARGET_CODE): Likewise.
3431         * config/stormy16/stormy16.c (IN_TARGET_CODE): Likewise.
3432         * config/tilegx/mul-tables.c (IN_TARGET_CODE): Likewise.
3433         * config/tilegx/tilegx-c.c (IN_TARGET_CODE): Likewise.
3434         * config/tilegx/tilegx.c (IN_TARGET_CODE): Likewise.
3435         * config/tilepro/mul-tables.c (IN_TARGET_CODE): Likewise.
3436         * config/tilepro/tilepro-c.c (IN_TARGET_CODE): Likewise.
3437         * config/tilepro/tilepro.c (IN_TARGET_CODE): Likewise.
3438         * config/v850/v850-c.c (IN_TARGET_CODE): Likewise.
3439         * config/v850/v850.c (IN_TARGET_CODE): Likewise.
3440         * config/vax/vax.c (IN_TARGET_CODE): Likewise.
3441         * config/visium/visium.c (IN_TARGET_CODE): Likewise.
3442         * config/vms/vms-c.c (IN_TARGET_CODE): Likewise.
3443         * config/vms/vms-f.c (IN_TARGET_CODE): Likewise.
3444         * config/vms/vms.c (IN_TARGET_CODE): Likewise.
3445         * config/xtensa/xtensa.c (IN_TARGET_CODE): Likewise.
3447 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3448             Alan Hayward  <alan.hayward@arm.com>
3449             David Sherwood  <david.sherwood@arm.com>
3451         * tree-ssa-sccvn.c (vn_reference_lookup_3): Avoid repeated
3452         checks for MEM_REF.
3454 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3455             Alan Hayward  <alan.hayward@arm.com>
3456             David Sherwood  <david.sherwood@arm.com>
3458         * doc/generic.texi (VEC_SERIES_EXPR): Document.
3459         * doc/md.texi (vec_series@var{m}): Document.
3460         * tree.def (VEC_SERIES_EXPR): New tree code.
3461         * tree.h (build_vec_series): Declare.
3462         * tree.c (build_vec_series): New function.
3463         * cfgexpand.c (expand_debug_expr): Handle VEC_SERIES_EXPR.
3464         * tree-pretty-print.c (dump_generic_node): Likewise.
3465         * gimple-pretty-print.c (dump_binary_rhs): Likewise.
3466         * tree-inline.c (estimate_operator_cost): Likewise.
3467         * expr.c (expand_expr_real_2): Likewise.
3468         * optabs-tree.c (optab_for_tree_code): Likewise.
3469         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
3470         * fold-const.c (const_binop): Fold VEC_SERIES_EXPRs of constants.
3471         * expmed.c (make_tree): Handle VEC_SERIES.
3472         * optabs.def (vec_series_optab): New optab.
3473         * optabs.h (expand_vec_series_expr): Declare.
3474         * optabs.c (expand_vec_series_expr): New function.
3475         * tree-vect-generic.c (expand_vector_operations_1): Check that
3476         the operands also have vector type.
3478 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3479             Alan Hayward  <alan.hawyard@arm.com>
3480             David Sherwood  <david.sherwood@arm.com>
3482         * doc/generic.texi (VEC_DUPLICATE_EXPR): Document.
3483         (VEC_COND_EXPR): Add missing @tindex.
3484         * doc/md.texi (vec_duplicate@var{m}): Document.
3485         * tree.def (VEC_DUPLICATE_EXPR): New tree codes.
3486         * tree.c (build_vector_from_val): Add stubbed-out handling of
3487         variable-length vectors, using VEC_DUPLICATE_EXPR.
3488         (uniform_vector_p): Handle VEC_DUPLICATE_EXPR.
3489         * cfgexpand.c (expand_debug_expr): Likewise.
3490         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
3491         * tree-inline.c (estimate_operator_cost): Likewise.
3492         * tree-pretty-print.c (dump_generic_node): Likewise.
3493         * tree-vect-generic.c (ssa_uniform_vector_p): Likewise.
3494         * fold-const.c (const_unop): Fold VEC_DUPLICATE_EXPRs of a constant.
3495         (test_vec_duplicate_folding): New function.
3496         (fold_const_c_tests): Call it.
3497         * optabs.def (vec_duplicate_optab): New optab.
3498         * optabs-tree.c (optab_for_tree_code): Handle VEC_DUPLICATE_EXPR.
3499         * optabs.h (expand_vector_broadcast): Declare.
3500         * optabs.c (expand_vector_broadcast): Make non-static.  Try using
3501         vec_duplicate_optab.
3502         * expr.c (store_constructor): Try using vec_duplicate_optab for
3503         uniform vectors.
3504         (expand_expr_real_2): Handle VEC_DUPLICATE_EXPR.
3506 2017-12-15  Markus Trippelsdorf  <markus@trippelsdorf.de>
3508         PR target/83358
3509         * config/i386/x86-tune-costs.h (skylake_cost, core_cost): Increase
3510         div/mod latencies a bit.
3512 2017-12-15  Jeff Law  <law@redhat.com>
3514         PR tree-optimization/36550
3515         * tree-ssa-threadupdate.c (count_stmts_and_phis_in_block): New.
3516         (mark_threaded_blocks): Rewrite code to avoid block copying when
3517         optimizing for size.  Don't pessimize blocks which will be
3518         copied, but all the statements will be dead.
3520 2017-12-15  Alexandre Oliva  <aoliva@redhat.com>
3522         PR tree-optimization/81165
3523         * tree-ssa-threadupdate.c (uses_in_bb): New.
3524         (estimate_threading_killed_stmts): New.
3525         * tree-ssa-threadupdate.h (estimate_threading_killed_stmts): Prototype.
3526         * tree-ssa-threadedge.c
3527         (record_temporary_equivalences_from_stmts_at_dest): Expand limit
3528         when its hit.
3530 2017-12-15  Jeff Law  <law@redhat.com>
3532         PR tree-optimization/83410
3533         * tree-ssa-threadupdate.c (thread_block_1): Avoid certain jump
3534         threads when parallelizing loops.
3536 2017-12-15  Jakub Jelinek  <jakub@redhat.com>
3538         * tree-core.h (struct attribute_spec): Swap affects_type_identity and
3539         handler fields.
3540         * config/alpha/alpha.c (vms_attribute_table): Swap
3541         affects_type_identity and handler fields, adjust comments.
3542         * config/mips/mips.c (mips_attribute_table): Likewise.
3543         * config/visium/visium.c (visium_attribute_table): Likewise.
3544         * config/epiphany/epiphany.c (epiphany_attribute_table): Likewise.
3545         * config/microblaze/microblaze.c (microblaze_attribute_table):
3546         Likewise.
3547         * config/spu/spu.c (spu_attribute_table): Likewise.
3548         * config/mcore/mcore.c (mcore_attribute_table): Likewise.
3549         * config/arc/arc.c (arc_attribute_table): Likewise.
3550         * config/m68k/m68k.c (m68k_attribute_table): Likewise.
3551         * config/v850/v850.c (v850_handle_interrupt_attribute,
3552         v850_handle_data_area_attribute): Formatting fixes.
3553         (v850_attribute_table): Swap affects_type_identity and handler
3554         fields, adjust comments.
3555         * config/m32r/m32r.c (m32r_attribute_table): Likewise.
3556         * config/arm/arm.c (arm_attribute_table): Likewise.
3557         * config/avr/avr.c (avr_attribute_table): Likewise.
3558         * config/s390/s390.c (s390_attribute_table): Likewise.
3559         * config/sh/sh.c (sh_attribute_table): Likewise.
3560         * config/i386/i386.c (ix86_handle_cconv_attribute,
3561         ix86_handle_callee_pop_aggregate_return): Formatting fixes.
3562         (ix86_attribute_table): Swap affects_type_identity and handler
3563         fields, adjust comments.
3564         * config/i386/cygming.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
3565         * config/sparc/sparc.c (sparc_attribute_table): Likewise.
3566         * config/m32c/m32c.c (m32c_attribute_table): Likewise.
3567         * config/sol2.h (SOLARIS_ATTRIBUTE_TABLE): Likewise.
3568         * config/ia64/ia64.c (ia64_attribute_table): Likewise.
3569         * config/msp430/msp430.c (msp430_attribute_table): Likewise.
3570         * config/rx/rx.c (rx_attribute_table): Likewise.
3571         * config/cr16/cr16.c (cr16_attribute_table): Likewise.
3572         * config/h8300/h8300.c (h8300_attribute_table): Likewise.
3573         * config/nvptx/nvptx.c (nvptx_attribute_table): Likewise.
3574         * config/powerpcspe/powerpcspe.c (rs6000_attribute_table): Likewise.
3575         * config/darwin.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
3576         * config/stormy16/stormy16.c (xstormy16_attribute_table): Likewise.
3577         * config/bfin/bfin.c (bfin_attribute_table): Likewise.
3578         * config/rs6000/rs6000.c (rs6000_attribute_table): Likewise.
3579         * config/rl78/rl78.c (rl78_attribute_table): Likewise.
3580         * config/nds32/nds32.c (nds32_attribute_table): Likewise.
3581         * doc/plugins.texi (user_attr): Likewise.  Add NULL for
3582         exclude.
3583         * attribs.c (empty_attribute_table): Swap affects_type_identity and
3584         handler fields.
3585         (register_scoped_attributes, decl_attributes): Formatting fixes.
3587         PR tree-optimization/83269
3588         * fold-const.c (fold_binary_loc): Perform (-A) - B -> (-B) - A
3589         subtraction in arg0's type if type is signed and arg0 is unsigned.
3590         Formatting fix.
3592         PR sanitizer/81281
3593         * match.pd ((T)(P + A) - (T)P -> (T) A): Use @@0 instead of @0 and
3594         convert? on @0 instead of convert.  Check type of @1, not @0.
3595         ((T)P - (T)(P + A) -> -(T) A): Use @@0 instead of @0 and
3596         convert? on @0 instead of convert.  Check type of @1, not @0.
3597         ((T)(P + A) - (T)(P + B) -> (T)A - (T)B): Use @@0 instead of @0,
3598         only optimize if either both @1 and @2 types are narrower
3599         precision, or both are wider or equal precision, and in the former
3600         case only if both have undefined overflow.
3602 2017-12-15  Richard Biener  <rguenther@suse.de>
3604         PR lto/83388
3605         * internal-fn.def (IFN_NOP): Add.
3606         * internal-fn.c (expand_NOP): Do nothing.
3607         * lto-streamer-in.c (input_function): Instead of removing
3608         sanitizer calls replace them with IFN_NOP calls.
3610 2017-12-15  Richard Sandiford  <richard.sandiford@linaro.org>
3611             Alan Hayward  <alan.hayward@arm.com>
3612             David Sherwood  <david.sherwood@arm.com>
3614         * dse.c (store_info, read_info_type): Replace begin and end with
3615         offset and width.
3616         (print_range): New function.
3617         (set_all_positions_unneeded, any_positions_needed_p)
3618         (check_mem_read_rtx, scan_stores, scan_reads, dse_step5): Update
3619         accordingly.
3620         (record_store): Likewise.  Optimize the case in which all positions
3621         are unneeded.
3622         (get_stored_val): Replace read_begin and read_end with read_offset
3623         and read_width.
3624         (replace_read): Update call accordingly.
3626 2017-12-15  Bin Cheng  <bin.cheng@arm.com>
3628         * gimple-loop-interchange.cc (STMT_COST_RATIO): New macro.
3629         (loop_cand::m_num_stmts, loop_cand::m_const_init_reduc): New members.
3630         (loop_cand::loop_cand): Initialize above members.
3631         (loop_cand::supported_operations): Delete.
3632         (loop_cand::can_interchange_p): Inline above function.
3633         (loop_cand::classify_simple_reduction): Record number of constant
3634         initialized simple reductions.
3635         (should_interchange_loops): New parameters.  Check stmt cost of loops
3636         to be interchange.
3637         (tree_loop_interchange::interchange): Prepare stmt cost of outer loop.
3638         Update call to should_interchange_loops.
3639         (should_interchange_loop_nest): Update call to
3640         should_interchange_loops.
3642 2017-12-15  Eric Botcazou  <ebotcazou@adacore.com>
3644         PR target/66488
3645         * ggc-page.c (HOST_BITS_PER_PTR): Do not define here...
3646         * hwint.h (HOST_BITS_PER_PTR): ...but here instead.
3647         * config/i386/xm-mingw32.h (HOST_BITS_PER_PTR): Delete.
3649 2017-12-15  Julia Koval  <julia.koval@intel.com>
3651         * config/i386/i386-builtin.def (__builtin_ia32_vaesenclast_v16qi,
3652         __builtin_ia32_vaesenclast_v32qi, __builtin_ia32_vaesenclast_v64qi):
3653         New.
3654         * config/i386/sse.md (vaesenclast_<mode>): New pattern.
3655         * config/i386/vaesintrin.h (_mm256_aesenclast_epi128,
3656         _mm512_aesenclast_epi128, _mm_aesenclast_epi128): New intrinsics.
3658 2017-12-15  Julia Koval  <julia.koval@intel.com>
3660         * config/i386/i386-builtin.def (__builtin_ia32_vaesenc_v16qi,
3661         __builtin_ia32_vaesenc_v32qi, __builtin_ia32_vaesenc_v64qi): New.
3662         * config/i386/sse.md (vaesenc_<mode>): New pattern.
3663         * config/i386/vaesintrin.h (_mm256_aesenc_epi128, _mm512_aesenc_epi128,
3664         _mm_aesenc_epi128): New intrinsics.
3666 2017-12-15  Julia Koval  <julia.koval@intel.com>
3668         * config/i386/i386-builtin.def (__builtin_ia32_vaesdeclast_v16qi,
3669         __builtin_ia32_vaesdeclast_v32qi, __builtin_ia32_vaesdeclast_v64qi):
3670         New.
3671         * config/i386/sse.md (vaesdeclast_<mode>): New pattern.
3672         * config/i386/vaesintrin.h (_mm256_aesdeclast_epi128,
3673         _mm512_aesdeclast_epi128, _mm_aesdeclast_epi128): New intrinsics.
3675 2017-12-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3677         * gimple-ssa-strength-reduction.c (analyze_increments):
3678         Distinguish replacement costs for constant strides from those for
3679         unknown strides.
3681 2017-12-14  Jakub Jelinek  <jakub@redhat.com>
3683         * var-tracking.c (variable_tracking_main_1): Formatting fix.
3685 2017-12-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3687         * doc/invoke.texi: Document -Wcast-function-type.
3688         * recog.h (stored_funcptr): Change signature.
3689         * tree-dump.c (dump_node): Avoid warning.
3690         * typed-splay-tree.h (typed_splay_tree): Avoid warning.
3692 2017-12-14  Qing Zhao  <qing.zhao@oracle.com>
3694         PR middle_end/79538
3695         * gimple-fold.c (get_range_strlen): Add the handling of non-member 
3696         array.
3698 2017-12-14  David Malcolm  <dmalcolm@redhat.com>
3700         PR tree-optimization/83312
3701         * domwalk.h (dom_walker::dom_walker): Fix typo in comment.
3702         * tree-cfg.c (find_taken_edge): Update to handle NULL_TREE for
3703         "val" param, and to cope with arbitrary basic blocks.
3704         (find_taken_edge_cond_expr): Add "cond_stmt" param and use it to
3705         handle NULL_TREE for "val", dropping "bb" param.
3706         (find_taken_edge_switch_expr): Make "switch_stmt" param const and
3707         drop "bb" param.  Handle NULL_TREE for "val".
3708         (find_case_label_for_value): Make "switch_stmt" param const.
3709         * tree-vrp.c (class check_array_bounds_dom_walker): New subclass
3710         of dom_walker.
3711         (vrp_prop::check_all_array_refs): Reimplement as...
3712         (check_array_bounds_dom_walker::before_dom_children): ...this new
3713         vfunc.  Replace linear search through BB block list, excluding
3714         those with non-executable in-edges via dominator walk.
3716 2017-12-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3718         * config/arm/arm.opt (mverbose-cost-dump): New option.
3719         * config/arm/arm.c (arm_rtx_costs): Use it.
3721 2017-12-14  Andreas Schwab  <schwab@linux-m68k.org>
3723         PR bootstrap/83396
3724         * reload1.c (emit_input_reload_insns): Skip debug markers.
3726 2017-12-14  Alexandre Oliva  <aoliva@redhat.com>
3728         * config/i386/i386.c (rest_of_insert_endbranch): Use call loc
3729         for its nop_endbr.
3731         PR bootstrap/83396
3732         * config/arc/arc.c (hwloop_optimize): Skip debug insns.
3733         * config/sh/sh-protos.h (sh_find_set_of_reg): Adjust.
3734         * config/sh/sh.c: Skip debug insns besides notes.
3735         * config/sh/sh.md: Likewise.
3736         * config/sh/sh_treg_combine.cc: Likewise.
3737         * config/sh/sync.md: Likewise.
3739 2017-12-14  Tom de Vries  <tom@codesourcery.com>
3741         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
3742         Add item for weak_undefined.
3744 2017-12-14  Richard Biener  <rguenther@suse.de>
3746         PR tree-optimization/67842
3747         * sese.h (bb_in_region): Remove #if 0'ed code.
3749 2017-12-14  Richard Biener  <rguenther@suse.de>
3751         PR tree-optimization/83326
3752         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Add
3753         may_be_zero parameter and handle it by not marking the first
3754         peeled copy as not exiting the loop.
3755         (try_peel_loop): Likewise.
3756         (canonicalize_loop_induction_variables): Use number_of_iterations_exit
3757         to handle the case of constant or zero iterations and perform
3758         loop header copying on-the-fly.
3760 2017-12-14  Richard Biener  <rguenther@suse.de>
3762         PR tree-optimization/83418
3763         * vr-values.c (vr_values::extract_range_for_var_from_comparison_expr):
3764         Instead of asserting we don't get unfolded comparisons deal with them.
3766 2017-12-14  Jakub Jelinek  <jakub@redhat.com>
3768         PR bootstrap/83396
3769         * var-tracking.c (vt_initialize): Ignore non-DEBUG_INSNs outside of
3770         basic blocks.  Assert debug bind insns don't appear outside of bbs,
3771         don't reset them.  Assert insns without BLOCK_FOR_INSN are outside of
3772         bb.  Simplify.
3774         PR tree-optimization/83198
3775         * gimple-ssa-sprintf.c (format_floating): Set type solely based on
3776         dir.modifier, regardless of TREE_TYPE (arg).  Assume non-REAL_CST
3777         value if arg is a REAL_CST with incompatible type.
3779 2017-12-14  Sudakshina Das  <sudi.das@arm.com>
3780             Bin Cheng  <bin.cheng@arm.com>
3782         PR target/81228
3783         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Move LTGT
3784         to CCFPEmode.
3785         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Add
3786         LTGT.
3788 2017-12-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3790         * config/aarch64/aarch64-cores.def (cortex-a55, cortex-a75,
3791         cortex-a75.cortex-a55): Specify AARCH64_FL_F16 in the arch features.
3793 2017-12-14  Richard Sandiford  <richard.sandiford@linaro.org>
3794             Alan Hayward  <alan.hayward@arm.com>
3795             David Sherwood  <david.sherwood@arm.com>
3797         * poly-int.h: New file.
3798         * poly-int-types.h: Likewise.
3799         * coretypes.h: Include them.
3800         (POLY_INT_CONVERSION): Define.
3801         * target.def (estimated_poly_value): New hook.
3802         * doc/tm.texi.in (TARGET_ESTIMATED_POLY_VALUE): New hook.
3803         * doc/tm.texi: Regenerate.
3804         * doc/poly-int.texi: New file.
3805         * doc/gccint.texi: Include it.
3806         * doc/rtl.texi: Describe restrictions on subreg modes.
3807         * Makefile.in (TEXI_GCCINT_FILES): Add poly-int.texi.
3808         * genmodes.c (NUM_POLY_INT_COEFFS): Provide a default definition.
3809         (emit_insn_modes_h): Emit a definition of NUM_POLY_INT_COEFFS.
3810         * targhooks.h (default_estimated_poly_value): Declare.
3811         * targhooks.c (default_estimated_poly_value): New function.
3812         * target.h (estimated_poly_value): Likewise.
3813         * wide-int.h (WI_UNARY_RESULT): Use wi::binary_traits.
3814         (wi::unary_traits): Delete.
3815         (wi::binary_traits::signed_shift_result_type): Define for
3816         offset_int << HOST_WIDE_INT, etc.
3817         (generic_wide_int::operator <<=): Define for all types and use
3818         wi::lshift instead of <<.
3819         (wi::hwi_with_prec): Add a default constructor.
3820         (wi::ints_for): New class.
3821         (operator <<): Define for all wide-int types.
3822         (operator /): New function.
3823         (operator %): Likewise.
3824         * selftest.h (ASSERT_KNOWN_EQ, ASSERT_KNOWN_EQ_AT, ASSERT_MAYBE_NE)
3825         (ASSERT_MAYBE_NE_AT): New macros.
3827 2017-12-13  Eric Botcazou  <ebotcazou@adacore.com>
3828             Dominik Vogt  <vogt@linux.vnet.ibm.com>
3830         PR middle-end/78468
3831         * emit-rtl.c (init_emit): Remove ??? comment.
3832         * explow.c (get_dynamic_stack_size): Take known alignment of stack
3833         pointer + STACK_DYNAMIC_OFFSET into account in lieu of STACK_BOUNDARY.
3834         * config/sparc/sparc.h (INIT_EXPANDERS): In 32-bit mode, lower the
3835         alignment of 3 virtual registers to BITS_PER_WORD.
3837         * config/sparc/sparc.c (sparc_compute_frame_size): Simplify.
3839 2017-12-13  Peter Bergner  <bergner@vnet.ibm.com>
3841         * config/rs6000/ppc-auxv.h (PPC_FEATURE2_HTM_NO_SUSPEND): New define.
3842         * config/rs6000/rs6000.c (cpu_supports_info): Use it.
3844 2017-12-13  Alexandre Oliva  <aoliva@redhat.com>
3846         PR bootstrap/83396
3847         * reload1.c (eliminate_regs_in_insn): Skip debug markers.
3849 2017-12-13  Jakub Jelinek  <jakub@redhat.com>
3851         * tree-cfg.c (verify_gimple_in_cfg): Verify no non-label stmts
3852         with the exception of debug begin stmt markers appear before
3853         labels.
3855         PR bootstrap/83396
3856         * final.c (rest_of_handle_final): Call variable_tracking_main only
3857         if !flag_var_tracking.
3859 2017-12-13  Alexandre Oliva  <aoliva@redhat.com>
3860             Jakub Jelinek  <jakub@redhat.com>
3862         PR bootstrap/83396
3863         PR debug/83391
3864         * tree-cfgcleanup.c (remove_forwarder_block): Keep after
3865         labels debug stmts that can only appear after labels.
3867 2017-12-13  Alexander Monakov  <amonakov@ispras.ru>
3869         PR rtl-optimization/82398
3870         * sel-sched.c (sel_rank_for_schedule): Fix check for zero
3871         EXPR_USEFULNESS in priority comparison.
3873 2017-12-13  Segher Boessenkool  <segher@kernel.crashing.org>
3875         PR rtl-optimization/83393
3876         * combine.c (move_deaths): If reg_stat points to a too new insn in
3877         last_death, do not use it: find the proper insn instead.
3879 2017-12-12  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
3881         PR target/82190
3882         * config/rs6000/rs6000-string.c (expand_block_compare,
3883         expand_strn_compare): Fix set_mem_size() calls.
3885 2017-12-12  Jeff Law  <law@redhat.com>
3887         PR tree-optimization/83298
3888         PR tree-optimization/83362
3889         PR tree-optimization/83383
3890         * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Make
3891         push_value_range a public interface.  Add new argument to
3892         record_ranges_from_stmt.
3893         * gimple-ssa-evrp-analyze.c
3894         (evrp_range_analyzer::record_ranges_from_stmt): Add new argument.
3895         Update comments.  Handle recording temporary equivalences.
3896         * tree-ssa-dom.c (dom_opt_opt_walker::before_dom_children): Add
3897         new argument to call to evrp_range_analyzer::record_ranges_from_stmt.
3898         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Likewise.
3899         * tree-ssa-threadedge.c: Include alloc-pool.h, vr-values.h and
3900         gimple-ssa-evrp-analyze.h.
3901         (record_temporary_equivalences_from_phis): Add new argument.  When
3902         the PHI arg is an SSA_NAME, set the result's range to the range
3903         of the PHI arg.
3904         (record_temporary_equivalences_from_stmts_at_dest): Record ranges
3905         from statements too.
3906         (thread_through_normal_block): Accept new argument, evrp_range_analyzer.
3907         Pass it down to children as needed.
3908         (thread_outgoing_edges): Likewise.
3909         (thread_across_edge): Likewise.   Push/pop range state as needed.
3910         * tree-ssa-threadedge.h (thread_outgoing_edges): Update prototype.
3912 2017-12-12  Julia Koval  <julia.koval@intel.com>
3914         * config/i386/i386.c (PTA_SKYLAKE_AVX512): Add PTA_CLWB.
3915         (PTA_CANNONLAKE): Remove PTA_CLWB.
3917 2017-12-12  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3919         PR target/83332
3920         * config/rs6000/vector.md (vcondv2dfv2di): New define_expand.
3921         (vcondv2div2df): Likewise.
3922         (vconduv2dfv2di): Likewise.
3924 2017-12-12  Jakub Jelinek  <jakub@redhat.com>
3926         * builtins.def (BUILT_IN_NEXTAFTER, BUILT_IN_NEXTAFTERF,
3927         BUILT_IN_NEXTAFTERL, BUILT_IN_NEXTTOWARD, BUILT_IN_NEXTTOWARDF,
3928         BUILT_IN_NEXTTOWARDL): Use ATTR_MATHFN_ERRNO instead of
3929         ATTR_MATHFN_FPROUNDING_ERRNO.
3931 2017-12-12  Richard Biener  <rguenther@suse.de>
3933         PR tree-optimization/83385
3934         * graphite-scop-detection.c (get_order, order): Remove.
3935         (bb_to_rpo): New global.
3936         (cmp_pbbs): Adjust.
3937         (build_scops): Sort pbbs in RPO order.
3939 2017-12-12  James Greenhalgh  <james.greenhalgh@arm.com>
3941         * combine.c (simplify_set): Do not transform subregs to zero_extends
3942         if the destination mode is a vector mode.
3944 2017-12-12  Jakub Jelinek  <jakub@redhat.com>
3946         PR tree-optimization/83359
3947         * tree-cfg.h (fold_loop_internal_call): Declare.
3948         * tree-vectorizer.c (fold_loop_internal_call): Moved to ...
3949         * tree-cfg.c (fold_loop_internal_call): ... here.  No longer static.
3950         (find_loop_dist_alias): New function.
3951         (move_sese_region_to_fn): If any dloop->orig_loop_num value is
3952         updated, also adjust any corresponding LOOP_DIST_ALIAS internal
3953         calls.
3955         PR tree-optimization/80631
3956         * tree-vect-loop.c (get_initial_def_for_reduction): Fix comment typo.
3957         (vect_create_epilog_for_reduction): Add INDUC_VAL and INDUC_CODE
3958         arguments, for INTEGER_INDUC_COND_REDUCTION use INDUC_VAL instead of
3959         hardcoding zero as the value if COND_EXPR is never true.  For
3960         INTEGER_INDUC_COND_REDUCTION don't emit the final COND_EXPR if
3961         INDUC_VAL is equal to INITIAL_DEF, and use INDUC_CODE instead of
3962         hardcoding MAX_EXPR as the reduction operation.
3963         (is_nonwrapping_integer_induction): Allow negative step.
3964         (vectorizable_reduction): Compute INDUC_VAL and INDUC_CODE for
3965         vect_create_epilog_for_reduction, if no value is suitable, don't
3966         use INTEGER_INDUC_COND_REDUCTION for now.  Formatting fixes.
3968 2017-12-12  Richard Biener  <rguenther@suse.de>
3970         PR tree-optimization/81889
3971         * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness): Use
3972         range info from the non-wrapping IV instead of just the range
3973         of the type.
3975 2017-12-12  Julia Koval  <julia.koval@intel.com>
3977         * config.gcc: Add vaesintrin.h.
3978         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI_V64QI):
3979         New type.
3980         * config/i386/i386-builtin.def (__builtin_ia32_vaesdec_v16qi,
3981         __builtin_ia32_vaesdec_v32qi, __builtin_ia32_vaesdec_v64qi):
3982         New builtins.
3983         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
3984         * config/i386/immintrin.h: Include vaesintrin.h.
3985         * config/i386/sse.md (vaesdec_<mode>): New pattern.
3986         * config/i386/vaesintrin.h (_mm256_aesdec_epi128, _mm512_aesdec_epi128,
3987         _mm_aesdec_epi128): New intrinsics.
3989 2017-12-12  Julia Koval  <julia.koval@intel.com>
3991         * common/config/i386/i386-common.c (OPTION_MASK_ISA_VAES_SET,
3992         OPTION_MASK_ISA_VAES_UNSET): New.
3993         (ix86_handle_option): Handle -mvaes.
3994         * config/i386/cpuid.h: Define bit_VAES.
3995         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mvaes.
3996         * config/i386/i386-c.c (__VAES__): New.
3997         * config/i386/i386.c (ix86_target_string): Add -mvaes.
3998         (ix86_valid_target_attribute_inner_p): Ditto.
3999         * config/i386/i386.h (TARGET_VAES, TARGET_VAES_P): New.
4000         * config/i386/i386.opt: Add -mvaes.
4001         * doc/invoke.texi: Ditto.
4003 2017-12-12  Alexandre Oliva  <aoliva@redhat.com>
4005         * debug.h (gcc_debug_hooks): Add inline_entry.
4006         * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Likewise.
4007         * debug.c (do_nothing_debug_hooks): Likewise.
4008         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
4009         * dwarf2out.c (dwarf2_debug_hooks): Likewise.
4010         (dwarf2_lineno_debug_hooks): Likewise.
4012         * common.opt (gstatement-frontiers): New, setting
4013         debug_nonbind_markers_p.
4014         * rtl.h (MAY_HAVE_DEBUG_MARKER_INSNS): Activate.
4015         * toplev.c (process_options): Autodetect value for debug statement
4016         frontiers option.
4017         * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): Activate.
4018         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers): New.
4020         * cfgexpand.c (expand_gimple_basic_block): Handle begin stmt
4021         markers.  Integrate source bind into debug stmt expand loop.
4022         (pass_expand::execute): Check debug marker limit.  Avoid deep
4023         TER and expand debug locations for debug bind insns only.
4024         * cse.c (insn_live_p): Keep nonbind markers and debug bindings
4025         followed by them.
4026         * df-scan.c (df_insn_delete): Accept out-of-block debug insn.
4027         * final.c (reemit_insn_block_notes): Take current block from
4028         nonbind markers.  Declare note where it's first set.
4029         (final_scan_insn): Handle begin stmt notes.  Emit is_stmt according to
4030         begin stmt markers if enabled.
4031         (notice_source_line): Handle nonbind markers.  Fail if their
4032         location is unknown or that of builtins.
4033         (rest_of_handle_final): Convert begin stmt markers to notes if
4034         var-tracking didn't run.
4035         (rest_of_clean_state): Skip begin stmt markers.
4036         * gimple-pretty-print.c (dump_gimple_debug): Handle begin stmt
4037         markers.
4038         * function.c (allocate_struct_function): Set begin_stmt_markers.
4039         * function.h (struct function): Add debug_marker_count counter
4040         and debug_nonbind_markers flag.
4041         * gimple-iterator.c (gsi_remove): Adjust debug_marker_count.
4042         * gimple-low.c (lower_function_body): Adjust
4043         debug_nonbind_markers.
4044         (lower_stmt): Drop or skip gimple debug stmts.
4045         (lower_try_catch): Skip debug stmts.
4046         * gimple.c (gimple_build_debug_begin_stmt): New.
4047         (gimple_copy): Increment debug_marker_count if copying one.
4048         * gimple.h (gimple_build_debug_begin_stmt): Declare.
4049         * gimplify.c (rexpr_location): New.
4050         (rexpr_has_location): New.
4051         (warn_switch_unreachable_r): Handle gimple debug stmts.
4052         (shortcut_cond_r): Call expr_location.
4053         (find_goto): New.
4054         (find_goto_label): New.
4055         (shortcut_cond_expr): Call expr_has_location, expr_location, and
4056         find_goto_label.
4057         (gimplify_cond_expr): Call find_goto_label, expr_has_location, and
4058         expr_location.
4059         (gimplify_expr): Handle begin stmt markers.  Reject debug expr decls.
4060         * langhooks-def.h (LANG_HOOKS_EMITS_BEGIN_STMT): New.  Add to...
4061         (LANG_HOOKS_INITIALIZER): ... this.
4062         * langhooks.h (struct lang_hooks): Add emits_begin_stmt.
4063         * lra-contraints.c (inherit_reload_reg): Tolerate between-blocks
4064         debug insns.
4065         (update_ebb_live_info): Skip debug insn markers.
4066         * lra.c (debug_insn_static_data): Rename to...
4067         (debug_bind_static_data): ... this.
4068         (debug_marker_static_data): New.
4069         (lra_set_insn_recog_data): Select one of the above depending
4070         on debug insn kind.
4071         (lra_update_isn_regno_info): Don't assume debug insns have
4072         freqs.
4073         (push_insns): Skip debug insns.
4074         * lto-streamer-in.c (input_function): Drop debug stmts
4075         depending on active options.  Adjust debug_nonbind_markers.
4076         * params.def (PARAM_MAX_DEBUG_MARKER_COUNT): New.
4077         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
4078         begin stmt marker notes.
4079         (print_insn): Likewise.
4080         * recog.c (extract_insn): Recognize rtl for debug markers.
4081         * rtl.def (DEBUG_MARKER): New.
4082         * tree-inline.c: Include params.h.
4083         (remap_gimple_stmt): Handle nonbind markers.
4084         (maybe_move_debug_stmts_to_successors): Likewise.
4085         (copy_debug_stmt): Likewise.
4086         * tree-iterator.c (append_to_statement_list_1): Append begin stmt
4087         markers regardless of no side effects.
4088         (tsi_link_before): Don't update container's side effects when adding
4089         a begin stmt marker.
4090         (tsi_link_after): Likewise.
4091         (expr_first): Skip begin stmt markers.
4092         (expr_last): Likewise.
4093         * tree-pretty-print (dump_generic_node): Handle begin stmt markers.
4094         * tree-ssa-threadedge.c (propagate_threaded_block_debug_info):
4095         Disregard nonbind markers.
4096         * tree.c (make_node_stat): Don't set side effects for begin stmt
4097         markers.
4098         (build1_stat): Likewise.
4099         * tree.def (DEBUG_BEGIN_STMT): New.
4100         * tree.h (GOTO_DESTINATION): Require a GOTO_EXPR.
4101         * var-tracking.c (delete_debug_insns): Renamed to...
4102         (delete_vta_debug_insns): ... this.
4103         (reemit_marker_as_note): New.
4104         (vt_initialize): Reemit markers.
4105         (delete_vta_debug_insns): Likewise.
4106         (vt_debug_insns_local): Reemit or delete markers.
4107         (variable_tracking_main_1): Likewise.
4108         * doc/generic.texi (DEBUG_BEGIN_STMT): Document.
4109         * doc/gimple.texi (gimple_debug_begin_stmt_p): New.
4110         (gimple_debug_nonbind_marker_p): New.
4111         (gimple_build_debug_bind): Adjust.
4112         (gimple_build_debug_begin_stmt): New.
4113         * doc/invoke.texi (max-debug-marker-count): New param.
4114         * doc/rtl.texi (debug_implicit_ptr, entry_value): New.
4115         (debug_parameter_ref, debug_marker): New.
4116         (NOTE_INSN_BEGIN_STMT): New.
4117         (DEBUG_INSN): Describe begin stmt markers.
4119         * cfgbuild.c (find_bb_boundaries): Don't purge dead edges if,
4120         without debug insns, we wouldn't, but clean up debug insns
4121         after a control flow insn nevertheless.
4123         * cfgcleanup.c (delete_unreachable_blocks): Use alternate
4124         block removal order if MAY_HAVE_DEBUG_BIND_INSNS.
4125         * cfgexpand.c (label_rtx_for_bb): Skip debug insns.
4126         * cfgrtl.c (try_redirect_by_replacing_jump): Skip debug insns.
4127         (rtl_tidy_fallthru_edge): Likewise.
4128         (rtl_verify_fallthru): Likewise.
4129         (rtl_verify_bb_layout): Likewise.
4130         (skip_insns_after_block): Likewise.
4131         (duplicate_insn_chain): Use DEBUG_BIND_INSN_P.
4132         * dwarf2out.c: Include print-rtl.h.
4133         (dwarf2out_next_real_insn): New.
4134         (dwarf2out_var_location): Call it.  Disregard begin stmt markers.
4135         Dump debug binds in asm comments.
4136         * gimple-iterator.c (gimple_find_edge_insert_loc): Skip debug stmts.
4137         * gimple-iterator.h (gsi_start_bb_nondebug): Remove; adjust
4138         callers to use gsi_start_nondebug_bb instead.
4139         (gsi_after_labels): Skip gimple debug stmts.
4140         (gsi_start_nondebug): New.
4141         * gimple-loop-interchange.c (find_deps_in_bb_for_stmt): Adjust.
4142         (proper_loop_form_for_interchange): Adjust.
4143         * gimple-low.c (gimple_seq_may_fallthru): Take last nondebug stmt.
4144         * gimple.h (gimple_seq_last_nondebug_stmt): New.
4145         * gimplify.c (last_stmt_in_scope): Skip debug stmts.
4146         (collect_fallthrough_labels): Likewise.
4147         (should_warn_for_implicit_fallthrough): Likewise.
4148         (warn_implicit_fallthrough_r): Likewise.
4149         (expand_FALLTHROUGH_r): Likewise.
4150         * graphite-isl-ast-to-gimple.c (gsi_insert_earliest): Adjust.
4151         (graphite_copy_stmts_from_block): Skip nonbind markers.
4152         * haifa-sched.c (sched_extend_bb): Skip debug insns.
4153         * ipa-icf-gimple.c (func_checker::compare_bb): Adjust.
4154         * jump.c (clean_barriers): Skip debug insns.
4155         * omp-expand.c (expand_parallel_call): Skip debug insns.
4156         (expand_task_call): Likewise.
4157         (remove_exit_barrier): Likewise.
4158         (expand_omp_taskreg): Likewise.
4159         (expand_omp_for_init_counts): Likewise.
4160         (expand_omp_for_generic): Likewise.
4161         (expand_omp_for_static_nochunk): Likewise.
4162         (expand_omp_for_static_chunk): Likewise.
4163         (expand_omp_simd): Likewise.
4164         (expand_omp_taskloop_for_outer): Likewise.
4165         (expand_omp_taskloop_for_inner): Likewise.
4166         (expand_oacc_for): Likewise.
4167         (expand_omp_sections): Likewise.
4168         (expand_omp_single): Likewise.
4169         (expand_omp_synch): Likewise.
4170         (expand_omp_atomic_load): Likewise.
4171         (expand_omp_atomic_store): Likewise.
4172         (expand_omp_atomic_fetch_op): Likewise.
4173         (expand_omp_atomic_pipeline): Likewise.
4174         (expand_omp_atomic_mutex): Likewise.
4175         (expand_omp_target): Likewise.
4176         (grid_expand_omp_for_loop): Likewise.
4177         (grid_expand_target_grid_body): Likewise.
4178         (build_omp_regions_1): Likewise.
4179         * omp-low.c (check_combined_parallel): Skip debug stmts.
4180         * postreload.c (fixup_debug_insns): Skip nonbind debug insns.
4181         * regcprop.c (find_oldest_value_reg): Ensure REGNO is not a pseudo.
4182         * sese.c (sese_trivially_empty_bb_p): Call is_gimple_debug in
4183         test.
4184         * tree-cfg.c (make_blobs_1): Skip debug stmts.
4185         (make_edges): Likewise.
4186         (cleanup_dead_labels): Likewise.
4187         (gimple_can_merge_blocks_p): Likewise.
4188         (stmt_starts_bb_p): Likewise.
4189         (gimple_block_label): Likewise.
4190         (gimple_redirect_edge_and_branch): Likewise.
4191         * tree-cfgcleanup.c (remove_forwarder_block): Rearrange skipping
4192         of debug stmts.
4193         (execute_cleanup_cfg_post_optimizing): Dump enumerated decls with
4194         TDF_SLIM.
4195         * tree-pretty-print (print_declaration): Omit initializer in slim
4196         dumps.
4197         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Mark begin stmt
4198         markers.
4199         (eliminate_unnecessary_stmts): Stabilize block removal order.
4200         * tree-ssa-tail-merge.c (find_duplicate): Skip debug stmts.
4201         * var-tracking.c (get_first_insn): New.
4202         (vt_emit_notes): Call it.
4203         (vt_initialize): Walk any insns before the first BB.
4204         (delete_debug_insns): Likewise.
4206         * gimple.h (enum gimple_debug_subcode): Add
4207         GIMPLE_DEBUG_BEGIN_STMT.
4208         (gimple_debug_begin_stmt_p): New.
4209         (gimple_debug_nonbind_marker_p): New.
4210         * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): New.
4211         (MAY_HAVE_DEBUG_BIND_STMTS): Renamed from....
4212         (MAY_HAVE_DEBUG_STMTS): ... this.  Check both.
4213         * insn-notes.def (BEGIN_STMT): New.
4214         * rtl.h (MAY_HAVE_DEBUG_MARKER_INSNS): New.
4215         (MAY_HAVE_DEBUG_BIND_INSNS): Renamed from....
4216         (MAY_HAVE_DEBUG_INSNS): ... this.  Check both.
4217         (NOTE_MARKER_LOCATION, NOTE_MARKER_P): New.
4218         (DEBUG_BIND_INSN_P, DEBUG_MARKER_INSN_P): New.
4219         (INSN_DEBUG_MARKER_KIND): New.
4220         (GEN_RTX_DEBUG_MARKER_BEGIN_STMT_PAT): New.
4221         (INSN_VAR_LOCATION): Check for VAR_LOCATION.
4222         (INSN_VAR_LOCATION_PTR): New.
4223         * cfgexpand.c (expand_debug_locations): Handle debug bind insns
4224         only.
4225         (expand_gimple_basic_block): Likewise.  Emit debug temps for TER
4226         deps only if debug bind insns are enabled.
4227         (pass_expand::execute): Avoid deep TER and expand
4228         debug locations for debug bind insns only.
4229         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Narrow
4230         debug stmts special handling down to debug bind stmts.
4231         * combine.c (try_combine): Narrow debug insns special handling
4232         down to debug bind insns.
4233         * cse.c (delete_trivially_dead_insns): Handle debug bindings.
4234         Narrow debug insns preexisting special handling down to debug
4235         bind insns.
4236         * dce.c (rest_of_handle_ud_dce): Narrow debug insns special
4237         handling down to debug bind insns.
4238         * function.c (instantiate_virtual_regs): Skip debug markers,
4239         adjust handling of debug binds.
4240         * gimple-ssa-backprop.c (backprop::prepare_change): Try debug
4241         temp insertion iff MAY_HAVE_DEBUG_BIND_STMTS.
4242         * haifa-sched.c (schedule_insn): Narrow special handling of debug
4243         insns to debug bind insns.
4244         * ipa-param-manipulation.c (ipa_modify_call_arguments): Narrow
4245         special handling of debug stmts to debug bind stmts.
4246         * ipa-split.c (split_function): Likewise.
4247         * ira.c (combine_and_move_insns): Adjust debug bind insns only.
4248         * loop-unroll.c (apply_opt_in_copies): Adjust tests on bind
4249         debug insns.
4250         * reg-stack.c (convert_regs_1): Use DEBUG_BIND_INSN_P.
4251         * regrename.c (build_def_use): Likewise.
4252         * regcprop.c (copyprop_hardreg_forward_1): Likewise.
4253         (pass_cprop_hardreg): Narrow special casing of debug insns to
4254         debug bind insns.
4255         * regstat.c (regstat_init_n_sets_and_refs): Likewise.
4256         * reload1.c (reload): Likewise.
4257         * sese.c (sese_insert_phis_for_liveouts): Narrow special
4258         casing of debug stmts to debug bind stmts.
4259         * shrink-wrap.c (move_insn_for_shrink_wrap): Likewise.
4260         * ssa-iterators.h (num_imm_uses): Likewise.
4261         * tree-cfg.c (gimple_merge_blocks): Narrow special casing of
4262         debug stmts to debug bind stmts.
4263         * tree-inline.c (tree_function_versioning): Narrow special casing
4264         of debug stmts to debug bind stmts.
4265         * tree-loop-distribution.c (generate_loops_for_partition):
4266         Narrow special casing of debug stmts to debug bind stmts.
4267         * tree-sra.c (analyze_access_subtree): Narrow special casing
4268         of debug stmts to debug bind stmts.
4269         * tree-ssa-dce.c (remove_dead_stmt): Narrow special casing of debug
4270         stmts to debug bind stmts.
4271         * tree-ssa-loop-ivopt.c (remove_unused_ivs): Narrow special
4272         casing of debug stmts to debug bind stmts.
4273         * tree-ssa-reassoc.c (reassoc_remove_stmt): Likewise.
4274         * tree-ssa-tail-merge.c (tail_merge_optimize): Narrow special
4275         casing of debug stmts to debug bind stmts.
4276         * tree-ssa-threadedge.c (propagate_threaded_block_debug_info):
4277         Likewise.
4278         * tree-ssa.c (flush_pending_stmts): Narrow special casing of
4279         debug stmts to debug bind stmts.
4280         (gimple_replace_ssa_lhs): Likewise.
4281         (insert_debug_temp_for_var_def): Likewise.
4282         (insert_debug_temps_for_defs): Likewise.
4283         (reset_debug_uses): Likewise.
4284         * tree-ssanames.c (release_ssa_name_fn): Likewise.
4285         * tree-vect-loop-manip.c (adjust_debug_stmts_now): Likewise.
4286         (adjust_debug_stmts): Likewise.
4287         (adjust_phi_and_debug_stmts): Likewise.
4288         (vect_do_peeling): Likewise.
4289         * tree-vect-loop.c (vect_transform_loop): Likewise.
4290         * valtrack.c (propagate_for_debug): Use BIND_DEBUG_INSN_P.
4291         * var-tracking.c (adjust_mems): Narrow special casing of debug
4292         insns to debug bind insns.
4293         (dv_onepart_p, dataflow_set_clar_at_call, use_type): Likewise.
4294         (compute_bb_dataflow, vt_find_locations): Likewise.
4295         (vt_expand_loc, emit_notes_for_changes): Likewise.
4296         (vt_init_cfa_base): Likewise.
4297         (vt_emit_notes): Likewise.
4298         (vt_initialize): Likewise.
4299         (vt_finalize): Likewise.
4301         * emit-rtl.c (next_nondebug_insn, prev_nondebug_insn): Reorder.
4302         (next_nonnote_nondebug_insn, prev_nonnote_nondebug_insn): Reorder.
4303         (next_nonnote_nondebug_insn_bb): New.
4304         (prev_nonnote_nondebug_insn_bb): New.
4305         (prev_nonnote_insn_bb, next_nonnote_insn_bb): Remove.
4306         * rtl.h (prev_nonnote_insn_bb, next_nonnote_insn_bb): Remove decls.
4307         (prev_nonnote_nondebug_insn_bb): Declare.
4308         (next_nonnote_nondebug_insn_bb): Declare.
4309         * cfgbuild.c (find_bb_boundaries): Adjust to skip debug insns.
4310         * cfgrtl.c (get_last_bb_insn): Likewise.
4311         * lra.c (push_insns): Likewise.
4313 2017-12-11  David Malcolm  <dmalcolm@redhat.com>
4315         PR c/82050
4316         * selftest-run-tests.c (selftest::run_tests): Move start/finish code
4317         to...
4318         * selftest.c (selftest::test_runner::test_runner): New ctor.
4319         (selftest::test_runner::~test_runner): New dtor.
4320         * selftest.h (class selftest::test_runner): New class.
4322 2017-12-11  Carl Love  <cel@us.ibm.com>
4324         * config/rs6000/altivec.h (vec_extract_fp32_from_shorth,
4325         vec_extract_fp32_from_shortl]): Add #defines.
4326         * config/rs6000/rs6000-builtin.def (VSLDOI_2DI): Add macro expansion.
4327         * config/rs6000/rs6000-c.c (ALTIVEC_BUILTIN_VEC_UNPACKH,
4328         ALTIVEC_BUILTIN_VEC_UNPACKL, ALTIVEC_BUILTIN_VEC_AND,
4329         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SRL,
4330         ALTIVEC_BUILTIN_VEC_SRO, ALTIVEC_BUILTIN_VEC_SLD,
4331         ALTIVEC_BUILTIN_VEC_SLL): Add expansions.
4332         * doc/extend.texi: Add documentation for the added builtins.
4334 2017-12-11  Alexandre Oliva  <aoliva@redhat.com>
4336         PR rtl-optimization/80693
4337         PR rtl-optimization/81019
4338         PR rtl-optimization/81020
4339         * combine.c (distribute_notes): Reset any REG_UNUSED REGs that
4340         are not mentioned in i3.  Place the REG_UNUSED note on i2,
4341         possibly modified to REG_DEAD, if it did not originate in i3.
4343 2017-12-11  Jakub Jelinek  <jakub@redhat.com>
4345         * recog.c (store_data_bypass_p_1): New function.
4346         (store_data_bypass_p): Handle USE in a PARALLEL like CLOBBER.  Use
4347         store_data_bypass_p_1 to avoid code duplication.  Formatting fixes.
4349 2017-12-11  Segher Boessenkool  <segher@kernel.crashing.org>
4351         PR rtl-optimization/83361
4352         * ifcvt.c (if_convert): Call fixup_partitions.
4354 2017-12-11  Will Schmidt  <will_schmidt@vnet.ibm.com>
4356         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
4357         early folding of splat_u{8,16,32}.
4359 2017-12-11  Jakub Jelinek  <jakub@redhat.com>
4361         * config/aarch64/aarch64.c (aarch64_print_operand): Don't start
4362         output_operand_lossage first argument with capital letter.
4363         (aarch64_override_options): Don't start error and sorry first argument
4364         with capital letter.
4366 2017-12-11  Andi Kleen  <ak@linux.intel.com>
4368         PR gcov-profile/83355
4369         * auto-profile.c (string_table::get_index_by_decl): Don't
4370         recurse when abstract origin points to itself.
4372 2017-12-11  Bin Cheng  <bin.cheng@arm.com>
4374         PR tree-optimization/83320
4375         * gimple-loop-interchange.cc (free_data_refs_with_aux): Use delete.
4376         (prune_datarefs_not_in_loop): Ditto.
4378 2017-12-10  Gerald Pfeifer  <gerald@pfeifer.com>
4380         * doc/install.texi (Specific): Tweak link to mkssoftware.com.
4382 2017-12-10  Jakub Jelinek  <jakub@redhat.com>
4384         PR tree-optimization/83337
4385         * gimple-loop-interchange.cc (compute_access_stride): Handle
4386         bitfield DRs properly.
4388 2017-12-09  Jakub Jelinek  <jakub@redhat.com>
4390         PR tree-optimization/83338
4391         * tree-vect-stmts.c (vectorizable_operation): Handle POINTER_DIFF_EXPR
4392         vectorization as MINUS_EXPR with a subsequent VIEW_CONVERT_EXPR from
4393         vector of unsigned integers to vector of signed integers.
4395 2017-12-08  Vladimir Makarov  <vmakarov@redhat.com>
4397         PR rtl-optimization/83317
4398         * lra-constraints.c (process_address_1): Add insn code check.
4400 2017-12-08  Michael Matz  <matz@suse.de>
4402         Fix PR tree-optimization/83323
4403         * gimple-loop-jam (unroll_jam_possible_p): Correct test for
4404         head-controlled loops and loop BBs.
4405         * common.opt (funroll-and-jam): Remove, instead ...
4406         (floop-unroll-and-jam): ... reuse this option.
4407         * opts.c (default_options_table): Use OPT_floop_unroll_and_jam.
4408         * doc/invoke.texi (-funroll-and-jam): Move docu to ...
4409         (-floop-unroll-and-jam): ... this option.
4411 2017-12-08  Jakub Jelinek  <jakub@redhat.com>
4413         * ipa-polymorphic-call.c (noncall_stmt_may_be_vtbl_ptr_store): Fix
4414         a comment typo, get_base_ref_and_offset -> get_ref_base_and_extent.
4415         * ipa-prop.c (stmt_may_be_vtbl_ptr_store): Likewise.
4417 2017-12-08  Richard Biener  <rguenther@suse.de>
4419         PR middle-end/81782
4420         * tree-ssa-uninit.c (warn_uninitialized_vars): Properly
4421         handle accesses outside of zero-sized vars.
4423 2017-12-08  Martin Jambor  <mjambor@suse.cz>
4425         PR tree-optimization/83141
4426         * tree-sra.c (contains_vce_or_bfcref_p): Move up in the file, also
4427         test for MEM_REFs implicitely changing types with padding.  Remove
4428         inline keyword.
4429         (build_accesses_from_assign): Added contains_vce_or_bfcref_p checks.
4431 2017-12-08  Jakub Jelinek  <jakub@redhat.com>
4433         * config/arc/arc.c (arc_attribute_table): Add exclusions to
4434         the comment.
4435         * config/avr/avr.c (avr_attribute_table): Likewise.
4436         * config/msp430/msp430.c (msp430_attribute_table): Likewise.
4437         * config/rl78/rl78.c (rl78_attribute_table): Likewise.
4438         * config/nds32/nds32.c (nds32_attribute_table): Likewise.
4439         * config/darwin.h (SUBTARGET_ATTRIBUTE_TABLE): Initialize new member
4440         of struct attribute_spec.
4441         * config/i386/cygming.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
4443 2017-12-08  Ulrich Weigand  <uweigand@de.ibm.com>
4445         PR target/82960
4446         * config/spu/spu.c (pad_bb): Only check INSN_CODE when INSN_P is true.
4448 2017-12-08  Jan Hubicka  <hubicka@ucw.cz>
4450         PR middle-end/83609
4451         * profile-count.c (profile_count::from_gcov_type): Move from
4452         profile-count.h; handle overflow.
4453         * profile-count.h (profile_count::from_gcov_type): Move offline.
4455 2017-12-08  Segher Boessenkool  <segher@kernel.crashing.org>
4457         PR rtl-optimization/83304
4458         * combine.c (move_deaths): If we do not know where a register died,
4459         search for it.
4461 2017-12-08  Richard Biener  <rguenther@suse.de>
4463         * gimple-loop-interchange.cc (tree_loop_interchange::interchange):
4464         Provide -fopt-info-loop feedback when we interchange in a nest.
4466 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4468         * config/arm/driver-arm.c (arm_cpu_table): Use fp-capable product names
4469         for armv6 ARM CPU IDs.
4471 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4473         * common/config/arm/arm-common.c: Include <algorithm>.
4474         (INCLUDE_VECTOR): Define.
4475         (compare_opt_names): New function.
4476         (arm_rewrite_selected_arch): Only strip out extensions that can be
4477         expressed through -mfpu.  Sort the remaining extensions
4478         alphabetically.
4480 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4482         * config/arm/arm.h (arm_asm_auto_mfpu): Declare.
4483         (ASM_CPU_SPEC_FUNCTIONS): Add new rule asm_auto_mfpu.
4484         (ASM_CPU_SPEC): Use it if -mfpu is set to auto.
4485         * common/config/arm/arm-common.c (arm_asm_auto_mfpu): New function.
4487 2017-06-08  Tristan Gingold  <gindold@adacore.com>
4489         PR ada/81470
4490         * dwarf2out.c (dwarf2out_do_cfi_startproc): Only emit
4491         .cfi_personality or .cfi_lsda if the eh data format is dwarf2.
4493 2017-12-08  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4495         * config/sol2.h (SOLARIS_ATTRIBUTE_TABLE): Initialize new member
4496         of struct attribute_spec.
4498 2017-12-08  Julia Koval  <julia.koval@intel.com>
4500         * config/i386/avx512vnniintrin.h (_mm512_dpwssds_epi32,
4501         _mm512_mask_dpwssds_epi32, _mm512_maskz_dpwssds_epi32): New intrinsics.
4502         * config/i386/avx512vnnivlintrin.h (_mm256_dpwssds_epi32,
4503         _mm256_mask_dpwssds_epi32, _mm256_maskz_dpwssds_epi32,
4504         _mm_dpwssds_epi32, _mm_mask_dpwssds_epi32,
4505         _mm_maskz_dpwssds_epi32): Ditto.
4507 2017-12-08  Richard Biener  <rguenther@suse.de>
4509         PR tree-optimization/81303
4510         * tree-vect-stmts.c (vect_is_simple_cond): For invariant
4511         conditions try to create a comparison vector type matching
4512         the data vector type.
4513         (vectorizable_condition): Adjust.
4514         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
4515         Leave invariant conditions alone in case we can vectorize those.
4517 2017-12-08  Julia Koval  <julia.koval@intel.com>
4519         * config/i386/avx512vnniintrin.h (_mm512_dpwssd_epi32,
4520         _mm512_mask_dpwssd_epi32, _mm512_maskz_dpwssd_epi32): New intrinsics.
4521         * config/i386/avx512vnnivlintrin.h (_mm256_dpwssd_epi32,
4522         _mm256_mask_dpwssd_epi32, _mm256_maskz_dpwssd_epi32, _mm_dpwssd_epi32,
4523         _mm_mask_dpwssd_epi32, _mm_maskz_dpwssd_epi32): Ditto.
4525 2017-12-08  Julia Koval  <julia.koval@intel.com>
4527         * config/i386/avx512vnniintrin.h (_mm512_dpbusds_epi32,
4528         _mm512_mask_dpbusds_epi32, _mm512_maskz_dpbusds_epi32): New.
4529         * config/i386/avx512vnnivlintrin.h (_mm256_dpbusds_epi32,
4530         _mm256_mask_dpbusds_epi32, _mm256_maskz_dpbusds_epi32,
4531         _mm_dpbusds_epi32, _mm_mask_dpbusds_epi32,
4532         _mm_maskz_dpbusds_epi32): New intrinsics.
4534 2017-12-07  Sandra Loosemore  <sandra@codesourcery.com>
4536         * config/nios2/nios2.md (ld<bhw_uns>io): Add splitter for memory
4537         operand.
4538         (ld<bh>io_signed): Likewise.
4539         (st<bhw>io): Likewise.
4540         * config/nios2/predicates.md (ldstio_memory_operand): Allow
4541         SMALL_INT12 constant integer operand.
4543 2017-12-07  Sandra Loosemore  <sandra@codesourcery.com>
4545         * config/nios2/nios2.c (nios2_symbolic_constant_allowed):
4546         Rename to...
4547         (nios2_large_constant_allowed): ...this.  Adjust uses.
4548         (nios2_plus_symbolic_constant_p): Rename to...
4549         (nios2_plus_large_constant_p): ...this.  Adjust uses.
4550         (nios2_legitimate_address_p): Correct CONST_INT handling.
4551         (nios2_symbolic_memory_operand_p): Rename to...
4552         (nios2_large_constant_memory_operand_p): ...this.  Adjust uses.
4553         (nios2_large_constant_p): Check for large constant integers too.
4554         (nios2_split_large_constant): Handle constant integers.
4555         (nios2_split_symbolic_memory_operand): Rename to...
4556         (nios2_split_large_constant_memory_operand): ...this.  Adjust uses.
4557         (nios2_legitimize_constant_address): Handle constant integers.
4558         (r0rel_constant_p): Handle small constant integers.
4559         (nios2_print_operand_address): Handle r0-relative integer addresses.
4560         * config/nios2/nios2-protos.h: Adjust for renamed functions.
4561         * config/nios2/nios2.md: Adjust for renamed functions.
4563 2017-12-07  Andrew Waterman  <andrew@sifive.com>
4565         * config/riscv/riscv.c (TARGET_ASM_SELECT_SECTION): New define.
4566         (TARGET_HAVE_SRODATA_SECTION): New define.
4567         (riscv_select_section): New function.
4569 2017-12-08  Joseph Myers  <joseph@codesourcery.com>
4570             Alexander Monakov  <amonakov@ispras.ru>
4571             Jakub Jelinek  <jakub@redhat.com>
4573         PR target/81906
4574         * config/i386/i386.c (ix86_expand_rint): Handle flag_rounding_math.
4576 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4578         * config/aarch64/aarch64.c (aarch64_print_address_internal): Return
4579         a bool success value.  Don't call output_operand_lossage here.
4580         (aarch64_print_ldpstp_address): Return a bool success value.
4581         (aarch64_print_operand_address): Call output_addr_const if
4582         aarch64_print_address_internal fails.
4583         (aarch64_print_operand): Don't assert that the mode is 16 bytes for
4584         'y'; call output_operand_lossage instead.  Call output_operand_lossage
4585         if aarch64_print_ldpstp_address fails.
4587 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4589         * tree-vector-builder.h
4590         (tree_vector_builder::binary_encoded_nelts): Declare.
4591         * tree-vector-builder.c
4592         (tree_vector_builder::binary_encoded_nelts): New function.
4593         * fold-const.c (negate_expr_p): Likewise.
4594         (operand_equal_p, fold_checksum_tree): Likewise.
4595         * tree-loop-distribution.c (const_with_all_bytes_same): Likewise.
4596         * tree.c (integer_zerop, integer_onep, integer_all_onesp, real_zerop)
4597         (real_onep, real_minus_onep, add_expr, initializer_zerop): Likewise.
4598         (uniform_vector_p): Likewise.
4599         * varasm.c (const_hash_1, compare_constant): Likewise.
4600         * tree-ssa-ccp.c: Include tree-vector-builder.h.
4601         (valid_lattice_transition): Operate directly on the VECTOR_CST
4602         encoding.
4603         * ipa-icf.c: Include tree-vector-builder.h.
4604         (sem_variable::equals): Operate directly on the VECTOR_CST encoding.
4605         * print-tree.c (print_node): Print encoding of VECTOR_CSTs.
4607 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4609         * tree.c (build_vector): Delete.
4610         * tree.h (build_vector): Make static and move into the self-testing
4611         block.
4613 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4615         * vector-builder.h (vector_builder::derived): New const overload.
4616         (vector_builder::elt): New function.
4617         * tree-vector-builder.h (tree_vector_builder::type): New function.
4618         (tree_vector_builder::apply_step): Declare.
4619         * tree-vector-builder.c (tree_vector_builder::apply_step): New
4620         function.
4621         * gimple-fold.h (tree_vector_builder): Declare.
4622         (gimple_build_vector): Take a tree_vector_builder instead of a
4623         type and vector of elements.
4624         * gimple-fold.c (gimple_build_vector): Likewise.
4625         * tree-vect-loop.c (get_initial_def_for_reduction): Update call
4626         accordingly.
4627         (get_initial_defs_for_reduction): Likewise.
4628         (vectorizable_induction): Likewise.
4630 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4632         * tree-vector-builder.h
4633         (tree_vector_builder::new_binary_operation): Declare.
4634         * tree-vector-builder.c
4635         (tree_vector_builder::new_binary_operation): New function.
4636         * fold-const.c (fold_relational_const): Use it.
4637         (const_binop): Likewise.  Check that both input vectors have
4638         the same number of elements, thus excluding things like WIDEN_SUM.
4639         Check whether it is possible to operate directly on the encodings
4640         of stepped inputs.
4642 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4644         * fold-const.c (fold_negate_expr_1): Use tree_vector_builder and
4645         new_unary_operation, operating only on the encoded elements.
4646         (const_unop): Likewise.
4647         (exact_inverse): Likewise.
4648         (distributes_over_addition_p): New function.
4649         (const_binop): Use tree_vector_builder and new_unary_operation
4650         for combinations of VECTOR_CST and INTEGER_CST.  Operate only
4651         on the encoded elements unless the encoding is strided and the
4652         operation does not distribute over addition.
4653         (fold_convert_const):  Use tree_vector_builder and
4654         new_unary_operation.  Operate only on the encoded elements
4655         for truncating integer conversions, or for non-stepped encodings.
4657 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4659         * config/sparc/sparc.c: Include tree-vector-builder.h.
4660         (sparc_fold_builtin): Use tree_vector_builder instead of build_vector.
4661         * expmed.c: Include tree-vector-builder.h.
4662         (make_tree): Use tree_vector_builder instead of build_vector.
4663         * fold-const.c: Include tree-vector-builder.h.
4664         (const_binop): Use tree_vector_builder instead of build_vector.
4665         (const_unop): Likewise.
4666         (native_interpret_vector): Likewise.
4667         (fold_vec_perm): Likewise.
4668         (fold_ternary_loc): Likewise.
4669         * gimple-fold.c: Include tree-vector-builder.h.
4670         (gimple_fold_stmt_to_constant_1): Use tree_vector_builder instead
4671         of build_vector.
4672         * tree-ssa-forwprop.c: Include tree-vector-builder.h.
4673         (simplify_vector_constructor): Use tree_vector_builder instead
4674         of build_vector.
4675         * tree-vect-generic.c: Include tree-vector-builder.h.
4676         (add_rshift): Use tree_vector_builder instead of build_vector.
4677         (expand_vector_divmod): Likewise.
4678         (optimize_vector_constructor): Likewise.
4679         * tree-vect-loop.c: Include tree-vector-builder.h.
4680         (vect_create_epilog_for_reduction): Use tree_vector_builder instead
4681         of build_vector.  Explicitly use a stepped encoding for
4682         { 1, 2, 3, ... }.
4683         * tree-vect-slp.c: Include tree-vector-builder.h.
4684         (vect_get_constant_vectors): Use tree_vector_builder instead
4685         of build_vector.
4686         (vect_transform_slp_perm_load): Likewise.
4687         (vect_schedule_slp_instance): Likewise.
4688         * tree-vect-stmts.c: Include tree-vector-builder.h.
4689         (vectorizable_bswap): Use tree_vector_builder instead of build_vector.
4690         (vect_gen_perm_mask_any): Likewise.
4691         (vectorizable_call): Likewise.  Explicitly use a stepped encoding.
4692         * tree.c: (build_vector_from_ctor): Use tree_vector_builder instead
4693         of build_vector.
4694         (build_vector_from_val): Likewise.  Explicitly use a duplicate
4695         encoding.
4697 2017-12-07  Richard Sandiford  <richard.sandiford@arm.com>
4699         * doc/generic.texi (VECTOR_CST): Describe new representation of
4700         vector constants.
4701         * vector-builder.h: New file.
4702         * tree-vector-builder.h: Likewise.
4703         * tree-vector-builder.c: Likewise.
4704         * Makefile.in (OBJS): Add tree-vector-builder.o.
4705         * tree.def (VECTOR_CST): Update comment to refer to generic.texi.
4706         * tree-core.h (tree_base): Add a vector_cst field to the u union.
4707         (tree_vector): Change the number of elements to
4708         vector_cst_encoded_nelts.
4709         * tree.h (VECTOR_CST_NELTS): Redefine using TYPE_VECTOR_SUBPARTS.
4710         (VECTOR_CST_ELTS): Delete.
4711         (VECTOR_CST_ELT): Redefine using vector_cst_elt.
4712         (VECTOR_CST_LOG2_NPATTERNS, VECTOR_CST_NPATTERNS): New macros.
4713         (VECTOR_CST_NELTS_PER_PATTERN, VECTOR_CST_DUPLICATE_P): Likewise.
4714         (VECTOR_CST_STEPPED_P, VECTOR_CST_ENCODED_ELTS): Likewise.
4715         (VECTOR_CST_ENCODED_ELT): Likewise.
4716         (vector_cst_encoded_nelts): New function.
4717         (make_vector): Take the values of VECTOR_CST_LOG2_NPATTERNS and
4718         VECTOR_CST_NELTS_PER_PATTERN as arguments.
4719         (vector_cst_int_elt, vector_cst_elt): Declare.
4720         * tree.c: Include tree-vector-builder.h.
4721         (tree_code_size): Abort if passed VECTOR_CST.
4722         (tree_size): Update for new VECTOR_CST layout.
4723         (make_vector): Take the values of VECTOR_CST_LOG2_NPATTERNS and
4724         VECTOR_CST_NELTS_PER_PATTERN as arguments.
4725         (build_vector): Use tree_vector_builder.
4726         (vector_cst_int_elt, vector_cst_elt): New functions.
4727         (drop_tree_overflow): For VECTOR_CST, drop the TREE_OVERFLOW from the
4728         encoded elements and then create the vector in the canonical form.
4729         (check_vector_cst, check_vector_cst_duplicate, check_vector_cst_fill)
4730         (check_vector_cst_stepped, test_vector_cst_patterns): New functions.
4731         (tree_c_tests): Call test_vector_cst_patterns.
4732         * lto-streamer-out.c (DFS::DFS_write_tree_body): Handle the new
4733         VECTOR_CST fields.
4734         (hash_tree): Likewise.
4735         * tree-streamer-out.c (write_ts_vector_tree_pointers): Likewise.
4736         (streamer_write_tree_header): Likewise.
4737         * tree-streamer-in.c (lto_input_ts_vector_tree_pointers): Likewise.
4738         (streamer_alloc_tree): Likewise.  Update call to make_vector.
4739         * fold-const.c (fold_ternary_loc): Avoid using VECTOR_CST_ELTS.
4741 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4743         * selftest.h (ASSERT_TRUE_AT, ASSERT_FALSE_AT, ASSERT_EQ_AT)
4744         (ASSERT_NE, ASSERT_PRED1): Add underscores to local variable names
4745         * selftest-rtl.h (ASSERT_RTX_EQ, ASSERT_RTX_PTR_EQ): Likewise.
4747 2017-12-07  Bin Cheng  <bin.cheng@arm.com>
4748             Richard Biener  <rguenther@suse.de>
4750         PR tree-optimization/81303
4751         * Makefile.in (gimple-loop-interchange.o): New object file.
4752         * common.opt (floop-interchange): Reuse the option from graphite.
4753         * doc/invoke.texi (-floop-interchange): Ditto.  New document for
4754         -floop-interchange and mention it for -O3.
4755         * opts.c (default_options_table): Enable -floop-interchange at -O3.
4756         * gimple-loop-interchange.cc: New file.
4757         * params.def (PARAM_LOOP_INTERCHANGE_MAX_NUM_STMTS): New parameter.
4758         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): New parameter.
4759         * passes.def (pass_linterchange): New pass.
4760         * timevar.def (TV_LINTERCHANGE): New time var.
4761         * tree-pass.h (make_pass_linterchange): New declaration.
4762         * tree-ssa-loop-ivcanon.c (create_canonical_iv): Change to external
4763         interchange.  Record IV before/after increment in new parameters.
4764         * tree-ssa-loop-ivopts.h (create_canonical_iv): New declaration.
4765         * tree-vect-loop.c (vect_is_simple_reduction): Factor out reduction
4766         path check into...
4767         (check_reduction_path): ...New function here.
4768         * tree-vectorizer.h (check_reduction_path): New declaration.
4770 2017-12-07  Vladimir Makarov  <vmakarov@redhat.com>
4772         PR target/83252
4773         PR rtl-optimization/80818
4774         * lra.c (add_regs_to_insn_regno_info): Make a hard reg in CLOBBER
4775         always early clobbered.
4776         * lra-lives.c (process_bb_lives): Check input hard regs for early
4777         clobbered non-operand hard reg.
4779 2017-12-07  Jakub Jelinek  <jakub@redhat.com>
4781         PR middle-end/83164
4782         * tree-cfg.c (verify_gimple_assign_binary): Don't require
4783         types_compatible_p, just that TYPE_MODE is the same.
4785 2017-12-07  Martin Sebor  <msebor@redhat.com>
4787         PR c/81544
4788         * attribs.c (empty_attribute_table): Initialize new member of
4789         struct attribute_spec.
4790         (decl_attributes): Add argument.  Handle mutually exclusive
4791         combinations of attributes.
4792         (selftests::test_attribute_exclusions): New function.
4793         (selftests::attribute_c_tests): Ditto.
4794         * attribs.h (decl_attributes): Add default argument.
4795         * selftest.h (attribute_c_tests): Declare.
4796         * selftest-run-tests.c (selftest::run_tests): Call attribute_c_tests.
4797         * tree-core.h (attribute_spec::exclusions, exclude): New type and
4798         member.
4799         * doc/extend.texi (Common Function Attributes): Update const and pure.
4800         * config/alpha/alpha.c (vms_attribute_table): Initialize new member
4801         of struct attribute_spec.
4802         * config/arc/arc.c (arc_attribute_table): Same.
4803         * config/arm/arm.c (arm_attribute_table): Same.
4804         * config/avr/avr.c ( avr_attribute_table): Same.
4805         * config/bfin/bfin.c (bfin_attribute_table): Same.
4806         * config/cr16/cr16.c (cr16_attribute_table): Same.
4807         * config/epiphany/epiphany.c (epiphany_attribute_table): Same.
4808         * config/h8300/h8300.c (h8300_attribute_table): Same.
4809         * config/i386/i386.c (ix86_attribute_table): Same.
4810         * config/ia64/ia64.c (ia64_attribute_table): Same.
4811         * config/m32c/m32c.c (m32c_attribute_table): Same.
4812         * config/m32r/m32r.c (m32r_attribute_table): Same.
4813         * config/m68k/m68k.c (m68k_attribute_table): Same.
4814         * config/mcore/mcore.c (mcore_attribute_table): Same.
4815         * config/microblaze/microblaze.c (microblaze_attribute_table): Same.
4816         * config/mips/mips.c (mips_attribute_table): Same.
4817         * config/msp430/msp430.c (msp430_attribute_table): Same.
4818         * config/nds32/nds32.c (nds32_attribute_table): Same.
4819         * config/nvptx/nvptx.c (nvptx_attribute_table): Same.
4820         * config/powerpcspe/powerpcspe.c (rs6000_attribute_table): Same.
4821         * config/rl78/rl78.c (rl78__attribute_table): Same.
4822         * config/rs6000/rs6000.c (rs6000_attribute_table): Same.
4823         * onfig/rx/rx.c (rx_attribute_table): Same.
4824         * config/s390/s390.c (s390_handle_vectorbool_attribute): Same.
4825         * config/sh/sh.c (sh_attribute_table): Same.
4826         * config/sparc/sparc.c (sparc_attribute_table): Same.
4827         * config/spu/spu.c (spu_attribute_table): Same.
4828         * config/stormy16/stormy16.c (xstormy16_attribute_table): Same.
4829         * config/v850/v850.c (v850_attribute_table): Same.
4830         * config/visium/visium.c (visium_attribute_table): Same.
4832 2017-12-07  Tamar Christina  <tamar.christina@arm.com>
4834         PR target/82641
4835         * config/arm/arm.c (INCLUDE_STRING): Define.
4836         (arm_last_printed_arch_string, arm_last_printed_fpu_string): New.
4837         (arm_declare_function_name): Conservatively emit .arch, .arch_extensions
4838         and .fpu.
4840 2017-12-07  Michael Matz  <matz@suse.de>
4842         Add unroll and jam pass
4844         * gimple-loop-jam.c: New file.
4845         * Makefile.in (OBJS): Add gimple-loop-jam.o.
4846         * common.opt (funroll-and-jam): New option.
4847         * opts.c (default_options_table): Add unroll-and-jam at -O3.
4848         * params.def (PARAM_UNROLL_JAM_MIN_PERCENT): New param.
4849         (PARAM_UNROLL_JAM_MAX_UNROLL): Ditto.
4850         * passes.def: Add pass_loop_jam.
4851         * timevar.def (TV_LOOP_JAM): Add.
4852         * tree-pass.h (make_pass_loop_jam): Declare.
4853         * cfgloop.c (flow_loop_tree_node_add): Add AT argument.
4854         * cfgloop.h (flow_loop_tree_node_add): Adjust declaration.
4855         * cfgloopmanip.c (duplicate_loop): Add AT argument, adjust call
4856         to flow_loop_tree_node_add.
4857         (duplicate_subloops, copy_loops_to): Append to sibling list.
4858         * cfgloopmanip.h: (duplicate_loop): Adjust declaration.
4859         * doc/invoke.texi (-funroll-and-jam): Document new option.
4860         (unroll-jam-min-percent, unroll-jam-max-unroll): Document new params.
4862 2017-12-07  Richard Biener  <rguenther@suse.de>
4864         PR tree-optimization/83296
4865         PR tree-optimization/67769
4866         * tree-ssa-phiopt.c (conditional_replacement): Do not reset
4867         flow sensitive info in an unrelated BB.
4868         (value_replacement): Use reset_flow_sensitive_info.
4869         (minmax_replacement): Reset flow sensitive info on the def
4870         we move.  Do not reset flow sensitive info in the whole BB
4871         we move the stmt to.
4872         (abs_replacement): Likewise.
4874 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4876         PR target/43871
4877         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
4878         rs6000_cpu to the given -mcpu=, or to the default processor.
4880 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4882         * config/rs6000/rs6000.h (rs6000_cpu_attr): Delete.
4883         * config/rs6000/rs6000.c (rs6000_variable_issue_1): Use rs6000_tune
4884         instead of rs6000_cpu_attr.
4885         (rs6000_adjust_cost): Ditto.
4886         (is_microcoded_insn): Ditto.
4887         (rs6000_adjust_priority): Ditto.
4888         (rs6000_issue_rate): Ditto.
4889         (rs6000_use_sched_lookahead): Ditto.
4890         (rs6000_use_sched_lookahead_guard): Ditto.
4891         (rs6000_sched_reorder): Ditto.
4892         (force_new_group): Ditto.
4893         * config/rs6000/rs6000.md (cpu attribute): Ditto.
4894         (group_ending_nop): Ditto.
4896 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4898         * config/rs6000/rs6000.opt (rs6000_tune): New variable.
4899         * config/rs6000/rs6000.c (rs6000_option_override_internal): Also set
4900         rs6000_tune.  Use rs6000_tune instead of rs6000_cpu where appropriate.
4901         (rs6000_loop_align): Use rs6000_tune instead of rs6000_cpu where
4902         appropriate.
4903         (rs6000_reassociation_width): Ditto.
4904         (rs6000_emit_epilogue): Ditto.
4905         (rs6000_adjust_cost): Ditto.
4906         (is_microcoded_insn): Ditto.
4907         (is_cracked_insn): Ditto.
4908         (rs6000_adjust_priority): Ditto.
4909         (rs6000_sched_reorder): Ditto.
4910         (rs6000_sched_reorder2): Ditto.
4911         (insn_must_be_first_in_group): Ditto.
4912         (insn_must_be_last_in_group): Ditto.
4913         (rs6000_register_move_cost): Ditto.
4914         * config/rs6000/rs6000.h (rs6000_cpu_attr): Use rs6000_tune instead of
4915         rs6000_cpu.
4917 2017-12-07  Julia Koval  <julia.koval@intel.com>
4919         * config.gcc: Add vaesintrin.h.
4920         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI_V64QI):
4921         New type.
4922         * config/i386/i386-builtin.def (__builtin_ia32_vaesdec_v16qi,
4923         __builtin_ia32_vaesdec_v32qi, __builtin_ia32_vaesdec_v64qi):
4924         New builtins.
4925         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
4926         * config/i386/immintrin.h: Include vaesintrin.h.
4927         * config/i386/sse.md (vaesdec_<mode>): New pattern.
4928         * config/i386/vaesintrin.h (_mm256_aesdec_epi128, _mm512_aesdec_epi128,
4929         _mm_aesdec_epi128): New intrinsics.
4931 2017-12-06  David Malcolm  <dmalcolm@redhat.com>
4933         * Makefile.in (C_COMMON_OBJS): Add c-family/c-spellcheck.o.
4934         * spellcheck-tree.c (find_closest_macro_cpp_cb): Move to
4935         c-family/c-spellcheck.cc.
4936         (best_macro_match::best_macro_match): Likewise.
4937         * spellcheck-tree.h
4938         (struct edit_distance_traits<cpp_hashnode *>): Move to
4939         c-family/c-spellcheck.h.
4940         (class best_macro_match): Likewise.
4942 2017-12-06  Jakub Jelinek  <jakub@redhat.com>
4944         PR tree-optimization/83293
4945         * gimple-ssa-strength-reduction.c (insert_initializers): Use
4946         GSI_NEW_STMT instead of GSI_SAME_STMT in gsi_insert_after that
4947         might insert into empty bb.
4949         PR sanitizer/81281
4950         * match.pd ((T)(P + A) - (T)P -> (T) A): Split into separate
4951         simplify for plus with :c added, and pointer_plus without that.
4952         ((T)P - (T)(P + A) -> -(T) A): Likewise.  If type is integral
4953         with undefined overflow and the conversion is not widening,
4954         perform negation in utype and only convert to type afterwards.
4955         ((T)(P + A) - (T)(P + B) -> (T)A - (T)B): Split into separate
4956         simplify for plus with :c added, and pointer_plus without that.
4957         If type is integral with undefined overflow and the conversion is
4958         not widening, perform minus in utype and only convert to type
4959         afterwards.  Move the last pointer_diff_expr simplify into the
4960         two outermost ifs.
4962 2017-12-06  Martin Sebor  <msebor@redhat.com>
4964         PR tree-optimization/82646
4965         * builtins.c (maybe_emit_chk_warning): Use size as the bound for
4966         strncpy, not maxlen.
4968 2017-12-06  Martin Sebor  <msebor@redhat.com>
4970         * doc/invoke.texi (-Wstringop-truncation): Mention attribute
4971         nonstring.
4973         PR tree-optimization/83075
4974         * tree-ssa-strlen.c (handle_builtin_stxncpy): Avoid assuming
4975         strncat/strncpy don't change length of source string.
4977 2017-12-06  Eric Botcazou  <ebotcazou@adacore.com>
4979         Revert
4980         2017-11-29  Martin Aberg  <maberg@gaisler.com>
4982         * config/sparc/sparc.md (divdf3_fix): Add NOP and adjust length
4983         to prevent b2bst errata sequence.
4984         (sqrtdf2_fix): Likewise.
4986 2017-12-06  Jakub Jelinek  <jakub@redhat.com>
4988         PR tree-optimization/81945
4989         * cfgloop.h (FOR_EACH_LOOP_FN): Use FN instead of hardcoding fn.
4990         * tree-cfg.c (move_sese_region_to_fn): If any of the loops moved
4991         to dest_cfun has orig_loop_num set, either remap it to the new
4992         loop number if the loop got moved too, or clear it.
4994 2017-12-05  Steve Ellcey  <sellcey@cavium.com>
4996         * config/aarch64/thunderx2-t99.md (thunderx2t99_branch): Add trap
4997         to reservation.
4998         (thunderx2t99_nothing): New insn reservation.
4999         (thunderx2t99_mrs): New insn reservation.
5000         (thunderx2t99_multiple): New insn reservation.
5001         (thunderx2t99_alu_basi): Add bfx to reservation.
5002         (thunderx2t99_fp_cmp): Add fccmps and fccmpd to reservation.
5004 2017-12-05  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
5006         PR target/82248
5007         * config/arm/arm.md (probe_stack) : Use the 'o' constraint.
5009 2017-12-05  Bin Cheng  <bin.cheng@arm.com>
5011         * tree-ssa-dce.c (simple_dce_from_worklist): Move and rename from
5012         tree-ssa-pre.c::remove_dead_inserted_code.
5013         * tree-ssa-dce.h: New file.
5014         * tree-ssa-pre.c (tree-ssa-dce.h): Include new header file.
5015         (remove_dead_inserted_code): Move and rename to function
5016         tree-ssa-dce.c::simple_dce_from_worklist.
5017         (pass_pre::execute): Update use.
5019 2017-12-05  Richard Biener  <rguenther@suse.de>
5021         PR tree-optimization/83277
5022         * graphite-isl-ast-to-gimple.c (should_copy_to_new_region): Make sure
5023         to code-gen liveout vars.
5025 2017-12-05  Richard Sandiford  <richard.sandiford@linaro.org>
5027         * config/aarch64/aarch64-simd.md (aarch64_simd_bsldi_internal)
5028         (aarch64_simd_bsldi_alt): Check REG_P before GP_REGNUM_P.
5029         (aarch64_cm<optab>di, aarch64_cmtstdi): Add leading "&&" to
5030         split condition.
5032 2017-12-05  Max Filippov  <jcmvbkbc@gmail.com>
5034         * config/xtensa/xtensa.c (xtensa_asan_shadow_offset): New
5035         function.
5036         (TARGET_ASAN_SHADOW_OFFSET): New macro definition.
5037         * config/xtensa/xtensa.h (FRAME_GROWS_DOWNWARD): Set to 1 if
5038         ASAN is enabled.
5040 2017-12-05  Richard Biener  <rguenther@suse.de>
5042         * timevar.def (TV_TREE_RECIP, TV_TREE_SINCOS, TV_TREE_WIDEN_MUL): Add.
5043         * tree-ssa-math-opts.c (pass_data_cse_reciprocal): Use TV_TREE_RECIP.
5044         (pass_data_cse_sincos): Use TV_TREE_SINCOS.
5045         (pass_data_optimize_widening_mul): Use TV_TREE_WIDEN_MUL.
5047 2017-12-05  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
5049         * dbxout.c (dbxout_block): Grow buf to 30 bytes.
5051 2017-12-05  Martin Liska  <mliska@suse.cz>
5052             Jakub Jelinek  <jakub@redhat.com>
5054         * doc/invoke.texi: Document the options.
5055         * flag-types.h (enum sanitize_code): Add
5056         SANITIZE_POINTER_COMPARE and SANITIZE_POINTER_SUBTRACT.
5057         * ipa-inline.c (sanitize_attrs_match_for_inline_p): Add handling
5058         of SANITIZE_POINTER_COMPARE and SANITIZE_POINTER_SUBTRACT.
5059         * opts.c: Define new sanitizer options.
5060         * sanitizer.def (BUILT_IN_ASAN_POINTER_COMPARE): Likewise.
5061         (BUILT_IN_ASAN_POINTER_SUBTRACT): Likewise.
5063 2017-12-05  Julia Koval  <julia.koval@intel.com>
5065         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VNNI_SET,
5066         OPTION_MASK_ISA_AVX512VNNI_UNSET): New.
5067         (ix86_handle_option): Handle -mavx512vnni.
5068         * config/i386/cpuid.h (bit_AVX512VNNI): New bit.
5069         * config/i386/driver-i386.c (host_detect_local_cpu): Handle new bit.
5070         * config/i386/i386-c (__AVX512VNNI__): New.
5071         * config/i386/i386.c (ix86_target_string): Handle new option.
5072         (ix86_valid_target_attribute_inner_p): Handle new option.
5073         * config/i386/i386.h (TARGET_AVX512VNNI, TARGET_AVX512VNNI_P): New.
5074         * config/i386/i386.opt (mavx512vnni): New option.
5076 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
5078         PR target/81616
5079         * athlon.md: Disable for generic.
5080         * haswell.md: Enable for generic.
5081         * i386.c (ix86_sched_init_global): Add core hooks for generic.
5082         * x86-tune-sched.c (ix86_issue_rate): Increase issue rate for generic
5083         to 4.
5084         (ix86_adjust_cost): Move generic to haswell path.
5086 2017-12-04  Eric Botcazou  <ebotcazou@adacore.com>
5088         * config/sparc/sparc.c (sparc_do_work_around_errata): Use mem_ref
5089         instead of MEM_P in a couple more places.  Fix formatting issues.
5091 2017-12-04  Jim Wilson  <jimw@sifive.com>
5093         * config/riscv/riscv.c (riscv_for_each_saved_reg): Use GP_REG_LAST
5094         instead of GP_REG_LAST-1.
5095         (riscv_adjust_libcall_cfi_prologue): Likewise.
5096         (riscv_adjust_libcall_cri_epilogue): Likewise.
5097         * config/riscv/riscv.h (CALL_USED_REGISTERS): Change a6 to t6 in
5098         comment.
5100 2017-12-04  Luis Machado  <luis.machado@linaro.org>
5102         * ipa-pure-const.c (check_decl): Add missing newline.
5103         (state_from_flags): Likewise.
5105 2017-12-04  Jeff Law  <law@redhat.com>
5107         PR tree-optimizatin/78496
5108         * gimple-ssa-evrp-analyze.h
5109         (evrp_range_analyzer::get_vr_values): Simplify.
5110         * gimple-ssa-evrp-analyze.c: Corresponding changes.
5111         * tree-ssa-dom.c: Include alloc-pool.h, tree-vrp.h, vr-values.h
5112         and gimple-ssa-evrp-analyze.h.
5113         (dom_opt_dom_walker class): Add evrp_range_analyzer member.
5114         (simplify_stmt_for_jump_threading): Copy a blob of code from
5115         tree-vrp.c to use ranges to simplify statements.
5116         (dom_opt_dom_walker::before_dom_children): Call
5117         evrp_range_analyzer::{enter,record_ranges_from_stmt} methods.
5118         (dom_opt_dom_walker::after_dom_children): Similarly for
5119         evrp_range_analyzer::leave.
5120         (dom_opt_dom_walker::optimize_stmt): Use EVRP ranges to optimize
5121         conditionals.
5123         * gimple-ssa-evrp-analyze.c
5124         (evrp_range_analyzer::extract_range_from_stmt):  Always use
5125         vr_values::update_value_range so preexisting range info is
5126         medged with new range info, even if the new range is VR_VARYING.
5128 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
5130         * combine.c: Adjust comment.
5131         (use_crosses_set_p): Delete.
5132         (can_combine_p): Use modified_between_p instead of use_crosses_set_p.
5133         (try_combine): Ditto.
5135 2017-12-04  Richard Biener  <rguenther@suse.de>
5137         PR tree-optimization/83255
5138         * graphite-isl-ast-to-gimple.c (translate_isl_ast_node_for):
5139         Re-add zero-iteration check.
5141 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
5143         PR rtl-optimization/83245
5144         * lra.c (collect_non_operand_hard_regs): Treat clobbers of non-operand
5145         hard registers as earlyclobber, also if not in an asm.
5147 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
5149         PR bootstrap/83265
5150         Revert
5151         2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5153         PR target/43871
5154         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
5155         rs6000_cpu based on cpu_index, not tune_index.
5157 2017-12-04  Richard Biener  <rguenther@suse.de>
5159         PR tree-optimization/83238
5160         * graphite-scop-detection.c (scop_detection::merge_sese): Make
5161         code match comment, rejecting invalid SESE regions.
5163 2017-12-03  John David Anglin  <danglin@gcc.gnu.org>
5165         * config/pa/pa.c (pa_legitimate_address_p): For scaled indexing,
5166         require base operand is a REG_POINTER prior to reload on targets
5167         with non-equivalent space registers.
5169 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
5171         * ipa-cp.c (ipcp_lattice<valtype>::print): Update dumping.
5172         (update_specialized_profile): Fix updating of counts.
5173         (perhaps_add_new_callers): Likewise.
5175 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
5177         PR target/81616
5178         * x86-tune.def: Remove obsolette FIXMEs.
5179         (X86_TUNE_PARTIAL_FLAG_REG_STALL): Disable for generic
5180         (X86_TUNE_FUSE_CMP_AND_BRANCH_32, X86_TUNE_FUSE_CMP_AND_BRANCH_64,
5181         X86_TUNE_FUSE_CMP_AND_BRANCH_SOFLAGS, X86_TUNE_FUSE_ALU_AND_BRANCH):
5182         Enable for generic.
5183         (X86_TUNE_PAD_RETURNS): Disable for generic.
5185 2017-12-02  Jakub Jelinek  <jakub@redhat.com>
5187         PR tree-optimization/83170
5188         PR tree-optimization/83241
5189         * gimple-ssa-store-merging.c
5190         (imm_store_chain_info::try_coalesce_bswap): Update vuse field from
5191         gimple_vuse (ins_stmt) in case it has changed.
5192         (imm_store_chain_info::output_merged_store): Likewise.
5194         * tree-chkp.c (chkp_compute_bounds_for_assignment): Handle
5195         POINTER_DIFF_EXPR.
5197         PR c++/81212
5198         * tree-cfg.c (pass_warn_function_return::execute): Handle
5199         __builtin_ubsan_handle_missing_return like __builtin_unreachable
5200         with BUILTINS_LOCATION.
5202         PR target/78643
5203         PR target/80583
5204         * expr.c (get_inner_reference): If DECL_MODE of a non-bitfield
5205         is BLKmode for vector field with vector raw mode, use TYPE_MODE
5206         instead of DECL_MODE.
5208         * config/i386/i386-protos.h (standard_sse_constant_opcode): Change
5209         last argument to rtx pointer.
5210         * config/i386/i386.c (standard_sse_constant_opcode): Replace X argument
5211         with OPERANDS.  For AVX+ 128-bit VEX encoded instructions over 256-bit
5212         or 512-bit.  If setting EXT_REX_SSE_REG_P, use EVEX encoded insn
5213         depending on the chosen ISAs.
5214         * config/i386/i386.md (*movxi_internal_avx512f, *movoi_internal_avx,
5215         *movti_internal, *movdi_internal, *movsi_internal, *movtf_internal,
5216         *movdf_internal, *movsf_internal): Adjust standard_sse_constant_opcode
5217         callers.
5218         * config/i386/sse.md (mov<mode>_internal): Likewise.
5219         * config/i386/mmx.md (*mov<mode>_internal): Likewise.
5221 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5223         * doc/invoke.texi (-dp): Say that instruction cost is printed as well.
5225 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5227         * config/rs6000/rs6000-protos.h (emit_fusion_addis): Remove last two
5228         parameters from prototype.
5229         * config/rs6000/rs6000.c (emit_fusion_addis): Remove last two
5230         parameters.  Don't print a comment.
5231         (emit_fusion_gpr_load): Adjust.
5232         (emit_fusion_load_store): Adjust.
5233         * config/rs6000/rs6000.md (*fusion_p9_<mode>_constant): Adjust.
5234         * config/rs6000/vsx.md (two peepholes): Print the "vector load fusion"
5235         comment on the second line.
5237 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5239         PR target/43871
5240         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
5241         rs6000_cpu based on cpu_index, not tune_index.
5243 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5245         * final.c (output_asm_name): Print insn_cost.  Shorten output.  Print
5246         which_alternative instead of which_alternative + 1.
5247         (output_asm_insn): Print an extra tab if the template is short.
5249 2017-12-01  Jim Wilson  <jimw@sifive.com>
5251         * common.opt (use_gnu_debug_info_extensions): Delete DWARF_DEBUG from
5252         comment.
5253         * config/vx-common.h (DWARF_DEBUGGING_INFO): Delete undef.
5254         * doc/tm.texi.in (PREFERRED_DEBUGGING_TYPE): Delete DWARF_DEBUG
5255         reference.
5256         * doc/tm.texi: Regenerate.
5258 2017-12-01  Michael Meissner  <meissner@linux.vnet.ibm.com>
5260         PR target/81959
5261         * config/rs6000/rs6000.c (rs6000_address_for_fpconvert): Check for
5262         whether we can allocate pseudos before trying to fix an address.
5263         * config/rs6000/rs6000.md (float_<mode>si2_hw): Make sure the
5264         memory address is indexed or indirect.
5265         (floatuns_<mode>si2_hw2): Likewise.
5267 2017-12-01  Jason Merrill  <jason@redhat.com>
5269         * Makefile.in (TAGS): Add c-family/*.cc.
5271 2017-12-01  Wilco Dijkstra  <wdijkstr@arm.com>
5273         * config/aarch64/aarch64.md (call_insn): Use %c rather than %a.
5274         (call_value_insn): Likewise.
5275         (sibcall_insn): Likewise.
5276         (sibcall_value_insn): Likewise.
5277         (movsi_aarch64): Likewise.
5278         (movdi_aarch64): Likewise.
5279         (add_losym_): Likewise.
5280         (ldr_got_small_): Likewise.
5281         (ldr_got_small_sidi): Likewise.
5282         (ldr_got_small_28k_): Likewise.
5283         (ldr_got_small_28k_sidi): Likewise.
5284         * config/aarch64/aarch64.c (aarch64_print_address_internal):
5285         Move output_addr_const to symbolic case. Add error check.
5287 2017-12-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
5289         * config/s390/predicates.md (plus16_Q_operand): New predicate.
5290         * config/s390/s390.md: Disable MVC merging peephole if it would
5291         disable operand forwarding.
5292         (new peephole2): Split MVCs if it would turn them into up to 2
5293         forwardable MVCs.
5295 2017-12-01  Richard Biener  <rguenther@suse.de>
5297         PR tree-optimization/83232
5298         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Fix
5299         detection of same access. Instead of breaking the group here
5300         do not consider the duplicate.  Add comment explaining real fix.
5302 2017-12-01  Jonathan Wakely  <jwakely@redhat.com>
5304         * doc/md.texi (Insn Splitting): Fix "central flowgraph" typo.
5306 2017-12-01  Sudakshina Das  <sudi.das@arm.com>
5308         * config/arm/vfp.md (*movhf_vfp_fp16): Add conds attribute.
5310 2017-12-01  Jakub Jelinek  <jakub@redhat.com>
5312         * function.h (struct function): Remove cilk_frame_decl,
5313         is_cilk_function and calls_cilk_spawn fields.
5314         * tree-inline.h (struct copy_body_data): Remove remap_var_for_cilk
5315         field.
5316         * omp-simd-clone.c (simd_clone_clauses_extract): Don't clear
5317         cilk_elemental field.
5318         * cgraph.h (struct cgraph_simd_clone): Remove cilk_elemental field.
5319         * target.def: Adjust comment.
5320         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
5321         Don't test cilk_elemental.
5323         PR tree-optimization/83233
5324         * gimple-ssa-store-merging.c (nop_stats, bswap_stats): Use
5325         bswap_stat name for the struct.
5327         PR c/79153
5328         * tree.h (SWITCH_BREAK_LABEL_P): Define.
5329         * gimplify.c (collect_fallthrough_labels): Handle GIMPLE_BIND
5330         starting with a GIMPLE_SWITCH and ending with GIMPLE_LABEL with
5331         SWITCH_BREAK_LABEL_P set on the label.
5332         (gimplify_switch_expr): Set SWITCH_BREAK_LABEL_P on the label
5333         added for default case if it was missing and not all cases covered.
5334         Wrap GIMPLE_SWITCH and the switch_body_seq into a GIMPLE_BIND if
5335         switch_body_seq ends with a GIMPLE_LABEL with SWITCH_BREAK_LABEL_P
5336         set on the label.
5337         * tree-chrec.c (evolution_function_is_univariate_p): Add return true;
5338         to avoid -Wimplicit-fallthrough warning.
5339         * config/i386/i386.c (ix86_expand_special_args_builtin): Add
5340         FALLTHRU comment to avoid -Wimplicit-fallthrough warning.
5342         PR tree-optimization/83221
5343         * tree-ssa-reassoc.c (sort_by_operand_rank): Shift bb_rank
5344         down by 16.
5345         (init_reassoc): Formatting fix.
5347         PR sanitizer/81275
5348         * tree-cfg.c (group_case_labels_stmt): Don't optimize away
5349         C++ FE implicitly added __builtin_unreachable () until -Wreturn-type
5350         is diagnosed.
5352         PR sanitizer/83219
5353         * tree-cfg.c: Include asan.h.
5354         (gimple_seq_unreachable_p): Return false for -fsanitize=unreachable.
5356 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5358         * config/i386/i386.md: Fix AVX512 register width in AVX512 instruction.
5360 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5362         * config/i386/i386.c (standard_sse_constant_opcode): Fix wrong form for
5363         vpcmpeqd instruction.
5365 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5367         * config/i386/i386.c (standard_sse_constant_opcode): Fix
5368         registers type for 128bit mode.
5370 2017-11-30  Jakub Jelinek  <jakub@redhat.com>
5372         * spellcheck-tree.c (test_find_closest_identifier): Use ; instead
5373         of ;;.
5374         * gengtype-state.c (read_state_pair): Likewise.
5375         * gimple-fold.c (gimple_fold_builtin_string_compare): Likewise.
5376         * sel-sched-dump.c (dump_insn_rtx_1): Likewise.
5377         * ipa-cp.c (intersect_aggregates_with_edge): Likewise.
5378         * ifcvt.c (noce_try_store_flag_constants): Likewise.
5379         * tree-ssa-ccp.c (ccp_finalize): Likewise.
5380         * omp-grid.c (grid_process_kernel_body_copy): Likewise.
5381         * builtins.c (fold_builtin_3): Likewise.
5382         * graphite-scop-detection.c
5383         (scop_detection::stmt_has_simple_data_refs_p): Likewise.
5384         * hsa-gen.c (hsa_function_representation::hsa_function_representation):
5385         Likewise.
5387 2017-12-01  Maxim Ostapenko  <m.ostapenko@samsung.com>
5389         PR sanitizer/81697
5390         * asan.c (asan_protect_global): Add new ignore_decl_rtl_set_p
5391         parameter. Return true if ignore_decl_rtl_set_p is true and other
5392         conditions are satisfied.
5393         * asan.h (asan_protect_global): Add new parameter.
5394         * varasm.c (categorize_decl_for_section): Pass true as second parameter
5395         to asan_protect_global calls.
5397 2017-11-30  Jim Wilson  <jimw@sifive.com>
5399         * doc/invoke.texi (RISC-V Options): Delete nonexistent -mmemcpy and
5400         -mno-memcpy options.  For -mplt, -mfdiv, -mdiv, -msave-restore, and
5401         -mstrict-align, add info on default value.  Delete redundant lines for
5402         -mabi.  Add missing -mexplicit-relocs docs.
5404 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5406         * config/arc/arc.md (trap): New pattern.
5408 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5410         * config/arc/arc.c (hwloop_optimize): Prevent the last ZOL
5411         instruction to end into a delay slot.
5412         * config/arc/arc.md (cond_delay_insn): Check if the instruction
5413         can be placed into a delay slot against reg_note.
5415 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5417         * config/arc/arc.c (hwloop_optimize): Update hw-loop's end/start
5418         labels number of usages.
5420 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5422         * config/arc/arc.c (arc_cannot_substitue_mem_equiv_p): New
5423         function.
5424         (TARGET_CANNOT_SUBSTITUTE_MEM_EQUIV_P): Define.
5426 2017-11-30  Jakub Jelinek  <jakub@redhat.com>
5428         PR target/83210
5429         * internal-fn.c (expand_mul_overflow): Optimize unsigned
5430         multiplication by power of 2 constant into two shifts + comparison.
5432 2017-11-30  Jan Hubicka  <hubicka@ucw.cz>
5434         PR target/81616
5435         * config/i386/x86-tune-costs.h (generic_cost): Revise for modern CPUs.
5437 2017-11-30  Richard Biener  <rguenther@suse.de>
5439         PR tree-optimization/83202
5440         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Add
5441         allow_peel argument and guard peeling.
5442         (canonicalize_loop_induction_variables): Likewise.
5443         (canonicalize_induction_variables): Pass false.
5444         (tree_unroll_loops_completely_1): Pass unroll_outer to disallow
5445         peeling from cunrolli.
5447 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5449         * combine.c (try_combine): Print a message to dump file whenever
5450         I0, I1, or I2 cannot be combined into I3.
5452 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5454         PR rtl-optimization/83156
5455         PR rtl-optimization/82621
5456         * combine.c (try_combine): Don't split an I2 if one of the dests is
5457         set again before I3.  Allow unused dests.
5459 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5461         * config/rs6000/rs6000.md (*add<mode>3_carry_in_internal2): New.
5463 2017-11-29  Vladimir Makarov  <vmakarov@redhat.com>
5465         PR rtl-optimization/80818
5466         * lra.c (collect_non_operand_hard_regs): New arg insn.  Pass it
5467         recursively.  Use insn code for clobber.
5468         (lra_set_insn_recog_data): Pass the new arg to
5469         collect_non_operand_hard_regs.
5470         (add_regs_to_insn_regno_info): Pass insn instead of uid.  Use insn
5471         code for clobber.
5472         (lra_update_insn_regno_info): Pass insn to
5473         add_regs_to_insn_regno_info.
5475 2017-11-29  Jim Wilson  <jimw@sifive.com>
5476             Andrew Waterman  <andrew@sifive.com>
5478         * config/riscv/riscv.c (SINGLE_SHIFT_COST): New.
5479         (riscv_rtx_costs): Case ZERO_EXTRACT, match new pattern, and return
5480         SINGLE_SHIFT_COST.  Case LT and ZERO_EXTEND, likewise.  Case ASHIFT,
5481         use SINGLE_SHIFT_COST.
5482         * config/riscv/riscv.md (lshrsi3_zero_extend_1): New.
5483         (lshrsi3_zero_extend_2, lshrsi3_zero_extend_3): New.
5485 2017-11-29  Julia Koval  <julia.koval@intel.com>
5487         * config/i386/avx512vbmi2intrin.h (_mm512_shldv_epi16,
5488         _mm512_mask_shldv_epi16, _mm512_maskz_shldv_epi16, _mm512_shldv_epi32,
5489         _mm512_mask_shldv_epi32, _mm512_maskz_shldv_epi32, _mm512_shldv_epi64,
5490         _mm512_mask_shldv_epi64, _mm512_maskz_shldv_epi64): New intrinsics.
5491         * config/i386/avx512vbmi2vlintrin.h (_mm256_shldv_epi16,
5492         _mm256_mask_shldv_epi16, _mm256_maskz_shldv_epi16, _mm256_shldv_epi32,
5493         _mm256_mask_shldv_epi32, _mm256_maskz_shldv_epi32, _mm256_shldv_epi64,
5494         _mm256_mask_shldv_epi64, _mm256_maskz_shldv_epi64, _mm_shldv_epi16,
5495         _mm_mask_shldv_epi16, _mm_maskz_shldv_epi16, _mm_shldv_epi32,
5496         _mm_mask_shldv_epi32, _mm_maskz_shldv_epi32, _mm_shldv_epi64,
5497         _mm_mask_shldv_epi64, _mm_maskz_shldv_epi64): Ditto.
5498         * config/i386/i386-builtin.def (__builtin_ia32_vpshldv_v32hi,
5499         __builtin_ia32_vpshldv_v32hi_mask, __builtin_ia32_vpshldv_v32hi_maskz,
5500         __builtin_ia32_vpshldv_v16hi, __builtin_ia32_vpshldv_v16hi_mask,
5501         __builtin_ia32_vpshldv_v16hi_maskz, __builtin_ia32_vpshldv_v8hi,
5502         __builtin_ia32_vpshldv_v8hi_mask, __builtin_ia32_vpshldv_v8hi_maskz,
5503         __builtin_ia32_vpshldv_v16si, __builtin_ia32_vpshldv_v16si_mask,
5504         __builtin_ia32_vpshldv_v16si_maskz, __builtin_ia32_vpshldv_v8si,
5505         __builtin_ia32_vpshldv_v8si_mask, __builtin_ia32_vpshldv_v8si_maskz,
5506         __builtin_ia32_vpshldv_v4si, __builtin_ia32_vpshldv_v4si_mask,
5507         __builtin_ia32_vpshldv_v4si_maskz, __builtin_ia32_vpshldv_v8di,
5508         __builtin_ia32_vpshldv_v8di_mask, __builtin_ia32_vpshldv_v8di_maskz,
5509         __builtin_ia32_vpshldv_v4di, __builtin_ia32_vpshldv_v4di_mask,
5510         __builtin_ia32_vpshldv_v4di_maskz, __builtin_ia32_vpshldv_v2di,
5511         __builtin_ia32_vpshldv_v2di_mask,
5512         __builtin_ia32_vpshldv_v2di_maskz): New builtins.
5513         * config/i386/sse.md (vpshldv_<mode>, vpshldv_<mode>_mask,
5514         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): New patterns.
5516 2017-11-29  Julia Koval  <julia.koval@intel.com>
5518         * config/i386/avx512vbmi2intrin.h (_mm512_shrdv_epi16,
5519         _mm512_mask_shrdv_epi16, _mm512_maskz_shrdv_epi16, _mm512_shrdv_epi32,
5520         _mm512_mask_shrdv_epi32, _mm512_maskz_shrdv_epi32, _mm512_shrdv_epi64,
5521         _mm512_mask_shrdv_epi64, _mm512_maskz_shrdv_epi64): New intrinsics.
5522         * config/i386/avx512vbmi2vlintrin.h (_mm256_shrdv_epi16,
5523         _mm256_mask_shrdv_epi16, _mm256_maskz_shrdv_epi16, _mm256_shrdv_epi32,
5524         _mm256_mask_shrdv_epi32, _mm256_maskz_shrdv_epi32, _mm256_shrdv_epi64,
5525         _mm256_mask_shrdv_epi64, _mm256_maskz_shrdv_epi64, _mm_shrdv_epi16,
5526         _mm_mask_shrdv_epi16, _mm_maskz_shrdv_epi16, _mm_shrdv_epi32,
5527         _mm_mask_shrdv_epi32, _mm_maskz_shrdv_epi32, _mm_shrdv_epi64,
5528         _mm_mask_shrdv_epi64, _mm_maskz_shrdv_epi64): Ditto.
5529         * config/i386/i386-builtin-types.def (V32HI_FTYPE_V32HI_V32HI_V32HI,
5530         V32HI_FTYPE_V32HI_V32HI_V32HI_INT, V16HI_FTYPE_V16HI_V16HI_V16HI_INT,
5531         V8HI_FTYPE_V8HI_V8HI_V8HI_INT, V8SI_FTYPE_V8SI_V8SI_V8SI_INT,
5532         V4SI_FTYPE_V4SI_V4SI_V4SI_INT, V8DI_FTYPE_V8DI_V8DI_V8DI,
5533         V8DI_FTYPE_V8DI_V8DI_V8DI_INT, V4DI_FTYPE_V4DI_V4DI_V4DI_INT,
5534         V16SI_FTYPE_V16SI_V16SI_V16SI, V16SI_FTYPE_V16SI_V16SI_V16SI_INT,
5535         V2DI_FTYPE_V2DI_V2DI_V2DI_INT): New types.
5536         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
5537         * config/i386/sse.md (vpshrdv_<mode>, vpshrdv_<mode>_mask,
5538         vpshrdv_<mode>_maskz, vpshrdv_<mode>_maskz_1): New pattern.
5540 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5542         * config/sparc/sparc.c (sparc_do_work_around_errata): Treat the
5543         movsi_pic_gotdata_op instruction as a load for the UT699 errata
5544         workaround.
5546 2017-11-29  Martin Aberg  <maberg@gaisler.com>
5548         * config/sparc/sparc.md (divdf3_fix): Add NOP and adjust length
5549         to prevent b2bst errata sequence.
5550         (sqrtdf2_fix): Likewise.
5552 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5554         * config/sparc/sparc.c (fpop_reg_depend_p): New function.
5555         (div_sqrt_insn_p): New function.
5556         (sparc_do_work_around_errata): Insert NOP instructions to
5557         prevent sequences that could trigger the TN-0013 errata for
5558         certain LEON3 processors.
5559         (pass_work_around_errata::gate): Also test sparc_fix_lost_divsqrt.
5560         (sparc_option_override): Set sparc_fix_lost_divsqrt appropriately.
5561         * config/sparc/sparc.md (fix_lost_divsqrt): New attribute.
5562         (in_branch_delay): Prevent div and sqrt in delay slot if
5563         fix_lost_divsqrt.
5564         * config/sparc/sparc.opt (sparc_fix_lost_divsqrt): New variable.
5566 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5568         * config/sparc/sparc.c (atomic_insn_p): New function.
5569         (sparc_do_work_around_errata): Insert NOP instructions to
5570         prevent sequences that could trigger the TN-0010 errata for
5571         UT700.
5572         * config/sparc/sync.md (atomic_compare_and_swap_leon3_1): Make
5573         instruction referable in atomic_insns_p.
5575 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5577         * config/sparc/sync.md (swapsi): 16-byte align if sparc_fix_gr712rc.
5578         (atomic_compare_and_swap_leon3_1): Likewise.
5579         (ldstub): Likewise.
5581 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5583         * config/sparc/sparc.c (fpop_insn_p): New function.
5584         (sparc_do_work_around_errata): Insert NOP instructions to
5585         prevent sequences that could trigger the TN-0012 errata for
5586         GR712RC.
5587         (pass_work_around_errata::gate): Also test sparc_fix_gr712rc.
5588         * config/sparc/sparc.md (fix_gr712rc): New attribute.
5589         (in_branch_annul_delay): Prevent floating-point instructions
5590         in delay slot of annulled integer branch.
5592 2017-11-29  Richard Biener  <rguenther@suse.de>
5594         PR tree-optimization/83202
5595         * tree-vect-slp.c (scalar_stmts_set_t): New typedef.
5596         (bst_fail): Use it.
5597         (vect_analyze_slp_cost_1): Add visited set, do not account SLP
5598         nodes vectorized to the same stmts multiple times.
5599         (vect_analyze_slp_cost): Allocate a visited set and pass it down.
5600         (vect_analyze_slp_instance): Adjust.
5601         (scalar_stmts_to_slp_tree_map_t): New typedef.
5602         (vect_schedule_slp_instance): Add a map recording the SLP node
5603         representing the vectorized stmts for a set of scalar stmts.
5604         Avoid code-generating redundancies.
5605         (vect_schedule_slp): Allocate map and pass it down.
5607 2017-11-29  Nathan Sidwell  <nathan@acm.org>
5609         PR c++/83187
5610         * tree.c (build_complex_type): Fix canonicalization.  Only fill in
5611         type if it is new.
5613 2017-11-29  Wilco Dijkstra  <wdijkstr@arm.com>
5615         * config/aarch64/aarch64.c (aarch64_print_operand): Add new
5616         cases for printing LDP/STP memory addresses.
5617         (aarch64_print_address_internal): Renamed from
5618         aarch64_print_operand_address, added parameter, add Pmode check.
5619         (aarch64_print_ldpstp_address): New function for LDP/STP addresses.
5620         (aarch64_print_operand_address): Indirect to
5621         aarch64_print_address_internal.
5622         * config/aarch64/aarch64-simd.md (store_pair_lanes): Use new
5623         'y' operand output specifier.
5625 2017-11-29  Jakub Jelinek  <jakub@redhat.com>
5627         PR middle-end/83185
5628         * tree.c (build_simple_mem_ref_loc): Handle
5629         get_addr_base_and_unit_offset returning a MEM_REF.
5631         PR middle-end/80929
5632         * rtlanal.c (seq_cost): For non-single_set insns try to use insn_cost.
5634         PR target/80819
5635         * config/i386/sse.md (vec_concatv2di): Remove * from (=Yr,0,*rm)
5636         alternative.
5638 2017-11-28  David Malcolm  <dmalcolm@redhat.com>
5640         * diagnostic-show-locus.c (layout::print_trailing_fixits): Handle
5641         m_x_offset.
5642         (layout::move_to_column): Likewise.
5644 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5646         PR sanitizer/81275
5647         * tree.c (block_may_fallthru): Return false if SWITCH_ALL_CASES_P
5648         is set on SWITCH_EXPR and !block_may_fallthru (SWITCH_BODY ()).
5650 2017-11-28  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
5651             Martin Jambor  <mjambor@suse.cz>
5653         PR ipa/82808
5654         * tree.h (expr_type_first_operand_type_p): Declare
5655         * tree.c (expr_type_first_operand_type_p): New function.
5656         * ipa-prop.h (ipa_get_type): Allow i to be out of bounds.
5657         (ipa_value_from_jfunc): Adjust declaration.
5658         * ipa-cp.c (ipa_get_jf_pass_through_result): New parameter RES_TYPE.
5659         Use it as result type for arithmetics, unless it is NULL in which case
5660         be more conservative.
5661         (ipa_value_from_jfunc): New parameter PARM_TYPE, pass it to
5662         ipa_get_jf_pass_through_result.
5663         (propagate_vals_across_pass_through): Likewise.
5664         (propagate_scalar_across_jump_function): New parameter PARM_TYPE, pass
5665         is to propagate_vals_across_pass_through.
5666         (propagate_constants_across_call): Pass PARM_TYPE to
5667         propagate_scalar_across_jump_function.
5668         (find_more_scalar_values_for_callers_subset): Pass parameter type to
5669         ipa_value_from_jfunc.
5670         (cgraph_edge_brings_all_scalars_for_node): Likewise.
5671         * ipa-fnsummary.c (evaluate_properties_for_edge): Renamed parms_info
5672         to caller_parms_info, pass parameter type to ipa_value_from_jfunc.
5673         * ipa-prop.c (try_make_edge_direct_simple_call): New parameter
5674         target_type, pass it to ipa_value_from_jfunc.
5675         (update_indirect_edges_after_inlining): Pass parameter type to
5676         try_make_edge_direct_simple_call.
5678 2017-11-28  Jeff Law  <law@redhat.com>
5680         * gimple-ssa-evrp-analyze.c
5681         (evrp_range_analyzer::record_ranges_from_phis): Only use SCEV to
5682         refine ranges if scev_initialized_p returns true.
5683         * vr-values.c (vr_values::extract_range_from_phi_node): Likewise.
5685 2017-11-28  Julia Koval  <julia.koval@intel.com>
5687         * config/i386/avx512vbmi2intrin.h (_mm512_shrdi_epi16,
5688         _mm512_mask_shrdi_epi16, _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
5689         _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32, _mm512_shrdi_epi64,
5690         _mm512_mask_shrdi_epi64, _mm512_maskz_shrdi_epi64): New intrinsics.
5691         * config/i386/avx512vbmi2vlintrin.h (_mm256_shrdi_epi16,
5692         _mm256_mask_shrdi_epi16, _mm256_maskz_shrdi_epi16,
5693         _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32, _mm256_shrdi_epi32,
5694         _mm256_mask_shrdi_epi64, _mm256_maskz_shrdi_epi64, _mm256_shrdi_epi64,
5695         _mm_mask_shrdi_epi16, _mm_maskz_shrdi_epi16, _mm_shrdi_epi16,
5696         _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32, _mm_shrdi_epi32,
5697         _mm_mask_shrdi_epi64, _mm_maskz_shrdi_epi64, _mm_shrdi_epi64): Ditto.
5698         * config/i386/i386-builtin.def (__builtin_ia32_vpshrd_v32hi,
5699         __builtin_ia32_vpshrd_v32hi_mask, __builtin_ia32_vpshrd_v16hi,
5700         __builtin_ia32_vpshrd_v16hi_mask, __builtin_ia32_vpshrd_v8hi,
5701         __builtin_ia32_vpshrd_v8hi_mask, __builtin_ia32_vpshrd_v16si,
5702         __builtin_ia32_vpshrd_v16si_mask, __builtin_ia32_vpshrd_v8si,
5703         __builtin_ia32_vpshrd_v8si_mask, __builtin_ia32_vpshrd_v4si,
5704         __builtin_ia32_vpshrd_v4si_mask, __builtin_ia32_vpshrd_v8di,
5705         __builtin_ia32_vpshrd_v8di_mask, __builtin_ia32_vpshrd_v4di,
5706         __builtin_ia32_vpshrd_v4di_mask, __builtin_ia32_vpshrd_v2di,
5707         __builtin_ia32_vpshrd_v2di_mask): New builtins.
5708         * config/i386/sse.md (vpshrd_<mode><mask_name>): New pattern.
5710 2017-11-28  Julia Koval  <julia.koval@intel.com>
5712         * config/i386/avx512vbmi2intrin.h (_mm512_shldi_epi16,
5713         _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16, _mm512_shldi_epi32,
5714         _mm512_mask_shldi_epi32, _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
5715         _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): New intrinsics.
5716         * config/i386/avx512vbmi2vlintrin.h (_mm256_shldi_epi16,
5717         _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
5718         _mm256_mask_shldi_epi32, _mm256_maskz_shldi_epi32, _mm256_shldi_epi32,
5719         _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64, _mm256_shldi_epi64,
5720         _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16, _mm_shldi_epi16,
5721         _mm_mask_shldi_epi32, _mm_maskz_shldi_epi32, _mm_shldi_epi32,
5722         _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64, _mm_shldi_epi64): Ditto.
5723         * config/i386/i386-builtin-types.def (V32HI_FTYPE_V32HI_V32HI_INT,
5724         V32HI_FTYPE_V32HI_V32HI_INT_V32HI_INT, V16SI_FTYPE_V16SI_V16SI_INT,
5725         V16SI_FTYPE_V16SI_V16SI_INT_V16SI_INT,
5726         V8DI_FTYPE_V8DI_V8DI_INT_V8DI_INT, V8SI_FTYPE_V8SI_V8SI_INT_V8SI_INT,
5727         V16HI_FTYPE_V16HI_V16HI_INT_V16HI_INT,
5728         V4DI_FTYPE_V4DI_V4DI_INT_V4DI_INT,
5729         V8HI_FTYPE_V8HI_V8HI_INT_V8HI_INT,
5730         V4SI_FTYPE_V4SI_V4SI_INT_V4SI_INT,
5731         V2DI_FTYPE_V2DI_V2DI_INT_V2DI_INT): New types.
5732         * config/i386/i386-builtin.def (__builtin_ia32_vpshld_v32hi,
5733         __builtin_ia32_vpshld_v32hi_mask, __builtin_ia32_vpshld_v16hi,
5734         __builtin_ia32_vpshld_v16hi_mask, __builtin_ia32_vpshld_v8hi,
5735         __builtin_ia32_vpshld_v8hi_mask, __builtin_ia32_vpshld_v16si,
5736         __builtin_ia32_vpshld_v16si_mask, __builtin_ia32_vpshld_v8si,
5737         __builtin_ia32_vpshld_v8si_mask, __builtin_ia32_vpshld_v4si,
5738         __builtin_ia32_vpshld_v4si_mask, __builtin_ia32_vpshld_v8di,
5739         __builtin_ia32_vpshld_v8di_mask, __builtin_ia32_vpshld_v4di,
5740         __builtin_ia32_vpshld_v4di_mask, __builtin_ia32_vpshld_v2di,
5741         __builtin_ia32_vpshld_v2di_mask): New builtins.
5742         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
5743         * config/i386/sse.md (vpshld_<mode><mask_name>): New pattern.
5745 2017-11-28  Richard Biener  <rguenther@suse.de>
5747         PR tree-optimization/80776
5748         * gimple-ssa-evrp-analyze.h (evrp_range_analyzer::set_ssa_range_info):
5749         Declare.
5750         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
5751         New function.
5752         (evrp_range_analyzer::record_ranges_from_incoming_edges):
5753         If the incoming edge is an effective fallthru because the other
5754         edge only reaches a __builtin_unreachable () then record ranges
5755         derived from the controlling condition in SSA info.
5756         (evrp_range_analyzer::record_ranges_from_phis): Use set_ssa_range_info.
5757         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5759 2017-11-28  Olivier Hainque  <hainque@adacore.com>
5761         * Makefile.in (SELFTEST_FLAGS): Use nul instead of /dev/null
5762         on mingw build hosts.
5764 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5766         PR debug/81307
5767         * dbxout.c (dbx_block_with_cold_children): Fix function comment.
5769 2017-11-28  Richard Biener  <rguenther@suse.de>
5771         PR middle-end/83141
5772         * gimple-fold.c (gimple_fold_builtin_memory_op): For aggregate
5773         copies generated from memcpy use a character array as reference
5774         type.
5776 2017-11-28  Julia Koval  <julia.koval@intel.com>
5777             Sebastian Peryt  <sebastian.peryt@intel.com>
5779         * Makefile.in (cilkplus.def, cilk-builtins.def, c-family/cilk.o,
5780         c-family/c-cilkplus.o, c-family/array-notation-common.o,
5781         cilk-common.o, cilk.h, cilk-common.c): Remove.
5782         * builtin-types.def
5783         (BT_FN_INT_PTR_PTR_PTR_FTYPE_BT_INT_BT_PTR_BT_PTR_BT_PTR): Remove.
5784         * builtins.c (is_builtin_name): Remove cilkplus condition.
5785         (BUILT_IN_CILK_DETACH, BUILT_IN_CILK_POP_FRAME): Remove.
5786         * builtins.def (DEF_CILK_BUILTIN_STUB, DEF_CILKPLUS_BUILTIN,
5787         cilk-builtins.def, cilkplus.def): Remove.
5788         * cif-code.def (CILK_SPAWN): Remove.
5789         * cilk-builtins.def: Delete.
5790         * cilk-common.c: Ditto.
5791         * cilk.h: Ditto.
5792         * cilkplus.def: Ditto.
5793         * config/darwin.h (fcilkplus): Delete.
5794         * cppbuiltin.c: Ditto.
5795         * doc/extend.texi: Remove cilkplus doc.
5796         * doc/generic.texi: Ditto.
5797         * doc/invoke.texi: Ditto.
5798         * doc/passes.texi: Ditto.
5799         * gcc.c (fcilkplus): Remove.
5800         * gengtype.c (cilk.h): Remove.
5801         * gimple-pretty-print.c (dump_gimple_omp_for): Remove cilkplus
5802         support.
5803         * gimple.h (GF_OMP_FOR_KIND_CILKFOR, GF_OMP_FOR_KIND_CILKSIMD):
5804         Remove.
5805         * gimplify.c (gimplify_return_expr, maybe_fold_stmt,
5806         gimplify_call_expr,
5807         is_gimple_stmt, gimplify_modify_expr, gimplify_scan_omp_clauses,
5808         gimplify_adjust_omp_clauses, gimplify_omp_for, gimplify_expr): Remove
5809         cilkplus conditions.
5810         * ipa-fnsummary.c (ipa_dump_fn_summary, compute_fn_summary,
5811         inline_read_section): Ditto.
5812         * ipa-inline-analysis.c (cilk.h): Remove.
5813         * ira.c (ira_setup_eliminable_regset): Remove cilkplus support.
5814         * lto-wrapper.c (merge_and_complain, append_compiler_options,
5815         append_linker_options): Remove condition for fcilkplus.
5816         * lto/lto-lang.c (cilk.h): Remove.
5817         (lto_init): Remove condition for fcilkplus.
5818         * omp-expand.c (expand_cilk_for_call): Delete.
5819         (expand_omp_taskreg, expand_omp_for_static_chunk,
5820         expand_omp_for): Remove cilkplus
5821         conditions.
5822         (expand_cilk_for): Delete.
5823         * omp-general.c (omp_extract_for_data): Remove cilkplus support.
5824         * omp-low.c (scan_sharing_clauses, create_omp_child_function,
5825         execute_lower_omp, diagnose_sb_0): Ditto.
5826         * omp-simd-clone.c (simd_clone_clauses_extract): Ditto.
5827         * tree-core.h (OMP_CLAUSE__CILK_FOR_COUNT_): Delete.
5828         * tree-nested.c: Ditto.
5829         * tree-pretty-print.c (dump_omp_clause): Remove cilkplus support.
5830         (dump_generic_node): Ditto.
5831         * tree.c (OMP_CLAUSE__CILK_FOR_COUNT_): Delete.
5832         * tree.def (cilk_simd, cilk_for, cilk_spawn_stmt, cilk_sync_stmt):
5833         Delete.
5834         * tree.h (CILK_SPAWN_FN, EXPR_CILK_SPAWN): Delete.
5836 2017-11-28  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
5838         * config/aarch64/aarch64.md (div<mode>3): Change check to TARGET_FLOAT.
5839         * config/aarch64/aarch64.c (aarch64_emit_approx_div): Add early exit
5840         for vector mode and !TARGET_SIMD.
5842 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5844         * tree.def (SWITCH_EXPR): Change from 3 operand to 2 operand tree.
5845         Adjust comment.
5846         * tree.h (SWITCH_LABELS): Remove.
5847         * gimplify.c (gimplify_switch_expr): Don't test SWITCH_LABELS,
5848         assert SWITCH_BODY is non-NULL.
5849         * tree-pretty-print.c (dump_generic_node): Remove SWITCH_LABELS
5850         handling.
5851         * tree.c (block_may_fallthru): Always return true; for SWITCH_EXPR.
5853         PR tree-optimization/80788
5854         * match.pd (X +- C1 CMP C2 -> X CMP C2 -+ C1): If res
5855         has TREE_OVERFLOW set, call drop_tree_overflow.
5857 2017-11-28  Richard Biener  <rguenther@suse.de>
5859         PR tree-optimization/83158
5860         * tree-vrp.c (intersect_ranges): Prefer ~[0, 0] in a few more cases.
5862 2017-11-28  Segher Boessenkool  <segher@kernel.crashing.org>
5864         PR 81288/target
5865         * config/rs6000/rs6000.c (rs6000_rtx_costs): Do not handle
5866         TARGET_ISEL && !TARGET_MFCRF differently.  Simplify code.
5868 2017-11-27  Segher Boessenkool  <segher@kernel.crashing.org>
5870         * config/rs6000/rs6000.md (<code><GPR:mode><GPR2:mode>2_isel): Change
5871         LT/GT/LTU/GTU to LE/GE/LEU/GEU where possible.
5873 2017-11-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
5875         PR middle_end/82333
5876         * varasm.c (compare_constant): Take the mode of the constants into
5877         account when comparing floating point constants.
5879 2017-11-27  Gerald Pfeifer  <gerald@pfeifer.com>
5881         * hash-set.h (DEFINE_DEBUG_HASH_SET): Remove static qualifier
5882         from explicit instantiation of debug_helper.
5883         * vec.h (DEFINE_DEBUG_VEC): Ditto.
5885 2017-11-27  Richard Biener  <rguenther@suse.de>
5887         * gimple-fold.c (gimple_fold_builtin_memory_op): Remove dead code,
5888         refactor a bit.
5890 2017-11-27  Richard Biener  <rguenther@suse.de>
5892         * tree.c (wide_int_to_tree): Free discarded INTEGER_CST.
5893         (type_hash_canon): Also clear int_cst_hash_table entry for
5894         TYPE_MIN/MAX_VALUE.
5895         (build_nonstandard_integer_type): Hash all TYPE_MAX_VALUEs.
5897 2017-11-27  Tamar Christina  <tamar.christina@arm.com>
5899         * doc/extend.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5900         * doc/invoke.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5901         * doc/sourcebuild.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5903 2017-11-27  Markus Trippelsdorf  <markus@trippelsdorf.de>
5905         * hash-map.h (gt_cleare_cache): Avoid UB.
5907 2017-11-27  Eric Botcazou  <ebotcazou@adacore.com>
5909         * cfgloop.h (struct loop): Document usage of USHRT_MAX for unroll.
5910         * loop-unroll.c (decide_unroll_constant_iterations): Implement it.
5911         (decide_unroll_runtime_iterations): Likewise.
5912         (decide_unroll_stupid): Likewise.
5914 2017-11-27  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
5916         PR target/83109
5917         * config/i386/i386.md: Add a loop with incssp.
5919 2017-11-27  Martin Jambor  <mjambor@suse.cz>
5921         PR tree-optimization/81248
5922         * tree-sra.c (splice_param_accesses): Remove size check.
5923         (decide_one_param_reduction): Fix size check.
5924         * gimple-pretty-print.c (dump_profile): Silence warning.
5925         * params.def (PARAM_IPA_SRA_PTR_GROWTH_FACTOR): Adjust description.
5927 2017-11-27  Jakub Jelinek  <jakub@redhat.com>
5929         PR debug/81307
5930         * dbxout.c (lastlineno): New variable.
5931         (dbx_debug_hooks): Use dbxout_switch_text_section as
5932         switch_text_section debug hook.
5933         (dbxout_function_end): Switch to current_function_section
5934         rather than function_section.  If crtl->has_bb_partition,
5935         output just one N_FUN, depending on in_cold_section_p.
5936         (dbxout_source_line): Remember last lineno in lastlineno.
5937         (dbxout_switch_text_section): New function.
5938         (dbxout_function_decl): Adjust dbxout_block caller.
5939         (dbx_block_with_cold_children): New function.
5940         (dbxout_block): Return true if any LBRAC/RBRAC have been
5941         emitted.  Use dbx_block_with_cold_children at depth == 0
5942         in second partition.  Add PARENT_BLOCKNUM argument, pass
5943         it optionally adjusted to children.  Output LBRAC/RBRAC
5944         around recursive call only if the block is in the current
5945         partition, if not and anything was output, emit empty
5946         range LBRAC/RBRAC.
5947         * final.c (final_scan_insn): Compute cold_function_name
5948         before calling switch_text_section debug hook.  Call
5949         that hook even if dwarf2out_do_frame if not emitting
5950         dwarf debug info.
5952         PR target/83100
5953         * varasm.c (bss_initializer_p): Return true for DECL_COMMON
5954         TREE_READONLY decls.
5956 2017-11-27  Markus Trippelsdorf  <markus@trippelsdorf.de>
5958         PR rtl-optimization/82488
5959         * expr.c (fixup_args_size_notes): Avoid signed integer overflow.
5961 2017-11-26  Julia Koval  <julia.koval@intel.com>
5963         * config/i386/i386.c (processor_target_table): Add skylake_cost for
5964         skylake-avx512.
5965         * config/i386/x86-tune-costs.h (skylake_memcpy, skylake_memset,
5966         skylake_cost): New.
5968 2017-11-26  Julia Koval  <julia.koval@intel.com>
5970         * config/i386/driver-i386.c (host_detect_local_cpu):
5971         Detect skylake-avx512.
5973 2017-11-26  Julia Koval  <julia.koval@intel.com>
5975         * config.gcc: Add -march=cannonlake.
5976         * config/i386/driver-i386.c (host_detect_local_cpu): Detect cannonlake.
5977         * config/i386/i386-c.c (ix86_target_macros_internal): Handle cannonlake.
5978         * config/i386/i386.c (processor_costs): Add m_CANNONLAKE.
5979         (PTA_CANNONLAKE): New.
5980         (processor_target_table): Add cannonlake.
5981         (ix86_option_override_internal): Ditto.
5982         (fold_builtin_cpu): Ditto.
5983         (get_builtin_code_for_version): Handle cannonlake.
5984         (M_INTEL_COREI7_CANNONLAKE): New.
5985         * config/i386/i386.h (TARGET_CANNONLAKE, PROCESSOR_CANNONLAKE): New.
5986         * doc/invoke.texi: Add -march=cannonlake.
5988 2017-11-14  Boris Kolpackov  <boris@codesynthesis.com>
5990         * plugin.c (add_new_plugin): Use platform-specific library extensions.
5991         (try_init_one_plugin): Alternative implementation for MinGW.
5992         * Makefile.in (plugin_implib): New.
5993         (gengtype-lex.c): Fix broken AIX workaround.
5994         * configure: Regenerate.
5995         * doc/plugins.texi: Document support for MinGW.
5997 2017-11-25  Jakub Jelinek  <jakub@redhat.com>
5999         PR rtl-optimization/81553
6000         * combine.c (simplify_if_then_else): In (if_then_else COND (OP Z C1) Z)
6001         to (OP Z (mult COND (C1 * STORE_FLAG_VALUE))) optimization, if OP
6002         is a shift where C1 has different mode than the whole shift, use C1's
6003         mode for MULT rather than the shift's mode.
6005         PR target/82848
6006         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Don't fold
6007         builtins not enabled in the currently selected ISA.
6009 2017-11-24  Jackson Woodruff  <jackson.woodruff@arm.com>
6011         PR tree-optimization/71026
6012         * tree-ssa-math-opts (is_division_by_square, is_square_of): New.
6013         (insert_reciprocals): Change to insert reciprocals before a division
6014         by a square and to insert the square of a reciprocal.
6015         (execute_cse_reciprocals_1): Change to consider division by a square.
6016         (register_division_in): Add importance parameter.
6018 2017-11-24  Richard Biener  <rguenther@suse.de>
6020         PR tree-optimization/82402
6021         * tree-vect-loop-manip.c (create_lcssa_for_virtual_phi): Properly
6022         set SSA_NAME_OCCURS_IN_ABNORMAL_PHI.
6024 2017-11-24  Marc Glisse  <marc.glisse@inria.fr>
6026         * match.pd (0-ptr): New transformation.
6028 2017-11-24  Jan Hubicka  <hubicka@ucw.cz>
6030         PR bootstrap/83015
6031         * ipa-inline.c (inline_small_functions): Set current badnes correctly
6032         when skipping checking.
6034 2017-11-24  Richard Biener  <rguenther@suse.de>
6036         PR tree-optimization/83128
6037         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Handle STRING_CSTs.
6038         (vn_reference_lookup_3): Likewise.
6040 2017-11-24  Jakub Jelinek  <jakub@redhat.com>
6042         PR sanitizer/83014
6043         * ubsan.c (ubsan_type_descriptor): Use pp_unsigned_wide_integer
6044         instead of pp_printf with HOST_WIDE_INT_PRINT_DEC.  Avoid calling
6045         tree_to_uhwi twice.
6047         * tree-object-size.c (pass_through_call): Use gimple_call_return_flags
6048         ERF_RETURN*ARG* for builtins other than BUILT_IN_ASSUME_ALIGNED,
6049         check for the latter with gimple_call_builtin_p.  Do not handle
6050         BUILT_IN_STPNCPY_CHK which is not a pass through call.
6052 2017-11-24  Christophe Lyon  <christophe.lyon@linaro.org>
6054         * config/arm/arm_neon.h: Fix pragma GCC push_options before
6055         vdot_u32.
6057 2017-11-23  Julia Koval  <julia.koval@intel.com>
6059         * config/i386/avx512vbmi2intrin.h (_mm512_mask_expand_epi8,
6060         _mm512_maskz_expand_epi8, _mm512_mask_expandloadu_epi8,
6061         _mm512_maskz_expandloadu_epi8, _mm512_mask_expand_epi16,
6062         _mm512_maskz_expand_epi16, _mm512_mask_expandloadu_epi16,
6063         _mm512_maskz_expandloadu_epi16): New intrinsics.
6064         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_expand_epi8,
6065         _mm_maskz_expand_epi8, _mm_mask_expandloadu_epi8,
6066         _mm_maskz_expandloadu_epi8, _mm_mask_expand_epi16,
6067         _mm_maskz_expand_epi16, _mm_mask_expandloadu_epi16,
6068         _mm_maskz_expandloadu_epi16, _mm256_mask_expand_epi16,
6069         _mm256_maskz_expand_epi16, _mm256_mask_expandloadu_epi16,
6070         _mm256_maskz_expandloadu_epi16, _mm256_mask_expand_epi8,
6071         _mm256_maskz_expand_epi8, _mm256_mask_expandloadu_epi8,
6072         _mm256_maskz_expandloadu_epi8): New intrinsics.
6073         * config/i386/i386-builtin-types.def (V64QI_FTYPE_PCV64QI_V64QI_UDI,
6074         V32HI_FTYPE_PCV32HI_V32HI_USI, V32QI_FTYPE_PCV32QI_V32QI_USI,
6075         V16HI_FTYPE_PCV16HI_V16HI_UHI, V16QI_FTYPE_PCV16QI_V16QI_UHI,
6076         V8HI_FTYPE_PCV8HI_V8HI_UQI): New types.
6077         * config/i386/i386.c (ix86_expand_special_args_builtin): Use new types.
6078         * config/i386/sse.md (VI248_VLBW): New iterator.
6079         (expand<mode>_mask, expand<mode>_maskz): New patterns.
6081 2017-11-23  Julia Koval  <julia.koval@intel.com>
6083         * config.gcc (avx512vbmi2intrin.h, avx512vbmi2vlintrin): New headers.
6084         * config/i386/avx512vbmi2intrin.h (_mm512_mask_compress_epi8,
6085         _mm512_maskz_compress_epi8, _mm512_mask_compressstoreu_epi8,
6086         _mm512_mask_compress_epi16, _mm512_maskz_compress_epi16,
6087         _mm512_mask_compressstoreu_epi16): New.
6088         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_compress_epi8,
6089         _mm_maskz_compress_epi8, _mm256_mask_compressstoreu_epi16,
6090         _mm_mask_compress_epi16, _mm_maskz_compress_epi16,
6091         _mm256_mask_compress_epi16, _mm256_maskz_compress_epi16,
6092         _mm_mask_compressstoreu_epi8, _mm_mask_compressstoreu_epi16,
6093         _mm256_mask_compress_epi8, _mm256_maskz_compress_epi8,
6094         _mm256_mask_compressstoreu_epi8): New.
6095         * config/i386/i386-builtin-types.def (VOID_FTYPE_PV64QI_V64QI_UDI,
6096         VOID_FTYPE_PV32HI_V32HI_USI, VOID_FTYPE_PV32QI_V32QI_USI,
6097         VOID_FTYPE_PV16QI_V16QI_UHI, VOID_FTYPE_PV16HI_V16HI_UHI,
6098         VOID_FTYPE_PV8HI_V8HI_UQI): New types.
6099         * config/i386/i386-builtin.def (__builtin_ia32_compressqi512_mask,
6100         __builtin_ia32_compresshi512_mask, __builtin_ia32_compressqi256_mask,
6101         __builtin_ia32_compressqi128_mask, __builtin_ia32_compresshi256_mask,
6102         __builtin_ia32_compresshi128_mask,
6103         __builtin_ia32_compressstoreuqi512_mask,
6104         __builtin_ia32_compressstoreuhi512_mask,
6105         __builtin_ia32_compressstoreuqi256_mask,
6106         __builtin_ia32_compressstoreuqi128_mask,
6107         __builtin_ia32_compressstoreuhi256_mask,
6108         __builtin_ia32_compressstoreuhi128_mask): New builtins.
6109         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Create special args
6110         array for flags2.
6111         (ix86_expand_special_args_builtin): Handle new types.
6112         (s4fma_expand): Handle new builtin array.
6113         * config/i386/immintrin.h: Include new headers.
6114         * config/i386/sse.md (VI12_AVX512VLBW): New iterator.
6115         (compress<mode>_mask, compressstore<mode>_mask): New patterns.
6117 2017-11-23  Jakub Jelinek  <jakub@redhat.com>
6119         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Formatting
6120         fixes.  Declare temp and g variables at the top in order to avoid
6121         {} in most of the cases.
6123 2017-11-23  Marc Glisse  <marc.glisse@inria.fr>
6125         * match.pd (ptr-0): New transformation.
6127 2017-11-23  Charles Baylis  <charles.baylis@linaro.org>
6129         * config/arm/arm-protos.h (enum arm_addr_mode_op): New.
6130         (struct addr_mode_cost_table): New.
6131         (struct tune_params): Add field addr_mode_costs.
6132         * config/arm/arm.c (generic_addr_mode_costs): New.
6133         (arm_slowmul_tune): Initialise addr_mode_costs field.
6134         (arm_fastmul_tune): Likewise.
6135         (arm_strongarm_tune): Likewise.
6136         (arm_xscale_tune): Likewise.
6137         (arm_9e_tune): Likewise.
6138         (arm_marvell_pj4_tune): Likewise.
6139         (arm_v6t2_tune): Likewise.
6140         (arm_cortex_tune): Likewise.
6141         (arm_cortex_a8_tune): Likewise.
6142         (arm_cortex_a7_tune): Likewise.
6143         (arm_cortex_a15_tune): Likewise.
6144         (arm_cortex_a35_tune): Likewise.
6145         (arm_cortex_a53_tune): Likewise.
6146         (arm_cortex_a57_tune): Likewise.
6147         (arm_exynosm1_tune): Likewise.
6148         (arm_xgene1_tune): Likewise.
6149         (arm_cortex_a5_tune): Likewise.
6150         (arm_cortex_a9_tune): Likewise.
6151         (arm_cortex_a12_tune): Likewise.
6152         (arm_cortex_a73_tune): Likewise.
6153         (arm_v7m_tune): Likewise.
6154         (arm_cortex_m7_tune): Likewise.
6155         (arm_v6m_tune): Likewise.
6156         (arm_fa726te_tune): Likewise.
6157         (arm_mem_costs): Use table lookup to calculate cost of addressing mode.
6159 2017-11-23  Charles Baylis  <charles.baylis@linaro.org>
6161         * config/arm/arm.c (arm_mem_costs): New function.
6162         (arm_rtx_costs_internal): Use arm_mem_costs.
6164 2017-11-23  Mark Wielaard  <mark@klomp.org>
6166         * dwarf2out.c (init_sections_and_labels): Use generation to create
6167         unique ranges_section_label and ranges_base_label. Return generation.
6168         (output_rnglists): Add generation argument. Use generation to create
6169         unique ranges labels.
6170         (dwarf2out_finish): Get generation from init_sections_and_labels
6171         and pass generation to output_rnglists.
6173 2017-11-23  Mike Stump  <mikestump@comcast.net>
6174             Eric Botcazou  <ebotcazou@adacore.com>
6176         * doc/generic.texi (ANNOTATE_EXPR): Document 3rd operand.
6177         * cfgloop.h (struct loop): Add unroll field.
6178         * function.h (struct function): Add has_unroll bitfield.
6179         * gimplify.c (gimple_boolify) <ANNOTATE_EXPR>: Deal with unroll kind.
6180         (gimplify_expr) <ANNOTATE_EXPR>: Propagate 3rd operand.
6181         * loop-init.c (pass_loop2::gate): Return true if cfun->has_unroll.
6182         (pass_rtl_unroll_loops::gate): Likewise.
6183         * loop-unroll.c (decide_unrolling): Tweak note message.  Skip loops
6184         for which loop->unroll==1.
6185         (decide_unroll_constant_iterations): Use note for consistency and
6186         take loop->unroll into account.  Return early if loop->unroll is set.
6187         Fix thinko in existing test.
6188         (decide_unroll_runtime_iterations): Use note for consistency and
6189         take loop->unroll into account.
6190         (decide_unroll_stupid): Likewise.
6191         * lto-streamer-in.c (input_cfg): Read loop->unroll.
6192         * lto-streamer-out.c (output_cfg): Write loop->unroll.
6193         * tree-cfg.c (replace_loop_annotate_in_block) <annot_expr_unroll_kind>:
6194         New case.
6195         (replace_loop_annotate) <annot_expr_unroll_kind>: Likewise.
6196         (print_loop): Print loop->unroll if set.
6197         * tree-core.h (enum annot_expr_kind): Add annot_expr_unroll_kind.
6198         * tree-inline.c (copy_loops): Copy unroll and set cfun->has_unroll.
6199         * tree-pretty-print.c (dump_generic_node) <annot_expr_unroll_kind>:
6200         New case.
6201         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Bail out if
6202         loop->unroll is set and smaller than the trip count.  Otherwise bypass
6203         entirely the heuristics if loop->unroll is set.  Remove dead note.
6204         Fix off-by-one bug in other note.
6205         (try_peel_loop): Bail out if loop->unroll is set.  Fix formatting.
6206         (tree_unroll_loops_completely_1): Force unrolling if loop->unroll
6207         is greater than 1.
6208         (tree_unroll_loops_completely): Make static.
6209         (pass_complete_unroll::execute): Use correct type for variable.
6210         (pass_complete_unrolli::execute): Fix formatting.
6211         * tree.def (ANNOTATE_EXPR): Add 3rd operand.
6213 2017-11-23  Sergey Shalnov  <Sergey.Shalnov@intel.com>
6215         * config/i386/i386.h (TARGET_PREFER_AVX256): Also
6216         enable when TARGET_PREFER_AVX128 is set.
6218 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
6220         * ipa-profile.c (ipa_propagate_frequency_1): Use count instead of
6221         frequency.
6222         * cgraph.c (cgraph_edge::dump_edge_flags): Dump sreal frequencies.
6223         (cgraph_edge::maybe_hot_p): Use sreal frequencies.
6225 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
6227         * ipa-fnsummary.c (record_modified_bb_info): Use sreal
6228         frequencies. Fix estimation of aggregate parameters.
6230 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
6232         * cgraphclones.c (cgraph_node::create_clone): Fix updating of profile
6233         when inlining.
6235 2017-11-23  Tom de Vries  <tom@codesourcery.com>
6237         * config/sh/sh.h (ASM_OUTPUT_ADDR_VEC_ELT): Wrap in "do {} while (0)".
6239 2017-11-23  Tom de Vries  <tom@codesourcery.com>
6241         * config/ft32/ft32.h (ASM_OUTPUT_ADDR_VEC_ELT): Remove semicolon after
6242         macro.
6244 2017-11-23  Oleg Endo  <olegendo@gcc.gnu.org>
6246         PR target/83111
6247         * config/sh/sh.md (udivsi3, divsi3, sibcall_value_pcrel,
6248         sibcall_value_pcrel_fdpic): Use local variable instead of
6249         operands[3].
6250         (calli_tbr_rel): Add missing operand 2.
6251         (call_valuei_tbr_rel): Add missing operand 3.
6253 2017-11-23  Jakub Jelinek  <jakub@redhat.com>
6255         PR middle-end/82253
6256         * expr.c (expand_assignment): For CONCAT to_rtx, complex type from and
6257         bitpos/bitsize covering the whole destination, use store_expr only if
6258         the complex mode is the same.  Otherwise, use expand_normal and if
6259         it returns CONCAT, subreg each part separately instead of trying to
6260         subreg the whole result.
6262 2017-11-23  Richard Biener  <rguenther@suse.de>
6264         PR tree-optimization/23094
6265         * tree-ssa-sccvn.c (vuse_ssa_val): Handle VN_TOP when we
6266         come here from walking over backedges in the first iteration.
6267         (vn_reference_lookup_3): Skip clobbers that store the same value.
6269 2017-11-23  Richard Biener  <rguenther@suse.de>
6271         PR tree-optimization/81403
6272         * tree-ssa-pre.c (get_representative_for): Add parameter specifying
6273         a block we need a leader relative to.
6274         (phi_translate_1): For nary processing require a leader from
6275         get_representative_for given we run expression simplification
6276         using match-and-simplify.  Remove previous fix.
6278 2017-11-22  Jeff Law  <law@redhat.com>
6280         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::try_find_new_range):
6281         Use new method allocate_value_range rather than accessing the
6282         vrp_value_range_pool data member directly.
6283         * tree-vrp.c (simplify_stmt_for_jump_threading): Tweak slightly
6284         to use extract_range_from_stmt method to avoid need for
6285         extract_range_from_assignment method.
6286         (vrp_prop::vrp_finalize): Use set_lattice_propagation_complete
6287         method rather than setting values_propgated data member directly.
6288         * vr-values.h (class vr_values): Privatize vrp_value_range_pool,
6289         and values propagated data members and extract_range_from_assignment
6290         method.  Reorder private data members to conform to standards.
6291         Add new methods set_lattice_propagation_complete and
6292         allocate_value_range.
6294 2017-11-22  Eric Botcazou  <ebotcazou@adacore.com>
6296         PR rtl-optimization/83030
6297         * doc/rtl.texi (Flags in an RTL Expression): Alphabetize, add entry
6298         for CROSSING_JUMP_P and mention usage of 'jump' for JUMP_INSNs.
6299         (Insns): Delete entry for REG_CROSSING_JUMP in register notes.
6300         * bb-reorder.c (update_crossing_jump_flags): Do not test whether the
6301         CROSSING_JUMP_P flag is already set before setting it.
6302         * cfgrtl.c (fixup_partition_crossing): Likewise.
6303         * reorg.c (relax_delay_slots): Do not consider a CROSSING_JUMP_P
6304         insn as useless.
6306 2017-11-22  Jakub Jelinek  <jakub@redhat.com>
6308         * simplify-rtx.c (simplify_binary_operation_1) <case VEC_SERIES>:
6309         Handle the case where both arguments are using gen_const_vec_series.
6311 2017-11-22  David Malcolm  <dmalcolm@redhat.com>
6313         PR c++/62170
6314         * pretty-print.c (pp_format): Move quoting implementation to
6315         pp_begin_quote and pp_end_quote.  Update pp_format_decoder call
6316         to pass address of "quote" local.
6317         (pp_begin_quote): New function.
6318         (pp_end_quote): New function.
6319         * pretty-print.h (printer_fn): Convert penultimate param from bool
6320         to bool *.
6321         (pp_begin_quote): New decl.
6322         (pp_end_quote): New decl.
6323         * tree-diagnostic.c (default_tree_printer): Convert penultimate
6324         param from bool to bool *.
6325         * tree-diagnostic.h (default_tree_printer): Likewise.
6327 2017-11-22  Jeff Law  <law@redhat.com>
6329         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::evrp_range_analyzer)
6330         Initialize vr_values.
6331         (evrp_range_analyzer::try_find_new_range): Call methods attached to
6332         vr_values via vr_values class instance rather than delegators.
6333         (evrp_range_analyzer::record_ranges_from_phis): Likewise.
6334         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6335         (evrp_range_analyzer::push_value_range): Likewise.
6336         (evrp_range_analyzer::pop_value_range): Likewise.
6337         * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
6338         most delegators.  Those remaining are exposed as public interfaces.
6339         Make vr_values a pointer and private.
6340         (evrp_range_analyzer::~evrp_range_analyzer): Delete the attached
6341         vr_values.
6342         (evrp_range_analyzer::get_vr_value): New method.
6343         * gimple-ssa-evrp.c (class evrp_folder): Use DISABLE_COPY_AND_ASSIGN.
6344         (evrp_folder::evrp_folder): New ctor to initialize vr_values.
6345         (class evrp_dom_walker): Attach evrp_folder class, initialize
6346         it in the ctor.  Remove temporary delegators.
6347         (evrp_dom_walker::before_dom_children): Call methods in attached
6348         evrp_range_analyzer class via class instance pointer.  Use
6349         free value_range_constant_singleton to remove need for
6350         op_with_constant_singleton_value delegator method.  Do not
6351         create a vrp_prop class instance for every call!  Narrow
6352         scope of a couple variables.
6353         (evrp_dom_walker::cleanup): Call methods in attached
6354         evrp_range_analyzer class via class instance pointer.
6355         * vr-values.h (class vr_values): Privatize many methods and
6356         data members.
6358 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6360         * tree.def (REDUC_MAX_EXPR, REDUC_MIN_EXPR, REDUC_PLUS_EXPR): Delete.
6361         * cfgexpand.c (expand_debug_expr): Remove handling for them.
6362         * expr.c (expand_expr_real_2): Likewise.
6363         * fold-const.c (const_unop): Likewise.
6364         * optabs-tree.c (optab_for_tree_code): Likewise.
6365         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
6366         * tree-inline.c (estimate_operator_cost): Likewise.
6367         * tree-pretty-print.c (dump_generic_node): Likewise.
6368         (op_code_prio): Likewise.
6369         (op_symbol_code): Likewise.
6370         * internal-fn.def (DEF_INTERNAL_SIGNED_OPTAB_FN): Define.
6371         (IFN_REDUC_PLUS, IFN_REDUC_MAX, IFN_REDUC_MIN): New internal functions.
6372         * internal-fn.c (direct_internal_fn_optab): New function.
6373         (direct_internal_fn_array, direct_internal_fn_supported_p
6374         (internal_fn_expanders): Handle DEF_INTERNAL_SIGNED_OPTAB_FN.
6375         * fold-const-call.c (fold_const_reduction): New function.
6376         (fold_const_call): Handle CFN_REDUC_PLUS, CFN_REDUC_MAX and
6377         CFN_REDUC_MIN.
6378         * tree-vect-loop.c: Include internal-fn.h.
6379         (reduction_code_for_scalar_code): Rename to...
6380         (reduction_fn_for_scalar_code): ...this and return an internal
6381         function.
6382         (vect_model_reduction_cost): Take an internal_fn rather than
6383         a tree_code.
6384         (vect_create_epilog_for_reduction): Likewise.  Build calls rather
6385         than assignments.
6386         (vectorizable_reduction): Use internal functions rather than tree
6387         codes for the reduction operation.  Update calls to the functions
6388         above.
6389         * config/aarch64/aarch64-builtins.c (aarch64_gimple_fold_builtin):
6390         Use calls to internal functions rather than REDUC tree codes.
6391         * config/aarch64/aarch64-simd.md: Update comment accordingly.
6393 2017-11-22  Olivier Hainque  <hainque@adacore.com>
6395         * config/vxworks.c (vxworks_override_options): Pick default
6396         dwarf version from macro value, VXWORKS_DWARF_VERSION_DEFAULT.
6397         * config/vxworks.h: Define VXWORKS_DWARF_VERSION_DEFAULT and
6398         DWARF_GNAT_ENCODINGS_DEFAULT.
6399         * config/vxworksae.h: Likewise.
6401 2017-11-22  Marc Glisse  <marc.glisse@inria.fr>
6403         PR tree-optimization/83104
6404         * vr-values.c (simplify_stmt_using_ranges): Check integral argument,
6405         not result.
6407 2017-11-22  Marek Polacek  <polacek@redhat.com>
6408             H.J. Lu  <hongjiu.lu@intel.com>
6409             Jason Merrill  <jason@redhat.com>
6411         PR c++/60336
6412         PR middle-end/67239
6413         PR target/68355
6414         * calls.c (initialize_argument_information): Call
6415         warn_parameter_passing_abi target hook.
6416         (store_one_arg): Use 0 for empty record size.  Don't push 0 size
6417         argument onto stack.
6418         (must_pass_in_stack_var_size_or_pad): Return false for empty types.
6419         * common.opt: Update -fabi-version description.
6420         * config/i386/i386.c (init_cumulative_args): Set cum->warn_empty.
6421         (ix86_gimplify_va_arg): Call arg_int_size_in_bytes instead of
6422         int_size_in_bytes.
6423         (ix86_is_empty_record): New function.
6424         (ix86_warn_parameter_passing_abi): New function.
6425         (TARGET_EMPTY_RECORD_P): Redefine.
6426         (TARGET_WARN_PARAMETER_PASSING_ABI): Redefine.
6427         * config/i386/i386.h (CUMULATIVE_ARGS): Add warn_empty.
6428         * doc/tm.texi: Regenerated.
6429         * doc/tm.texi.in (TARGET_EMPTY_RECORD_P,
6430         TARGET_WARN_PARAMETER_PASSING_ABI): Add.
6431         * dwarf2out.c (get_ultimate_context): Move to tree.c.
6432         * explow.c (hard_function_value): Call arg_int_size_in_bytes
6433         instead of int_size_in_bytes.
6434         * expr.c (copy_blkmode_to_reg): Likewise.
6435         * function.c (aggregate_value_p): Return 0 for empty types.
6436         (assign_parm_find_entry_rtl): Call warn_parameter_passing_abi
6437         target hook.
6438         (locate_and_pad_parm): Call arg size_in_bytes instead
6439         size_in_bytes.
6440         * lto-streamer-out.c (hash_tree): Hash TYPE_EMPTY_P and DECL_PADDING_P.
6441         * stor-layout.c (finalize_type_size): Set TYPE_EMPTY_P.
6442         * target.def (empty_record_p, warn_parameter_passing_abi): New target
6443         hooks.
6444         * targhooks.c (hook_void_CUMULATIVE_ARGS_tree): New hook.
6445         (std_gimplify_va_arg_expr): Skip empty records.  Call
6446         arg_size_in_bytes instead size_in_bytes.
6447         * targhooks.h (hook_void_CUMULATIVE_ARGS_tree): Declare.
6448         * tree-core.h (tree_type_common): Add empty_flag.
6449         (tree_decl_common): Update comments.
6450         * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Stream
6451         DECL_PADDING_P.
6452         (unpack_ts_type_common_value_fields): Stream TYPE_EMPTY_P.
6453         * tree-streamer-out.c (pack_ts_decl_common_value_fields): Stream
6454         DECL_PADDING_P.
6455         (pack_ts_type_common_value_fields): Stream TYPE_EMPTY_P.
6456         * tree.c (default_is_empty_type): New function.
6457         (default_is_empty_record): New function.
6458         (arg_int_size_in_bytes): New function.
6459         (arg_size_in_bytes): New function.
6460         (get_ultimate_context): New function.
6461         * tree.h: Define TYPE_EMPTY_P, DECL_PADDING_P and
6462         TRANSLATION_UNIT_WARN_EMPTY_P.
6463         (default_is_empty_record, arg_int_size_in_bytes,
6464         arg_size_in_bytes, get_ultimate_context): Declare.
6466 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6468         * config/arm/arm.c (cmse_clear_registers): New function.
6469         (cmse_nonsecure_call_clear_caller_saved): Replace register clearing
6470         code by call to cmse_clear_registers.
6471         (cmse_nonsecure_entry_clear_before_return): Likewise.
6473 2017-11-22  Tamar Christina  <tamar.christina@arm.com>
6475         * config/arm/arm_neon.h (vdot_u32, vdotq_u32)
6476         (vdot_s32, vdotq_s32): New.
6477         (vdot_lane_u32, vdotq_lane_u32): New.
6478         (vdot_lane_s32, vdotq_lane_s32): New.
6481 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6483         PR middle-end/82547
6484         * wide-int.cc (wi::add_large, wi::sub_large): Fix overflow detection
6485         for unsigned values with fewer HWIs than the precision.
6486         (test_overflow): New function.
6487         (wide_int_cc_tests): Call it.
6489 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6491         * emit-rtl.c (init_derived_machine_modes): Make sure ptr_mode
6492         has the same mode class as Pmode.
6494 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6496         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Get rid of
6497         padding_bits_to_clear_ptr.
6498         (cmse_nonsecure_entry_clear_before_return): Likewise.
6500 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6502         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Use
6503         auto_sbitap instead of integer bitfield to control register needing
6504         clearing.
6506 2017-11-22  Jakub Jelinek  <jakub@redhat.com>
6508         PR tree-optimization/83044
6509         * tree-vrp.c (vrp_prop::check_array_ref): If eltsize is not
6510         INTEGER_CST or is 0, clear up_bound{,_p1} and later ignore tests
6511         that need the upper bound.  Subtract offset from
6512         get_addr_base_and_unit_offset only if positive and subtract it
6513         before division by eltsize rather than after it.
6515         PR debug/83084
6516         * valtrack.c (propagate_for_debug_subst, propagate_for_debug): Reset
6517         debug insns if they would contain UNSPEC_VOLATILE or volatile asm.
6518         (dead_debug_insert_temp): Likewise, but also ignore even non-volatile
6519         asm.
6521         PR middle-end/82875
6522         * optabs.c (expand_doubleword_mult, expand_binop): Before calling
6523         expand_binop with *mul_widen_optab, make sure at least one of the
6524         operands doesn't have VOIDmode.
6526         PR debug/83034
6527         * dwarf2out.c (mem_loc_descriptor): Handle VEC_SERIES.
6529         PR rtl-optimization/82044
6530         PR tree-optimization/82042
6531         * dse.c (record_store): Check for overflow.
6532         (check_mem_read_rtx): Properly check for overflow if width == -1, call
6533         add_wild_read instead of clear_rhs_from_active_local_stores on
6534         overflow and log it into dump_file.
6536 2017-11-22  Richard Biener  <rguenther@suse.de>
6538         * gimple-iterator.c (gimple_find_edge_insert_loc): Ignore
6539         fake edges to exit when looking for a place to insert.
6540         * tree-ssa-pre.c (clear_expression_ids): Inline into callers
6541         and remove.
6542         (insert_into_preds_of_block): Commit edge insertion immediately,
6543         assert that doesn't require new BBs.
6544         (fini_pre): Release expressions.
6545         (pass_pre::execute): Shuffle things around a bit, if the fn
6546         is too large do not compute AVAIL either as this is really the
6547         quadratic bit.
6549 2017-11-22  Richard Biener  <rguenther@suse.de>
6551         PR tree-optimization/83089
6552         * tree-if-conv.c (pass_if_conversion::execute): If anything
6553         changed reset SCEV and free the number of iteration estimates.
6555 2017-11-21  Martin Sebor  <msebor@redhat.com>
6557         PR tree-optimization/82945
6558         * calls.h (warn_nonstring_bound): Remove unused function.
6560 2017-11-21  Martin Sebor  <msebor@redhat.com>
6562         PR tree-optimization/82945
6563         * builtins.c (expand_builtin_strlen): Call maybe_warn_nonstring_arg.
6564         * calls.h (maybe_warn_nonstring_arg): Declare new function.
6565         * calls.c (get_attr_nonstring_decl, maybe_warn_nonstring_arg): New
6566         functions.
6567         (initialize_argument_information): Call maybe_warn_nonstring_arg.
6568         * calls.h (get_attr_nonstring_decl): Declare new function.
6569         * doc/extend.texi (attribute nonstring): Update.
6570         * gimple-fold.c (gimple_fold_builtin_strncpy): Call
6571         get_attr_nonstring_decl and handle it.
6572         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Same.  Improve
6573         detection of nul-termination.
6574         (strlen_to_stridx): Change to a pointer.
6575         (handle_builtin_strlen, handle_builtin_stxncpy): Adjust.
6576         (pass_strlen::execute): Same.
6578 2017-11-21  Sergey Shalnov  <Sergey.Shalnov@intel.com>
6580         * config/i386/i386-opts.h (enum prefer_vector_width): Added new enum
6581         for the new option -mprefer-vector-width=[none|128|256|512].
6582         * config/i386/i386.c (ix86_target_string): remove old style options
6583         -mprefer-avx256 and make -mprefer-avx128 as alias.
6584         (ix86_option_override_internal):  Apply defaults for the
6585         -mprefer-vector-width=[128|256] option.
6586         * config/i386/i386.h (TARGET_PREFER_AVX128, TARGET_PREFER_AVX256):
6587         Implement macros to work with -mprefer-vector-width=.
6588         * config/i386/i386.opt: Implemented option
6589         -mprefer-vector-width=[none|128|256|512].
6590         * doc/invoke.texi: Documentation for
6591         -mprefer-vector-width=[none|128|256|512].
6593 2017-11-21  Pat Haugen  <pthaugen@us.ibm.com>
6595         * config/rs6000/ppc-asm.h (f50, vs50): Fix values.
6597 2017-11-21  Jonathan Wakely  <jwakely@redhat.com>
6599         * doc/invoke.texi (-Wmaybe-uninitialized): Rephrase for clarity.
6601 2017-11-21  Marc Glisse  <marc.glisse@inria.fr>
6603         * doc/generic.texi: Document POINTER_DIFF_EXPR, update
6604         POINTER_PLUS_EXPR.
6605         * cfgexpand.c (expand_debug_expr): Handle POINTER_DIFF_EXPR.
6606         * expr.c (expand_expr_real_2): Likewise.
6607         * fold-const.c (const_binop, fold_addr_of_array_ref_difference,
6608         fold_binary_loc): Likewise.
6609         * match.pd (X-X, P+(Q-P), &D-P, (P+N)-P, P-(P+N), (P+M)-(P+N),
6610         P-Q==0, -(A-B), X-Z<Y-Z, (X-Z)-(Y-Z), Z-X<Z-Y, (Z-X)-(Z-Y),
6611         (A-B)+(C-A)): New transformations for POINTER_DIFF_EXPR, based on
6612         MINUS_EXPR transformations.
6613         * optabs-tree.c (optab_for_tree_code): Handle POINTER_DIFF_EXPR.
6614         * tree-cfg.c (verify_expr, verify_gimple_assign_binary): Likewise.
6615         * tree-inline.c (estimate_operator_cost): Likewise.
6616         * tree-pretty-print.c (dump_generic_node, op_code_prio,
6617         op_symbol_code): Likewise.
6618         * tree-vect-stmts.c (vectorizable_operation): Likewise.
6619         * vr-values.c (extract_range_from_binary_expr): Likewise.
6620         * varasm.c (initializer_constant_valid_p_1): Likewise.
6621         * tree.def: New tree code POINTER_DIFF_EXPR.
6623 2017-11-21  Uros Bizjak  <ubizjak@gmail.com>
6625         * config/i386/i386.md (*bswap<mode>2_movbe): Add
6626         integer suffix to movbe mnemonic.
6627         (*bswaphi2_movbe): Ditto.
6628         (bswaphi_lowpart): Merge with *bswaphi_lowpart_1.
6630 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6632         PR c++/83045
6633         * tree-cfg.c (pass_warn_function_return::execute): Formatting fix.
6634         Also warn if seen __builtin_unreachable () call with BUILTINS_LOCATION.
6635         Use LOCATION_LOCUS when comparing against UNKNOWN_LOCATION.
6637 2017-11-21  Martin Liska  <mliska@suse.cz>
6639         * tree-inline.c (expand_call_inline): Remove not needed
6640         xstrdup_for_dump.
6642 2017-11-21  James Cowgill  <James.Cowgill@imgtec.com>
6643             Jakub Jelinek  <jakub@redhat.com>
6645         PR target/82880
6646         * config/mips/frame-header-opt.c (mips_register_frame_header_opt):
6647         Remove static keyword from f variable.
6649 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6651         PR tree-optimization/83086
6652         * gimple-ssa-store-merging.c
6653         (imm_store_chain_info::try_coalesce_bswap): Test this_n.base_addr
6654         rather than n.base_addr.
6656 2017-11-21  Martin Liska  <mliska@suse.cz>
6658         PR rtl-optimization/82044
6659         PR tree-optimization/82042
6660         * dse.c (check_mem_read_rtx): Check for overflow.
6662 2017-11-21  Aldy Hernandez  <aldyh@redhat.com>
6664         * tree-ssa-threadbackward.c (find_jump_threads_backwards): Fix
6665         typo in comment.
6667 2017-11-21  Claudiu Zissulescu  <claziss@synopsys.com>
6669         * cfgrtl.c (force_nonfallthru_and_redirect): Don't split a call
6670         and its corresponding call arg location note.
6672 2017-11-21  Claudiu Zissulescu  <claziss@synopsys.com>
6673             Andrew Burgess  <andrew.burgess@embecosm.com>
6675         * config/arc/arc-protos.h (arc_compute_frame_size): Delete
6676         declaration.
6677         (arc_return_slot_offset): Likewise.
6678         (arc_eh_return_address_location): New declaration.
6679         * config/arc/arc.c (TARGET_BUILTIN_SETJMP_FRAME_VALUE): Define.
6680         (MUST_SAVE_REGISTER): Add exception handler case.
6681         (MUST_SAVE_RETURN_ADDR): Likewise.
6682         (arc_frame_pointer_required): Likewise.
6683         (arc_frame_pointer_needed): New function.
6684         (arc_compute_frame_size): Changed.
6685         (arc_expand_prologue): Likewise.
6686         (arc_expand_epilogue): Likewise.
6687         (arc_initial_elimination_offset): Likewise.
6688         (arc_return_slot_offset): Delete.
6689         (arc_eh_return_address_location): New function.
6690         (arc_builtin_setjmp_frame_value): Likewise.
6691         * config/arc/arc.h (EH_RETURN_DATA_REGNO): Use 2 registers.
6692         (EH_RETURN_STACKADJ_RTX): Define.
6693         (EH_RETURN_HANDLER_RTX): Likewise.
6694         * config/arc/arc.md (eh_return): Delete.
6696 2017-11-21  Aldy Hernandez  <aldyh@redhat.com>
6698         * print-tree.h (debug_vec_tree): Remove prototype.
6699         * gdbinit.in (pvt): Remove macro.
6701 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6703         PR tree-optimization/83047
6704         * gimple-ssa-store-merging.c
6705         (imm_store_chain_info::output_merged_store): If the loads with the
6706         same vuse are in different basic blocks, for load_gsi pick a load
6707         location that is dominated by the other loads.
6709         PR c++/83059
6710         * config/i386/i386.c (ix86_memmodel_check): Start
6711         -Winvalid-memory-model diagnostics with lowercase letter.
6713         PR debug/82718
6714         * dwarf2out.c (dw_loc_list): If crtl->has_bb_partition, temporarily
6715         set in_cold_section_p to the partition containing loc_list->first.
6716         When seeing loc_list->last_before_switch node, update secname and
6717         perform range_across_switch second partition handling only after that.
6719         PR debug/82933
6720         * run-rtl-passes.c: Include debug.h.
6721         (run_rtl_passes): Call debug_hooks->assembly_start.
6722         * dwarf2out.c (dwarf2out_assembly_start): Return early if invoked
6723         multiple times.
6725         PR target/82981
6726         * internal-fn.c (expand_mul_overflow): Use OPTAB_WIDEN instead of
6727         OPTAB_DIRECT in calls to expand_simple_binop.
6729 2017-11-20  David Malcolm  <dmalcolm@redhat.com>
6731         PR c/81404
6732         * Makefile.in (C_COMMON_OBJS): Add c-family/known-headers.o.
6734 2017-11-20  David Malcolm  <dmalcolm@redhat.com>
6736         PR c++/72786
6737         * spellcheck.h (best_match::blithely_get_best_candidate): New
6738         accessor.
6740 2017-11-20  Jakub Jelinek  <jakub@redhat.com>
6742         * config/i386/i386.c (parse_mtune_ctrl_str): Start diagnostics
6743         with lower case letter.
6745 2017-11-20  Uros Bizjak  <ubizjak@gmail.com>
6747         * config/i386/i386.md (bswaphi2): New expander.
6748         (*bswaphi2_movbe): New insn pattern.
6749         (bswaphi -> rorhi pepehole2): New peephole pattern.
6751 2017-11-20  Jeff Law  <law@redhat.com>
6753         * Makefile.in (OBJS): Add gimple-ssa-evrp-analyze.o.
6754         * gimple-ssa-evrp-analyze.c: New file pulled from gimple-ssa-evrp.c.
6755         * gimple-ssa-evrp-analyze.h: New file pulled from gimple-ssa-evrp.c.
6756         * gimple-ssa-evrp.c: Remove bits moved into new files.  Include
6757         gimple-ssa-evrp-analyze.h.
6759         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Do not
6760         set BB_VISITED here.
6761         (evrp_range_analyzer::enter): Set BB_VISITED here instead.
6763 2017-11-20  Bin Cheng  <bin.cheng@arm.com>
6765         * tree-predcom.c: Add general comment on Store-Store chains.
6766         (split_data_refs_to_components): Postpone clearing eliminate_store_p
6767         flag in component.
6768         (get_chain_last_ref_at): Rename into...
6769         (get_chain_last_write_at): ...this.
6770         (get_chain_last_write_before_load): New function.
6771         (add_ref_to_chain): Promote type of chain from CT_STORE_LOAD to
6772         CT_STORE_STORE when write reference is added.
6773         (determine_roots_comp): Support load ref in CT_STORE_STORE chains.
6774         (is_inv_store_elimination_chain): Update get_chain_last_write_at call.
6775         (initialize_root_vars_store_elim_1): Ditto.
6776         (initialize_root_vars_store_elim_2): Ditto.  Replace rhs once default
6777         definition is created.
6778         (execute_pred_commoning_chain): Support load ref in CT_STORE_STORE
6779         chain by replacing it with dominant stored value.
6781 2017-11-20  Bin Cheng  <bin.cheng@arm.com>
6783         * tree-predcom.c (add_ref_to_chain): Remove check on distance.
6785 2017-11-20  Marc Glisse  <marc.glisse@inria.fr>
6787         * vr-values.c (extract_range_from_binary_expr): Use a full range
6788         for VR_VARYING.
6790 2017-11-20  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6792         * config/arm/arm.md (R4_REGNUM): Define constant.
6793         (nonsecure_call_internal): Remove r4 clobber.
6794         (nonsecure_call_value_internal): Likewise.
6795         * config/arm/thumb1.md (nonsecure_call_reg_thumb1_v5): Remove second
6796         clobber and resequence match_operands.
6797         (nonsecure_call_value_reg_thumb1_v5): Likewise.
6798         * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Likewise.
6799         (nonsecure_call_value_reg_thumb2): Likewise.
6801 2017-11-20  Jakub Jelinek  <jakub@redhat.com>
6803         PR tree-optimization/78821
6804         * gimple-ssa-store-merging.c (find_bswap_or_nop_load): Give up
6805         if base is TARGET_MEM_REF.  If base is not MEM_REF, set base_addr
6806         to the address of the base rather than the base itself.
6807         (find_bswap_or_nop_1): Just use pointer comparison for vuse check.
6808         (find_bswap_or_nop_finalize): New function.
6809         (find_bswap_or_nop): Use it.
6810         (bswap_replace): Return a tree rather than bool, change first
6811         argument from gimple * to gimple_stmt_iterator, allow inserting
6812         into an empty sequence, allow ins_stmt to be NULL - then emit
6813         all stmts into gsi.  Fix up MEM_REF address gimplification.
6814         (pass_optimize_bswap::execute): Adjust bswap_replace caller.
6815         (struct store_immediate_info): Add N and INS_STMT non-static
6816         data members.
6817         (store_immediate_info::store_immediate_info): Initialize them
6818         from newly added ctor args.
6819         (merged_store_group::apply_stores): Formatting fixes.  Sort by
6820         bitpos at the end.
6821         (stmts_may_clobber_ref_p): For stores call also
6822         refs_anti_dependent_p.
6823         (gather_bswap_load_refs): New function.
6824         (imm_store_chain_info::try_coalesce_bswap): New method.
6825         (imm_store_chain_info::coalesce_immediate_stores): Use it.
6826         (split_group): Handle LROTATE_EXPR and NOP_EXPR rhs_code specially.
6827         (imm_store_chain_info::output_merged_store): Fail if number of
6828         new estimated stmts is bigger or equal than old.  Handle LROTATE_EXPR
6829         and NOP_EXPR rhs_code.
6830         (pass_store_merging::process_store): Compute n and ins_stmt, if
6831         ins_stmt is non-NULL and the store rhs is otherwise invalid, use
6832         LROTATE_EXPR rhs_code.  Pass n and ins_stmt to store_immediate_info
6833         ctor.
6834         (pass_store_merging::execute): Calculate dominators.
6836         * tree-ssa-math-opts.c (nop_stats, bswap_stats, struct symbolic_number,
6837         BITS_PER_MARKER, MARKER_MASK, MARKER_BYTE_UNKNOWN, HEAD_MARKER, CMPNOP,
6838         CMPXCHG, do_shift_rotate, verify_symbolic_number_p,
6839         init_symbolic_number, find_bswap_or_nop_load, perform_symbolic_merge,
6840         find_bswap_or_nop_1, find_bswap_or_nop, pass_data_optimize_bswap,
6841         class pass_optimize_bswap, bswap_replace,
6842         pass_optimize_bswap::execute): Moved to ...
6843         * gimple-ssa-store-merging.c: ... this file.
6844         Include optabs-tree.h.
6845         (nop_stats, bswap_stats, do_shift_rotate, verify_symbolic_number_p,
6846         init_symbolic_number, find_bswap_or_nop_load, perform_symbolic_merge,
6847         find_bswap_or_nop_1, find_bswap_or_nop, bswap_replace): Put into
6848         anonymous namespace, remove static keywords.
6849         (pass_optimize_bswap::gate): Test BITS_PER_UNIT == 8 here...
6850         (pass_optimize_bswap::execute): ... rather than here.  Formatting fix.
6852 2017-11-20  Jan Hubicka  <hubicka@ucw.cz>
6854         PR bootstrap/83062
6855         * ipa-inline.c (can_inline_edge_p): Fix typo in previous patch.
6857 2017-11-20  Aldy Hernandez  <aldyh@redhat.com>
6859         * vec.h (debug_helper): New function.
6860         (DEFINE_DEBUG_VEC): New macro.
6861         * hash-set.h (debug_helper): New function.
6862         (DEFINE_DEBUG_HASH_SET): New macro.
6863         * cfg.c (debug_slim (edge)): New function.
6864         Call DEFINE_DEBUG_VEC for edges.
6865         Call DEFINE_DEBUG_HASH_SET for edges.
6866         * cfghooks.c (debug_slim (basic_block)): New function.
6867         Call DEFINE_DEBUG_VEC for basic blocks.
6868         Call DEFINE_DEBUG_HASH_SET for basic blocks.
6869         * print-tree.c (debug_slim): New function to handle trees.
6870         Call DEFINE_DEBUG_VEC for trees.
6871         Call DEFINE_DEBUG_HASH_SET for trees.
6872         (debug (vec<tree, va_gc>) &): Remove.
6873         (debug (<vec<tree, va_gc>) *): Remove.
6874         * print-rtl.c (debug_slim): New function to handle const_rtx.
6875         Call DEFINE_DEBUG_VEC for rtx_def.
6876         Call DEFINE_DEBUG_VEC for rtx_insn.
6877         Call DEFINE_DEBUG_HASH_SET for rtx_def.
6878         Call DEFINE_DEBUG_HASH_SET for rtx_insn.
6879         * sel-sched-dump.c (debug (vec<rtx_insn *> &): Remove.
6880         (debug (vec<rtx_insn *> *ptr): Remove.
6881         (debug_insn_vector): Remove.
6882         * stor-layout.c (debug_rli): Call debug() instead of debug_vec_tree.
6884 2017-11-20  Tom de Vries  <tom@codesourcery.com>
6886         PR rtl-optimization/82020
6887         * simplify-rtx.c (simplify_ternary_operation): Fix comparison mode of
6888         IF_THEN_ELSE condition.
6890 2017-11-19  Jeff Law  <law@redhat.com>
6892         * tree-ssa-dom.c (record_equivalences_from_phis): Fix handling
6893         of degenerates resulting from ignoring an edge.
6895 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6897         PR ipa/81360
6898         * ipa-inline.c (can_inline_edge_p): Also check that caller is optimized
6900 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6902         PR ipa/83001
6903         * profile-count.c (profile_count::to_sreal_scale): Fix return value
6904         for uninitialied counts.
6906 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6908         PR ipa/60243
6909         * tree-inline.c (estimate_num_insns): Set to 1 at least.
6911 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6913         PR target/82713
6914         * i386.c (ix86_builtin_vectorization_cost): Be ready for insane types.
6916 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6918         * config/arc/arc.h (FUNCTION_PROFILER): Remove semicolon after
6919         "do while (0)".
6921 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6923         * config/phoenix.h (TARGET_OS_CPP_BUILTINS): Remove semicolon after
6924         "do {} while (0)".
6926 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6928         * config/visium/visium.h (ASM_OUTPUT_CASE_END): Remove semicolon after
6929         macro body.
6931 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6933         * config/ft32/ft32.h (REGISTER_TARGET_PRAGMAS): Remove semicolon after
6934         "do {} while (0)".
6935         * config/spu/spu.h (REGISTER_TARGET_PRAGMAS): Same.
6937 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6939         * config/mcore/mcore-elf.h (MCORE_EXPORT_NAME): Remove semicolon after
6940         "do {} while (0)".
6941         * config/mcore/mcore.h (ASM_OUTPUT_ALIGNED_COMMON): After missing
6942         semicolon after MCORE_EXPORT_NAME call.
6944 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6946         PR target/82961
6947         * vmsdbgout.c (vmsdbgout_early_finish): New function.
6948         (vmsdbg_debug_hooks): Set early_finish field to vmsdbgout_early_finish.
6950 2017-11-18  Jan Hubicka  <hubicka@ucw.cz>
6952         * cgraphclones.c (cgraph_edge::clone): Rename gcov_count to prof_count.
6953         (cgraph_edge::clone): Cleanup updating of profile.
6954         * ipa-cp.c (update_profiling_info): Likewise.
6955         * ipa-inline-transform.c (inline_transform): Likewise.
6956         * ipa-inline.c (inline_small_functions): Add missing space to dump.
6957         * ipa-split.c (execute_split_functions): Do not split when function
6958         is cold.
6959         * predict.c (estimate_bb_frequencies): Cleanup updating of profile.
6960         * profile-count.c (profile_count::dump): Add global0.
6961         (profile_count::to_cgraph_frequency): Do not ICE when entry is
6962         undefined.
6963         (profile_count::to_sreal_scale): Likewise.
6964         (profile_count::adjust_for_ipa_scaling): Fix typo in comment.
6965         (profile_count::combine_with_ipa_count): New function.
6966         * profile-count.h (profile_guessed_global0adjusted): New.
6967         (profile_count::adjusted_zero): New.
6968         (profile_count::global0adjusted): New.
6969         (profile_count::combine_with_ipa_count): New.
6970         * tree-inline.c (copy_edges_for_bb): Add NUM/DEN arugment;
6971         correct profile of return block of split functions.
6972         (copy_cfg_body): Remove unused profile_count.
6973         (copy_body): Likewise.
6974         (expand_call_inline): Update.
6975         (tree_function_versioning): Update.
6977 2017-11-18  Aldy Hernandez  <aldyh@redhat.com>
6979         * hash-set.h (hash_set::empty): New.
6980         * tree-ssa-threadbackward.h: Delete.
6981         * tree-ssa-threadbackward.c (class thread_jumps): New.
6982         Move max_threaded_paths into class.
6983         (fsm_find_thread_path): Remove arguments that are now in class.
6984         (profitable_jump_thread_path): Rename to...
6985         (thread_jumps::profitable_jump_thread_path): ...this.
6986         (convert_and_register_jump_thread_path): Rename to...
6987         (thread_jumps::convert_and_register_current_path): ...this.
6988         (check_subpath_and_update_thread_path): Rename to...
6989         (thread_jumps::check_subpath_and_update_thread_path): ...this.
6990         (register_jump_thread_path_if_profitable): Rename to...
6991         (thread_jumps::register_jump_thread_path_if_profitable): ...this.
6992         (handle_phi): Rename to...
6993         (thread_jumps::handle_phi): ...this.
6994         (handle_assignment): Rename to...
6995         (thread_jumps::handle_assignment): ...this.
6996         (fsm_find_control_statement_thread_paths): Rename to...
6997         (thread_jumps::fsm_find_control_statement_thread_paths): ...this.
6998         (find_jump_threads_backwards): Rename to...
6999         (thread_jumps::find_jump_threads_backwards): ...this.
7000         Initialize path local data.
7001         (pass_thread_jumps::execute): Call find_jump_threads_backwards
7002         from within thread_jumps class.
7003         (pass_early_thread_jumps::execute): Same.
7005 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
7007         * cgraph.c (cgraph_node::dump): Do IPA sanity checking on IPA counts.
7009 2017-11-17  Steve Ellcey  <sellcey@cavium.com>
7011         * config/aarch64/aarch64-simd.md (fnma<mode>4): Move neg operator
7012         to canonical location.
7014 2017-11-17  Steve Ellcey  <sellcey@cavium.com>
7016         PR target/81356
7017         * config/aarch64/aarch64.c (aarch64_use_by_pieces_infrastructure_p):
7018         Remove.
7019         (TARGET_USE_BY_PIECES_INFRASTRUCTURE_P): Remove define.
7021 2017-11-17  H.J. Lu  <hongjiu.lu@intel.com>
7023         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Use
7024         rcrt1.o%s/grcrt1.o%s for -static-pie.
7026 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
7028         * i386.c (ix86_multiplication_cost, ix86_division_cost,
7029         ix86_shift_rotate_cost): Break out from ...
7030         (ix86_rtx_costs): ... here.
7031         (ix86_add_stmt_cost): Use rtx cost machinery to compute cost of
7032         vector operations.
7034 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
7036         * predict.c (determine_unlikely_bbs): Set cgraph node count to 0
7037         when entry block was promoted unlikely.
7038         (estimate_bb_frequencies): Increase frequency scale.
7039         * profile-count.h (profile_count): Export precision info.
7041 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
7043         * tree-tailcall.c (eliminate_tail_call): Be more careful about not
7044         disturbin profile of entry block.
7046 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
7048         * ipa-fnsummary.c (estimate_node_size_and_time): Be more tolerant for
7049         roundoff errors.
7051 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
7053         * ipa-cp.c (update_profiling_info): Handle conversion to local profile.
7054         * tree-cfg.c (execute_fixup_cfg): Do fixup same way as inliner does.
7056 2017-11-17  Jeff Law  <law@redhat.com>
7058         * gimple-ssa-evrp.c (class evrp_range_analyzer): New class extracted
7059         from evrp_dom_walker class.  Various methods moved into new class.
7060         (evrp_range_analyzer::evrp_range_analyzer): Constructor for new class.
7061         (evrp_range_analyzer::enter): New method.
7062         (evrp_range_analyzer::leave): New method.
7063         (evrp_dom_walker): Remove delegators no longer needed by this class.
7064         Replace vr_values data member with evrp_range_analyzer
7066         * gimple-ssa-evrp.c (evrp_dom_walker::record_ranges_from_phis): New
7067         method extracted from evrp_dom_walker::before_dom_children.
7068         (evrp_dom_walker::record_ranges_from_stmt): Likewise.
7069         (evrp_dom_walker::record_ranges_from_incoming_edge): Likewise.
7071         * gimple-ssa-evrp.c (evrp_dom_walker): Add cleanup method.
7072         Add private copy constructor and move assignment operators.
7073         Privatize methods and class data where trivially possible.
7074         (evrp_dom_walker::cleanup): New function, extracted from
7075         execute_early_vrp.  Simplify access to class data.
7077         * vr-values.h (get_output_for_vrp): Prototype.
7078         * vr-values.c (get_output_for_vrp): New function extracted from
7079         vrp_visit_assignment_or_call and extract_range_from_stmt.
7080         (vrp_visit_assignment_or_call): Use get_output_for_vrp.  Simplify.
7082 2017-11-17  Luis Machado  <luis.machado@linaro.org>
7084         * config/aarch64/aarch64.c
7085         (qdf24xx_prefetch_tune) <default_opt_level>: Set to -1.
7086         (qdf24xx_tunings) <autoprefetcher_model>: Set to
7087         tune_params::AUTOPREFETCHER_WEAK.
7089 2017-11-17  Tamar Christina  <tamar.christina@arm.com>
7091         PR target/82641
7092         * config/arm/arm.c (arm_valid_target_attribute_rec):
7093         Parse "arch=" and "+<ext>".
7094         (arm_valid_target_attribute_tree): Re-init global options.
7095         (arm_option_override): Make non-static.
7096         (arm_options_perform_arch_sanity_checks): Make errors fatal.
7097         * gcc/config/arm/arm-c.c (__ARM_FEATURE_CMSE): Support undef.
7098         (__ARM_FEATURE_CRC32): Support undef.
7099         * config/arm/arm_acle.h (__ARM_FEATURE_CRC32): Replace with pragma.
7100         * doc/extend.texi (ARM Function Attributes): Add pragma and target.
7102 2017-11-17  David Malcolm  <dmalcolm@redhat.com>
7104         * gdbinit.in (break-on-diagnostic): New command.
7106 2017-11-17  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
7108         * config/i386/i386.c (ix86_expand_epilogue): Change simple
7109         return to indirect jump for EH return if control-flow
7110         protection is enabled. Change explicit 'false' argument in
7111         pro_epilogue_adjust_stack with a value of
7112         flag_cf_protection.
7113         * config/i386/i386.md (simple_return_indirect_internal):
7114         Remove SImode restriction to support 64-bit.
7116 2017-11-17  Segher Boessenkool  <segher@kernel.crashing.org>
7118         * combine.c (added_notes_insn): New.
7119         (try_combine): Handle added_notes_insn like added_links_insn.
7120         Rewrite return value code.
7121         (distribute_notes): Set added_notes_insn to the earliest insn we added
7122         a note to.
7124 2017-11-17  Segher Boessenkool  <segher@kernel.crashing.org>
7126         PR rtl-optimization/82621
7127         * combine.c (try_combine): Do not split PARALLELs of two SETs if the
7128         dest of one of those SETs is unused.
7130 2017-11-17  Richard Biener  <rguenther@suse.de>
7132         PR fortran/83017
7133         * tree-core.h (enum annot_expr_kind): Add annot_expr_parallel_kind.
7134         * tree-pretty-print.c (dump_generic_node): Handle
7135         annot_expr_parallel_kind.
7136         * tree-cfg.c (replace_loop_annotate_in_block): Likewise.
7137         * gimplify.c (gimple_boolify): Likewise.
7139 2017-11-17  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
7141         * config.gcc (extra_headers): Add cet.h for x86 targets.
7142         * config/i386/cet.h: New file.
7143         * doc/install.texi: Add --enable-cet/--disable-cet.
7145 2017-11-17  Richard Biener  <rguenther@suse.de>
7147         PR tree-optimization/83017
7148         * tree-parloops.c (MIN_PER_THREAD): Use --param parloops-min-per-thread.
7149         (gen_parallel_loop): Properly count iterations.
7150         (parallelize_loops): Handle loop->can_be_parallel independent
7151         of flag_loop_parallelize_all.  Make static profitability test match
7152         the runtime one.
7153         * params.def (PARAM_PARLOOPS_MIN_PER_THREAD): New.
7154         * invoke.texi (parloops-min-per-thread): Document.
7156 2017-11-17  Vineet Gupta  <vgupta@synopsys.com>
7158         * config/arc/linux.h: GLIBC_DYNAMIC_LINKER update per glibc
7159         upstreaming review comments.
7161 2017-11-17  Tamar Christina  <tamar.christina@arm.com>
7163         * expr.c (copy_blkmode_to_reg): Fix bitsize for targets
7164         with fast unaligned access.
7165         * doc/sourcebuild.texi (word_mode_no_slow_unalign): New.
7167 2017-11-17  Thomas Preud'homme  <thomas.preudhomme@arm.com>
7169         * config/arm/arm.c (cmse_nonsecure_entry_clear_before_return): Allocate
7170         to_clear_arg_regs_bitmap to the same size as to_clear_bitmap.
7172 2017-11-17  Richard Biener  <rguenther@suse.de>
7174         * tree-ssa-pre.c (phi_translate_1): Remove redundant constant
7175         folding of references.
7177 2017-11-17  Qing Zhao  <qing.zhao@oracle.com>
7179         PR middle-end/78809
7180         * gimple-fold.c (gimple_fold_builtin_string_compare): Add handling
7181         of replacing call to strncmp with corresponding call to strcmp when
7182         meeting conditions.
7184 2017-11-17  Sergey Shalnov  <Sergey.Shalnov@intel.com>
7186         * config/i386/x86-tune.def (X86_TUNE_AVX256_OPTIMAL): Add tuning
7187         option prefer-avx256 for skylake-avx512 configuration.
7188         * config/i386/i386.c (ix86_option_override_internal): Ditto.
7189         (get_builtin_code_for_version): Ditto.
7191 2017-11-17  Chung-Ju Wu  <jasonwucj@gmail.com>
7192             Monk Chiang  <sh.chiang04@gmail.com>
7194         * config/nds32/nds32.h (FIRST_PSEUDO_REGISTER): Modify.
7195         (FIXED_REGISTERS): Reserve more register numbers.
7196         (CALL_USED_REGISTERS): Likewise.
7197         (REG_ALLOC_ORDER): Likewise.
7198         (REG_CLASS_CONTENTS): Likewise.
7199         (REGISTER_NAMES): Likewise.
7201 2017-11-17  Chung-Ju Wu  <jasonwucj@gmail.com>
7202             Kito Cheng  <kito.cheng@gmail.com>
7204         * config/nds32/nds32-modes.def: Add vector mode V4QI V2HI V8QI V4HI
7205         V2SI.
7206         * config/nds32/iterators.md: Add vector mode iterators and attributes.
7208 2017-11-16  Steven Munroe  <munroesj@gcc.gnu.org>
7210         * config/rs6000/mmintrin.h (_mm_add_pi32[_ARCH_PWR]): Correct
7211         parameter list for vec_splats.
7213 2017-11-16  Joseph Myers  <joseph@codesourcery.com>
7215         * doc/invoke.texi (-std=c17): Refer to 2018 expected publication
7216         date of C17.
7217         (-std=c18, -std=iso9899:2018, -std=gnu18): Document option aliases.
7219 2017-11-16  Thomas Koenig  <tkoenig@gcc.gnu.org>
7221         PR bootstrap/82856
7222         * doc/install.texi: Document incompatibility of Perl >=5.6.26
7223         with the required version of automake 1.11.6.
7225 2017-11-16  Pat Haugen  <pthaugen@us.ibm.com>
7227         * rs6000/power9.md (power9fpdiv): New automaton and cpu_unit defined
7228         for it.
7229         (DU_C2_3_power9): Correct reservation combinations.
7230         (FP_DIV_power9, VEC_DIV_power9): New.
7231         (power9-alu): Split out rotate/shift...
7232         (power9-rot): ...to here, correct dispatch resource.
7233         (power9-cracked-alu, power9-mul, power9-mul-compare): Correct dispatch
7234         resource.
7235         (power9-fp): Correct latency.
7236         (power9-sdiv): Add div/sqrt resource.
7237         (power9-ddiv): Correct latency, add div/sqrt resource.
7238         (power9-sqrt, power9-dsqrt): Add div/sqrt resource.
7239         (power9-vecfdiv, power9-vecdiv): Correct latency, add div/sqrt
7240         resource.
7241         (power9-qpdiv, power9-qpmul): Adjust resource usage.
7243 2017-11-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
7245         * config/rs6000/rs6000.c (rs6000_expand_builtin): Do not do the
7246         switch statement mapping KF built-ins to TF built-ins if we don't
7247         have the proper ISA 3.0 assembler support.
7249 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7251         * tree-emutls.c (lower_emutls_data): Remove unused bb_freq.
7252         (lower_emutls_function_body): Do not compute it.
7254 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7256         * ipa-split.c (split_bb_info): Turn time to sreal.
7257         (split_point): Likewise.
7258         (dump_split_point): Likewise.
7259         (fine_split_points): Likewise.
7260         (execute_split_functions): Only zero split_bbs; turn time to sreals.
7262 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7264         * ipa-fnsummary.c (analyze_function_body): Accumulate time consistently
7265         in sreal.
7267 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7269         * predict.c (combine_predictions_for_bb): Preserve zero predicted
7270         edges.
7271         (expensive_function_p): Remove useless assert.
7272         (determine_unlikely_bbs): Propagate also forward; determine cold blocks
7274 2017-11-16  Martin Sebor  <msebor@redhat.com>
7276         PR tree-optimization/82588
7277         PR tree-optimization/82583
7278         * tree-vrp.c (check_array_ref): Handle flexible array members,
7279         string literals, and inner indices.
7280         (search_for_addr_array): Add detail to diagnostics.
7282 2017-11-16  Nathan Sidwell  <nathan@acm.org>
7284         PR c++/82836
7285         PR c++/82737
7286         * tree.h (COPY_DECL_RTL): Rename parms for clarity.
7287         (SET_DECL_ASSEMBLER_NAME): Forward to
7288         overwrite_decl_assembler_name.
7289         (COPY_DECL_ASSEMBLER_NAME): Rename parms for clarity.
7290         (overwrite_decl_assembler_name): Declare.
7291         * tree.c (overwrite_decl_assembler_name): New.
7292         * langhooks-def.h (lhd_overwrite_decl_assembler_name): Declare.
7293         (LANG_HOOKS_OVERWRITE_DECL_ASSEMBLER_NAME): Provide default.
7294         (LANG_HOOKS_INITIALIZER): Add it.
7295         * langhooks.h (struct lang_hooks): Add overwrite_decl_assembler_name.
7296         * langhooks.c (lhd_set_decl_assembler_name): Use
7297         SET_DECL_ASSEMBLER_NAME.
7298         (lhd_overwrite_decl_assembler_name): Default implementation.
7300 2017-11-16  Wilco Dijkstra  <wdijkstr@arm.com>
7301             Jackson Woodruff  <jackson.woodruff@arm.com>
7303         PR tree-optimization/71026
7304         * match.pd: Canonicalize constant multiplies in division.
7306 2017-11-16  Wilco Dijkstra  <wdijkstr@arm.com>
7308         * opts.c (default_options_table): Add OPT_fomit_frame_pointer entry.
7309         * common/config/alpha/alpha-common.c (TARGET_OPTION_OPTIMIZATION_TABLE):
7310         Remove OPT_fomit_frame_pointer entry.
7311         * common/config/arc/arc-common.c: Likewise.
7312         * common/config/arm/arm-common.c: Likewise.
7313         * common/config/avr/avr-common.c: Likewise.
7314         * common/config/c6x/c6x-common.c: Likewise.
7315         * common/config/cr16/cr16-common.c: Likewise.
7316         * common/config/cris/cris-common.c: Likewise.
7317         * common/config/epiphany/epiphany-common.c: Likewise.
7318         * common/config/fr30/fr30-common.c: Likewise.
7319         * common/config/frv/frv-common.c: Likewise.
7320         * common/config/ia64/ia64-common.c: Likewise.
7321         * common/config/iq2000/iq2000-common.c: Likewise.
7322         * common/config/lm32/lm32-common.c: Likewise.
7323         * common/config/m32r/m32r-common.c: Likewise.
7324         * common/config/mcore/mcore-common.c: Likewise.
7325         * common/config/microblaze/microblaze-common.c: Likewise.
7326         * common/config/mips/mips-common.c: Likewise.
7327         * common/config/mmix/mmix-common.c: Likewise.
7328         * common/config/mn10300/mn10300-common.c: Likewise.
7329         * common/config/nios2/nios2-common.c: Likewise.
7330         * common/config/pa/pa-common.c: Likewise.
7331         * common/config/pdp11/pdp11-common.c: Likewise.
7332         * common/config/powerpcspe/powerpcspe-common.c: Likewise.
7333         * common/config/riscv/riscv-common.c: Likewise.
7334         * common/config/rs6000/rs6000-common.c: Likewise.
7335         * common/config/rx/rx-common.c: Likewise.
7336         * common/config/s390/s390-common.c: Likewise.
7337         * common/config/sh/sh-common.c: Likewise.
7338         * common/config/sparc/sparc-common.c: Likewise.
7339         * common/config/tilegx/tilegx-common.c: Likewise.
7340         * common/config/tilepro/tilepro-common.c: Likewise.
7341         * common/config/v850/v850-common.c: Likewise.
7342         * common/config/visium/visium-common.c: Likewise.
7343         * common/config/xstormy16/xstormy16-common.c: Likewise.
7344         * common/config/xtensa/xtensa-common.c: Likewise.
7345         * invoke.texi (-fomit-frame-pointer): Update documentation.
7347 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7349         * tree-cfg.c (gimple_find_sub_bbs): Do not compute freq.
7351 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7353         * cfg.c (scale_bbs_frequencies_int,
7354         cale_bbs_frequencies_gcov_type): Remove.
7355         * cfg.h (scale_bbs_frequencies_int,
7356         cale_bbs_frequencies_gcov_type): Remove.
7358 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7360         * tree-ssa-loop-manip.c
7361         (scale_dominated_blocks_in_loop): Update to profile counts.
7362         (tree_transform_and_unroll_loop): Likewise.
7364 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7366         * tree-vect-loop-manip.c (vect_do_peeling): Do not use
7367         scale_bbs_frequencies_int.
7369 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7371         * final.c (compute_alignments): Use counts rather than frequencies.
7373 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7375         * cfgloopanal.c: Include sreal.h
7376         (average_num_loop_insns): Use counts and sreal for accounting.
7378 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7380         * cfgloopmanip.c (duplicate_loop_to_header_edge): Cleanup profile
7381         manipulation.
7383 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7384             Kito Cheng  <kito.cheng@gmail.com>
7386         * config/nds32/constraints.md: Provide more constraints.
7387         * config/nds32/nds32.h (enum reg_class, REG_CLASS_NAMES,
7388         REG_CLASS_CONTENTS): Define R5_REG, R8_REG, and FRAME_POINTER_REG to
7389         support constraints usage.
7391 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7393         * config/nds32/constants.md (UNSPEC_VOLATILE_FUNC_RETURN): Remove.
7395 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7396             Kito Cheng  <kito.cheng@gmail.com>
7398         * config/nds32/nds32.opt: Add mext-perf, mext-perf2, mext-string.
7399         * config/nds32/nds32.opt: Refine the layout.
7400         * config/nds32/nds32.c (TARGET_EXT_PERF, TARGET_EXT_PERF2,
7401         TARGET_EXT_STRING): Support new options.
7402         * config/nds32/nds32.h: Likewise.
7403         * config/nds32/nds32.md: Likewise.
7404         * config/nds32/nds32-predicates.c: Likewise.
7405         * config/nds32/constraints.md: Likewise.
7406         * common/config/nds32/nds32-common.c: Likewise.
7408 2017-11-16  Julia Koval  <julia.koval@intel.com>
7410         PR target/82983
7411         * config/i386/gfniintrin.h: Add sse check.
7412         * config/i386/i386.c (ix86_expand_builtin): Fix gfni check.
7414 2017-11-16  Julia Koval  <julia.koval@intel.com>
7416         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VBMI2_SET,
7417         OPTION_MASK_ISA_AVX512VBMI2_UNSET): New.
7418         (ix86_handle_option): Handle -mavx512vbmi2.
7419         * config/i386/cpuid.h: Add bit_AVX512VBMI2.
7420         * config/i386/driver-i386.c (host_detect_local_cpu): Handle new bit.
7421         * config/i386/i386-c.c (__AVX512VBMI2__): New.
7422         * config/i386/i386.c (ix86_target_string): Handle -mavx512vbmi2.
7423         (ix86_valid_target_attribute_inner_p): Ditto.
7424         * config/i386/i386.h (TARGET_AVX512VBMI2, TARGET_AVX512VBMI2_P): New.
7425         * config/i386/i386.opt (mavx512vbmi2): New option.
7426         * doc/invoke.texi: Add new option.
7428 2017-11-16  Julia Koval  <julia.koval@intel.com>
7430         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8, _mm256_gf2p8mul_epi8,
7431         _mm_mask_gf2p8mul_epi8, _mm_maskz_gf2p8mul_epi8,
7432         _mm256_mask_gf2p8mul_epi8, _mm256_maskz_gf2p8mul_epi8,
7433         _mm512_mask_gf2p8mul_epi8, _mm512_maskz_gf2p8mul_epi8,
7434         _mm512_gf2p8mul_epi8): New intrinsics.
7435         * config/i386/i386-builtin-types.def
7436         (V64QI_FTYPE_V64QI_V64QI): New type.
7437         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8mulb_v64qi,
7438         __builtin_ia32_vgf2p8mulb_v64qi_mask, __builtin_ia32_vgf2p8mulb_v32qi,
7439         __builtin_ia32_vgf2p8mulb_v32qi_mask, __builtin_ia32_vgf2p8mulb_v16qi,
7440         __builtin_ia32_vgf2p8mulb_v16qi_mask): New builtins.
7441         * config/i386/sse.md (vgf2p8mulb_*): New pattern.
7442         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
7444 2017-11-15  Uros Bizjak  <ubizjak@gmail.com>
7446         * config/i386/i386.c (x86_print_call_or_nop): Emit 5 byte nop
7447         explicitly as a stream of bytes.
7449 2017-11-15  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
7451         * config/rs6000/altivec.h (vec_xst_be): New #define.
7452         * config/rs6000/altivec.md (altivec_vperm_<mode>_direct): Rename
7453         and externalize from *altivec_vperm_<mode>_internal.
7454         * config/rs6000/rs6000-builtin.def (XL_BE_V16QI): Remove macro
7455         instantiation.
7456         (XL_BE_V8HI): Likewise.
7457         (XL_BE_V4SI): Likewise.
7458         (XL_BE_V4SI): Likewise.
7459         (XL_BE_V2DI): Likewise.
7460         (XL_BE_V4SF): Likewise.
7461         (XL_BE_V2DF): Likewise.
7462         (XST_BE): Add BU_VSX_OVERLOAD_X macro instantiation.
7463         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Correct
7464         all array entries with these keys: VSX_BUILTIN_VEC_XL,
7465         VSX_BUILTIN_VEC_XL_BE, VSX_BUILTIN_VEC_XST.  Add entries for key
7466         VSX_BUILTIN_VEC_XST_BE.
7467         * config/rs6000/rs6000.c (altivec_expand_xl_be_builtin): Remove.
7468         (altivec_expand_builtin): Remove handling for VSX_BUILTIN_XL_BE_*
7469         built-ins.
7470         (altivec_init_builtins): Replace conditional calls to def_builtin
7471         for __builtin_vsx_ld_elemrev_{v8hi,v16qi} and
7472         __builtin_vsx_st_elemrev_{v8hi,v16qi} based on TARGET_P9_VECTOR
7473         with unconditional calls.  Remove calls to def_builtin for
7474         __builtin_vsx_le_be_<mode>.  Add a call to def_builtin for
7475         __builtin_vec_xst_be.
7476         * config/rs6000/vsx.md (vsx_ld_elemrev_v8hi): Convert define_insn
7477         to define_expand, and add alternate RTL generation for P8.
7478         (*vsx_ld_elemrev_v8hi_internal): New define_insn based on
7479         vsx_ld_elemrev_v8hi.
7480         (vsx_ld_elemrev_v16qi): Convert define_insn to define_expand, and
7481         add alternate RTL generation for P8.
7482         (*vsx_ld_elemrev_v16qi_internal): New define_insn based on
7483         vsx_ld_elemrev_v16qi.
7484         (vsx_st_elemrev_v8hi): Convert define_insn
7485         to define_expand, and add alternate RTL generation for P8.
7486         (*vsx_st_elemrev_v8hi_internal): New define_insn based on
7487         vsx_st_elemrev_v8hi.
7488         (vsx_st_elemrev_v16qi): Convert define_insn to define_expand, and
7489         add alternate RTL generation for P8.
7490         (*vsx_st_elemrev_v16qi_internal): New define_insn based on
7491         vsx_st_elemrev_v16qi.
7493 2017-11-15  H.J. Lu  <hongjiu.lu@intel.com>
7495         PR target/82990
7496         * config/i386/i386.c (pass_insert_vzeroupper::gate): Remove
7497         TARGET_AVX512ER check.
7498         (ix86_option_override_internal): Set MASK_VZEROUPPER if
7499         neither -mzeroupper nor -mno-zeroupper is used and
7500         TARGET_EMIT_VZEROUPPER is set.
7501         * config/i386/i386.h (TARGET_EMIT_VZEROUPPER): New.
7502         * config/i386/x86-tune.def: Add X86_TUNE_EMIT_VZEROUPPER.
7504 2017-11-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
7506         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
7507         folding of vector compares.
7508         (fold_build_vec_cmp): New helper function.
7509         (fold_compare_helper): New helper function.
7510         (builtin_function_type): Add compare builtins to the list of functions
7511         having unsigned arguments.  Cosmetic updates to comment indentation.
7512         * config/rs6000/vsx.md (vcmpneb, vcmpneh, vcmpnew): Update to specify
7513         the not+eq combination.
7515 2017-11-15  Bin Cheng  <bin.cheng@arm.com>
7517         PR tree-optimization/82726
7518         PR tree-optimization/70754
7519         * tree-predcom.c (order_drefs_by_pos): New function.
7520         (combine_chains): Move code setting has_max_use_after to...
7521         (try_combine_chains): ...here.  New parameter.  Sort combined chains
7522         according to position information.
7523         (tree_predictive_commoning_loop): Update call to above function.
7524         (update_pos_for_combined_chains, pcom_stmt_dominates_stmt_p): New.
7526 2017-11-15  Bin Cheng  <bin.cheng@arm.com>
7528         PR tree-optimization/82726
7529         Revert
7530         2017-01-23  Bin Cheng  <bin.cheng@arm.com>
7532         PR tree-optimization/70754
7533         * tree-predcom.c (stmt_combining_refs): New parameter INSERT_BEFORE.
7534         (reassociate_to_the_same_stmt): New parameter INSERT_BEFORE.  Insert
7535         combined stmt before it if not NULL.
7536         (combine_chains): Process refs reversely and compute dominance point
7537         for root ref.
7539         Revert
7540         2017-02-23  Bin Cheng  <bin.cheng@arm.com>
7542         PR tree-optimization/79663
7543         * tree-predcom.c (combine_chains): Process refs in reverse order
7544         only for ZERO length chains, and add explaining comment.
7546 2017-11-15  Tamar Christina  <tamar.christina@arm.com>
7548         * config/arm/arm-cpus.in (armv8_3, ARMv8_3a, armv8.3-a): New
7549         * config/arm/arm-tables.opt (armv8.3-a): Regenerated.
7550         * doc/invoke.texi (ARM Options): Add armv8.3-a.
7552 2017-11-15  Tamar Christina  <tamar.christina@arm.com>
7554         * config/arm/arm.h (TARGET_DOTPROD): Add arm_arch8_2.
7556 2017-11-15  Martin Liska  <mliska@suse.cz>
7558         * tree-cfg.c (pass_warn_function_return::execute):
7559         Compare warn_return_type for greater than zero.
7561 2017-11-15  Sebastian Peryt  <sebastian.peryt@intel.com>
7563         PR target/82941
7564         PR target/82942
7565         * config/i386/i386.c (pass_insert_vzeroupper): Modify gate condition
7566         to return true on Xeon and not on Xeon Phi.
7567         (ix86_check_avx256_register): Changed to ...
7568         (ix86_check_avx_upper_register): ... this. Add extra check for
7569         VALID_AVX512F_REG_OR_XI_MODE.
7570         (ix86_avx_u128_mode_needed): Changed
7571         ix86_check_avx256_register to ix86_check_avx_upper_register.
7572         (ix86_check_avx256_stores): Changed to ...
7573         (ix86_check_avx_upper_stores): ... this. Changed
7574         ix86_check_avx256_register to ix86_check_avx_upper_register.
7575         (ix86_avx_u128_mode_after): Changed
7576         avx_reg256_found to avx_upper_reg_found. Changed
7577         ix86_check_avx256_stores to ix86_check_avx_upper_stores.
7578         (ix86_avx_u128_mode_entry): Changed
7579         ix86_check_avx256_register to ix86_check_avx_upper_register.
7580         (ix86_avx_u128_mode_exit): Ditto.
7581         * config/i386/i386.h: (host_detect_local_cpu): New define.
7583 2017-11-15  Dominik Infuehr  <dominik.infuehr@theobroma-systems.com>
7585         * config/arm/xgene1.md (xgene1): Split into automatons
7586         xgene1_main, xgene1_decoder, xgene1_div, xgene1_simd.
7587         (xgene1_f_load): Adjust reservations and/or types.
7588         (xgene1_f_store): Likewise.
7589         (xgene1_load_pair): Likewise.
7590         (xgene1_store_pair): Likewise.
7591         (xgene1_fp_load1): Likewise.
7592         (xgene1_load1): Likewise.
7593         (xgene1_store1): Likewise.
7594         (xgene1_move): Likewise.
7595         (xgene1_alu): Likewise.
7596         (xgene1_simd): Likewise.
7597         (xgene1_bfm): Likewise.
7598         (xgene1_neon_load1): Likewise.
7599         (xgene1_neon_store1): Likewise.
7600         (xgene1_neon_logic): Likewise.
7601         (xgene1_neon_st1): Likewise.
7602         (xgene1_neon_ld1r): Likewise.
7603         (xgene1_alu_cond): Added.
7604         (xgene1_shift_reg): Likwise.
7605         (xgene1_bfx): Likewise.
7606         (xgene1_mul): Split into xgene1_mul32, xgene1_mul64.
7608 2017-11-15  Jakub Jelinek  <jakub@redhat.com>
7610         PR target/82981
7611         * internal-fn.c: Include gimple-ssa.h, tree-phinodes.h and
7612         ssa-iterators.h.
7613         (can_widen_mult_without_libcall): New function.
7614         (expand_mul_overflow): If only checking unsigned mul overflow,
7615         not result, and can do efficiently MULT_HIGHPART_EXPR, emit that.
7616         Don't use WIDEN_MULT_EXPR if it would involve a libcall, unless
7617         no other way works.  Add MULT_HIGHPART_EXPR + MULT_EXPR support.
7618         (expand_DIVMOD): Formatting fix.
7619         * expmed.h (expand_mult): Add NO_LIBCALL argument.
7620         * expmed.c (expand_mult): Likewise.  Use OPTAB_WIDEN rather
7621         than OPTAB_LIB_WIDEN if NO_LIBCALL is true, and allow it to fail.
7623         PR tree-optimization/82977
7624         * tree-ssa-strlen.c (strlen_optimize_stmt): Pass a reference to a copy
7625         constructed temporary to strlen_to_stridx.put.
7627 2017-11-15  Martin Liska  <mliska@suse.cz>
7629         * configure.ac: Remove -fkeep-inline-functions from coverage_flags.
7630         * configure: Regenerate.
7632 2017-11-15  Martin Liska  <mliska@suse.cz>
7634         PR target/82927
7635         * config/sh/sh-mem.cc: Use proper probability for
7636         REG_BR_PROB_NOTE.
7638 2017-11-14  Jeff Law  <law@redhat.com>
7640         * explow.c (anti_adjust_stack_and_probe_stack_clash): Avoid probing
7641         the red zone for stack_clash_protection_final_dynamic_probe targets
7642         when the total dynamic stack size is zero bytes.
7644         * tree-ssa-threadupdate.c (thread_through_all_blocks): Thread
7645         blocks is post order.
7647 2017-11-15  Alexandre Oliva  <aoliva@redhat.com>
7649         * dumpfile.h (TDF_COMPARE_DEBUG): New.
7650         * final.c (rest_of_clean_state): Set it for the
7651         -fcompare-debug dump.
7652         * tree-pretty-print.c (dump_generic_node): Omit OBJ_TYPE_REF
7653         class when TDF_COMPARE_DEBUG is set.
7655         * dwarf2out.c (gen_producer_string): Discard
7656         OPT_fcompare_debug.
7658 2017-11-15  Joseph Myers  <joseph@codesourcery.com>
7660         PR c/81156
7661         * doc/extend.texi (Other Builtins): Document __builtin_tgmath.
7662         * ginclude/tgmath.h (__tg_cplx, __tg_ldbl, __tg_dbl, __tg_choose)
7663         (__tg_choose_2, __tg_choose_3, __TGMATH_REAL_1_2)
7664         (__TGMATH_REAL_2_3): Remove macros.
7665         (__TGMATH_CPLX, __TGMATH_CPLX_2, __TGMATH_REAL, __TGMATH_REAL_2)
7666         (__TGMATH_REAL_3, __TGMATH_CPLX_ONLY): Define using
7667         __builtin_tgmath.
7668         (frexp, ldexp, nexttoward, scalbn, scalbln): Define using
7669         __TGMATH_REAL_2.
7670         (remquo): Define using __TGMATH_REAL_3.
7672 2017-11-14  Jeff Law  <law@redhat.com>
7674         * vr-values.c: New file with contents extracted from tree-vrp.c.
7675         * Makefile.in (OBJS): Add vr-values.o
7676         * tree-vrp.h (set_value_range_to_nonnull): Prototype.
7677         (set_value_range, set_and_canonicalize_value_range): Likewise.
7678         (vrp_bitmap_equal_p, range_is_nonnull): Likewise.
7679         (value_range_constant_singleton, symbolic_range_p): Likewise.
7680         (compare_values, compare_values_warnv, vrp_val_is_min): Likewise.
7681         (vrp_val_is_max, copy_value_range, set_value_range_to_value): Likewise.
7682         (extract_range_from_binary_expr_1, vrp_val_min, vrp_val_max): Likewise.
7683         (set_value_range_to_null, range_int_cst_p, opreand_less_p): Likewise.
7684         (find_case_label_range, find_case_label_index): Likewise.
7685         (zero_nonzero_bits_from_vr, overflow_comparison_p): Likewise.
7686         (range_int_cst_singleton_p, value_inside_range): Likewise.
7687         (get_single_symbol): Likewise.
7688         (switch_update): Move structure definition here.
7689         (to_remove_edges, to_update_switch_stmts): Provide externs.
7690         * tree-vrp.c: Move all methods for vr-values class to vr-values.c
7691         (vrp_val_max, vrp_val_min, vrp_val_is_max): Make externally visible.
7692         (vrp_val_is_min, set_value_range): Likewise.
7693         (set_and_canonicalize_value_range, copy_value_range): Likewise.
7694         (set_value_range_to_value, set_value_range_to_nonnull): Likewise.
7695         (set_value_range_to_null, vrp_bitmap_equal_p): Likewise.
7696         (range_is_nonnull, range_int_cst_p): Likewwise.
7697         (range_int_cst_singleton_p, symbolic_range_p): Likewise.
7698         (get_single_symbol, operand_less_p): Likewise
7699         (compare_values_warnv, compare_values): Likewise.
7700         (value_inside_range, value_range_constant_singleton): Likewise.
7701         (zero_nonzero_bitgs_from_vr): Likewise.
7702         (extract_range_from_binary_expr_1): Likewise.
7703         (overflow_comparison_p): Likewise.
7704         (to_remove_edges, to_update_switch_stmts): Likewise.
7705         (find_case_label-index, find_case_label_range): Likewise.
7706         (switch_update, set_value_range_to_nonnegative): Remove.
7707         (set_value_range_to_truthvalue): Likewise.
7708         (symbolic_range_based_on_p, gimple_assign_nonzero_p): Likewise.
7709         (gimple_stmt_nonzero_p, compare_ranges): Likewise.
7710         (compare_range_with_value, vrp_valueize, vrp_valueize_1): Likewise.
7711         (find_case_label_ranges, test_for_singularity): Likewise.
7712         (range_fits_type_p, simplify_conversion_using_ranges): LIkewise.
7713         (x_vr_values): Move to its remaining use site.
7715 2017-11-10  Jeff Law  <law@redhat.com>
7717         * vr-values.h (VR_INITIALIZER): Move #define here.
7718         * gimple-ssa-evrp.c: New file with contents extracted from tree-vrp.c
7719         * Makefile.in (OBJS): Add tree-evrp.o
7720         * tree-vrp.h (assert_info): Move structure definition here.
7721         (set_value_range_to_varying): Prototype.
7722         (vrp_operand_equal_p, range_includes_zero_p): Likewise.
7723         (infer_value_range, register_edge_assert_for): Likewise.
7724         (stmt_interesting_for_vrp): Likewise.
7725         * tree-vrp.c: Move all methods for evrp class into gimple-ssa-evrp.c.
7726         (set_value_range_to_varying): No longer static.
7727         (vrp_operand_equal_p, range_includes_zero_p): Likewise.
7728         (infer_value_range, register_edge_assert_for): Likewise.
7730 2017-11-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
7732         * config/rs6000/rs6000.md (bswapdi2): On 32-bit ISA 3.0, don't
7733         generate the XXBRD instruction.
7735         * config/rs6000/rs6000-c.c (is_float128_p): New helper function.
7736         (rs6000_builtin_type_compatible): Treat _Float128 and long double
7737         as being compatible if -mabi=ieeelongdouble.
7738         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_HW_1): New macros
7739         to setup float128 built-ins with hardware support.
7740         (BU_FLOAT128_HW_2): Likewise.
7741         (BU_FLOAT128_HW_3): Likewise.
7742         (BU_FLOAT128_HW_VSX_1): Likewise.
7743         (BU_FLOAT128_HW_VSX_2): Likewise.
7744         (scalar_extract_expq): Change float128 built-in functions to
7745         accommodate having both KFmode and TFmode functions.  Use the
7746         KFmode variant as the default.
7747         (scalar_extract_sigq): Likewise.
7748         (scalar_test_neg_qp): Likewise.
7749         (scalar_insert_exp_q): Likewise.
7750         (scalar_insert_exp_qp): Likewise.
7751         (scalar_test_data_class_qp): Likewise.
7752         (sqrtf128_round_to_odd): Delete processing the round to odd
7753         built-in functions as special built-in functions, and define them
7754         as float128 built-ins.  Use the KFmode variant as the default.
7755         (truncf128_round_to_odd): Likewise.
7756         (addf128_round_to_odd): Likewise.
7757         (subf128_round_to_odd): Likewise.
7758         (mulf128_round_to_odd): Likewise.
7759         (divf128_round_to_odd): Likewise.
7760         (fmaf128_round_to_odd): Likewise.
7761         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Add
7762         support for KFmode and TFmode xststdcqp calls.
7763         (rs6000_expand_builtin): If long double is IEEE 128-bit floating
7764         point, switch the built-in handlers for the get/set float128
7765         exponent, get float128 mantissa, float128 test built-ins, and the
7766         float128 round to odd built-in functions.  Eliminate creating the
7767         float128 round to odd built-in functions as special built-ins.
7768         (rs6000_init_builtins): Eliminate special creation of the float128
7769         round to odd built-in functions.
7770         * config/rs6000/vsx.md (xsxexpqp_<mode>): Change float128 built-in
7771         function insns to support both TFmode and KFmode variants.
7772         (xsxsigqp_<mode>): Likewise.
7773         (xsiexpqpf_<mode>): Likewise.
7774         (xsiexpqp_<mode>): Likewise.
7775         (xststdcqp_<mode>): Likewise.
7776         (xststdcnegqp_<mode>): Likewise.
7777         (xststdcqp_<mode>): Likewise.
7779 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7781         * tree-ssa-threadupdate.c (compute_path_counts): Remove
7782         unused path_in_freq_ptr parameter.
7783         (ssa_fix_duplicate_block_edges): Do not pass around path_in_freq
7785 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7787         * ipa-inline.c (edge_badness): Dump sreal frequency.
7788         (compute_inlined_call_time): Match natural implementaiton ...
7789         * ipa-fnsummary.c (estimate_edge_size_and_time): ... here; remove
7790         forgotten division by CGRAPH_FREQ_BASE.
7792 2017-11-14  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
7794         * config.gcc (*-*-solaris2*): Enable default_use_cxa_atexit since
7795         Solaris 11.  Update comment.
7796         * configure.ac (gcc_cv_ld_pid): Adapt comment for Solaris 12
7797         renaming.
7798         * config/sol2.h (STARTFILE_SPEC): Likewise.
7799         * configure: Regenerate.
7801 2017-11-14  Carl Love  <cel@us.ibm.com>
7803         * config/rs6000/rs6000.c (swap_endian_selector_for_mode): Remove
7804         le_ and be_ prefixes to swap* variables.  Remove
7805         if (VECTOR_ELT_ORDER_BIG) statement. Remove E_V16QImode case
7806         statements.
7808 2017-11-14  Jason Merrill  <jason@redhat.com>
7810         Support GTY((cache)) on hash_map.
7811         * hash-traits.h (ggc_remove): Add ggc_maybe_mx member function.
7812         (ggc_cache_remove): Override it instead of ggc_mx.
7813         * hash-table.h (gt_ggc_mx): Call it instead of ggc_mx.
7814         (gt_cleare_cache): Call ggc_mx instead of gt_ggc_mx.
7815         * hash-map-traits.h (simple_hashmap_traits): Add maybe_mx member.
7816         (simple_cache_map_traits): Override maybe_mx.
7817         * hash-map.h (hash_entry): Add ggc_maybe_mx and keep_cache_entry.
7818         (hash_map): Friend gt_cleare_cache.
7819         (gt_cleare_cache): New.
7820         * tree.h (tree_cache_traits): New hash_map traits class.
7821         (tree_cache_map): New typedef.
7823 2017-11-14  Richard Biener  <rguenther@suse.de>
7825         * tree-cfgcleanup.c (cleanup_control_expr_graph): Remove first_p
7826         paramter and handling.
7827         (cleanup_control_flow_bb): Likewise.
7828         (cleanup_control_flow_pre): New helper performing a DFS walk
7829         to call cleanup_control_flow_bb in PRE order.
7830         (cleanup_tree_cfg_1): Do the first phase of cleanup_control_flow_bb
7831         via cleanup_control_flow_pre.
7833 2017-11-14  James Greenhalgh  <james.greenhalgh@arm.com>
7835         * config/aarch64/aarch64-simd.md
7836         (aarch64_simd_bsl<mode>_internal): Remove DImode.
7837         (*aarch64_simd_bsl<mode>_alt): Likewise.
7838         (aarch64_simd_bsldi_internal): New.
7839         (aarch64_simd_bsldi_alt): Likewise.
7841 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7843         * tracer.c (better_p): Do not compare frequencies.
7844         * reg-stack.c (better_edge): Likewise.
7845         * shrink-wrap.c (try_shrink_wrapping): Do not convert to gcov counts
7846         and back.
7848 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7850         * auto-profile.c (afdo_annotate_cfg): Use update_max_bb_count.
7851         * cgraphunit.c (cgraph_node::expand_thunk): Use update_max_bb_count.
7852         * ipa-utils.c (ipa_merge_profiles): Use update_max_bb_count.
7853         * lto-streamer-in.c (input_function): Use update_max_bb_count.
7854         * omp-expand.c (expand_omp_taskreg): Use update_max_bb_count.
7855         * predict.c (maybe_hot_frequency_p): Inline to ...
7856         (maybe_hot_count_p): ... here; rewrite to counts.
7857         (counts_to_freqs): Rename to ...
7858         (update_max_bb_count): ... this one.
7859         (expensive_function_p): Use counts.
7860         (estimate_bb_frequencies): Update.
7861         (rebuild_frequencies): Update.
7862         * predict.h (counts_to_freqs): Rename to ...
7863         (update_max_bb_count): ... this one.
7864         * profile.c (compute_branch_probabilities): Add debug info
7865         * tree-inline.c (expand_call_inline): Update debug info.
7866         (optimize_inline_calls): Use update_max_bb_count..
7867         (tree_function_versioning): Use update_max_bb_count..
7868         * value-prof.c (gimple_value_profile_transformations):
7869         Do not use update_max_bb_count.
7871 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7873         * ipa-inline.c (compute_uninlined_call_time, compute_inlined_call_time):
7874         always use frequencies.
7876 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7878         * bb-reorder.c: Remove frequencies from comments.
7879         (better_edge_p): Use profile counts.
7880         (find_traces): Dump profile counts.
7881         (rotate_loop): Use profile counts.
7882         (find_traces_1_round): Likewise.
7883         (connect_better_edge_p): Use counts instead of probabilities for
7884         reverse walk.
7885         (copy_bb_p): Drop early check for non-0 frequency.
7886         (sanitize_hot_paths): Update comments.
7888 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7890         * ipa-split.c (struct split_point): Add count.
7891         (consider_split): Do not compute incoming frequency; compute incoming
7892         count and store it to split_point.
7893         (split_function): Set count of the call to split part correctly.
7895 2017-11-13  Carl Love  <cel@us.ibm.com>
7897         * config/rs6000/altivec.md (altivec_vsumsws_be): Add define_expand.
7899 2017-11-13  Tom Tromey  <tom@tromey.com>
7901         * doc/cpp.texi (Variadic Macros): Document __VA_OPT__.
7903 2017-11-13  Carl Love  <cel@us.ibm.com>
7905         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
7906         Add support for builtins:
7907         unsigned int vec_first_{,miss}_match_{,or_eos}index,
7908         vector {un,}signed {char,int,short},
7909         vector {un,}signed {char,int,short}) arguments.
7910         * config/rs6000/rs6000-builtin.def (VFIRSTMATCHINDEX,
7911         VFIRSTMATCHOREOSINDEX, VFIRSTMISMATCHINDEX, VFIRSTMISMATCHOREOSINDEX):
7912         Add BU_P9V_AV_2 expansions for the builtins.
7913         * config/rs6000/altivec.h (vec_first_match_index,
7914         vec_first_mismatch_index, vec_first_match_or_eos_index,
7915         vec_first_mismatch_or_eos_index): Add #defines for the builtins.
7916         * config/rs6000/rs6000-protos.h (bytes_in_mode): Add
7917         new extern declaration.
7918         * config/rs6000/rs6000.c (bytes_in_mode): Add new function.
7919         * config/rs6000/vsx.md (first_match_index_<mode>,
7920         first_match_or_eos_index_<mode>, first_mismatch_index_<mode>,
7921         first_mismatch_or_eos_index_<mode>): Add define expand.
7922         (vctzlsbb_<mode>): Add mode field to define_insn for vctzlsbb.
7923         * doc/extend.texi: Update the built-in documenation file for the new
7924         built-in functions.
7926 2017-11-13  Michael Meissner  <meissner@linux.vnet.ibm.com>
7928         * match.pd: Convert fminf<N>, fminf<N>x, fmax<N>, and fmax<N>x
7929         into the min/max operations for _Float<N> and _Float<N>X types.
7931 2017-11-13  Eric Botcazou  <ebotcazou@adacore.com>
7933         PR lto/81351
7934         * dwarf2out.c (do_eh_frame): New static variable.
7935         (dwarf2out_begin_prologue): Set it.
7936         (dwarf2out_frame_finish): Test it instead of dwarf2out_do_eh_frame.
7938 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7940         * tree-ssa-coalesce.c (coalesce_cost): Fix formating.
7942         * tree-ssa-sink.c (select_best_block): Do not use frequencies.
7944 2017-11-13  Eric Botcazou  <ebotcazou@adacore.com>
7946         PR lto/81351
7947         * debug.h (dwarf2out_do_eh_frame): Declare.
7948         * dwarf2cfi.c (dwarf2out_do_eh_frame): New predicate.
7949         (dwarf2out_do_frame): Use it.
7950         (dwarf2out_do_cfi_asm): Likewise.
7951         * dwarf2out.c (dwarf2out_frame_finish): Likewise.
7952         (dwarf2out_assembly_start): Likewise.
7953         (dwarf2out_begin_prologue): Fix comment.
7954         * toplev.c (compile_file): Always call dwarf2out_frame_finish
7955         if the target needs either debug or unwind DWARF2 info.
7956         * lto-opts.c (lto_write_options): Do not save -fexceptions,
7957         -fnon-call-exceptions, -ffp-contract, -fmath-errno, -fsigned-zeros,
7958         -ftrapping-math, -ftrapv and -fwrapv.
7960 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7962         * cgraph.c (cgraph_edge::sreal_frequency): New function.
7963         * cgraph.h (cgraph_edge::sreal_frequency): Declare.
7964         * ipa-fnsummary.c (dump_ipa_call_summary): Use sreal_frequency.
7965         (estimate_edge_size_and_time): Likewise.
7966         (ipa_merge_fn_summary_after_inlining): Likewise.
7967         * ipa-inline.c (cgraph_freq_base_rec): Remove.
7968         (compute_uninlined_call_time): Use sreal_frequency.
7969         (compute_inlined_call_time): Likewise.
7970         (ipa_inline): Do not initialize cgraph_freq_base_rec.
7971         * profile-count.c: Include sreal.h.
7972         (profile_count::to_sreal_scale): New.
7973         * profile-count.h: Forward declare sreal.
7974         (profile_count::to_sreal_scale): Declare.
7976 2017-11-13  Nathan Sidwell  <nathan@acm.org>
7978         * diagnostic.c (maybe_line_and_column): New.
7979         (diagnostic_get_location_text): Use it.
7980         (diagnostic_report_current_module): Likewise.
7981         (test_diagnostic_get_location_text): Add tests.
7983 2017-11-13  Luis Machado  <luis.machado@linaro.org>
7985         * doc/md.texi (Specifying processor pipeline description): Fix
7986         incorrect latency for the div instruction example.
7988 2017-11-13  Jakub Jelinek  <jakub@redhat.com>
7990         PR tree-optimization/78821
7991         * gimple-ssa-store-merging.c (compatible_load_p): Don't require
7992         that bit_not_p is the same.
7993         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
7994         (split_group): Count precisely bit_not_p bits in each statement.
7995         (invert_op): New function.
7996         (imm_store_chain_info::output_merged_store): Use invert_op to
7997         emit BIT_XOR_EXPR with a xor_mask instead of BIT_NOT_EXPR if some
7998         but not all orig_stores have BIT_NOT_EXPR in the corresponding spots.
8000 2017-11-13  Martin Liska  <mliska@suse.cz>
8002         * gcov.c (struct coverage_info): Remove typedef of coverage_t.
8003         (struct source_info): Likewise.
8004         (add_branch_counts): Likewise.
8005         (add_line_counts): Likewise.
8006         (function_summary): Likewise.
8007         (output_intermediate_line): Likewise.
8008         (generate_results): Likewise.
8010 2017-11-13  Martin Liska  <mliska@suse.cz>
8012         * gcov.c (struct block_info): Remove typedef for block_t.
8013         (struct line_info): Likewise.
8014         (line_info::has_block): Likewise.
8015         (EXIT_BLOCK): Likewise.
8016         (unblock): Likewise.
8017         (circuit): Likewise.
8018         (get_cycles_count): Likewise.
8019         (process_file): Likewise.
8020         (read_graph_file): Likewise.
8021         (solve_flow_graph): Likewise.
8022         (find_exception_blocks): Likewise.
8023         (add_line_counts): Likewise.
8024         (accumulate_line_info): Likewise.
8025         (output_line_details): Likewise.
8027 2017-11-13  Martin Liska  <mliska@suse.cz>
8029         * gcov.c (struct arc_info): Remove typedef for arc_t.
8030         (struct line_info): Likewise.
8031         (add_branch_counts): Likewise.
8032         (output_branch_count): Likewise.
8033         (function_info::~function_info): Likewise.
8034         (circuit): Likewise.
8035         (output_intermediate_line): Likewise.
8036         (read_graph_file): Likewise.
8037         (solve_flow_graph): Likewise.
8038         (find_exception_blocks): Likewise.
8039         (add_line_counts): Likewise.
8040         (accumulate_line_info): Likewise.
8041         (output_line_details): Likewise.
8042         (output_function_details): Likewise.
8044 2017-11-13  Martin Liska  <mliska@suse.cz>
8046         * gcov.c (struct function_info): Remove typedef for function_t.
8047         (struct source_info): Likewise.
8048         (source_info::get_functions_at_location): Likewise.
8049         (solve_flow_graph): Likewise.
8050         (find_exception_blocks): Likewise.
8051         (add_line_counts): Likewise.
8052         (output_intermediate_file): Likewise.
8053         (process_file): Likewise.
8054         (generate_results): Likewise.
8055         (release_structures): Likewise.
8056         (read_graph_file): Likewise.
8057         (read_count_file): Likewise.
8058         (accumulate_line_counts): Likewise.
8059         (output_lines): Likewise.
8061 2017-11-13  Martin Liska  <mliska@suse.cz>
8063         * gcov.c (function_info::function_info): Remove num_counts
8064         and add vector<gcov_type>.
8065         (function_info::~function_info): Use the vector.
8066         (process_file): Likewise.
8067         (read_graph_file): Likewise.
8068         (read_count_file): Likewise.
8069         (solve_flow_graph): Likewise.
8071 2017-11-13  Martin Liska  <mliska@suse.cz>
8073         * gcov.c (function_info::is_artificial): New function.
8074         (process_file): Erase all artificial early.
8075         (generate_results): Skip as all artificial are already
8076         removed.
8078 2017-11-13  Martin Liska  <mliska@suse.cz>
8080         * gcov.c (read_graph_file): Store to global vector of functions.
8081         (read_count_file): Iterate the vector.
8082         (process_file): Likewise.
8083         (generate_results): Likewise.
8084         (release_structures): Likewise.
8086 2017-11-13  Jakub Jelinek  <jakub@redhat.com>
8088         PR tree-optimization/82954
8089         * gimple-ssa-store-merging.c
8090         (imm_store_chain_info::coalesce_immediate_stores): If
8091         !infof->ops[N].base_addr, split group if info->ops[N].base_addr.
8093 2017-11-13  Richard Sandiford  <richard.sandiford@linaro.org>
8095         * config/aarch64/aarch64-simd.md (aarch64_store_lane0<mode>):
8096         Upddate call to ENDIAN_LANE_N.
8097         (aarch64_<sur>dot_lane<vsi2qi>): Use aarch64_endian_lane_rtx.
8098         (aarch64_<sur>dot_laneq<vsi2qi>): Likewise.
8099         (*aarch64_simd_vec_copy_lane<mode>): Update calls to ENDIAN_LANE_N
8100         and use aarch64_endian_lane_rtx.
8101         (*aarch64_simd_vec_copy_lane_<vswap_width_name><mode>): Likewise.
8103 2017-11-12  Tom de Vries  <tom@codesourcery.com>
8105         * config/riscv/riscv.h (ASM_OUTPUT_LABELREF): Wrap in do {} while (0).
8107 2017-11-12  Tom de Vries  <tom@codesourcery.com>
8109         * config/elfos.h (ASM_OUTPUT_ASCII): Remove semicolon after macro body.
8111 2017-11-12  Tom de Vries  <tom@codesourcery.com>
8113         * config/cr16/cr16.h (ASM_OUTPUT_LABELREF): Remove semicolon after macro
8114         body.
8115         * config/powerpcspe/xcoff.h (ASM_OUTPUT_LABELREF): Same.
8116         * config/rs6000/xcoff.h (ASM_OUTPUT_LABELREF): Same.
8117         * defaults.h (ASM_OUTPUT_LABELREF): Same.
8119 2017-11-11  Martin Sebor  <msebor@redhat.com>
8121         PR c/81117
8122         * doc/extend.texi (attribute nonstring): Remove spurious argument.
8124         PR bootstrap/82948
8125         * prefic.c (translate_name): Replace strncpy with memcpy to
8126         avoid -Wstringop-truncation.
8128 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
8130         * tree-ssa-loop-im.c (execute_sm_if_changed): Do not compute freq_sum.
8132 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
8134         * predict.c (maybe_hot_frequency_p): Do not use cfun.
8136 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
8138         * tree-ssa-tail-merge.c (replace_block_by): Fix and re-enable profile
8139         merging.
8141 2017-11-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
8143         * config/rs6000/rs6000.md (bswaphi2_reg): On ISA 3.0 systems,
8144         enable generating XXBRH if the value is in a vector register.
8145         (bswapsi2_reg): On ISA 3.0 systems, enable generating XXBRW if the
8146         value is in a vector register.
8147         (bswapdi2_reg): On ISA 3.0 systems, always use XXBRD to do
8148         register to register bswap64's instead of doing the GPR sequence
8149         used on previous machines.
8150         (bswapdi2_xxbrd): New insn.
8151         (bswapdi2_reg): Disallow on ISA 3.0.
8152         (register to register bswap64 splitter): Do not split the insn on
8153         ISA 3.0 systems that use XXBRD.
8155 2017-11-10  Martin Sebor  <msebor@redhat.com>
8157         PR c/81117
8158         * config/darwin-c.c (framework_construct_pathname): Replace strncpy
8159         with memcpy.
8160         (find_subframework_file): Same.
8162 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
8164         * auto-profile.c (afdo_indirect_call): Drop frequency.
8165         * cgraph.c (symbol_table::create_edge): Drop frequency argument.
8166         (cgraph_node::create_edge): Drop frequency argument.
8167         (cgraph_node::create_indirect_edge): Drop frequency argument.
8168         (cgraph_edge::make_speculative): Drop frequency arguments.
8169         (cgraph_edge::resolve_speculation): Do not update frequencies
8170         (cgraph_edge::dump_edge_flags): Do not dump frequency.
8171         (cgraph_node::dump): Check consistency in IPA mode.
8172         (cgraph_edge::maybe_hot_p): Use IPA counter.
8173         (cgraph_edge::verify_count_and_frequency): Rename to ...
8174         (cgraph_edge::verify_count): ... this one; drop frequency checking.
8175         (cgraph_node::verify_node): Update.
8176         * cgraph.h (struct cgraph_edge): Drop frequency.
8177         (cgraph_edge::frequency): New function.
8178         * cgraphbuild.c (pass_build_cgraph_edges::execute): Donot pass
8179         frequencies.
8180         (cgraph_edge::rebuild_edges): Likewise.
8181         * cgraphclones.c (cgraph_edge::clone): Scale only counts.
8182         (duplicate_thunk_for_node): Do not pass frequency.
8183         (cgraph_node::create_clone): Scale only counts.
8184         (cgraph_node::create_virtual_clone): Do not pass frequency.
8185         (cgraph_node::create_edge_including_clones): Do not pass frequency.
8186         (cgraph_node::create_version_clone): Do not pass frequency.
8187         * cgraphunit.c (cgraph_node::analyze): Do not pass frequency.
8188         (cgraph_node::expand_thunk): Do not pass frequency.
8189         (cgraph_node::create_wrapper): Do not pass frequency.
8190         * gimple-iterator.c (update_call_edge_frequencies): Do not pass
8191         frequency.
8192         * gimple-streamer-in.c (input_bb): Scale only IPA counts.
8193         * ipa-chkp.c (chkp_produce_thunks): Do not pass frequency.
8194         * ipa-cp.c (ipcp_lattice::print): Use frequency function.
8195         (gather_caller_stats): Use frequency function.
8196         (ipcp_cloning_candidate_p): Use frequency function.
8197         (ipcp_propagate_stage): Use frequency function.
8198         (get_info_about_necessary_edges): Use frequency function.
8199         (update_profiling_info): Update only IPA profile.
8200         (update_specialized_profile): Use frequency functoin.
8201         (perhaps_add_new_callers): Update only IPA profile.
8202         * ipa-devirt.c (ipa_devirt): Use IPA profile.
8203         * ipa-fnsummary.c (redirect_to_unreachable): Do not set frequrency.
8204         (dump_ipa_call_summary): Use frequency function.
8205         (estimate_edge_size_and_time): Use frequency function.
8206         (ipa_merge_fn_summary_after_inlining): Use frequency function.
8207         * ipa-inline-analysis.c (do_estimate_edge_time): Use IPA profile.
8208         * ipa-inline-transform.c (update_noncloned_frequencies): Rename to ..
8209         (update_noncloned_counts): ... ths one; scale counts only.
8210         (clone_inlined_nodes): Do not scale frequency.
8211         (inline_call): Do not pass frequency.
8212         * ipa-inline.c (compute_uninlined_call_time): Use IPA profile.
8213         (compute_inlined_call_time): Use IPA profile.
8214         (want_inline_small_function_p): Use IPA profile.
8215         (want_inline_self_recursive_call_p): Use IPA profile.
8216         (edge_badness): Use IPA profile.
8217         (lookup_recursive_calls): Use IPA profile.
8218         (recursive_inlining): Do not pass frequency.
8219         (resolve_noninline_speculation): Do not update frequency.
8220         (inline_small_functions): Collect max of IPA profile.
8221         (dump_overall_stats): Dump IPA porfile.
8222         (dump_inline_stats): Dump IPA porfile.
8223         (ipa_inline): Collect IPA stats.
8224         * ipa-inline.h (clone_inlined_nodes): Update prototype.
8225         * ipa-profile.c (ipa_propagate_frequency_1): Use frequency function.
8226         (ipa_propagate_frequency): Use frequency function.
8227         (ipa_profile): Cleanup.
8228         * ipa-prop.c (ipa_make_edge_direct_to_target): Do not pass frequency
8229         * ipa-utils.c (ipa_merge_profiles): Merge all profiles.
8230         * lto-cgraph.c (lto_output_edge): Do not stream frequency.
8231         (input_node): Do not stream frequency.
8232         (input_edge): Do not stream frequency.
8233         (merge_profile_summaries): Scale only IPA profiles.
8234         * omp-simd-clone.c (simd_clone_adjust): Do not pass frequency.
8235         * predict.c (drop_profile): Do not recompute frequency.
8236         * trans-mem.c (ipa_tm_insert_irr_call): Do not pass frequency.
8237         (ipa_tm_insert_gettmclone_call): Do not pass frequency.
8238         * tree-cfg.c (execute_fixup_cfg): Drop profile to global0 if needed.
8239         * tree-chkp.c (chkp_copy_bounds_for_assign): Do not pass frequency.
8240         * tree-emutls.c (gen_emutls_addr): Do not pass frequency.
8241         * tree-inline.c (copy_bb): Do not scale frequency.
8242         (expand_call_inline): Do not scale frequency.
8243         (tree_function_versioning): Do not scale frequency.
8244         * ubsan.c (ubsan_create_edge): Do not pass frequency.
8246 2017-11-10  Julia Koval  <julia.koval@intel.com>
8248         * config/i386/gfniintrin.h (_mm_gf2p8affine_epi64_epi8)
8249         (_mm256_gf2p8affine_epi64_epi8, _mm_mask_gf2p8affine_epi64_epi8)
8250         (_mm_maskz_gf2p8affine_epi64_epi8, _mm256_mask_gf2p8affine_epi64_epi8)
8251         (_mm256_maskz_gf2p8affine_epi64_epi8)
8252         (_mm512_mask_gf2p8affine_epi64_epi8, _mm512_gf2p8affine_epi64_epi8)
8253         (_mm512_maskz_gf2p8affine_epi64_epi8): New intrinsics.
8254         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8affineqb_v64qi)
8255         (__builtin_ia32_vgf2p8affineqb_v32qi)
8256         (__builtin_ia32_vgf2p8affineqb_v16qi): New builtins.
8257         * config/i386/sse.md (vgf2p8affineqb_<mode><mask_name>): New pattern.
8259 2017-11-10  Tamar Christina  <tamar.christina@arm.com>
8261         PR target/82641
8262         * config/arm/arm.c
8263         (arm_option_override): Refactor.
8264         (arm_option_reconfigure_globals): New.
8265         (arm_options_perform_arch_sanity_checks): New.
8266         * config/arm/arm-protos.h (arm_option_reconfigure_globals):
8267         New prototype.
8268         (arm_options_perform_arch_sanity_checks): Likewise
8270 2017-11-10  Pat Haugen  <pthaugen@us.ibm.com>
8272         * rs6000/power9.md (power9-qpdiv): Correct DFU pipe usage.
8273         (power9-qpmul): New.
8274         * rs6000/rs6000.md ("type" attr): Add qmul.
8275         (mul<mode>3, fma<mode>4_hw, *fms<mode>4_hw, *nfma<mode>4_hw,
8276         *nfms<mode>4_hw, mul<mode>3_odd, fma<mode>4_odd, *fms<mode>4_odd,
8277         *nfma<mode>4_odd, *nfms<mode>4_odd): Change type to qmul.
8279 2017-11-10  Martin Sebor  <msebor@redhat.com>
8281         PR c/81117
8282         * builtins.c (compute_objsize): Handle arrays that
8283         compute_builtin_object_size likes to fail for.  Make extern.
8284         * builtins.h (compute_objsize): Declare.
8285         (check_strncpy_sizes): New function.
8286         (expand_builtin_strncpy): Call check_strncpy_sizes.
8287         * gimple-fold.c (gimple_fold_builtin_strncpy): Implement
8288         -Wstringop-truncation.
8289         (gimple_fold_builtin_strncat): Same.
8290         * gimple.c (gimple_build_call_from_tree): Set call location.
8291         * tree-ssa-strlen.c (strlen_to_stridx): New global variable.
8292         (maybe_diag_bound_equal_length, is_strlen_related_p): New functions.
8293         (handle_builtin_stxncpy, handle_builtin_strncat): Same.
8294         (handle_builtin_strlen): Use strlen_to_stridx.
8295         (strlen_optimize_stmt): Handle flavors of strncat, strncpy, and
8296         stpncpy.
8297         Use strlen_to_stridx.
8298         (pass_strlen::execute): Release strlen_to_stridx.
8299         * doc/invoke.texi (-Wsizeof-pointer-memaccess): Document enhancement.
8300         (-Wstringop-truncation): Document new option.
8302 2017-11-10  Martin Liska  <mliska@suse.cz>
8304         PR gcov-profile/82702
8305         * gcov.c (main): Handle intermediate files in a different
8306         way.
8307         (get_gcov_intermediate_filename): New function.
8308         (output_gcov_file): Remove support of intermediate files.
8309         (generate_results): Allocate intermediate file.
8310         (release_structures): Clean-up properly fn_end.
8311         (output_intermediate_file): Start iterating with line 1.
8313 2017-11-10  Jakub Jelinek  <jakub@redhat.com>
8315         PR tree-optimization/82929
8316         * gimple-ssa-store-merging.c (struct store_immediate_info): Add
8317         ops_swapped_p non-static data member.
8318         (store_immediate_info::store_immediate_info): Clear it.
8319         (imm_store_chain_info::coalesce_immediate_stores): If swapping
8320         ops set ops_swapped_p.
8321         (count_multiple_uses): Handle ops_swapped_p.
8323 2017-11-10  Martin Liska  <mliska@suse.cz>
8325         * coverage.c (coverage_init): Stream information about
8326         support of has_unexecuted_blocks.
8327         * doc/gcov.texi: Document that.
8328         * gcov-dump.c (dump_gcov_file): Support it in gcov_dump tool.
8329         * gcov.c (read_graph_file): Likewise.
8330         (output_line_beginning): Fix a small issue with
8331         color output.
8333 2017-11-10  Bin Cheng  <bin.cheng@arm.com>
8335         * tree-predcom.c (determine_roots_comp): Avoid memory leak by freeing
8336         reference of trivial component.
8338 2017-11-10  Jakub Jelinek  <jakub@redhat.com>
8340         PR bootstrap/82916
8341         * gimple-ssa-store-merging.c
8342         (pass_store_merging::terminate_all_aliasing_chains): For
8343         gimple_store_p stmts also call refs_output_dependent_p.
8345         PR rtl-optimization/82913
8346         * compare-elim.c (try_merge_compare): Punt if def_insn is not
8347         single set.
8349 2017-11-09  Jeff Law  <law@redhat.com>
8351         * vr-values.h: New file with vr_values class.
8352         * tree-vrp.c: Include vr-values.h
8353         (vrp_value_range_pool, vrp_equiv_obstack, num_vr_values): Move static
8354         data objects into the vr_values class.
8355         (vr_value, values_propagated, vr_phi_edge_counts): Likewise.
8356         (get_value_range): Make it a member function within vr_values class.
8357         (set_defs_to_varying, update_value_range, add_equivalence): Likewise.
8358         (vrp_stmt_computes_nonzero_p, op_with_boolean_value_range_p): Likewise.
8359         (op_with_constant_singleton_value_range): Likewise.
8360         (extract_range_for_var_from_comparison_expr): Likewise.
8361         (extract_range_from_assert, extract_range_from_ssa_name): Likewise.
8362         (extract_range_from_binary_expr): Likewise.
8363         (extract_range_from_unary_expr): Likewise.
8364         (extract_range_from_cond_expr, extrat_range_from_comparison): Likewise.
8365         (check_for_binary_op_overflow, extract_range_basic): Likewise.
8366         (extract_range_from_assignment, adjust_range_with_scev): Likewise.
8367         (dump_all_value_ranges, get_vr_for_comparison): Likewise.
8368         (compare_name_with_value, compare_names): Likewise.
8369         (vrp_evaluate_conditional_warnv_with_ops_using_ranges): Likewise.
8370         (vrp_evaluate_conditional_warnv_with_ops): Likewise.  Remove prototype.
8371         (vrp_evaluate_conditional, vrp_visit_cond_stmt): Likewise.
8372         (vrp_visit_switch_stmt, extract_range_from_stmt): Likewise.
8373         (extract_range_from_phi_node): Likewise.
8374         (simplify_truth_ops_using_ranges): Likewise.
8375         (simplify_div_or_mod_using_ranges): Likewise.
8376         (simplify_min_or_max_using_ranges, simplify_abs_using_ranges): Likewise.
8377         (simplify_bit_ops_using_ranges, simplify_cond_using_ranges_1): Likewise.
8378         (simplify_cond_using_ranges_2, simplify_switch_using_ranges): Likewise.
8379         (simplify_float_conversion_using_ranges): Likewise.
8380         (simplify_internal_call_using_ranges): Likewise.
8381         (two_valued_val_range_p, simplify_stmt_using_ranges): Likewise.
8382         (vrp_visit_assignment_or_call): Likewise.  Smuggle class instance
8383         poitner via x_vr_values for calls into gimple folder.
8384         (vrp_initialize_lattice): Make this the vr_values ctor.
8385         (vrp_free_lattice): Make this the vr_values dtor.
8386         (set_vr_value): New function.
8387         (class vrp_prop): Add vr_values data member.  Add various member
8388         functions as well as member functions that delegate to vr_values.
8389         (check_array_ref): Make a member function within vrp_prop class.
8390         (search_for_addr_array, vrp_initialize): Likewise.
8391         (vrp_finalize): Likewise.  Revamp to avoid direct access to
8392         vr_value, values_propagated, etc.
8393         (check_array_bounds): Extract vrp_prop class instance pointer from
8394         walk info structure.  Use it to call member functions.
8395         (check_all_array_refs): Make a member function within vrp_prop class.
8396         Smuggle class instance pointer via walk info structure.
8397         (x_vr_values): New local static.
8398         (vrp_valueize): Use x_vr_values to get class instance.
8399         (vr_valueize_1): Likewise.
8400         (class vrp_folder): Add vr_values data member.  Add various member
8401         functions as well as member functions that delegate to vr_values.
8402         (fold_predicate_in): Make a mber fucntion within vrp_folder class.
8403         (simplify_stmt_for_jump_threading): Extract smuggled vr_values
8404         class instance from vr_values.  Use it to call member functions.
8405         (vrp_dom_walker): Add vr_values data member.
8406         (vrp_dom_walker::after_dom_children): Smuggle vr_values class
8407         instance via x_vr_values.
8408         (identify_jump_threads): Accept vr_values as argument.  Store
8409         it into the walker structure.
8410         (evrp_dom_walker): Add vr_values class data member.  Add various
8411         delegators.
8412         (evrp_dom_walker::try_find_new_range): Use vr_values data
8413         member to access the memory allocator.
8414         (evrp_dom_walker::before_dom_children): Store vr_values class
8415         instance into the vrp_folder class.
8416         (evrp_dom_walker::push_value_range): Rework to avoid direct
8417         access to num_vr_values and vr_value.
8418         (evrp_dom_walker::pop_value_range): Likewise.
8419         (execute_early_vrp): Remove call to vrp_initialize_lattice.
8420         Use vr_values to get to dump_all_value_ranges member function.
8421         Remove call to vrp_free_lattice.  Call vrp_initialize, vrp_finalize,
8422         and simplify_cond_using_ranges_2 via vrp_prop class instance.
8423         Pass vr_values class instance down to identify_jump_threads.
8424         Remove call to vrp_free_lattice.
8425         (debug_all_value_ranges): Remove.
8427         * tree-vrp.c (vrp_prop): Move class to earlier point in the file.
8428         (vrp_folder): Likewise.
8430         * tree-vrp.c (set_value_range): Do not reference vrp_equiv_obstack.
8431         Get it from the existing bitmap instead.
8432         (vrp_intersect_ranges_1): Likewise.
8434 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8436         * gimple-ssa-store-merging.c (struct store_immediate_info): Add
8437         bit_not_p field.
8438         (store_immediate_info::store_immediate_info): Add bitnotp argument,
8439         set bit_not_p to it.
8440         (imm_store_chain_info::coalesce_immediate_stores): Break group
8441         if bit_not_p is different.
8442         (count_multiple_uses, split_group,
8443         imm_store_chain_info::output_merged_store): Handle info->bit_not_p.
8444         (handled_load): Avoid multiple chained BIT_NOT_EXPRs.
8445         (pass_store_merging::process_store): Handle BIT_{AND,IOR,XOR}_EXPR
8446         result inverted using BIT_NOT_EXPR, compute bit_not_p, pass it
8447         to store_immediate_info ctor.
8449 2017-11-09  Jim Wilson  <jimw@sifive.com>
8451         * collect2.c (OBJECT_FORMAT_COFF): Remove EXTENDED_COFF support.
8452         (scan_prog_file): Likewise.
8454 2017-11-09  Jan Hubicka  <hubicka@ucw.cz>
8456         * bb-reorder.c (max_entry_frequency): Remove.
8457         (find_traces, rotate_loop, mark_bb_visited, connect_better_edge_p,
8458         connect_traces, push_to_next_round_p): Remove prototypes.
8459         (find_traces_1_round): Use counts only.
8460         (push_to_next_round_p): Likewise.
8461         (find_traces): Likewise.
8462         (rotate_loop): Likewise.
8463         (find_traces_1_round): Likewise.
8464         (connect_traces): Likewise.
8465         (edge_order): Likewise.
8467 2017-11-09  Thomas Preud'homme  <thomas.preudhomme@arm.com>
8469         * config/arm/arm.c (output_return_instruction): Add comments to
8470         indicate requirement for cmse_nonsecure_entry return to account
8471         for the size of clearing instruction output here.
8472         (thumb_exit): Likewise.
8473         * config/arm/thumb2.md (thumb2_cmse_entry_return): Fix length for
8474         return in hardfloat mode.
8476 2017-11-09  Segher Boessenkool  <segher@kernel.crashing.org>
8478         * config/rs6000/rs6000.c (machine_function): Add a bool,
8479         "toc_is_wrapped_separately".
8480         (rs6000_option_override_internal): Enable OPTION_MASK_SAVE_TOC_INDIRECT
8481         if it wasn't explicitly set or unset, we are optimizing for speed, and
8482         doing separate shrink-wrapping.
8483         (rs6000_get_separate_components): Enable the TOC component if
8484         saving the TOC register in the prologue.
8485         (rs6000_components_for_bb): Handle the TOC component.
8486         (rs6000_emit_prologue_components): Store the TOC register where needed.
8487         (rs6000_set_handled_components): Mark TOC as handled, if handled.
8488         (rs6000_emit_prologue): Don't save the TOC if that is already done.
8490 2017-11-09  Martin Jambor  <mjambor@suse.cz>
8492         * ipa-param-manipulation.c: New file.
8493         * ipa-param-manipulation.h: Likewise.
8494         * Makefile.in (OBJS): Add ipa-param-manipulation.o.
8495         (PLUGIN_HEADERS): Addded ipa-param-manipulation.h
8496         * ipa-param.h (ipa_parm_op): Moved to ipa-param-manipulation.h.
8497         (ipa_parm_adjustment): Likewise.
8498         (ipa_parm_adjustment_vec): Likewise.
8499         (ipa_get_vector_of_formal_parms): Moved declaration to
8500         ipa-param-manipulation.h.
8501         (ipa_get_vector_of_formal_parm_types): Likewise.
8502         (ipa_modify_formal_parameters): Likewise.
8503         (ipa_modify_call_arguments): Likewise.
8504         (ipa_combine_adjustments): Likewise.
8505         (ipa_dump_param_adjustments): Likewise.
8506         (ipa_modify_expr): Likewise.
8507         (ipa_get_adjustment_candidate): Likewise.
8508         * ipa-prop.c (ipa_get_vector_of_formal_parms): Moved to
8509         ipa-param-manipulation.c.
8510         (ipa_get_vector_of_formal_parm_types): Likewise.
8511         (ipa_modify_formal_parameters): Likewise.
8512         (ipa_modify_call_arguments): Likewise.
8513         (ipa_modify_expr): Likewise.
8514         (get_ssa_base_param): Likewise.
8515         (ipa_get_adjustment_candidate): Likewise.
8516         (index_in_adjustments_multiple_times_p): Likewise.
8517         (ipa_combine_adjustments): Likewise.
8518         (ipa_dump_param_adjustments): Likewise.
8519         * tree-sra.c: Also include ipa-param-manipulation.h
8520         * omp-simd-clone.c: Include ipa-param-manipulation.h instead of
8521         ipa-param.h.
8523 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8524             Alan Hayward  <alan.hayward@arm.com>
8525             David Sherwood  <david.sherwood@arm.com>
8527         * doc/sourcebuild.texi (vect_masked_store): Document.
8529 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8530             Alan Hayward  <alan.hayward@arm.com>
8531             David Sherwood  <david.sherwood@arm.com>
8533         * doc/sourcebuild.texi (vect_align_stack_vars): Document.
8535 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8536             Alan Hayward  <alan.hayward@arm.com>
8537             David Sherwood  <david.sherwood@arm.com>
8539         * doc/sourcebuild.texi (vect_variable_length): Document.
8541 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8542             Alan Hayward  <alan.hayward@arm.com>
8543             David Sherwood  <david.sherwood@arm.com>
8545         * doc/sourcebuild.texi (vect_unaligned_possible): Document.
8547 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8548             Alan Hayward  <alan.hayward@arm.com>
8549             David Sherwood  <david.sherwood@arm.com>
8551         * doc/sourcebuild.texi (vect_element_align_preferred): Document.
8553 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8554             Alan Hayward  <alan.hayward@arm.com>
8555             David Sherwood  <david.sherwood@arm.com>
8557         * doc/sourcebuild.texi (vect_perm_short, vect_perm_byte): Document
8558         previously undocumented selectors.
8559         (vect_perm3_byte, vect_perm3_short, vect_perm3_int): Document.
8561 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8563         * doc/rtl.texi (const_vector): Say that elements can be
8564         const_wide_ints too.
8565         * emit-rtl.h (valid_for_const_vec_duplicate_p): Declare.
8566         * emit-rtl.c (valid_for_const_vec_duplicate_p): New function.
8567         (gen_vec_duplicate): Use it instead of CONSTANT_P.
8568         * optabs.c (expand_vector_broadcast): Likewise.
8570 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8571             Alan Hayward  <alan.hayward@arm.com>
8572             David Sherwood  <david.sherwood@arm.com>
8574         * tree-ssa-loop-ivopts.c (get_address_cost): Try using a
8575         scaled index even if the unscaled address was invalid.
8576         Don't increase the complexity of using a scale in that case.
8578 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8579             Alan Hayward  <alan.hayward@arm.com>
8580             David Sherwood  <david.sherwood@arm.com>
8582         * doc/rtl.texi: Rewrite the subreg rules so that they partition
8583         the inner register into REGMODE_NATURAL_SIZE bytes rather than
8584         UNITS_PER_WORD bytes.
8585         * emit-rtl.c (validate_subreg): Divide subregs into blocks
8586         based on REGMODE_NATURAL_SIZE of the inner mode.
8587         (gen_lowpart_common): Split the SCALAR_FLOAT_MODE_P and
8588         !SCALAR_FLOAT_MODE_P cases.  Use REGMODE_NATURAL_SIZE for the latter.
8589         * expmed.c (lowpart_bit_field_p): Divide the value up into
8590         chunks of REGMODE_NATURAL_SIZE rather than UNITS_PER_WORD.
8591         * expr.c (store_constructor): Use REGMODE_NATURAL_SIZE to test
8592         whether something is likely to occupy more than one register.
8594 2017-11-09  Jan Hubicka  <hubicka@ucw.cz>
8596         PR ipa/82879
8597         * ipa-inline-transform.c (update_noncloned_frequencies): Use
8598         profile_count::adjust_for_ipa_scaling.
8599         * tree-inline.c (copy_bb, copy_cfg_body): Likewise.
8600         * profile-count.c (profile_count::adjust_for_ipa_scaling): New member
8601         function.
8602         * profile-count.h (profile_count::adjust_for_ipa_scaling): Declare.
8604 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8606         * gimple-ssa-store-merging.c (count_multiple_uses): New function.
8607         (split_group): Add total_orig and total_new arguments, estimate the
8608         number of statements related to the store group without store merging
8609         and with store merging.
8610         (imm_store_chain_info::output_merged_store): Adjust split_group
8611         callers, punt if estimated number of statements with store merging
8612         is not smaller than estimated number of statements without it.
8613         Formatting fix.
8614         (handled_load): Remove has_single_use checks.
8615         (pass_store_merging::process_store): Likewise.
8617 2017-11-09  Richard Biener  <rguenther@suse.de>
8619         PR tree-optimization/82902
8620         * tree-ssa-phiprop.c (propagate_with_phi): Test proper type.
8622 2017-11-09  Martin Liska  <mliska@suse.cz>
8624         PR target/82863
8625         * emit-rtl.c (init_emit_regs): Initialize split_branch_probability to
8626         uninitialized.
8628 2017-11-09  Martin Liska  <mliska@suse.cz>
8630         PR tree-optimization/82669
8631         * sbitmap.h (bmp_iter_set_init): Remove non needed check.
8633 2017-11-09  Martin Liska  <mliska@suse.cz>
8635         PR gcov-profile/48463
8636         * coverage.c (coverage_begin_function): Output also end locus
8637         of a function and information whether the function is
8638         artificial.
8639         * gcov-dump.c (tag_function): Parse and print the information.
8640         * gcov.c (INCLUDE_MAP): Add include.
8641         (INCLUDE_SET): Likewise.
8642         (struct line_info): Move earlier in the source file because
8643         of vector<line_info> in function_info structure.
8644         (line_info::line_info): Likewise.
8645         (line_info::has_block): Likewise.
8646         (struct source_info): Add new member index.
8647         (source_info::get_functions_at_location): New function.
8648         (function_info::group_line_p): New function.
8649         (output_intermediate_line): New function.
8650         (output_intermediate_file): Use the mentioned function.
8651         (struct function_start): New.
8652         (struct function_start_pair_hash): Likewise.
8653         (process_file): Add code that identifies group functions.
8654         Assign lines either to global or function scope.
8655         (generate_results): Skip artificial functions.
8656         (find_source): Assign index for each source file.
8657         (read_graph_file): Read new flag artificial and end_line.
8658         (add_line_counts): Assign it either to global of function scope.
8659         (accumulate_line_counts): Isolate core of the function to
8660         accumulate_line_info and call it for both function and global
8661         scope lines.
8662         (accumulate_line_info): New function.
8663         (output_line_beginning): Fix GNU coding style.
8664         (print_source_line): New function.
8665         (output_line_details): Likewise.
8666         (output_function_details): Likewise.
8667         (output_lines): Iterate both source (global) scope and function
8668         scope.
8669         (struct function_line_start_cmp): New class.
8670         * doc/gcov.texi: Reflect changes in documentation.
8672 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8674         PR debug/82837
8675         * dwarf2out.c (const_ok_for_output_1): Reject NEG in addition to NOT.
8676         (mem_loc_descriptor): Handle (const (neg (...))) as (neg (const (...)))
8677         and similarly for not instead of neg.
8679 2017-11-08  Andi Kleen  <ak@linux.intel.com>
8681         * config/i386/i386.opt: Add -mforce-indirect-call.
8682         * config/i386/predicates.md: Check for flag_force_indirect_call.
8683         * doc/invoke.texi: Document -mforce-indirect-call
8685 2017-11-08  Kito Cheng  <kito.cheng@gmail.com>
8687         * config/riscv/riscv-protos.h (riscv_slow_unaligned_access_p):
8688         New extern.
8689         (MOVE_RATIO): Use riscv_slow_unaligned_access_p.
8690         config/riscv/riscv.c (predict.h): New include.
8691         (riscv_slow_unaligned_access_p): No longer static.
8692         (riscv_block_move_straight): Add require.
8693         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): Delete.
8695 2017-11-08  Jakub Jelinek  <jakub@redhat.com>
8697         PR target/82855
8698         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>,
8699         <avx512>_eq<mode>3<mask_scalar_merge_name>_1): Use
8700         nonimmediate_operand predicate for operand 1 instead of
8701         register_operand.
8703 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8705         * config/aarch64/aarch64-simd.md (store_pair_lanes<mode>):
8706         New pattern.
8707         * config/aarch64/constraints.md (Uml): New constraint.
8708         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): New
8709         predicate.
8711 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8713         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
8714         of two vec_duplicates into a vec_concat.
8716 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8718         * simplify-rtx.c (simplify_ternary_operation, VEC_MERGE):
8719         Simplify vec_merge of vec_duplicate and vec_concat.
8720         * config/aarch64/constraints.md (Utq): New constraint.
8721         * config/aarch64/aarch64-simd.md (load_pair_lanes<mode>): New
8722         define_insn.
8724 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8726         * simplify-rtx.c (simplify_ternary_operation, VEC_MERGE):
8727         Simplify vec_merge of vec_duplicate and const_vector.
8728         * config/aarch64/predicates.md (aarch64_simd_or_scalar_imm_zero):
8729         New predicate.
8730         * config/aarch64/aarch64-simd.md (*aarch64_combinez<mode>): Use VDC
8731         mode iterator.  Update predicate on operand 1 to
8732         handle non-const_vec constants.  Delete constraints.
8733         (*aarch64_combinez_be<mode>): Likewise for operand 2.
8735 2017-11-08  Jakub Jelinek  <jakub@redhat.com>
8737         PR tree-optimization/78821
8738         * gimple-ssa-store-merging.c (struct store_operand_info): Add bit_not_p
8739         data member.
8740         (store_operand_info::store_operand_info): Initialize it to false.
8741         (pass_store_merging::terminate_all_aliasing_chains): Rewritten to use
8742         ref_maybe_used_by_stmt_p and stmt_may_clobber_ref_p on lhs of each
8743         store in the group, and if chain_info is non-NULL, to ignore altogether
8744         that chain.
8745         (compatible_load_p): Fail if bit_not_p does not match.
8746         (imm_store_chain_info::output_merged_store): Handle bit_not_p loads.
8747         (handled_load): Fill in bit_not_p.  Handle BIT_NOT_EXPR.
8748         (pass_store_merging::process_store): Adjust
8749         terminate_all_aliasing_chains calls to pass NULL in all current spots,
8750         call terminate_all_aliasing_chains newly when adding a store into
8751         a chain with non-NULL chain_info.
8753 2017-11-08  Wilco Dijkstra  <wdijkstr@arm.com>
8755         * config/aarch64/aarch64.c (aarch64_can_eliminate): Simplify logic.
8757 2017-11-08  Wilco Dijkstra  <wdijkstr@arm.com>
8759         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
8760         Remove.
8761         (aarch64_layout_frame): Initialise emit_frame_chain.
8762         (aarch64_can_eliminate): Remove omit leaf frame pointer code.
8763         (TARGET_FRAME_POINTER_REQUIRED): Remove define.
8765 2017-11-08  Martin Liska  <mliska@suse.cz>
8767         * gimplify.c (expand_FALLTHROUGH_r): Simplify usage
8768         of gimple_call_internal_p.
8770 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8772         * config/mips/mips.h (ASM_OUTPUT_LABELREF): Wrap in "do {} while (0)".
8774 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8776         * config/mips/mips.h (ASM_OUTPUT_CASE_END): Remove semicolon after
8777         "do {} while (0)".
8779 2017-11-08  Martin Liska  <mliska@suse.cz>
8781         PR sanitizer/82792
8782         * gimplify.c (expand_FALLTHROUGH_r): Skip IFN_ASAN_MARK.
8784 2017-11-07  Eric Botcazou  <ebotcazou@adacore.com>
8786         * gimple-pretty-print.c (dump_profile): Return "" instead of NULL.
8788 2017-11-07  Jakub Jelinek  <jakub@redhat.com>
8790         PR target/82855
8791         * config/i386/i386.md (SWI1248_AVX512BWDQ2_64): New mode iterator.
8792         (*cmp<mode>_ccz_1): New insn with $k alternative.
8794         PR target/82855
8795         * config/i386/i386.c (ix86_swap_binary_operands_p): Treat
8796         RTX_COMM_COMPARE as commutative as well.
8797         (ix86_binary_operator_ok): Formatting fix.
8798         * config/i386/sse.md (*mul<mode>3<mask_name><round_name>,
8799         *<code><mode>3<mask_name><round_saeonly_name>,
8800         *<code><mode>3<mask_name>, *<code>tf3, *mul<mode>3<mask_name>,
8801         *<s>mul<mode>3_highpart<mask_name>,
8802         *vec_widen_umult_even_v16si<mask_name>,
8803         *vec_widen_umult_even_v8si<mask_name>,
8804         *vec_widen_umult_even_v4si<mask_name>,
8805         *vec_widen_smult_even_v16si<mask_name>,
8806         *vec_widen_smult_even_v8si<mask_name>, *sse4_1_mulv2siv2di3<mask_name>,
8807         *avx2_pmaddwd, *sse2_pmaddwd, *<sse4_1_avx2>_mul<mode>3<mask_name>,
8808         *avx2_<code><mode>3, *avx512f_<code><mode>3<mask_name>,
8809         *sse4_1_<code><mode>3<mask_name>, *<code>v8hi3,
8810         *sse4_1_<code><mode>3<mask_name>, *<code>v16qi3, *avx2_eq<mode>3,
8811         <avx512>_eq<mode>3<mask_scalar_merge_name>_1, *sse4_1_eqv2di3,
8812         *sse2_eq<mode>3, <mask_codefor><code><mode>3<mask_name>,
8813         *<code><mode>3, *<sse2_avx2>_uavg<mode>3<mask_name>,
8814         *<ssse3_avx2>_pmulhrsw<mode>3<mask_name>, *ssse3_pmulhrswv4hi3): Use
8815         !(MEM_P (operands[1]) && MEM_P (operands[2])) condition instead of
8816         ix86_binary_operator_ok.  Formatting fixes.
8817         (*<plusminus_insn><mode>3<mask_name><round_name>,
8818         *<plusminus_insn><mode>3, *<plusminus_insn><mode>3_m): Formatting
8819         fixes.
8821 2017-11-07  Segher Boessenkool  <segher@kernel.crashing.org>
8823         * config/rs6000/rs6000.md (GPR2): New mode_iterator.
8824         ("cstore<mode>4"): Don't always expand with rs6000_emit_int_cmove for
8825         eq and ne if TARGET_ISEL.
8826         (cmp): New code_iterator.
8827         (UNS, UNSU_, UNSIK): New code_attrs.
8828         (<code><GPR:mode><GPR2:mode>2_isel): New define_insn_and_split.
8829         ("eq<mode>3"): New define_expand, rename the define_insn_and_split
8830         to...
8831         ("eq<mode>3"): ... this.
8832         ("ne<mode>3"): New define_expand, rename the define_insn_and_split
8833         to...
8834         ("ne<mode>3"): ... this.
8836 2017-11-07  Julia Koval  <julia.koval@intel.com>
8838         PR target/82812
8839         * common/config/i386/i386-common.c
8840         (OPTION_MASK_ISA_GENERAL_REGS_ONLY_UNSET): Remove MPX from flag.
8841         (ix86_handle_option): Move MPX to isa_flags2 and GFNI to isa_flags.
8842         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto.
8843         * config/i386/i386.opt: Ditto.
8844         * config/i386/i386.c (ix86_target_string): Ditto.
8845         (ix86_option_override_internal): Ditto.
8846         (ix86_init_mpx_builtins): Move MPX to args2.
8847         (ix86_expand_builtin): Special handling for OPTION_MASK_ISA_GFNI.
8848         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8affineinvqb_v64qi,
8849         __builtin_ia32_vgf2p8affineinvqb_v64qi_mask,
8850         __builtin_ia32_vgf2p8affineinvqb_v32qi,
8851         __builtin_ia32_vgf2p8affineinvqb_v32qi_mask,
8852         __builtin_ia32_vgf2p8affineinvqb_v16qi,
8853         __builtin_ia32_vgf2p8affineinvqb_v16qi_mask): Move to ARGS array.
8855 2017-11-07  Uros Bizjak  <ubizjak@gmail.com>
8857         PR target/80425
8858         * config/i386.i386.md (*zero_extendsidi2): Change (?r,*Yj), (?*Yi,r)
8859         and (*x,m) to ($r,Yj), ($Yi,r) and ($x,m).
8860         (zero-extendsidi peephole2): Remove peephole.
8862 2017-11-07  Eric Botcazou  <ebotcazou@adacore.com>
8864         PR c/53037
8865         * stor-layout.c: Include attribs.h.
8866         (handle_warn_if_not_align): Replace test on TYPE_USER_ALIGN with
8867         explicit lookup of "aligned" attribute.
8869 2017-11-07  Andrew Waterman  <andrew@sifive.com>
8871         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): New prototype.
8872         (riscv_expand_block_move): Likewise.
8873         * config/riscv/riscv.h (MOVE_RATIO): Tune cost to movmemsi
8874         implementation.
8875         (RISCV_MAX_MOVE_BYTES_PER_LOOP_ITER): New define.
8876         (RISCV_MAX_MOVE_BYTES_STRAIGHT): New define.
8877         * config/riscv/riscv.c (riscv_block_move_straight): New function.
8878         (riscv_adjust_block_mem): Likewise.
8879         (riscv_block_move_loop): Likewise.
8880         (riscv_expand_block_move): Likewise.
8881         * config/riscv/riscv.md (movmemsi): New pattern.
8883 2017-11-07  Michael Clark  <michaeljclark@mac.com>
8885         * config/riscv/linux.h (MUSL_ABI_SUFFIX): New define.
8886         (MUSL_DYNAMIC_LINKER): Likewise.
8888 2017-11-07  Richard Sandiford  <richard.sandiford@linaro.org>
8890         * config/aarch64/aarch64.md (ashl<mode>3, ashr<mode>3, lshr<mode>3)
8891         (rotr<mode>3, rotl<mode>3): Use aarch64_reg_or_imm instead of
8892         nonmmory_operand.
8894 2017-11-07  Richard Biener  <rguenther@suse.de>
8896         * match.pd: Fix build.
8898 2017-11-07  Wilco Dijkstra  <wdijkstr@arm.com>
8899             Jackson Woodruff  <jackson.woodruff@arm.com>
8901         PR tree-optimization/71026
8902         * match.pd: Canonicalize negate in division.
8904 2017-11-07  Sudakshina Das  <sudi.das@arm.com>
8906         PR middle-end/80131
8907         * match.pd: Simplify 1 << (C - x) where C = precision (x) - 1.
8909 2017-11-07  Marc Glisse  <marc.glisse@inria.fr>
8911         * match.pd ((a&~b)|(a^b),(a&~b)^~a,(a|b)&~(a^b),a|~(a^b),
8912         (a|b)|(a&^b),(a&b)|~(a^b),~(~a&b),~X^Y): New transformations.
8914 2017-11-07  Marc Glisse  <marc.glisse@inria.fr>
8916         * fold-const.c (negate_expr_p) [PLUS_EXPR, MINUS_EXPR]: Handle
8917         non-scalar integral types.
8918         * match.pd (negate_expr_p): Handle MINUS_EXPR.
8919         (-(A-B), -(~A)): New transformations.
8921 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8923         * config/powerpcspe/aix43.h (SUBTARGET_OVERRIDE_OPTIONS): Remove
8924         semicolon after "do {} while (0)".
8925         * config/powerpcspe/aix51.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8926         * config/powerpcspe/aix52.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8927         * config/powerpcspe/aix53.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8928         * config/powerpcspe/aix61.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8929         * config/powerpcspe/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8931 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8933         * config/rs6000/aix43.h (SUBTARGET_OVERRIDE_OPTIONS): Remove semicolon
8934         after "do {} while (0)".
8935         * config/rs6000/aix51.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8936         * config/rs6000/aix52.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8937         * config/rs6000/aix53.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8938         * config/rs6000/aix61.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8939         * config/rs6000/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8941 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8943         PR other/82784
8944         * config/arm/arm.c (HANDLE_NARROW_SHIFT_ARITH): Remove semicolon after
8945         "while {} do (0)".
8946         (arm_rtx_costs_internal): Add missing semicolon after
8947         HANDLE_NARROW_SHIFT_ARITH call.
8949 2017-11-06  Segher Boessenkool  <segher@kernel.crashing.org>
8951         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
8952         disable isel if it was not set explicitly.
8954 2017-11-06  James Bowman  <james.bowman@ftdichip.com>
8956         * gcc/dwarf2out.c (modified_type_die): Retain ADDR_SPACE qualifiers.
8957         (add_type_attribute) likewise.
8959 2017-11-06  H.J. Lu  <hongjiu.lu@intel.com>
8961         * config/i386/i386.c (ix86_can_use_return_insn_p): Use reference
8962         of struct ix86_frame.
8963         (ix86_initial_elimination_offset): Likewise.
8964         (ix86_expand_split_stack_prologue): Likewise.
8966 2017-11-06  Marc Glisse  <marc.glisse@inria.fr>
8968         * tree-vrp.h (enum value_range_type): Update stale comment.
8970 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8971             Alan Hayward  <alan.hayward@arm.com>
8972             David Sherwood  <david.sherwood@arm.com>
8974         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm)
8975         (aarch64_expand_vec_perm_const): Take the number of units too.
8976         * config/aarch64/aarch64.c (aarch64_expand_vec_perm)
8977         (aarch64_expand_vec_perm_const): Likewise.
8978         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>)
8979         (vec_perm<mode>): Update accordingly.
8981 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8982             Alan Hayward  <alan.hayward@arm.com>
8983             David Sherwood  <david.sherwood@arm.com>
8985         * config/aarch64/aarch64-protos.h (aarch64_simd_vect_par_cnst_half):
8986         Take the number of units too.
8987         * config/aarch64/aarch64.c (aarch64_simd_vect_par_cnst_half): Likewise.
8988         (aarch64_simd_check_vect_par_cnst_half): Update call accordingly,
8989         but check for a vector mode before rather than after the call.
8990         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
8991         (move_hi_quad_<mode>, vec_unpack<su>_hi_<mode>)
8992         (vec_unpack<su>_lo_<mode, vec_widen_<su>mult_lo_<mode>)
8993         (vec_widen_<su>mult_hi_<mode>, vec_unpacks_lo_<mode>)
8994         (vec_unpacks_hi_<mode>, aarch64_saddl2<mode>, aarch64_uaddl2<mode>)
8995         (aarch64_ssubl2<mode>, aarch64_usubl2<mode>, widen_ssum<mode>3)
8996         (widen_usum<mode>3, aarch64_saddw2<mode>, aarch64_uaddw2<mode>)
8997         (aarch64_ssubw2<mode>, aarch64_usubw2<mode>, aarch64_sqdmlal2<mode>)
8998         (aarch64_sqdmlsl2<mode>, aarch64_sqdmlal2_lane<mode>)
8999         (aarch64_sqdmlal2_laneq<mode>, aarch64_sqdmlsl2_lane<mode>)
9000         (aarch64_sqdmlsl2_laneq<mode>, aarch64_sqdmlal2_n<mode>)
9001         (aarch64_sqdmlsl2_n<mode>, aarch64_sqdmull2<mode>)
9002         (aarch64_sqdmull2_lane<mode>, aarch64_sqdmull2_laneq<mode>)
9003         (aarch64_sqdmull2_n<mode>): Update accordingly.
9005 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
9006             Alan Hayward  <alan.hayward@arm.com>
9007             David Sherwood  <david.sherwood@arm.com>
9009         * config/aarch64/aarch64-protos.h (aarch64_reverse_mask): Take
9010         the number of units too.
9011         * config/aarch64/aarch64.c (aarch64_reverse_mask): Likewise.
9012         * config/aarch64/aarch64-simd.md (vec_load_lanesoi<mode>)
9013         (vec_store_lanesoi<mode>, vec_load_lanesci<mode>)
9014         (vec_store_lanesci<mode>, vec_load_lanesxi<mode>)
9015         (vec_store_lanesxi<mode>): Update accordingly.
9017 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
9018             Alan Hayward  <alan.hayward@arm.com>
9019             David Sherwood  <david.sherwood@arm.com>
9021         * config/aarch64/aarch64-protos.h (aarch64_endian_lane_rtx): Declare.
9022         * config/aarch64/aarch64.c (aarch64_endian_lane_rtx): New function.
9023         * config/aarch64/aarch64.h (ENDIAN_LANE_N): Take the number
9024         of units rather than the mode.
9025         * config/aarch64/iterators.md (nunits): New mode attribute.
9026         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args):
9027         Use aarch64_endian_lane_rtx instead of GEN_INT (ENDIAN_LANE_N ...).
9028         * config/aarch64/aarch64-simd.md (aarch64_dup_lane<mode>)
9029         (aarch64_dup_lane_<vswap_width_name><mode>, *aarch64_mul3_elt<mode>)
9030         (*aarch64_mul3_elt_<vswap_width_name><mode>): Likewise.
9031         (*aarch64_mul3_elt_to_64v2df, *aarch64_mla_elt<mode>): Likewise.
9032         (*aarch64_mla_elt_<vswap_width_name><mode>, *aarch64_mls_elt<mode>)
9033         (*aarch64_mls_elt_<vswap_width_name><mode>, *aarch64_fma4_elt<mode>)
9034         (*aarch64_fma4_elt_<vswap_width_name><mode>):: Likewise.
9035         (*aarch64_fma4_elt_to_64v2df, *aarch64_fnma4_elt<mode>): Likewise.
9036         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
9037         (*aarch64_fnma4_elt_to_64v2df, reduc_plus_scal_<mode>): Likewise.
9038         (reduc_plus_scal_v4sf, reduc_<maxmin_uns>_scal_<mode>): Likewise.
9039         (reduc_<maxmin_uns>_scal_<mode>): Likewise.
9040         (*aarch64_get_lane_extend<GPI:mode><VDQQH:mode>): Likewise.
9041         (*aarch64_get_lane_zero_extendsi<mode>): Likewise.
9042         (aarch64_get_lane<mode>, *aarch64_mulx_elt_<vswap_width_name><mode>)
9043         (*aarch64_mulx_elt<mode>, *aarch64_vgetfmulx<mode>): Likewise.
9044         (aarch64_sq<r>dmulh_lane<mode>, aarch64_sq<r>dmulh_laneq<mode>)
9045         (aarch64_sqrdml<SQRDMLH_AS:rdma_as>h_lane<mode>): Likewise.
9046         (aarch64_sqrdml<SQRDMLH_AS:rdma_as>h_laneq<mode>): Likewise.
9047         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>): Likewise.
9048         (aarch64_sqdml<SBINQOPS:as>l_laneq<mode>): Likewise.
9049         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
9050         (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>_internal): Likewise.
9051         (aarch64_sqdmull_lane<mode>, aarch64_sqdmull_laneq<mode>): Likewise.
9052         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
9053         (aarch64_sqdmull2_laneq<mode>_internal): Likewise.
9054         (aarch64_vec_load_lanesoi_lane<mode>): Likewise.
9055         (aarch64_vec_store_lanesoi_lane<mode>): Likewise.
9056         (aarch64_vec_load_lanesci_lane<mode>): Likewise.
9057         (aarch64_vec_store_lanesci_lane<mode>): Likewise.
9058         (aarch64_vec_load_lanesxi_lane<mode>): Likewise.
9059         (aarch64_vec_store_lanesxi_lane<mode>): Likewise.
9060         (aarch64_simd_vec_set<mode>): Update use of ENDIAN_LANE_N.
9061         (aarch64_simd_vec_setv2di): Likewise.
9063 2017-11-06  Carl Love  <cel@us.ibm.com>
9065         * config/rs6000/rs6000-c.c (P8V_BUILTIN_VEC_REVB): Add power 8
9066         definitions.
9067         (P9V_BUILTIN_VEC_REVB): Remove the power 9 instance definitions.
9068         * config/rs6000/altivec.h (vec_revb): Change the #define from power 9
9069         to power 8.
9070         * config/rs6000/r6000-protos.h (swap_endian_selector_for_mode): Add new
9071         extern declaration.
9072         * config/rs6000/rs6000.c (swap_endian_selector_for_mode): Add function.
9073         * config/rs6000/rs6000-builtin.def (BU_P8V_VSX_1, BU_P8V_OVERLOAD_1):
9074         Add power 8 macro expansions.
9075         (BU_P9V_OVERLOAD_1): Remove power 9 overload expansion.
9076         * config/rs6000/vsx.md (revb_<mode>): Add define_expand to generate
9077         power 8 instructions.  (VSX_XXBR): Add iterator.
9079 2017-11-06  Wilco Dijkstra  <wdijkstr@arm.com>
9081         * config/arm/arm.md (predicable_short_it): Change default to "no",
9082         improve documentation, remove uses that are identical to the default.
9083         (enabled_for_depr_it): Rename to enabled_for_short_it.
9084         * gcc/config/arm/arm-fixed.md (predicable_short_it):
9085         Remove default uses.
9086         * gcc/config/arm/ldmstm.md (predicable_short_it): Likewise.
9087         * gcc/config/arm/sync.md (predicable_short_it): Likewise.
9088         * gcc/config/arm/thumb2.md (predicable_short_it): Likewise.
9089         * gcc/config/arm/vfp.md (predicable_short_it): Likewise.
9091 2017-11-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
9093         PR target/82748
9094         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_1): Delete
9095         float128 helper macros, which are no longer used after deleting
9096         the old 'q' built-in functions, and moving the round to odd
9097         built-in functions to being special built-in functions.
9098         (BU_FLOAT128_2): Likewise.
9099         (BU_FLOAT128_1_HW): Likewise.
9100         (BU_FLOAT128_2_HW): Likewise.
9101         (BU_FLOAT128_3_HW): Likewise.
9102         (FABSQ): Delete old 'q' built-in functions.
9103         (COPYSIGNQ): Likewise.
9104         (SQRTF128_ODD): Move round to odd built-in functions to be
9105         special built-in functions, so that we can handle
9106         -mabi=ieeelongdouble.
9107         (TRUNCF128_ODD): Likewise.
9108         (ADDF128_ODD): Likewise.
9109         (SUBF128_ODD): Likewise.
9110         (MULF128_ODD): Likewise.
9111         (DIVF128_ODD): Likewise.
9112         (FMAF128_ODD): Likewise.
9113         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Map old 'q'
9114         built-in names to 'f128'.
9115         * config/rs6000/rs6000.c (rs6000_fold_builtin): Remove folding the
9116         old 'q' built-in functions, as the machine independent code for
9117         'f128' built-in functions handles this.
9118         (rs6000_expand_builtin): Add expansion for float128 round to odd
9119         functions, keying off on -mabi=ieeelongdouble of whether to use
9120         the KFmode or TFmode variant.
9121         (rs6000_init_builtins): Initialize the _Float128 round to odd
9122         built-in functions.
9123         * doc/extend.texi (PowerPC Built-in Functions): Document the old
9124         _Float128 'q' built-in functions are now mapped into the new
9125         'f128' built-in functions.
9127 2017-11-06  David Edelsohn  <dje.gcc@gmail.com>
9129         * collect2.c (add_lto_object): Compile for OBJECT_COFF.
9130         (scan_prog_file): Don't skip PASS_LTOINFO. Scan for LTO objects.
9132 2017-11-06  David Malcolm  <dmalcolm@redhat.com>
9134         PR jit/82826
9135         * ipa-fnsummary.c (ipa_fnsummary_c_finalize): New function.
9136         * ipa-fnsummary.h (ipa_fnsummary_c_finalize): New decl.
9137         * toplev.c: Include "ipa-fnsummary.h".
9138         (toplev::finalize): Call ipa_fnsummary_c_finalize.
9140 2017-11-06  Jakub Jelinek  <jakub@redhat.com>
9142         PR tree-optimization/82838
9143         * gimple-ssa-store-merging.c
9144         (imm_store_chain_info::output_merged_store): Call force_gimple_operand_1
9145         on a separate gimple_seq which is then appended to seq.
9147 2017-11-06  Jeff Law  <law@redhat.com>
9149         PR target/82788
9150         * config/i386/i386.c (PROBE_INTERVAL): Remove.
9151         (get_probe_interval): New functions.
9152         (ix86_adjust_stack_and_probe_stack_clash): Use get_probe_interval.
9153         (ix86_adjust_stack_and_probe): Likewise.
9154         (output_adjust_stack_and_probe): Likewise.
9155         (ix86_emit_probe_stack_range): Likewise.
9156         (ix86_expand_prologue): Likewise.
9158 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
9160         PR tree-optimization/82816
9161         * tree-ssa-math-opts.c (convert_mult_to_widen): Return false
9162         if the modes of the two types are the same.
9163         (convert_plusminus_to_widen): Likewise.
9165 2017-11-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
9167         * config/rs6000/altivec.md (*p9_vadu<mode>3) Rename to
9168         p9_vadu<mode>3.
9169         (usadv16qi): New define_expand.
9170         (usadv8hi): New define_expand.
9172 2017-11-06  Jan Hubicka  <hubicka@ucw.cz>
9174         PR bootstrap/82832
9175         * ipa-inline-transform.c (update_noncloned_frequencies): Always
9176         scale.
9177         (inline_transform): Likewise.
9178         * predict.c (counts_to_freqs): Remove useless conditional.
9179         * profile-count.h (profile_count::apply_scale): Move sanity check.
9180         * tree-inline.c (copy_bb): Always scale.
9181         (copy_cfg_body): Likewise.
9183 2017-11-06  Christophe Lyon  <christophe.lyon@linaro.org>
9185         PR target/67591
9186         * config/arm/arm.md (*sub_shiftsi): Add predicable_short_it
9187         attribute.
9188         (*cmp_ite0): Add enabled_for_depr_it attribute.
9189         (*cmp_ite1): Likewise.
9191 2017-11-06  Segher Boessenkool  <segher@kernel.crashing.org>
9193         * config/rs6000/rs6000.c (rs6000_insn_cost): Handle TYPE_MFCR and
9194         TYPE_MFCRF.
9196 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
9198         * tree-vrp.c (vrp_int_const_binop): Return true on success and
9199         return the value by pointer.
9200         (extract_range_from_multiplicative_op_1): Update accordingly.
9201         Return as soon as an operation fails.
9203 2017-11-05  Tom de Vries  <tom@codesourcery.com>
9205         PR other/82784
9206         * asan.c (DEF_SANITIZER_BUILTIN_1): Factor out of ...
9207         (DEF_SANITIZER_BUILTIN): ... here.
9208         (initialize_sanitizer_builtins): Use DEF_SANITIZER_BUILTIN_1 instead of
9209         DEF_SANITIZER_BUILTIN in if stmt.  Add missing semicolon.
9211 2017-11-05  Tom de Vries  <tom@codesourcery.com>
9213         PR other/82784
9214         * config/elfos.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove semicolon after
9215         macro body.
9216         (ASM_OUTPUT_CASE_LABEL): Add semicolon after
9217         ASM_OUTPUT_BEFORE_CASE_LABEL call.
9218         * config/arc/arc.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove semicolon
9219         after macro body.
9220         * config/m68k/m68kelf.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
9221         * config/mips/mips.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
9222         * config/v850/v850.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
9224 2017-11-05  Tom de Vries  <tom@codesourcery.com>
9226         PR other/82784
9227         * graphite-scop-detection.c (DEBUG_PRINT): Remove semicolon after
9228         "do {} while (0)".
9230 2017-11-04  Michael Clark  <michaeljclark@mac.com>
9232         * config/riscv/riscv.c (riscv_print_operand): Add a 'i' format.
9233         config/riscv/riscv.md (addsi3): Use 'i' for immediates.
9234         (adddi3): Likewise.
9235         (*addsi3_extended): Likewise.
9236         (*addsi3_extended2): Likewise.
9237         (<optab>si3): Likewise.
9238         (<optab>di3): Likewise.
9239         (<optab><mode>3): Likewise.
9240         (<*optabe>si3_internal): Likewise.
9241         (zero_extendqi<SUPERQI:mode>2): Likewise.
9242         (*add<mode>hi3): Likewise.
9243         (*xor<mode>hi3): Likewise.
9244         (<optab>di3): Likewise.
9245         (*<optab>si3_extend): Likewise.
9246         (*sge<u>_<X:mode><GPR:mode>): Likewise.
9247         (*slt<u>_<X:mode><GPR:mode>): Likewise.
9248         (*sle<u>_<X:mode><GPR:mode>): Likewise.
9250 2017-11-04  Andrew Waterman  <andrew@sifive.com>
9252         * config/riscv/riscv.c (riscv_option_override): Conditionally set
9253         TARGET_STRICT_ALIGN based upon -mtune argument.
9255 2017-11-04  Andrew Waterman  <andrew@sifive.com>
9257         * config/riscv/riscv.h (SLOW_BYTE_ACCESS): Change to 1.
9259 2017-11-04  Daniel Santos  <daniel.santos@pobox.com>
9261         * config/i386/i386.c (choose_basereg): Use optional scratch
9262         register and add assertion.
9263         (x86_emit_outlined_ms2sysv_save): Use scratch register when
9264         needed, and don't allocate stack.
9265         (ix86_expand_prologue): Rearrange where SSE saves/stub call is
9266         emitted, correct wrong allocation with -mcall-ms2sysv-xlogues.
9267         (ix86_emit_outlined_ms2sysv_restore): Fix non-immediate offsets.
9269 2017-11-03  Jeff Law  <law@redhat.com>
9271         * config/i386/i386.c (ix86_emit_restore_reg_using_pop): Prototype.
9272         (ix86_adjust_stack_and_probe_stack_clash): Use a push/pop sequence
9273         to probe at the start of a noreturn function.
9275 2017-11-03  Jakub Jelinek  <jakub@redhat.com>
9277         PR tree-optimization/78821
9278         * gimple-ssa-store-merging.c: Update the file comment.
9279         (MAX_STORE_ALIAS_CHECKS): Define.
9280         (struct store_operand_info): New type.
9281         (store_operand_info::store_operand_info): New constructor.
9282         (struct store_immediate_info): Add rhs_code and ops data members.
9283         (store_immediate_info::store_immediate_info): Add rhscode, op0r
9284         and op1r arguments to the ctor, initialize corresponding data members.
9285         (struct merged_store_group): Add load_align_base and load_align
9286         data members.
9287         (merged_store_group::merged_store_group): Initialize them.
9288         (merged_store_group::do_merge): Update them.
9289         (merged_store_group::apply_stores): Pick the constant for
9290         encode_tree_to_bitpos from one of the two operands, or skip
9291         encode_tree_to_bitpos if neither operand is a constant.
9292         (class pass_store_merging): Add process_store method decl.  Remove
9293         bool argument from terminate_all_aliasing_chains method decl.
9294         (pass_store_merging::terminate_all_aliasing_chains): Remove
9295         var_offset_p argument and corresponding handling.
9296         (stmts_may_clobber_ref_p): New function.
9297         (compatible_load_p): New function.
9298         (imm_store_chain_info::coalesce_immediate_stores): Terminate group
9299         if there is overlap and rhs_code is not INTEGER_CST.  For
9300         non-overlapping stores terminate group if rhs is not mergeable.
9301         (get_alias_type_for_stmts): Change first argument from
9302         auto_vec<gimple *> & to vec<gimple *> &.  Add IS_LOAD, CLIQUEP and
9303         BASEP arguments.  If IS_LOAD is true, look at rhs1 of the stmts
9304         instead of lhs.  Compute *CLIQUEP and *BASEP in addition to the
9305         alias type.
9306         (get_location_for_stmts): Change first argument from
9307         auto_vec<gimple *> & to vec<gimple *> &.
9308         (struct split_store): Remove orig_stmts data member, add orig_stores.
9309         (split_store::split_store): Create orig_stores rather than orig_stmts.
9310         (find_constituent_stmts): Renamed to ...
9311         (find_constituent_stores): ... this.  Change second argument from
9312         vec<gimple *> * to vec<store_immediate_info *> *, push pointers
9313         to info structures rather than the statements.
9314         (split_group): Rename ALLOW_UNALIGNED argument to
9315         ALLOW_UNALIGNED_STORE, add ALLOW_UNALIGNED_LOAD argument and handle
9316         it.  Adjust find_constituent_stores caller.
9317         (imm_store_chain_info::output_merged_store): Handle rhs_code other
9318         than INTEGER_CST, adjust split_group, get_alias_type_for_stmts and
9319         get_location_for_stmts callers.  Set MR_DEPENDENCE_CLIQUE and
9320         MR_DEPENDENCE_BASE on the MEM_REFs if they are the same in all stores.
9321         (mem_valid_for_store_merging): New function.
9322         (handled_load): New function.
9323         (pass_store_merging::process_store): New method.
9324         (pass_store_merging::execute): Use process_store method.  Adjust
9325         terminate_all_aliasing_chains caller.
9327 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9329         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9330         Return true for more constants, symbols and label references.
9331         (aarch64_valid_floating_const): Remove unused function.
9333 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9335         PR target/82786
9336         * config/aarch64/aarch64.c (aarch64_layout_frame):
9337         Undo forcing of LR at bottom of frame.
9339 2017-11-03  Jeff Law  <law@redhat.com>
9341         PR target/82823
9342         * config/i386/i386.c (ix86_expand_prologue): Tighten assert
9343         for int_registers_saved.
9345         * cfganal.c (single_pred_edge_ignoring_loop_edges): New function
9346         extracted from tree-ssa-dom.c.
9347         * cfganal.h (single_pred_edge_ignoring_loop_edges): Prototype.
9348         * tree-ssa-dom.c (single_incoming_edge_ignoring_loop_edges): Remove.
9349         (record_equivalences_from_incoming_edge): Add additional argument
9350         to single_pred_edge_ignoring_loop_edges call.
9351         * tree-ssa-uncprop.c (single_incoming_edge_ignoring_loop_edges): Remove.
9352         (uncprop_dom_walker::before_dom_children): Add additional argument
9353         to single_pred_edge_ignoring_loop_edges call.
9354         * tree-ssa-sccvn.c (sccvn_dom_walker::before_dom_children): Use
9355         single_pred_edge_ignoring_loop_edges rather than open coding.
9356         * tree-vrp.c (evrp_dom_walker::before_dom_children): Similarly.
9358 2017-11-03  Marc Glisse  <marc.glisse@inria.fr>
9360         * match.pd (-(-A)): Rewrite.
9362 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9364         * config/rs6000/rs60000-protos.h (rs6000_emit_sISEL): Delete.
9365         (rs6000_emit_int_cmove): New declaration.
9366         * config/rs6000/rs6000.c (rs6000_emit_int_cmove): Delete declaration.
9367         (rs6000_emit_sISEL): Delete.
9368         (rs6000_emit_int_cmove): Make non-static.
9369         * config/rs6000/rs6000.md (cstore<mode>4): Use rs6000_emit_int_cmove
9370         instead of rs6000_emit_sISEL.
9372 2017-11-03  Jan Hubicka  <hubicka@ucw.cz>
9374         * asan.c (create_cond_insert_point): Maintain profile.
9375         * ipa-utils.c (ipa_merge_profiles): Be sure only IPA profiles are
9376         merged.
9377         * basic-block.h (struct basic_block_def): Remove frequency.
9378         (EDGE_FREQUENCY): Use to_frequency
9379         * bb-reorder.c (push_to_next_round_p): Use only IPA counts for global
9380         heuristics.
9381         (find_traces): Update to use to_frequency.
9382         (find_traces_1_round): Likewise; use only IPA counts.
9383         (bb_to_key): Likewise.
9384         (connect_traces): Use IPA counts only.
9385         (copy_bb_p): Update to use to_frequency.
9386         (fix_up_crossing_landing_pad): Likewise.
9387         (sanitize_hot_paths): Likewise.
9388         * bt-load.c (basic_block_freq): Likewise.
9389         * cfg.c (init_flow): Set count_max to uninitialized.
9390         (check_bb_profile): Remove frequencies; check counts.
9391         (dump_bb_info): Do not dump frequencies.
9392         (update_bb_profile_for_threading): Update counts only.
9393         (scale_bbs_frequencies_int): Likewise.
9394         (MAX_SAFE_MULTIPLIER): Remove.
9395         (scale_bbs_frequencies_gcov_type): Update counts only.
9396         (scale_bbs_frequencies_profile_count): Update counts only.
9397         (scale_bbs_frequencies): Update counts only.
9398         * cfg.h (struct control_flow_graph): Add count-max.
9399         (update_bb_profile_for_threading): Update prototype.
9400         * cfgbuild.c (find_bb_boundaries): Do not update frequencies.
9401         (find_many_sub_basic_blocks): Likewise.
9402         * cfgcleanup.c (try_forward_edges): Likewise.
9403         (try_crossjump_to_edge): Likewise.
9404         * cfgexpand.c (expand_gimple_cond): Likewise.
9405         (expand_gimple_tailcall): Likewise.
9406         (construct_init_block): Likewise.
9407         (construct_exit_block): Likewise.
9408         * cfghooks.c (verify_flow_info): Check consistency of counts.
9409         (dump_bb_for_graph): Do not dump frequencies.
9410         (split_block_1): Do not update frequencies.
9411         (split_edge): Do not update frequencies.
9412         (make_forwarder_block): Do not update frequencies.
9413         (duplicate_block): Do not update frequencies.
9414         (account_profile_record): Do not update frequencies.
9415         * cfgloop.c (find_subloop_latch_edge_by_profile): Use IPA counts
9416         for global heuristics.
9417         * cfgloopanal.c (average_num_loop_insns): Update to use to_frequency.
9418         (expected_loop_iterations_unbounded): Use counts only.
9419         * cfgloopmanip.c (scale_loop_profile): Simplify.
9420         (create_empty_loop_on_edge): Simplify
9421         (loopify): Simplify
9422         (duplicate_loop_to_header_edge): Simplify
9423         * cfgrtl.c (force_nonfallthru_and_redirect): Update profile.
9424         (update_br_prob_note): Take care of removing note when profile
9425         becomes undefined.
9426         (relink_block_chain): Do not dump frequency.
9427         (rtl_account_profile_record): Use to_frequency.
9428         * cgraph.c (symbol_table::create_edge): Convert count to ipa count.
9429         (cgraph_edge::redirect_call_stmt_to_calle): Conver tcount to ipa count.
9430         (cgraph_update_edges_for_call_stmt_node): Likewise.
9431         (cgraph_edge::verify_count_and_frequency): Update.
9432         (cgraph_node::verify_node): Temporarily disable frequency verification.
9433         * cgraphbuild.c (compute_call_stmt_bb_frequency): Use
9434         to_cgraph_frequency.
9435         (cgraph_edge::rebuild_edges): Convert to ipa counts.
9436         * cgraphunit.c (init_lowered_empty_function): Do not initialize
9437         frequencies.
9438         (cgraph_node::expand_thunk): Update profile.
9439         * except.c (dw2_build_landing_pads): Do not update frequency.
9440         * final.c (compute_alignments): Use to_frequency.
9441         (dump_basic_block_info): Do not dump frequency.
9442         * gimple-pretty-print.c (dump_profile): Do not dump frequency.
9443         (dump_gimple_bb_header): Do not dump frequency.
9444         * gimple-ssa-isolate-paths.c (isolate_path): Do not update frequency;
9445         do update count.
9446         * gimple-streamer-in.c (input_bb): Do not stream frequency.
9447         * gimple-streamer-out.c (output_bb): Do not stream frequency.
9448         * haifa-sched.c (sched_pressure_start_bb): Use to_freuqency.
9449         (init_before_recovery): Do not update frequency.
9450         (sched_create_recovery_edges): Do not update frequency.
9451         * hsa-gen.c (convert_switch_statements): Do not update frequency.
9452         * ipa-cp.c (ipcp_propagate_stage): Update search for max_count.
9453         (ipa_cp_c_finalize): Set max_count to uninitialized.
9454         * ipa-fnsummary.c (get_minimal_bb): Use counts.
9455         (param_change_prob): Use counts.
9456         * ipa-profile.c (ipa_profile_generate_summary): Do not summarize
9457         local profiles.
9458         * ipa-split.c (consider_split): Use to_frequency.
9459         (split_function): Use to_frequency.
9460         * ira-build.c (loop_compare_func): Likewise.
9461         (mark_loops_for_removal): Likewise.
9462         (mark_all_loops_for_removal): Likewise.
9463         * loop-doloop.c (doloop_modify): Do not update frequency.
9464         * loop-unroll.c (unroll_loop_runtime_iterations): Do not update
9465         frequency.
9466         * lto-streamer-in.c (input_function): Update count_max.
9467         * omp-expand.c (expand_omp_taskreg): Update count_max.
9468         * omp-simd-clone.c (simd_clone_adjust): Update profile.
9469         * predict.c (maybe_hot_frequency_p): Use to_frequency.
9470         (maybe_hot_count_p): Use ipa counts only.
9471         (maybe_hot_bb_p): Simplify.
9472         (maybe_hot_edge_p): Simplify.
9473         (probably_never_executed): Do not take frequency argument.
9474         (probably_never_executed_bb_p): Do not pass frequency.
9475         (probably_never_executed_edge_p): Likewise.
9476         (combine_predictions_for_bb): Check that profile is nonzero.
9477         (propagate_freq): Do not set frequency.
9478         (drop_profile): Simplify.
9479         (counts_to_freqs): Simplify.
9480         (expensive_function_p): Use to_frequency.
9481         (propagate_unlikely_bbs_forward): Simplify.
9482         (determine_unlikely_bbs): Simplify.
9483         (estimate_bb_frequencies): Add hack to silence graphite issues.
9484         (compute_function_frequency): Use ipa counts.
9485         (pass_profile::execute): Update.
9486         (rebuild_frequencies): Use counts only.
9487         (force_edge_cold): Use counts only.
9488         * profile-count.c (profile_count::dump): Dump new count types.
9489         (profile_count::differs_from_p): Check compatiblity.
9490         (profile_count::to_frequency): New function.
9491         (profile_count::to_cgraph_frequency): New function.
9492         * profile-count.h (struct function): Declare.
9493         (enum profile_quality): Add profile_guessed_local and
9494         profile_guessed_global0.
9495         (class profile_proability): Decrease number of bits to 29;
9496         update from_reg_br_prob_note and to_reg_br_prob_note.
9497         (class profile_count: Update comment; decrease number of bits
9498         to 61. Check compatibility.
9499         (profile_count::compatible_p): New private member function.
9500         (profile_count::ipa_p): New member function.
9501         (profile_count::operator<): Handle global zero correctly.
9502         (profile_count::operator>): Handle global zero correctly.
9503         (profile_count::operator<=): Handle global zero correctly.
9504         (profile_count::operator>=): Handle global zero correctly.
9505         (profile_count::nonzero_p): New member function.
9506         (profile_count::force_nonzero): New member function.
9507         (profile_count::max): New member function.
9508         (profile_count::apply_scale): Handle IPA scalling.
9509         (profile_count::guessed_local): New member function.
9510         (profile_count::global0): New member function.
9511         (profile_count::ipa): New member function.
9512         (profile_count::to_frequency): Declare.
9513         (profile_count::to_cgraph_frequency): Declare.
9514         * profile.c (OVERLAP_BASE): Delete.
9515         (compute_frequency_overlap): Delete.
9516         (compute_branch_probabilities): Do not use compute_frequency_overlap.
9517         * regs.h (REG_FREQ_FROM_BB): Use to_frequency.
9518         * sched-ebb.c (rank): Use counts only.
9519         * shrink-wrap.c (handle_simple_exit): Use counts only.
9520         (try_shrink_wrapping): Use counts only.
9521         (place_prologue_for_one_component): Use counts only.
9522         * tracer.c (find_best_predecessor): Use to_frequency.
9523         (find_trace): Use to_frequency.
9524         (tail_duplicate): Use to_frequency.
9525         * trans-mem.c (expand_transaction): Do not update frequency.
9526         * tree-call-cdce.c: Do not update frequency.
9527         * tree-cfg.c (gimple_find_sub_bbs): Likewise.
9528         (gimple_merge_blocks): Likewise.
9529         (gimple_split_edge): Likewise.
9530         (gimple_duplicate_sese_region): Likewise.
9531         (gimple_duplicate_sese_tail): Likewise.
9532         (move_sese_region_to_fn): Likewise.
9533         (gimple_account_profile_record): Likewise.
9534         (insert_cond_bb): Likewise.
9535         * tree-complex.c (expand_complex_div_wide): Likewise.
9536         * tree-eh.c (lower_resx): Update profile.
9537         * tree-inline.c (copy_bb): Simplify count scaling; do not scale
9538         frequencies.
9539         (initialize_cfun): Do not initialize frequencies
9540         (freqs_to_counts): Delete.
9541         (copy_cfg_body): Ignore count parameter.
9542         (copy_body): Update.
9543         (expand_call_inline): Update count_max.
9544         (optimize_inline_calls): Update count_max.
9545         (tree_function_versioning): Update count_max.
9546         * tree-ssa-coalesce.c (coalesce_cost_bb): Use to_frequency.
9547         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Do not update
9548         frequency.
9549         * tree-ssa-loop-im.c (execute_sm_if_changed): Use counts only.
9550         * tree-ssa-loop-ivcanon.c (unloop_loops): Do not update freuqency.
9551         (try_peel_loop): Likewise.
9552         * tree-ssa-loop-ivopts.c (get_scaled_computation_cost_at): Use
9553         to_frequency.
9554         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): Pass -1.
9555         (tree_transform_and_unroll_loop): Do not use frequencies
9556         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations):
9557         Use reliable prediction only.
9558         * tree-ssa-loop-unswitch.c (hoist_guard): Do not use frequencies.
9559         * tree-ssa-sink.c (select_best_block): Use to_frequency.
9560         * tree-ssa-tail-merge.c (replace_block_by): Temporarily disable
9561         probability scaling.
9562         * tree-ssa-threadupdate.c (create_block_for_threading): Do
9563         not update frequency
9564         (any_remaining_duplicated_blocks): Likewise.
9565         (update_profile): Likewise.
9566         (estimated_freqs_path): Delete.
9567         (freqs_to_counts_path): Delete.
9568         (clear_counts_path): Delete.
9569         (ssa_fix_duplicate_block_edges): Likewise.
9570         (duplicate_thread_path): Likewise.
9571         * tree-switch-conversion.c (gen_inbound_check): Use counts.
9572         * tree-tailcall.c (decrease_profile): Do not update frequency.
9573         (eliminate_tail_call): Likewise.
9574         * tree-vect-loop-manip.c (vect_do_peeling): Likewise.
9575         * tree-vect-loop.c (scale_profile_for_vect_loop): Likewise.
9576         (optimize_mask_stores): Likewise.
9577         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
9578         * ubsan.c (ubsan_expand_null_ifn): Update profile.
9579         (ubsan_expand_ptr_ifn): Update profile.
9580         * value-prof.c (gimple_ic): Simplify.
9581         * value-prof.h (gimple_ic): Update prototype.
9582         * ipa-inline-transform.c (inline_transform): Fix scaling conditoins.
9583         * ipa-inline.c (compute_uninlined_call_time): Be sure that
9584         counts are nonzero.
9585         (want_inline_self_recursive_call_p): Likewise.
9586         (resolve_noninline_speculation): Only cummulate defined counts.
9587         (inline_small_functions): Use nonzero_p.
9588         (ipa_inline): Do not access freed node.
9590 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9592         * config/aarch64/aarch64.c (aarch64_override_options_internal):
9593         Set PARAM_SCHED_PRESSURE_ALGORITHM to SCHED_PRESSURE_MODEL.
9595 2017-11-03  Kito Cheng  <kito.cheng@gmail.com>
9597         * config/riscv/riscv.c (riscv_legitimize_move): Handle
9598         non-legitimate address.
9600 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9602         * config/rs6000/rs6000.md (*lt0_disi): Delete.
9603         (*lt0_<mode>di, *lt0_<mode>si): New.
9605 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9607         * config/rs6000/rs6000.md (move_from_CR_ov_bit): Change condition to
9608         TARGET_PAIRED_FLOAT.
9610 2017-11-03  Siddhesh Poyarekar  <siddhesh.poyarekar@linaro.org>
9611             Jim Wilson  <jim.wilson@linaro.org>
9613         * config/aarch64/aarch64-cores.def (saphira): New CPU.
9614         * config/aarch64/aarch64-tune.md: Regenerated.
9615         * doc/invoke.texi (AArch64 Options/-mtune): Add "saphira".
9616         * gcc/config/aarch64/aarch64.c (saphira_tunings): New tuning table.
9618 2017-11-03  Cupertino Miranda  <cmiranda@synopsys.com>
9620         * config/arc/arc.c (arc_save_restore): Corrected CFA note.
9621         (arc_expand_prologue): Restore blink for millicode.
9622         * config/arc/linux.h (LINK_EH_SPEC): Defined.
9624 2017-11-03  Richard Sandiford  <richard.sandiford@linaro.org>
9626         PR target/82809
9627         * config/i386/i386.c (ix86_vector_duplicate_value): Use
9628         gen_vec_duplicate after forcing the scalar into a register.
9630 2017-11-02  Segher Boessenkool  <segher@kernel.crashing.org>
9632         * combine (try_combine): Print the insns input to try_combine to the
9633         dump file.
9635 2017-11-02  Steve Ellcey  <sellcey@cavium.com>
9637         PR target/79868
9638         * config/aarch64/aarch64-c.c (aarch64_pragma_target_parse):
9639         Remove second argument from aarch64_process_target_attr call.
9640         * config/aarch64/aarch64-protos.h (aarch64_process_target_attr):
9641         Ditto.
9642         * config/aarch64/aarch64.c (aarch64_attribute_info): Change
9643         field type.
9644         (aarch64_handle_attr_arch): Remove second argument.
9645         (aarch64_handle_attr_cpu): Ditto.
9646         (aarch64_handle_attr_tune): Ditto.
9647         (aarch64_handle_attr_isa_flags): Ditto.
9648         (aarch64_process_one_target_attr): Ditto.
9649         (aarch64_process_target_attr): Ditto.
9650         (aarch64_option_valid_attribute_p): Remove second argument.
9651         on aarch64_process_target_attr call.
9653 2017-11-02  David Malcolm  <dmalcolm@redhat.com>
9655         * diagnostic.c: Include "selftest-diagnostic.h".
9656         (selftest::assert_location_text): New function.
9657         (selftest::test_diagnostic_get_location_text): New function.
9658         (selftest::diagnostic_c_tests): Call it.
9660 2017-11-02  David Malcolm  <dmalcolm@redhat.com>
9662         * Makefile.in (OBJS-libcommon): Add selftest-diagnostic.o.
9663         * diagnostic-show-locus.c: Include "selftest-diagnostic.h".
9664         (class selftest::test_diagnostic_context): Move to...
9665         * selftest-diagnostic.c: New file.
9666         * selftest-diagnostic.h: New file.
9668 2017-11-02  James Bowman  <james.bowman@ftdichip.com>
9670         * config/ft32/ft32.c (ft32_addr_space_legitimate_address_p): increase
9671         offset range for FT32B.
9672         * config/ft32/ft32.h: option "mcompress" enables relaxation.
9673         * config/ft32/ft32.md: Add TARGET_NOPM.
9674         * config/ft32/ft32.opt: Add mft32b, mcompress, mnopm.
9675         * gcc/doc/invoke.texi: Add mft32b, mcompress, mnopm.
9677 2017-11-02  Wilco Dijkstra  <wdijkstr@arm.com>
9679         * config/aarch64/aarch64.h (MALLOC_ABI_ALIGNMENT): New define.
9681 2017-11-02  Jeff Law  <law@redhat.com>
9683         * gimple-ssa-sprintf.c (sprintf_dom_walker): Remove
9684         virtual keyword on FINAL OVERRIDE members.
9686         * tree-ssa-propagate.h (ssa_propagation_engine): Group
9687         virtuals together.  Add virtual destructor.
9688         (substitute_and_fold_engine): Similarly.
9690 2017-11-02  Jan Hubicka  <hubicka@ucw.cz>
9692         * x86-tune.def (X86_TUNE_USE_INCDEC): Enable for Haswell+.
9694 2017-11-02  Richard Biener  <rguenther@suse.de>
9696         PR tree-optimization/82795
9697         * tree-if-conv.c (predicate_mem_writes): Remove bogus assert.
9699 2017-11-02  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
9701         * acinclude.m4 (gcc_AC_INITFINI_ARRAY): Don't require
9702         gcc_SUN_LD_VERSION.
9703         (gcc_GAS_CHECK_FEATURE): Remove.
9704         * configure.ac (ld_vers) <*-*-solaris2*>: Move comments from
9705         gcc_AC_INITFINI_ARRAY here.  Update for Solaris 11.4 changes.
9706         * configure: Regenerate.
9708 2017-11-02  Claudiu Zissulescu <claziss@synopsys.com>
9710         * config/arc/arc.c (hwloop_optimize): Account for empty
9711         body loops.
9713 2017-11-02  Richard Biener  <rguenther@suse.de>
9715         PR middle-end/82765
9716         * varasm.c (decode_addr_const): Make offset HOST_WIDE_INT.
9717         Truncate ARRAY_REF index and element size.
9719 2017-11-01  Palmer Dabbelt  <palmer@dabbelt.com>
9721         * doc/invoke.texi (RISC-V Options): Use "@minus{}2 GB", not "-2 GB".
9723 2017-11-01  Jeff Law  <law@redhat.com>
9725         * tree-ssa-ccp.c (ccp_folder): New class derived from
9726         substitute_and_fold_engine.
9727         (ccp_folder::get_value): New member function.
9728         (ccp_folder::fold_stmt): Renamed from ccp_fold_stmt.
9729         (ccp_fold_stmt): Remove prototype.
9730         (ccp_finalize): Call substitute_and_fold from the ccp_class.
9731         * tree-ssa-copy.c (copy_folder): New class derived from
9732         substitute_and_fold_engine.
9733         (copy_folder::get_value): Renamed from get_value.
9734         (fini_copy_prop): Call substitute_and_fold from copy_folder class.
9735         * tree-vrp.c (vrp_folder): New class derived from
9736         substitute_and_fold_engine.
9737         (vrp_folder::fold_stmt): Renamed from vrp_fold_stmt.
9738         (vrp_folder::get_value): New member function.
9739         (vrp_finalize): Call substitute_and_fold from vrp_folder class.
9740         (evrp_dom_walker::before_dom_children): Similarly for replace_uses_in.
9741         * tree-ssa-propagate.h (substitute_and_fold_engine): New class to
9742         provide a class interface to folder/substitute routines.
9743         (ssa_prop_fold_stmt_fn): Remove typedef.
9744         (ssa_prop_get_value_fn): Likewise.
9745         (subsitute_and_fold): Remove prototype.
9746         (replace_uses_in): Likewise.
9747         * tree-ssa-propagate.c (substitute_and_fold_engine::replace_uses_in):
9748         Renamed from replace_uses_in.  Call the virtual member function
9749         (substitute_and_fold_engine::replace_phi_args_in): Similarly.
9750         (substitute_and_fold_dom_walker): Remove initialization of
9751         data member entries for calbacks.  Add substitute_and_fold_engine
9752         member and initialize it.
9753         (substitute_and_fold_dom_walker::before_dom_children0: Use the
9754         member functions for get_value, replace_phi_args_in c
9755         replace_uses_in, and fold_stmt calls.
9756         (substitute_and_fold_engine::substitute_and_fold): Renamed from
9757         substitute_and_fold.  Remove assert.   Update ctor call.
9759         * tree-ssa-propagate.h (ssa_prop_visit_stmt_fn): Remove typedef.
9760         (ssa_prop_visit_phi_fn): Likewise.
9761         (class ssa_propagation_engine): New class to provide an interface
9762         into ssa_propagate.
9763         * tree-ssa-propagate.c (ssa_prop_visit_stmt): Remove file scoped
9764         variable.
9765         (ssa_prop_visit_phi): Likewise.
9766         (ssa_propagation_engine::simulate_stmt): Moved into class.
9767         Call visit_phi/visit_stmt from the class rather than via
9768         file scoped static variables.
9769         (ssa_propagation_engine::simulate_block): Moved into class.
9770         (ssa_propagation_engine::process_ssa_edge_worklist): Similarly.
9771         (ssa_propagation_engine::ssa_propagate): Similarly.  No longer
9772         set file scoped statics for the visit_stmt/visit_phi callbacks.
9773         * tree-complex.c (complex_propagate): New class derived from
9774         ssa_propagation_engine.
9775         (complex_propagate::visit_stmt): Renamed from complex_visit_stmt.
9776         (complex_propagate::visit_phi): Renamed from complex_visit_phi.
9777         (tree_lower_complex): Call ssa_propagate via the complex_propagate
9778         class.
9779         * tree-ssa-ccp.c: (ccp_propagate): New class derived from
9780         ssa_propagation_engine.
9781         (ccp_propagate::visit_phi): Renamed from ccp_visit_phi_node.
9782         (ccp_propagate::visit_stmt): Renamed from ccp_visit_stmt.
9783         (do_ssa_ccp): Call ssa_propagate from the ccp_propagate class.
9784         * tree-ssa-copy.c (copy_prop): New class derived from
9785         ssa_propagation_engine.
9786         (copy_prop::visit_stmt): Renamed from copy_prop_visit_stmt.
9787         (copy_prop::visit_phi): Renamed from copy_prop_visit_phi_node.
9788         (execute_copy_prop): Call ssa_propagate from the copy_prop class.
9789         * tree-vrp.c (vrp_prop): New class derived from ssa_propagation_engine.
9790         (vrp_prop::visit_stmt): Renamed from vrp_visit_stmt.
9791         (vrp_prop::visit_phi): Renamed from vrp_visit_phi_node.
9792         (execute_vrp): Call ssa_propagate from the vrp_prop class.
9794 2017-11-01  Jakub Jelinek  <jakub@redhat.com>
9796         PR rtl-optimization/82778
9797         PR rtl-optimization/82597
9798         * compare-elim.c (struct comparison): Add in_a_setter field.
9799         (find_comparison_dom_walker::before_dom_children): Remove killed
9800         bitmap and df_simulate_find_defs call, instead walk the defs.
9801         Compute last_setter and initialize in_a_setter.  Merge definitions
9802         with first initialization for a few variables.
9803         (try_validate_parallel): Use insn_invalid_p instead of
9804         recog_memoized.  Return insn rather than just the pattern.
9805         (try_merge_compare): Fix up comment.  Don't uselessly test if
9806         in_a is a REG_P.  Use cmp->in_a_setter instead of walking UD
9807         chains.
9808         (execute_compare_elim_after_reload): Remove df_chain_add_problem
9809         call.
9811 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9812             Alan Hayward  <alan.hayward@arm.com>
9813             David Sherwood  <david.sherwood@arm.com>
9815         * config/aarch64/aarch64.c (aarch64_rtx_costs): Use
9816         aarch64_hard_regno_nregs to get the number of registers
9817         in a mode.
9819 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9820             Alan Hayward  <alan.hayward@arm.com>
9821             David Sherwood  <david.sherwood@arm.com>
9823         * config/aarch64/constraints.md (Upl): Rename to...
9824         (Uaa): ...this.
9825         * config/aarch64/aarch64.md
9826         (*zero_extend<SHORT:mode><GPI:mode>2_aarch64, *addsi3_aarch64_uxtw):
9827         Update accordingly.
9829 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9830             Alan Hayward  <alan.hayward@arm.com>
9831             David Sherwood  <david.sherwood@arm.com>
9833         * config/aarch64/aarch64.c (aarch64_add_constant_internal)
9834         (aarch64_add_constant, aarch64_add_sp, aarch64_sub_sp): Move
9835         earlier in file.
9837 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9838             Alan Hayward  <alan.hayward@arm.com>
9839             David Sherwood  <david.sherwood@arm.com>
9841         * config/aarch64/aarch64.c (aarch64_evpc_trn, aarch64_evpc_uzp)
9842         (aarch64_evpc_zip, aarch64_evpc_ext, aarch64_evpc_rev)
9843         (aarch64_evpc_dup): Generate rtl direcly, rather than using
9844         named expanders.
9845         (aarch64_expand_vec_perm_const_1): Explicitly check for permutes
9846         of a single element.
9847         * config/aarch64/iterators.md: Add a comment above the permute
9848         unspecs to say that they are generated directly by
9849         aarch64_expand_vec_perm_const.
9850         * config/aarch64/aarch64-simd.md: Likewise the permute instructions.
9852 2017-11-01  Nathan Sidwell  <nathan@acm.org>
9854         * tree-dump.c (dequeue_and_dump): Use HAS_DECL_ASSEMBLER_NAME_P.
9856 2017-11-01  Palmer Dabbelt  <palmer@dabbelt.com>
9858         * doc/invoke.texi (RISC-V Options): Explicitly name the medlow
9859         and medany code models, and describe what they do.
9861 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9863         Revert accidental duplicate:
9865         * combine.c (can_change_dest_mode): Reject changes in
9866         REGMODE_NATURAL_SIZE.
9868 2017-11-01  Segher Boessenkool  <segher@kernel.crashing.org>
9870         PR rtl-optimization/64682
9871         PR rtl-optimization/69567
9872         PR rtl-optimization/69737
9873         PR rtl-optimization/82683
9874         * combine.c (distribute_notes) <REG_DEAD>: If the new I2 sets the same
9875         register mentioned in the note, drop the note, unless it came from I3,
9876         in which case it should go to I3 again.
9878 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9880         * tree-ssa-dse.c (normalize_ref): Check whether the ranges overlap
9881         and return false if not.
9882         (clear_bytes_written_by, live_bytes_read): Update accordingly.
9884 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9886         * tree-ssa-alias.h (ranges_overlap_p): Return false if either
9887         range is known to be empty.
9889 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9890             Alan Hayward  <alan.hayward@arm.com>
9891             David Sherwood  <david.sherwood@arm.com>
9893         * simplify-rtx.c (simplify_const_unary_operation): Use GET_MODE_NUNITS
9894         and CONST_VECTOR_NUNITS instead of computing the number of units from
9895         the byte sizes of the vector and element.
9896         (simplify_binary_operation_1): Likewise.
9897         (simplify_const_binary_operation): Likewise.
9898         (simplify_ternary_operation): Likewise.
9900 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9901             Alan Hayward  <alan.hayward@arm.com>
9902             David Sherwood  <david.sherwood@arm.com>
9904         * var-tracking.c (INT_MEM_OFFSET): Replace with...
9905         (int_mem_offset): ...this new function.
9906         (var_mem_set, var_mem_delete_and_set, var_mem_delete)
9907         (find_mem_expr_in_1pdv, dataflow_set_preserve_mem_locs)
9908         (same_variable_part_p, use_type, add_stores, vt_get_decl_and_offset):
9909         Update accordingly.
9911 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9912             Alan Hayward  <alan.hayward@arm.com>
9913             David Sherwood  <david.sherwood@arm.com>
9915         * lower-subreg.c (interesting_mode_p): New function.
9916         (compute_costs, find_decomposable_subregs, decompose_register)
9917         (simplify_subreg_concatn, can_decompose_p, resolve_simple_move)
9918         (resolve_clobber, dump_choices): Use it.
9920 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9921             Alan Hayward  <alan.hayward@arm.com>
9922             David Sherwood  <david.sherwood@arm.com>
9924         * rtlhash.c (add_rtx): Use add_hwi for 'w' and add_int for 'i'.
9926 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9927             Alan Hayward  <alan.hayward@arm.com>
9928             David Sherwood  <david.sherwood@arm.com>
9930         * alias.c (find_base_value, find_base_term): Only process integer
9931         truncations.  Check the precision rather than the size.
9933 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9934             Alan Hayward  <alan.hayward@arm.com>
9935             David Sherwood  <david.sherwood@arm.com>
9937         * machmode.h (is_narrower_int_mode): New function
9938         * optabs.c (expand_float, expand_fix): Use it.
9939         * dwarf2out.c (rotate_loc_descriptor): Likewise.
9941 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9942             Alan Hayward  <alan.hayward@arm.com>
9943             David Sherwood  <david.sherwood@arm.com>
9945         * rtl.h (narrower_subreg_mode): New function.
9946         * ira-color.c (update_costs_from_allocno): Use it.
9948 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9949             Alan Hayward  <alan.hayward@arm.com>
9950             David Sherwood  <david.sherwood@arm.com>
9952         * optabs-query.h (convert_optab_p): New function, split out from...
9953         (convert_optab_handler): ...here.
9954         (widening_optab_handler): Delete.
9955         (find_widening_optab_handler): Remove permit_non_widening parameter.
9956         (find_widening_optab_handler_and_mode): Likewise.  Provide an
9957         override that operates on mode class wrappers.
9958         * optabs-query.c (widening_optab_handler): Delete.
9959         (find_widening_optab_handler_and_mode): Remove permit_non_widening
9960         parameter.  Assert that the two modes are the same class and that
9961         the "from" mode is narrower than the "to" mode.  Use
9962         convert_optab_handler instead of widening_optab_handler.
9963         * expmed.c (expmed_mult_highpart_optab): Use convert_optab_handler
9964         instead of widening_optab_handler.
9965         * expr.c (expand_expr_real_2): Update calls to
9966         find_widening_optab_handler.
9967         * optabs.c (expand_widen_pattern_expr): Likewise.
9968         (expand_binop_directly): Take the insn_code as a parameter.
9969         (expand_binop): Only call find_widening_optab_handler for
9970         conversion optabs; use optab_handler otherwise.  Update calls
9971         to find_widening_optab_handler and expand_binop_directly.
9972         Use convert_optab_handler instead of widening_optab_handler.
9973         * tree-ssa-math-opts.c (convert_mult_to_widen): Update calls to
9974         find_widening_optab_handler and use scalar_mode rather than
9975         machine_mode.
9976         (convert_plusminus_to_widen): Likewise.
9978 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9979             Alan Hayward  <alan.hayward@arm.com>
9980             David Sherwood  <david.sherwood@arm.com>
9982         * machmode.h (fixed_size_mode): New class.
9983         * rtl.h (get_pool_mode): Return fixed_size_mode.
9984         * gengtype.c (main): Add fixed_size_mode.
9985         * target.def (get_raw_result_mode): Return a fixed_size_mode.
9986         (get_raw_arg_mode): Likewise.
9987         * doc/tm.texi: Regenerate.
9988         * targhooks.h (default_get_reg_raw_mode): Return a fixed_size_mode.
9989         * targhooks.c (default_get_reg_raw_mode): Likewise.
9990         * config/ia64/ia64.c (ia64_get_reg_raw_mode): Likewise.
9991         * config/mips/mips.c (mips_get_reg_raw_mode): Likewise.
9992         * config/msp430/msp430.c (msp430_get_raw_arg_mode): Likewise.
9993         (msp430_get_raw_result_mode): Likewise.
9994         * config/avr/avr-protos.h (regmask): Use as_a <fixed_side_mode>
9995         * dbxout.c (dbxout_parms): Require fixed-size modes.
9996         * expr.c (copy_blkmode_from_reg, copy_blkmode_to_reg): Likewise.
9997         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Likewise.
9998         * omp-low.c (lower_oacc_reductions): Likewise.
9999         * simplify-rtx.c (simplify_immed_subreg): Take fixed_size_modes.
10000         (simplify_subreg): Update accordingly.
10001         * varasm.c (constant_descriptor_rtx::mode): Change to fixed_size_mode.
10002         (force_const_mem): Update accordingly.  Return NULL_RTX for modes
10003         that aren't fixed-size.
10004         (get_pool_mode): Return a fixed_size_mode.
10005         (output_constant_pool_2): Take a fixed_size_mode.
10007 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
10008             Alan Hayward  <alan.hayward@arm.com>
10009             David Sherwood  <david.sherwood@arm.com>
10011         * doc/rtl.texi (vec_series): Document.
10012         (const): Say that the operand can be a vec_series.
10013         * rtl.def (VEC_SERIES): New rtx code.
10014         * rtl.h (const_vec_series_p_1): Declare.
10015         (const_vec_series_p): New function.
10016         * emit-rtl.h (gen_const_vec_series): Declare.
10017         (gen_vec_series): Likewise.
10018         * emit-rtl.c (const_vec_series_p_1, gen_const_vec_series)
10019         (gen_vec_series): Likewise.
10020         * optabs.c (expand_mult_highpart): Use gen_const_vec_series.
10021         * simplify-rtx.c (simplify_unary_operation): Handle negations
10022         of vector series.
10023         (simplify_binary_operation_series): New function.
10024         (simplify_binary_operation_1): Use it.  Handle VEC_SERIES.
10025         (test_vector_ops_series): New function.
10026         (test_vector_ops): Call it.
10027         * config/powerpcspe/altivec.md (altivec_lvsl): Use
10028         gen_const_vec_series.
10029         (altivec_lvsr): Likewise.
10030         * config/rs6000/altivec.md (altivec_lvsl, altivec_lvsr): Likewise.
10032 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
10033             Alan Hayward  <alan.hayward@arm.com>
10034             David Sherwood  <david.sherwood@arm.com>
10036         * doc/rtl.texi (const): Update description of address constants.
10037         Say that vector constants are allowed too.
10038         * common.md (E, F): Use CONSTANT_P instead of checking for
10039         CONST_VECTOR.
10040         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
10041         checking for CONST_VECTOR.
10042         * expmed.c (make_tree): Use build_vector_from_val for a CONST
10043         VEC_DUPLICATE.
10044         * expr.c (expand_expr_real_2): Check for vector modes instead
10045         of checking for CONST_VECTOR.
10046         * rtl.h (const_vec_p): New function.
10047         (const_vec_duplicate_p): Check for a CONST VEC_DUPLICATE.
10048         (unwrap_const_vec_duplicate): Handle them here too.
10050 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
10051             David Malcolm  <dmalcolm@redhat.com>
10052             Alan Hayward  <alan.hayward@arm.com>
10053             David Sherwood  <david.sherwood@arm.com>
10055         * rtl.h (vec_duplicate_p): New function.
10056         * selftest-rtl.c (assert_rtx_eq_at): New function.
10057         * selftest-rtl.h (ASSERT_RTX_EQ): New macro.
10058         (assert_rtx_eq_at): Declare.
10059         * selftest.h (selftest::simplify_rtx_c_tests): Declare.
10060         * selftest-run-tests.c (selftest::run_tests): Call it.
10061         * simplify-rtx.c: Include selftest.h and selftest-rtl.h.
10062         (simplify_unary_operation_1): Recursively handle vector duplicates.
10063         (simplify_binary_operation_1): Likewise.  Handle VEC_SELECTs of
10064         vector duplicates.
10065         (simplify_subreg): Handle subregs of vector duplicates.
10066         (make_test_reg, test_vector_ops_duplicate, test_vector_ops)
10067         (selftest::simplify_rtx_c_tests): New functions.
10069 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
10070             Alan Hayward  <alan.hayward@arm.com>
10071             David Sherwood  <david.sherwood@arm.com>
10073         * emit-rtl.h (gen_const_vec_duplicate): Declare.
10074         (gen_vec_duplicate): Likewise.
10075         * emit-rtl.c (gen_const_vec_duplicate_1): New function, split
10076         out from...
10077         (gen_const_vector): ...here.
10078         (gen_const_vec_duplicate, gen_vec_duplicate): New functions.
10079         (gen_rtx_CONST_VECTOR): Use gen_const_vec_duplicate for constants
10080         whose elements are all equal.
10081         * optabs.c (expand_vector_broadcast): Use gen_const_vec_duplicate.
10082         * simplify-rtx.c (simplify_const_unary_operation): Likewise.
10083         (simplify_relational_operation): Likewise.
10084         * config/aarch64/aarch64.c (aarch64_simd_gen_const_vector_dup):
10085         Likewise.
10086         (aarch64_simd_dup_constant): Use gen_vec_duplicate.
10087         (aarch64_expand_vector_init): Likewise.
10088         * config/arm/arm.c (neon_vdup_constant): Likewise.
10089         (neon_expand_vector_init): Likewise.
10090         (arm_expand_vec_perm): Use gen_const_vec_duplicate.
10091         (arm_block_set_unaligned_vect): Likewise.
10092         (arm_block_set_aligned_vect): Likewise.
10093         * config/arm/neon.md (neon_copysignf<mode>): Likewise.
10094         * config/i386/i386.c (ix86_expand_vec_perm): Likewise.
10095         (expand_vec_perm_even_odd_pack): Likewise.
10096         (ix86_vector_duplicate_value): Use gen_vec_duplicate.
10097         * config/i386/sse.md (one_cmpl<mode>2): Use CONSTM1_RTX.
10098         * config/ia64/ia64.c (ia64_expand_vecint_compare): Use
10099         gen_const_vec_duplicate.
10100         * config/ia64/vect.md (addv2sf3, subv2sf3): Use CONST1_RTX.
10101         * config/mips/mips.c (mips_gen_const_int_vector): Use
10102         gen_const_vec_duplicate.
10103         (mips_expand_vector_init): Use CONST0_RTX.
10104         * config/powerpcspe/altivec.md (abs<mode>2, nabs<mode>2): Likewise.
10105         (define_split): Use gen_const_vec_duplicate.
10106         * config/rs6000/altivec.md (abs<mode>2, nabs<mode>2): Use CONST0_RTX.
10107         (define_split): Use gen_const_vec_duplicate.
10108         * config/s390/vx-builtins.md (vec_genmask<mode>): Likewise.
10109         (vec_ctd_s64, vec_ctd_u64, vec_ctsl, vec_ctul): Likewise.
10110         * config/spu/spu.c (spu_const): Likewise.
10112 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
10113             Alan Hayward  <alan.hayward@arm.com>
10114             David Sherwood  <david.sherwood@arm.com>
10116         * combine.c (can_change_dest_mode): Reject changes in
10117         REGMODE_NATURAL_SIZE.
10119 2017-10-31  Sandra Loosemore  <sandra@codesourcery.com>
10121         * configure.ac (--enable-libssp): New.
10122         (gcc_cv_libc_provides_ssp): Check for explicit setting before
10123         trying to determine target-specific default.  Adjust indentation.
10124         * configure: Regenerated.
10125         * doc/install.texi (Configuration): Expand --disable-libssp
10126         documentation.
10128 2017-10-31  Daniel Santos  <daniel.santos@pobox.com>
10130         config/i386/i386.c (ix86_expand_epilogue): Correct stack
10131         calculation.
10133 2017-10-31  Martin Jambor  <mjambor@suse.cz>
10135         PR c++/81702
10136         * gimple-fold.c (gimple_get_virt_method_for_vtable): Remove assert.
10138 2017-10-31  David Malcolm  <dmalcolm@redhat.com>
10140         * auto-profile.c (autofdo_source_profile::read): Use
10141         UNKNOWN_LOCATION rather than 0.
10142         * diagnostic-core.h (warning_at_rich_loc): Rename to...
10143         (warning_at): ...this overload.
10144         (warning_at_rich_loc_n): Rename to...
10145         (warning_n): ...this overload.
10146         (error_at_rich_loc): Rename to...
10147         (error_at): ...this overload.
10148         (pedwarn_at_rich_loc): Rename to...
10149         (pedwarn): ...this overload.
10150         (permerror_at_rich_loc): Rename to...
10151         (permerror): ...this overload.
10152         (inform_at_rich_loc): Rename to...
10153         (inform): ...this overload.
10154         * diagnostic.c: (diagnostic_n_impl): Delete location_t-based decl.
10155         (diagnostic_n_impl_richloc): Rename to...
10156         (diagnostic_n_impl): ...this rich_location *-based decl.
10157         (inform_at_rich_loc): Rename to...
10158         (inform): ...this, and add an assertion.
10159         (inform_n): Update for removal of location_t-based diagnostic_n_impl.
10160         (warning_at_rich_loc): Rename to...
10161         (warning_at): ...this, and add an assertion.
10162         (warning_at_rich_loc_n): Rename to...
10163         (warning_n): ...this, and add an assertion.
10164         (warning_n): Update location_t-based implementation for removal of
10165         location_t-based diagnostic_n_impl.
10166         (pedwarn_at_rich_loc): Rename to...
10167         (pedwarn): ...this, and add an assertion.
10168         (permerror_at_rich_loc): Rename to...
10169         (permerror): ...this, and add an assertion.
10170         (error_n): Update for removal of location_t-based diagnostic_n_impl.
10171         (error_at_rich_loc): Rename to...
10172         (error_at): ...this, and add an assertion.
10173         * gcc.c (do_spec_1): Use UNKNOWN_LOCATION rather than 0.
10174         (driver::do_spec_on_infiles): Likewise.
10175         * substring-locations.c (format_warning_va): Update for renaming
10176         of inform_at_rich_loc.
10178 2017-10-31  Michael Meissner  <meissner@linux.vnet.ibm.com>
10180         * builtins.def (DEF_FLOATN_BUILTIN): Change most _Float<N> and
10181         _Float<N>X built-in functions so that the variant without the
10182         "__builtin_" prefix is only enabled for the GNU C and Objective C
10183         languages when they are in non-strict ANSI/ISO mode.
10184         (DEF_EXT_LIB_FLOATN_NX_BUILTINS): Likewise.
10185         * target.def (floatn_builtin_p): Add a target hook to control
10186         whether _Float<N> and _Float<N>X built-in functions without the
10187         "__builtin_" prefix are enabled, and return true for C and
10188         Objective C in the default hook.  Include langhooks.h in
10189         targhooks.c.
10190         * targhooks.h (default_floatn_builtin_p): Likewise.
10191         * targhooks.c (default_floatn_builtin_p): Likewise.
10192         * doc/tm.texi.in (TARGET_FLOATN_BUILTIN_P): Document the
10193         floatn_builtin_p target hook.
10194         * doc/tm.texi (TARGET_FLOATN_BUILTIN_P): Likewise.
10196 2017-10-31  Matthew Fortune  <matthew.fortune@imgtec.com>
10197             Eric Botcazou  <ebotcazou@adacore.com>
10199         PR rtl-optimization/81803
10200         * lra-constraints.c (curr_insn_transform): Also reload the whole
10201         register for a strict subreg no wider than a word if this is for
10202         a WORD_REGISTER_OPERATIONS target.
10204 2017-10-31  Jason Merrill  <jason@redhat.com>
10206         * gdbinit.in: Skip over inlines from timevar.h.
10208 2017-10-31  Martin Liska  <mliska@suse.cz>
10210         * doc/gcov.texi: Document new option.
10211         * gcov.c (print_usage): Likewise print it.
10212         (process_args): Support the argument.
10213         (format_count): New function.
10214         (format_gcov): Use the function.
10216 2017-10-31  Martin Liska  <mliska@suse.cz>
10218         * gcov.c (struct name_map): do not use typedef.
10219         Define operator== and operator<.
10220         (name_search): Remove.
10221         (name_sort): Remove.
10222         (main): Do not allocate names.
10223         (process_file): Add vertical space.
10224         (generate_results): Use std::find.
10225         (release_structures): Do not release memory.
10226         (find_source): Use std::find.
10228 2017-10-31  Martin Liska  <mliska@suse.cz>
10230         * gcov.c (struct line_info): Remove it's typedef.
10231         (line_info::line_info): Add proper ctor.
10232         (line_info::has_block): Do not use a typedef.
10233         (struct source_info): Do not use typedef.
10234         (circuit): Likewise.
10235         (get_cycles_count): Likewise.
10236         (output_intermediate_file): Iterate via vector iterator.
10237         (add_line_counts): Use std::vector methods.
10238         (accumulate_line_counts): Likewise.
10239         (output_lines): Likewise.
10241 2017-10-31  Martin Liska  <mliska@suse.cz>
10243         * gcov.c (struct source_info): Remove typedef.
10244         (source_info::source_info): Add proper ctor.
10245         (accumulate_line_counts): Use struct, not it's typedef.
10246         (output_gcov_file): Likewise.
10247         (output_lines): Likewise.
10248         (main): Do not allocate an array.
10249         (output_intermediate_file): Use size of vector container.
10250         (process_file): Resize the vector.
10251         (generate_results): Do not preallocate, use newly added vector
10252         lines.
10253         (release_structures): Do not release sources.
10254         (find_source): Use vector methods.
10255         (add_line_counts): Do not use typedef.
10257 2017-10-31  Martin Liska  <mliska@suse.cz>
10259         * doc/gcov.texi: Document that.
10260         * gcov.c (add_line_counts): Mark lines with a non-executed
10261         statement.
10262         (output_line_beginning): Handle such lines.
10263         (output_lines): Pass new argument.
10264         (output_intermediate_file): Print it in intermediate format.
10266 2017-10-31  Martin Liska  <mliska@suse.cz>
10268         * color-macros.h: New file.
10269         * diagnostic-color.c: Factor out color related to macros to
10270         color-macros.h.
10271         * doc/gcov.texi: Document -k option.
10272         * gcov.c (INCLUDE_STRING): Include string.h.
10273         (print_usage): Add -k option.
10274         (process_args): Parse it.
10275         (pad_count_string): New function.
10276         (output_line_beginning): Likewise.
10277         (DEFAULT_LINE_START): New macro.
10278         (output_lines): Support color output.
10280 2017-10-31  Martin Liska  <mliska@suse.cz>
10282         PR gcov-profile/82633
10283         * doc/gcov.texi: Document -fkeep-{static,inline}-functions and
10284         their interaction with GCOV infrastructure.
10285         * configure.ac: Add -fkeep-{inline,static}-functions to
10286         coverage_flags.
10287         * configure: Regenerate.
10289 2017-10-31  Uros Bizjak  <ubizjak@gmail.com>
10291         PR target/82772
10292         * config/alpha/sync.md (fetchop_constr) <and>: Change to "rINM".
10294 2017-10-31  Segher Boessenkool  <segher@kernel.crashing.org>
10296         PR target/82674
10297         * config/rs6000/rs6000.md (allocate_stack): Force update interval
10298         into a register if it does not fit into an immediate offset field.
10300 2017-10-31  Olivier Hainque  <hainque@adacore.com>
10302         * gcc/Makefile.in (FLAGS_TO_PASS): Pass libsubdir as well.
10304 2017-10-31  Julia Koval  <julia.koval@intel.com>
10306         * config.gcc: Add gfniintrin.h.
10307         * config/i386/gfniintrin.h: New.
10308         * config/i386/i386-builtin-types.def
10309         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
10310         __builtin_ia32_vgf2p8affineinvqb_v64qi_mask,
10311         __builtin_ia32_vgf2p8affineinvqb_v32qi,
10312         __builtin_ia32_vgf2p8affineinvqb_v32qi_mask,
10313         __builtin_ia32_vgf2p8affineinvqb_v16qi,
10314         __builtin_ia32_vgf2p8affineinvqb_v16qi_mask): New builtins.
10315         * config/i386/i386-builtin.def (V64QI_FTYPE_V64QI_V64QI_INT_V64QI_UDI,
10316         V32QI_FTYPE_V32QI_V32QI_INT_V32QI_USI,
10317         V16QI_FTYPE_V16QI_V16QI_INT_V16QI_UHI,
10318         V64QI_FTYPE_V64QI_V64QI_INT): New types.
10319         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
10320         * config/i386/immintrin.h: Include gfniintrin.h.
10321         * config/i386/sse.md (vgf2p8affineinvqb_*) New pattern.
10323 2017-10-30  Eric Botcazou  <ebotcazou@adacore.com>
10325         * gcc.c (HAVE_TARGET_EXECUTABLE_SUFFIX): Remove old kludge.
10327 2017-10-30  Wilco Dijkstra  <wdijkstr@arm.com>
10329         * config/arm/arm.md (ashldi3): Remove shift by 1 expansion.
10330         (arm_ashldi3_1bit): Remove pattern.
10331         (ashrdi3): Remove shift by 1 expansion.
10332         (arm_ashrdi3_1bit): Remove pattern.
10333         (lshrdi3): Remove shift by 1 expansion.
10334         (arm_lshrdi3_1bit): Remove pattern.
10335         * config/arm/arm.c (arm_rtx_costs_internal): Slightly increase
10336         cost of ashldi3 by 1.
10337         * config/arm/neon.md (ashldi3_neon): Remove shift by 1 expansion.
10338         (<shift>di3_neon): Likewise.
10340 2017-10-30  Dominik Infuehr  <dominik.infuehr@theobroma-systems.com>
10342         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov): Rename
10343         both identically named patterns to (*aarch64_simd_mov<VD:mode>)
10344         and (*aarch64_simd_mov<VQ:mode>).
10345         (*aarch64_simd_mov<VD:mode>): Change type attribute to match
10346         pattern alternative.
10347         (*aarch64_simd_mov<VQ:mode>): Re-order and change type
10348         attributes to match pattern alternative.
10350 2017-10-30  Steven Munroe  <munroesj@gcc.gnu.org>
10352         * config.gcc (powerpc*-*-*): Add emmintrin.h.
10353         * config/rs6000/emmintrin.h: New file.
10354         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include emmintrin.h.
10356 2017-10-30  Wilco Dijkstra  <wdijkstr@arm.com>
10358         * config/arm/vfp.md (movdi_vfp): Merge changes from movdi_vfp_cortexa8.
10359         * (movdi_vfp_cortexa8): Remove pattern.
10361 2017-10-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10363         * doc/install.texi (Specific, alpha*-*-*): Remove DEC OSF/1
10364         etc. reference.
10365         (Specific, alpha*-dec-osf5.1): Remove.
10366         (Specific, mips-sgi-irix5): Remove.
10367         (Specific, mips-sgi-irix6): Remove.
10369 2017-10-30  Jakub Jelinek  <jakub@redhat.com>
10371         PR middle-end/22141
10372         * gimple-ssa-store-merging.c (merged_store_group::apply_stores): Fix
10373         arguments to clear_bit_region_be.
10375 2017-10-30  Jim Wilson  <wilson@tuliptree.org>
10377         * gimplify.c: Include memmodel.h.
10379 2017-10-30  Martin Jambor  <mjambor@suse.cz>
10381         * omp-grid.c (grid_attempt_target_gridification): Also insert a
10382         condition whether loop should be executed at all.
10384 2017-10-30  Will Schmidt  <will_schmidt@vnet.ibm.com>
10386         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
10387         gimple folding of vec_madd() intrinsics.
10388         * config/rs6000/altivec.md (mulv8hi3): Rename altivec_vmladduhm to
10389         fmav8hi4.  (altivec_vmladduhm): Rename to fmav8hi4.
10390         * config/rs6000/rs6000-builtin.def: Rename vmladduhm to fmav8hi4
10392 2017-10-30  Richard Biener  <rguenther@suse.de>
10394         PR tree-optimization/82762
10395         Revert
10396         2017-10-23  Richard Biener  <rguenther@suse.de>
10398         PR tree-optimization/82129
10399         Revert
10400         2017-08-01  Richard Biener  <rguenther@suse.de>
10402         PR tree-optimization/81181
10403         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
10404         (compute_antic): ... end of iteration here.
10406 2017-10-30  Joseph Myers  <joseph@codesourcery.com>
10408         * doc/invoke.texi (C Dialect Options): Document -std=c17,
10409         -std=iso9899:2017 and -std=gnu17.
10410         * doc/standards.texi (C Language): Document C17 support.
10411         * doc/cpp.texi (Overview): Mention -std=c17.
10412         (Standard Predefined Macros): Document C11 and C17 values of
10413         __STDC_VERSION__.  Do not refer to C99 support as incomplete.
10414         * doc/extend.texi (Inline): Do not list individual options for
10415         standards newer than C99.
10416         * dwarf2out.c (highest_c_language, gen_compile_unit_die): Handle
10417         "GNU C17".
10418         * config/rl78/rl78.c (rl78_option_override): Handle "GNU C17"
10419         language name.
10421 2017-10-30  Maxim Ostapenko  <m.ostapenko@samsung.com>
10423         * asan.c (asan_finish_file): Align asan globals array by shadow
10424         granularity.
10426 2017-10-30  Jakub Jelinek  <jakub@redhat.com>
10428         PR middle-end/22141
10429         * gimple-ssa-store-merging.c: Include rtl.h and expr.h.
10430         (struct store_immediate_info): Add bitregion_start and bitregion_end
10431         fields.
10432         (store_immediate_info::store_immediate_info): Add brs and bre
10433         arguments and initialize bitregion_{start,end} from those.
10434         (struct merged_store_group): Add bitregion_start, bitregion_end,
10435         align_base and mask fields.  Drop unnecessary struct keyword from
10436         struct store_immediate_info.  Add do_merge method.
10437         (clear_bit_region_be): Use memset instead of loop storing zeros.
10438         (merged_store_group::do_merge): New method.
10439         (merged_store_group::merge_into): Use do_merge.  Allow gaps in between
10440         stores as long as the surrounding bitregions have no gaps.
10441         (merged_store_group::merge_overlapping): Use do_merge.
10442         (merged_store_group::apply_stores): Test that bitregion_{start,end}
10443         is byte aligned, rather than requiring that start and width are
10444         byte aligned.  Drop unnecessary struct keyword from
10445         struct store_immediate_info.  Allocate and populate also mask array.
10446         Make start of the arrays relative to bitregion_start rather than
10447         start and size them according to bitregion_{end,start} difference.
10448         (struct imm_store_chain_info): Drop unnecessary struct keyword from
10449         struct store_immediate_info.
10450         (pass_store_merging::gate): Punt if BITS_PER_UNIT or CHAR_BIT is not 8.
10451         (pass_store_merging::terminate_all_aliasing_chains): Drop unnecessary
10452         struct keyword from struct store_immediate_info.
10453         (imm_store_chain_info::coalesce_immediate_stores): Allow gaps in
10454         between stores as long as the surrounding bitregions have no gaps.
10455         Formatting fixes.
10456         (struct split_store): Add orig non-static data member.
10457         (split_store::split_store): Initialize orig to false.
10458         (find_constituent_stmts): Return store_immediate_info *, non-NULL
10459         if there is exactly a single original stmt.  Change stmts argument
10460         to pointer from reference, if NULL, don't push anything to it.  Add
10461         first argument, use it to optimize skipping over orig stmts that
10462         are known to be before bitpos already.  Simplify.
10463         (split_group): Return unsigned int count how many stores are or
10464         would be needed rather than a bool.  Add allow_unaligned argument.
10465         Change split_stores argument from reference to pointer, if NULL,
10466         only do a dry run computing how many stores would be produced.
10467         Rewritten algorithm to use both alignment and misalign if
10468         !allow_unaligned and handle bitfield stores with gaps.
10469         (imm_store_chain_info::output_merged_store): Set start_byte_pos
10470         from bitregion_start instead of start.  Compute allow_unaligned
10471         here, if true, do 2 split_group dry runs to compute which one
10472         produces fewer stores and prefer aligned if equal.  Punt if
10473         new count is bigger or equal than original before emitting any
10474         statements, rather than during that.  Remove no longer needed
10475         new_ssa_names tracking.  Replace num_stmts with
10476         split_stores.length ().  Use 32-bit stack allocated entries
10477         in split_stores auto_vec.  Try to reuse original store lhs/rhs1
10478         if possible.  Handle bitfields with gaps.
10479         (pass_store_merging::execute): Ignore bitsize == 0 stores.
10480         Compute bitregion_{start,end} for the stores and construct
10481         store_immediate_info with that.  Formatting fixes.
10483 2017-10-30  Uros Bizjak  <ubizjak@gmail.com>
10485         PR target/82725
10486         * config/i386/i386.c (legitimate_pic_address_disp_p): Allow
10487         UNSPEC_DTPOFF and UNSPEC_NTPOFF with SImode immediate offset.
10489 2017-10-29  Jim Wilson  <wilson@tuliptree.org>
10491         * gimplify.c: Include tm_p.h.
10493         * common.opt (gcoff): Re-add as ignored option.
10494         (gcoff1, gcoff2, gcoff3): Likewise.
10496         * Makefile.in (OBJS): Delete sdbout.o.
10497         (GTFILES): Delete $(srcdir)/sdbout.c.
10498         * debug.h: Delete sdb_debug_hooks.
10499         * final.c: Delete sdbout.h include.
10500         (final_scan_insn): Delete SDB_DEBUG check.
10501         (rest_of_clean_state): Likewise.
10502         * output.h: Delete sdb_begin_function_line.
10503         * sdbout.c: Delete.
10504         * sdbout.h: Delete.
10505         * toplev.c: Delete sdbout.h include.
10506         (process_options): Delete SDB_DEBUG check.
10507         * tree-core.h (tree_type_common): Delete pointer field of
10508         tree_type_symtab.
10509         * tree.c (copy_node): Clear TYPE_SYMTAB_DIE instead of
10510         TYPE_SYMTAB_POINTER.
10511         * tree.h (TYPE_SYMTAB_POINTER): Delete.
10512         (TYPE_SYMTAB_IS_POINTER): Delete.
10513         (TYPE_SYMTAB_IS_DIE): Renumber.
10514         * xcoffout.c: Refer to former sdbout.c file.
10515         (xcoffout_begin_prologue): Use past tense for sdbout.c reference.
10517         * doc/install.texi (--with-stabs): Delete COFF and ECOFF info.
10518         * doc/invoke.texi (SEEALSO): Delete adb and sdb references.
10519         (Debugging Options): Delete -gcoff.
10520         (-gstabs): Delete SDB reference.
10521         (-gcoff): Delete.
10522         (-gcoff@var{level}): Delete.
10523         * doc/passes.texi (Debugging information output): Delete SDB and
10524         sdbout.c references.
10525         * doc/tm.texi: Regenerate.
10526         * doc/tm.texi.in (DWARF_CIE_DATA_ALIGNMENT): Delete SDB from xref.
10527         (SDB and DWARF): Change node name to DWARF and delete SDB and COFF
10528         references.
10529         (DEBUGGER_AUTO_OFFSET): Delete COFF and SDB references.
10530         (PREFERRED_DEBUGGING_TYPE): Delete SDB_DEBUG and -gcoff references.
10531         (SDB_DEBUGGING_INFO): Delete.
10532         (PUT_SDB_@dots{}, SDB_DELIM, SDB_ALLOW_UNKNOWN_REFERENCES)
10533         SDB_ALLOW_FORWARD_REFERENCES, SDB_OUTPUT_SOURCE_LINE): Delete.
10534         * target.def (output_source_filename): Delete COFF reference.
10536         * common.opt (gcoff): Delete.
10537         (gxcoff+): Update Negative chain.
10538         * defaults.h: Delete all references to SDB_DEBUGGING_INFO and
10539         SDB_DEBUG.
10540         * dwarf2out.c (gen_array_type_die): Change SDB to debuggers.
10541         * flag-types.h (enum debug_info_type): Delete SDB_DEBUG.
10542         * function.c (number_blocks): Delete SDB_DEBUGGING_INFO, SDB_DEBUG,
10543         and SDB references.
10544         (expand_function_start): Change sdb reference to past tense.
10545         (expand_function_end): Change sdb reference to past tense.
10546         * gcc.c (cpp_unique_options): Delete gcoff3 reference.
10547         * opts.c (debug_type_names): Delete coff entry.
10548         (common_handle_option): Delete OPT_gcoff case.
10549         * system.h (SDB_DEBUG, SDB_DEBUGGING_INFO): Poison.
10551         * config/dbxcoff.h (PREFERRED_DEBUGGING_TYPE): Set to DBX_DEBUG.
10552         * config/cris/cris.h: Delete SDB reference in comment.
10553         * config/i386/cygming.h: Don't define SDB_DEBUGGING_INFO.
10554         (ASM_DECLARE_FUNCTION_NAME): Delete SDB reference from comment.
10555         * config/i386/gas.h: Don't define SDB_DEBUGGING_INFO.
10556         * config/i386/i386.c (svr4_dbx_register_map): Change SDB references
10557         to past tense.
10558         (ix86_expand_prologue): Likewise.
10559         * config/i386/winnt.c (i386_pe_start_function): Don't check SDB_DEBUG.
10560         * config/ia64/ia64.h: Likewise.
10561         * config/m68k/m68kelf.h (DBX_REGISTER_NUMBER): Delete SDB reference.
10562         * config/mips/mips.h (SUBTARGET_ASM_DEBUGGING_SPEC): Delete gcoff*
10563         support.
10564         * config/mmix/mmix.h: Likewise.
10565         * config/nds32/nds32.c: Likewise.
10566         * config/stormy/storym16.h: Likewise.
10567         * config/visium/visium.h: Likewise.
10568         * config/vx-common.h (SDB_DEBUGGING_INFO): Delete undef.
10570 2017-10-28  Sandra Loosemore  <sandra@codesourcery.com>
10572         * config/nios2/nios2.h (FRAME_GROWS_DOWNWARD): Define to 1.
10573         * config/nios2/nios2.c (nios2_initial_elimination_offset):  Make
10574         FRAME_POINTER_REGNUM point at high end of local var area.
10576 2017-10-27  Eric Botcazou  <ebotcazou@adacore.com>
10578         * bb-reorder.c (find_traces_1_round): Fix off-by-one index.
10579         Move comment around.  Do not reset best_edge for a copiable
10580         destination if the copy would cause a partition change.
10581         (better_edge_p): Remove redundant check.
10583 2017-10-27  Uros Bizjak  <ubizjak@gmail.com>
10585         * config/i386/i386-protos.h (ix86_fp_compare_mode): Remove prototype.
10587 2017-10-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
10589         * builtins.c (CASE_MATHFN_FLOATN): New helper macro to add cases
10590         for math functions that have _Float<N> and _Float<N>X variants.
10591         (mathfn_built_in_2): Add support for math functions that have
10592         _Float<N> and _Float<N>X variants.
10593         (DEF_INTERNAL_FLT_FLOATN_FN): New helper macro.
10594         (expand_builtin_mathfn_ternary): Add support for fma with
10595         _Float<N> and _Float<N>X variants.
10596         (expand_builtin): Likewise.
10597         (fold_builtin_3): Likewise.
10598         * builtins.def (DEF_EXT_LIB_FLOATN_NX_BUILTINS): New macro to
10599         create math function _Float<N> and _Float<N>X variants as external
10600         library builtins.
10601         (BUILT_IN_COPYSIGN _Float<N> and _Float<N>X variants) Use
10602         DEF_EXT_LIB_FLOATN_NX_BUILTINS to make built-in functions using
10603         the __builtin_ prefix and if not strict ansi, without the prefix.
10604         (BUILT_IN_FABS _Float<N> and _Float<N>X variants): Likewise.
10605         (BUILT_IN_FMA _Float<N> and _Float<N>X variants): Likewise.
10606         (BUILT_IN_FMAX _Float<N> and _Float<N>X variants): Likewise.
10607         (BUILT_IN_FMIN _Float<N> and _Float<N>X variants): Likewise.
10608         (BUILT_IN_NAN _Float<N> and _Float<N>X variants): Likewise.
10609         (BUILT_IN_SQRT _Float<N> and _Float<N>X variants): Likewise.
10610         * builtin-types.def (BT_FN_FLOAT16_FLOAT16_FLOAT16_FLOAT16): New
10611         function signatures for fma _Float<N> and _Float<N>X variants.
10612         (BT_FN_FLOAT32_FLOAT32_FLOAT32_FLOAT32): Likewise.
10613         (BT_FN_FLOAT64_FLOAT64_FLOAT64_FLOAT64): Likewise.
10614         (BT_FN_FLOAT128_FLOAT128_FLOAT128_FLOAT128): Likewise.
10615         (BT_FN_FLOAT32X_FLOAT32X_FLOAT32X_FLOAT32X): Likewise.
10616         (BT_FN_FLOAT64X_FLOAT64X_FLOAT64X_FLOAT64X): Likewise.
10617         (BT_FN_FLOAT128X_FLOAT128X_FLOAT128X_FLOAT128X): Likewise.
10618         * gencfn-macros.c (print_case_cfn): Add support for math functions
10619         that have _Float<N> and _Float<N>X variants.
10620         (print_define_operator_list): Likewise.
10621         (fltfn_suffixes): Likewise.
10622         (main): Likewise.
10623         * internal-fn.def (DEF_INTERNAL_FLT_FLOATN_FN): New helper macro
10624         for math functions that have _Float<N> and _Float<N>X variants.
10625         (SQRT): Add support for sqrt, copysign, fmin and fmax _Float<N>
10626         and _Float<N>X variants.
10627         (COPYSIGN): Likewise.
10628         (FMIN): Likewise.
10629         (FMAX): Likewise.
10630         * fold-const.c (tree_call_nonnegative_warnv_p): Add support for
10631         copysign, fma, fmax, fmin, and sqrt _Float<N> and _Float<N>X
10632         variants.
10633         (integer_valued_read_call_p): Likewise.
10634         * fold-const-call.c (fold_const_call_ss): Likewise.
10635         (fold_const_call_sss): Add support for copysign, fmin, and fmax
10636         _Float<N> and _Float<N>X variants.
10637         (fold_const_call_ssss): Add support for fma _Float<N> and
10638         _Float<N>X variants.
10639         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Add
10640         support for copysign and fma _Float<N> and _Float<N>X variants.
10641         (backprop::process_builtin_call_use): Likewise.
10642         * tree-call-cdce.c (can_test_argument_range); Add support for
10643         sqrt _Float<N> and _Float<N>X variants.
10644         (edom_only_function): Likewise.
10645         (get_no_error_domain): Likewise.
10646         * tree-ssa-math-opts.c (internal_fn_reciprocal): Likewise.
10647         * tree-ssa-reassoc.c (attempt_builtin_copysign): Add support for
10648         copysign _Float<N> and _Float<N>X variants.
10649         * config/rs6000/rs6000-builtin.def (SQRTF128): Delete, this is now
10650         handled by machine independent code.
10651         (FMAF128): Likewise.
10652         * doc/cpp.texi (Common Predefined Macros): Document defining
10653         __FP_FAST_FMAF<N> and __FP_FAST_FMAF<N>X if the backend supports
10654         fma _Float<N> and _Float<N>X variants.
10656 2017-10-27  Uros Bizjak  <ubizjak@gmail.com>
10658         PR target/82692
10659         * config/i386/i386-modes.def (CCFPU): Remove definition.
10660         * config/i386/i386.c (put_condition_mode): Remove CCFPU mode handling.
10661         (ix86_cc_modes_compatible): Ditto.
10662         (ix86_expand_carry_flag_compare): Ditto.
10663         (ix86_expand_int_movcc): Ditto.
10664         (ix86_expand_int_addcc): Ditto.
10665         (ix86_reverse_condition): Ditto.
10666         (ix86_unordered_fp_compare): Rename from ix86_fp_compare_mode.
10667         Return true/false for unordered/ordered fp comparisons.
10668         (ix86_cc_mode): Always return CCFPmode for float mode comparisons.
10669         (ix86_prepare_fp_compare_args): Update for rename.
10670         (ix86_expand_fp_compare): Update for rename.  Generate unordered
10671         compare RTXes wrapped with UNSPEC_NOTRAP unspec.
10672         (ix86_expand_sse_compare_and_jump): Ditto.
10673         * config/i386/predicates.md (fcmov_comparison_operator):
10674         Remove CCFPU mode handling.
10675         (ix86_comparison_operator): Ditto.
10676         (ix86_carry_flag_operator): Ditto.
10677         * config/i386/i386.md (UNSPEC_NOTRAP): New unspec.
10678         (*cmpu<mode>_i387): Wrap compare RTX with UNSPEC_NOTRAP unspec.
10679         (*cmpu<mode>_cc_i387): Ditto.
10680         (FPCMP): Remove mode iterator.
10681         (unord): Remove mode attribute.
10682         (unord_subst): New define_subst transformation
10683         (unord): New define_subst attribute.
10684         (unordered): Ditto.
10685         (*cmpi<unord><MODEF:mode>): Rewrite using unord_subst transformation.
10686         (*cmpi<unord>xf_i387): Ditto.
10687         * config/i386/sse.md (<sse>_<unord>comi<round_saeonly_name>): Merge
10688         from <sse>_comi<round_saeonly_name> and <sse>_ucomi<round_saeonly_name>
10689         using unord_subst transformation.
10690         * config/i386/subst.md (SUBST_A): Remove CCFP and CCFPU modes.
10691         (round_saeonly): Also handle CCFP mode.
10692         * reg-stack.c (subst_stack_regs_pat): Handle UNSPEC_NOTRAP unspec.
10693         Remove UNSPEC_SAHF unspec handling.
10695 2017-10-27  Jan Hubicka  <hubicka@ucw.cz>
10697         * x86-tune.def (X86_TUNE_INTER_UNIT_MOVES_TO_VEC): Disable for Zen.
10699 2017-10-27  Jeff Law  <law@redhat.com>
10701         * gimple-ssa-sprintf.c: Include domwalk.h.
10702         (class sprintf_dom_walker): New class, derived from dom_walker.
10703         (sprintf_dom_walker::before_dom_children): New function.
10704         (struct call_info): Moved into sprintf_dom_walker class
10705         (compute_formath_length, handle_gimple_call): Likewise.
10706         (sprintf_length::execute): Call the dominator walker rather
10707         than walking the statements.
10709         * tree-vrp.c (check_all_array_refs): Do not use wi->info to smuggle
10710         gimple statement locations.
10711         (check_array_bounds): Corresponding changes.  Get the statement's
10712         location directly from wi->stmt.
10714 2017-10-27  Palmer Dabbelt  <palmer@dabbelt.com>
10716         PR target/82717
10717         * doc/invoke.texi (RISC-V) <-mabi>: Correct and improve.
10719 2017-10-27  Jan Hubicka  <hubicka@ucw.cz>
10721         * config/i386/x86-tune.def (X86_TUNE_PARTIAL_REG_DEPENDENCY,
10722         X86_TUNE_MOVX): Disable for Haswell and newer CPUs.
10724 2017-10-27  Jakub Jelinek  <jakub@redhat.com>
10726         PR target/82703
10727         * config/i386/i386-protos.h (maybe_get_pool_constant): Removed.
10728         * config/i386/i386.c (maybe_get_pool_constant): Removed.
10729         (ix86_split_to_parts): Use avoid_constant_pool_reference instead of
10730         maybe_get_pool_constant.
10731         * config/i386/predicates.md (zero_extended_scalar_load_operand):
10732         Likewise.
10734 2017-10-27  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10736         * doc/install.texi (Specific, i?86-*-solaris2.10): Simplify gas
10737         2.26 caveat.  Update gas and gld versions.
10738         (Specific, *-*-solaris2*): Update binutils version.  Remove caveat
10739         reference.
10741 2017-10-27  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
10743         * cgraph.h (set_malloc_flag): Declare.
10744         * cgraph.c (set_malloc_flag_1): New function.
10745         (set_malloc_flag): Likewise.
10746         * ipa-fnsummary.h (ipa_call_summary): Add new field is_return_callee.
10747         * ipa-fnsummary.c (ipa_call_summary::reset): Set is_return_callee to
10748         false.
10749         (read_ipa_call_summary): Add support for reading is_return_callee.
10750         (write_ipa_call_summary): Stream is_return_callee.
10751         * ipa-inline.c (ipa_inline): Remove call to ipa_free_fn_summary.
10752         * ipa-pure-const.c: Add headers ssa.h, alloc-pool.h, symbol-summary.h,
10753         ipa-prop.h, ipa-fnsummary.h.
10754         (pure_const_names): Change to static.
10755         (malloc_state_e): Define.
10756         (malloc_state_names): Define.
10757         (funct_state_d): Add field malloc_state.
10758         (varying_state): Set malloc_state to STATE_MALLOC_BOTTOM.
10759         (check_retval_uses): New function.
10760         (malloc_candidate_p): Likewise.
10761         (analyze_function): Add support for malloc attribute.
10762         (pure_const_write_summary): Stream malloc_state.
10763         (pure_const_read_summary): Add support for reading malloc_state.
10764         (dump_malloc_lattice): New function.
10765         (propagate_malloc): New function.
10766         (warn_function_malloc): New function.
10767         (ipa_pure_const::execute): Call propagate_malloc and
10768         ipa_free_fn_summary.
10769         (pass_local_pure_const::execute): Add support for malloc attribute.
10770         * ssa-iterators.h (RETURN_FROM_IMM_USE_STMT): New macro.
10771         * doc/invoke.texi: Document Wsuggest-attribute=malloc.
10773 2017-10-27  Martin Liska  <mliska@suse.cz>
10775         PR gcov-profile/82457
10776         * doc/invoke.texi: Document that one needs a non-strict ISO mode
10777         for fork-like functions to be properly instrumented.
10779 2017-10-27  Richard Biener  <rguenther@suse.de>
10781         PR middle-end/81659
10782         * tree-eh.c (pass_lower_eh_dispatch::execute): Free dominator
10783         info when we redirected EH.
10785 2017-10-26  Michael Collison  <michael.collison@arm.com>
10787         * config/aarch64/aarch64.md(<optab>_trunc><vf><GPI:mode>2):
10788         New pattern.
10789         (<optab>_trunchf<GPI:mode>2: New pattern.
10790         (<optab>_trunc<vgp><GPI:mode>2: New pattern.
10791         * config/aarch64/iterators.md (wv): New mode attribute.
10792         (vf, VF): New mode attributes.
10793         (vgp, VGP): New mode attributes.
10794         (s): Update attribute with SImode and DImode prefixes.
10796 2017-10-26  Sandra Loosemore  <sandra@codesourcery.com>
10798         * config/nios2/constraints.md ("S"): Match r0rel_constant_p too.
10799         * config/nios2/nios2-protos.h (r0rel_constant_p): Declare.
10800         * config/nios2/nios2.c: (nios2_r0rel_sec_regex): New.
10801         (nios2_option_overide): Initialize it.  Don't allow R0-relative
10802         addressing with PIC.
10803         (nios2_rtx_costs): Handle r0rel_constant_p like gprel_constant_p.
10804         (nios2_symbolic_constant_p): Likewise.
10805         (nios2_legitimate_address_p): Likewise.
10806         (nios2_r0rel_section_name_p): New.
10807         (nios2_symbol_ref_in_r0rel_data_p): New.
10808         (nios2_emit_move_sequence): Handle r0rel_constant_p.
10809         (r0rel_constant_p): New.
10810         (nios2_print_operand_address): Handle r0rel_constant_p.
10811         (nios2_cdx_narrow_form_p): Likewise.
10812         * config/nios2/nios2.opt (mr0rel-sec=): New option.
10813         * doc/invoke.texi (Option Summary): Add -mr0rel-sec.
10814         (Nios II Options): Document -mr0rel-sec.
10816 2017-10-26  Sandra Loosemore  <sandra@codesourcery.com>
10818         * config/nios2/nios2.c: Include xregex.h.
10819         (nios2_gprel_sec_regex): New.
10820         (nios2_option_overide): Initialize it.  Don't allow GP-relative
10821         addressing with PIC.
10822         (nios2_small_section_name_p): Check for regex match.
10823         * config/nios2/nios2.opt (mgprel-sec=): New option.
10824         * doc/invoke.texi (Option Summary): Add -mgprel-sec.
10825         (Nios II Options): Document -mgprel-sec.
10827 2017-10-26  Jim Wilson  <wilson@tuliptree.org>
10829         * doc/invoke.texi (-fdebug-prefix-map): Expand documentation.
10831 2017-10-26  Tom de Vries  <tom@codesourcery.com>
10833         PR tree-optimization/82707
10834         * gimple.c (gimple_copy): Fix unsharing of
10835         GIMPLE_OMP_{SINGLE,TARGET,TEAMS}.
10837 2017-10-26  Olga Makhotina  <olga.makhotina@intel.com>
10839         * config/i386/avx512fintrin.h (_mm512_cmpeq_pd_mask,
10840         _mm512_cmple_pd_mask, _mm512_cmplt_pd_mask,
10841         _mm512_cmpneq_pd_mask, _mm512_cmpnle_pd_mask,
10842         _mm512_cmpnlt_pd_mask, _mm512_cmpord_pd_mask,
10843         _mm512_cmpunord_pd_mask, _mm512_mask_cmpeq_pd_mask,
10844         _mm512_mask_cmple_pd_mask, _mm512_mask_cmplt_pd_mask,
10845         _mm512_mask_cmpneq_pd_mask, _mm512_mask_cmpnle_pd_mask,
10846         _mm512_mask_cmpnlt_pd_mask, _mm512_mask_cmpord_pd_mask,
10847         _mm512_mask_cmpunord_pd_mask, _mm512_cmpeq_ps_mask,
10848         _mm512_cmple_ps_mask, _mm512_cmplt_ps_mask,
10849         _mm512_cmpneq_ps_mask, _mm512_cmpnle_ps_mask,
10850         _mm512_cmpnlt_ps_mask, _mm512_cmpord_ps_mask,
10851         _mm512_cmpunord_ps_mask, _mm512_mask_cmpeq_ps_mask,
10852         _mm512_mask_cmple_ps_mask, _mm512_mask_cmplt_ps_mask,
10853         _mm512_mask_cmpneq_ps_mask, _mm512_mask_cmpnle_ps_mask,
10854         _mm512_mask_cmpnlt_ps_mask, _mm512_mask_cmpord_ps_mask,
10855         _mm512_mask_cmpunord_ps_mask): New intrinsics.
10857 2017-10-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
10859         * config/rs6000/aix.h (TARGET_IEEEQUAD_DEFAULT): Set long double
10860         default to IBM.
10861         * config/rs6000/darwin.h (TARGET_IEEEQUAD_DEFAULT): Likewise.
10862         * config/rs6000/rs6000.opt (-mabi=ieeelongdouble): Move the
10863         warning to rs6000.c.  Remove the Undocumented flag, since it has
10864         been documented.
10865         (-mabi=ibmlongdouble): Likewise.
10866         * config/rs6000/rs6000.c (TARGET_IEEEQUAD_DEFAULT): If it is not
10867         already set, set the default format for long double.
10868         (rs6000_debug_reg_global): Print whether long double is IBM or
10869         IEEE.
10870         (rs6000_option_override_internal): Rework setting long double
10871         format.  Only warn if the user is changing the long double default
10872         and they did not use -Wno-psabi.
10873         * doc/invoke.texi (PowerPC options): Update the documentation for
10874         -mabi=ieeelongdouble and -mabi=ibmlongdouble.
10876 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10877             Alan Hayward  <alan.hayward@arm.com>
10878             David Sherwood  <david.sherwood@arm.com>
10880         * rtl.h (wider_subreg_mode): New function.
10881         * ira.h (ira_sort_regnos_for_alter_reg): Take a machine_mode *
10882         rather than an unsigned int *.
10883         * ira-color.c (regno_max_ref_width): Replace with...
10884         (regno_max_ref_mode): ...this new variable.
10885         (coalesced_pseudo_reg_slot_compare): Update accordingly.
10886         Use wider_subreg_mode.
10887         (ira_sort_regnos_for_alter_reg): Likewise.  Take a machine_mode *
10888         rather than an unsigned int *.
10889         * lra-constraints.c (uses_hard_regs_p): Use wider_subreg_mode.
10890         (process_alt_operands): Likewise.
10891         (invariant_p): Likewise.
10892         * lra-spills.c (assign_mem_slot): Likewise.
10893         (add_pseudo_to_slot): Likewise.
10894         * lra.c (collect_non_operand_hard_regs): Likewise.
10895         (add_regs_to_insn_regno_info): Likewise.
10896         * reload1.c (regno_max_ref_width): Replace with...
10897         (regno_max_ref_mode): ...this new variable.
10898         (reload): Update accordingly.  Update call to
10899         ira_sort_regnos_for_alter_reg.
10900         (alter_reg): Update to use regno_max_ref_mode.  Call wider_subreg_mode.
10901         (init_eliminable_invariants): Update to use regno_max_ref_mode.
10902         (scan_paradoxical_subregs): Likewise.
10904 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10906         * config/aarch64/aarch64.h (EXIT_IGNORE_STACK): Set if alloca is used.
10907         (aarch64_frame): Add emit_frame_chain boolean.
10908         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
10909         Move eh_return case to aarch64_layout_frame.
10910         (aarch64_layout_frame): Initialize emit_frame_chain.
10911         (aarch64_expand_prologue): Use emit_frame_chain.
10913 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10915         * config/aarch64/aarch64.c (aarch64_layout_frame):
10916         Ensure LR is always stored at the bottom of the callee-saves.
10917         Remove rarely used frame layout which saves callee-saves at top of
10918         frame, so the store of LR can be used as a valid probe in all cases.
10920 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10922         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10923         Improve unaligned TImode/TFmode base/offset split.
10925 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10926             Alan Hayward  <alan.hayward@arm.com>
10927             David Sherwood  <david.sherwood@arm.com>
10929         * caller-save.c (mark_referenced_regs):  Use read_modify_subreg_p.
10930         * combine.c (find_single_use_1): Likewise.
10931         (expand_field_assignment): Likewise.
10932         (move_deaths): Likewise.
10933         * lra-constraints.c (simplify_operand_subreg): Likewise.
10934         (curr_insn_transform): Likewise.
10935         * lra.c (collect_non_operand_hard_regs): Likewise.
10936         (add_regs_to_insn_regno_info): Likewise.
10937         * rtlanal.c (reg_referenced_p): Likewise.
10938         (covers_regno_no_parallel_p): Likewise.
10940 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10942         * wide-int-print.cc (print_hex): Loop based on extract_uhwi.
10943         Don't print any bits outside the precision of the value.
10944         * wide-int.cc (test_printing): Add some new tests.
10946 2017-10-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10948         * configure.ac (gcc_cv_as_ix86_xbrace_comment): Check if assembler
10949         supports -xbrace_comment option.
10950         * configure: Regenerate.
10951         * config.in: Regenerate.
10952         * config/i386/sol2.h (ASM_XBRACE_COMMENT_SPEC): Define.
10953         (ASM_CPU_SPEC): Use it.
10955 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10957         * target.def (static_rtx_alignment): New hook.
10958         * targhooks.h (default_static_rtx_alignment): Declare.
10959         * targhooks.c (default_static_rtx_alignment): New function.
10960         * doc/tm.texi.in (TARGET_STATIC_RTX_ALIGNMENT): New hook.
10961         * doc/tm.texi: Regenerate.
10962         * varasm.c (force_const_mem): Use targetm.static_rtx_alignment
10963         instead of targetm.constant_alignment.  Remove call to
10964         set_mem_attributes.
10965         * config/cris/cris.c (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10966         (cris_preferred_mininum_alignment): New function, split out from...
10967         (cris_constant_alignment): ...here.
10968         (cris_static_rtx_alignment): New function.
10969         * config/i386/i386.c (ix86_static_rtx_alignment): New function,
10970         split out from...
10971         (ix86_constant_alignment): ...here.
10972         (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10973         * config/mmix/mmix.c (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10974         (mmix_static_rtx_alignment): New function.
10975         * config/spu/spu.c (spu_static_rtx_alignment): New function.
10976         (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10978 2017-10-26  Tamar Christina  <tamar.christina@arm.com>
10980         PR target/81800
10981         * config/aarch64/aarch64.md (lrint<GPF:mode><GPI:mode>2):
10982         Add flag_trapping_math and flag_fp_int_builtin_inexact.
10984 2017-10-25  Palmer Dabbelt  <palmer@dabbelt.com>
10986         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Define.
10987         * config/riscv/pic.md (local_pic_load): Rename to local_pic_load_s,
10988         mark as a sign-extending load.
10989         (local_pic_load_u): Define.
10991 2017-10-25  Eric Botcazou  <ebotcazou@adacore.com>
10993         PR middle-end/82062
10994         * fold-const.c (operand_equal_for_comparison_p): Also return true
10995         if ARG0 is a simple variant of ARG1 with narrower precision.
10996         (fold_ternary_loc): Always pass unstripped operands to the predicate.
10998 2017-10-25  Jan Hubicka  <hubicka@ucw.cz>
11000         * i386.c (ix86_builtin_vectorization_cost): Compute scatter/gather
11001         cost correctly.
11002         * i386.h (processor_costs): Add gather_static, gather_per_elt,
11003         scatter_static, scatter_per_elt.
11004         * x86-tune-costs.h: Add new cost entries.
11006 2017-10-25  Richard Biener  <rguenther@suse.de>
11008         * tree-ssa-sccvn.h (vn_eliminate): Declare.
11009         * tree-ssa-pre.c (class eliminate_dom_walker, eliminate,
11010         class pass_fre): Move to ...
11011         * tree-ssa-sccvn.c (class eliminate_dom_walker, vn_eliminate,
11012         class pass_fre): ... here and adjust for statistics.
11014 2017-10-25  Jakub Jelinek  <jakub@redhat.com>
11016         PR libstdc++/81706
11017         * attribs.c (attribute_value_equal): Use omp_declare_simd_clauses_equal
11018         for comparison of OMP_CLAUSEs regardless of flag_openmp{,_simd}.
11019         (duplicate_one_attribute, copy_attributes_to_builtin): New functions.
11020         * attribs.h (duplicate_one_attribute, copy_attributes_to_builtin): New
11021         declarations.
11023 2017-10-25  Richard Biener  <rguenther@suse.de>
11025         * tree-ssa-pre.c (need_eh_cleanup, need_ab_cleanup, el_to_remove,
11026         el_to_fixup, el_todo, el_avail, el_avail_stack, eliminate_avail,
11027         eliminate_push_avail, eliminate_insert): Move inside...
11028         (class eliminate_dom_walker): ... this class in preparation
11029         of move.
11030         (fini_eliminate): Remove by merging with ...
11031         (eliminate): ... this function.  Adjust for class changes.
11032         (pass_pre::execute): Remove fini_eliminate call.
11033         (pass_fre::execute): Likewise.
11035 2017-10-24  Jakub Jelinek  <jakub@redhat.com>
11037         PR target/82460
11038         * config/i386/sse.md (UNSPEC_VPERMI2, UNSPEC_VPERMI2_MASK): Remove.
11039         (VPERMI2, VPERMI2I): New mode iterators.
11040         (<avx512>_vpermi2var<mode>3_maskz): Remove 3 define_expand patterns.
11041         (<avx512>_vpermi2var<mode>3<sd_maskz_name>): Remove 3 define_insn
11042         patterns.
11043         (<avx512>_vpermi2var<mode>3_mask): New define_expand using VPERMI2
11044         mode iterator.  Remove 3 old define_insn patterns.
11045         (*<avx512>_vpermi2var<mode>3_mask): 2 new define_insn patterns.
11046         (<avx512>_vpermt2var<mode>3_maskz): Adjust 1 define_expand to use
11047         VPERMI2 mode iterator, remove the other two expanders.
11048         (<avx512>_vpermt2var<mode>3<sd_maskz_name>): Adjust 1 define_insn
11049         to use VPERMI2 mode iterator, add another alternative for vpermi2*
11050         instructions, remove the other two patterns.
11051         (<avx512>_vpermt2var<mode>3_mask): Adjust 1 define_insn to use VPERMI2
11052         mode iterator, remove the other two patterns.
11053         * config/i386/i386.c (ix86_expand_vec_perm_vpermi2): Renamed to ...
11054         (ix86_expand_vec_perm_vpermt2): ... this.  Swap mask and op0
11055         arguments, use gen_*vpermt2* expanders instead of gen_*vpermi2*
11056         and adjust argument order accordingly.
11057         (ix86_expand_vec_perm): Adjust caller.
11058         (expand_vec_perm_1): Likewise.
11059         (expand_vec_perm_vpermi2_vpshub2): Rename to ...
11060         (expand_vec_perm_vpermt2_vpshub2): ... this.
11061         (ix86_expand_vec_perm_const_1): Adjust caller.
11062         (ix86_vectorize_vec_perm_const_ok): Adjust comments.
11064         PR target/82370
11065         * config/i386/sse.md (VIMAX_AVX2): Remove V4TImode.
11066         (VIMAX_AVX2_AVX512BW, VIMAX_AVX512VL): New mode iterators.
11067         (vec_shl_<mode>): Remove unused expander.
11068         (avx512bw_<shift_insn><mode>3): New define_insn.
11069         (<sse2_avx2>_ashl<mode>3, <sse2_avx2>_lshr<mode>3): Replaced by ...
11070         (<sse2_avx2>_<shift_insn><mode>3): ... this.  New define_insn.
11072 2017-10-24  Paolo Carlini  <paolo.carlini@oracle.com>
11074         PR c++/82466
11075         * doc/invoke.texi ([Wbuiltin-declaration-mismatch]): Extend
11076         description.
11078 2017-10-24  Wilco Dijkstra  <wdijkstr@arm.com>
11080         PR rtl-optimization/82396
11081         * gcc/haifa-sched.c (ready_sort_real): Remove qsort workaround.
11082         (autopref_multipass_init): Simplify initialization.
11083         (autopref_rank_data): Simplify sort order.
11084         * gcc/sched-int.h (autopref_multipass_data_): Remove
11085         multi_mem_insn_p, min_offset and max_offset.
11087 2017-10-24  Wilco Dijkstra  <wdijkstr@arm.com>
11089         PR middle-end/60580
11090         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
11091         Check special value of flag_omit_frame_pointer.
11092         (aarch64_can_eliminate): Likewise.
11093         (aarch64_override_options_after_change_1): Simplify handling of
11094         -fomit-frame-pointer and -fomit-leaf-frame-pointer.
11096 2017-10-24  Richard Biener  <rguenther@suse.de>
11098         PR tree-optimization/82697
11099         * tree-ssa-phiopt.c (cond_store_replacement): Use alias-set
11100         zero for conditional load and unconditional store.
11102 2017-10-24  H.J. Lu  <hongjiu.lu@intel.com>
11104         * doc/install.texi: Document bootstrap-cet.
11106 2017-10-24  H.J. Lu  <hongjiu.lu@intel.com>
11108         PR target/82659
11109         * config/i386/i386.c (rest_of_insert_endbranch): Don't insert
11110         ENDBR instruction at function entrance if function is only
11111         called directly.
11113 2017-10-24  Jakub Jelinek  <jakub@redhat.com>
11115         PR target/82628
11116         * config/i386/i386.md (addcarry<mode>, subborrow<mode>): Change
11117         patterns to better describe from which operation the CF is computed.
11118         (addcarry<mode>_0, subborrow<mode>_0): New patterns.
11119         * config/i386/i386.c (ix86_expand_builtin) <case handlecarry>: Pass
11120         one LTU with [DT]Imode and another one with [SD]Imode.  If arg0
11121         is 0, use _0 suffixed expanders instead of emitting a comparison
11122         before it.
11124 2017-10-06  Sergey Shalnov  <Sergey.Shalnov@intel.com>
11126         * config/i386/i386.md(*movsf_internal, *movdf_internal):
11127         Avoid 512-bit AVX modes for TARGET_PREFER_AVX256.
11129 2017-10-24  Eric Botcazou  <ebotcazou@adacore.com>
11131         PR middle-end/82569
11132         * tree-outof-ssa.h (always_initialized_rtx_for_ssa_name_p): Delete.
11133         * expr.c (expand_expr_real_1) <expand_decl_rtl>: Revert latest change.
11134         * loop-iv.c (iv_get_reaching_def): Likewise.
11135         * cfgexpand.c (expand_one_ssa_partition): Initialize the RTX if the
11136         variable is promoted and the partition contains undefined values.
11138 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
11140         * config/nios2/nios2.c (nios2_rtx_costs): Make costs better
11141         reflect reality.
11142         (nios2_address_cost): Define.
11143         (nios2_legitimize_address): Recognize (exp + constant) directly.
11144         (TARGET_ADDRESS_COST): Define.
11146 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
11148         * config/nios2/nios2-protos.h (nios2_large_constant_p): Declare.
11149         (nios2_symbolic_memory_operand_p): Declare.
11150         (nios2_split_large_constant): Declare.
11151         (nios2_split_symbolic_memory_operand): Declare.
11152         * config/nios2/nios2.c: Adjust includes.
11153         (nios2_symbolic_constant_allowed): New.
11154         (nios2_symbolic_constant_p): New.
11155         (nios2_plus_symbolic_constant_p): New.
11156         (nios2_valid_addr_expr_p): Recognize addresses involving
11157         symbolic constants.
11158         (nios2_legitimate_address_p): Likewise, also LO_SUM.
11159         (nios2_symbolic_memory_operand_p): New.
11160         (nios2_large_constant_p): New.
11161         (nios2_split_large_constant): New.
11162         (nios2_split_plus_large_constant): New.
11163         (nios2_split_symbolic_memory_operand): New.
11164         (nios2_legitimize_address): Code refactoring.  Handle addresses
11165         involving symbolic constants.
11166         (nios2_emit_move_sequence): Likewise.
11167         (nios2_print_operand): Improve error output.
11168         (nios2_print_operand_address): Handle LO_SUM.
11169         (nios2_cdx_narrow_form_p): Likewise.
11170         * config/nios2/nios2.md (movqi_internal): Add splitter for memory
11171         operands involving symbolic constants.
11172         (movhi_internal, movsi_internal): Likewise.
11173         (zero_extendhisi2, zero_extendqi<mode>2): Likewise.
11174         (extendhisi2, extendqi<mode>2): Likewise.
11176 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
11178         * tree-pass.h (PROP_rtl_split_insns): Define.
11179         * recog.c (pass_data_split_all_insns): Provide PROP_rtl_split_insns.
11181 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
11183         * config/nios2/nios2.c (TARGET_LRA_P): Don't override.
11185 2017-10-23  Jakub Jelinek  <jakub@redhat.com>
11187         PR debug/82630
11188         * target.def (const_not_ok_for_debug_p): Default to
11189         default_const_not_ok_for_debug_p instead of hook_bool_rtx_false.
11190         * targhooks.h (default_const_not_ok_for_debug_p): New declaration.
11191         * targhooks.c (default_const_not_ok_for_debug_p): New function.
11192         * dwarf2out.c (const_ok_for_output_1): Only reject UNSPECs for
11193         which targetm.const_not_ok_for_debug_p returned true.
11194         * config/arm/arm.c (arm_const_not_ok_for_debug_p): Return true
11195         for UNSPECs.
11196         * config/powerpcspe/powerpcspe.c (rs6000_const_not_ok_for_debug_p):
11197         Likewise.
11198         * config/rs6000/rs6000.c (rs6000_const_not_ok_for_debug_p): Likewise.
11199         * config/i386/i386.c (ix86_delegitimize_address_1): Don't delegitimize
11200         UNSPEC_GOTOFF with addend into addend - _GLOBAL_OFFSET_TABLE_ + symbol
11201         if !base_term_p.
11202         (ix86_const_not_ok_for_debug_p): New function.
11203         (i386_asm_output_addr_const_extra): Handle UNSPEC_GOTOFF.
11204         (TARGET_CONST_NOT_OK_FOR_DEBUG_P): Redefine.
11206 2017-10-23  David Malcolm  <dmalcolm@redhat.com>
11208         PR bootstrap/82610
11209         * system.h: Conditionally include "unique-ptr.h" if
11210         INCLUDE_UNIQUE_PTR is defined.
11211         * unique-ptr-tests.cc: Remove include of "unique-ptr.h" in favor
11212         of defining INCLUDE_UNIQUE_PTR before including "system.h".
11214 2017-10-23  Sebastian Perta  <sebastian.perta@renesas.com>
11216         * config/rl78/rl78.md: New define_expand "subdi3".
11218 2017-10-23  H.J. Lu  <hongjiu.lu@intel.com>
11220         PR target/82673
11221         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Skip
11222         DF_REF_INSN if DF_REF_INSN_INFO is false.
11224 2017-10-23  Jan Hubicka  <hubicka@ucw.cz>
11226         * i386.c (dimode_scalar_chain::compute_convert_gain): Use
11227         xmm_move instead of sse_move.
11228         (sse_store_index): New function.
11229         (ix86_register_move_cost): Be more sensible about mismatch stall;
11230         model AVX moves correctly; make difference between sse->integer and
11231         integer->sse.
11232         (ix86_builtin_vectorization_cost): Model correctly aligned and unaligned
11233         moves; make difference between SSE and AVX.
11234         * i386.h (processor_costs): Remove sse_move; add xmm_move, ymm_move
11235         and zmm_move. Increase size of sse load and store tables;
11236         add unaligned load and store tables; add ssemmx_to_integer.
11237         * x86-tune-costs.h: Update all entries according to real
11238         move latencies from Agner Fog's manual and chip documentation.
11240 2017-10-23  Jakub Jelinek  <jakub@redhat.com>
11242         PR target/82628
11243         * config/i386/predicates.md (x86_64_dwzext_immediate_operand): New.
11244         * config/i386/constraints.md (Wf): New constraint.
11245         * config/i386/i386.md (UNSPEC_SBB): New unspec.
11246         (cmp<dwi>_doubleword): Removed.
11247         (sub<mode>3_carry_ccc, *sub<mode>3_carry_ccc_1): New patterns.
11248         (sub<mode>3_carry_ccgz): Use unspec instead of compare.
11249         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>: Don't
11250         expand with cmp<dwi>_doubleword.  For LTU and GEU use
11251         sub<mode>3_carry_ccc instead of sub<mode>3_carry_ccgz and use CCCmode.
11253         * common.opt (gcolumn-info): Enable by default.
11254         * doc/invoke.texi (gcolumn-info): Document new default.
11256 2017-10-23  Richard Biener  <rguenther@suse.de>
11258         PR tree-optimization/82672
11259         * graphite-isl-ast-to-gimple.c (graphite_copy_stmts_from_block):
11260         Fold the stmt if we propagated into it.
11262 2017-10-23  Richard Biener  <rguenther@suse.de>
11264         * tree-ssa-pre.c (bitmap_remove_from_set): Rename to...
11265         (bitmap_remove_expr_from_set): ... this.  All callers call this
11266         for non-constant values.
11267         (bitmap_set_subtract): Rename to...
11268         (bitmap_set_subtract_expressions): ... this.  Adjust and
11269         optimize.
11270         (bitmap_set_contains_value): Remove superfluous check.
11271         (bitmap_set_replace_value): Inline into single caller ...
11272         (bitmap_value_replace_in_set): ... here and simplify.
11273         (dependent_clean): Merge into ...
11274         (clean): ... this using an overload.  Adjust.
11275         (prune_clobbered_mems): Adjust.
11276         (compute_antic_aux): Likewise.
11277         (compute_partial_antic_aux): Likewise.
11279 2017-10-23  Richard Biener  <rguenther@suse.de>
11281         PR tree-optimization/82129
11282         Revert
11283         2017-08-01  Richard Biener  <rguenther@suse.de>
11285         PR tree-optimization/81181
11286         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
11287         (compute_antic): ... end of iteration here.
11289 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11291         * target.def (starting_frame_offset): New hook.
11292         * doc/tm.texi (STARTING_FRAME_OFFSET): Remove in favor of...
11293         (TARGET_STARTING_FRAME_OFFSET): ...this new hook.
11294         * doc/tm.texi.in: Regenerate.
11295         * hooks.h (hook_hwi_void_0): Declare.
11296         * hooks.c (hook_hwi_void_0): New function.
11297         * doc/rtl.texi: Refer to TARGET_STARTING_FRAME_OFFSET instead of
11298         STARTING_FRAME_OFFSET.
11299         * builtins.c (expand_builtin_setjmp_receiver): Likewise.
11300         * reload1.c (reload): Likewise.
11301         * cfgexpand.c (expand_used_vars): Use targetm.starting_frame_offset
11302         instead of STARTING_FRAME_OFFSET.
11303         * function.c (try_fit_stack_local): Likewise.
11304         (assign_stack_local_1): Likewise
11305         (instantiate_virtual_regs): Likewise.
11306         * rtlanal.c (rtx_addr_can_trap_p_1): Likewise.
11307         * config/avr/avr.md (nonlocal_goto_receiver): Likewise.
11308         * config/aarch64/aarch64.h (STARTING_FRAME_OFFSET): Delete.
11309         * config/alpha/alpha.h (STARTING_FRAME_OFFSET): Likewise.
11310         * config/arc/arc.h (STARTING_FRAME_OFFSET): Likewise.
11311         * config/arm/arm.h (STARTING_FRAME_OFFSET): Likewise.
11312         * config/bfin/bfin.h (STARTING_FRAME_OFFSET): Likewise.
11313         * config/c6x/c6x.h (STARTING_FRAME_OFFSET): Likewise.
11314         * config/cr16/cr16.h (STARTING_FRAME_OFFSET): Likewise.
11315         * config/cris/cris.h (STARTING_FRAME_OFFSET): Likewise.
11316         * config/fr30/fr30.h (STARTING_FRAME_OFFSET): Likewise.
11317         * config/frv/frv.h (STARTING_FRAME_OFFSET): Likewise.
11318         * config/ft32/ft32.h (STARTING_FRAME_OFFSET): Likewise.
11319         * config/h8300/h8300.h (STARTING_FRAME_OFFSET): Likewise.
11320         * config/i386/i386.h (STARTING_FRAME_OFFSET): Likewise.
11321         * config/ia64/ia64.h (STARTING_FRAME_OFFSET): Likewise.
11322         * config/m32c/m32c.h (STARTING_FRAME_OFFSET): Likewise.
11323         * config/m68k/m68k.h (STARTING_FRAME_OFFSET): Likewise.
11324         * config/mcore/mcore.h (STARTING_FRAME_OFFSET): Likewise.
11325         * config/mn10300/mn10300.h (STARTING_FRAME_OFFSET): Likewise.
11326         * config/moxie/moxie.h (STARTING_FRAME_OFFSET): Likewise.
11327         * config/msp430/msp430.h (STARTING_FRAME_OFFSET): Likewise.
11328         * config/nds32/nds32.h (STARTING_FRAME_OFFSET): Likewise.
11329         * config/nios2/nios2.h (STARTING_FRAME_OFFSET): Likewise.
11330         * config/nvptx/nvptx.h (STARTING_FRAME_OFFSET): Likewise.
11331         * config/pdp11/pdp11.h (STARTING_FRAME_OFFSET): Likewise.
11332         * config/riscv/riscv.h (STARTING_FRAME_OFFSET): Likewise.
11333         * config/rl78/rl78.h (STARTING_FRAME_OFFSET): Likewise.
11334         * config/rx/rx.h (STARTING_FRAME_OFFSET): Likewise.
11335         * config/s390/s390.h (STARTING_FRAME_OFFSET): Likewise.
11336         * config/sh/sh.h (STARTING_FRAME_OFFSET): Likewise.
11337         * config/sparc/sparc.c (sparc_compute_frame_size): Likewise.
11338         * config/sparc/sparc.h (STARTING_FRAME_OFFSET): Likewise.
11339         * config/spu/spu.h (STARTING_FRAME_OFFSET): Likewise.
11340         * config/stormy16/stormy16.h (STARTING_FRAME_OFFSET): Likewise.
11341         * config/tilegx/tilegx.h (STARTING_FRAME_OFFSET): Likewise.
11342         * config/tilepro/tilepro.h (STARTING_FRAME_OFFSET): Likewise.
11343         * config/v850/v850.h (STARTING_FRAME_OFFSET): Likewise.
11344         * config/visium/visium.h (STARTING_FRAME_OFFSET): Likewise.
11345         * config/avr/avr.h (STARTING_FRAME_OFFSET): Likewise.
11346         * config/avr/avr-protos.h (avr_starting_frame_offset): Likewise.
11347         * config/avr/avr.c (avr_starting_frame_offset): Make static and
11348         return a HOST_WIDE_INT.
11349         (avr_builtin_setjmp_frame_value): Use it instead of
11350         STARTING_FRAME_OFFSET.
11351         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11352         * config/epiphany/epiphany.h (STARTING_FRAME_OFFSET): Delete.
11353         * config/epiphany/epiphany.c (epiphany_starting_frame_offset):
11354         New function.
11355         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11356         * config/iq2000/iq2000.h (STARTING_FRAME_OFFSET): Delete.
11357         * config/iq2000/iq2000.c (iq2000_starting_frame_offset): New function.
11358         (TARGET_CONSTANT_ALIGNMENT): Redefine.
11359         * config/lm32/lm32.h (STARTING_FRAME_OFFSET): Delete.
11360         * config/lm32/lm32.c (lm32_starting_frame_offset): New function.
11361         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11362         * config/m32r/m32r.h (STARTING_FRAME_OFFSET): Delete.
11363         * config/m32r/m32r.c (m32r_starting_frame_offset): New function.
11364         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11365         * config/microblaze/microblaze.h (STARTING_FRAME_OFFSET): Delete.
11366         * config/microblaze/microblaze.c (microblaze_starting_frame_offset):
11367         New function.
11368         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11369         * config/mips/mips.h (STARTING_FRAME_OFFSET): Delete.
11370         * config/mips/mips.c (mips_compute_frame_info): Refer to
11371         TARGET_STARTING_FRAME_OFFSET instead of STARTING_FRAME_OFFSET.
11372         (mips_starting_frame_offset): New function.
11373         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11374         * config/mmix/mmix.h (STARTING_FRAME_OFFSET): Delete.
11375         * config/mmix/mmix-protos.h (mmix_starting_frame_offset): Delete.
11376         * config/mmix/mmix.c (mmix_starting_frame_offset): Make static
11377         and return a HOST_WIDE_INT.
11378         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11379         (mmix_initial_elimination_offset): Refer to
11380         TARGET_STARTING_FRAME_OFFSET instead of STARTING_FRAME_OFFSET.
11381         * config/pa/pa.h (STARTING_FRAME_OFFSET): Delete.
11382         * config/pa/pa.c (pa_starting_frame_offset): New function.
11383         (pa_compute_frame_size): Use it instead of STARTING_FRAME_OFFSET.
11384         (pa_expand_prologue): Likewise.
11385         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11386         * config/powerpcspe/aix.h (STARTING_FRAME_OFFSET): Split out
11387         !FRAME_GROWS_DOWNWARD handling to...
11388         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11389         * config/powerpcspe/darwin.h (STARTING_FRAME_OFFSET): Split out
11390         !FRAME_GROWS_DOWNWARD handling to...
11391         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11392         * config/powerpcspe/powerpcspe.h (STARTING_FRAME_OFFSET): Split out
11393         !FRAME_GROWS_DOWNWARD handling to...
11394         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11395         * config/powerpcspe/powerpcspe.c (TARGET_STARTING_FRAME_OFFSET):
11396         Redefine.
11397         (rs6000_starting_frame_offset): New function.
11398         * config/rs6000/aix.h (STARTING_FRAME_OFFSET): Split out
11399         !FRAME_GROWS_DOWNWARD handling to...
11400         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11401         * config/rs6000/darwin.h (STARTING_FRAME_OFFSET): Split out
11402         !FRAME_GROWS_DOWNWARD handling to...
11403         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11404         * config/rs6000/rs6000.h (STARTING_FRAME_OFFSET): Split out
11405         !FRAME_GROWS_DOWNWARD handling to...
11406         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11407         * config/rs6000/rs6000.c (TARGET_STARTING_FRAME_OFFSET): Refine.
11408         (rs6000_starting_frame_offset): New function.
11409         * config/vax/elf.h (STARTING_FRAME_OFFSET): Delete.
11410         * config/vax/vax.h (STARTING_FRAME_OFFSET): Delete.
11411         * config/vax/vax.c (vax_starting_frame_offset): New function.
11412         (vax_expand_prologue): Use it instead of STARTING_FRAME_OFFSET.
11413         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11414         * config/xtensa/xtensa.h (STARTING_FRAME_OFFSET): Delete.
11415         * config/xtensa/xtensa.c (xtensa_starting_frame_offset): New function.
11416         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11417         * system.h (STARTING_FRAME_OFFSET): Poison.
11419 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11421         * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
11422         SCALAR_TYPE_MODE instead of TYPE_MODE.
11424 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11425             Alan Hayward  <alan.hayward@arm.com>
11426             David Sherwood  <david.sherwood@arm.com>
11428         * dwarf2out.c (loc_list_from_tree_1): Use SCALAR_INT_TYPE_MODE
11430 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11431             Alan Hayward  <alan.hayward@arm.com>
11432             David Sherwood  <david.sherwood@arm.com>
11434         * expmed.c (expand_shift_1): Use scalar_mode for scalar_mode.
11436 2017-10-23  Richard Biener  <rguenther@suse.de>
11438         PR tree-optimization/82129
11439         * tree-ssa-pre.c (bitmap_set_and): Remove.
11440         (compute_antic_aux): Compute ANTIC_OUT intersection in a way
11441         canonicalizing expressions in the set to those with lowest
11442         ID rather than taking that from the first edge.
11444 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11446         * combine.c (rtx_equal_for_field_assignment_p): Use
11447         byte_lowpart_offset.
11449 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11450             Alan Hayward  <alan.hayward@arm.com>
11451             David Sherwood  <david.sherwood@arm.com>
11453         * internal-fn.c (expand_direct_optab_fn): Don't assign directly
11454         to a SUBREG_PROMOTED_VAR.
11456 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11457             Alan Hayward  <alan.hayward@arm.com>
11458             David Sherwood  <david.sherwood@arm.com>
11460         * cfgexpand.c (expand_debug_expr): Use GET_MODE_UNIT_PRECISION.
11461         (expand_debug_source_expr): Likewise.
11462         * combine.c (combine_simplify_rtx): Likewise.
11463         * cse.c (fold_rtx): Likewise.
11464         * optabs.c (expand_float): Likewise.
11465         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11466         (simplify_binary_operation_1): Likewise.
11468 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11469             Alan Hayward  <alan.hayward@arm.com>
11470             David Sherwood  <david.sherwood@arm.com>
11472         * combine.c (simplify_comparison): Use HWI_COMPUTABLE_MODE_P.
11473         (record_promoted_value): Likewise.
11474         * expr.c (expand_expr_real_2): Likewise.
11475         * ree.c (update_reg_equal_equiv_notes): Likewise.
11476         (combine_set_extension): Likewise.
11477         * rtlanal.c (low_bitmask_len): Likewise.
11478         * simplify-rtx.c (neg_const_int): Likewise.
11479         (simplify_binary_operation_1): Likewise.
11481 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11482             Alan Hayward  <alan.hayward@arm.com>
11483             David Sherwood  <david.sherwood@arm.com>
11485         * lra-spills.c (assign_mem_slot): Use subreg_size_lowpart_offset.
11486         * regcprop.c (maybe_mode_change): Likewise.
11487         * reload1.c (alter_reg): Likewise.
11489 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11491         * inchash.h (inchash::hash::add_wide_int): New function.
11492         * lto-streamer-out.c (hash_tree): Use it.
11494 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11496         * inchash.h (inchash::hash::add_wide_int): Rename to...
11497         (inchash::hash::add_hwi): ...this.
11498         * ipa-devirt.c (hash_odr_vtable): Update accordingly.
11499         (polymorphic_call_target_hasher::hash): Likewise.
11500         * ipa-icf.c (sem_function::get_hash, sem_function::init): Likewise.
11501         (sem_item::add_expr, sem_item::add_type, sem_variable::get_hash)
11502         (sem_item_optimizer::update_hash_by_addr_refs): Likewise.
11503         * lto-streamer-out.c (hash_tree): Likewise.
11504         * optc-save-gen.awk: Likewise.
11505         * tree.c (add_expr): Likewise.
11507 2017-10-22  Uros Bizjak  <ubizjak@gmail.com>
11509         PR target/52451
11510         * config/i386/i386.c (ix86_fp_compare_mode): Return CCFPmode
11511         for ordered inequality comparisons even with TARGET_IEEE_FP.
11513 2017-10-22  Uros Bizjak  <ubizjak@gmail.com>
11515         PR target/82628
11516         * config/i386/i386.md (cmp<dwi>_doubleword): New pattern.
11517         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>:
11518         Expand with cmp<dwi>_doubleword.
11520 2017-10-21  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11522         * extend.texi: Add x86 specific to 'nocf_check' attribute.
11523         List CET intrinsics.
11524         * invoke.texi: Add -mcet, -mibt, -mshstk options.  Add x86
11525         specific to -fcf-protection option.
11527 2017-10-21  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11529         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET): New.
11530         (OPTION_MASK_ISA_SHSTK_SET): Likewise.
11531         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
11532         (OPTION_MASK_ISA_SHSTK_UNSET): Likewise.
11533         (ix86_handle_option): Add -mibt, -mshstk, -mcet handling.
11534         * config.gcc (extra_headers): Add cetintrin.h for x86 targets.
11535         (extra_objs): Add cet.o for Linux/x86 targets.
11536         (tmake_file): Add i386/t-cet for Linux/x86 targets.
11537         * config/i386/cet.c: New file.
11538         * config/i386/cetintrin.h: Likewise.
11539         * config/i386/t-cet: Likewise.
11540         * config/i386/cpuid.h (bit_SHSTK): New.
11541         (bit_IBT): Likewise.
11542         * config/i386/driver-i386.c (host_detect_local_cpu): Detect and
11543         pass IBT and SHSTK bits.
11544         * config/i386/i386-builtin-types.def
11545         (VOID_FTYPE_UNSIGNED_PVOID): New.
11546         (VOID_FTYPE_UINT64_PVOID): Likewise.
11547         * config/i386/i386-builtin.def: Add CET intrinsics.
11548         * config/i386/i386-c.c (ix86_target_macros_internal): Add
11549         OPTION_MASK_ISA_IBT, OPTION_MASK_ISA_SHSTK handling.
11550         * config/i386/i386-passes.def: Add pass_insert_endbranch pass.
11551         * config/i386/i386-protos.h (make_pass_insert_endbranch): New
11552         prototype.
11553         * config/i386/i386.c (rest_of_insert_endbranch): New.
11554         (pass_data_insert_endbranch): Likewise.
11555         (pass_insert_endbranch): Likewise.
11556         (make_pass_insert_endbranch): Likewise.
11557         (ix86_notrack_prefixed_insn_p): Likewise.
11558         (ix86_target_string): Add -mibt, -mshstk flags.
11559         (ix86_option_override_internal): Add flag_cf_protection
11560         processing.
11561         (ix86_valid_target_attribute_inner_p): Set OPT_mibt, OPT_mshstk.
11562         (ix86_print_operand): Add 'notrack' prefix output.
11563         (ix86_init_mmx_sse_builtins): Add CET intrinsics.
11564         (ix86_expand_builtin): Expand CET intrinsics.
11565         (x86_output_mi_thunk): Add 'endbranch' instruction.
11566         * config/i386/i386.h (TARGET_IBT): New.
11567         (TARGET_IBT_P): Likewise.
11568         (TARGET_SHSTK): Likewise.
11569         (TARGET_SHSTK_P): Likewise.
11570         * config/i386/i386.md (unspecv): Add UNSPECV_NOP_RDSSP,
11571         UNSPECV_INCSSP, UNSPECV_SAVEPREVSSP, UNSPECV_RSTORSSP,
11572         UNSPECV_WRSS, UNSPECV_WRUSS, UNSPECV_SETSSBSY, UNSPECV_CLRSSBSY.
11573         (builtin_setjmp_setup): New pattern.
11574         (builtin_longjmp): Likewise.
11575         (rdssp<mode>): Likewise.
11576         (incssp<mode>): Likewise.
11577         (saveprevssp): Likewise.
11578         (rstorssp): Likewise.
11579         (wrss<mode>): Likewise.
11580         (wruss<mode>): Likewise.
11581         (setssbsy): Likewise.
11582         (clrssbsy): Likewise.
11583         (nop_endbr): Likewise.
11584         * config/i386/i386.opt: Add -mcet, -mibt, -mshstk and -mcet-switch
11585         options.
11586         * config/i386/immintrin.h: Include <cetintrin.h>.
11587         * config/i386/linux-common.h
11588         (file_end_indicate_exec_stack_and_cet): New prototype.
11589         (TARGET_ASM_FILE_END): New.
11591 2017-10-20  Jan Hubicka  <hubicka@ucw.cz>
11593         * i386.c (ix86_builtin_vectorization_cost): Use existing rtx_cost
11594         latencies instead of having separate table; make difference between
11595         integer and float costs.
11596         * i386.h (processor_costs): Remove scalar_stmt_cost,
11597         scalar_load_cost, scalar_store_cost, vec_stmt_cost, vec_to_scalar_cost,
11598         scalar_to_vec_cost, vec_align_load_cost, vec_unalign_load_cost,
11599         vec_store_cost.
11600         * x86-tune-costs.h: Remove entries which has been removed in
11601         procesor_costs from all tables; make cond_taken_branch_cost
11602         and cond_not_taken_branch_cost COST_N_INSNS based.
11604 2017-10-20  Jan Hubicka  <hubicka@ucw.cz>
11606         * x86-tune-costs.h (intel_cost, generic_cost): Fix move costs.
11608 2017-10-20  Jakub Jelinek  <jakub@redhat.com>
11610         * config/i386/i386.md (isa): Remove fma_avx512f.
11611         * config/i386/sse.md (<avx512>_fmadd_<mode>_mask<round_name>,
11612         <avx512>_fmadd_<mode>_mask3<round_name>,
11613         <avx512>_fmsub_<mode>_mask<round_name>,
11614         <avx512>_fmsub_<mode>_mask3<round_name>,
11615         <avx512>_fnmadd_<mode>_mask<round_name>,
11616         <avx512>_fnmadd_<mode>_mask3<round_name>,
11617         <avx512>_fnmsub_<mode>_mask<round_name>,
11618         <avx512>_fnmsub_<mode>_mask3<round_name>,
11619         <avx512>_fmaddsub_<mode>_mask<round_name>,
11620         <avx512>_fmaddsub_<mode>_mask3<round_name>,
11621         <avx512>_fmsubadd_<mode>_mask<round_name>,
11622         <avx512>_fmsubadd_<mode>_mask3<round_name>): Remove isa attribute.
11623         (*vec_widen_umult_even_v16si<mask_name>,
11624         *vec_widen_smult_even_v16si<mask_name>): Likewise.
11625         (<mask_codefor>avx512bw_dbpsadbw<mode><mask_name>): Likewise.
11627 2017-10-20  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11629         * extend.texi: Add 'nocf_check' documentation.
11630         * gimple.texi: Add second parameter to
11631         gimple_build_call_from_tree.
11632         * invoke.texi: Add -fcf-protection documentation.
11633         * rtl.texi: Add REG_CALL_NOTRACK documenation.
11635 2017-10-20  Richard Biener  <rguenther@suse.de>
11637         PR tree-optimization/82473
11638         * tree-vect-loop.c (vectorizable_reduction): Properly get at
11639         the largest input type.
11641 2017-10-20  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11643         * c-attribs.c (handle_nocf_check_attribute): New function.
11644         (c_common_attribute_table): Add 'nocf_check' handling.
11645         * gimple-parser.c: Add second argument NULL to
11646         gimple_build_call_from_tree.
11647         * attrib.c (comp_type_attributes): Check nocf_check attribute.
11648         * cfgexpand.c (expand_call_stmt): Set REG_CALL_NOCF_CHECK for
11649         call insn.
11650         * combine.c (distribute_notes): Add REG_CALL_NOCF_CHECK handling.
11651         * common.opt: Add fcf-protection flag.
11652         * emit-rtl.c (try_split): Add REG_CALL_NOCF_CHECK handling.
11653         * flag-types.h: Add enum cf_protection_level.
11654         * gimple.c (gimple_build_call_from_tree): Add second parameter.
11655         Add 'nocf_check' attribute propagation to gimple call.
11656         * gimple.h (gf_mask): Add GF_CALL_NOCF_CHECK.
11657         (gimple_build_call_from_tree): Update prototype.
11658         (gimple_call_nocf_check_p): New function.
11659         (gimple_call_set_nocf_check): Likewise.
11660         * gimplify.c: Add second argument to gimple_build_call_from_tree.
11661         * ipa-icf.c: Add nocf_check attribute in statement hash.
11662         * recog.c (peep2_attempt): Add REG_CALL_NOCF_CHECK handling.
11663         * reg-notes.def: Add REG_NOTE (CALL_NOCF_CHECK).
11664         * toplev.c (process_options): Add flag_cf_protection handling.
11666 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11668         * x86-tune-costs.h (core_cost): Fix div, move and sqrt latencies.
11670 2017-10-20  Richard Biener  <rguenther@suse.de>
11672         PR tree-optimization/82603
11673         * tree-if-conv.c (predicate_mem_writes): Make sure to only
11674         remove false predicated stores.
11676 2017-10-20  Richard Biener  <rguenther@suse.de>
11678         * graphite-isl-ast-to-gimple.c
11679         (translate_isl_ast_to_gimple::graphite_copy_stmts_from_block):
11680         Remove return value and simplify, dump copied stmt after lhs
11681         adjustment.
11682         (translate_isl_ast_to_gimple::translate_isl_ast_node_user):
11683         Reduce dump verbosity.
11684         (gsi_insert_earliest): Likewise.
11685         (translate_isl_ast_to_gimple::copy_bb_and_scalar_dependences): Adjust.
11686         * graphite.c (print_global_statistics): Adjust dumping.
11687         (print_graphite_scop_statistics): Likewise.
11688         (print_graphite_statistics): Do not dump loops here.
11689         (graphite_transform_loops): But here.
11691 2017-10-20  Nicolas Roche  <roche@adacore.com>
11693         * configure.ac (ACX_PROG_GNAT): Append "libgnat" to include search dir.
11694         * configure: Regenerate.
11696 2017-10-20  Jakub Jelinek  <jakub@redhat.com>
11698         PR target/82158
11699         * tree-cfg.c (pass_warn_function_return::execute): In noreturn
11700         functions when optimizing replace GIMPLE_RETURN stmts with
11701         calls to __builtin_unreachable ().
11703         PR sanitizer/82595
11704         * config/gnu-user.h (LIBTSAN_EARLY_SPEC): Add libtsan_preinit.o
11705         for -fsanitize=thread link of executables.
11706         (LIBLSAN_EARLY_SPEC): Add liblsan_preinit.o for -fsanitize=leak
11707         link of executables.
11709         PR target/82370
11710         * config/i386/sse.md (VI248_AVX2, VI248_AVX512BW, VI248_AVX512BW_2):
11711         New mode iterators.
11712         (<shift_insn><mode>3<mask_name>): Change the last of the 3
11713         define_insns for logical vector shifts to use VI248_AVX512BW
11714         iterator instead of VI48_AVX512, remove <mask_mode512bit_condition>
11715         condition, useless isa and prefix attributes.  Change the first
11716         2 of these define_insns to ...
11717         (<mask_codefor><shift_insn><mode>3<mask_name>): ... this, new
11718         define_insn for avx512vl.
11719         (<shift_insn><mode>3): ... and this, new define_insn without
11720         masking for non-avx512vl.
11722         PR target/82370
11723         * config/i386/sse.md (*andnot<mode>3,
11724         <mask_codefor><code><mode>3<mask_name>, *<code><mode>3): Split
11725         (=v,v,vm) alternative into (=x,x,xm) and (=v,v,vm), for 128-bit
11726         and 256-bit vectors, the (=x,x,xm) alternative and when mask is
11727         not applied use empty suffix even for TARGET_AVX512VL.
11728         * config/i386/subst.md (mask_prefix3, mask_prefix4): When mask
11729         is applied, supply evex,evex or evex,evex,evex instead of just
11730         evex.
11732 2017-10-20  Julia Koval  <julia.koval@intel.com>
11734         * common/config/i386/i386-common.c (OPTION_MASK_ISA_GFNI_SET,
11735         (OPTION_MASK_ISA_GFNI_UNSET): New.
11736         (ix86_handle_option): Handle OPT_mgfni.
11737         * config/i386/cpuid.h (bit_GFNI): New.
11738         * config/i386/driver-i386.c (host_detect_local_cpu): Detect gfni.
11739         * config/i386/i386-c.c (ix86_target_macros_internal): Define __GFNI__.
11740         * config/i386/i386.c (ix86_target_string): Add -mgfni.
11741         (ix86_valid_target_attribute_inner_p): Add OPT_mgfni.
11742         * config/i386/i386.h (TARGET_GFNI, TARGET_GFNI_P): New.
11743         * config/i386/i386.opt: Add mgfni.
11745 2017-10-20  Orlando Arias  <oarias@knights.ucf.edu>
11747         * config/msp430/msp430.c (msp430_option_override): Disable
11748         -fdelete-null-pointer-checks.
11749         * doc/invoke.text (-fdelete-null-pointer-checks): Document that.
11751 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11753         * x86-tune-costs.h (generic_cost, core_cost): Correct costs
11754         of x87 and SSE instructions.
11756 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11758         * asan.c (create_cond_insert_point): Do not update edge count.
11759         * auto-profile.c (afdo_propagate_edge): Update for edge count removal.
11760         (afdo_propagate_circuit): Likewise.
11761         (afdo_calculate_branch_prob): Likewise.
11762         (afdo_annotate_cfg): Likewise.
11763         * basic-block.h (struct edge_def): Remove count.
11764         (edge_def::count): New accessor.
11765         * bb-reorder.c (rotate_loop): Update.
11766         (find_traces_1_round): Update.
11767         (connect_traces): Update.
11768         (sanitize_hot_paths): Update.
11769         * cfg.c (unchecked_make_edge): Update.
11770         (make_single_succ_edge): Update.
11771         (check_bb_profile): Update.
11772         (dump_edge_info): Update.
11773         (update_bb_profile_for_threading): Update.
11774         (scale_bbs_frequencies_int): Update.
11775         (scale_bbs_frequencies_gcov_type): Update.
11776         (scale_bbs_frequencies_profile_count): Update.
11777         (scale_bbs_frequencies): Update.
11778         * cfganal.c (connect_infinite_loops_to_exit): Update.
11779         * cfgbuild.c (compute_outgoing_frequencies): Update.
11780         (find_many_sub_basic_blocks): Update.
11781         * cfgcleanup.c (try_forward_edges): Update.
11782         (try_crossjump_to_edge): Update
11783         * cfgexpand.c (expand_gimple_cond): Update
11784         (expand_gimple_tailcall): Update
11785         (construct_exit_block): Update
11786         * cfghooks.c (verify_flow_info): Update
11787         (redirect_edge_succ_nodup): Update
11788         (split_edge): Update
11789         (make_forwarder_block): Update
11790         (duplicate_block): Update
11791         (account_profile_record): Update
11792         * cfgloop.c (find_subloop_latch_edge_by_profile): Update.
11793         * cfgloopanal.c (expected_loop_iterations_unbounded): Update.
11794         * cfgloopmanip.c (scale_loop_profile): Update.
11795         (loopify): Update.
11796         (lv_adjust_loop_entry_edge): Update.
11797         * cfgrtl.c (try_redirect_by_replacing_jump): Update.
11798         (force_nonfallthru_and_redirect): Update.
11799         (purge_dead_edges): Update.
11800         (rtl_flow_call_edges_add): Update.
11801         * cgraphunit.c (init_lowered_empty_function): Update.
11802         (cgraph_node::expand_thunk): Update.
11803         * gimple-pretty-print.c (dump_probability): Update.
11804         (dump_edge_probability): Update.
11805         * gimple-ssa-isolate-paths.c (isolate_path): Update.
11806         * haifa-sched.c (sched_create_recovery_edges): Update.
11807         * hsa-gen.c (convert_switch_statements): Update.
11808         * ifcvt.c (dead_or_predicable): Update.
11809         * ipa-inline-transform.c (inline_transform): Update.
11810         * ipa-split.c (split_function): Update.
11811         * ipa-utils.c (ipa_merge_profiles): Update.
11812         * loop-doloop.c (add_test): Update.
11813         * loop-unroll.c (unroll_loop_runtime_iterations): Update.
11814         * lto-streamer-in.c (input_cfg): Update.
11815         (input_function): Update.
11816         * lto-streamer-out.c (output_cfg): Update.
11817         * modulo-sched.c (sms_schedule): Update.
11818         * postreload-gcse.c (eliminate_partially_redundant_load): Update.
11819         * predict.c (maybe_hot_edge_p): Update.
11820         (unlikely_executed_edge_p): Update.
11821         (probably_never_executed_edge_p): Update.
11822         (dump_prediction): Update.
11823         (drop_profile): Update.
11824         (propagate_unlikely_bbs_forward): Update.
11825         (determine_unlikely_bbs): Update.
11826         (force_edge_cold): Update.
11827         * profile.c (compute_branch_probabilities): Update.
11828         * reg-stack.c (better_edge): Update.
11829         * shrink-wrap.c (handle_simple_exit): Update.
11830         * tracer.c (better_p): Update.
11831         * trans-mem.c (expand_transaction): Update.
11832         (split_bb_make_tm_edge): Update.
11833         * tree-call-cdce.c: Update.
11834         * tree-cfg.c (gimple_find_sub_bbs): Update.
11835         (gimple_split_edge): Update.
11836         (gimple_duplicate_sese_region): Update.
11837         (gimple_duplicate_sese_tail): Update.
11838         (gimple_flow_call_edges_add): Update.
11839         (insert_cond_bb): Update.
11840         (execute_fixup_cfg): Update.
11841         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update.
11842         * tree-complex.c (expand_complex_div_wide): Update.
11843         * tree-eh.c (lower_resx): Update.
11844         (unsplit_eh): Update.
11845         (cleanup_empty_eh_move_lp): Update.
11846         * tree-inline.c (copy_edges_for_bb): Update.
11847         (freqs_to_counts): Update.
11848         (copy_cfg_body): Update.
11849         * tree-ssa-dce.c (remove_dead_stmt): Update.
11850         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
11851         * tree-ssa-loop-im.c (execute_sm_if_changed): Update.
11852         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update.
11853         (unloop_loops): Update.
11854         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
11855         * tree-ssa-loop-split.c (connect_loops): Update.
11856         (split_loop): Update.
11857         * tree-ssa-loop-unswitch.c (hoist_guard): Update.
11858         * tree-ssa-phionlycprop.c (propagate_rhs_into_lhs): Update.
11859         * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Update.
11860         * tree-ssa-reassoc.c (branch_fixup): Update.
11861         * tree-ssa-tail-merge.c (replace_block_by): Update.
11862         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges): Update.
11863         (compute_path_counts): Update.
11864         (update_profile): Update.
11865         (recompute_probabilities): Update.
11866         (update_joiner_offpath_counts): Update.
11867         (estimated_freqs_path): Update.
11868         (freqs_to_counts_path): Update.
11869         (clear_counts_path): Update.
11870         (ssa_fix_duplicate_block_edges): Update.
11871         (duplicate_thread_path): Update.
11872         * tree-switch-conversion.c (hoist_edge_and_branch_if_true): Update.
11873         (case_bit_test_cmp): Update.
11874         (collect_switch_conv_info): Update.
11875         (gen_inbound_check): Update.
11876         (do_jump_if_equal): Update.
11877         (emit_cmp_and_jump_insns): Update.
11878         * tree-tailcall.c (decrease_profile): Update.
11879         (eliminate_tail_call): Update.
11880         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
11881         (vect_do_peeling): Update.
11882         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
11883         * ubsan.c (ubsan_expand_null_ifn): Update.
11884         (ubsan_expand_ptr_ifn): Update.
11885         * value-prof.c (gimple_divmod_fixed_value): Update.
11886         (gimple_mod_pow2): Update.
11887         (gimple_mod_subtract): Update.
11888         (gimple_ic): Update.
11889         (gimple_stringop_fixed_value): Update.
11891 2017-10-19  Uros Bizjak  <ubizjak@gmail.com>
11893         PR target/82618
11894         * config/i386/i386.md (sub to cmp): New peephole2 pattern.
11896 2017-10-19  Alexander Monakov  <amonakov@ispras.ru>
11898         PR rtl-optimization/82395
11899         * ira-color.c (allocno_priority_compare_func): Fix comparison step
11900         based on non_spilled_static_chain_regno_p.
11902 2017-10-19  Uros Bizjak  <ubizjak@gmail.com>
11904         * config/i386/i386.c (output_387_binary_op): Rewrite SSE part.
11905         (ix86_emit_mode_set): Rewrite insn mnemonic construction.
11906         (ix86_prepare_fp_compare_args): Redefine is_sse as bool.
11908 2017-10-19  Martin Sebor  <msebor@redhat.com>
11910         PR tree-optimization/82596
11911         * tree.c (array_at_struct_end_p): Handle STRING_CST.
11913 2017-10-19  Eric Botcazou  <ebotcazou@adacore.com>
11915         * asan.c (handle_builtin_alloca): Deal with all alloca variants.
11916         (get_mem_refs_of_builtin_call): Likewise.
11917         * builtins.c (expand_builtin_apply): Adjust call to
11918         allocate_dynamic_stack_space.
11919         (expand_builtin_alloca): For __builtin_alloca_with_align_and_max, pass
11920         the third argument to allocate_dynamic_stack_space, otherwise -1.
11921         (expand_builtin): Deal with all alloca variants.
11922         (is_inexpensive_builtin): Likewise.
11923         * builtins.def (BUILT_IN_ALLOCA_WITH_ALIGN_AND_MAX): New.
11924         * calls.c (special_function_p): Deal with all alloca variants.
11925         (initialize_argument_information): Adjust call to
11926         allocate_dynamic_stack_space.
11927         (expand_call): Likewise.
11928         * cfgexpand.c (expand_call_stmt): Deal with all alloca variants.
11929         * doc/extend.texi (Built-ins): Add __builtin_alloca_with_align_and_max
11930         * explow.c (allocate_dynamic_stack_space): Add MAX_SIZE parameter and
11931         use it for the stack usage computation.
11932         * explow.h (allocate_dynamic_stack_space): Adjust prototype.
11933         * function.c (gimplify_parameters): Call build_alloca_call_expr.
11934         * gimple-ssa-warn-alloca.c (alloca_call_type): Simplify control flow.
11935         Take into account 3rd argument of __builtin_alloca_with_align_and_max.
11936         (in_loop_p): Remove first argument and useless check.
11937         (pass_walloca::execute): Remove useless test and adjust call to above.
11938         * gimple.c (gimple_build_call_from_tree): Deal with all alloc variants
11939         * gimplify.c (gimplify_vla_decl): Call build_alloca_call_expr.
11940         (gimplify_call_expr): Deal with all alloca variants.
11941         * hsa-gen.c (gen_hsa_alloca): Likewise.
11942         (gen_hsa_insns_for_call): Likewise.
11943         * ipa-pure-const.c (special_builtin_state): Likewise.
11944         * tree-chkp.c (chkp_build_returned_bound): Likewise.
11945         * tree-object-size.c (alloc_object_size): Likewise.
11946         * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
11947         (call_may_clobber_ref_p_1): Likewise.
11948         * tree-ssa-ccp.c (evaluate_stmt): Likewise.
11949         (ccp_fold_stmt): Likewise.
11950         (optimize_stack_restore): Likewise.
11951         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
11952         (mark_all_reaching_defs_necessary_1): Likewise.
11953         (propagate_necessity): Likewise.
11954         (eliminate_unnecessary_stmts): Likewise.
11955         * tree.c (build_common_builtin_nodes): Build
11956         BUILT_IN_ALLOCA_WITH_ALIGN_AND_MAX.
11957         (build_alloca_call_expr): New function.
11958         * tree.h (ALLOCA_FUNCTION_CODE_P): New macro.
11959         (CASE_BUILT_IN_ALLOCA): Likewise.
11960         (build_alloca_call_expr): Declare.
11961         * varasm.c (incorporeal_function_p): Deal with all alloca variants.
11963 2017-10-19  Eric Botcazou  <ebotcazou@adacore.com>
11965         PR debug/82509
11966         * dwarf2out.c (new_die_raw): New static inline function.
11967         (new_die): Use it to create the DIE.
11968         (add_AT_external_die_ref): Likewise.
11969         (clone_die): Likewise.
11970         (clone_as_declaration): Likewise.
11971         (dwarf2out_vms_debug_main_pointer): Likewise.
11972         (base_type_die): Likewise.  Remove early return for corner cases.
11973         Do not call add_pubtype on the DIE here.
11974         (is_base_type): Remove ERROR_MARK and return 0 for VOID_TYPE.
11975         (modified_type_die): Adjust the lookup for reverse order DIEs.  Skip
11976         typedefs for base types with DW_AT_endianity.  Make sure a DIE with
11977         native order exists for base types, attach the DIE manually and call
11978         add_pubtype on it.  Do not equate a reverse order DIE to the type.
11980 2017-10-19  Richard Earnshaw  <rearnsha@arm.com>
11982         * config/arm/arm.c (align_ok_ldrd_strd): New function.
11983         (mem_ok_for_ldrd_strd): New parameter align.  Extract the alignment of
11984         the mem into it.
11985         (gen_operands_ldrd_strd): Validate the alignment of the accesses.
11987 2017-10-19  Jakub Jelinek  <jakub@redhat.com>
11989         * flag-types.h (enum sanitize_code): Add SANITIZE_BUILTIN.  Or
11990         SANITIZE_BUILTIN into SANITIZE_UNDEFINED.
11991         * sanitizer.def (BUILT_IN_UBSAN_HANDLE_INVALID_BUILTIN,
11992         BUILT_IN_UBSAN_HANDLE_INVALID_BUILTIN_ABORT): New builtins.
11993         * opts.c (sanitizer_opts): Add builtin.
11994         * ubsan.c (instrument_builtin): New function.
11995         (pass_ubsan::execute): Call it.
11996         (pass_ubsan::gate): Enable even for SANITIZE_BUILTIN.
11997         * doc/invoke.texi: Document -fsanitize=builtin.
11999         * ubsan.c (ubsan_expand_null_ifn): Use _v1 suffixed type mismatch
12000         builtins, store max (log2 (align), 0) into uchar field instead of
12001         align into uptr field.
12002         (ubsan_expand_objsize_ifn): Use _v1 suffixed type mismatch builtins,
12003         store uchar 0 field instead of uptr 0 field.
12004         (instrument_nonnull_return): Use _v1 suffixed nonnull return builtin,
12005         instead of passing one address of struct with 2 locations pass
12006         two addresses of structs with 1 location each.
12007         * sanitizer.def (BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH,
12008         BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_ABORT,
12009         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN,
12010         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_ABORT): Removed.
12011         (BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_V1,
12012         BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_V1_ABORT,
12013         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_V1,
12014         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_V1_ABORT): New builtins.
12016 2017-10-19  Martin Liska  <mliska@suse.cz>
12018         PR driver/81829
12019         * file-find.c (remove_prefix): Remove.
12020         * file-find.h (remove_prefix): Likewise.
12021         * gcc-ar.c: Remove smartness of lookup.
12023 2017-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
12025         * config/rs6000/rs6000.md (*call_indirect_aix<mode>,
12026         *call_value_indirect_aix<mode>, *call_indirect_elfv2<mode>,
12027         *call_value_indirect_elfv2<mode>): Add correct mode to the unspec.
12029 2017-10-19  Jakub Jelinek  <jakub@redhat.com>
12031         PR target/82580
12032         * config/i386/i386.md (setcc + movzbl to xor + setcc): New peephole2.
12033         (setcc + and to xor + setcc): New peephole2.
12035 2017-10-19  Tom de Vries  <tom@codesourcery.com>
12037         * doc/sourcebuild.texi (Test Directives, Variants of
12038         dg-require-support): Add dg-require-stack-size.
12040 2017-10-19  Martin Liska  <mliska@suse.cz>
12042         PR sanitizer/82517
12043         * gimplify.c (gimplify_decl_expr): Do not instrument variables
12044         that have a large alignment.
12045         (gimplify_target_expr): Likewise.
12047 2017-10-18  Segher Boessenkool  <segher@kernel.crashing.org>
12049         PR rtl-optimization/82602
12050         * ira.c (rtx_moveable_p): Return false for volatile asm.
12052 2017-10-18  Uros Bizjak  <ubizjak@gmail.com>
12054         PR target/82580
12055         * config/i386/i386-modes.def (CCGZ): New CC mode.
12056         * config/i386/i386.md (sub<mode>3_carry_ccgz): New insn pattern.
12057         * config/i386/predicates.md (ix86_comparison_operator):
12058         Handle CCGZmode.
12059         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>:
12060         Emulate LE, LEU, GT, GTU, LT, LTU, GE and GEU double-word comparisons
12061         with double-word subtraction.
12062         (put_condition_code): Handle CCGZmode.
12064 2017-10-18  Aldy Hernandez  <aldyh@redhat.com>
12066         * wide-int.cc (debug (const wide_int &)): New.
12067         (debug (const wide_int *)): New.
12068         (debug (const widest_int &)): New.
12069         (debug (const widest_int *)): New.
12071 2017-10-18  Vladimir Makarov  <vmakarov@redhat.com>
12073         PR middle-end/82556
12074         * lra-constraints.c (curr_insn_transform): Use non-input operand
12075         instead of output one for matched reload.
12077 2017-10-18  Bin Cheng  <bin.cheng@arm.com>
12079         * tree-loop-distribution.c (INCLUDE_ALGORITHM): New header file.
12080         (tree-ssa-loop-ivopts.h): New header file.
12081         (struct builtin_info): New fields.
12082         (classify_builtin_1): Compute and record base and offset parts for
12083         memset builtin partition by calling strip_offset.
12084         (offset_cmp, fuse_memset_builtins): New functions.
12085         (finalize_partitions): Fuse adjacent memset partitions by calling
12086         above function.
12087         * tree-ssa-loop-ivopts.c (strip_offset): Delete static declaration.
12088         Expose the interface.
12089         * tree-ssa-loop-ivopts.h (strip_offset): New declaration.
12091 2017-10-18  Bin Cheng  <bin.cheng@arm.com>
12093         PR tree-optimization/82574
12094         * tree-loop-distribution.c (find_single_drs): New parameter.  Check
12095         that data reference must be executed exactly once per iteration
12096         against the outermost loop in nest.
12097         (classify_partition): Update call to above function.
12099 2017-10-18  Richard Biener  <rguenther@suse.de>
12101         PR tree-optimization/82591
12102         * graphite.c (graphite_transform_loops): Move code gen message
12103         printing ...
12104         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
12105         Here.  Handle scop_to_isl_ast failing.
12106         (scop_to_isl_ast): Limit the number of ISL operations.
12108 2017-10-18  Richard Biener  <rguenther@suse.de>
12110         * graphite-isl-ast-to-gimple.c
12111         (translate_isl_ast_to_gimple::set_rename): Simplify.
12112         (translate_isl_ast_to_gimple::set_rename_for_each_def): Inline...
12113         (graphite_copy_stmts_from_block): ... here.
12114         (copy_bb_and_scalar_dependences): Simplify.
12115         (add_parameters_to_ivs_params): Canonicalize.
12116         (generate_entry_out_of_ssa_copies): Simplify.
12117         * graphite-sese-to-poly.c (extract_affine_name): Simplify
12118         by passing in ISL dimension.
12119         (parameter_index_in_region_1): Rename to ...
12120         (parameter_index_in_region): ... this.
12121         (extract_affine): Adjust assert, pass down parameter index.
12122         (add_param_constraints): Use range-info when available.
12123         (build_scop_context): Adjust.
12124         * sese.c (new_sese_info): Adjust.
12125         (free_sese_info): Likewise.
12126         * sese.h (bb_map_t, rename_map_t, phi_rename, init_back_edge_pair_t):
12127         Remove unused typedefs.
12128         (struct sese_info_t): Simplify rename_map, remove incomplete_phis.
12130 2017-10-18  Martin Liska  <mliska@suse.cz>
12132         * combine.c (simplify_compare_const): Add gcc_fallthrough.
12134 2017-10-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
12136         * config/s390/s390.c (s390_bb_fallthru_entry_likely): New function.
12137         (s390_sched_init): Do not reset s390_sched_state if we entered the
12138         current basic block via a fallthru edge and all others are unlikely.
12140 2017-10-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
12142         * config/s390/s390.c (NUM_SIDES): New variable.
12143         (LONGRUNNING_THRESHOLD): New variable.
12144         (LATENCY_FACTOR): New variable.
12145         (s390_sched_score): Decrease score for long-running instructions on
12146         wrong side.
12147         (s390_sched_variable_issue): Perform bookkeeping for long-running
12148         instructions.
12150 2017-10-18  Richard Biener  <rguenther@suse.de>
12152         * graphite-isl-ast-to-gimple.c (gcc_expression_from_isl_ast_expr_id):
12153         Simplify with removal of the parameter rename map.
12154         (set_rename): Likewise.
12155         (should_copy_to_new_region): Likewise.
12156         (graphite_copy_stmts_from_block): Likewise.
12157         (copy_bb_and_scalar_dependences): Remove initialization of
12158         unused copied_bb_map.
12159         (copy_def): Remove.
12160         (copy_internal_parameters): Likewise.
12161         (graphite_regenerate_ast_isl): Do not call copy_internal_parameters.
12162         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
12163         Use INTEGRAL_TYPE_P.
12164         (parameter_index_in_region_1): Rename to ...
12165         (assign_parameter_index_in_region): ... this.  Assert we have
12166         a parameter we handle.
12167         (scan_tree_for_params): Adjust.
12168         * sese.h (parameter_rename_map_t): Remove.
12169         (struct sese_info_t): Remove unused parameter_rename_map and
12170         copied_bb_map members.
12171         * sese.c (new_sese_info): Adjust.
12172         (free_sese_info): Likewise.
12174 2017-10-18  Martin Liska  <mliska@suse.cz>
12176         PR sanitizer/82545
12177         * asan.c (asan_expand_poison_ifn): Do not put gimple stmt
12178         on an abnormal edge.
12180 2017-10-18  Sebastian Huber  <sebastian.huber@embedded-brains.de>
12182         * doc/invoke.texi (ffunction-sections and fdata-sections):
12183         Update.
12185 2017-10-17  Eric Botcazou  <ebotcazou@adacore.com>
12187         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Bail out only if
12188         the use statement can throw internally.
12190 2017-10-17  Eric Botcazou  <ebotcazou@adacore.com>
12192         * config/visium/visium.c (visium_select_cc_mode): Return CCmode for
12193         any RTX present on the RHS of a SET.
12194         * compare-elim.c (try_eliminate_compare): Restore comment.
12196 2017-10-17  Jakub Jelinek  <jakub@redhat.com>
12198         * langhooks.h (struct lang_hooks): Document that tree_size langhook
12199         may be also called on tcc_type nodes.
12200         * langhooks.c (lhd_tree_size): Likewise.
12202 2017-10-17  David Malcolm  <dmalcolm@redhat.com>
12204         * gimple-ssa-sprintf.c (fmtwarn): Update for changed signature of
12205         format_warning_at_substring.
12206         (maybe_warn): Convert source_range * param to a location_t.  Pass
12207         UNKNOWN_LOCATION rather than NULL to fmtwarn.
12208         (format_directive): Remove code to extract source_ranges and
12209         source_range * in favor of just a location_t.
12210         (parse_directive): Pass UNKNOWN_LOCATION rather than NULL to
12211         fmtwarn.
12212         * substring-locations.c (format_warning_va): Convert
12213         source_range * param to a location_t.
12214         (format_warning_at_substring): Likewise.
12215         * substring-locations.h (format_warning_va): Likewise.
12216         (format_warning_at_substring): Likewise.
12218 2017-10-17  Jan Hubicka  <hubicka@ucw.cz>
12220         * target.h (enum vect_cost_for_stmt): Add vec_gather_load and
12221         vec_scatter_store
12222         * tree-vect-stmts.c (record_stmt_cost): Make difference between normal
12223         and scatter/gather ops.
12225         * aarch64/aarch64.c (aarch64_builtin_vectorization_cost): Add
12226         vec_gather_load and vec_scatter_store.
12227         * arm/arm.c (arm_builtin_vectorization_cost): Likewise.
12228         * powerpcspe/powerpcspe.c (rs6000_builtin_vectorization_cost): Likewise.
12229         * rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Likewise.
12230         * s390/s390.c (s390_builtin_vectorization_cost): Likewise.
12231         * spu/spu.c (spu_builtin_vectorization_cost): Likewise.
12232         * i386/i386.c (x86_builtin_vectorization_cost): Likewise.
12234 2017-10-17  Uros Bizjak  <ubizjak@gmail.com>
12236         * reg-stack.c (compare_for_stack_reg): Add bool argument.
12237         Detect FTST instruction and handle its register pops.  Only pop
12238         second operand if can_pop_second_op is true.
12239         (subst_stack_regs_pat) <case COMPARE>: Detect FCOMI instruction to
12240         set can_pop_second_op to false in the compare_for_stack_reg call.
12242         * config/i386/i386.md (*cmpi<FPCMP:unord><MODEF:mode>): Only call
12243         output_fp_compare for stack register operands.
12244         * config/i386/i386.c (output_fp_compare): Do not output SSE compare
12245         instructions here.  Do not emit stack register pops here.  Assert
12246         that FCOMPP pops next to top stack register.  Rewrite function.
12248 2017-10-17  Nathan Sidwell  <nathan@acm.org>
12250         PR middle-end/82577
12251         * alias.c (compare_base_decls): Check HAS_DECL_ASSEMBLER_NAME_P,
12252         use DECL_ASSEMBLER_NAME_RAW.
12254         PR middle-end/82546
12255         * tree.c (tree_code_size): Reformat.  Punt to lang hook for unknown
12256         TYPE nodes.
12258 2017-10-17  Qing Zhao <qing.zhao@oracle.com>
12259             Wilco Dijkstra <wilco.dijkstra@arm.com>
12261         * builtins.c (expand_builtin_update_setjmp_buf): Add a
12262         converstion to Pmode from the buf_addr.
12264 2017-10-17  Richard Biener  <rguenther@suse.de>
12266         * graphite-dependences.c (scop_get_reads_and_writes): Change
12267         output parameters to references.
12269 2017-10-17  Jackson Woodruff  <jackson.woodruff@arm.com>
12271         PR 71026/tree-optimization
12272         * fold-const.c (distribute_real_division): Removed.
12273         (fold_binary_loc): Remove calls to distribute_real_divison.
12275 2017-10-17  Richard Biener  <rguenther@suse.de>
12277         * graphite-scop-detection.c
12278         (scop_detection::stmt_has_simple_data_refs_p): Always use
12279         the full nest as region.
12280         (try_generate_gimple_bb): Likewise.
12281         * sese.c (scalar_evolution_in_region): Simplify now that
12282         SCEV can handle instantiation in regions.
12283         * tree-scalar-evolution.c (instantiate_scev_name): Also instantiate
12284         in the non-loop part of a function if requested.
12286 2017-10-17  Richard Biener  <rguenther@suse.de>
12288         PR tree-optimization/82563
12289         * graphite-isl-ast-to-gimple.c (generate_entry_out_of_ssa_copies):
12290         New function.
12291         (graphite_regenerate_ast_isl): Call it.
12292         * graphite-scop-detection.c (build_scops): Remove entry edge split.
12294 2017-10-17  Jakub Jelinek  <jakub@redhat.com>
12296         PR tree-optimization/82549
12297         * fold-const.c (optimize_bit_field_compare, fold_truth_andor_1):
12298         Formatting fixes.  Instead of calling make_bit_field_ref with negative
12299         bitpos return 0.
12301 2017-10-17  Olga Makhotina  <olga.makhotina@intel.com>
12303         * config/i386/avx512dqintrin.h (_mm_mask_reduce_sd,
12304         _mm_maskz_reduce_sd, _mm_mask_reduce_ss,=20
12305         _mm_maskz_reduce_ss): New.
12306         * config/i386/i386-builtin.def (__builtin_ia32_reducesd_mask,
12307         __builtin_ia32_reducess_mask): Ditto..
12308         (__builtin_ia32_reducesd, __builtin_ia32_reducess): Remove.
12309         * config/i386/sse.md (reduces<mode>): Renamed to ...
12310         (reduces<mode><mask_scalar_name>): ... this.
12311         (vreduce<ssescalarmodesuffix>\t{%3, %2, %1, %0|%0, %1, %2, %3}):
12312         Changed to ...
12313         (vreduce<ssescalarmodesuffix>\t{%3, %2, %1, %0<mask_scalar_operand4>|
12314         %0<mask_scalar_operand4>, %1, %2, %3}): ... this.
12316 2017-10-16  David Malcolm  <dmalcolm@redhat.com>
12318         * Makefile.in (OBJS): Add unique-ptr-tests.o.
12319         * selftest-run-tests.c (selftest::run_tests): Call
12320         selftest::unique_ptr_tests_cc_tests.
12321         * selftest.h (selftest::unique_ptr_tests_cc_tests): New decl.
12322         * unique-ptr-tests.cc: New file.
12324 2017-10-16  Vladimir Makarov  <vmakarov@redhat.com>
12326         PR sanitizer/82353
12327         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12328         locations.
12329         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12330         (make_hard_regno_born, make_hard_regno_dead): Update
12331         bb_killed_pseudos and bb_gen_pseudos for fixed regs.
12333 2017-10-16  Jeff Law  <law@redhat.com>
12335         * tree-ssa-dse.c (live_bytes_read): Fix thinko.
12337 2017-10-16  Jan Hubicka  <hubicka@ucw.cz>
12339         * x86-tune-costs.h (znver1_cost): Fix move cost tables.
12341 2017-10-16  Olivier Hainque  <hainque@adacore.com>
12343         * gcc/config.gcc (powerpc*-*-*spe*): Pick 8548 as the default
12344         with_cpu if we were configured for an e500v2 target cpu name.
12346 2017-10-16  Thomas Preud'homme  <thomas.preudhomme@arm.com>
12348         * config/arm/arm-cpus.in (cortex-m33): Add nodsp option.
12349         * doc/invoke.texi: Document +nodsp as a valid extension for
12350         -mcpu=cortex-m33.
12352 2017-10-16  Martin Liska  <mliska@suse.cz>
12354         * sbitmap.c (bitmap_bit_in_range_p_checking): New function.
12355         (test_set_range): Likewise.
12356         (test_range_functions): Rename to ...
12357         (test_bit_in_range): ... this.
12358         (sbitmap_c_tests): Add new test.
12360 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12362         * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
12363         New.
12364         (vdot_lane_u32, vdot_laneq_u32, vdotq_lane_u32, vdotq_laneq_u32): New.
12365         (vdot_lane_s32, vdot_laneq_s32, vdotq_lane_s32, vdotq_laneq_s32): New.
12367 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12369         * config/aarch64/aarch64-builtins.c
12370         (aarch64_types_quadopu_lane_qualifiers): New.
12371         (TYPES_QUADOPU_LANE): New.
12372         * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): New.
12373         (<sur>dot_prod<vsi2qi>, aarch64_<sur>dot_lane<vsi2qi>): New.
12374         (aarch64_<sur>dot_laneq<vsi2qi>): New.
12375         * config/aarch64/aarch64-simd-builtins.def (sdot, udot): New.
12376         (sdot_lane, udot_lane, sdot_laneq, udot_laneq): New.
12377         * config/aarch64/iterators.md (sur): Add UNSPEC_SDOT, UNSPEC_UDOT.
12378         (Vdottype, DOTPROD): New.
12379         (sur): Add SDOT and UDOT.
12381 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12383         * config/aarch64/aarch64.h (AARCH64_FL_DOTPROD): New.
12384         (AARCH64_ISA_DOTPROD, TARGET_DOTPROD): New.
12385         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12386         Add TARGET_DOTPROD.
12387         * config/aarch64/aarch64-option-extensions.def (dotprod): New.
12388         * config/aarch64/aarch64-cores.def (cortex-a55, cortex-a75):
12389         Enable TARGET_DOTPROD.
12390         (cortex-a75.cortex-a55): Likewise.
12391         * doc/invoke.texi (aarch64-feature-modifiers): Document dotprod.
12393 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12395         * config/arm/arm-builtins.c (arm_unsigned_uternop_qualifiers): New.
12396         (UTERNOP_QUALIFIERS, arm_umac_lane_qualifiers, UMAC_LANE_QUALIFIERS):
12397         New.
12398         * config/arm/arm_neon_builtins.def (sdot, udot, sdot_lane, udot_lane):
12399         New.
12400         * config/arm/iterators.md (DOTPROD, VSI2QI, vsi2qi): New.
12401         (UNSPEC_DOT_S, UNSPEC_DOT_U, opsuffix): New.
12402         * config/arm/neon.md (neon_<sup>dot<vsi2qi>): New.
12403         (neon_<sup>dot_lane<vsi2qi>, <sup>dot_prod<vsi2qi>): New.
12404         * config/arm/types.md (neon_dot, neon_dot_q): New.
12405         * config/arm/unspecs.md (sup): Add UNSPEC_DOT_S, UNSPEC_DOT_U.
12407 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12409         * config/arm/arm.h (TARGET_DOTPROD): New.
12410         * config/arm/arm.c (arm_arch_dotprod): New.
12411         (arm_option_reconfigure_globals): Add arm_arch_dotprod.
12412         * config/arm/arm-c.c (__ARM_FEATURE_DOTPROD): New.
12413         * config/arm/arm-cpus.in (armv8.2-a): Enabled +dotprod.
12414         (feature dotprod, group dotprod, ALL_SIMD_INTERNAL): New.
12415         (ALL_FPU_INTERNAL): Use ALL_SIMD_INTERNAL.
12416         * config/arm/t-multilib (v8_2_a_simd_variants): Add dotprod.
12417         * doc/invoke.texi (armv8.2-a): Document dotprod
12419 2017-10-14  Jan Hubicka  <hubicka@ucw.cz>
12421         * i386.c (ix86_vec_cost): New function.
12422         (ix86_rtx_costs): Handle vector operations better.
12423         * i386.h (struct processor_costs): Add sse_op, fmasd, fmass.
12424         * x86-tune-costs.h: Add new costs to all tables.
12426 2017-10-14  Jan Hubicka  <hubicka@ucw.cz>
12428         * i386.c (ix86_rtx_costs): Make difference between x87 and SSE
12429         operations.
12430         * i386.h (struct processor_costs): Add addss, mulss, mulsd, divss,
12431         divsd, sqrtss and sqrtsd
12432         * x86-tune-costs.h: Add new entries to all costs.
12433         (znver1_cost): Fix to match real instruction latencies.
12435 2017-10-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12436             Michael Collison <michael.collison@arm.com>
12438         * compare-elim.c: Include emit-rtl.h.
12439         (can_merge_compare_into_arith): New function.
12440         (try_validate_parallel): Likewise.
12441         (try_merge_compare): Likewise.
12442         (try_eliminate_compare): Call the above when no previous clobber
12443         is available.
12444         (execute_compare_elim_after_reload): Add DF_UD_CHAIN and DF_DU_CHAIN
12445         dataflow problems.
12447 2017-10-14  Jakub Jelinek  <jakub@redhat.com>
12449         PR middle-end/62263
12450         PR middle-end/82498
12451         * tree-ssa-phiopt.c (value_replacement): Comment fix.  Handle
12452         up to 2 preparation statements for ASSIGN in MIDDLE_BB.
12454         PR middle-end/62263
12455         PR middle-end/82498
12456         * tree-ssa-forwprop.c (simplify_rotate): Allow def_arg1[N]
12457         to be any operand_equal_p operands.  For & (B - 1) require
12458         B to be power of 2.  Recognize
12459         (X << (Y & (B - 1))) | (X >> ((-Y) & (B - 1))) and similar patterns.
12461 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12463         PR bootstrap/82553
12464         * optabs.c (expand_memory_blockage): Fix call of
12465         targetm.have_memory_blockage.
12467 2017-10-14  Jakub Jelinek  <jakub@redhat.com>
12469         PR bootstrap/82548
12470         * config.gcc (*-*-solaris2*, i[34567]86-*-cygwin*,
12471         x86_64-*-cygwin*, i[34567]86-*-mingw* | x86_64-*-mingw*): Append
12472         objects to extra_objs instead of overwriting it.
12474 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12476         * config/i386/sync.md (FILD_ATOMIC/FIST_ATOMIC FP load peephole2):
12477         Use any_fp_register_operand as operand[3] predicate.  Simplify
12478         equality test for operands[2] and operands[4] memory location.
12479         (LDX_ATOMIC/STX_ATOMIC FP load peephole2): Ditto.
12480         (FILD_ATOMIC/FIST_ATOMIC FP load peephole2 with mem blockage): New.
12481         (LDX_ATOMIC/LDX_ATOMIC FP load peephole2 with mem blockage): Ditto.
12482         (FILD_ATOMIC/FIST_ATOMIC FP store peephole2): Use
12483         any_fp_register_operand as operand[1] predicate.  Simplify
12484         equality test for operands[0] and operands[3] memory location.
12485         (LDX_ATOMIC/STX_ATOMIC FP store peephole2): Ditto.
12486         (FILD_ATOMIC/FIST_ATOMIC FP store peephole2 with mem blockage): New.
12487         (LDX_ATOMIC/LDX_ATOMIC FP storepeephole2 with mem blockage): Ditto.
12489 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12491         * target-insns.def: Add memory_blockage.
12492         * optabs.c (expand_memory_blockage): New function.
12493         (expand_asm_memory_barrier): Rename ...
12494         (expand_asm_memory_blockage): ... to this.
12495         (expand_mem_thread_fence): Call expand_memory_blockage
12496         instead of expand_asm_memory_barrier.
12497         (expand_mem_singnal_fence): Ditto.
12498         (expand_atomic_load): Ditto.
12499         (expand_atomic_store): Ditto.
12500         * doc/md.texi (Standard Pattern Names For Generation):
12501         Document memory_blockage instruction pattern.
12503 2017-10-13  Sebastian Perta  <sebastian.perta@renesas.com>
12505         * config/rl78/rl78.c (rl78_emit_libcall): New function.
12506         * config/rl78/rl78-protos.h (rl78_emit_libcall): New function.
12507         * config/rl78/rl78.md: New define_expand "adddi3".
12509 2017-10-13  Jan Hubicka  <hubicka@ucw.cz>
12511         * cfghooks.c (verify_flow_info): Disable check that all probabilities
12512         are set correctly.
12514 2017-10-13  Jeff Law  <law@redhat.com>
12516         * tree-ssa-reassoc.c (reassociate_bb): Clarify code slighly.
12518 2017-10-13  Jakub Jelinek  <jakub@redhat.com>
12520         PR target/82274
12521         * internal-fn.c (expand_mul_overflow): If both operands have
12522         the same highpart of -1 or 0 and the topmost bit of lowpart
12523         is different, overflow is if res <= 0 rather than res < 0.
12525 2017-10-13  Pat Haugen  <pthaugen@us.ibm.com>
12527         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Remove
12528         TARGET_P9_VECTOR code for unaligned_load case.
12530 2017-10-13  Jan Hubicka  <hubicka@ucw.cz>
12532         * cfghooks.c (verify_flow_info): Check that edge probabilities are set.
12534 2017-10-13  Nathan Sidwell  <nathan@acm.org>
12536         * tree-core.h (tree_contains_struct): Make bool.
12537         * tree.c (tree_contains_struct): Likewise.
12538         * tree.h (MARK_TS_BASE): Remove do ... while (0) idiom.
12539         (MARK_TS_TYPED, MARK_TS_COMMON, MARK_TS_TYPE_COMMON,
12540         MARK_TS_TYPE_WITH_LANG_SPECIFIC, MARK_TS_DECL_MINIMAL,
12541         MARK_TS_DECL_COMMON, MARK_TS_DECL_WRTL, MARK_TS_DECL_WITH_VIS,
12542         MARK_TS_DECL_NON_COMMON): Likewise, use comma operator.
12544 2017-10-13  Richard Biener  <rguenther@suse.de>
12546         * graphite-isl-ast-to-gimple.c
12547         (translate_isl_ast_to_gimple::get_rename_from_scev): Remove unused
12548         parameters and dominance check.
12549         (translate_isl_ast_to_gimple::graphite_copy_stmts_from_block): Adjust.
12550         (translate_isl_ast_to_gimple::copy_bb_and_scalar_dependences): Likewise.
12551         (translate_isl_ast_to_gimple::graphite_regenerate_ast_isl):
12552         Do not update SSA form here or do intermediate IL verification.
12553         * graphite.c: Include tree-ssa.h and tree-into-ssa.h.
12554         (graphite_initialize): Remove check on the number of loops in
12555         the function and inline into graphite_transform_loops.
12556         (graphite_finalize): Inline into graphite_transform_loops.
12557         (graphite_transform_loops): Perform SSA update and IL verification
12558         here.
12559         * params.def (PARAM_GRAPHITE_MIN_LOOPS_PER_FUNCTION): Remove.
12561 2017-10-13  Richard Biener  <rguenther@suse.de>
12563         * graphite-isl-ast-to-gimple.c (max_mode_int_precision,
12564         graphite_expression_type_precision): Avoid global constructor
12565         by moving ...
12566         (translate_isl_ast_to_gimple::translate_isl_ast_to_gimple): Here.
12567         (translate_isl_ast_to_gimple::graphite_expr_type): Add type member.
12568         (translate_isl_ast_to_gimple::translate_isl_ast_node_for): Use it.
12569         (translate_isl_ast_to_gimple::build_iv_mapping): Likewise.
12570         (translate_isl_ast_to_gimple::graphite_create_new_guard): Likewise.
12571         * graphite-sese-to-poly.c (build_original_schedule): Return nothing.
12573 2017-10-13  H.J. Lu  <hongjiu.lu@intel.com>
12575         PR target/82499
12576         * config/i386/i386.h (ix86_red_zone_size): New.
12577         * config/i386/i386.md (push peephole2s): Replace
12578         "!ix86_using_red_zone ()" with "ix86_red_zone_size == 0".
12580 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12581             Alan Hayward  <alan.hayward@arm.com>
12582             David Sherwood  <david.sherwood@arm.com>
12584         * combine.c (can_change_dest_mode): Reject changes in
12585         REGMODE_NATURAL_SIZE.
12587 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12588             Alan Hayward  <alan.hayward@arm.com>
12589             David Sherwood  <david.sherwood@arm.com>
12591         * cfgexpand.c (expand_debug_expr): Use GET_MODE_UNIT_BITSIZE.
12592         (expand_debug_source_expr): Likewise.
12593         * combine.c (combine_simplify_rtx): Likewise.
12594         * cse.c (fold_rtx): Likewise.
12595         * fwprop.c (canonicalize_address): Likewise.
12596         * targhooks.c (default_shift_truncation_mask): Likewise.
12598 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12599             Alan Hayward  <alan.hayward@arm.com>
12600             David Sherwood  <david.sherwood@arm.com>
12602         * optabs.c (add_equal_note): Use GET_MODE_UNIT_SIZE.
12603         (widened_mode): Likewise.
12604         (expand_unop): Likewise.
12605         * ree.c (transform_ifelse): Likewise.
12606         (merge_def_and_ext): Likewise.
12607         (combine_reaching_defs): Likewise.
12608         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12610 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12611             Alan Hayward  <alan.hayward@arm.com>
12612             David Sherwood  <david.sherwood@arm.com>
12614         * caller-save.c (replace_reg_with_saved_mem): Use byte_lowpart_offset.
12615         * combine.c (gen_lowpart_for_combine): Likewise.
12616         * dwarf2out.c (rtl_for_decl_location): Likewise.
12617         * final.c (alter_subreg): Likewise.
12618         * rtlhooks.c (gen_lowpart_general): Likewise.
12619         (gen_lowpart_if_possible): Likewise.
12621 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12622             Alan Hayward  <alan.hayward@arm.com>
12623             David Sherwood  <david.sherwood@arm.com>
12625         * calls.c (expand_call): Use subreg_lowpart_offset.
12626         * cse.c (cse_insn): Likewise.
12627         * regcprop.c (copy_value): Likewise.
12628         (copyprop_hardreg_forward_1): Likewise.
12630 2017-10-13  Jakub Jelinek  <jakub@redhat.com>
12632         PR target/82524
12633         * config/i386/i386.md (addqi_ext_1, andqi_ext_1,
12634         *andqi_ext_1_cc, *<code>qi_ext_1, *xorqi_ext_1_cc): Change
12635         =Q constraints to +Q and into insn condition add check
12636         that operands[0] and operands[1] are equal.
12637         (*addqi_ext_2, *andqi_ext_2, *<code>qi_ext_2): Change
12638         =Q constraints to +Q and into insn condition add check
12639         that operands[0] is equal to either operands[1] or operands[2].
12641         PR target/82498
12642         * fold-const.c (fold_binary_loc) <bit_rotate>: Code cleanups,
12643         instead of handling MINUS_EXPR twice (once for each argument),
12644         canonicalize operand order and handle just once, use rtype where
12645         possible.  Handle (A << B) | (A >> (-B & (Z - 1))).
12647         PR target/82498
12648         * config/i386/ia32intrin.h (__rold, __rord, __rolq, __rorq): Allow
12649         any values of __C while still being pattern recognizable as a simple
12650         rotate instruction.
12652 2017-10-13  Richard Biener  <rguenther@suse.de>
12654         PR tree-optimization/82451
12655         Revert
12656         2017-10-02  Richard Biener  <rguenther@suse.de>
12658         PR tree-optimization/82355
12659         * graphite-isl-ast-to-gimple.c (build_iv_mapping): Also build
12660         a mapping for the enclosing loop but avoid generating one for
12661         the loop tree root.
12662         (copy_bb_and_scalar_dependences): Remove premature codegen
12663         error on PHIs in blocks duplicated into multiple places.
12664         * graphite-scop-detection.c
12665         (scop_detection::stmt_has_simple_data_refs_p): For a loop not
12666         in the region use it as loop and nest to analyze the DR in.
12667         (try_generate_gimple_bb): Likewise.
12668         * graphite-sese-to-poly.c (extract_affine_chrec): Adjust.
12669         (add_loop_constraints): For blocks in a loop not in the region
12670         create a dimension with a single iteration.
12671         * sese.h (gbb_loop_at_index): Remove assert.
12673         * cfgloop.c (loop_preheader_edge): For the loop tree root
12674         return the single successor of the entry block.
12675         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
12676         Reset the SCEV hashtable and niters.
12677         * graphite-scop-detection.c
12678         (scop_detection::graphite_can_represent_scev): Add SCOP parameter,
12679         assert that we only have POLYNOMIAL_CHREC that vary in loops
12680         contained in the region.
12681         (scop_detection::graphite_can_represent_expr): Adjust.
12682         (scop_detection::stmt_has_simple_data_refs_p): For loops
12683         not in the region set loop to NULL.  The nest is now the
12684         entry edge to the region.
12685         (try_generate_gimple_bb): Likewise.
12686         * sese.c (scalar_evolution_in_region): Adjust for
12687         instantiate_scev change.
12688         * tree-data-ref.h (graphite_find_data_references_in_stmt):
12689         Make nest parameter the edge into the region.
12690         (create_data_ref): Likewise.
12691         * tree-data-ref.c (dr_analyze_indices): Make nest parameter an
12692         entry edge into a region and adjust instantiate_scev calls.
12693         (create_data_ref): Likewise.
12694         (graphite_find_data_references_in_stmt): Likewise.
12695         (find_data_references_in_stmt): Pass the loop preheader edge
12696         from the nest argument.
12697         * tree-scalar-evolution.h (instantiate_scev): Make instantiate_below
12698         parameter the edge into the region.
12699         (instantiate_parameters): Use the loop preheader edge as entry.
12700         * tree-scalar-evolution.c (analyze_scalar_evolution): Handle
12701         NULL loop.
12702         (get_instantiated_value_entry): Make instantiate_below parameter
12703         the edge into the region.
12704         (instantiate_scev_name): Likewise.  Adjust dominance checks,
12705         when we cannot use loop-based instantiation instantiate by
12706         walking use-def chains.
12707         (instantiate_scev_poly): Adjust.
12708         (instantiate_scev_binary): Likewise.
12709         (instantiate_scev_convert): Likewise.
12710         (instantiate_scev_not): Likewise.
12711         (instantiate_array_ref): Remove.
12712         (instantiate_scev_3): Likewise.
12713         (instantiate_scev_2): Likewise.
12714         (instantiate_scev_1): Likewise.
12715         (instantiate_scev_r): Do not blindly handle N-operand trees.
12716         Do not instantiate array-refs.  Handle all constants and invariants.
12717         (instantiate_scev): Make instantiate_below parameter
12718         the edge into the region.
12719         (resolve_mixers): Use the loop preheader edge for the region
12720         parameter to instantiate_scev_r.
12721         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Adjust.
12723 2017-10-13  Richard Biener  <rguenther@suse.de>
12725         PR tree-optimization/82525
12726         * graphite-isl-ast-to-gimple.c
12727         (translate_isl_ast_to_gimple::widest_int_from_isl_expr_int): Split
12728         out from ...
12729         (translate_isl_ast_to_gimple::gcc_expression_from_isl_expr_int): Here.
12730         Fail code generation when we cannot represent the isl integer.
12731         (binary_op_to_tree): Elide modulo operations that are no-ops
12732         in the type we code generate.  Remove now superfluous code
12733         generation errors.
12735 2017-10-13  Richard Biener  <rguenther@suse.de>
12737         * graphite-scop-detection.c (loop_ivs_can_be_represented): Remove.
12738         (scop_detection::harmful_loop_in_region): Remove premature
12739         IV type restriction.
12740         (scop_detection::graphite_can_represent_scev): We can handle
12741         pointer IVs just fine.
12743 2017-10-13  Alan Modra  <amodra@gmail.com>
12745         * doc/extend.texi (Extended Asm <Clobbers>): Rename to
12746         "Clobbers and Scratch Registers".  Add paragraph on
12747         alternative to clobbers for scratch registers and OpenBLAS
12748         example.
12750 2017-10-13  Alan Modra  <amodra@gmail.com>
12752         * doc/extend.texi (Clobbers): Correct vax example.  Delete old
12753         example of a memory input for a string of known length.  Move
12754         commentary out of table.  Add a number of new examples
12755         covering array memory inputs.
12757 2017-10-12  Martin Liska  <mliska@suse.cz>
12759         PR tree-optimization/82493
12760         * sbitmap.c (bitmap_bit_in_range_p): Fix the implementation.
12761         (test_range_functions): New function.
12762         (sbitmap_c_tests): Likewise.
12763         * selftest-run-tests.c (selftest::run_tests): Run new tests.
12764         * selftest.h (sbitmap_c_tests): New function.
12766         * tree-ssa-dse.c (live_bytes_read): Fix thinko.
12768 2017-10-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
12770         * config/rs6000/amo.h: Fix spacing issue.
12772 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12774         PR target/82498
12775         * config/i386/i386.md (*ashl<mode>3_mask_1,
12776         *<shift_insn><mode>3_mask_1, *<rotate_insn><mode>3_mask_1,
12777         *<btsc><mode>_mask_1, *btr<mode>_mask_1): New define_insn_and_split
12778         patterns.
12780 2017-10-12  Jan Hubicka  <hubicka@ucw.cz>
12782         * profile-count.h (safe_scale_64bit): Fix GCC4.x path.
12783         (profile_probability): Set max_probability
12784         to (uint32_t) 1 << (n_bits - 2) and update accessors to avoid overlfows
12785         in temporaries.
12786         * profile-count.c (profile_probability::differs_from_p): Do not
12787         rely on max_probaiblity == 10000
12789 2017-10-12  Jeff Law  <law@redhat.com>
12791         * tree-ssa-dse.c (valid_ao_ref_for_dse): Reject ao_refs with
12792         negative offsets.
12794 2017-10-12  Martin Sebor  <msebor@redhat.com>
12796         PR other/82301
12797         PR c/82435
12798         * cgraphunit.c (maybe_diag_incompatible_alias): New function.
12799         (handle_alias_pairs): Call it.
12800         * common.opt (-Wattribute-alias): New option.
12801         * doc/extend.texi (ifunc attribute): Discuss C++ specifics.
12802         * doc/invoke.texi (-Wattribute-alias): Document.
12804 2017-10-12  Vladimir Makarov  <vmakarov@redhat.com>
12806         Revert
12807         2017-10-11  Vladimir Makarov  <vmakarov@redhat.com>
12808         PR sanitizer/82353
12809         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12810         locations.
12811         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12812         (make_hard_regno_born, make_hard_regno_dead): Update
12813         bb_killed_pseudos and bb_gen_pseudos.
12815 2017-10-12  Jan Hubicka  <hubicka@ucw.cz>
12817         * config/i386/x86-tune-sched.c (ix86_adjust_cost): Fix Zen support.
12819 2017-10-12  Uros Bizjak  <ubizjak@gmail.com>
12821         * config/alpha/alpha.c (alpha_split_conditional_move):
12822         Use std::swap instead of manually swapping.
12823         (alpha_stdarg_optimize_hook): Ditto.
12824         (alpha_canonicalize_comparison): Ditto.
12826 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12828         * tree-loop-distribution.c (struct builtin_info): New struct.
12829         (struct partition): Refactor fields into struct builtin_info.
12830         (partition_free): Free struct builtin_info.
12831         (build_size_arg_loc, build_addr_arg_loc): Delete.
12832         (generate_memset_builtin, generate_memcpy_builtin): Get memory range
12833         information from struct builtin_info.
12834         (find_single_drs): New function refactored from classify_partition.
12835         Also moved builtin validity checks to this function.
12836         (compute_access_range, alloc_builtin): New functions.
12837         (classify_builtin_st, classify_builtin_ldst): New functions.
12838         (classify_partition): Refactor code into functions find_single_drs,
12839         classify_builtin_st and classify_builtin_ldst.
12840         (distribute_loop): Don't do runtime alias check when distributing
12841         loop nest.
12842         (find_seed_stmts_for_distribution): New function.
12843         (pass_loop_distribution::execute): Refactor code finding seed
12844         stmts into above function.  Support distribution for the innermost
12845         two-level loop nest.  Adjust dump information.
12847 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12849         * tree-loop-distribution.c: Adjust the general comment.
12850         (NUM_PARTITION_THRESHOLD): New macro.
12851         (ssa_name_has_uses_outside_loop_p): Support loop nest distribution.
12852         (classify_partition): Skip builtin pattern of loop nest's inner loop.
12853         (merge_dep_scc_partitions): New parameter ignore_alias_p and use it
12854         in call to build_partition_graph.
12855         (finalize_partitions): New parameter.  Make loop distribution more
12856         conservative by fusing more partitions.
12857         (distribute_loop): Don't do runtime alias check in case of loop nest
12858         distribution.
12859         (find_seed_stmts_for_distribution): New function.
12860         (prepare_perfect_loop_nest): New function.
12861         (pass_loop_distribution::execute): Refactor code finding seed stmts
12862         and loop nest into above functions.  Support loop nest distribution.
12863         Adjust dump information accordingly.
12865 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12867         * tree-loop-distribution.c (break_alias_scc_partitions): Add comment
12868         and set PTYPE_SEQUENTIAL for merged partition.
12870 2017-10-12  Richard Biener  <rguenther@suse.de>
12872         PR tree-optimization/69728
12873         Revert
12874         2017-09-19  Richard Biener  <rguenther@suse.de>
12876         PR tree-optimization/69728
12877         * graphite-sese-to-poly.c (schedule_error): New global.
12878         (add_loop_schedule): Handle empty domain by failing the
12879         schedule.
12880         (build_original_schedule): Handle schedule_error.
12882         * graphite-sese-to-poly.c (add_loop_schedule): Handle empty
12883         domain by returning an unchanged schedule.
12885 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12887         * genrecog.c (validate_pattern): For VEC_SELECT verify that
12888         CONST_INT selectors are 0 to GET_MODE_NUNITS (imode) - 1.
12890 2017-10-12  Aldy Hernandez  <aldyh@redhat.com>
12892         * Makefile.in (TAGS): Merge all the *.def files into one pattern.
12893         Handle params.def.
12895 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12897         PR c++/82159
12898         * expr.c (store_field): Don't optimize away bitsize == 0 store
12899         from CALL_EXPR with addressable return type.
12901 2017-10-11  Segher Boessenkool  <segher@kernel.crashing.org>
12903         * config/rs6000/rs6000.h (TARGET_ISEL64): Delete.
12904         * config/rs6000/rs6000.md (sel): Delete mode attribute.
12905         (mov<mode>cc, isel_signed_<mode>, isel_unsigned_<mode>,
12906         *isel_reversed_signed_<mode>, *isel_reversed_unsigned_<mode>): Use
12907         TARGET_ISEL instead of TARGET_ISEL<sel>.
12909 2017-10-11  David Edelsohn  <dje.gcc@gmail.com>
12911         * config/rs6000/rs6000.c
12912         (rs6000_xcoff_asm_output_aligned_decl_common): Test for NULL decl.
12914 2017-10-11  Segher Boessenkool  <segher@kernel.crashing.org>
12916         * config/rs6000/predicates.md (zero_constant, all_ones_constant):
12917         Move up in file.
12918         (reg_or_cint_operand): Fix comment.
12919         (reg_or_zero_operand): New predicate.
12920         * config/rs6000/rs6000-protos.h (output_isel): Delete.
12921         * config/rs6000/rs6000.c (output_isel): Delete.
12922         * config/rs6000/rs6000.md (isel_signed_<mode>): Use reg_or_zero_operand
12923         instead of reg_or_cint_operand.  Output instruction directly (not via
12924         output_isel).
12925         (isel_unsigned_<mode>): Ditto.
12926         (*isel_reversed_signed_<mode>): Use reg_or_zero_operand instead of
12927         gpc_reg_operand.  Add an instruction alternative for this.  Output
12928         instruction directly.
12929         (*isel_reversed_unsigned_<mode>): Ditto.
12931 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12933         * config/i386/i386.c (ix86_canonicalize_comparison): New function.
12934         (TARGET_CANONICALIZE_COMPARISON): Define.
12936 2017-10-11  Qing Zhao  <qing.zhao@oracle.com>
12938         PR target/81422
12939         * config/aarch64/aarch64.c (aarch64_load_symref_appropriately):
12940         Check whether the dest is REG before adding REG_EQUIV note.
12942 2017-10-11  Vladimir Makarov  <vmakarov@redhat.com>
12944         PR sanitizer/82353
12945         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12946         locations.
12947         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12948         (make_hard_regno_born, make_hard_regno_dead): Update
12949         bb_killed_pseudos and bb_gen_pseudos.
12951 2017-10-11  Nathan Sidwell  <nathan@acm.org>
12953         * incpath.h (enum incpath_kind): Name enum, prefix values.
12954         (add_path, add_cpp_dir_path, get_added_cpp_dirs): Use incpath_kind.
12955         * incpath.c (heads, tails): Use INC_MAX.
12956         (add_env_var_paths, add_standard_paths): Use incpath_kind.
12957         (merge_include_chains, split_quote_chain,
12958         register_include_chains): Update incpath_kind names.
12959         (add_cpp_dir_path, add_path, get_added_cpp_dirs): Use incpath_kind.
12960         * config/darwin-c.c (add_system_framework_path): Update incpath_kind
12961         names.
12962         (add_framework_path, darwin_register_objc_includes): Likewise.
12963         * config/vms/vms-c.c (vms_c_register_includes): Likewise.
12965 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12967         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
12968         Do not use float_operator operator predicate.
12969         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
12970         * config/i386/predicates.md (float_operator): Remove predicate.
12972 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12974         * config/i386/i386.md (*jcc<mode>_0_i387): Remove insn pattern.
12975         (*jccxf_i387): Ditto.
12976         (*jcc<mode>_i387): Ditto.
12977         (*jccu<mode>_i387): Ditto.
12978         (*jcc<X87MODEF:mode>_<SWI24:mode>_i387): Ditto.
12979         (*jcc_*_i387 splitters): Remove.
12980         * config/i386/i386-protos.h (ix86_split_fp_branch): Remove prototype.
12981         * config/i386/i386.c (ix86_split_fp_branch): Remove.
12982         * config/i386/predicates.md (ix86_swapped_fp_comparison_operator):
12983         Remove predicate.
12985 2017-10-11  Jan Hubicka  <hubicka@ucw.cz>
12987         * profile-count.h (slow_safe_scale_64bit): New function.
12988         (safe_scale_64bit): New inline.
12989         (profile_count::max_safe_multiplier): Remove; use safe_scale_64bit.
12990         * profile-count.c: Include wide-int.h
12991         (slow_safe_scale_64bit): New.
12993 2017-10-11  Nathan Sidwell  <nathan@acm.org>
12995         * tree.h (DECL_ASSEMBLER_NAME_SET_P): Don't check
12996         HAS_DECL_ASSEMBLER_NAME_P.
12997         * gimple-expr.c (gimple_decl_printable_name: Check
12998         HAS_DECL_ASSEMBLER_NAME_P too.
12999         * ipa-utils.h (type_in_anonymous_namespace_p): Check
13000         DECL_ASSEMBLER_NAME_SET_P of TYPE_NAME.
13001         (odr_type_p): No need to assert TYPE_NAME is a TYPE_DECL.
13002         * passes.c (rest_of_decl_compilation): Check
13003         HAS_DECL_ASSEMBLER_NAME_P too.
13004         * recog.c (verify_changes): Likewise.
13005         * tree-pretty-print.c (dump_decl_name): Likewise.
13006         * tree-ssa-structalias.c (alias_get_name): Likewise.  Reimplement.
13008         * tree.h (DECL_ASSEMBLER_NAME_RAW): New.
13009         (SET_DECL_ASSEMBLER_NAME): Use it.
13010         (DECL_ASSEMBLER_NAME_SET_P): Likewise.
13011         (COPY_DECL_ASSEMBLER_NAME): Likewise.
13012         * tree.c (decl_assembler_name): Use DECL_ASSEMBLER_NAME_RAW.
13014 2017-10-11  Jan Hubicka  <hubicka@ucw.cz>
13016         * config.gcc (i386, x86_64): Add extra objects.
13017         * config/i386/i386-protos.h (ix86_rip_relative_addr_p): Declare.
13018         (ix86_min_insn_size): Declare.
13019         (ix86_issue_rate): Declare.
13020         (ix86_adjust_cost): Declare.
13021         (ia32_multipass_dfa_lookahead): Declare.
13022         (ix86_macro_fusion_p): Declare.
13023         (ix86_macro_fusion_pair_p): Declare.
13024         (ix86_bd_has_dispatch): Declare.
13025         (ix86_bd_do_dispatch): Declare.
13026         (ix86_core2i7_init_hooks): Declare.
13027         (ix86_atom_sched_reorder): Declare.
13028         * config/i386/i386.c Move all CPU cost tables to x86-tune-costs.h.
13029         (COSTS_N_BYTES): Move to x86-tune-costs.h.
13030         (DUMMY_STRINGOP_ALGS):Move to x86-tune-costs.h.
13031         (rip_relative_addr_p): Rename to ...
13032         (ix86_rip_relative_addr_p): ... this one; export.
13033         (memory_address_length): Update.
13034         (ix86_issue_rate): Move to x86-tune-sched.c.
13035         (ix86_flags_dependent): Move to x86-tune-sched.c.
13036         (ix86_agi_dependent): Move to x86-tune-sched.c.
13037         (exact_dependency_1): Move to x86-tune-sched.c.
13038         (exact_store_load_dependency): Move to x86-tune-sched.c.
13039         (ix86_adjust_cost): Move to x86-tune-sched.c.
13040         (ia32_multipass_dfa_lookahead): Move to x86-tune-sched.c.
13041         (ix86_macro_fusion_p): Move to x86-tune-sched.c.
13042         (ix86_macro_fusion_pair_p): Move to x86-tune-sched.c.
13043         (do_reorder_for_imul): Move to x86-tune-sched-atom.c.
13044         (swap_top_of_ready_list): Move to x86-tune-sched-atom.c.
13045         (ix86_sched_reorder): Move to x86-tune-sched-atom.c.
13046         (core2i7_first_cycle_multipass_init): Move to x86-tune-sched-core.c.
13047         (core2i7_dfa_post_advance_cycle): Move to x86-tune-sched-core.c.
13048         (min_insn_size): Rename to ...
13049         (ix86_min_insn_size): ... this one; export.
13050         (core2i7_first_cycle_multipass_begin): Move to x86-tune-sched-core.c.
13051         (core2i7_first_cycle_multipass_issue): Move to x86-tune-sched-core.c.
13052         (core2i7_first_cycle_multipass_backtrack): Move to
13053         x86-tune-sched-core.c.
13054         (core2i7_first_cycle_multipass_end): Move to x86-tune-sched-core.c.
13055         (core2i7_first_cycle_multipass_fini): Move to x86-tune-sched-core.c.
13056         (ix86_sched_init_global): Break up logic to ix86_core2i7_init_hooks.
13057         (ix86_avoid_jump_mispredicts): Update.
13058         (TARGET_SCHED_DISPATCH): Move to ix86-tune-sched-bd.c.
13059         (TARGET_SCHED_DISPATCH_DO): Move to ix86-tune-sched-bd.c.
13060         (TARGET_SCHED_REORDER): Move to ix86-tune-sched-bd.c.
13061         (DISPATCH_WINDOW_SIZE): Move to ix86-tune-sched-bd.c.
13062         (MAX_DISPATCH_WINDOWS): Move to ix86-tune-sched-bd.c.
13063         (MAX_INSN): Move to ix86-tune-sched-bd.c.
13064         (MAX_IMM): Move to ix86-tune-sched-bd.c.
13065         (MAX_IMM_SIZE): Move to ix86-tune-sched-bd.c.
13066         (MAX_IMM_32): Move to ix86-tune-sched-bd.c.
13067         (MAX_IMM_64): Move to ix86-tune-sched-bd.c.
13068         (MAX_LOAD): Move to ix86-tune-sched-bd.c.
13069         (MAX_STORE): Move to ix86-tune-sched-bd.c.
13070         (BIG): Move to ix86-tune-sched-bd.c.
13071         (enum dispatch_group): Move to ix86-tune-sched-bd.c.
13072         (enum insn_path): Move to ix86-tune-sched-bd.c.
13073         (get_mem_group): Move to ix86-tune-sched-bd.c.
13074         (is_cmp): Move to ix86-tune-sched-bd.c.
13075         (dispatch_violation): Move to ix86-tune-sched-bd.c.
13076         (is_branch): Move to ix86-tune-sched-bd.c.
13077         (is_prefetch): Move to ix86-tune-sched-bd.c.
13078         (init_window): Move to ix86-tune-sched-bd.c.
13079         (allocate_window): Move to ix86-tune-sched-bd.c.
13080         (init_dispatch_sched): Move to ix86-tune-sched-bd.c.
13081         (is_end_basic_block): Move to ix86-tune-sched-bd.c.
13082         (process_end_window): Move to ix86-tune-sched-bd.c.
13083         (allocate_next_window): Move to ix86-tune-sched-bd.c.
13084         (find_constant): Move to ix86-tune-sched-bd.c.
13085         (get_num_immediates): Move to ix86-tune-sched-bd.c.
13086         (has_immediate): Move to ix86-tune-sched-bd.c.
13087         (get_insn_path): Move to ix86-tune-sched-bd.c.
13088         (get_insn_group): Move to ix86-tune-sched-bd.c.
13089         (count_num_restricted): Move to ix86-tune-sched-bd.c.
13090         (fits_dispatch_window): Move to ix86-tune-sched-bd.c.
13091         (add_insn_window): Move to ix86-tune-sched-bd.c.
13092         (add_to_dispatch_window): Move to ix86-tune-sched-bd.c.
13093         (debug_dispatch_window_file): Move to ix86-tune-sched-bd.c.
13094         (debug_dispatch_window): Move to ix86-tune-sched-bd.c.
13095         (debug_insn_dispatch_info_file): Move to ix86-tune-sched-bd.c.
13096         (debug_ready_dispatch): Move to ix86-tune-sched-bd.c.
13097         (do_dispatch): Move to ix86-tune-sched-bd.c.
13098         (has_dispatch): Move to ix86-tune-sched-bd.c.
13099         * config/i386/t-i386: Add new object files.
13100         * config/i386/x86-tune-costs.h: New file.
13101         * config/i386/x86-tune-sched-atom.c: New file.
13102         * config/i386/x86-tune-sched-bd.c: New file.
13103         * config/i386/x86-tune-sched-core.c: New file.
13104         * config/i386/x86-tune-sched.c: New file.
13106 2017-10-11  Liu Hao  <lh_mouse@126.com>
13108         * pretty-print.c [_WIN32] (colorize_init): Remove.  Use
13109         the generic version below instead.
13110         (should_colorize): Recognize Windows consoles as terminals
13111         for MinGW targets.
13112         * pretty-print.c [__MINGW32__] (write_all): New function.
13113         [__MINGW32__] (find_esc_head): Likewise.
13114         [__MINGW32__] (find_esc_terminator): Likewise.
13115         [__MINGW32__] (eat_esc_sequence): Likewise.
13116         [__MINGW32__] (mingw_ansi_fputs): New function that handles
13117         ANSI escape codes.
13118         (pp_write_text_to_stream): Use mingw_ansi_fputs instead of fputs
13119         for MinGW targets.
13121 2017-10-11  Richard Biener  <rguenther@suse.de>
13123         * tree-ssa-loop-niter.c (infer_loop_bounds_from_pointer_arith):
13124         Properly call analyze_scalar_evolution with the loop of the stmt.
13126 2017-10-11  Richard Biener  <rguenther@suse.de>
13128         * tree.def (POLYNOMIAL_CHREC): Remove CHREC_VARIABLE tree operand.
13129         * tree-core.h (tree_base): Add chrec_var union member.
13130         * tree.h (CHREC_VAR): Remove.
13131         (CHREC_LEFT, CHREC_RIGHT, CHREC_VARIABLE): Adjust.
13132         * tree-chrec.h (build_polynomial_chrec): Adjust.
13133         * tree-chrec.c (reset_evolution_in_loop): Use build_polynomial_chrec.
13134         * tree-pretty-print.c (dump_generic_node): Use CHREC_VARIABLE.
13136 2017-10-11  Marc Glisse  <marc.glisse@inria.fr>
13138         * fold-const.c (fold_binary_loc) [X +- Y CMP X]: Move ...
13139         * match.pd: ... here.
13140         ((T) X == (T) Y): Relax condition.
13142 2017-10-11  Bin Cheng  <bin.cheng@arm.com>
13144         PR tree-optimization/82472
13145         * tree-loop-distribution.c (sort_partitions_by_post_order): Refine
13146         comment.
13147         (break_alias_scc_partitions): Update postorder number.
13149 2017-10-11  Martin Liska  <mliska@suse.cz>
13151         PR sanitizer/82490
13152         * opts.c (parse_no_sanitize_attribute): Do not use error_value
13153         variable.
13154         * opts.h (parse_no_sanitize_attribute): Remove last argument.
13156 2017-10-11  Martin Liska  <mliska@suse.cz>
13158         * print-rtl.c (print_insn): Move declaration of idbuf
13159         to same scope as name.
13161 2017-10-11  Martin Liska  <mliska@suse.cz>
13163         Revert r253637:
13165         PR sanitizer/82484
13166         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
13167         volatile arguments.
13169 2017-10-11  Martin Liska  <mliska@suse.cz>
13171         PR sanitizer/82484
13172         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
13173         volatile arguments.
13175 2017-10-11  Adhemerval Zanella  <adhemerval.zanella@linaro.org>
13177         * config.gcc (default_gnu_indirect_function): Default to yes for
13178         arm*-*-linux* with glibc.
13180 2017-10-11  Richard Biener  <rguenther@suse.de>
13182         * tree-scalar-evolution.c (get_scalar_evolution): Handle
13183         default-defs and types we do not want to analyze.
13184         (interpret_loop_phi): Replace unreachable code with an assert.
13185         (compute_scalar_evolution_in_loop): Remove and inline ...
13186         (analyze_scalar_evolution_1): ... here, replacing condition with
13187         what makes the intent clearer.  Remove handling of cases
13188         get_scalar_evolution now handles.
13190 2017-10-10  Jim Wilson  <wilson@tuliptree.org>
13192         PR rtl-optimization/81434
13193         * haifa-sched.c (prune_ready_list): Init min_cost_group to 0.  Update
13194         comment for main loop.  In sched_group_found if, also add checks for
13195         pass and min_cost_group.
13197 2017-10-10  Segher Boessenkool  <segher@kernel.crashing.org>
13199         * config/rs6000/rs6000.c (TARGET_INSN_COST): New.
13200         (rs6000_insn_cost): New function.
13201         * config/rs6000/rs6000.md (cost): New attribute.
13203 2017-10-10  Jakub Jelinek  <jakub@redhat.com>
13204             H.J. Lu  <hongjiu.lu@intel.com>
13206         PR target/79565
13207         PR target/82483
13208         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Add
13209         OPTION_MASK_ISA_MMX for __builtin_ia32_maskmovq,
13210         __builtin_ia32_vec_ext_v4hi and __builtin_ia32_vec_set_v4hi.
13211         (ix86_expand_builtin): Treat OPTION_MASK_ISA_MMX similarly
13212         to OPTION_MASK_ISA_AVX512VL - builtins that have both
13213         OPTION_MASK_ISA_MMX and some other bit set require both
13214         mmx and the ISAs without the mmx bit.
13215         * config/i386/i386-builtin.def (__builtin_ia32_cvtps2pi,
13216         __builtin_ia32_cvttps2pi, __builtin_ia32_cvtpi2ps,
13217         __builtin_ia32_pavgb, __builtin_ia32_pavgw, __builtin_ia32_pmulhuw,
13218         __builtin_ia32_pmaxub, __builtin_ia32_pmaxsw, __builtin_ia32_pminub,
13219         __builtin_ia32_pminsw, __builtin_ia32_psadbw, __builtin_ia32_pmovmskb,
13220         __builtin_ia32_pshufw, __builtin_ia32_cvtpd2pi,
13221         __builtin_ia32_cvttpd2pi, __builtin_ia32_cvtpi2pd,
13222         __builtin_ia32_pmuludq, __builtin_ia32_pabsb, __builtin_ia32_pabsw,
13223         __builtin_ia32_pabsd, __builtin_ia32_phaddw, __builtin_ia32_phaddd,
13224         __builtin_ia32_phaddsw, __builtin_ia32_phsubw, __builtin_ia32_phsubd,
13225         __builtin_ia32_phsubsw, __builtin_ia32_pmaddubsw,
13226         __builtin_ia32_pmulhrsw, __builtin_ia32_pshufb, __builtin_ia32_psignb,
13227         __builtin_ia32_psignw, __builtin_ia32_psignd, __builtin_ia32_movntq,
13228         __builtin_ia32_paddq, __builtin_ia32_psubq, __builtin_ia32_palignr):
13229         Add OPTION_MASK_ISA_MMX.
13231 2017-10-10  Andreas Tobler  <andreast@gcc.gnu.org>
13233         * config.gcc (armv7*-*-freebsd*): New target.
13234         (armv6*-*-freebsd*): Remove obsolete TARGET_FREEBSD_ARMv6 define.
13236 2017-10-10  Jan Hubicka  <hubicka@ucw.cz>
13238         * x86-tune.def (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI,
13239         X86_TUNE_ADJUST_UNROLL, X86_TUNE_ONE_IF_CONV_INSN): Move to right
13240         spot in the file.
13242 2017-10-10  Richard Sandiford  <richard.sandiford@linaro.org>
13244         * wide-int.h (wide_int_ref_storage): Make host_dependent_precision
13245         a template parameter.
13246         (WIDE_INT_REF_FOR): Update accordingly.
13247         * tree.h (wi::int_traits <const_tree>): Delete.
13248         (wi::tree_to_widest_ref, wi::tree_to_offset_ref): New typedefs.
13249         (wi::to_widest, wi::to_offset): Use them.  Expand commentary.
13250         (wi::tree_to_wide_ref): New typedef.
13251         (wi::to_wide): New function.
13252         * calls.c (get_size_range): Use wi::to_wide when operating on
13253         trees as wide_ints.
13254         * cgraph.c (cgraph_node::create_thunk): Likewise.
13255         * config/i386/i386.c (ix86_data_alignment): Likewise.
13256         (ix86_local_alignment): Likewise.
13257         * dbxout.c (stabstr_O): Likewise.
13258         * dwarf2out.c (add_scalar_info, gen_enumeration_type_die): Likewise.
13259         * expr.c (const_vector_from_tree): Likewise.
13260         * fold-const-call.c (host_size_t_cst_p, fold_const_call_1): Likewise.
13261         * fold-const.c (may_negate_without_overflow_p, negate_expr_p)
13262         (fold_negate_expr_1, int_const_binop_1, const_binop)
13263         (fold_convert_const_int_from_real, optimize_bit_field_compare)
13264         (all_ones_mask_p, sign_bit_p, unextend, extract_muldiv_1)
13265         (fold_div_compare, fold_single_bit_test, fold_plusminus_mult_expr)
13266         (pointer_may_wrap_p, expr_not_equal_to, fold_binary_loc)
13267         (fold_ternary_loc, multiple_of_p, fold_negate_const, fold_abs_const)
13268         (fold_not_const, round_up_loc): Likewise.
13269         * gimple-fold.c (gimple_fold_indirect_ref): Likewise.
13270         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Likewise.
13271         (alloca_call_type): Likewise.
13272         * gimple.c (preprocess_case_label_vec_for_gimple): Likewise.
13273         * godump.c (go_output_typedef): Likewise.
13274         * graphite-sese-to-poly.c (tree_int_to_gmp): Likewise.
13275         * internal-fn.c (get_min_precision): Likewise.
13276         * ipa-cp.c (ipcp_store_vr_results): Likewise.
13277         * ipa-polymorphic-call.c
13278         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Likewise.
13279         * ipa-prop.c (ipa_print_node_jump_functions_for_edge): Likewise.
13280         (ipa_modify_call_arguments): Likewise.
13281         * match.pd: Likewise.
13282         * omp-low.c (scan_omp_1_op, lower_omp_ordered_clauses): Likewise.
13283         * print-tree.c (print_node_brief, print_node): Likewise.
13284         * stmt.c (expand_case): Likewise.
13285         * stor-layout.c (layout_type): Likewise.
13286         * tree-affine.c (tree_to_aff_combination): Likewise.
13287         * tree-cfg.c (group_case_labels_stmt): Likewise.
13288         * tree-data-ref.c (dr_analyze_indices): Likewise.
13289         (prune_runtime_alias_test_list): Likewise.
13290         * tree-dump.c (dequeue_and_dump): Likewise.
13291         * tree-inline.c (remap_gimple_op_r, copy_tree_body_r): Likewise.
13292         * tree-predcom.c (is_inv_store_elimination_chain): Likewise.
13293         * tree-pretty-print.c (dump_generic_node): Likewise.
13294         * tree-scalar-evolution.c (iv_can_overflow_p): Likewise.
13295         (simple_iv_with_niters): Likewise.
13296         * tree-ssa-address.c (addr_for_mem_ref): Likewise.
13297         * tree-ssa-ccp.c (ccp_finalize, evaluate_stmt): Likewise.
13298         * tree-ssa-loop-ivopts.c (constant_multiple_of): Likewise.
13299         * tree-ssa-loop-niter.c (split_to_var_and_offset)
13300         (refine_value_range_using_guard, number_of_iterations_ne_max)
13301         (number_of_iterations_lt_to_ne, number_of_iterations_lt)
13302         (get_cst_init_from_scev, record_nonwrapping_iv)
13303         (scev_var_range_cant_overflow): Likewise.
13304         * tree-ssa-phiopt.c (minmax_replacement): Likewise.
13305         * tree-ssa-pre.c (compute_avail): Likewise.
13306         * tree-ssa-sccvn.c (vn_reference_fold_indirect): Likewise.
13307         (vn_reference_maybe_forwprop_address, valueized_wider_op): Likewise.
13308         * tree-ssa-structalias.c (get_constraint_for_ptr_offset): Likewise.
13309         * tree-ssa-uninit.c (is_pred_expr_subset_of): Likewise.
13310         * tree-ssanames.c (set_nonzero_bits, get_nonzero_bits): Likewise.
13311         * tree-switch-conversion.c (collect_switch_conv_info, array_value_type)
13312         (dump_case_nodes, try_switch_expansion): Likewise.
13313         * tree-vect-loop-manip.c (vect_gen_vector_loop_niters): Likewise.
13314         (vect_do_peeling): Likewise.
13315         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13316         * tree-vect-stmts.c (vectorizable_load): Likewise.
13317         * tree-vrp.c (compare_values_warnv, vrp_int_const_binop): Likewise.
13318         (zero_nonzero_bits_from_vr, ranges_from_anti_range): Likewise.
13319         (extract_range_from_binary_expr_1, adjust_range_with_scev): Likewise.
13320         (overflow_comparison_p_1, register_edge_assert_for_2): Likewise.
13321         (is_masked_range_test, find_switch_asserts, maybe_set_nonzero_bits)
13322         (vrp_evaluate_conditional_warnv_with_ops, intersect_ranges): Likewise.
13323         (range_fits_type_p, two_valued_val_range_p, vrp_finalize): Likewise.
13324         (evrp_dom_walker::before_dom_children): Likewise.
13325         * tree.c (cache_integer_cst, real_value_from_int_cst, integer_zerop)
13326         (integer_all_onesp, integer_pow2p, integer_nonzerop, tree_log2)
13327         (tree_floor_log2, tree_ctz, mem_ref_offset, tree_int_cst_sign_bit)
13328         (tree_int_cst_sgn, get_unwidened, int_fits_type_p): Likewise.
13329         (get_type_static_bounds, num_ending_zeros, drop_tree_overflow)
13330         (get_range_pos_neg): Likewise.
13331         * ubsan.c (ubsan_expand_ptr_ifn): Likewise.
13332         * config/darwin.c (darwin_mergeable_constant_section): Likewise.
13333         * config/aarch64/aarch64.c (aapcs_vfp_sub_candidate): Likewise.
13334         * config/arm/arm.c (aapcs_vfp_sub_candidate): Likewise.
13335         * config/avr/avr.c (avr_fold_builtin): Likewise.
13336         * config/bfin/bfin.c (bfin_local_alignment): Likewise.
13337         * config/msp430/msp430.c (msp430_attr): Likewise.
13338         * config/nds32/nds32.c (nds32_insert_attributes): Likewise.
13339         * config/powerpcspe/powerpcspe-c.c
13340         (altivec_resolve_overloaded_builtin): Likewise.
13341         * config/powerpcspe/powerpcspe.c (rs6000_aggregate_candidate)
13342         (rs6000_expand_ternop_builtin): Likewise.
13343         * config/rs6000/rs6000-c.c
13344         (altivec_resolve_overloaded_builtin): Likewise.
13345         * config/rs6000/rs6000.c (rs6000_aggregate_candidate): Likewise.
13346         (rs6000_expand_ternop_builtin): Likewise.
13347         * config/s390/s390.c (s390_handle_hotpatch_attribute): Likewise.
13349 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13351         * tree-vect-loop-manip.c (rename_variables_in_bb): Rename PHI nodes
13352         when copying loop nest with only one inner loop.
13354 2017-10-10  Richard Biener  <rguenther@suse.de>
13356         * tree-cfgcleanup.c (cleanup_tree_cfg_noloop): Avoid compacting
13357         blocks if SCEV is active.
13358         * tree-scalar-evolution.c (analyze_scalar_evolution_1): Remove
13359         dead code.
13360         (analyze_scalar_evolution): Handle cached evolutions the obvious way.
13361         (scev_initialize): Assert we are not yet initialized.
13363 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13365         * tree-loop-distribution.c (generate_loops_for_partition): Remove
13366         inner loop's exit stmt by making it always exit the loop, otherwise
13367         we would generate an infinite empty loop.
13369 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13371         * tree-vect-loop-manip.c (slpeel_tree_duplicate_loop_to_edge_cfg): Skip
13372         renaming variables in new preheader if it's deleted.
13374 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13376         * tree-loop-distribution.c (struct partition): Remove unused field
13377         loops of the structure.
13378         (partition_alloc, partition_free): Ditto.
13379         (build_rdg_partition_for_vertex): Ditto.
13381 2017-10-09  Jeff Law  <law@redhat.com>
13383         * targhooks.c (default_stack_clash_protection_final_dynamic_probe): Fix
13384         return type to match prototype and documentation.
13386 2010-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13388         * config/rs6000/rs6000.c (processor_costs): Move to ...
13389         * config/rs6000/rs6000.h: ... here.
13390         (rs6000_cost): Declare.
13392 2017-10-09  Eric Botcazou  <ebotcazou@adacore.com>
13394         * except.c (setjmp_fn): New global variable.
13395         (init_eh): Initialize it if DONT_USE_BUILTIN_SETJMP is defined.
13396         (sjlj_emit_function_enter): Call it instead of BUILTIN_SETJMP
13397         if DONT_USE_BUILTIN_SETJMP is defined.
13399 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13401         * target.def (insn_cost): New hook.
13402         * doc/tm.texi.in (TARGET_INSN_COST): New hook.
13403         * doc/tm.texi: Regenerate.
13404         * rtlanal.c (insn_cost): Use the new hook.
13406 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13408         * combine.c (combine_validate_cost): Compute the new insn_cost,
13409         not just pattern_cost.
13410         (try_combine): Adjust comment.
13412 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13414         * cfgrtl.c (rtl_account_profile_record): Replace insn_rtx_cost with
13415         insn_cost.
13416         * combine.c (uid_insn_cost): Adjust comment.
13417         (combine_validate_cost): Adjust comment.  Use pattern_cost instead
13418         of insn_rtx_cost
13419         (combine_instructions): Use insn_cost instead of insn_rtx_cost.
13420         * dse.c (find_shift_sequence): Ditto.
13421         * ifcvt.c (cheap_bb_rtx_cost_p): Ditto.
13422         (bb_valid_for_noce_process_p): Use pattern_cost.
13423         * rtl.h (insn_rtx_cost): Delete.
13424         (pattern_cost): New prototype.
13425         (insn_cost): New prototype.
13426         * rtlanal.c (insn_rtx_cost): Rename to...
13427         (pattern_cost): ... this.
13428         (insn_cost): New.
13430 2017-10-09  Uros Bizjak  <ubizjak@gmail.com>
13432         * config/i386/i386.md (*jcc_2): Remove insn pattern.
13433         (*jcc<mode>_0_r_i387): Ditto.
13434         (*jccxf_r_i387): Ditto.
13435         (*jcc<mode>_r_i387): Ditto.
13436         (*jccu<mode>_r_i387): Ditto.
13437         (*jcc<X87MODEF:mode>_<SWI24:mode>_r_i387): Ditto.
13438         (*jcc): Rename from *jcc_1.
13440 2017-10-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
13442         * config/rs6000/rs6000-p8swap.c (rs6000_analyze_swaps): Process
13443         deferred rescans after the lvx/stvx recombination pre-pass.
13445 2017-10-09  Michael Meissner  <meissner@linux.vnet.ibm.com>
13447         * config/rs6000/amo.h: New include file to provide ISA 3.0 atomic
13448         memory operation instruction support.
13449         * config.gcc (powerpc*-*-*): Include amo.h as an extra header.
13450         (rs6000-ibm-aix[789]*): Likewise.
13451         * doc/extend.texi (PowerPC Atomic Memory Operation Functions):
13452         Document new functions.
13454 2017-10-09  Richard Biener  <rguenther@suse.de>
13456         PR tree-optimization/82397
13457         * tree-data-ref.c (data_ref_compare_tree): Make sure to return
13458         equality only for semantically equal trees.
13460 2017-10-09  Richard Biener  <rguenther@suse.de>
13462         PR tree-optimization/82449
13463         * sese.c (scev_analyzable_p): Check whether the SCEV is linear.
13464         * tree-chrec.h (evolution_function_is_constant_p): Adjust to
13465         allow constant addresses.
13466         * tree-chrec.c (scev_is_linear_expression): Constant evolutions
13467         are linear.
13469 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13471         * config/s390/s390-builtins.def (vec_nabs, vec_vfi): Fix builtin
13472         flags.
13474 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13476         PR target/82463
13477         * config/s390/vecintrin.h (vec_madd, vec_msub): Fix macro
13478         definitions.
13480 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13482         PR target/82465
13483         * config/s390/s390-builtins.def (vec_sqrt): Fix builtin flags.
13485 2017-10-09  Jakub Jelinek  <jakub@redhat.com>
13487         PR target/82464
13488         * config/s390/s390-builtins.def (s390_vec_xor_flt_a,
13489         s390_vec_xor_flt_b, s390_vec_xor_flt_c): New.
13491 2017-10-09  Richard Sandiford  <richard.sandiford@linaro.org>
13493         * wide-int.h (WI_BINARY_OPERATOR_RESULT): New macro.
13494         (WI_BINARY_PREDICATE_RESULT): Likewise.
13495         (wi::binary_traits::operator_result): New type.
13496         (wi::binary_traits::predicate_result): Likewise.
13497         (generic_wide_int::operator~, unary generic_wide_int::operator-)
13498         (generic_wide_int::operator==, generic_wide_int::operator!=)
13499         (generic_wide_int::operator&, generic_wide_int::and_not)
13500         (generic_wide_int::operator|, generic_wide_int::or_not)
13501         (generic_wide_int::operator^, generic_wide_int::operator+
13502         (binary generic_wide_int::operator-, generic_wide_int::operator*):
13503         Delete.
13504         (operator~, unary operator-, operator==, operator!=, operator&)
13505         (operator|, operator^, operator+, binary operator-, operator*): New
13506         functions.
13507         * expr.c (get_inner_reference): Use wi::bit_and_not.
13508         * fold-const.c (fold_binary_loc): Likewise.
13509         * ipa-prop.c (ipa_compute_jump_functions_for_edge): Likewise.
13510         * tree-ssa-ccp.c (get_value_from_alignment): Likewise.
13511         (bit_value_binop): Likewise.
13512         * tree-ssa-math-opts.c (find_bswap_or_nop_load): Likewise.
13513         * tree-vrp.c (zero_nonzero_bits_from_vr): Likewise.
13514         (extract_range_from_binary_expr_1): Likewise.
13515         (masked_increment): Likewise.
13516         (simplify_bit_ops_using_ranges): Likewise.
13518 2017-10-09  Martin Jambor  <mjambor@suse.cz>
13520         PR hsa/82416
13521         * hsa-common.h (hsa_op_with_type): New method extend_int_to_32bit.
13522         * hsa-gen.c (hsa_extend_inttype_to_32bit): New function.
13523         (hsa_type_for_scalar_tree_type): Use it.  Always force min32int for
13524         COMPLEX types.
13525         (hsa_fixup_mov_insn_type): New function.
13526         (hsa_op_with_type::get_in_type): Use it.
13527         (hsa_build_append_simple_mov): Likewise.  Allow sub-32bit
13528         immediates in an assert.
13529         (hsa_op_with_type::extend_int_to_32bit): New method.
13530         (gen_hsa_insns_for_bitfield): Fixup instruction and intermediary
13531         types.  Convert to dest type if necessary.
13532         (gen_hsa_insns_for_bitfield_load): Fixup load type if necessary.
13533         (reg_for_gimple_ssa): Pass false as min32int to
13534         hsa_type_for_scalar_tree_type.
13535         (gen_hsa_addr): Fixup type when creating addresable temporary.
13536         (gen_hsa_cmp_insn_from_gimple): Extend operands if necessary.
13537         (gen_hsa_unary_operation): Extend operands and convert to dest type if
13538         necessary.  Call hsa_fixup_mov_insn_type.
13539         (gen_hsa_binary_operation): Changed operand types to hsa_op_with_type,
13540         extend operands and convert to dest type if necessary.
13541         (gen_hsa_insns_for_operation_assignment): Extend operands and convert
13542         to dest type if necessary.
13543         (set_output_in_type): Call hsa_fixup_mov_insn_type.  Just ude dest
13544         if conversion nt necessary and size matches.
13545         (gen_hsa_insns_for_load): Call hsa_fixup_mov_insn_type, convert
13546         to dest type if necessary.
13547         (gen_hsa_insns_for_store): Call hsa_fixup_mov_insn_type.
13548         (gen_hsa_insns_for_switch_stmt): Likewise. Also extend operands if
13549         necessary.
13550         (gen_hsa_clrsb): Likewise.
13551         (gen_hsa_ffs): Likewise.
13552         (gen_hsa_divmod): Extend operands and convert to dest type if
13553         necessary.
13554         (gen_hsa_atomic_for_builtin): Change type of op to hsa_op_with_type.
13556 2017-10-08  Segher Boessenkool  <segher@kernel.crashing.org>
13558         * config/rs6000/rs6000.md (conditional branch): Clean up formatting.
13559         Remove empty default arguments.  Use a brace block as output
13560         statement.
13561         (conditional return): Ditto.
13562         (jump): Ditto.
13563         (indirect_jump): Ditto.  Use b%T0 instead of bctr/blr.
13564         (tablejump, tablejumpsi, tablejumpdi, *tablejump<mode>_internal1):
13565         Ditto.
13566         (group_ending_nop): Ditto.
13567         (doloop_end): Ditto.
13568         (ctr<mode>, ctr<mode>_internal1, ctr<mode>_internal2): Ditto.
13569         (splitters for those): Ditto.
13571 2017-10-08  Segher Boessenkool  <segher@kernel.crashing.org>
13573         * config/rs6000/rs6000-string.c (expand_strncmp_align_check): Invert
13574         a conditional jump (and the compare for it) so that pc_rtx is the
13575         last operand.
13576         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Adjust
13577         for the deleted and renamed ctr<mode>_internal[234] patterns.
13578         * config/rs6000/rs6000.md: Delete second conditional branch pattern.
13579         Delete second conditional return pattern.
13580         (ctr<mode>_internal2): Delete this second bdnz pattern.
13581         (ctr<mode>_internal3): Rename to ctr<mode>_internal2.
13582         (ctr<mode>_internal4): Delete this second bdz pattern.
13584 2017-10-08  Eric Botcazou  <ebotcazou@adacore.com>
13586         * tree-outof-ssa.h (ssaexpand): Add partitions_for_undefined_values.
13587         (always_initialized_rtx_for_ssa_name_p): New predicate.
13588         * tree-outof-ssa.c (remove_ssa_form): Initialize new field of SA.
13589         (finish_out_of_ssa): Free new field of SA.
13590         * tree-ssa-coalesce.h (get_undefined_value_partitions): Declare.
13591         * tree-ssa-coalesce.c: Include tree-ssa.h.
13592         (get_parm_default_def_partitions): Remove extern keyword.
13593         (get_undefined_value_partitions): New function.
13594         * expr.c (expand_expr_real_1) <expand_decl_rtl>: For a SSA_NAME, do
13595         not set SUBREG_PROMOTED_VAR_P on the sub-register if it may contain
13596         uninitialized bits.
13597         * loop-iv.c (iv_get_reaching_def): Disqualify all subregs.
13599 2017-10-08  Eric Botcazou  <ebotcazou@adacore.com>
13601         * builtins.def (BUILT_IN_SETJMP): Revert latest change.
13603 2017-10-08  Jan Hubicka  <hubicka@ucw.cz>
13605         * config/i386/i386.c (ix86_expand_set_or_movmem): Disable 512bit loops
13606         for targets that preffer 128bit.
13608 2017-10-08  Jan Hubicka  <hubicka@ucw.cz>
13610         * config/i386/i386.c (has_dispatch): Disable for Ryzen.
13612 2017-10-08  Olivier Hainque  <hainque@adacore.com>
13614         * config/arm/arm.c (arm_set_return_address): Use MEM_VOLATILE_P
13615         on the target mem instead of RTX_FRAME_RELATED_P on the insn to
13616         prevent DSE.
13617         (thumb_set_return_address): Likewise.
13619 2017-10-08  Olivier Hainque  <hainque@adacore.com>
13621         * common/config/arm/arm-common.c (arm_except_unwind_info):
13622         Handle DWARF2_UNWIND_INFO.
13624 2017-10-07  Michael Collison <michael.collison@arm.com>
13626         * config/aarch64/aarch64.md (*aarch64_reg_<optab>_minus<mode>3):
13627         New pattern.
13629 2017-10-07  Eric Botcazou  <ebotcazou@adacore.com>
13631         * builtins.def (BUILT_IN_SETJMP): Declare as library builtin instead
13632         of GCC builtin if DONT_USE_BUILTIN_SETJMP is defined.
13633         * except.c (sjlj_emit_function_enter): If DONT_USE_BUILTIN_SETJMP is
13634         defined, force the creation of a new block for a dispatch label.
13636 2017-10-07  Jan Hubicka  <hubicka@ucw.cz>
13638         * invoke.texi (Wsuggest-attribute=cold): Document.
13639         * common.opt (Wsuggest-attribute=cold): New
13640         * ipa-pure-const.c (warn_function_cold): New function.
13641         * predict.c (compute_function_frequency): Use it.
13642         * predict.h (warn_function_cold): Declare.
13644 2017-10-06  Jan Hubicka  <hubicka@ucw.cz>
13646         * tree-switch-conversion.c (do_jump_if_equal, emit_cmp_and_jump_insns):
13647         Update profile.
13649 2017-10-06  Martin Liska  <mliska@suse.cz>
13651         * sanopt.c (struct sanopt_tree_triplet_hash): Remove inline
13652         keyword for member functions.
13653         (struct sanopt_tree_couple): New struct.
13654         (struct sanopt_tree_couple_hash): New function.
13655         (struct sanopt_ctx): Add new hash_map.
13656         (has_dominating_ubsan_ptr_check): New function.
13657         (record_ubsan_ptr_check_stmt): Likewise.
13658         (maybe_optimize_ubsan_ptr_ifn): Likewise.
13659         (sanopt_optimize_walker): Handle IFN_UBSAN_PTR.
13660         (pass_sanopt::execute): Handle also SANITIZE_POINTER_OVERFLOW.
13662 2017-10-06  Sudakshina Das  <sudi.das@arm.com>
13664         PR target/82440
13665         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Only call
13666         aarch64_simd_valid_immediate on CONST_VECTORs.
13667         (aarch64_reg_or_bic_imm): Likewise.
13669 2017-10-06  Wilco Dijkstra  <wdijkstr@arm.com>
13671         PR rtl-optimization/82396
13672         * haifa-sched.c (ready_sort_real): Disable qsort checking.
13674 2017-10-06  Sebastian Pop  <sebpop@gmail.com>
13676         * graphite-dependences.c (scop_get_reads): Move code to...
13677         (scop_get_must_writes): Move code to...
13678         (scop_get_may_writes): Move code to...
13679         (scop_get_reads_and_writes): ... here.
13680         (scop_get_dependences): Call scop_get_reads_and_writes.
13682 2017-10-06  Jakub Jelinek  <jakub@redhat.com>
13684         PR tree-optimization/82434
13685         * fold-const.h (can_native_encode_type_p,
13686         can_native_encode_string_p): Remove.
13687         * fold-const.c (native_encode_int): Formatting fixes.  If ptr is NULL,
13688         don't encode anything, just return what would be otherwise returned.
13689         (native_encode_fixed, native_encode_complex, native_encode_vector):
13690         Likewise.
13691         (native_encode_string): Likewise.  Inline by hand
13692         can_native_encode_string_p.
13693         (can_native_encode_type_p): Remove.
13694         (can_native_encode_string_p): Remove.
13695         * tree-vect-stmts.c (vectorizable_store): Instead of testing just
13696         STRING_CSTs using can_native_encode_string_p, test all
13697         CONSTANT_CLASS_P values using native_encode_expr with NULL ptr.
13698         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Remove last
13699         argument from native_encode_expr.
13700         (rhs_valid_for_store_merging_p): Use native_encode_expr with NULL ptr.
13701         (pass_store_merging::execute): Don't unnecessarily look for 3 stmts,
13702         but just 2.
13704 2017-10-06  Richard Biener  <rguenther@suse.de>
13706         PR tree-optimization/82397
13707         * tree-vect-data-refs.c (dr_group_sort_cmp): Do not use
13708         operand_equal_p but rely on data_ref_compare_tree for detecting
13709         equalities.
13710         (vect_analyze_data_ref_accesses): Use data_ref_compare_tree
13711         to match up with dr_group_sort_cmp.
13713 2017-10-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13715         PR target/82322
13716         * config/s390/s390-builtins.def (s390_vfi): Define new overloaded
13717         builtin.
13718         * config/s390/s390-builtin-types.def: Regenerate.
13720 2017-10-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13722         PR target/82317
13723         * config/s390/s390-builtin-types.def: Regenerate.
13724         * config/s390/s390-builtins.def (s390_vfmaxdb_4, s390_vfmindb_4):
13725         Change flag from B_VXE to B_VX.
13726         (s390_vec_min_dbl): Remove B_VXE flag.
13728 2017-10-06  Richard Biener  <rguenther@suse.de>
13730         * graphite-isl-ast-to-gimple.c: Include ssa.h and tree-ssa.h.
13731         (translate_isl_ast_to_gimple::translate_pending_phi_nodes,
13732         translate_isl_ast_to_gimple::is_valid_rename,
13733         translate_isl_ast_to_gimple::get_rename,
13734         translate_isl_ast_to_gimple::get_def_bb_for_const,
13735         translate_isl_ast_to_gimple::get_new_name,
13736         translate_isl_ast_to_gimple::collect_all_ssa_names,
13737         translate_isl_ast_to_gimple::copy_loop_phi_args,
13738         translate_isl_ast_to_gimple::collect_all_ssa_names,
13739         translate_isl_ast_to_gimple::copy_loop_phi_args,
13740         translate_isl_ast_to_gimple::copy_loop_phi_nodes,
13741         translate_isl_ast_to_gimple::add_close_phis_to_merge_points,
13742         translate_isl_ast_to_gimple::add_close_phis_to_outer_loops,
13743         translate_isl_ast_to_gimple::copy_loop_close_phi_args,
13744         translate_isl_ast_to_gimple::copy_loop_close_phi_nodes,
13745         translate_isl_ast_to_gimple::copy_cond_phi_args,
13746         translate_isl_ast_to_gimple::copy_cond_phi_nodes,
13747         translate_isl_ast_to_gimple::edge_for_new_close_phis,
13748         translate_isl_ast_to_gimple::add_phi_arg_for_new_expr,
13749         translate_isl_ast_to_gimple::rename_uses,
13750         translate_isl_ast_to_gimple::rename_all_uses): Remove.
13751         (translate_isl_ast_to_gimple::get_rename_from_scev): Simplify.
13752         (set_rename_for_each_def): Likewise.
13753         (graphite_copy_stmts_from_block): Handle debug stmt resetting
13754         here.  Handle rewriting SCEV analyzable uses here.
13755         (copy_bb_and_scalar_dependences): Generate code for PHI
13756         copy-in/outs.
13757         (graphite_regenerate_ast_isl): Adjust.
13758         * graphite-scop-detection.c (trivially_empty_bb_p): Move to sese.[ch].
13759         (add_write, add_read): New functions.
13760         (build_cross_bb_scalars_def): Use it and simplify.
13761         (build_cross_bb_scalars_use): Likewise.
13762         (graphite_find_cross_bb_scalar_vars): Inline into...
13763         (try_generate_gimple_bb): ...here.  Add dependences for PHIs,
13764         simulating out-of-SSA.  Compute liveout and add dependencies.
13765         (build_scops): Force an empty entry block.
13766         * sese.h (sese_info_t::liveout, sese_info_t::debug_liveout): New
13767         members.
13768         (sese_build_liveouts): Declare.
13769         (sese_trivially_empty_bb_p): Likewise.
13770         * sese.c (sese_build_liveouts_bb): Properly handle PHIs,
13771         compute liveout and debug_liveout.
13772         (sese_bad_liveouts_use): Remove.
13773         (sese_reset_debug_liveouts_bb): Likewise.
13774         (sese_reset_debug_liveouts): Rewrite in terms of debug_liveout.
13775         (sese_build_liveouts): Build liveout and debug_liveout and store
13776         it in region.
13777         (new_sese_info): Adjust.
13778         (free_sese_info): Likewise.
13779         (sese_insert_phis_for_liveouts): Reset debug stmts from here,
13780         do not build liveout here.
13781         (move_sese_in_condition): Adjust region entry.
13782         (scev_analyzable_p): Match up with chrec_apply requirements.
13783         (sese_trivially_empty_bb_p): New.
13784         * tree-into-ssa.c (get_reaching_def): Properly support generating
13785         default-defs for incremental rewrite of anonymous names.
13787 2017-10-06  Richard Biener  <rguenther@suse.de>
13789         * graphite-sese-to-poly.c (extract_affine): For casts increasing
13790         precision do not perform modulo reduction.
13792 2017-10-06  Richard Biener  <rguenther@suse.de>
13794         PR tree-optimization/82436
13795         * tree-vect-slp.c (vect_supported_load_permutation_p): More
13796         conservatively choose the vectorization factor when checking
13797         whether we can perform the required load permutation.
13798         (vect_transform_slp_perm_load): Assert when we may not fail.
13800 2017-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
13802         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Correct error
13803         message for incompatible -msdata=* and -mcall-* options.
13805 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13807         * config/i386/i386.c (ia32_multipass_dfa_lookahead): Default to issue
13808         rate for post-reload scheduling.
13810 2017-10-05  Tamar Christina  <tamar.christina@arm.com>
13812         * doc/sourcebuild.texi (vect_sizes_16B_8B, vect_sizes_32B_16B): New.
13814 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13816         * config/i386/i386.c (znver1_cost): Set branch_cost to 3 (instead of 2)
13817         to improve monte carlo in scimark.
13819 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13821         * config/i386/i386.c (ix86_size_cost, i386_cost, i486_cost,
13822         pentium_cost, lakemont_cost, pentiumpro_cost, geode_cost, k6_cost,
13823         athlon_cost, k8_cost, amdfam10_cost, btver1_cost, btver2_cost,
13824         pentium4_cost, nocona_cost): Set reassociation width to 1.
13825         (bdver1_cost, bdver2_cost, bdver3_cost, bdver4_cost): Set reassociation
13826         width to 2 for fp operations and 1 otherwise.
13827         (znver1_cost): Set scalar reassoc width to 4 and vector to 3 and 6
13828         for int and fp.
13829         (atom_cost): Set reassociation width to 2.
13830         (slm_cost, generic_cost): Set fp reassociation width
13831         to 2 and 1 otherwise.
13832         (intel_cost): Set fp reassociation width to 4 and 1 otherwise.
13833         (core_cost): Set fp reassociation width to 4 and vector to 2.
13834         (ix86_reassociation_width): Rewrite using cost table; special case
13835         plus/minus on Zen; honor X86_TUNE_SSE_SPLIT_REGS
13836         and TARGET_AVX128_OPTIMAL.
13837         * config/i386/i386.h (processor_costs): Add
13838         reassoc_int, reassoc_fp, reassoc_vec_int, reassoc_vec_fp.
13839         (TARGET_VECTOR_PARALLEL_EXECUTION, TARGET_REASSOC_INT_TO_PARALLEL,
13840         TARGET_REASSOC_FP_TO_PARALLEL): Remove.
13841         * x86-tune.def (X86_TUNE_REASSOC_INT_TO_PARALLEL): Remove.
13842         (X86_TUNE_REASSOC_FP_TO_PARALLEL): Remove.
13843         (X86_TUNE_VECTOR_PARALLEL_EXECUTION):  Remove.
13845 2017-10-05  Nathan Sidwell  <nathan@acm.org>
13847         * doc/invoke.texi (Wparentheses): Document C++ MVP behaviour.
13849 2017-10-05  Tamar Christina  <tamar.christina@arm.com>
13851         * config/arm/arm.c (arm_test_fpu_data): New.
13852         (arm_run_selftests): Call arm_test_fpu_data.
13854 2017-10-04  Nathan Sidwell  <nathan@acm.org>
13856         * toplev.c (toplev::main): Remove excess parens on pretty_printer
13857         decl.
13858         * caller-save.c (insert_save): Remove excess parens on TO_SAVE parm.
13860 2017-10-04  Sudakshina Das  <sudi.das@arm.com>
13862         * config/aarch64/aarch64-protos.h (enum simd_immediate_check): New
13863         check type for aarch64_simd_valid_immediate.
13864         (aarch64_output_simd_mov_immediate): Update prototype.
13865         (aarch64_simd_valid_immediate): Update prototype.
13866         * config/aarch64/aarch64-simd.md (orr<mode>3): modified pattern to add
13867         support for ORR-immediate.
13868         (and<mode>3): modified pattern to add support for BIC-immediate.
13869         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Function
13870         now checks for valid immediate for BIC and ORR based on new enum
13871         argument.
13872         (aarch64_output_simd_mov_immediate): Function now used to output
13873         BIC/ORR imm as well based on new enum argument.
13874         * config/aarch64/constraints.md (Do): New vector immediate constraint.
13875         (Db) : Likewise.
13876         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): New predicate.
13877         (aarch64_reg_or_bic_imm): Likewise.
13879 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13881         * config/s390/vx-builtins.md ("vec_mergeh<mode>")
13882         ("vec_mergel<mode>"): Change mode iterator to V_128_NOSINGLE.
13884 2017-10-04  Wilco Dijkstra  <wdijkstr@arm.com>
13886         Revert r253399:
13888         PR rtl-optimization/82396
13889         * haifa-sched.c (autopref_multipass_init): Simplify
13890         initialization.
13891         (autopref_rank_data): Simplify sort order.
13892         * sched-int.h (autopref_multipass_data_): Remove
13893         multi_mem_insn_p, min_offset and max_offset.
13895 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13897         * doc/sourcebuild.texi: Document vect_peeling_profitable.
13899 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13901         * doc/sourcebuild.texi: Document vect_intdouble_cvt and
13902         vect_doubleint_cvt.
13904 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13906         * doc/sourcebuild.texi: Document vect_long_mult.
13908 2017-10-04  Richard Sandiford  <richard.sandiford@linaro.org>
13910         PR tree-optimization/82413
13911         * fold-const.c (build_range_check): Use widest_int when comparing
13912         the maximum ETYPE value with HIGH.
13914 2017-10-04  Wilco Dijkstra  <wdijkstr@arm.com>
13916         PR rtl-optimization/82396
13917         * haifa-sched.c (autopref_multipass_init): Simplify
13918         initialization.
13919         (autopref_rank_data): Simplify sort order.
13920         * sched-int.h (autopref_multipass_data_): Remove
13921         multi_mem_insn_p, min_offset and max_offset.
13923 2017-10-04  Jakub Jelinek  <jakub@redhat.com>
13925         PR tree-optimization/82381
13926         * tree-ssa-reassoc.c (sort_by_operand_rank): Check for different
13927         oeN->rank first.  Return 1 or -1 if one op is SSA_NAME and the other
13928         is not.
13930         PR tree-optimization/82374
13931         * omp-low.c (create_omp_child_function): Copy DECL_ATTRIBUTES,
13932         DECL_FUNCTION_SPECIFIC_OPTIMIZATION,
13933         DECL_FUNCTION_SPECIFIC_TARGET and DECL_FUNCTION_VERSIONED from
13934         current_function_decl to the new decl.
13936 2017-10-03  Michael Meissner  <meissner@linux.vnet.ibm.com>
13938         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_2_HW): Define new
13939         helper macro for IEEE float128 hardware built-in functions.
13940         (SQRTF128_ODD): Add built-in functions with the round-to-odd
13941         semantics.
13942         (TRUNCF128_ODD): Likewise.
13943         (ADDF128_ODD): Likewise.
13944         (SUBF128_ODD): Likewise.
13945         (MULF128_ODD): Likewise.
13946         (DIVF128_ODD): Likewise.
13947         (FMAF128_ODD): Likewise.
13948         * config/rs6000/rs6000.md (UNSPEC_ROUND_TO_ODD): Rename to
13949         UNSPEC_TRUNC_ROUND_TO_ODD.
13950         (UNSPEC_TRUNC_ROUND_TO_ODD): Likewise.
13951         (UNSPEC_ADD_ROUND_TO_ODD): New unspec codes for the IEEE 128-bit
13952         floating point round to odd instructions.
13953         (UNSPEC_SUB_ROUND_TO_ODD): Likewise.
13954         (UNSPEC_MUL_ROUND_TO_ODD): Likewise.
13955         (UNSPEC_DIV_ROUND_TO_ODD): Likewise.
13956         (UNSPEC_FMA_ROUND_TO_ODD): Likewise.
13957         (UNSPEC_SQRT_ROUND_TO_ODD): Likewise.
13958         (trunc<mode>sf2_hw): Change the truncate with round to odd
13959         expansion to use UNSPEC_TRUNC_ROUND_TO_ODD.
13960         (add<mode>3_odd): Add insns for IEEE 128-bit floating point round
13961         to odd hardware instructions.
13962         (sub<mode>3_odd): Likewise.
13963         (mul<mode>3_odd): Likewise.
13964         (div<mode>3_odd): Likewise.
13965         (sqrt<mode>2_odd): Likewise.
13966         (fma<mode>4_odd): Likewise.
13967         (fms<mode>4_odd): Likewise.
13968         (nfma<mode>4_odd): Likewise.
13969         (nfms<mode>4_odd): Likewise.
13970         (trunc<mode>df2_odd): Change the truncate with round to odd
13971         expansion to use UNSPEC_TRUNC_ROUND_TO_ODD.  Add a generator
13972         function.
13973         * doc/extend.texi (PowerPC built-in functions): Update documentation
13974         for existing IEEE float128-bit built-in functions.  Add built-in
13975         functions that generate the IEEE 128-bit floating point round to
13976         odd instructions.
13978 2017-10-03  Segher Boessenkool  <segher@kernel.crashing.org>
13980         PR rtl-optimization/77729
13981         * simplify-rtx.c (simplify_binary_operation_1): Delete the (X&C1)|C2
13982         to (X&(C1&~C2))|C2 transformations.
13984 2017-10-03  Martin Jambor  <mjambor@suse.cz>
13986         PR tree-optimization/82363
13987         * tree-sra.c (propagate_subaccesses_across_link): In unrecoverable
13988         mismatch, mark lacc written regardless of racc.
13990 2017-10-03  Jakub Jelinek  <jakub@redhat.com>
13992         PR tree-optimization/82381
13993         * tree-ssa-reassoc.c (sort_by_operand_rank): Don't check
13994         stmt_to_insert nor wheather SSA_NAMEs are default defs.
13995         Return 1 or -1 if one of bba and bbb is NULL. If bb_rank is equal,
13996         fallthrough into reassoc_stmt_dominates_stmt_p.
13998         PR target/82386
13999         * combine.c (combine_instructions): Don't combine in unreachable
14000         basic blocks.
14002 2017-10-02  Peter Bergner  <bergner@vnet.ibm.com>
14004         PR target/80210
14005         * config/rs6000/rs6000.c (rs6000_option_override_internal): Rewrite
14006         function to not use the have_cpu variable.  Do not set cpu_index,
14007         rs6000_cpu_index or rs6000_tune_index if we end up using TARGET_DEFAULT
14008         or the default cpu.
14009         (rs6000_valid_attribute_p): Remove duplicate initializations of
14010         old_optimize and func_optimize.
14011         (rs6000_pragma_target_parse): Call rs6000_activate_target_options ().
14012         (rs6000_activate_target_options): Make global.
14013         * config/rs6000/rs6000-protos.h (rs6000_activate_target_options): Add
14014         prototype.
14016 2017-10-02  Jakub Jelinek  <jakub@redhat.com>
14018         * tree-dfa.c (get_ref_base_and_extent): Set *pmax_size to -1
14019         if *poffset + *pmax_size overflows in HOST_WIDE_INT.
14020         Set *poffset to 0 and *psize and *pmax_size to -1 if
14021         *poffset + *psize overflows in HOST_WIDE_INT.
14023         PR tree-optimization/82387
14024         PR tree-optimization/82388
14025         PR tree-optimization/82389
14026         * tree-ssa-dse.c (dse_classify_store): Test byte_tracking_enabled
14027         instead of live_bytes non-NULL.
14029 2017-10-02  Georg-Johann Lay  <avr@gjlay.de>
14031         PR target/41076
14032         * confg/avr/avr.md (*iorhi3.ashift8-ext.zerox): Add "r,r,0"
14033         alternative.
14035 2017-10-02  Richard Biener  <rguenther@suse.de>
14037         * graphite-isl-ast-to-gimple.c (set_codegen_error): With
14038         -fchecking and --param graphite-allow-codegen-errors=0 ICE.
14039         * params.def (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): New param.
14041 2017-10-02  Richard Sandiford  <richard.sandiford@linaro.org>
14043         * tree.h (wi::int_traits <const_tree>::decompose): Assert that the
14044         requested precision matches the type's.
14045         * calls.c (alloc_max_size): Calculate the new candidate size as
14046         a widest_int and use wi::to_widest when comparing it with the
14047         current candidate size.
14048         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Compare with
14049         zero rather than integer_zero_node.
14050         * match.pd: Check for a no-op conversion before using wi::add
14051         rather than after.  Use tree_to_uhwi when summing small shift
14052         counts into an unsigned int.
14054 2017-10-02  Richard Sandiford  <richard.sandiford@linaro.org>
14055             Alan Hayward  <alan.hayward@arm.com>
14056             David Sherwood  <david.sherwood@arm.com>
14058         PR target/71307
14059         * config/aarch64/aarch64.h (POINTER_AND_FP_REGS): New reg class.
14060         (REG_CLASS_NAMES, REG_CLASS_CONTENTS): Update accordingly.
14061         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
14062         POINTER_AND_FP_REGS.
14064 2017-10-02  Richard Biener  <rguenther@suse.de>
14066         PR tree-optimization/82355
14067         * graphite-isl-ast-to-gimple.c (build_iv_mapping): Also build
14068         a mapping for the enclosing loop but avoid generating one for
14069         the loop tree root.
14070         (copy_bb_and_scalar_dependences): Remove premature codegen
14071         error on PHIs in blocks duplicated into multiple places.
14072         * graphite-scop-detection.c
14073         (scop_detection::stmt_has_simple_data_refs_p): For a loop not
14074         in the region use it as loop and nest to analyze the DR in.
14075         (try_generate_gimple_bb): Likewise.
14076         * graphite-sese-to-poly.c (extract_affine_chrec): Adjust.
14077         (add_loop_constraints): For blocks in a loop not in the region
14078         create a dimension with a single iteration.
14079         * sese.h (gbb_loop_at_index): Remove assert.
14081 2017-10-01  Kevin Buettner  <kevinb@redhat.com>
14083         * omp-expand.c (adjust_context_scope): New function.
14084         (expand_parallel_call): Call adjust_context_scope.
14086 2017-10-01  Jeff Law  <law@redhat.com>
14088         * tree-ssa-dom.c (optimize_stmt): Make this a method within the
14089         dom_opt_dom_walker class with direct access to private members.
14090         Add comments.  Call test_for_singularity.
14091         (dom_opt_dom_walker::before_dom_children): Corresponding changes.
14092         (dom_opt_dom_walker::after_dom_children): Do not lazily initialize
14093         m_dummy_cond anymore.
14094         (class dom_opt_dom_walker): Initialize m_dummy_cond member in the
14095         class ctor.
14096         (pass_dominator:execute): Build the dummy_cond here and pass it
14097         to the dom_opt_dom_walker ctor.
14098         (test_for_singularity): New function.
14100 2017-09-30  Krister Walfridsson  <krister.walfridsson@gmail.com>
14101             Maya Rashish  <coypu@sdf.org>
14103         * config.gcc (*-*-netbsd*): New variable nbsd_tm_file containing
14104         netbsd.h, netbsd-stdint.h, and netbsd-elf.h.
14105         (alpha*-*-netbsd*) Use nbsd_tm_file.
14106         (arm*-*-netbsdelf*) Likewise.
14107         (i[34567]86-*-netbsdelf*) Likewise.
14108         (x86_64-*-netbsd*) Likewise.
14109         (mips*-*-netbsd*) Likewise.
14110         (powerpc-*-netbsd*) Likewise.
14111         (sh*-*-netbsd*) Likewise.
14112         (sparc-*-netbsdelf*) Likewise.
14113         (sparc64-*-netbsd*) Likewise.
14114         (m68k*-*-netbsdelf*) Use nbsd_tm_file and add CHAR_FAST8/SHORT_FAST16
14115         to tm_defines.
14116         (vax-*-netbsdelf*) Likewise.
14117         * config/netbsd-stdint.h (INT_FAST8_TYPE): Check CHAR_FAST8.
14118         (UINT_FAST8_TYPE) Likewise.
14119         (INT_FAST16_TYPE) Check CHAR_FAST16.
14120         (UINT_FAST16_TYPE) Likewise.
14122 2017-09-30  Jakub Jelinek  <jakub@redhat.com>
14124         PR target/82361
14125         * config/i386/i386.md
14126         (TARGET_USE_8BIT_IDIV zext divmodsi4 splitter): New define_split.
14127         (divmodsi4_zext_1, divmodsi4_zext_2, *divmodsi4_zext_1,
14128         *divmodsi4_zext_2): New define_insn_and_split.
14129         (*divmodsi4_noext_zext_1, *divmodsi4_noext_zext_2): New define_insn.
14130         (TARGET_USE_8BIT_IDIV zext udivmodsi4 splitter): New define_split.
14131         (udivmodsi4_zext_1, udivmodsi4_zext_2, *udivmodsi4_zext_1,
14132         *udivmodsi4_zext_2, *udivmodsi4_pow2_zext_1, *udivmodsi4_pow2_zext_2):
14133         New define_insn_and_split.
14134         (*udivmodsi4_noext_zext_1, *udivmodsi4_noext_zext_2): New define_insn.
14135         * config/i386/i386.c (ix86_split_idivmod): Handle operands[0] or
14136         operands[1] having DImode when mode is SImode.
14138         * config/i386/i386.c (ix86_split_idivmod): Use mode instead of
14139         always SImode for DIV and MOD in REG_EQUAL notes.
14141 2017-09-29  Yury Gribov  <tetra2005@gmail.com>
14143         PR middle-end/82319
14144         * match.pd: Fix handling of NaNs in pattern.
14146 2017-09-29  Jeff Law  <law@redhat.com>
14148         * sbitmap.c (bitmap_bit_in_range_p): New function.
14149         * sbitmap.h (bitmap_bit_in_range_p): Prototype.
14150         * tree-ssa-dse.c (live_bytes_read): New function.
14151         (dse_classify_store): Ignore reads of dead bytes.
14153         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Fix
14154         typos and whitespace errors.
14155         * config/i386/predicates.md (address_no_seg_operand): Likewise.
14156         * config/s390/s390.c (s390_emit_prologue): Likewise.
14158 2017-09-29  Vladimir Makarov  <vmakarov@redhat.com>
14160         PR target/81481
14161         * ira-costs.c (scan_one_insn): Don't take into account PIC equiv
14162         with a symbol for LRA.
14164 2017-09-29  Vladimir Makarov  <vmakarov@redhat.com>
14166         PR rtl-optimization/82338
14167         * lra-constraints.c (inherit_in_ebb): Check usage_insns check.
14169 2017-09-29  Alexander Monakov  <amonakov@ispras.ru>
14171         * genmodes.c (calc_wider_mode): Suppress qsort macro.
14172         * system.h [CHECKING_P] (qsort): Redirect to qsort_chk.
14173         (qsort_chk): Declare.
14174         * vec.c [CHECKING_P] (qsort_chk_error): New static function.
14175         (qsort_chk): New function.
14177 2017-09-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
14179         PR tree-optimization/82337
14180         * gimple-ssa-strength-reduction.c (find_phi_def): Don't record a
14181         phi definition if the PHI result appears in an abnormal PHI.
14182         (find_basis_for_base_expr): Don't record a basis if the LHS of the
14183         basis appears in an abnormal PHI.
14185 2017-09-29  Richard Biener  <rguenther@suse.de>
14187         * graphite-isl-ast-to-gimple.c
14188         (translate_isl_ast_to_gimple::set_codegen_error): New function.
14189         (binary_op_to_tree): Use it.
14190         (get_rename_from_scev): Likewise.
14191         (copy_loop_phi_nodes): Likewise.
14192         (copy_bb_and_scalar_dependences): Likewise.
14193         (translate_pending_phi_nodes): Likewise.
14195 2017-09-29  Jakub Jelinek  <jakub@redhat.com>
14197         PR target/82339
14198         * config/i386/i386.md (*movdi_internal peephole2): New -Os peephole
14199         for movabsq $(i32 << shift), r64.
14201 2017-09-28  Uros Bizjak  <ubizjak@gmail.com>
14203         * config/i386/i386.c (ix86_print_operand_address_as): Do not check
14204         index when encoding %esp as %rsp to avoid 0x67 prefix.
14206 2017-09-28  Sergey Shalnov  <Sergey.Shalnov@intel.com>
14208         * config/i386/i386.md (*movsf_internal, *movdf_internal):
14209         Return 256-bit AVX modes for TARGET_PREFER_AVX256.
14211 2017-09-28  Thomas Preud'homme  <thomas.preudhomme@arm.com>
14213         * config/arm/arm.c (arm_option_override): Forbid ARMv8-M Security
14214         Extensions with more than 16 double VFP registers.
14215         (cmse_nonsecure_entry_clear_before_return): Remove second entry of
14216         to_clear_mask and all code related to it.  Replace the remaining
14217         entry by a sbitmap and adapt code accordingly.
14219 2017-09-28  Henry Linjamäki  <henry.linjamaki@parmance.com>
14221         * brig-builtins.def: Change pure attributes to const.
14223 2017-09-28  Joseph Myers  <joseph@codesourcery.com>
14225         * config.gcc (default_gnu_indirect_function): Default to yes for
14226         sparc*-*-linux* with glibc.
14228 2017-09-28  Joseph Myers  <joseph@codesourcery.com>
14230         * config/aarch64/aarch64.c (aarch64_elf_asm_constructor)
14231         (aarch64_elf_asm_destructor): Pass SECTION_NOTYPE to get_section
14232         when creating .init_array and .fini_array sections with priority
14233         specified.
14235 2017-09-27  Christophe Lyon  <christophe.lyon@linaro.org>
14237         PR target/71727
14238         * config/aarch64/aarch64.c
14239         (aarch64_builtin_support_vector_misalignment): Always return false
14240         when misalignment is unknown.
14242 2017-09-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14244         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Revise
14245         this function to return false if the definition used by the swap
14246         instruction is artificial, or if the memory address from which the
14247         constant value is loaded is not represented by a base address held
14248         in a register or if the base address register is a frame or stack
14249         pointer.  Additionally, return false if the base address of the
14250         loaded constant is a SYMBOL_REF but is not considered to be a
14251         constant.
14252         (replace_swapped_load_constant): New function.
14253         (rs6000_analyze_swaps): Add a new pass to replace a swap of a
14254         loaded constant vector with a load of a swapped constant vector.
14256 2017-09-27  Carl Love  <cel@us.ibm.com>
14258         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
14259         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
14260         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
14261         fctiw instruction.
14263 2017-09-27  Alexander Monakov  <amonakov@ispras.ru>
14265         * haifa-sched.c (autopref_rank_for_schedule): Order 'irrelevant' insns
14266         first, always call autopref_rank_data otherwise.
14268 2017-09-27  Richard Biener  <rguenther@suse.de>
14270         * graphite-scop-detection.c (find_scop_parameters): Move
14271         loop bound handling ...
14272         (gather_bbs::before_dom_children): ... here, avoiding the need
14273         to build scop_info->loop_nest.
14274         (record_loop_in_sese): Remove.
14275         * sese.h (sese_info_t::loop_nest): Remove.
14276         * sese.c (new_sese_info): Do not allocate loop_nest.
14277         (free_sese_info): Do not free loop_nest.
14279 2017-09-27  Jakub Jelinek  <jakub@redhat.com>
14281         PR c++/82159
14282         * gimplify.c (gimplify_modify_expr): Don't optimize away zero sized
14283         lhs from calls if the lhs has addressable type.
14285 2017-09-27  Richard Biener  <rguenther@suse.de>
14287         * graphite.h (scop::max_alias_set): New member.
14288         * graphite-scop-detection.c: Remove references to non-existing
14289         --param in comments.
14290         (build_alias_sets): Record the maximum alias set used for drs.
14291         (build_scops): Support zero as unlimited for
14292         --param graphite-max-arrays-per-scop.
14293         * graphite-sese-to-poly.c (add_scalar_version_numbers): Remove
14294         and inline into ...
14295         (build_poly_sr_1): ... here.  Compute alias set based on the
14296         maximum alias set used for drs rather than
14297         PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP
14299 2017-09-27  Richard Biener  <rguenther@suse.de>
14301         * graphite-optimize-isl.c (get_schedule_for_node_st): Allow
14302         --param loop-block-tile-size=0 to disable tiling.
14304 2017-09-27  Richard Biener  <rguenther@suse.de>
14306         * doc/invoke.texi (graphite-max-bbs-per-function): Remove.
14307         (graphite-max-nb-scop-params): Document special value zero.
14308         * domwalk.h (dom_walker::STOP): New symbolical constant.
14309         (dom_walker::dom_walker): Add optional parameter for bb to
14310         RPO mapping.
14311         (dom_walker::~dom_walker): Declare.
14312         (dom_walker::before_dom_children): Document STOP return value.
14313         (dom_walker::m_user_bb_to_rpo): New member.
14314         (dom_walker::m_bb_to_rpo): Likewise.
14315         * domwalk.c (dom_walker::dom_walker): Compute bb to RPO
14316         mapping here if not provided by the user.
14317         (dom_walker::~dom_walker): Free bb to RPO mapping if not
14318         provided by the user.
14319         (dom_walker::STOP): Define.
14320         (dom_walker::walk): Do not compute bb to RPO mapping here.
14321         Support STOP return value from before_dom_children to stop
14322         walking.
14323         * graphite-optimize-isl.c (optimize_isl): If the schedule
14324         is the same still generate code if -fgraphite-identity
14325         or -floop-parallelize-all are given.
14326         * graphite-scop-detection.c: Include cfganal.h.
14327         (gather_bbs::gather_bbs): Get and pass through bb to RPO
14328         mapping.
14329         (gather_bbs::before_dom_children): Return STOP for BBs
14330         not in the region.
14331         (build_scops): Compute bb to RPO mapping and pass it to
14332         the domwalk.  Treat --param graphite-max-nb-scop-params=0
14333         as not limiting the number of params.
14334         * graphite.c (graphite_initialize): Remove limit on the
14335         number of basic-blocks in a function.
14336         * params.def (PARAM_GRAPHITE_MAX_BBS_PER_FUNCTION): Remove.
14337         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Adjust to documented
14338         default value of 10.
14340 2017-09-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
14342         * config/rs6000/vsx.md (peephole for optimizing move SF to GPR):
14343         Adjust code to eliminate needing to do the shift right 32-bits
14344         operation after XSCVDPSPN.
14346 2017-09-26  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
14348         * match.pd ((X / Y) == 0 -> X < Y): New pattern.
14349         ((X / Y) != 0 -> X >= Y): Likewise.
14351 2017-09-26  Carl Love  <cel@us.ibm.com>
14353         * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_XL_LEN_R,
14354         P9V_BUILTIN_VEC_XST_LEN_R): Add support for builtins
14355         vector unsigned char vec_xl_len_r (unsigned char *, size_t);
14356         void vec_xst_len_r (vector unsigned char, unsigned char *, size_t);
14357         * config/rs6000/altivec.h (vec_xl_len_r, vec_xst_len_r): Add defines.
14358         * config/rs6000/rs6000-builtin.def (XL_LEN_R, XST_LEN_R): Add
14359         definitions and overloading.
14360         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case
14361         statement for P9V_BUILTIN_XST_LEN_R.
14362         (altivec_init_builtins): Add def_builtin for P9V_BUILTIN_STXVLL.
14363         * config/rs6000/vsx.md (lxvll, stxvll, xl_len_r, xst_len_r): Add
14364         define_expand and define_insn for the instructions and builtins.
14365         * doc/extend.texi: Update the built-in documentation file for the new
14366         built-in functions.
14367         * config/rs6000/altivec.md (altivec_lvsl_reg, altivec_lvsr_reg): Add
14368         define_insn for the instructions
14370 2017-09-26  Krister Walfridsson  <krister.walfridsson@gmail.com>
14372         PR target/39570
14373         * gcc/config/netbsd-protos.h: New file.
14374         * gcc/config/netbsd.c: New file.
14375         * gcc/config/netbsd.h (SUBTARGET_INIT_BUILTINS): Define.
14376         * gcc/config/t-netbsd: New file.
14377         * gcc/config.gcc (tm_p_file): Add netbsd-protos.h.
14378         (tmake_file) Add t-netbsd.
14379         (extra_objs) Add netbsd.o.
14381 2017-09-26  Janus Weil  <janus@gcc.gnu.org>
14383         PR fortran/82143
14384         PR fortran/82324
14385         * doc/sourcebuild.texi: Document fortran_real_10 and fortran_real_16.
14387 2017-09-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
14389         * config/rs6000/rs6000.md (extendsi<mode>2): Add a splitter to do
14390         sign extension from a vector register to a GPR by doing a 32-bit
14391         direct move and then an EXTSW.
14392         (extendsi<mode>2 splitter): Likewise.
14393         (movsi_from_sf): Adjust code to eliminate doing a 32-bit shift
14394         right or vector extract after doing XSCVDPSPN.  Use
14395         zero_extendsidi2 instead of p8_mfvsrd_4_disf to move the value to
14396         the GPRs.
14397         (movdi_from_sf_zero_ext): Likewise.
14398         (reload_gpr_from_vsxsf): Likewise.
14399         (p8_mfvsrd_4_disf): Delete, no longer used.
14400         (movsi_from_df): Optimize converting a DFmode to a SFmode, and
14401         then needing to move the SFmode to a GPR to use the XSCVDPSP
14402         instruction instead of FRSP and XSCVDPSPN.
14403         * config/rs6000/vsx.md (vsx_xscvspdp_scalar2): Move insn so that
14404         it is adjacent to the other XSCVSPDP insns.
14405         (vsx_xscvdpsp_scalar): Use "ww" constraint instead of "f" to allow
14406         SFmode to be in traditional Altivec registers.
14407         (vsx_xscvdpspn): Eliminate useless alternative constraint.
14408         (vsx_xscvspdpn): Likewise.
14409         (vsx_xscvspdpn_scalar): Likewise.
14411 2017-09-26  Martin Jambor  <mjambor@suse.cz>
14413         * tree-sra.c (compare_access_positions): Put integral types first,
14414         stabilize sorting of integral types, remove conditions putting
14415         non-full-precision integers last.
14416         (sort_and_splice_var_accesses): Disable scalarization if a
14417         non-integert would be represented by a non-full-precision integer.
14419 2017-09-26  Joseph Myers  <joseph@codesourcery.com>
14421         * config/microblaze/linux.h (TARGET_ASM_FILE_END): Likewise.
14422         * config/pa/pa.h (NEED_INDICATE_EXEC_STACK): Likewise.
14423         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Likewise.
14424         * config/pa/pa.c (pa_hpux_file_end): Rename to pa_file_end.
14425         Define unconditionally, with [ASM_OUTPUT_EXTERNAL_REAL]
14426         conditionals inside the function instead of around it.  Call
14427         file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK.
14428         (TARGET_ASM_FILE_END): Define unconditionally to pa_file_end.
14430 2017-09-26  Richard Biener  <rguenther@suse.de>
14432         * graphite-scop-detection.c (scop_detection::build_scop_depth): Rewrite,
14433         fold in ...
14434         (scop_detection::build_scop_breadth): ... this.  Removed.
14435         (scop_detection::loop_is_valid_in_scop): Fold into single caller.
14436         (scop_detection::harmful_stmt_in_bb): Likewise.
14437         (scop_detection::graphite_can_represent_stmt): Likewise.
14438         (scop_detection::loop_body_is_valid_scop): Likewise.  Remove recursion.
14439         (scop_detection::can_represent_loop): Remove recursion, fold in ...
14440         (scop_detection::can_represent_loop_1): ... this.  Removed.
14441         (scop_detection::harmful_loop_in_region): Simplify after inlining
14442         the above and remove more quadraticness.
14443         (build_scops): Adjust.
14444         * tree-data-ref.c (loop_nest_has_data_refs): Remove pointless
14445         quadraticness.
14447 2017-09-26  Jakub Jelinek  <jakub@redhat.com>
14449         PR target/82267
14450         * config/i386/i386.c (ix86_print_operand_address_as): Only test
14451         REGNO (base) == SP_REG if base is a REG.
14453         PR middle-end/35691
14454         * tree-ssa-reassoc.c (update_range_test): Dump r->exp each time
14455         if it is different SSA_NAME.
14456         (optimize_range_tests_cmp_bitwise): New function.
14457         (optimize_range_tests): Call it.
14459 2017-09-26  Richard Biener  <rguenther@suse.de>
14461         PR tree-optimization/82321
14462         * graphite.c (canonicalize_loop_closed_ssa): Properly check
14463         for the def being inside the loop.
14465 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14467         * config/s390/vx-builtins.md ("vmslg"): Add missing operand in
14468         assembler output.
14469         * config/s390/s390-builtins.def: Fix constraint on op4.
14471 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14473         * config/s390/s390.c (s390_expand_vec_compare): Use the new mode
14474         independent expanders.
14475         * config/s390/vector.md ("vec_cmpuneq", "vec_cmpltgt")
14476         ("vec_ordered", "vec_unordered"): New expanders.
14478 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14480         * config/s390/s390.c (s390_preferred_simd_mode): Return V4SFmode
14481         for SFmode.
14483 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14485         * config/s390/vector.md ("vec_unpacks_low_v16qi"): Rename to
14486         vec_unpacks_lo_v16qi.
14487         ("vec_unpacku_low_v16qi"): Rename to vec_unpacku_lo_v16qi.
14489 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14491         * config/s390/vector.md ("vec_unpacks_lo_v4sf")
14492         ("vec_unpacks_hi_v4sf", "vec_unpacks_lo_v2df")
14493         ("vec_unpacks_hi_v2df", "vec_pack_trunc_v2df"): New expanders.
14495 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14497         * config/s390/predicates.md ("const_shift_by_byte_operand"): New
14498         predicate.
14499         * config/s390/vector.md ("*vec_srb<mode>"): Change modes to V_128
14500         and V16QI.
14501         ("*vec_slb<mode>"): New insn pattern.
14502         ("vec_shr_<mode>"): New expander.
14503         * config/s390/vx-builtins.md ("vec_slb<mode>"): Turn into expander
14504         and force the shift count operand to V16QImode.
14505         ("vec_srb<mode>"): Set shift count mode to V16QI.
14507 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14509         * config/s390/vector.md ("vec_widen_umult_lo_<mode>")
14510         ("vec_widen_umult_hi_<mode>", "vec_widen_smult_lo_<mode>")
14511         ("vec_widen_smult_hi_<mode>"): New expander definitions.
14513 2017-09-26  Richard Earnshaw  <rearnsha@arm.com>
14515         PR target/82175
14516         * config/arm/arm.h (DRIVER_SELF_SPECS): Separate sub-rules with commas.
14518 2017-09-26  Richard Biener  <rguenther@suse.de>
14520         PR tree-optimization/82320
14521         * tree-ssa-sccvn.c (set_ssa_val_to): Changing undef to undef
14522         isn't a change.
14524 2017-09-25  Jeff Law  <law@redhat.com>
14526         * config/rs6000/rs6000-protos.h (output_probe_stack_range): Update
14527         prototype for new argument.
14528         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack_1): New function,
14529         mostly extracted from rs6000_emit_allocate_stack.
14530         (rs6000_emit_probe_stack_range_stack_clash): New function.
14531         (rs6000_emit_allocate_stack): Call
14532         rs6000_emit_probe_stack_range_stack_clash as needed.
14533         (rs6000_emit_probe_stack_range): Add additional argument
14534         to call to gen_probe_stack_range{si,di}.
14535         (output_probe_stack_range): New.
14536         (output_probe_stack_range_1): Renamed from output_probe_stack_range.
14537         (output_probe_stack_range_stack_clash): New.
14538         (rs6000_emit_prologue): Emit notes into dump file as requested.
14539         * rs6000.md (allocate_stack): Handle -fstack-clash-protection.
14540         (probe_stack_range<P:mode>): Operand 0 is now early-clobbered.
14541         Add additional operand and pass it to output_probe_stack_range.
14543 2017-09-25  Bin Cheng  <bin.cheng@arm.com>
14545         PR tree-optimization/82163
14546         * tree-ssa-loop-manip.h (verify_loop_closed_ssa): New parameter.
14547         (checking_verify_loop_closed_ssa): New parameter.
14548         * tree-ssa-loop-manip.c (check_loop_closed_ssa_use): Delete.
14549         (check_loop_closed_ssa_stmt): Delete.
14550         (check_loop_closed_ssa_def, check_loop_closed_ssa_bb): New functions.
14551         (verify_loop_closed_ssa): Check loop closed ssa form for LOOP.
14552         (tree_transform_and_unroll_loop): Check loop closed ssa form only for
14553         changed loops.
14555 2017-09-25  Pekka Jaaskelainen <pekka@parmance.com>
14557         * brig-builtins.def: Treat HSAIL barrier builtins as
14558         setjmp/longjump style functions.
14560 2017-09-25  Richard Sandiford  <richard.sandiford@linaro.org>
14562         * target.def (constant_alignment): New hook.
14563         * defaults.h (CONSTANT_ALIGNMENT): Delete.
14564         * doc/tm.texi.in (CONSTANT_ALIGNMENT): Replace with...
14565         (TARGET_CONSTANT_ALIGNMENT): ...this new hook.
14566         * doc/tm.texi: Regenerate.
14567         * targhooks.h (default_constant_alignment): Declare.
14568         (constant_alignment_word_strings): Likewise.
14569         * targhooks.c (default_constant_alignment): New function.
14570         (constant_alignment_word_strings): Likewise.
14571         * builtins.c (get_object_alignment_2): Use targetm.constant_alignment
14572         instead of CONSTANT_ALIGNMENT.
14573         * varasm.c (align_variable, get_variable_align, build_constant_desc)
14574         (force_const_mem): Likewise.
14575         * config/aarch64/aarch64.h (CONSTANT_ALIGNMENT): Delete.
14576         * config/aarch64/aarch64.c (aarch64_constant_alignment): New function.
14577         (aarch64_classify_address): Call it instead of CONSTANT_ALIGNMENT.
14578         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14579         * config/alpha/alpha.h (CONSTANT_ALIGNMENT): Delete commented-out
14580         definition.
14581         * config/arc/arc.h (CONSTANT_ALIGNMENT): Delete.
14582         * config/arc/arc.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14583         constant_alignment_word_strings.
14584         * config/arm/arm.h (CONSTANT_ALIGNMENT_FACTOR): Delete.
14585         (CONSTANT_ALIGNMENT): Likewise.
14586         * config/arm/arm.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14587         (arm_constant_alignment): New function.
14588         * config/bfin/bfin.h (CONSTANT_ALIGNMENT): Delete.
14589         * config/bfin/bfin.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14590         constant_alignment_word_strings.
14591         * config/cr16/cr16.h (CONSTANT_ALIGNMENT): Delete.
14592         * config/cr16/cr16.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14593         constant_alignment_word_strings.
14594         * config/cris/cris.h (CONSTANT_ALIGNMENT): Delete.
14595         * config/cris/cris.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14596         (cris_constant_alignment): New function.
14597         * config/epiphany/epiphany.h (CONSTANT_ALIGNMENT): Delete.
14598         * config/epiphany/epiphany.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14599         (epiphany_constant_alignment): New function.
14600         * config/fr30/fr30.h (CONSTANT_ALIGNMENT): Delete.
14601         * config/fr30/fr30.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14602         constant_alignment_word_strings.
14603         * config/frv/frv.h (CONSTANT_ALIGNMENT): Delete.
14604         * config/frv/frv.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14605         constant_alignment_word_strings.
14606         * config/ft32/ft32.h (CONSTANT_ALIGNMENT): Delete.
14607         * config/ft32/ft32.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14608         constant_alignment_word_strings.
14609         * config/i386/i386.h (CONSTANT_ALIGNMENT): Delete.
14610         * config/i386/i386-protos.h (ix86_constant_alignment): Delete.
14611         * config/i386/i386.c (ix86_constant_alignment): Make static.
14612         Use the same interface as the target hook.
14613         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14614         * config/ia64/ia64.h (CONSTANT_ALIGNMENT): Delete.
14615         * config/ia64/ia64.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14616         constant_alignment_word_strings.
14617         * config/iq2000/iq2000.h (CONSTANT_ALIGNMENT): Delete.
14618         * config/iq2000/iq2000.c (iq2000_constant_alignment): New function.
14619         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14620         * config/lm32/lm32.h (CONSTANT_ALIGNMENT): Delete.
14621         * config/lm32/lm32.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14622         constant_alignment_word_strings.
14623         * config/m32r/m32r.h (CONSTANT_ALIGNMENT): Delete.
14624         * config/m32r/m32r.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14625         constant_alignment_word_strings.
14626         * config/mcore/mcore.h (CONSTANT_ALIGNMENT): Delete.
14627         * config/mcore/mcore.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14628         constant_alignment_word_strings.
14629         * config/microblaze/microblaze.h (CONSTANT_ALIGNMENT): Delete.
14630         * config/microblaze/microblaze.c (microblaze_constant_alignment):
14631         New function.
14632         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14633         * config/mips/mips.h (CONSTANT_ALIGNMENT): Delete.
14634         * config/mips/mips.c (mips_constant_alignment): New function.
14635         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14636         * config/mmix/mmix.h (CONSTANT_ALIGNMENT): Delete.
14637         * config/mmix/mmix-protos.h (mmix_constant_alignment): Delete.
14638         * config/mmix/mmix.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14639         (mmix_constant_alignment): Make static.  Use the same interface
14640         as the target hook.
14641         * config/moxie/moxie.h (CONSTANT_ALIGNMENT): Delete.
14642         * config/moxie/moxie.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14643         constant_alignment_word_strings.
14644         * config/nios2/nios2.h (CONSTANT_ALIGNMENT): Delete.
14645         * config/nios2/nios2.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14646         constant_alignment_word_strings.
14647         * config/pa/pa.h (CONSTANT_ALIGNMENT): Delete.
14648         * config/pa/pa.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14649         constant_alignment_word_strings.
14650         * config/powerpcspe/powerpcspe.h (CONSTANT_ALIGNMENT): Delete.
14651         * config/powerpcspe/powerpcspe.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14652         (rs6000_constant_alignment): New function.
14653         * config/riscv/riscv.h (CONSTANT_ALIGNMENT): Delete.
14654         * config/riscv/riscv.c (riscv_constant_alignment): New function.
14655         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14656         * config/rs6000/rs6000.h (CONSTANT_ALIGNMENT): Delete.
14657         * config/rs6000/rs6000.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14658         (rs6000_constant_alignment): New function.
14659         * config/s390/s390.h (CONSTANT_ALIGNMENT): Delete.
14660         * config/s390/s390.c (s390_constant_alignment): New function.
14661         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14662         * config/sh/sh.h (CONSTANT_ALIGNMENT): Delete.
14663         * config/sh/sh.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14664         constant_alignment_word_strings.
14665         * config/sparc/sparc.h (CONSTANT_ALIGNMENT): Delete.
14666         * config/sparc/sparc.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14667         (sparc_constant_alignment): New function.
14668         * config/spu/spu.h (CONSTANT_ALIGNMENT): Delete.
14669         * config/spu/spu.c (spu_constant_alignment): New function.
14670         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14671         * config/stormy16/stormy16.h (CONSTANT_ALIGNMENT): Delete.
14672         * config/stormy16/stormy16.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14673         constant_alignment_word_strings.
14674         * config/tilegx/tilegx.h (CONSTANT_ALIGNMENT): Delete.
14675         * config/tilegx/tilegx.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14676         constant_alignment_word_strings.
14677         * config/tilepro/tilepro.h (CONSTANT_ALIGNMENT): Delete.
14678         * config/tilepro/tilepro.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14679         constant_alignment_word_strings.
14680         * config/visium/visium.h (CONSTANT_ALIGNMENT): Delete.
14681         * config/visium/visium.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14682         (visium_constant_alignment): New function.
14683         * config/xtensa/xtensa.h (CONSTANT_ALIGNMENT): Delete.
14684         * config/xtensa/xtensa.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14685         (xtensa_constant_alignment): New function.
14686         * system.h (CONSTANT_ALIGNMENT): Poison.
14688 2017-09-25  Will Schmidt  <will_schmidt@vnet.ibm.com>
14690         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
14691         for early folding of vector stores (ALTIVEC_BUILTIN_ST_*).
14692         (rs6000_builtin_valid_without_lhs): New helper function.
14693         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
14694         Remove obsoleted code for handling ALTIVEC_BUILTIN_VEC_ST.
14696 2017-09-25  Richard Sandiford  <richard.sandiford@linaro.org>
14698         * target.h (vec_perm_indices): Use unsigned short rather than
14699         unsigned char.
14700         (auto_vec_perm_indices): Likewise.
14701         * config/aarch64/aarch64.c (aarch64_vectorize_vec_perm_const_ok):
14702         Use unsigned int rather than unsigned char.
14703         * config/arm/arm.c (arm_vectorize_vec_perm_const_ok): Likewise.
14705 2017-09-25  Richard Biener  <rguenther@suse.de>
14707         * cfgloop.h (sort_sibling_loops): Declare.
14708         * cfgloop.c (sort_sibling_loops_cmp): New helper.
14709         (sort_sibling_loops): New function sorting the sibling loop list
14710         in RPO order.
14711         * graphite.c (graphite_transform_loops): Sort sibling loops.
14713 2017-09-25  Richard Sandiford  <richard.sandifird@linaro.org>
14715         * target.def (vec_perm_const_ok): Change sel parameter to
14716         vec_perm_indices.
14717         * optabs-query.c (can_vec_perm_p): Update accordingly.
14718         * doc/tm.texi: Regenerate.
14719         * config/aarch64/aarch64.c (expand_vec_perm_d): Change perm to
14720         auto_vec_perm_indices and remove separate nelt field.
14721         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
14722         (aarch64_evpc_ext, aarch64_evpc_rev, aarch64_evpc_dup)
14723         (aarch64_evpc_tbl, aarch64_expand_vec_perm_const_1)
14724         (aarch64_expand_vec_perm_const): Update accordingly.
14725         (aarch64_vectorize_vec_perm_const_ok): Likewise.  Change sel
14726         to vec_perm_indices.
14727         * config/arm/arm.c (expand_vec_perm_d): Change perm to
14728         auto_vec_perm_indices and remove separate nelt field.
14729         (arm_evpc_neon_vuzp, arm_evpc_neon_vzip, arm_evpc_neon_vrev)
14730         (arm_evpc_neon_vtrn, arm_evpc_neon_vext, arm_evpc_neon_vtbl)
14731         (arm_expand_vec_perm_const_1, arm_expand_vec_perm_const): Update
14732         accordingly.
14733         (arm_vectorize_vec_perm_const_ok): Likewise.  Change sel
14734         to vec_perm_indices.
14735         * config/i386/i386.c (ix86_vectorize_vec_perm_const_ok): Change
14736         sel to vec_perm_indices.
14737         * config/ia64/ia64.c (ia64_vectorize_vec_perm_const_ok): Likewise.
14738         * config/mips/mips.c (mips_vectorize_vec_perm_const_ok): Likewise.
14739         * config/powerpcspe/powerpcspe.c (rs6000_vectorize_vec_perm_const_ok):
14740         Likewise.
14741         * config/rs6000/rs6000.c (rs6000_vectorize_vec_perm_const_ok):
14742         Likewise.
14744 2017-09-25  Pierre-Marie de Rodat  <derodat@adacore.com>
14746         PR debug/82155
14747         * dwarf2out.c (dwarf2out_early_global_decl): Call dwarf2out_decl
14748         on the FUNCTION_DECL function context if it has a DIE that is a
14749         declaration.
14751 2017-09-25  Richard Biener  <rguenther@suse.de>
14753         PR tree-optimization/82285
14754         * tree-vect-patterns.c (vect_recog_bool_pattern): Also handle
14755         enumeral types.
14757 2017-09-25  Tom de Vries  <tom@codesourcery.com>
14759         PR target/80035
14760         PR target/81069
14761         * config/nvptx/nvptx.c (nvptx_output_call_insn): Add exit after call to
14762         noreturn function.
14764 2017-09-25  Richard Biener  <rguenther@suse.de>
14766         * graphite-optimize-isl.c (optimize_isl): Fail and dump if
14767         ISL errors other than isl_error_quota happen.  Dump if the
14768         schedule is the same.
14769         * graphite-sese-to-poly.c (build_poly_scop): Fail on ISL
14770         errors instead of aborting inside ISL.
14772 2017-09-25  Iain Sandoe  <iain@codesourcery.com>
14774         PR target/80556
14775         * config/i386/darwin.h (REAL_LIB_SPEC): New; put libSystem ahead
14776         of libgcc_eh for m64.
14777         * config/i386/darwin64.h: Likewise.
14779 2017-09-25  Richard Biener  <rguenther@suse.de>
14781         PR middle-end/82144
14782         * dwarf2out.c (gen_enumeration_type_die): Do not add alignment
14783         attribute for incomplete types nor twice for complete ones.
14785 2017-09-24  Uros Bizjak  <ubizjak@gmail.com>
14787         PR target/82267
14788         * config/i386/i386.c (ix86_print_operand_address_as): Encode
14789         %esp as %rsp to avoid 0x67 prefix if there is no index or base
14790         register.
14792 2017-09-23  Uros Bizjak  <ubizjak@gmail.com>
14794         PR bootstrap/82306
14795         * config/i386/i386.opt (mprefer-avx256): Use
14796         ix86_target_flags variable.
14797         * config/i386/i386.c (ix86_target_string): Move
14798         -mprefer-avx256 to flag2_opts.
14800 2017-09-22  Jakub Jelinek  <jakub@redhat.com>
14802         PR middle-end/35691
14803         * match.pd: Simplify x == -1 & y == -1 into (x & y) == -1
14804         and x != -1 | y != -1 into (x & y) != -1.
14806 2017-09-22  Steve Ellcey  <sellcey@cavium.com>
14808         * config.gcc: Add new case statement to set
14809         default_gnu_indirect_function.  Remove it from x86_64-*-linux*,
14810         i[34567]86-*, powerpc*-*-linux*spe*, powerpc*-*-linux*, s390-*-linux*,
14811         s390x-*-linux* case statements.   Added aarch64 to the list of
14812         supported architectures.
14814 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14816         PR tree-optimization/82289
14817         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs): Check
14818         STMT_VINFO_RELEVANT_P.
14820 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14821             Alan Hayward  <alan.hayward@arm.com>
14822             David Sherwood  <david.sherwood@arm.com>
14824         * tree-vrp.c (extract_range_from_multiplicative_op_1): Assert
14825         for VR_RANGE only; don't allow VR_ANTI_RANGE.
14826         (extract_range_from_binary_expr_1): Don't call
14827         extract_range_from_multiplicative_op_1 if !range_int_cst_p.
14829 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14830             Alan Hayward  <alan.hayward@arm.com>
14831             David Sherwood  <david.sherwood@arm.com>
14833         * target.def (preferred_vector_alignment): New hook.
14834         * doc/tm.texi.in (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): New
14835         hook.
14836         * doc/tm.texi: Regenerate.
14837         * targhooks.h (default_preferred_vector_alignment): Declare.
14838         * targhooks.c (default_preferred_vector_alignment): New function.
14839         * tree-vectorizer.h (dataref_aux): Add a target_alignment field.
14840         Expand commentary.
14841         (DR_TARGET_ALIGNMENT): New macro.
14842         (aligned_access_p): Update commentary.
14843         (vect_known_alignment_in_bytes): New function.
14844         * tree-vect-data-refs.c (vect_calculate_required_alignment): New
14845         function.
14846         (vect_compute_data_ref_alignment): Set DR_TARGET_ALIGNMENT.
14847         Calculate the misalignment based on the target alignment rather than
14848         the vector size.
14849         (vect_update_misalignment_for_peel): Use DR_TARGET_ALIGMENT
14850         rather than TYPE_ALIGN / BITS_PER_UNIT to update the misalignment.
14851         (vect_enhance_data_refs_alignment): Mask the byte misalignment with
14852         the target alignment, rather than masking the element misalignment
14853         with the number of elements in a vector.  Also use the target
14854         alignment when calculating the maximum number of peels.
14855         (vect_find_same_alignment_drs): Use vect_calculate_required_alignment
14856         instead of TYPE_ALIGN_UNIT.
14857         (vect_duplicate_ssa_name_ptr_info): Remove stmt_info parameter.
14858         Measure DR_MISALIGNMENT relative to DR_TARGET_ALIGNMENT.
14859         (vect_create_addr_base_for_vector_ref): Update call accordingly.
14860         (vect_create_data_ref_ptr): Likewise.
14861         (vect_setup_realignment): Realign by ANDing with
14862         -DR_TARGET_MISALIGNMENT.
14863         * tree-vect-loop-manip.c (vect_gen_prolog_loop_niters): Calculate
14864         the number of peels based on DR_TARGET_ALIGNMENT.
14865         * tree-vect-stmts.c (get_group_load_store_type): Compare the gap
14866         with the guaranteed alignment boundary when deciding whether
14867         overrun is OK.
14868         (vectorizable_mask_load_store): Interpret DR_MISALIGNMENT
14869         relative to DR_TARGET_ALIGNMENT instead of TYPE_ALIGN_UNIT.
14870         (ensure_base_align): Remove stmt_info parameter.  Get the
14871         target base alignment from DR_TARGET_ALIGNMENT.
14872         (vectorizable_store): Update call accordingly.   Interpret
14873         DR_MISALIGNMENT relative to DR_TARGET_ALIGNMENT instead of
14874         TYPE_ALIGN_UNIT.
14875         (vectorizable_load): Likewise.
14877 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14878             Alan Hayward  <alan.hayward@arm.com>
14879             David Sherwood  <david.sherwood@arm.com>
14881         * tree-vectorizer.h (vect_get_scalar_dr_size): New function.
14882         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Use it.
14883         (vect_enhance_data_refs_alignment): Likewise.
14885 2017-09-22  Richard Earnshaw  <richard.earnshaw@arm.com>
14887         * config/arm/parsecpu.awk (fatal): Note that we've encountered an
14888         error.  Only quit immediately if parsing is complete.
14889         (BEGIN): Initialize fatal_err and parse_done.
14890         (begin fpu, end fpu): Check number of arguments.
14891         (begin arch, end arch): Likewise.
14892         (begin cpu, end cpu): Likewise.
14893         (cname, tune for, tune flags, architecture, fpu, option): Likewise.
14894         (optalias): Likewise.
14896 2017-09-22  Richard Earnshaw  <richard.earnshaw@arm.com>
14898         * config.gcc (arm*-*-*): Don't add arm-isa.h to tm_p_file.
14899         * config/arm/arm-isa.h: Delete.  Move definitions to ...
14900         * arm-cpus.in: ... here.  Use new feature and fgroup values.
14901         * config/arm/arm.c (arm_option_override): Use lower case for feature
14902         bit names.
14903         * config/arm/arm.h (TARGET_HARD_FLOAT): Likewise.
14904         (TARGET_VFP3, TARGET_VFP5, TARGET_FMA): Likewise.
14905         * config/arm/parsecpu.awk (END): Add new command 'isa'.
14906         (isa_pfx): Delete.
14907         (print_isa_bits_for): New function.
14908         (gen_isa): New function.
14909         (gen_comm_data): Use print_isa_bits_for.
14910         (define feature): New keyword.
14911         (define fgroup): New keyword.
14912         * config/arm/t-arm (TM_H): Remove.
14913         (GTM_H): Add arm-isa.h.
14914         (arm-isa.h): Add rule to generate file.
14915         * common/config/arm/arm-common.c: (arm_canon_arch_option): Use lower
14916         case for feature bit names.
14918 2017-09-22  Richard Biener  <rguenther@suse.de>
14920         * graphite-isl-ast-to-gimple.c (graphite_verify): Inline into
14921         single caller.
14922         (graphite_regenerate_ast_isl): Do not reset SCEV.  Move debug
14923         print of no dependency loops ...
14924         * graphite.c (graphite_transform_loops): ... here.
14925         (canonicalize_loop_closed_ssa_form): Work from inner to outer
14926         loops.
14927         (same_close_phi_node, remove_duplicate_close_phi,
14928         make_close_phi_nodes_unique, defined_in_loop_p): Fold into ...
14929         (canonicalize_loop_closed_ssa): ... here and simplify.
14930         * graphite-optimize-isl.c: Include tree-vectorizer.h.
14931         (optimize_isl): Use dump_printf_loc to tell when we stopped
14932         optimizing because of an ISL timeout.
14934 2017-09-22  Richard Biener  <rguenther@suse.de>
14936         PR tree-optimization/82291
14937         * tree-if-conv.c (predicate_mem_writes): Make sure to
14938         remove writes in blocks predicated with false.
14940 2017-09-22  Richard Biener  <rguenther@suse.de>
14942         * sese.c: Include cfganal.h.
14943         (if_region_set_false_region): Remove.
14944         (create_if_region_on_edge): Likewise.
14945         (move_sese_in_condition): Re-implement without destroying
14946         dominators.
14948 2017-09-22  Richard Biener  <rguenther@suse.de>
14950         * graphite-isl-ast-to-gimple.c (translate_pending_phi_nodes):
14951         Verify both BBs contain loop PHI nodes before dispatching to
14952         copy_loop_phi_args.
14953         (graphite_regenerate_ast_isl): Do not recompute dominators,
14954         do not verify three times.  Restructure for clarity.
14955         * graphite-scop-detection.c (same_close_phi_node,
14956         remove_duplicate_close_phi, make_close_phi_nodes_unique,
14957         defined_in_loop_p, canonicalize_loop_closed_ssa,
14958         canonicalize_loop_closed_ssa_form): Simplify, remove excess
14959         checking and SSA rewrite, move to ...
14960         * graphite.c: ... here.  Include ssa.h and tree-ssa-loop-manip.h.
14961         (graphite_initialize): Do not pass in ctx, do not reset the
14962         SCEV cache, compute only dominators.
14963         (graphite_transform_loops): Allocate ISL ctx after
14964         graphite_initialize.  Call canonicalize_loop_closed_ssa_form.
14965         Maintain post-dominators only around build_scops.
14966         * sese.c (if_region_set_false_region): Make static.  Free
14967         and recompute dominators.
14968         (move_sese_in_condition): Assert we don't get called with
14969         post-dominators computed.
14970         * sese.h (if_region_set_false_region): Remove.
14972 2017-09-22  Sergey Shalnov  <sergey.shalnov@intel.com>
14974         * config/i386/sse.md ("mov<mode>_internal"): Use <sseinsnmode>
14975         mode attribute for TARGET_AVX512VL.
14977 2017-09-21  Sergey Shalnov  <sergey.shalnov@intel.com>
14979         * config/i386/i386.opt (mprefer-avx256): New option.
14980         * config/i386/i386.c (ix86_target_string): Add -mprefer-avx256
14981         to flag_opts.
14982         (ix86_preferred_simd_mode): Return 256-bit AVX modes
14983         for TARGET_PREFER_AVX256.
14984         * doc/invoke.texi (x86 Options): Document -mprefer-avx256.
14986 2017-09-21  Jeff Law  <law@redhat.com>
14988         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash):
14989         Fix dump output if the only stack space is for pushed registers.
14991 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14993         * config/spu/spu.c (spu_sched_adjust_cost): Update after renaming
14994         of insn_cost.
14996 2017-09-21  Martin Sebor  <msebor@redhat.com>
14998         PR c/81882
14999         * doc/extend.texi (attribute ifunc): Avoid relying on ill-formed
15000         code (in C++) or code that triggers warnings.
15002 2017-09-21  Eric Botcazou  <ebotcazou@adacore.com>
15004         * stor-layout.c (bit_from_pos): Do not distribute the conversion.
15006 2017-09-21  Segher Boessenkool  <segher@kernel.crashing.org>
15008         * haifa-sched.c: Rename insn_cost to insn_sched_cost.
15009         * sched-rgn.c: Ditto.
15010         * sel-sched-ir.c: Ditto.
15012 2017-09-21  Alexander Monakov  <amonakov@ispras.ru>
15014         * toplev.h (set_random_seed): Adjust return type.
15015         * toplev.c (init_local_tick): Move eager initialization of random_seed
15016         to get_random_seed.  Adjust comment.
15017         (init_random_seed): Inline to get_random_seed, delete.
15018         (get_random_seed): Initialize random_seed lazily.
15019         (set_random_seed): Do not return previous value.
15020         (print_switch_value): Do not call get_random_seed.
15022 2017-09-21  Evgeny Kudryashov  <kudryashov@ispras.ru>
15024         * cgraph.c (delete_function_version): New, broken out from...
15025         (cgraph_node::delete_function_version): ...here.  Rename to
15026         cgraph_node::delete_function_version_by_decl.  Update all uses.
15027         (cgraph_node::remove): Call delete_function_version.
15029 2017-09-21  Jakub Jelinek  <jakub@redhat.com>
15031         PR sanitizer/81715
15032         * tree-inline.c (expand_call_inline): Emit clobber stmts for
15033         VAR_DECLs to which addressable non-volatile parameters are mapped
15034         and for id->retvar after the return value assignment.  Clear
15035         id->retval and id->retbnd after inlining.
15037 2017-09-21  Richard Biener  <rguenther@suse.de>
15039         PR tree-optimization/82276
15040         PR tree-optimization/82244
15041         * tree-vrp.c (build_assert_expr_for): Set
15042         SSA_NAME_OCCURS_IN_ABNORMAL_PHI if the variable we assert on
15043         has it set.
15044         (remove_range_assertions): Revert earlier change.
15046 2017-09-21  Wilco Dijkstra  <wdijkstr@arm.com>
15048         PR target/71951
15049         * config/aarch64/aarch64.h (LIBGCC2_UNWIND_ATTRIBUTE): Define.
15051 2017-09-21  Richard Biener  <rguenther@suse.de>
15053         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
15054         Restore valid IL after code generation errors.
15055         * graphite.c (graphite_transform_loops): Diagnose code
15056         generation issues as MSG_MISSED_OPTIMIZATION and continue
15057         with processing SCOPs.
15059 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
15060             Alan Hayward  <alan.hayward@arm.com>
15061             David Sherwood  <david.sherwood@arm.com>
15063         * calls.c (compute_argument_addresses): Use simplify_gen_binary
15064         rather than choosing between plus_constant and gen_rtx_<CODE>.
15065         * expr.c (emit_push_insn): Likewise.
15066         (expand_expr_real_2): Likewise.
15068 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
15069             Alan Hayward  <alan.hayward@arm.com>
15070             David Sherwood  <david.sherwood@arm.com>
15072         * loop-unroll.c (split_iv): Call copy_rtx on the step.
15074 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
15075             Alan Hayward  <alan.hayward@arm.com>
15076             David Sherwood  <david.sherwood@arm.com>
15078         * tree.c (find_atomic_core_type): Check tree_fits_uhwi_p before
15079         calling tree_to_uhwi.
15081 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
15082             Alan Hayward  <alan.hayward@arm.com>
15083             David Sherwood  <david.sherwood@arm.com>
15085         * tree-ssa-ccp.c (get_value_for_expr): Use a positive test for
15086         INTEGER_CST rather than a negative test for ADDR_EXPR.
15088 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
15089             Alan Hayward  <alan.hayward@arm.com>
15090             David Sherwood  <david.sherwood@arm.com>
15092         * tree-vrp.c (extract_range_from_binary_expr_1): Check
15093         int_cst_rangeN before calling value_range_constant_singleton (&vrN).
15095 2017-09-21  Richard Biener  <rguenther@suse.de>
15097         PR tree-optimization/71351
15098         * graphite-isl-ast-to-gimple.c (translate_isl_ast_to_gimple::
15099         graphite_create_new_loop_guard): Remove, fold remaining parts
15100         into caller ...
15101         (translate_isl_ast_node_for): ... here and simplify.
15103 2017-09-21  Jakub Jelinek  <jakub@redhat.com>
15105         PR target/82260
15106         * config/i386/i386.md (*movqi_internal): Replace (=q,q) alternative
15107         with (=Q,Q), (=R,R) and (=r,r) alternatives, only enable the
15108         latter two for 64-bit, renumber alternatives, for -Os imov (=q,n)
15109         alternative always use QI mode, for -Os imov (=R,R) alternative
15110         always use SI mode, for imov (=Q,Q) or (=r,r) alternatives
15111         ignore -Os.
15113 2017-09-20  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
15114             Jeff Law  <law@redhat.com>
15116         * config/s390/s390.c (MIN_UNROLL_PROBES): Define.
15117         (allocate_stack_space): New function, partially extracted from
15118         s390_emit_prologue.
15119         (s390_emit_prologue): Track offset to most recent stack probe.
15120         Code to allocate space moved into allocate_stack_space.
15121         Dump actions when no stack is allocated.
15122         (s390_prologue_plus_offset): New function.
15123         (s390_emit_stack_probe): Likewise.
15125 2017-09-20  Alexandre Oliva  <aoliva@redhat.com>
15127         * common.opt (Wa, Wl, Wp, g, gz=): Add
15128         RejectNegative.
15129         (gno-column-info): Remove.
15130         (gcolumn-info): Drop RejectNegative.
15131         (gno-): New prefix.
15132         (gno-record-gcc-switches): Remove.
15133         (grecord-gcc-switches): Drop RejectNegative.
15134         (gno-split-dwarf): Remove.
15135         (gsplit-dwarf): Drop RejectNegative.
15136         (gno-strict-dwarf): Remove.
15137         (gstrict-dwarf): Drop RejectNegative.
15138         * config/darwin.opt (gfull, gused): Add RejectNegative.
15139         * dwarf2out.c (gen_producer_string): Drop
15140         gno-record-gcc-switches handler.
15141         * optc-gen.awk: Add g to prefixes with negative forms.
15142         * opts-common.c (remapping_prefix_p): New.
15143         (find_opt): Check it.
15144         (generate_canonical_option): Test g prefix.
15145         (option_map): Add -gno- mapping.
15146         (add_misspelling_candidates): Check remapping_prefix_p.
15148 2017-09-20  Jeff Law  <law@redhat.com>
15150         * config/powerpcspe/powerpcspe.c (rs6000_expand_prologue): Fix
15151         thinko in stack clash protection support.
15153         * explow.c (compute_stack_clash_protection_loop_data): Use
15154         CONST_INT_P instead of explicit test.  Verify object is a
15155         CONST_INT_P before looking at INTVAL.
15156         (anti_adjust_stack_and_probe_stack_clash): Use CONST_INT_P
15157         instead of explicit test.
15159 2017-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
15161         PR target/77687
15162         * config/rs6000/rs6000.md (stack_restore_tie): Store to a scratch
15163         address instead of to r1 and r11.
15165 2017-09-20  Sebastian Peryt  <sebastian.peryt@intel.com>
15167         * config.gcc: Support "knm".
15168         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "knm".
15169         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15170         PROCESSOR_KNM.
15171         * config/i386/i386.c (m_KNM): Define.
15172         (processor_target_table): Add "knm".
15173         (PTA_KNM): Define.
15174         (ix86_option_override_internal): Add "knm".
15175         (ix86_issue_rate): Add PROCESSOR_KNM.
15176         (ix86_adjust_cost): Ditto.
15177         (ia32_multipass_dfa_lookahead): Ditto.
15178         (get_builtin_code_for_version): Handle PROCESSOR_KNM.
15179         (fold_builtin_cpu): Add M_INTEL_KNM.
15180         * config/i386/i386.h (processor_costs): Define TARGET_KNM.
15181         (processor_type): Add PROCESSOR_KNM.
15182         * config/i386/x86-tune.def: Add m_KNM.
15183         * doc/invoke.texi: Add knm as x86 -march=/-mtune= CPU type.
15185 2017-09-20  Richard Biener  <rguenther@suse.de>
15187         PR tree-optimization/80213
15188         * graphite-scop-detection.c (trivially_empty_bb_p): Labels
15189         are allowed in empty BBs as well.
15190         (canonicalize_loop_closed_ssa): Also look for other complex
15191         edges.
15192         (scop_detection::get_sese): Include the loop-closed PHI block
15193         in loop SESEs.
15194         (scop_detection::merge_sese): Remove code adding extra blocks.
15195         (scop_detection::region_has_one_loop): Adjust for get_sese changes.
15196         (build_scops): Assert the final returned scop is invalid.
15198 2017-09-20  Richard Biener  <rguenther@suse.de>
15200         PR tree-optimization/82264
15201         * tree-ssa-sccvn.c (vn_phi_eq): Use safe_dyn_cast to check
15202         for GIMPLE_CONDs.
15203         (vn_phi_lookup): Likewise.
15204         (vn_phi_insert): Likewise.
15206 2017-09-20  Jakub Jelinek  <jakub@redhat.com>
15208         * dwarf2out.c (tree_add_const_value_attribute): For INTEGER_CST
15209         that fits into uhwi or shwi, add DW_AT_const_value regardless
15210         of early_dwarf without going through RTL, using add_AT_unsigned
15211         or add_AT_int.
15213         * dwarf2out.c (DEBUG_LTO_DWO_INFO_SECTION): Reorder defines.
15214         (DEBUG_LTO_ABBREV_SECTION): Likewise.
15215         (DEBUG_LTO_MACINFO_SECTION): Likewise.
15216         (DEBUG_MACRO_SECTION): Likewise.
15217         (DEBUG_LTO_MACRO_SECTION): Likewise.
15218         (DEBUG_STR_DWO_SECTION): Likewise.
15219         (DEBUG_LTO_STR_DWO_SECTION): Likewise.
15220         (DEBUG_LTO_LINE_SECTION): Drop .dwo suffix from the name.
15221         (DEBUG_LTO_DWO_LINE_SECTION): Define.
15222         (DEBUG_LTO_LINE_STR_SECTION): Define.
15223         (init_sections_and_labels): Initialize debug_line_str_section
15224         variable.  Initialize debug_loc_section for -gdwarf-5 to
15225         DEBUG_LOCLISTS_SECTION.  Formatting fixes.
15227 2017-09-20  Richard Biener  <rguenther@suse.de>
15229         * graphite-sese-to-poly.c (extract_affine): Properly handle
15230         POINTER_PLUS_EXPR, BIT_NOT_EXPR and conversion to signed.
15232 2017-09-20  Richard Biener  <rguenther@suse.de>
15234         PR tree-optimization/81373
15235         * graphite-scop-detection.c (build_cross_bb_scalars_def):
15236         Force SESE live-out defs to be handled even if they are
15237         scev_analyzable_p.
15239 2017-09-19  Jeff Law  <law@redhat.com>
15241         * combine-stack-adj.c (combine_stack_adjustments_for_block): Do
15242         nothing for stack adjustments with REG_STACK_CHECK.
15243         * sched-deps.c (parse_add_or_inc): Reject insns with
15244         REG_STACK_CHECK from dependency breaking.
15245         * config/i386/i386.c (pro_epilogue_adjust_stack): Return insn.
15246         (ix86_adjust_satck_and_probe_stack_clash): Add REG_STACK_NOTEs.
15247         * reg-notes.def (STACK_CHECK): New note.
15249         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): New.
15250         (ix86_expand_prologue): Dump stack clash info as needed.
15251         Call ix86_adjust_stack_and_probe_stack_clash as needed.
15253         * function.c (dump_stack_clash_frame_info): New function.
15254         * function.h (dump_stack_clash_frame_info): Prototype.
15255         (enum stack_clash_probes): New enum.
15257         * config/alpha/alpha.c (alpha_expand_prologue): Also check
15258         flag_stack_clash_protection.
15259         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Likewise.
15260         (arm_expand_prologue, thumb1_expand_prologue): Likewise.
15261         (arm_frame_pointer_required): Likewise.
15262         * config/ia64/ia64.c (ia64_compute_frame_size): Likewise.
15263         (ia64_expand_prologue): Likewise.
15264         * config/mips/mips.c (mips_expand_prologue): Likewise.
15265         * config/powerpcspe/powerpcspe.c (rs6000_expand_prologue): Likewise.
15266         * config/sparc/sparc.c (sparc_expand_prologue): Likewise.
15267         (sparc_flat_expand_prologue): Likewise.
15268         * config/spu/spu.c (spu_expand_prologue): Likewise.
15270         * explow.c: Include "params.h".
15271         (anti_adjust_stack_and_probe_stack_clash): New function.
15272         (get_stack_check_protect): Likewise.
15273         (compute_stack_clash_protection_loop_data): Likewise.
15274         (emit_stack_clash_protection_loop_start): Likewise.
15275         (emit_stack_clash_protection_loop_end): Likewise.
15276         (allocate_dynamic_stack_space): Use get_stack_check_protect.
15277         Use anti_adjust_stack_and_probe_stack_clash.
15278         * explow.h (compute_stack_clash_protection_loop_data): Prototype.
15279         (emit_stack_clash_protection_loop_start): Likewise.
15280         (emit_stack_clash_protection_loop_end): Likewise.
15281         * rtl.h (get_stack_check_protect): Prototype.
15282         * target.def (stack_clash_protection_final_dynamic_probe): New hook.
15283         * targhooks.c (default_stack_clash_protection_final_dynamic_probe): New.
15284         * targhooks.h (default_stack_clash_protection_final_dynamic_probe):
15285         Prototype.
15286         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE):
15287         Add @hook.
15288         * doc/tm.texi: Rebuilt.
15289         * config/aarch64/aarch64.c (aarch64_expand_prologue): Use
15290         get_stack_check_protect.
15291         * config/alpha/alpha.c (alpha_expand_prologue): Likewise.
15292         * config/arm/arm.c (arm_expand_prologue): Likewise.
15293         (arm_frame_pointer_required): Likewise.
15294         * config/i386/i386.c (ix86_expand_prologue): Likewise.
15295         * config/ia64/ia64.c (ia64_expand_prologue): Likewise.
15296         * config/mips/mips.c (mips_expand_prologue): Likewise.
15297         * config/powerpcspe/powerpcspe.c (rs6000_emit_prologue): Likewise.
15298         * config/rs6000/rs6000.c (rs6000_emit_prologue): Likewise.
15299         * config/sparc/sparc.c (sparc_expand_prologue): Likewise.
15300         (sparc_flat_expand_prologue): Likewise.
15302         * common.opt (-fstack-clash-protection): New option.
15303         * flag-types.h (enum stack_check_type): Note difference between
15304         -fstack-check= and -fstack-clash-protection.
15305         * params.def (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE): New PARAM.
15306         (PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Likewise.
15307         * toplev.c (process_options): Issue warnings/errors for cases
15308         not handled with -fstack-clash-protection.
15309         * doc/invoke.texi (-fstack-clash-protection): Document new option.
15310         (-fstack-check): Note additional problem with -fstack-check=generic.
15311         Note that -fstack-check is primarily for Ada and refer users
15312         to -fstack-clash-protection for stack-clash-protection.
15313         Document new params for stack clash protection.
15315 2017-09-19  Uros Bizjak  <ubizjak@gmail.com>
15317         * config/i386/i386.c (ix86_split_long_move): Do not handle
15318         address used for LEA in a special way.
15320 2017-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
15322         * simplify-rtx.c (simplify_binary_operation_1): Fix typo in comment.
15324 2017-09-19  Martin Sebor  <msebor@redhat.com>
15326         PR c/81854
15327         * cgraphunit.c (handle_alias_pairs): Reject aliases between functions
15328         of incompatible types.
15330 2017-09-19  Will Schmidt  <will_schmidt@vnet.ibm.com>
15332         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
15333         for early folding of vector loads (ALTIVEC_BUILTIN_LVX_*).
15334         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
15335         Remove obsoleted code for handling ALTIVEC_BUILTIN_VEC_LD.
15337 2017-09-19  Richard Biener  <rguenther@suse.de>
15339         PR tree-optimization/82244
15340         * tree-vrp.c (remove_range_assertions): Do not propagate
15341         a constant to abnormals but replace the assert with a copy.
15343 2017-09-19  Alexander Monakov  <amonakov@ispras.ru>
15345         PR rtl-optimization/57878
15346         PR rtl-optimization/68988
15347         * lra-assigns.c (reload_pseudo_compare_func): Remove fragmentation
15348         avoidance test involving non_reload_pseudos.  Move frequency test
15349         below the general fragmentation avoidance test.
15351 2017-09-19  Richard Biener  <rguenther@suse.de>
15353         PR tree-optimization/69728
15354         * graphite-sese-to-poly.c (schedule_error): New global.
15355         (add_loop_schedule): Handle empty domain by failing the
15356         schedule.
15357         (build_original_schedule): Handle schedule_error.
15359 2017-09-19  Richard Biener  <rguenther@suse.de>
15361         * graphite-scop-detection.c (scop_detection::can_represent_loop):
15362         Do not iterate to sibling loops but only to siblings of inner
15363         loops.
15365 2017-09-18  Andreas Schwab  <schwab@linux-m68k.org>
15367         PR target/81613
15368         * config/m68k/m68k.md (moveq feeding equality comparison): Check
15369         that the registers are different.
15371 2017-09-18  Uros Bizjak  <ubizjak@gmail.com>
15373         * config/i386/i386.c (fold_builtin_cpu): Add M_AMDFAM17H
15374         to processor_model and "amdfam17h" to arch_names_table.
15375         * doc/extend.texi (__builtin_cpu_is): Document amdfam17h CPU name.
15377 2017-09-18  Jakub Jelinek  <jakub@redhat.com>
15379         PR c/82234
15380         * doc/extend.texi: Add @findex entry for __builtin_shuffle.
15382 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15383             Alan Hayward  <alan.hayward@arm.com>
15384             David Sherwood  <david.sherwood@arm.com>
15386         * tree-vectorizer.h (vect_slp_analyze_operations): Replace parameters
15387         with a vec_info *.
15388         * tree-vect-loop.c (vect_analyze_loop_operations): Update call
15389         accordingly.
15390         * tree-vect-slp.c (vect_slp_analyze_node_operations): Add vec_info *
15391         parameter.  Set SLP_TREE_NUMBER_OF_VEC_STMTS here rather than in
15392         vect_schedule_slp_instance.
15393         (vect_slp_analyze_operations): Replace parameters with a vec_info *.
15394         Update call to vect_slp_analyze_node_operations.  Simplify return
15395         value.
15396         (vect_slp_analyze_bb_1): Update call accordingly.
15397         (vect_schedule_slp_instance): Remove vectorization_factor parameter.
15398         Don't calculate SLP_TREE_NUMBER_OF_VEC_STMTS here.
15399         (vect_schedule_slp): Update call accordingly.
15401 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15402             Alan Hayward  <alan.hayward@arm.com>
15403             David Sherwood  <david.sherwood@arm.com>
15405         * tree-vect-data-refs.c (vect_get_smallest_scalar_type): Cope
15406         with types that aren't in fact scalar.
15408 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15410         * tree-vect-slp.c (vect_record_max_nunits): New function,
15411         split out from...
15412         (vect_build_slp_tree_1): ...here.
15413         (vect_build_slp_tree_2): Call it for phis too.
15415 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15417         * tree-vect-stmts.c (vectorizable_mask_load_store): Pass mask_vectype
15418         to vect_get_vec_def_for_operand when getting the mask operand.
15420 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15421             Alan Hayward  <alan.hayward@arm.com>
15422             David Sherwood  <david.sherwood@arm.com>
15424         * tree-vect-loop.c (vectorizable_live_operation): Fix type of
15425         bitstart.
15427 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15428             Alan Hayward  <alan.hayward@arm.com>
15429             David Sherwood  <david.sherwood@arm.com>
15431         * tree-vect-loop.c (vectorizable_live_operation): Fix element size
15432         calculation for vector booleans.
15434 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15435             Alan Hayward  <alan.hayward@arm.com>
15436             David Sherwood  <david.sherwood@arm.com>
15438         * tree-vect-stmts.c (can_vectorize_live_stmts): New function,
15439         split out from...
15440         (vect_transform_stmt): ...here.
15441         (vect_analyze_stmt): Use it instead of calling
15442         vectorizable_live_operation directly.
15444 2017-09-18  Cesar Philippidis  <cesar@codesourcery.com>
15446         * omp-offload.c (oacc_xform_loop): Enable SIMD vectorization on
15447         non-SIMT targets in acc vector loops.
15449 2017-09-18  Claudiu Zissulescu  <claziss@synopsys.com>
15451         * configure.ac: Add arc and check if assembler supports gdwarf2.
15452         * configure: Regenerate.
15454 2017-09-18  Richard Biener  <rguenther@suse.de>
15456         PR tree-optimization/82220
15457         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Exclude
15458         epilogue niters from the min_profitable_iters compute.
15460 2017-09-18  Jakub Jelinek  <jakub@redhat.com>
15462         PR target/82145
15463         * config/i386/i386.c (ix86_init_large_pic_reg): Revert 2017-09-01
15464         changes.  Turn CODE_LABEL into NOTE_INSN_DELETED_LABEL immediately.
15465         (ix86_init_pic_reg): Revert 2017-09-01 changes.
15467 2017-09-18  Eric Botcazou  <ebotcazou@adacore.com>
15469         PR target/81361
15470         * dwarf2cfi.c (add_cfis_to_fde): Do not generate DW_CFA_set_loc after
15471         switching to a new text section.
15473 2017-09-18  Richard Biener  <rguenther@suse.de>
15475         * graphite-scop-detection.c (scop_detection::stmt_has_simple_data_ref):
15476         Simplify.
15477         (build_alias_set): Reject aliases with no access function.
15479 2017-09-18  Richard Biener  <rguenther@suse.de>
15481         PR tree-optimization/79622
15482         * graphite-scop-detection.c (build_cross_bb_scalars_def): Properly
15483         handle PHIs.
15484         (build_cross_bb_scalars_use): Likewise.
15486 2017-09-18  Pierre-Marie de Rodat  <derodat@adacore.com>
15488         * cgraph.h (cgraph_thunk_info): Fix a typo in a comment.
15490 2017-09-18  Alan Modra  <amodra@gmail.com>
15492         PR target/81996
15493         * gcc/config/rs6000/rs6000.c (rs6000_return_addr): Use
15494         stack_pointer_rtx for count 0.  Update comments.  Break up
15495         large rtl expression.
15497 2017-09-17  Daniel Santos  <daniel.santos@pobox.com>
15499         * config/i386/i386.c (xlogue_layout::STUB_NAME_MAX_LEN):
15500         Increase to 20 bytes.
15501         (xlogue_layout::s_stub_names): Add an additional size-2 diminsion.
15502         (xlogue_layout::get_stub_name): Modify to select the appropairate sse
15503         or avx version of the stub.
15505 2017-09-17  H.J. Lu  <hongjiu.lu@intel.com>
15507         PR target/82166
15508         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Properly
15509         compute the minimum stack alignment.  Also update preferred stack
15510         boundary for leaf functions.
15512 2017-09-16  Richard Sandiford  <richard.sandiford@linaro.org>
15514         PR tree-optimization/82228
15515         * tree-vect-loop.c (vectorizable_live_operation): Move initialization
15516         of ncopies.
15518 2017-09-16  Chung-Ju Wu  <jasonwucj@gmail.com>
15520         * common/config/nds32/nds32-common.c
15521         (nds32_option_optimization_table): Refine formatting.
15522         (nds32_option_optimization_table): Use -fsched-pressure and
15523         -fomit-frame-pointer for specific optimization level.
15525 2017-09-16  Chung-Ju Wu  <jasonwucj@gmail.com>
15527         * config/nds32/nds32.c: Refine formatting and comments.
15528         * config/nds32/nds32.h: Likewise.
15529         * config/nds32/nds32.md: Likewise.
15530         * config/nds32/nds32-cost.c: Likewise.
15531         * config/nds32/nds32-isr.c: Likewise.
15532         * config/nds32/nds32-md-auxiliary.c: Likewise.
15533         * config/nds32/nds32-multiple.md: Likewise.
15534         * config/nds32/nds32-predicates.c: Likewise.
15536 2017-09-15  Andrew Sutton  <andrew.n.sutton@gmail.com>
15537             Jakub Jelinek  <jakub@redhat.com>
15539         Add support for -std=c++2a.
15540         * doc/cpp.texi (__cplusplus): Document value for -std=c++2a
15541         or -std=gnu+2a.
15542         * doc/invoke.texi: Document -std=c++2a and -std=gnu++2a.
15544 2017-09-15  Steve Ellcey  <sellcey@cavium.com>
15546         PR target/82066
15547         * doc/extend.texi (Common Function Attributes): Add 
15548         references to ARM, AArch64, and S/390 specific attributes.
15549         (Function Specific Option Pragmas): Add AArch64 and S/390
15550         to list of back ends that support the target pragma.
15552 2017-09-15  Nathan Sidwell  <nathan@acm.org>
15554         * doc/standards.texi: Fix C++17 description.  Update URLs for
15555         C++11 & 14.
15557 2017-09-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
15559         * common.opt (Wcast-align=strict): New warning option.
15560         * doc/invoke.texi: Document -Wcast-align=strict. 
15562 2017-09-15  Pierre-Marie de Rodat  <derodat@adacore.com>
15564         * cgraph.h (cgraph_thunk_info): Add comments.
15565         * cgraph.c (cgraph_node::create_thunk): Adjust comment, make
15566         assert for VIRTUAL_* arguments stricter.
15568 2017-09-15  Jackson Woodruff  <jackson.woodruff@arm.com>
15570         PR tree-optimization/71026
15571         * match.pd: Move RDIV patterns from fold-const.c
15572         * fold-const.c (distribute_real_division): Removed.
15573         (fold_binary_loc): Remove calls to distribute_real_divison.
15575 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15577         * doc/invoke.texi: Document -std=c++17 and -std=gnu++17 and document
15578         c++1z and gnu++1z as deprecated.  Change other references to
15579         -std=c++1z to -std=c++17 and -std=gnu++1z to -std=gnu++17.
15580         Change -Wc++1z-compat to -Wc++17-compat.
15581         * doc/cpp.texi: Document -std=c++17 defines __cplusplus 201703L.
15582         * dwarf2out.c (highest_c_language): Handle C++17.
15583         (gen_compile_unit_die): Likewise.
15585 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15587         PR rtl-optimization/82192
15588         * combine.c (make_extraction): Don't look through non-paradoxical
15589         SUBREGs or TRUNCATE if pos + len is or might be bigger than
15590         inner's mode.
15592 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15593             Alan Hayward  <alan.hayward@arm.com>
15594             David Sherwood  <david.sherwood@arm.com>
15596         * target.def (function_arg_offset): New hook.
15597         * targhooks.h (default_function_arg_offset): Declare.
15598         * targhooks.c (default_function_arg_offset): New function.
15599         * function.c (locate_and_pad_parm): Use
15600         targetm.calls.function_arg_offset instead of FUNCTION_ARG_OFFSET.
15601         * doc/tm.texi.in (FUNCTION_ARG_OFFSET): Replace with...
15602         (TARGET_FUNCTION_ARG_OFFSET): ...this.
15603         * doc/tm.texi: Regenerate.
15604         * config/spu/spu.h (FUNCTION_ARG_OFFSET): Delete.
15605         * config/spu/spu.c (spu_function_arg_offset): New function.
15606         (TARGET_FUNCTION_ARG_OFFSET): Redefine.
15607         * system.h (FUNCTION_ARG_OFFSET): Poison.
15609 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15610             Alan Hayard  <alan.hayward@arm.com>
15611             David Sherwood  <david.sherwood@arm.com>
15613         * target.def (truly_noop_truncation): New hook.
15614         (mode_rep_extended): Refer to TARGET_TRULY_NOOP_TRUNCATION rather
15615         than TRULY_NOOP_TRUNCATION.
15616         * hooks.h (hook_bool_uint_uint_true): Declare.
15617         * hooks.c (hook_bool_uint_uint_true): New function.
15618         * doc/tm.texi.in (TRULY_NOOP_TRUNCATION): Replace with...
15619         (TARGET_TRULY_NOOP_TRUNCATION): ...this.
15620         * doc/tm.texi: Regenerate.
15621         * combine.c (make_extraction): Refer to TARGET_TRULY_NOOP_TRUNCATION
15622         rather than TRULY_NOOP_TRUNCATION in comments.
15623         (simplify_comparison): Likewise.
15624         (record_truncated_value): Likewise.
15625         * expmed.c (extract_bit_field_1): Likewise.
15626         (extract_split_bit_field): Likewise.
15627         * convert.c (convert_to_integer_1): Use targetm.truly_noop_truncation
15628         instead of TRULY_NOOP_TRUNCATION.
15629         * function.c (assign_parm_setup_block): Likewise.
15630         * machmode.h (TRULY_NOOP_TRUNCATION_MODES_P): Likewise.
15631         * rtlhooks.c: Include target.h.
15632         * config/aarch64/aarch64.h (TRULY_NOOP_TRUNCATION): Delete.
15633         * config/alpha/alpha.h (TRULY_NOOP_TRUNCATION): Delete.
15634         * config/arc/arc.h (TRULY_NOOP_TRUNCATION): Delete.
15635         * config/arm/arm.h (TRULY_NOOP_TRUNCATION): Delete.
15636         * config/avr/avr.h (TRULY_NOOP_TRUNCATION): Delete.
15637         * config/bfin/bfin.h (TRULY_NOOP_TRUNCATION): Delete.
15638         * config/c6x/c6x.h (TRULY_NOOP_TRUNCATION): Delete.
15639         * config/cr16/cr16.h (TRULY_NOOP_TRUNCATION): Delete.
15640         * config/cris/cris.h (TRULY_NOOP_TRUNCATION): Delete.
15641         * config/epiphany/epiphany.h (TRULY_NOOP_TRUNCATION): Delete.
15642         * config/fr30/fr30.h (TRULY_NOOP_TRUNCATION): Delete.
15643         * config/frv/frv.h (TRULY_NOOP_TRUNCATION): Delete.
15644         * config/ft32/ft32.h (TRULY_NOOP_TRUNCATION): Delete.
15645         * config/h8300/h8300.h (TRULY_NOOP_TRUNCATION): Delete.
15646         * config/i386/i386.h (TRULY_NOOP_TRUNCATION): Delete.
15647         * config/ia64/ia64.h (TRULY_NOOP_TRUNCATION): Delete.
15648         * config/iq2000/iq2000.h (TRULY_NOOP_TRUNCATION): Delete.
15649         * config/lm32/lm32.h (TRULY_NOOP_TRUNCATION): Delete.
15650         * config/m32c/m32c.h (TRULY_NOOP_TRUNCATION): Delete.
15651         * config/m32r/m32r.h (TRULY_NOOP_TRUNCATION): Delete.
15652         * config/m68k/m68k.h (TRULY_NOOP_TRUNCATION): Delete.
15653         * config/mcore/mcore.h (TRULY_NOOP_TRUNCATION): Delete.
15654         * config/microblaze/microblaze.h (TRULY_NOOP_TRUNCATION): Delete.
15655         * config/mips/mips.h (TRULY_NOOP_TRUNCATION): Delete.
15656         * config/mips/mips.c (mips_truly_noop_truncation): New function.
15657         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15658         * config/mips/mips.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15659         rather than TRULY_NOOP_TRUNCATION in comments.
15660         * config/mmix/mmix.h (TRULY_NOOP_TRUNCATION): Delete.
15661         * config/mn10300/mn10300.h (TRULY_NOOP_TRUNCATION): Delete.
15662         * config/moxie/moxie.h (TRULY_NOOP_TRUNCATION): Delete.
15663         * config/msp430/msp430.h (TRULY_NOOP_TRUNCATION): Delete.
15664         * config/nds32/nds32.h (TRULY_NOOP_TRUNCATION): Delete.
15665         * config/nios2/nios2.h (TRULY_NOOP_TRUNCATION): Delete.
15666         * config/nvptx/nvptx.h (TRULY_NOOP_TRUNCATION): Delete.
15667         * config/pa/pa.h (TRULY_NOOP_TRUNCATION): Delete.
15668         * config/pdp11/pdp11.h (TRULY_NOOP_TRUNCATION): Delete.
15669         * config/powerpcspe/powerpcspe.h (TRULY_NOOP_TRUNCATION): Delete.
15670         * config/riscv/riscv.h (TRULY_NOOP_TRUNCATION): Delete.
15671         * config/riscv/riscv.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15672         rather than TRULY_NOOP_TRUNCATION in comments.
15673         * config/rl78/rl78.h (TRULY_NOOP_TRUNCATION): Delete.
15674         * config/rs6000/rs6000.h (TRULY_NOOP_TRUNCATION): Delete.
15675         * config/rx/rx.h (TRULY_NOOP_TRUNCATION): Delete.
15676         * config/s390/s390.h (TRULY_NOOP_TRUNCATION): Delete.
15677         * config/sh/sh.h (MAYBE_BASE_REGISTER_RTX_P): Remove
15678         TRULY_NOOP_TRUNCATION condition.
15679         (MAYBE_INDEX_REGISTER_RTX_P): Likewise.
15680         (TRULY_NOOP_TRUNCATION): Delete.
15681         * config/sparc/sparc.h (TRULY_NOOP_TRUNCATION): Delete.
15682         * config/spu/spu.h (TRULY_NOOP_TRUNCATION): Delete.
15683         * config/spu/spu.c (spu_truly_noop_truncation): New function.
15684         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15685         * config/stormy16/stormy16.h (TRULY_NOOP_TRUNCATION): Delete.
15686         * config/tilegx/tilegx.h (TRULY_NOOP_TRUNCATION): Delete.
15687         * config/tilegx/tilegx.c (tilegx_truly_noop_truncation): New fuction.
15688         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15689         * config/tilegx/tilegx.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15690         rather than TRULY_NOOP_TRUNCATION in comments.
15691         * config/tilepro/tilepro.h (TRULY_NOOP_TRUNCATION): Delete.
15692         * config/v850/v850.h (TRULY_NOOP_TRUNCATION): Delete.
15693         * config/vax/vax.h (TRULY_NOOP_TRUNCATION): Delete.
15694         * config/visium/visium.h (TRULY_NOOP_TRUNCATION): Delete.
15695         * config/xtensa/xtensa.h (TRULY_NOOP_TRUNCATION): Delete.
15696         * system.h (TRULY_NOOP_TRUNCATION): Poison.
15698 2017-09-15  Christophe Lyon  <christophe.lyon@linaro.org>
15700         PR target/67591
15701         * config/arm/arm.md (*cmp_and): Add enabled_for_depr_it attribute.
15702         (*cmp_ior): Likewise.
15703         (*ior_scc_scc): Add alternative for enabled_for_depr_it attribute.
15704         (*ior_scc_scc_cmp): Likewise.
15705         (*and_scc_scc): Likewise.
15706         (*and_scc_scc_cmp): Likewise.
15708 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15709             Alan Hayard  <alan.hayward@arm.com>
15710             David Sherwood  <david.sherwood@arm.com>
15712         * target.def (can_change_mode_class): New hook.
15713         (mode_rep_extended): Refer to it instead of CANNOT_CHANGE_MODE_CLASS.
15714         (hard_regno_nregs): Likewise.
15715         * hooks.h (hook_bool_mode_mode_reg_class_t_true): Declare.
15716         * hooks.c (hook_bool_mode_mode_reg_class_t_true): New function.
15717         * doc/tm.texi.in (CANNOT_CHANGE_MODE_CLASS): Replace with...
15718         (TARGET_CAN_CHANGE_MODE_CLASS): ...this.
15719         (LOAD_EXTEND_OP): Update accordingly.
15720         * doc/tm.texi: Regenerate.
15721         * doc/rtl.texi: Refer to TARGET_CAN_CHANGE_MODE_CLASS instead of
15722         CANNOT_CHANGE_MODE_CLASS.
15723         * hard-reg-set.h (REG_CANNOT_CHANGE_MODE_P): Replace with...
15724         (REG_CAN_CHANGE_MODE_P): ...this new macro.
15725         * combine.c (simplify_set): Update accordingly.
15726         * emit-rtl.c (validate_subreg): Likewise.
15727         * recog.c (general_operand): Likewise.
15728         * regcprop.c (mode_change_ok): Likewise.
15729         * reload1.c (choose_reload_regs): Likewise.
15730         (inherit_piecemeal_p): Likewise.
15731         * rtlanal.c (simplify_subreg_regno): Likewise.
15732         * postreload.c (reload_cse_simplify_set): Use REG_CAN_CHANGE_MODE_P
15733         instead of CANNOT_CHANGE_MODE_CLASS.
15734         (reload_cse_simplify_operands): Likewise.
15735         * reload.c (push_reload): Use targetm.can_change_mode_class
15736         instead of CANNOT_CHANGE_MODE_CLASS.
15737         (push_reload): Likewise.  Also use REG_CAN_CHANGE_MODE_P instead of
15738         REG_CANNOT_CHANGE_MODE_P.
15739         * config/alpha/alpha.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15740         * config/alpha/alpha.c (alpha_can_change_mode_class): New function.
15741         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15742         * config/arm/arm.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15743         * config/arm/arm.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15744         (arm_can_change_mode_class): New function.
15745         * config/arm/neon.md: Refer to TARGET_CAN_CHANGE_MODE_CLASS rather
15746         than CANNOT_CHANGE_MODE_CLASS in comments.
15747         * config/i386/i386.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15748         * config/i386/i386-protos.h (ix86_cannot_change_mode_class): Delete.
15749         * config/i386/i386.c (ix86_cannot_change_mode_class): Replace with...
15750         (ix86_can_change_mode_class): ...this new function, inverting the
15751         sense of the return value.
15752         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15753         * config/ia64/ia64.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15754         * config/ia64/ia64.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15755         (ia64_can_change_mode_class): New function.
15756         * config/m32c/m32c.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15757         * config/m32c/m32c-protos.h (m32c_cannot_change_mode_class): Delete.
15758         * config/m32c/m32c.c (m32c_cannot_change_mode_class): Replace with...
15759         (m32c_can_change_mode_class): ...this new function, inverting the
15760         sense of the return value.
15761         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15762         * config/mips/mips.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15763         * config/mips/mips-protos.h (mips_cannot_change_mode_class): Delete.
15764         * config/mips/mips.c (mips_cannot_change_mode_class): Replace with...
15765         (mips_can_change_mode_class): ...this new function, inverting the
15766         sense of the return value.
15767         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15768         * config/msp430/msp430.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15769         * config/msp430/msp430.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15770         (msp430_can_change_mode_class): New function.
15771         * config/nvptx/nvptx.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15772         * config/nvptx/nvptx.c (nvptx_can_change_mode_class): New function.
15773         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15774         * config/pa/pa32-regs.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15775         * config/pa/pa64-regs.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15776         * config/pa/pa-protos.h (pa_cannot_change_mode_class): Delete.
15777         * config/pa/pa.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15778         (pa_cannot_change_mode_class): Replace with...
15779         (pa_can_change_mode_class): ...this new function, inverting the
15780         sense of the return value.
15781         (pa_modes_tieable_p): Refer to TARGET_CAN_CHANGE_MODE_CLASS rather
15782         than CANNOT_CHANGE_MODE_CLASS in comments.
15783         * config/pdp11/pdp11.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15784         * config/pdp11/pdp11-protos.h (pdp11_cannot_change_mode_class): Delete.
15785         * config/pdp11/pdp11.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15786         (pdp11_cannot_change_mode_class): Replace with...
15787         (pdp11_can_change_mode_class): ...this new function, inverting the
15788         sense of the return value.
15789         * config/powerpcspe/powerpcspe.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15790         * config/powerpcspe/powerpcspe-protos.h
15791         (rs6000_cannot_change_mode_class_ptr): Delete.
15792         * config/powerpcspe/powerpcspe.c
15793         (rs6000_cannot_change_mode_class_ptr): Delete.
15794         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15795         (rs6000_option_override_internal): Assign to
15796         targetm.can_change_mode_class instead of
15797         rs6000_cannot_change_mode_class_ptr.
15798         (rs6000_cannot_change_mode_class): Replace with...
15799         (rs6000_can_change_mode_class): ...this new function, inverting the
15800         sense of the return value.
15801         (rs6000_debug_cannot_change_mode_class): Replace with...
15802         (rs6000_debug_can_change_mode_class): ...this new function.
15803         * config/riscv/riscv.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15804         * config/riscv/riscv.c (riscv_can_change_mode_class): New function.
15805         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15806         * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15807         * config/rs6000/rs6000-protos.h (rs6000_cannot_change_mode_class_ptr):
15808         Delete.
15809         * config/rs6000/rs6000.c (rs6000_cannot_change_mode_class_ptr): Delete.
15810         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15811         (rs6000_option_override_internal): Assign to
15812         targetm.can_change_mode_class instead of
15813         rs6000_cannot_change_mode_class_ptr.
15814         (rs6000_cannot_change_mode_class): Replace with...
15815         (rs6000_can_change_mode_class): ...this new function, inverting the
15816         sense of the return value.
15817         (rs6000_debug_cannot_change_mode_class): Replace with...
15818         (rs6000_debug_can_change_mode_class): ...this new function.
15819         * config/s390/s390.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15820         * config/s390/s390-protos.h (s390_cannot_change_mode_class): Delete.
15821         * config/s390/s390.c (s390_cannot_change_mode_class): Replace with...
15822         (s390_can_change_mode_class): ...this new function, inverting the
15823         sense of the return value.
15824         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15825         * config/sh/sh.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15826         * config/sh/sh-protos.h (sh_cannot_change_mode_class): Delete.
15827         * config/sh/sh.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15828         (sh_cannot_change_mode_class): Replace with...
15829         (sh_can_change_mode_class): ...this new function, inverting the
15830         sense of the return value.
15831         * config/sparc/sparc.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15832         * config/sparc/sparc.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15833         (sparc_can_change_mode_class): New function.
15834         * config/spu/spu.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15835         * config/spu/spu.c (spu_can_change_mode_class): New function.
15836         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15837         * config/visium/visium.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15838         * config/visium/visium.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15839         (visium_can_change_mode_class): New function.
15840         * system.h (CANNOT_CHANGE_MODE_CLASS): Poison.
15842 2017-09-15  Richard Biener  <rguenther@suse.de>
15844         PR tree-optimization/82217
15845         * tree-ssa-sccvn.c (visit_phi): Properly handle all VN_TOP
15846         but not undefined case.
15848 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15850         PR target/82145
15851         * postreload.c (reload_cse_simplify_operands): Skip
15852         NOTE_INSN_DELETED_LABEL similarly to skipping CODE_LABEL.
15854 2017-09-15  Richard Biener  <rguenther@suse.de>
15856         PR tree-optimization/68823
15857         * graphite-scop-detection.c (build_alias_set): If we have a
15858         possible dependence check whether we can handle them by just
15859         looking at the DRs DR_ACCESS_FNs.
15860         (build_scops): If build_alias_set fails, fail the SCOP.
15862 2017-09-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
15864         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_1_HW): New macros
15865         to support float128 built-in functions that require the ISA 3.0
15866         hardware.
15867         (BU_FLOAT128_3_HW): Likewise.
15868         (SQRTF128): Add support for the IEEE 128-bit square root and fma
15869         built-in functions.
15870         (FMAF128): Likewise.
15871         (FMAQ): Likewise.
15872         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add
15873         support for built-in functions that need the ISA 3.0 IEEE 128-bit
15874         floating point instructions.
15875         (rs6000_invalid_builtin): Likewise.
15876         (rs6000_builtin_mask_names): Likewise.
15877         * config/rs6000/rs6000.h (MASK_FLOAT128_HW): Likewise.
15878         (RS6000_BTM_FLOAT128_HW): Likewise.
15879         (RS6000_BTM_COMMON): Likewise.
15880         * config/rs6000/rs6000.md (fma<mode>4_hw): Add a generator
15881         function.
15882         * doc/extend.texi (RS/6000 built-in functions): Document the
15883         IEEE 128-bit floating point square root and fused multiply-add
15884         built-in functions.
15886 2017-09-14  Pat Haugen  <pthaugen@us.ibm.com>
15888         * config/rs6000/rs6000.c (rs6000_set_up_by_prologue): Make sure the TOC
15889         reg (r2) isn't in the set of registers defined in the prologue.
15891 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15892             Alan Hayward  <alan.hayward@arm.com>
15893             David Sherwood  <david.sherwood@arm.com>
15895         * tree-vectorizer.h (_loop_vec_info): Add max_vectorization_factor.
15896         (LOOP_VINFO_MAX_VECT_FACTOR): New macro.
15897         (LOOP_VINFO_ORIG_VECT_FACTOR): Replace with...
15898         (LOOP_VINFO_ORIG_MAX_VECT_FACTOR): ...this new macro.
15899         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Update
15900         accordingly.
15901         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15902         max_vectorization_factor.
15903         (vect_analyze_loop_2): Set LOOP_VINFO_MAX_VECT_FACTOR.
15905 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15906             Alan Hayward  <alan.hayward@arm.com>
15907             David Sherwood  <david.sherwood@arm.com>
15909         * tree-vectorizer.h (vect_min_worthwhile_factor): Delete.
15910         (vect_worthwhile_without_simd_p): Declare.
15911         * tree-vect-loop.c (vect_worthwhile_without_simd_p): New function.
15912         (vectorizable_reduction): Use it.
15913         * tree-vect-stmts.c (vectorizable_shift): Likewise.
15914         (vectorizable_operation): Likewise.
15916 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15917             Alan Hayward  <alan.hayward@arm.com>
15918             David Sherwood  <david.sherwood@arm.com>
15920         * tree-vectorizer.h (vect_get_num_copies): New function.
15921         * tree-vect-data-refs.c (vect_get_data_access_cost): Use it.
15922         * tree-vect-loop.c (vectorizable_reduction): Likewise.
15923         (vectorizable_induction): Likewise.
15924         (vectorizable_live_operation): Likewise.
15925         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
15926         (vectorizable_bswap): Likewise.
15927         (vectorizable_call): Likewise.
15928         (vectorizable_conversion): Likewise.
15929         (vectorizable_assignment): Likewise.
15930         (vectorizable_shift): Likewise.
15931         (vectorizable_operation): Likewise.
15932         (vectorizable_store): Likewise.
15933         (vectorizable_load): Likewise.
15934         (vectorizable_condition): Likewise.
15935         (vectorizable_comparison): Likewise.
15936         (vect_analyze_stmt): Pass the slp node to vectorizable_live_operation.
15938 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15939             Alan Hayward  <alan.hayward@arm.com>
15940             David Sherwood  <david.sherwood@arm.com>
15942         * tree-vect-loop.c (vectorizable_induction): Use gimple_build instead
15943         of vect_init_vector.
15945 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15946             Alan Hayward  <alan.hayward@arm.com>
15947             David Sherwood  <david.sherwood@arm.com>
15949         * gimple-fold.h (gimple_build_vector_from_val): Declare, and provide
15950         an inline wrapper that provides a location.
15951         (gimple_build_vector): Likewise.
15952         * gimple-fold.c (gimple_build_vector_from_val): New function.
15953         (gimple_build_vector): Likewise.
15954         * tree-vect-loop.c (get_initial_def_for_reduction): Use the new
15955         functions to build the initial value.  Always return a gimple value.
15956         (get_initial_defs_for_reduction): Likewise.  Only compute
15957         neutral_vec once.
15958         (vect_create_epilog_for_reduction): Don't call force_gimple_operand or
15959         vect_init_vector on the results from get_initial_def(s)_for_reduction.
15960         (vectorizable_induction): Use gimple_build_vector rather than
15961         vect_init_vector.
15963 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15964             Alan Hayward  <alan.hayward@arm.com>
15965             David Sherwood  <david.sherwood@arm.com>
15967         * target.h (vec_perm_indices): New typedef.
15968         (auto_vec_perm_indices): Likewise.
15969         * optabs-query.h: Include target.h
15970         (can_vec_perm_p): Take a vec_perm_indices *.
15971         * optabs-query.c (can_vec_perm_p): Likewise.
15972         (can_mult_highpart_p): Update accordingly.  Use auto_vec_perm_indices.
15973         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15974         * tree-vect-generic.c (lower_vec_perm): Likewise.
15975         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15976         (vect_grouped_load_supported): Likewise.
15977         (vect_shift_permute_load_chain): Likewise.
15978         (vect_permute_store_chain): Use auto_vec_perm_indices.
15979         (vect_permute_load_chain): Likewise.
15980         * fold-const.c (fold_vec_perm): Take vec_perm_indices.
15981         (fold_ternary_loc): Update accordingly.  Use auto_vec_perm_indices.
15982         Update uses of can_vec_perm_p.
15983         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Replace the
15984         mode with a number of elements.  Take a vec_perm_indices *.
15985         (vect_create_epilog_for_reduction): Update accordingly.
15986         Use auto_vec_perm_indices.
15987         (have_whole_vector_shift): Likewise.  Update call to can_vec_perm_p.
15988         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15989         (vect_transform_slp_perm_load): Likewise.
15990         (vect_schedule_slp_instance): Use auto_vec_perm_indices.
15991         * tree-vectorizer.h (vect_gen_perm_mask_any): Take a vec_perm_indices.
15992         (vect_gen_perm_mask_checked): Likewise.
15993         * tree-vect-stmts.c (vect_gen_perm_mask_any): Take a vec_perm_indices.
15994         (vect_gen_perm_mask_checked): Likewise.
15995         (vectorizable_mask_load_store): Use auto_vec_perm_indices.
15996         (vectorizable_store): Likewise.
15997         (vectorizable_load): Likewise.
15998         (perm_mask_for_reverse): Likewise.  Update call to can_vec_perm_p.
15999         (vectorizable_bswap): Likewise.
16001 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
16002             Alan Hayward  <alan.hayward@arm.com>
16003             David Sherwood  <david.sherwood@arm.com>
16005         * tree.h (build_vector): Take a vec<tree> instead of a tree *.
16006         * tree.c (build_vector): Likewise.
16007         (build_vector_from_ctor): Update accordingly.
16008         (build_vector_from_val): Likewise.
16009         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
16010         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16011         * tree-vect-generic.c (add_rshift): Likewise.
16012         (expand_vector_divmod): Likewise.
16013         (optimize_vector_constructor): Likewise.
16014         * tree-vect-slp.c (vect_get_constant_vectors): Likewise.
16015         (vect_transform_slp_perm_load): Likewise.
16016         (vect_schedule_slp_instance): Likewise.
16017         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
16018         (vectorizable_call): Likewise.
16019         (vect_gen_perm_mask_any): Likewise.  Add elements in order.
16020         * expmed.c (make_tree): Likewise.
16021         * fold-const.c (fold_negate_expr_1): Use auto_vec<tree> when building
16022         a vector passed to build_vector.
16023         (fold_convert_const): Likewise.
16024         (exact_inverse): Likewise.
16025         (fold_ternary_loc): Likewise.
16026         (fold_relational_const): Likewise.
16027         (const_binop): Likewise.  Use VECTOR_CST_ELT directly when operating
16028         on VECTOR_CSTs, rather than going through vec_cst_ctor_to_array.
16029         (const_unop): Likewise.  Store the reduction accumulator in a
16030         variable rather than an array.
16031         (vec_cst_ctor_to_array): Take the number of elements as a parameter.
16032         (fold_vec_perm): Update calls accordingly.  Use auto_vec<tree> for
16033         the new vector, rather than constructing it after the input arrays.
16034         (native_interpret_vector): Use auto_vec<tree> when building
16035         a vector passed to build_vector.  Add elements in order.
16036         * tree-vect-loop.c (get_initial_defs_for_reduction): Use
16037         auto_vec<tree> when building a vector passed to build_vector.
16038         (vect_create_epilog_for_reduction): Likewise.
16039         (vectorizable_induction): Likewise.
16040         (get_initial_def_for_reduction): Likewise.  Fix indentation of
16041         case statements.
16042         * config/sparc/sparc.c (sparc_handle_vis_mul8x16): Change n_elts
16043         to a vec<tree> *.
16044         (sparc_fold_builtin): Use auto_vec<tree> when building a vector
16045         passed to build_vector.
16047 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
16048             Alan Hayward  <alan.hayward@arm.com>
16049             David Sherwood  <david.sherwood@arm.com>
16051         * tree-core.h (tree_base::u): Add an "nelts" field.
16052         (tree_vector): Use VECTOR_CST_NELTS as the length.
16053         * tree.c (tree_size): Likewise.
16054         (make_vector): Initialize VECTOR_CST_NELTS.
16055         * tree.h (VECTOR_CST_NELTS): Use the u.nelts field.
16056         * cfgexpand.c (expand_debug_expr): Use VECTOR_CST_NELTS instead of
16057         TYPE_VECTOR_SUBPARTS.
16058         * expr.c (const_vector_mask_from_tree): Consistently use "units"
16059         as the number of units, setting it from VECTOR_CST_NELTS.
16060         (const_vector_from_tree): Likewise.
16061         * fold-const.c (negate_expr_p): Use VECTOR_CST_NELTS instead of
16062         TYPE_VECTOR_SUBPARTS for the number of elements in a VECTOR_CST.
16063         (fold_negate_expr_1): Likewise.
16064         (fold_convert_const): Likewise.
16065         (const_binop): Likewise.  Differentiate the number of output and
16066         input elements.
16067         (const_unop): Likewise.
16068         (fold_ternary_loc): Use VECTOR_CST_NELTS for the number of elements
16069         in a VECTOR_CST, asserting that it is the same as TYPE_VECTOR_SUBPARTS
16070         in cases that did the opposite.
16072 2017-09-14  Richard Biener  <rguenther@suse.de>
16074         * tree-ssa-sccvn.c (visit_phi): Merge undefined values similar
16075         to VN_TOP.
16077 2017-09-14  Eric Botcazou  <ebotcazou@adacore.com>
16079         * dwarf2out.c (dwarf2out_source_line): Remove superfluous test.
16081 2017-09-14  Jakub Jelinek  <jakub@redhat.com>
16083         PR target/81325
16084         * cfgbuild.c (find_bb_boundaries): Ignore debug insns in decisions
16085         if and where to split a bb, except for splitting before debug insn
16086         sequences followed by non-label real insn.  Delete debug insns
16087         in between basic blocks.
16089         * combine.c (make_compound_operation_int): Formatting fixes.
16091         * config/alpha/elf.h (LINK_EH_SPEC): Add -static-pie support.
16092         * config/alpha/linux.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
16093         * config/netbsd.h (LINK_EH_SPEC): Likewise.
16094         * config/sol2.h (LINK_EH_SPEC): Likewise.
16095         * config/arm/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
16096         * config/s390/linux.h (LINK_SPEC): Likewise.
16097         * config/freebsd.h (LINK_EH_SPEC): Likewise.
16098         * config/openbsd.h (LINK_EH_SPEC): Likewise.
16099         * config/lm32/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
16100         * config/aarch64/aarch64-linux.h (LINUX_TARGET_LINK_SPEC): Likewise.
16101         * config/powerpcspe/sysv4.h (LINK_EH_SPEC): Likewise.
16102         * config/bfin/linux.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
16103         * config/i386/gnu-user64.h (GNU_USER_TARGET_LINK_SPEC): Fix a typo.
16104         * config/i386/gnu-user.h (GNU_USER_TARGET_LINK_SPEC): Formatting fix.
16106 2017-09-13  Jakub Jelinek  <jakub@redhat.com>
16108         * config/rs6000/sysv4.h (STARTFILE_LINUX_SPEC): Add -static-pie
16109         support.
16110         (ENDFILE_LINUX_SPEC): Likewise.
16111         (LINK_EH_SPEC): Likewise.
16112         * config/rs6000/linux64.h (LINK_SHLIB_SPEC): Likewise.
16113         (LINK_OS_LINUX_SPEC32): Likewise.
16114         (LINK_OS_LINUX_SPEC64): Likewise.
16115         * config/rs6000/linux.h (LINK_SHLIB_SPEC): Likewise.
16116         (LINK_OS_LINUX_SPEC): Likewise.
16118 2017-09-13  Martin Liska  <mliska@suse.cz>
16120         PR middle-end/82154
16121         * stmt.c (expand_sjlj_dispatch_table): Use CASE_LOW when
16122         CASE_HIGH is NULL_TREE.
16124 2017-09-13  Richard Sandiford  <richard.sandiford@linaro.org>
16125             Alan Hayward  <alan.hayward@arm.com>
16126             David Sherwood  <david.sherwood@arm.com>
16128         * target.def (secondary_memory_needed): New hook.
16129         (secondary_reload): Refer to TARGET_SECONDARY_MEMORY_NEEDED
16130         instead of SECONDARY_MEMORY_NEEDED.
16131         (secondary_memory_needed_mode): Likewise.
16132         * hooks.h (hook_bool_mode_reg_class_t_reg_class_t_false): Declare.
16133         * hooks.c (hook_bool_mode_reg_class_t_reg_class_t_false): New function.
16134         * doc/tm.texi.in (SECONDARY_MEMORY_NEEDED): Replace with...
16135         (TARGET_SECONDARY_MEMORY_NEEDED): ...this.
16136         (SECONDARY_MEMORY_NEEDED_RTX): Update reference accordingly.
16137         * doc/tm.texi: Regenerate.
16138         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED): Delete.
16139         * config/alpha/alpha.c (alpha_secondary_memory_needed): New function.
16140         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16141         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED): Delete.
16142         * config/i386/i386-protos.h (ix86_secondary_memory_needed): Delete.
16143         * config/i386/i386.c (inline_secondary_memory_needed): Put the
16144         mode argument first and change the reg_class arguments to reg_class_t.
16145         (ix86_secondary_memory_needed): Likewise.  Remove the strict parameter.
16146         Make static.  Update the call to inline_secondary_memory_needed.
16147         (ix86_register_move_cost): Update the call to
16148         inline_secondary_memory_needed.
16149         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16150         * config/ia64/ia64.h (SECONDARY_MEMORY_NEEDED): Delete commented-out
16151         definition.
16152         * config/ia64/ia64.c (spill_xfmode_rfmode_operand): Refer to
16153         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
16154         in comment.
16155         * config/mips/mips.h (SECONDARY_MEMORY_NEEDED): Delete.
16156         * config/mips/mips-protos.h (mips_secondary_memory_needed): Delete.
16157         * config/mips/mips.c (mips_secondary_memory_needed): Make static
16158         and match hook interface.  Add comment from mips.h.
16159         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16160         * config/mmix/mmix.md (truncdfsf2): Refer to
16161         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
16162         in comment.
16163         * config/pa/pa-64.h (SECONDARY_MEMORY_NEEDED): Rename to...
16164         (PA_SECONDARY_MEMORY_NEEDED): ...this, and put the mode argument first.
16165         * config/pa/pa.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16166         (pa_secondary_memory_needed): New function.
16167         * config/pdp11/pdp11.h (SECONDARY_MEMORY_NEEDED): Delete.
16168         * config/pdp11/pdp11-protos.h (pdp11_secondary_memory_needed): Delete.
16169         * config/pdp11/pdp11.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16170         (pdp11_secondary_memory_needed): Make static and match hook interface.
16171         * config/powerpcspe/powerpcspe.h (SECONDARY_MEMORY_NEEDED): Delete.
16172         * config/powerpcspe/powerpcspe-protos.h
16173         (rs6000_secondary_memory_needed_ptr): Delete.
16174         * config/powerpcspe/powerpcspe.c (rs6000_secondary_memory_needed_ptr):
16175         Delete.
16176         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16177         (rs6000_option_override_internal): Assign to
16178         targetm.secondary_memory_needed rather than
16179         rs6000_secondary_memory_needed_ptr.
16180         (rs6000_secondary_memory_needed): Match hook interface.
16181         (rs6000_debug_secondary_memory_needed): Likewise.
16182         * config/riscv/riscv.h (SECONDARY_MEMORY_NEEDED): Delete.
16183         * config/riscv/riscv.c (riscv_secondary_memory_needed): New function.
16184         (riscv_register_move_cost): Use it instead of SECONDARY_MEMORY_NEEDED.
16185         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16186         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED): Delete.
16187         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_ptr):
16188         Delete.
16189         * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_ptr): Delete.
16190         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16191         (rs6000_option_override_internal): Assign to
16192         targetm.secondary_memory_needed rather than
16193         rs6000_secondary_memory_needed_ptr.
16194         (rs6000_secondary_memory_needed): Match hook interface.
16195         (rs6000_debug_secondary_memory_needed): Likewise.
16196         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED): Delete.
16197         * config/s390/s390.c (s390_secondary_memory_needed): New function.
16198         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16199         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED): Delete.
16200         * config/sparc/sparc.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16201         (sparc_secondary_memory_needed): New function.
16202         * lra-constraints.c (check_and_process_move): Refer to
16203         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
16204         in comment.
16205         (curr_insn_transform): Likewise.
16206         (process_alt_operands): Use targetm.secondary_memory_needed
16207         instead of TARGET_SECONDARY_MEMORY_NEEDED.
16208         (check_secondary_memory_needed_p): Likewise.
16209         (choose_split_class): Likewise.
16210         * reload.c: Unconditionally include code that was previously
16211         conditional on SECONDARY_MEMORY_NEEDED.
16212         (push_secondary_reload): Use targetm.secondary_memory_needed
16213         instead of TARGET_SECONDARY_MEMORY_NEEDED.
16214         (push_reload): Likewise.
16215         * reload1.c: Unconditionally include code that was previously
16216         conditional on SECONDARY_MEMORY_NEEDED.
16217         (choose_reload_regs): Use targetm.secondary_memory_needed
16218         instead of TARGET_SECONDARY_MEMORY_NEEDED.
16219         (gen_reload): Likewise.
16220         * system.h (SECONDARY_MEMORY_NEEDED): Poison.
16222 2017-09-13  Richard Sandiford  <richard.sandiford@linaro.org>
16223             Alan Hayward  <alan.hayward@arm.com>
16224             David Sherwood  <david.sherwood@arm.com>
16226         * target.def (secondary_memory_needed_mode): New hook:
16227         * targhooks.c (default_secondary_memory_needed_mode): Declare.
16228         * targhooks.h (default_secondary_memory_needed_mode): New function.
16229         * doc/tm.texi.in (SECONDARY_MEMORY_NEEDED_MODE): Replace with...
16230         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): ...this.
16231         * doc/tm.texi: Regenerate.
16232         * lra-constraints.c (check_and_process_move): Use
16233         targetm.secondary_memory_needed_mode instead of
16234         TARGET_SECONDARY_MEMORY_NEEDED_MODE.
16235         (curr_insn_transform): Likewise.
16236         * reload.c (get_secondary_mem): Likewise.
16237         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16238         * config/alpha/alpha.c (alpha_secondary_memory_needed_mode): New
16239         function.
16240         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16241         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16242         * config/i386/i386.c (ix86_secondary_memory_needed_mode): New function.
16243         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16244         * config/powerpcspe/powerpcspe.h (SECONDARY_MEMORY_NEEDED_MODE):
16245         Delete.
16246         * config/powerpcspe/powerpcspe-protos.h
16247         (rs6000_secondary_memory_needed_mode): Delete.
16248         * config/powerpcspe/powerpcspe.c
16249         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16250         (rs6000_secondary_memory_needed_mode): Make static.
16251         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16252         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_mode):
16253         Delete.
16254         * config/rs6000/rs6000.c (TARGET_SECONDARY_MEMORY_NEEDED_MODE):
16255         Redefine.
16256         (rs6000_secondary_memory_needed_mode): Make static.
16257         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16258         * config/s390/s390.c (s390_secondary_memory_needed_mode): New function.
16259         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16260         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16261         * config/sparc/sparc.c (TARGET_SECONDARY_MEMORY_NEEDED_MODE):
16262         Redefine.
16263         (sparc_secondary_memory_needed_mode): New function.
16264         * system.h (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Poison.
16266 2017-09-13  Jackson Woodruff  <jackson.woodruff@arm.com>
16268         * config/aarch64/constraints.md (Umq): New constraint.
16269         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>):
16270         Change to use Umq.
16271         (mov<mode>): Update condition.
16273 2017-09-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16275         * gimple-ssa-store-merging.c (sort_by_bitpos): Compare store order
16276         when bitposition is the same.
16278 2017-09-13  Richard Biener  <rguenther@suse.de>
16280         * dwarf2out.c (output_die_symbol): Remove.
16281         (output_die): Do not output a DIEs symbol.
16283 2017-09-13  Richard Biener  <rguenther@suse.de>
16285         PR middle-end/82128
16286         * gimple-fold.c (gimple_fold_call): Update SSA name in-place to
16287         default-def to avoid breaking iterator update with the weird
16288         interaction with cgraph_update_edges_for_call_stmt_node.
16290 2017-09-13  Richard Biener  <rguenther@suse.de>
16292         * tree-cfg.c (verify_gimple_assign_binary): Add verification
16293         for WIDEN_SUM_EXPR, VEC_WIDEN_MULT_{HI,LO,EVEN,ODD}_EXPR,
16294         VEC_PACK_{TRUNC,SAT,FIX_TRUNC}_EXPR.
16295         (verify_gimple_assign_ternary): Add verification for DOT_PROD_EXPR.
16297 2017-09-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
16299         * config/aarch64/aarch64.c (aarch64_override_options_after_change_1):
16300         Disable pc relative literal load irrespective of
16301         TARGET_FIX_ERR_A53_84341 for default.
16303 2017-09-12  Eric Botcazou  <ebotcazou@adacore.com>
16305         * config/sparc/sparc.c (output_return): Output the source location of
16306         the insn in the delay slot, if any.
16307         (output_sibcall): Likewise.
16309 2017-09-12  H.J. Lu  <hongjiu.lu@intel.com>
16311         PR driver/81498
16312         * common.opt (-static-pie): New alias.
16313         (shared): Negate static-pie.
16314         (-no-pie): Update help text.
16315         (-pie): Likewise.
16316         (static-pie): New option.
16317         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Add
16318         -static-pie support.
16319         (GNU_USER_TARGET_ENDFILE_SPEC): Likewise.
16320         (LINK_EH_SPEC): Likewise.
16321         (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
16322         * config/i386/gnu-user.h (GNU_USER_TARGET_LINK_SPEC): Likewise.
16323         * config/i386/gnu-user64.h (GNU_USER_TARGET_LINK_SPEC): Likewise.
16324         * gcc.c (LINK_COMMAND_SPEC): Likewise.
16325         (init_gcc_specs): Likewise.
16326         (init_spec): Likewise.
16327         (display_help): Update help message for -pie.
16328         * doc/invoke.texi: Update -pie, -no-pie and -static.  Document
16329         -static-pie.
16331 2017-09-12  Wilco Dijkstra  <wdijkstr@arm.com>
16333         * config/aarch64/aarch64.md (movsi_aarch64): Remove all '*'.
16334         (movdi_aarch64): Likewise.
16335         (movti_aarch64): Likewise.
16337 2017-09-12 Simon Wright <simon@pushface.org>
16339         PR target/80204
16340         * config/darwin-driver.c (darwin_find_version_from_kernel): Eliminate
16341         calculation of the minor version, always output as 0.
16343 2017-09-12  Jakub Jelinek  <jakub@redhat.com>
16345         PR target/82112
16346         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): For
16347         ALTIVEC_BUILTIN_VEC_LD if arg1 has array type call default_conversion
16348         on it early, rather than manual conversion late.  For
16349         ALTIVEC_BUILTIN_VEC_ST if arg2 has array type call default_conversion
16350         instead of performing manual conversion.
16352 2017-09-12  Carl Love  <cel@us.ibm.com>
16354         * config/rs6000/altivec.md (vec_widen_umult_even_v4si,
16355         vec_widen_smult_even_v4si): Add define expands for vmuleuw, vmulesw,
16356         vmulouw, vmulosw.
16357         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
16358         VMULOSW): Add definitions.
16359         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
16360         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
16361         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
16362         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
16363         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
16365 2017-09-12  James Greenhalgh  <james.greenhalgh@arm.com>
16367         * config/aarch64/aarch64.md (movdi_aarch64): Set load/store
16368         types correctly.
16369         (movti_aarch64): Likewise.
16370         (movdf_aarch64): Likewise.
16371         (movtf_aarch64): Likewise.
16372         (load_pairdi): Likewise.
16373         (store_pairdi): Likewise.
16374         (load_pairdf): Likewise.
16375         (store_pairdf): Likewise.
16376         (loadwb_pair<GPI:mode>_<P:mode>): Likewise.
16377         (storewb_pair<GPI:mode>_<P:mode>): Likewise.
16378         (ldr_got_small_<mode>): Likewise.
16379         (ldr_got_small_28k_<mode>): Likewise.
16380         (ldr_got_tiny): Likewise.
16381         * config/aarch64/iterators.md (ldst_sz): New.
16382         (ldpstp_sz): Likewise.
16383         * config/aarch64/thunderx.md (thunderx_storepair): Split store_8
16384         to store_16.
16385         (thunderx_load): Split load_8 to load_16.
16386         * config/aarch64/thunderx2t99.md (thunderx2t99_loadpair): Split
16387         load_8 to load_16.
16388         (thunderx2t99_storepair_basic): Split store_8 to store_16.
16389         * config/arm/xgene1.md (xgene1_load_pair): Split load_8 to load_16.
16390         (xgene1_store_pair): Split store_8 to store_16.
16391         * config/aarch64/falkor.md (falkor_ld_3_ld): Split load_8 to load_16.
16392         (falkor_st_0_st_sd): Split store_8 to store_16.
16394 2017-09-12  James Greenhalgh  <james.greenhalgh@arm.com>
16396         * config/arm/types.md (type): Rename load1/2/3/4 to load_4/8/12/16
16397         and store1/2/3/4 to store_4/8/12/16.
16398         * config/aarch64/aarch64.md: Update for rename.
16399         * config/arm/arm.md: Likewise.: Likewise.
16400         * config/arm/arm.c: Likewise.
16401         * config/arm/thumb1.md: Likewise.
16402         * config/arm/thumb2.md: Likewise.
16403         * config/arm/vfp.md: Likewise.
16404         * config/arm/arm-generic.md: Likewise.
16405         * config/arm/arm1020e.md: Likewise.
16406         * config/arm/arm1026ejs.md: Likewise.
16407         * config/arm/arm1136jfs.md: Likewise.
16408         * config/arm/arm926ejs.md: Likewise.
16409         * config/arm/cortex-a15.md: Likewise.
16410         * config/arm/cortex-a17.md: Likewise.
16411         * config/arm/cortex-a5.md: Likewise.
16412         * config/arm/cortex-a53.md: Likewise.
16413         * config/arm/cortex-a57.md: Likewise.
16414         * config/arm/cortex-a7.md: Likewise.
16415         * config/arm/cortex-a8.md: Likewise.
16416         * config/arm/cortex-a9.md: Likewise.
16417         * config/arm/cortex-m4.md: Likewise.
16418         * config/arm/cortex-m7.md: Likewise.
16419         * config/arm/cortex-r4.md: Likewise.
16420         * config/arm/exynos-m1.md: Likewise.
16421         * config/arm/fa526.md: Likewise.
16422         * config/arm/fa606te.md: Likewise.
16423         * config/arm/fa626te.md: Likewise.
16424         * config/arm/fa726te.md: Likewise.
16425         * config/arm/fmp626.md: Likewise.
16426         * config/arm/iwmmxt.md: Likewise.
16427         * config/arm/ldmstm.md: Likewise.
16428         * config/arm/marvell-pj4.md: Likewise.
16429         * config/arm/xgene1.md: Likewise.
16430         * config/aarch64/thunderx.md: Likewise.
16431         * config/aarch64/thunderx2t99.md: Likewise.
16432         * config/aarch64/falkor.md: Likewise.
16434 2017-09-12  Martin Liska  <mliska@suse.cz>
16436         * attribs.c (private_lookup_attribute): New function.
16437         * attribs.h (private_lookup_attribute): Declared here.
16438         (lookup_attribute): Called from this place.
16440 2017-09-12  Richard Biener  <rguenther@suse.de>
16442         PR tree-optimization/82157
16443         * tree-ssa-pre.c (remove_dead_inserted_code): Do not remove
16444         stmts with side-effects.
16446 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16447             Alan Hayward  <alan.hayward@arm.com>
16448             David Sherwood <david.sherwood@arm.com>
16450         * target.def (hard_regno_nregs): New hook.
16451         (class_max_nregs): Refer to it instead of HARD_REGNO_NREGS.
16452         * targhooks.h (default_hard_regno_nregs): Declare.
16453         * targhooks.c (default_hard_regno_nregs): New function.
16454         * doc/tm.texi.in (HARD_REGNO_NREGS): Replace with...
16455         (TARGET_HARD_REGNO_NREGS): ...this hook.
16456         (HARD_REGNO_NREGS_HAS_PADDING): Update accordingly.
16457         (CLASS_MAX_NREGS): Likewise.
16458         * doc/tm.texi: Regenerate.
16459         * reginfo.c (init_reg_modes_target): Use targetm.hard_regno_nregs
16460         instead of HARD_REGNO_NREGS.
16461         * rtl.h (REG_NREGS): Refer to TARGET_HARD_REGNO_NREGS rather than
16462         HARD_REGNO_NREGS in the comment.
16463         * config/aarch64/aarch64.h (HARD_REGNO_NREGS): Delete.
16464         * config/aarch64/aarch64-protos.h (aarch64_hard_regno_nregs): Delete.
16465         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs): Make static.
16466         Return an unsigned int.
16467         (TARGET_HARD_REGNO_NREGS): Redefine.
16468         * config/alpha/alpha.h (HARD_REGNO_NREGS): Delete.
16469         * config/arc/arc.h (HARD_REGNO_NREGS): Delete.
16470         * config/arc/arc.c (TARGET_HARD_REGNO_NREGS): Redefine.
16471         (arc_hard_regno_nregs): New function.
16472         * config/arm/arm.h (HARD_REGNO_NREGS): Delete.
16473         * config/arm/arm.c (TARGET_HARD_REGNO_NREGS): Redefine.
16474         (arm_hard_regno_nregs): New function.
16475         * config/avr/avr.h (HARD_REGNO_NREGS): Delete.
16476         * config/bfin/bfin.h (HARD_REGNO_NREGS): Delete.
16477         * config/bfin/bfin.c (bfin_hard_regno_nregs): New function.
16478         (TARGET_HARD_REGNO_NREGS): Redefine.
16479         * config/c6x/c6x.h (HARD_REGNO_NREGS): Delete.
16480         * config/cr16/cr16.h (LONG_REG_P): Use targetm.hard_regno_nregs.
16481         (HARD_REGNO_NREGS): Delete.
16482         * config/cr16/cr16.c (TARGET_HARD_REGNO_NREGS): Redefine.
16483         (cr16_hard_regno_nregs): New function.
16484         (cr16_memory_move_cost): Use it instead of HARD_REGNO_NREGS.
16485         * config/cris/cris.h (HARD_REGNO_NREGS): Delete.
16486         * config/cris/cris.c (TARGET_HARD_REGNO_NREGS): Redefine.
16487         (cris_hard_regno_nregs): New function.
16488         * config/epiphany/epiphany.h (HARD_REGNO_NREGS): Delete.
16489         * config/fr30/fr30.h (HARD_REGNO_NREGS): Delete.
16490         (CLASS_MAX_NREGS): Use targetm.hard_regno_nregs.
16491         * config/frv/frv.h (HARD_REGNO_NREGS): Delete.
16492         (CLASS_MAX_NREGS): Remove outdated copy of documentation.
16493         * config/frv/frv-protos.h (frv_hard_regno_nregs): Delete.
16494         * config/frv/frv.c (TARGET_HARD_REGNO_NREGS): Redefine.
16495         (frv_hard_regno_nregs): Make static.  Take and return an
16496         unsigned int.
16497         (frv_class_max_nregs): Remove outdated copy of documentation.
16498         * config/ft32/ft32.h (HARD_REGNO_NREGS): Delete.
16499         * config/h8300/h8300.h (HARD_REGNO_NREGS): Delete.
16500         * config/h8300/h8300-protos.h (h8300_hard_regno_nregs): Delete.
16501         * config/h8300/h8300.c (h8300_hard_regno_nregs): Delete.
16502         * config/i386/i386.h (HARD_REGNO_NREGS): Delete.
16503         * config/i386/i386.c (ix86_hard_regno_nregs): New function.
16504         (TARGET_HARD_REGNO_NREGS): Redefine.
16505         * config/ia64/ia64.h (HARD_REGNO_NREGS): Delete.
16506         (CLASS_MAX_NREGS): Update comment.
16507         * config/ia64/ia64.c (TARGET_HARD_REGNO_NREGS): Redefine.
16508         (ia64_hard_regno_nregs): New function.
16509         * config/iq2000/iq2000.h (HARD_REGNO_NREGS): Delete.
16510         * config/lm32/lm32.h (HARD_REGNO_NREGS): Delete.
16511         * config/m32c/m32c.h (HARD_REGNO_NREGS): Delete.
16512         * config/m32c/m32c-protos.h (m32c_hard_regno_nregs): Delete.
16513         * config/m32c/m32c.c (m32c_hard_regno_nregs_1): Take and return
16514         an unsigned int.
16515         (m32c_hard_regno_nregs): Likewise.  Make static.
16516         (TARGET_HARD_REGNO_NREGS): Redefine.
16517         * config/m32r/m32r.h (HARD_REGNO_NREGS): Delete.
16518         * config/m68k/m68k.h (HARD_REGNO_NREGS): Delete.
16519         * config/m68k/m68k.c (TARGET_HARD_REGNO_NREGS): Redefine.
16520         (m68k_hard_regno_nregs): New function.
16521         * config/mcore/mcore.h (HARD_REGNO_NREGS): Delete.
16522         * config/microblaze/microblaze.h (HARD_REGNO_NREGS): Delete.
16523         * config/mips/mips.h (HARD_REGNO_NREGS): Delete.
16524         * config/mips/mips-protos.h (mips_hard_regno_nregs): Delete.
16525         * config/mips/mips.c (mips_hard_regno_nregs): Make static.
16526         Take and return an unsigned int.
16527         (TARGET_HARD_REGNO_NREGS): Redefine.
16528         * config/mmix/mmix.h (HARD_REGNO_NREGS): Delete.
16529         (CLASS_MAX_NREGS): Use targetm.hard_regno_nregs.
16530         * config/mn10300/mn10300.h (HARD_REGNO_NREGS): Delete.
16531         * config/moxie/moxie.h (HARD_REGNO_NREGS): Delete.
16532         * config/msp430/msp430.h (HARD_REGNO_NREGS): Delete.
16533         * config/msp430/msp430-protos.h (msp430_hard_regno_nregs): Delete.
16534         * config/msp430/msp430.c (TARGET_HARD_REGNO_NREGS): Redefine.
16535         (msp430_hard_regno_nregs): Make static.  Take and return an
16536         unsigned int.
16537         * config/nds32/nds32.h (HARD_REGNO_NREGS): Delete.
16538         * config/nds32/nds32-protos.h (nds32_hard_regno_nregs): Delete.
16539         * config/nds32/nds32.c (nds32_hard_regno_nregs): Delete.
16540         (nds32_hard_regno_mode_ok): Use targetm.hard_regno_nregs.
16541         * config/nios2/nios2.h (HARD_REGNO_NREGS): Delete.
16542         * config/nvptx/nvptx.h (HARD_REGNO_NREGS): Delete.
16543         * config/nvptx/nvptx.c (nvptx_hard_regno_nregs): New function.
16544         (TARGET_HARD_REGNO_NREGS): Redefine.
16545         * config/pa/pa32-regs.h (HARD_REGNO_NREGS): Rename to...
16546         (PA_HARD_REGNO_NREGS): ...this.
16547         * config/pa/pa64-regs.h (HARD_REGNO_NREGS): Rename to...
16548         (PA_HARD_REGNO_NREGS): ...this.
16549         * config/pa/pa.c (TARGET_HARD_REGNO_NREGS): Redefine.
16550         (pa_hard_regno_nregs): New function.
16551         * config/pdp11/pdp11.h (HARD_REGNO_NREGS): Delete.
16552         * config/pdp11/pdp11.c (TARGET_HARD_REGNO_NREGS): Redefine.
16553         (pdp11_hard_regno_nregs): New function.
16554         * config/powerpcspe/powerpcspe.h (HARD_REGNO_NREGS): Delete.
16555         * config/powerpcspe/powerpcspe.c (TARGET_HARD_REGNO_NREGS): Redefine.
16556         (rs6000_hard_regno_nregs_hook): New function.
16557         * config/riscv/riscv.h (HARD_REGNO_NREGS): Delete.
16558         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): Delete.
16559         * config/riscv/riscv.c (riscv_hard_regno_nregs): Make static.
16560         Take and return an unsigned int.  Move earlier in file.
16561         (TARGET_HARD_REGNO_NREGS): Redefine.
16562         * config/rl78/rl78.h (HARD_REGNO_NREGS): Delete.
16563         * config/rl78/rl78-protos.h (rl78_hard_regno_nregs): Delete.
16564         * config/rl78/rl78.c (TARGET_HARD_REGNO_NREGS): Reefine.
16565         (rl78_hard_regno_nregs): Make static.  Take and return an
16566         unsigned int.
16567         * config/rs6000/rs6000.h (HARD_REGNO_NREGS): Delete.
16568         * config/rs6000/rs6000.c (TARGET_HARD_REGNO_NREGS): Redefine.
16569         (rs6000_hard_regno_nregs_hook): New function.
16570         * config/rx/rx.h (HARD_REGNO_NREGS): Delete.
16571         * config/rx/rx.c (rx_hard_regno_nregs): New function.
16572         (TARGET_HARD_REGNO_NREGS): Redefine.
16573         * config/s390/s390.h (HARD_REGNO_NREGS): Delete.
16574         * config/s390/s390.c (REGNO_PAIR_OK): Use s390_hard_regno_nregs
16575         instead of HARD_REGNO_NREGS.
16576         (s390_hard_regno_nregs): New function.
16577         (s390_hard_regno_mode_ok): Add comment from s390.h.
16578         (TARGET_HARD_REGNO_NREGS): Redefine.
16579         * config/sh/sh.h (HARD_REGNO_NREGS): Delete.
16580         * config/sh/sh.c (TARGET_HARD_REGNO_NREGS): Redefine.
16581         (sh_hard_regno_nregs): New function.
16582         (sh_pass_in_reg_p): Use it.
16583         * config/sparc/sparc.h (HARD_REGNO_NREGS): Delete.
16584         * config/sparc/sparc.c (TARGET_HARD_REGNO_NREGS): Redefine.
16585         (sparc_hard_regno_nregs): New function.
16586         * config/spu/spu.h (HARD_REGNO_NREGS): Delete.
16587         * config/spu/spu.c (spu_hard_regno_nregs): New function.
16588         (spu_function_arg_advance): Use it, supplying a valid register number.
16589         (TARGET_HARD_REGNO_NREGS): Redefine.
16590         * config/stormy16/stormy16.h (HARD_REGNO_NREGS): Delete.
16591         * config/tilegx/tilegx.h (HARD_REGNO_NREGS): Delete.
16592         * config/tilepro/tilepro.h (HARD_REGNO_NREGS): Delete.
16593         * config/v850/v850.h (HARD_REGNO_NREGS): Delete.
16594         * config/vax/vax.h (HARD_REGNO_NREGS): Delete.
16595         * config/visium/visium.h (HARD_REGNO_NREGS): Delete.
16596         (CLASS_MAX_NREGS): Remove copy of old documentation.
16597         * config/visium/visium.c (TARGET_HARD_REGNO_NREGS): Redefine.
16598         (visium_hard_regno_nregs): New function.
16599         (visium_hard_regno_mode_ok): Use it instead of HARD_REGNO_NREGS.
16600         * config/xtensa/xtensa.h (HARD_REGNO_NREGS): Delete.
16601         * config/xtensa/xtensa.c (TARGET_HARD_REGNO_NREGS): Redefine.
16602         xtensa_hard_regno_nregs): New function.
16603         * system.h (HARD_REGNO_NREGS): Poison.
16605 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16607         * config/arm/arm.h (THUMB_SECONDARY_INPUT_RELOAD_CLASS): Use
16608         hard_regno_nregs instead of HARD_REGNO_NREGS.
16609         (THUMB_SECONDARY_OUTPUT_RELOAD_CLASS): Likewise.
16610         * config/c6x/c6x.c (c6x_expand_prologue): Likewise.
16611         (c6x_expand_epilogue): Likewise.
16612         * config/frv/frv.c (frv_alloc_temp_reg): Likewise.
16613         (frv_read_iacc_argument): Likewise.
16614         * config/sh/sh.c: Include regs.h.
16615         (sh_print_operand): Use hard_regno_nregs instead of HARD_REGNO_NREGS.
16616         (regs_used): Likewise.
16617         (output_stack_adjust): Likewise.
16618         * config/xtensa/xtensa.c (xtensa_copy_incoming_a7): Likewise.
16619         * expmed.c: Include regs.h.
16620         (store_bit_field_1): Use hard_regno_nregs instead of HARD_REGNO_NREGS.
16621         * ree.c: Include regs.h.
16622         (combine_reaching_defs): Use hard_regno_nregs instead of
16623         HARD_REGNO_NREGS.
16624         (add_removable_extension): Likewise.
16626 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16628         * regs.h (hard_regno_nregs): Turn into a function.
16629         (end_hard_regno): Update accordingly.
16630         * caller-save.c (setup_save_areas): Likewise.
16631         (save_call_clobbered_regs): Likewise.
16632         (replace_reg_with_saved_mem): Likewise.
16633         (insert_restore): Likewise.
16634         (insert_save): Likewise.
16635         * combine.c (can_change_dest_mode): Likewise.
16636         (move_deaths): Likewise.
16637         (distribute_notes): Likewise.
16638         * config/mips/mips.c (mips_hard_regno_call_part_clobbered): Likewise.
16639         * config/powerpcspe/powerpcspe.c (rs6000_cannot_change_mode_class)
16640         (rs6000_split_multireg_move): Likewise.
16641         (rs6000_register_move_cost): Likewise.
16642         (rs6000_memory_move_cost): Likewise.
16643         * config/rs6000/rs6000.c (rs6000_cannot_change_mode_class): Likewise.
16644         (rs6000_split_multireg_move): Likewise.
16645         (rs6000_register_move_cost): Likewise.
16646         (rs6000_memory_move_cost): Likewise.
16647         * cselib.c (cselib_reset_table): Likewise.
16648         (cselib_lookup_1): Likewise.
16649         * emit-rtl.c (set_mode_and_regno): Likewise.
16650         * function.c (aggregate_value_p): Likewise.
16651         * ira-color.c (setup_profitable_hard_regs): Likewise.
16652         (check_hard_reg_p): Likewise.
16653         (calculate_saved_nregs): Likewise.
16654         (assign_hard_reg): Likewise.
16655         (improve_allocation): Likewise.
16656         (calculate_spill_cost): Likewise.
16657         * ira-emit.c (modify_move_list): Likewise.
16658         * ira-int.h (ira_hard_reg_set_intersection_p): Likewise.
16659         (ira_hard_reg_in_set_p): Likewise.
16660         * ira.c (setup_reg_mode_hard_regset): Likewise.
16661         (clarify_prohibited_class_mode_regs): Likewise.
16662         (check_allocation): Likewise.
16663         * lra-assigns.c (find_hard_regno_for_1): Likewise.
16664         (lra_setup_reg_renumber): Likewise.
16665         (setup_try_hard_regno_pseudos): Likewise.
16666         (spill_for): Likewise.
16667         (assign_hard_regno): Likewise.
16668         (setup_live_pseudos_and_spill_after_risky_transforms): Likewise.
16669         * lra-constraints.c (in_class_p): Likewise.
16670         (lra_constraint_offset): Likewise.
16671         (simplify_operand_subreg): Likewise.
16672         (lra_constraints): Likewise.
16673         (split_reg): Likewise.
16674         (split_if_necessary): Likewise.
16675         (invariant_p): Likewise.
16676         (inherit_in_ebb): Likewise.
16677         * lra-lives.c (process_bb_lives): Likewise.
16678         * lra-remat.c (reg_overlap_for_remat_p): Likewise.
16679         (get_hard_regs): Likewise.
16680         (do_remat): Likewise.
16681         * lra-spills.c (assign_spill_hard_regs): Likewise.
16682         * mode-switching.c (create_pre_exit): Likewise.
16683         * postreload.c (reload_combine_recognize_pattern): Likewise.
16684         * recog.c (peep2_find_free_register): Likewise.
16685         * regcprop.c (kill_value_regno): Likewise.
16686         (set_value_regno): Likewise.
16687         (copy_value): Likewise.
16688         (maybe_mode_change): Likewise.
16689         (find_oldest_value_reg): Likewise.
16690         (copyprop_hardreg_forward_1): Likewise.
16691         * regrename.c (check_new_reg_p): Likewise.
16692         (regrename_do_replace): Likewise.
16693         * reload.c (push_reload): Likewise.
16694         (combine_reloads): Likewise.
16695         (find_dummy_reload): Likewise.
16696         (operands_match_p): Likewise.
16697         (find_reloads): Likewise.
16698         (find_equiv_reg): Likewise.
16699         (reload_adjust_reg_for_mode): Likewise.
16700         * reload1.c (count_pseudo): Likewise.
16701         (count_spilled_pseudo): Likewise.
16702         (find_reg): Likewise.
16703         (clear_reload_reg_in_use): Likewise.
16704         (free_for_value_p): Likewise.
16705         (allocate_reload_reg): Likewise.
16706         (choose_reload_regs): Likewise.
16707         (reload_adjust_reg_for_temp): Likewise.
16708         (emit_reload_insns): Likewise.
16709         (delete_output_reload): Likewise.
16710         * rtlanal.c (subreg_get_info): Likewise.
16711         * sched-deps.c (sched_analyze_reg): Likewise.
16712         * sel-sched.c (init_regs_for_mode): Likewise.
16713         (mark_unavailable_hard_regs): Likewise.
16714         (choose_best_reg_1): Likewise.
16715         (verify_target_availability): Likewise.
16716         * valtrack.c (dead_debug_insert_temp): Likewise.
16717         * var-tracking.c (track_loc_p): Likewise.
16718         (emit_note_insn_var_location): Likewise.
16719         * varasm.c (make_decl_rtl): Likewise.
16720         * reginfo.c (choose_hard_reg_mode): Likewise.
16721         (init_reg_modes_target): Refer directly to
16722         this_target_regs->x_hard_regno_nregs.
16724 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16726         * ira-costs.c (record_operand_costs): Use in_hard_reg_set_p
16727         instead of hard_regno_nregs.
16729 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16731         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Use
16732         end_hard_regno instead of hard_regno_nregs.
16733         * config/s390/s390.c (s390_reg_clobbered_rtx): Likewise.
16734         * config/sparc/sparc.h (ASM_DECLARE_REGISTER_GLOBAL): Likewise.
16735         * config/visium/visium.c (visium_hard_regno_mode_ok): Likewise.
16736         * ira-color.c (improve_allocation): Likewise.
16737         * lra-assigns.c (find_hard_regno_for_1): Likewise.
16738         * lra-lives.c (mark_regno_live): Likewise.
16739         (mark_regno_dead): Likewise.
16740         * lra-remat.c (operand_to_remat): Likewise.
16741         * lra.c (collect_non_operand_hard_regs): Likewise.
16742         * postreload.c (reload_combine_note_store): Likewise.
16743         (move2add_valid_value_p): Likewise.
16744         * reload.c (regno_clobbered_p): Likewise.
16746 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16748         * config/frv/frv.c (FOR_EACH_REGNO): Use END_REGNO instead of
16749         hard_regno_nregs.
16750         * config/v850/v850.c (v850_reorg): Likewise.
16751         * reload.c (refers_to_regno_for_reload_p): Likewise.
16752         (find_equiv_reg): Likewise.
16753         * reload1.c (reload_reg_reaches_end_p): Likewise.
16755 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16757         * caller-save.c (add_used_regs): Use REG_NREGS instead of
16758         hard_regno_nregs.
16759         * config/aarch64/aarch64.c (aarch64_split_combinev16qi): Likewise.
16760         * config/arm/arm.c (output_move_neon): Likewise.
16761         (arm_attr_length_move_neon): Likewise.
16762         (neon_split_vcombine): Likewise.
16763         * config/c6x/c6x.c (c6x_mark_reg_read): Likewise.
16764         (c6x_mark_reg_written): Likewise.
16765         (c6x_dwarf_register_span): Likewise.
16766         * config/i386/i386.c (ix86_save_reg): Likewise.
16767         * config/ia64/ia64.c (mark_reg_gr_used_mask): Likewise.
16768         (rws_access_reg): Likewise.
16769         * config/s390/s390.c (s390_call_saved_register_used): Likewise.
16770         * mode-switching.c (create_pre_exit): Likewise.
16771         * ree.c (combine_reaching_defs): Likewise.
16772         (add_removable_extension): Likewise.
16773         * regcprop.c (find_oldest_value_reg): Likewise.
16774         (copyprop_hardreg_forward_1): Likewise.
16775         * reload.c (reload_inner_reg_of_subreg): Likewise.
16776         (push_reload): Likewise.
16777         (combine_reloads): Likewise.
16778         (find_dummy_reload): Likewise.
16779         (reload_adjust_reg_for_mode): Likewise.
16780         * reload1.c (find_reload_regs): Likewise.
16781         (forget_old_reloads_1): Likewise.
16782         (reload_reg_free_for_value_p): Likewise.
16783         (reload_adjust_reg_for_temp): Likewise.
16784         (emit_reload_insns): Likewise.
16785         (delete_output_reload): Likewise.
16786         * sel-sched.c (choose_best_reg_1): Likewise.
16787         (choose_best_pseudo_reg): Likewise.
16789 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16790             Alan Hayward  <alan.hayward@arm.com>
16791             David Sherwood <david.sherwood@arm.com>
16793         * defaults.h (SLOW_UNALIGNED_ACCESS): Delete.
16794         * target.def (slow_unaligned_access): New hook.
16795         * targhooks.h (default_slow_unaligned_access): Declare.
16796         * targhooks.c (default_slow_unaligned_access): New function.
16797         * doc/tm.texi.in (SLOW_UNALIGNED_ACCESS): Replace with...
16798         (TARGET_SLOW_UNALIGNED_ACCESS): ...this.
16799         * doc/tm.texi: Regenerate.
16800         * config/alpha/alpha.h (SLOW_UNALIGNED_ACCESS): Delete.
16801         * config/arm/arm.h (SLOW_UNALIGNED_ACCESS): Delete.
16802         * config/i386/i386.h (SLOW_UNALIGNED_ACCESS): Delete commented-out
16803         definition.
16804         * config/powerpcspe/powerpcspe.h (SLOW_UNALIGNED_ACCESS): Delete.
16805         * config/powerpcspe/powerpcspe.c (TARGET_SLOW_UNALIGNED_ACCESS):
16806         Redefine.
16807         (rs6000_slow_unaligned_access): New function.
16808         (rs6000_emit_move): Use it instead of SLOW_UNALIGNED_ACCESS.
16809         (expand_block_compare): Likewise.
16810         (expand_strn_compare): Likewise.
16811         (rs6000_rtx_costs): Likewise.
16812         * config/riscv/riscv.h (SLOW_UNALIGNED_ACCESS): Delete.
16813         (riscv_slow_unaligned_access): Likewise.
16814         * config/riscv/riscv.c (riscv_slow_unaligned_access): Rename to...
16815         (riscv_slow_unaligned_access_p): ...this and make static.
16816         (riscv_option_override): Update accordingly.
16817         (riscv_slow_unaligned_access): New function.
16818         (TARGET_SLOW_UNALIGNED_ACCESS): Redefine.
16819         * config/rs6000/rs6000.h (SLOW_UNALIGNED_ACCESS): Delete.
16820         * config/rs6000/rs6000.c (TARGET_SLOW_UNALIGNED_ACCESS): Redefine.
16821         (rs6000_slow_unaligned_access): New function.
16822         (rs6000_emit_move): Use it instead of SLOW_UNALIGNED_ACCESS.
16823         (rs6000_rtx_costs): Likewise.
16824         * config/rs6000/rs6000-string.c (expand_block_compare)
16825         (expand_strn_compare): Use targetm.slow_unaligned_access instead
16826         of SLOW_UNALIGNED_ACCESS.
16827         * config/tilegx/tilegx.h (SLOW_UNALIGNED_ACCESS): Delete.
16828         * config/tilepro/tilepro.h (SLOW_UNALIGNED_ACCESS): Delete.
16829         * calls.c (expand_call): Use targetm.slow_unaligned_access instead
16830         of SLOW_UNALIGNED_ACCESS.
16831         * expmed.c (simple_mem_bitfield_p): Likewise.
16832         * expr.c (alignment_for_piecewise_move): Likewise.
16833         (emit_group_load_1): Likewise.
16834         (emit_group_store): Likewise.
16835         (copy_blkmode_from_reg): Likewise.
16836         (emit_push_insn): Likewise.
16837         (expand_assignment): Likewise.
16838         (store_field): Likewise.
16839         (expand_expr_real_1): Likewise.
16840         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
16841         * lra-constraints.c (simplify_operand_subreg): Likewise.
16842         * stor-layout.c (bit_field_mode_iterator::next_mode): Likewise.
16843         * gimple-ssa-store-merging.c: Likewise in block comment at start
16844         of file.
16845         * tree-ssa-strlen.c: Include target.h.
16846         (handle_builtin_memcmp): Use targetm.slow_unaligned_access instead
16847         of SLOW_UNALIGNED_ACCESS.
16848         * system.h (SLOW_UNALIGNED_ACCESS): Poison.
16850 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16852         PR rtl-optimization/82185
16853         * expmed.c (emit_store_flag_int): Only test tem if it has been
16854         initialized.
16856 2017-09-12  Richard Biener  <rguenther@suse.de>
16858         PR middle-end/82149
16859         * match.pd ((FTYPE) N CMP CST): Fix typo.
16861 2017-09-12  Simon Atanasyan  <simon.atanasyan@imgtec.com>
16863         * config/mips/mips.c (mips_attribute_table): Add 'short_call'
16864         attribute.
16865         (mips_near_type_p): Add 'short_call' attribute as a synonym
16866         for 'near'.
16867         * doc/extend.texi (short_call): Document new function attribute.
16869 2017-09-12  Jakub Jelinek  <jakub@redhat.com>
16871         PR target/82112
16872         * c-common.c (sync_resolve_size): Instead of c_dialect_cxx ()
16873         assertion check that in the condition.
16874         (get_atomic_generic_size): Likewise.  Before testing if parameter
16875         has pointer type, if it has array type, call for C++
16876         default_conversion to perform array-to-pointer conversion.
16878 2017-09-12  Richard Biener  <rguenther@suse.de>
16880         * tree-vect-generic.c (expand_vector_operations_1): Do nothing
16881         for operations we cannot scalarize.
16883 2017-09-12  Aldy Hernandez  <aldyh@redhat.com>
16885         * tree-ssa-threadbackward.c (fsm_find_thread_path): Make GC
16886         vectors heap vectors.  Clean up comments.
16887         Make visited_bbs a reference.
16888         (profitable_jump_thread_path): Make GC
16889         vectors heap vectors.  Clean up comments.
16890         Misc cleanups.
16891         (convert_and_register_jump_thread_path): Make GC vectors heap
16892         vectors.
16893         (check_subpath_and_update_thread_path): Same.  Clean up comments.
16894         Make visited_bbs a reference.
16895         (handle_phi): Abstract common code to to
16896         register_jump_thread_path_if_profitable.
16897         Rename VAR_BB to DEF_BB.
16898         Update comments.
16899         Make GC vectors heap vectors.
16900         Make visited_bbs a reference.
16901         (handle_assignment): Same.
16902         (register_jump_thread_path_if_profitable): New.
16903         (fsm_find_control_statement_thread_paths): Rename VAR_BB to
16904         DEF_BB.
16905         Make GC vectors heap vectors.  Clean up comments.
16906         Make visited_bbs a reference.
16907         (find_jump_threads_backwards): Make visited_bbs live in the stack.
16908         * tree-ssa-threadupdate.c (delete_jump_thread_path): Fix typo in
16909         comment.
16911 2017-09-11  Max Filippov  <jcmvbkbc@gmail.com>
16913         PR target/82181
16914         * config/xtensa/xtensa.c (xtensa_mem_offset): Check that both
16915         words of E_DImode object are reachable by xtensa_uimm8x4 access.
16917 2017-09-11  Vidya Praveen  <vidyapraveen@arm.com>
16919         Revert r251800 and r251799.
16921 2017-09-11  Martin Jambor  <mjambor@suse.cz>
16923         PR hsa/82119
16924         * hsa-gen.c (gen_hsa_phi_from_gimple_phi): Process ADDR_EXPRs in
16925         arguments in advance.
16926         * hsa-regalloc.c (naive_process_phi): New parameter predecessors,
16927         use it to find predecessor edges.
16928         (naive_outof_ssa): Collect vector of predecessors.
16930 2017-09-08  Jason Merrill  <jason@redhat.com>
16932         PR c++/70029 - ICE with ref-qualifier and -flto
16933         * langhooks.h (struct lang_hooks_for_types): Add
16934         copy_lang_qualifiers.
16935         * attribs.c (build_type_attribute_qual_variant): Use it.
16936         * langhooks-def.h (LANG_HOOKS_COPY_LANG_QUALIFIERS): Default to
16937         NULL.
16938         (LANG_HOOKS_FOR_TYPES_INITIALIZER): Use it.
16939         * tree.c (verify_type): Re-enable TYPE_CANONICAL main variant check.
16941 2017-09-08  Eric Botcazou  <ebotcazou@adacore.com>
16943         PR target/81988
16944         * config/sparc/sparc.md (mulsi3): Rename into *mulsi3_sp32.
16945         (*mulsi3_sp64): New instruction.
16946         (mulsi3): New expander.
16948 2017-09-08  Uros Bizjak  <ubizjak@gmail.com>
16950         * config/alpha/alpha.c (alpha_print_operand) <case 'S'>: Remove.
16952 2017-09-08  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
16954         * sancov.c: Include memmodel.h.
16956 2017-09-07  Eric Botcazou  <ebotcazou@adacore.com>
16958         PR target/80897
16959         * config/sparc/sparc.c (sparc_emit_set_symbolic_const64): Deal with too
16960         large offsets.
16962 2017-09-07  Carl Love  <cel@us.ibm.com>
16964         * config/rs6000/vsx.md (define_insn "*stxvl"): Add missing argument to
16965         the sldi instruction.
16967 2017-09-07  David Edelsohn  <dje.gcc@gmail.com>
16969         * sancov.c: Include tm_p.h.
16971 2017-09-07  Jakub Jelinek  <jakub@redhat.com>
16973         PR target/81979
16974         * output.h (switch_to_other_text_partition): New declaration.
16975         * varasm.c (switch_to_other_text_partition): New function.
16976         * config/rs6000/rs6000.c (uses_TOC): Return 2 if
16977         NOTE_INSN_SWITCH_TEXT_SECTIONS is seen before finding load_toc_* insn.
16978         (rs6000_elf_declare_function_name): If uses_TOC returned 2, switch
16979         to the other text partition before emitting LCL label and switch back
16980         after emitting the word after it.
16982 2017-09-07  Richard Biener  <rguenther@suse.de>
16984         * passes.def (pass_split_crit_edges): Remove instance before PRE.
16985         * tree-ssa-pre.c (pass_pre::execute): Instead manually split
16986         critical edges here, after loop init.
16987         (pass_data_pre): Remove PROP_no_crit_edges flags.
16988         * tree-ssa-sccvn.c (vn_reference_lookup_3): Use vn_valueize
16989         for valueization of call args to avoid leaking VN_TOP.
16990         (visit_use): Assert we do not visit default defs.
16991         (init_scc_vn): Use build_decl for VN_TOP to make name nicer.
16992         Use error_mark_node to more easily detect leaking VN_TOP.
16993         All default-defs are varying, not VN_TOP.  Mark them visited.
16994         (run_scc_vn): Make code match comment.
16996 2017-09-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
16998         * config/rs6000/rs6000-cpus.def (OTHER_VSX_VECTOR_MASKS): Delete
16999         OPTION_MASK_FLOAT128_KEYWORD.
17000         (POWERPC_MASKS): Likewise.
17001         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Delete
17002         support for the -mfloat128-type option, and make -mfloat128
17003         default on PowerPC Linux systems.  Define or undefine
17004         __FLOAT128__ and  __FLOAT128_HARDWARE__ for the current options.
17005         Define __float128 to be __ieee128 if IEEE 128-bit support is
17006         enabled, or undefine it.
17007         (rs6000_cpu_cpp_builtins): Delete defining __FLOAT128__ here.
17008         Delete defining __FLOAT128_TYPE__.
17009         * config/rs6000/rs6000.opt (x_TARGET_FLOAT128_TYPE): Delete the
17010         -mfloat128-type option and make -mfloat128 default on PowerPC
17011         Linux systems.
17012         (TARGET_FLOAT128_TYPE): Likewise.
17013         (-mfloat128-type): Likewise.
17014         * config/rs6000/rs6000.c (rs6000_option_override_internal):
17015         Delete the -mfloat128-type option and make -mfloat128 default on
17016         PowerPC Linux systems.  Always use __ieee128 to be the keyword for
17017         the IEEE 128-bit type, and map __float128 to __ieee128 if IEEE
17018         128-bit floating point is enabled.  Change tests from using
17019         -mfloat128-type to -mfloat128.
17020         (rs6000_mangle_type): Use the correct mangling for the __float128
17021         type even if normal long double is restricted to 64-bits.
17022         (floatn_mode): Enable the _Float128 type by default on VSX Linux
17023         systems.
17024         * config/rs6000/rs6000.h (MASK_FLOAT128_TYPE): Delete.
17025         (MASK_FLOAT128_KEYWORD): Define new shortcut macro.
17026         (RS6000BTM_FLOAT128): Define in terms of -mfloat128, not
17027         -mfloat128-type.
17028         * doc/invoke.texi (RS/6000 and PowerPC Options): Update
17029         documentation for -mfloat128.
17031 2017-09-06  Olivier Hainque  <hainque@adacore.com>
17033         * config.gcc (powerpc-wrs-vxworksspe): Now match as vxworks*spe.
17035 2017-09-06  Wish Wu  <wishwu007@gmail.com>
17036             Jakub Jelinek  <jakub@redhat.com>
17038         * asan.c (initialize_sanitizer_builtins): Add
17039         BT_FN_VOID_UINT8_UINT8, BT_FN_VOID_UINT16_UINT16,
17040         BT_FN_VOID_UINT32_UINT32, BT_FN_VOID_UINT64_UINT64,
17041         BT_FN_VOID_FLOAT_FLOAT, BT_FN_VOID_DOUBLE_DOUBLE and
17042         BT_FN_VOID_UINT64_PTR variables.
17043         * builtin-types.def (BT_FN_VOID_UINT8_UINT8): New fn type.
17044         (BT_FN_VOID_UINT16_UINT16): Likewise.
17045         (BT_FN_VOID_UINT32_UINT32): Likewise.
17046         (BT_FN_VOID_FLOAT_FLOAT): Likewise.
17047         (BT_FN_VOID_DOUBLE_DOUBLE): Likewise.
17048         (BT_FN_VOID_UINT64_PTR): Likewise.
17049         * common.opt (flag_sanitize_coverage): New variable.
17050         (fsanitize-coverage=trace-pc): Remove.
17051         (fsanitize-coverage=): Add.
17052         * flag-types.h (enum sanitize_coverage_code): New enum.
17053         * fold-const.c (fold_range_test): Disable non-short-circuit
17054         optimization if flag_sanitize_coverage.
17055         (fold_truth_andor): Likewise.
17056         * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
17057         * opts.c (COVERAGE_SANITIZER_OPT): Define.
17058         (coverage_sanitizer_opts): New array.
17059         (get_closest_sanitizer_option): Add OPTS argument, handle also
17060         OPT_fsanitize_coverage_.
17061         (parse_sanitizer_options): Adjusted to also handle
17062         OPT_fsanitize_coverage_.
17063         (common_handle_option): Add OPT_fsanitize_coverage_.
17064         * sancov.c (instrument_comparison, instrument_switch): New function.
17065         (sancov_pass): Add trace-cmp support.
17066         * sanitizer.def (BUILT_IN_SANITIZER_COV_TRACE_CMP1,
17067         BUILT_IN_SANITIZER_COV_TRACE_CMP2, BUILT_IN_SANITIZER_COV_TRACE_CMP4,
17068         BUILT_IN_SANITIZER_COV_TRACE_CMP8,
17069         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP1,
17070         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP2,
17071         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP4,
17072         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP8,
17073         BUILT_IN_SANITIZER_COV_TRACE_CMPF, BUILT_IN_SANITIZER_COV_TRACE_CMPD,
17074         BUILT_IN_SANITIZER_COV_TRACE_SWITCH): New builtins.
17075         * doc/invoke.texi: Document -fsanitize-coverage=trace-cmp.
17077 2017-09-06  Richard Earnshaw  <rearnsha@arm.com>
17079         * config/arm/parsecpu.awk (fatal): Note that we've encountered an
17080         error.  Only quit immediately if parsing is complete.
17081         (BEGIN): Initialize fatal_err and parse_done.
17082         (begin fpu, end fpu): Check number of arguments.
17083         (begin arch, end arch): Likewise.
17084         (begin cpu, end cpu): Likewise.
17085         (cname, tune for, tune flags, architecture, fpu, option): Likewise.
17086         (optalias): Likewise.
17088 2017-09-06  Richard Earnshaw  <rearnsha@arm.com>
17090         * config.gcc (arm*-*-*): Don't add arm-isa.h to tm_p_file.
17091         * config/arm/arm-isa.h: Delete.  Move definitions to ...
17092         * arm-cpus.in: ... here.  Use new feature and fgroup values.
17093         * config/arm/arm.c (arm_option_override): Use lower case for feature
17094         bit names.
17095         * config/arm/arm.h (TARGET_HARD_FLOAT): Likewise.
17096         (TARGET_VFP3, TARGET_VFP5, TARGET_FMA): Likewise.
17097         * config/arm/parsecpu.awk (END): Add new command 'isa'.
17098         (isa_pfx): Delete.
17099         (print_isa_bits_for): New function.
17100         (gen_isa): New function.
17101         (gen_comm_data): Use print_isa_bits_for.
17102         (define feature): New keyword.
17103         (define fgroup): New keyword.
17104         * config/arm/t-arm (OPTIONS_H_EXTRA): Add arm-isa.h
17105         (arm-isa.h): Add rule to generate file.
17106         * common/config/arm/arm-common.c: (arm_canon_arch_option): Use lower
17107         case for feature bit names.
17109 2017-09-06  Richard Biener  <rguenther@suse.de>
17111         * tree-ssa-pre.c (NECESSARY): Remove.
17112         (create_expression_by_pieces): Do not touch pass-local flags.
17113         (insert_into_preds_of_block): Likewise.
17114         (do_pre_regular_insertion): Likewise.
17115         (eliminate_insert): Likewise.
17116         (eliminate_dom_walker::before_dom_children): Likewise.
17117         (fini_eliminate): Do not look at inserted_exprs.
17118         (mark_operand_necessary): Remove.
17119         (remove_dead_inserted_code): Replace with simple work-list
17120         algorithm based on inserted_exprs and SSA uses.
17121         (pass_pre::execute): Re-order fini_eliminate and
17122         remove_dead_inserted_code.
17124 2017-09-06  Olivier Hainque  <hainque@adacore.com>
17126         * config/powerpcspe/vxworks.h (VXCPU_FOR_8548): Correct definition
17127         for VxWorks 7.  Adjust surrounding comments.
17129 2017-09-06  Richard Biener  <rguenther@suse.de>
17131         * gimple-ssa-strength-reduction.c
17132         (find_candidates_dom_walker::before_dom_children): Also allow
17133         pointer types.
17135 2017-09-06  Richard Biener  <rguenther@suse.de>
17137         PR tree-optimization/82108
17138         * tree-vect-stmts.c (vectorizable_load): Fix pointer adjustment
17139         for gap in the non-permutation SLP case.
17141 2017-09-06  Martin Jambor  <mjambor@suse.cz>
17143         PR tree-optimization/82078
17144         * tree-sra.c (sort_and_splice_var_accesses): Move call to
17145         add_access_to_work_queue...
17146         (build_accesses_from_assign): ...here.
17147         (propagate_all_subaccesses): Make sure racc is the group
17148         representative, if there is one.
17150 2017-09-06  Jakub Jelinek  <jakub@redhat.com>
17152         PR middle-end/82095
17153         * varasm.c (categorize_decl_for_section): Use SECCAT_TBSS for TLS vars with
17154         NULL DECL_INITIAL.
17156 2017-09-06  Richard Biener  <rguenther@suse.de>
17158         * gimple-ssa-strength-reduction.c
17159         (find_candidates_dom_walker::before_doom_children): Use a
17160         type and not a mode check.
17162 2017-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
17164         PR target/77308
17165         * config/arm/predicates.md (arm_general_adddi_operand): Create new
17166         non-vfp predicate.
17167         * config/arm/arm.md (*arm_adddi3, *arm_subdi3): Use new predicates.
17169 2017-09-05  Jeff Law  <law@redhat.com>
17171         PR tree-optimization/64910
17172         * tree-ssa-reassoc.c (reassociate_bb): Restrict last change to
17173         cases where we have 3 or more operands.
17175 2017-09-05  Jakub Jelinek  <jakub@redhat.com>
17177         PR middle-end/81768
17178         * omp-low.c (lower_omp_for): Recompute tree invariant if
17179         gimple_omp_for_initial/final is ADDR_EXPR.
17181         PR middle-end/81768
17182         * omp-expand.c (expand_omp_simd): Force second operands of COND_EXPR
17183         into gimple val before gimplification fo the COND_EXPR.
17185 2017-09-05  Aldy Hernandez  <aldyh@redhat.com>
17187         * tree-ssa-threadupdate.c (duplicate_thread_path): Remove unused
17188         REGION_COPY argument.
17189         (thread_through_all_blocks): Remove unused argument to
17190         duplicate_thread_path.
17192 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17193             Alan Hayward  <alan.hayward@arm.com>
17194             David Sherwood  <david.sherwood@arm.com>
17196         * config/aarch64/aarch64-protos.h (aarch64_gen_adjusted_ldpstp):
17197         Take a scalar_mode rather than a machine_mode.
17198         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
17199         * config/aarch64/aarch64.c (aarch64_simd_container_mode): Likewise.
17200         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
17201         (aarch64_gen_adjusted_ldpstp): Likewise.
17202         (aarch64_expand_vector_init): Use scalar_mode instead of machine_mode.
17204 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17205             Alan Hayward  <alan.hayward@arm.com>
17206             David Sherwood  <david.sherwood@arm.com>
17208         * config/aarch64/aarch64-protos.h (aarch64_is_extend_from_extract):
17209         Take a scalar_int_mode instead of a machine_mode.
17210         (aarch64_mask_and_shift_for_ubfiz_p): Likewise.
17211         (aarch64_output_scalar_simd_mov_immediate): Likewise.
17212         (aarch64_simd_scalar_immediate_valid_for_move): Likewise.
17213         (aarch64_simd_attr_length_rglist): Delete.
17214         * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Take
17215         a scalar_int_mode instead of a machine_mode.
17216         (aarch64_add_offset): Likewise.
17217         (aarch64_internal_mov_immediate): Likewise
17218         (aarch64_add_constant_internal): Likewise.
17219         (aarch64_add_constant): Likewise.
17220         (aarch64_movw_imm): Likewise.
17221         (aarch64_rtx_arith_op_extract_p): Likewise.
17222         (aarch64_mask_and_shift_for_ubfiz_p): Likewise.
17223         (aarch64_simd_scalar_immediate_valid_for_move): Likewise.
17224         Remove assert that the mode isn't a vector.
17225         (aarch64_output_scalar_simd_mov_immediate): Likewise.
17226         (aarch64_expand_mov_immediate): Update calls after above changes.
17227         (aarch64_output_casesi): Use as_a <scalar_int_mode>.
17228         (aarch64_and_bitmask_imm): Check for scalar integer modes.
17229         (aarch64_move_imm): Likewise.
17230         (aarch64_can_const_movi_rtx_p): Likewise.
17231         (aarch64_strip_extend): Likewise.
17232         (aarch64_extr_rtx_p): Likewise.
17233         (aarch64_rtx_costs): Likewise, using wode_mode as the mode of
17234         a CONST_INT when the mode parameter is VOIDmode.
17235         (aarch64_float_const_rtx_p): Use scalar_int_mode for a temporary.
17237 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17239         * machmode.h (bitwise_mode_for_mode): Return opt_mode.
17240         * stor-layout.c (bitwise_mode_for_mode): Likewise.
17241         (bitwise_type_for_mode): Update accordingly.
17243 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17245         * stor-layout.h (mode_for_size_tree): Return an opt_mode.
17246         * stor-layout.c (mode_for_size_tree): Likewise.
17247         (mode_for_array): Update accordingly.
17248         (layout_decl): Likewise.
17249         (compute_record_mode): Likewise.  Only set the mode once.
17251 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17253         * target.def (get_mask_mode): Change return type to opt_mode.
17254         Expand commentary.
17255         * doc/tm.texi: Regenerate.
17256         * targhooks.h (default_get_mask_mode): Return an opt_mode.
17257         * targhooks.c (default_get_mask_mode): Likewise.
17258         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
17259         * optabs-query.c (can_vec_mask_load_store_p): Update use of
17260         targetm.get_mask_mode.
17261         * tree.c (build_truth_vector_type): Likewise.
17263 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17265         * machmode.h (mode_for_vector): Return an opt_mode.
17266         * stor-layout.c (mode_for_vector): Likewise.
17267         (mode_for_int_vector): Update accordingly.
17268         (layout_type): Likewise.
17269         * config/i386/i386.c (emit_memmov): Likewise.
17270         (ix86_expand_set_or_movmem): Likewise.
17271         (ix86_expand_vector_init): Likewise.
17272         (ix86_get_mask_mode): Likewise.
17273         * config/powerpcspe/powerpcspe.c (rs6000_expand_vec_perm_const_1):
17274         Likewise.
17275         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Likewise.
17276         * expmed.c (extract_bit_field_1): Likewise.
17277         * expr.c (expand_expr_real_2): Likewise.
17278         * optabs-query.c (can_vec_perm_p): Likewise.
17279         (can_vec_mask_load_store_p): Likewise.
17280         * optabs.c (expand_vec_perm): Likewise.
17281         * targhooks.c (default_get_mask_mode): Likewise.
17282         * tree-vect-stmts.c (vectorizable_store): Likewise.
17283         (vectorizable_load): Likewise.
17284         (get_vectype_for_scalar_type_and_size): Likewise.
17286 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17288         * machmode.h (mode_for_int_vector): New function.
17289         * stor-layout.c (mode_for_int_vector): Likewise.
17290         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Use it.
17291         * config/powerpcspe/powerpcspe.c (rs6000_do_expand_vec_perm): Likewise.
17292         * config/rs6000/rs6000.c (rs6000_do_expand_vec_perm): Likewise.
17293         * config/s390/s390.c (s390_expand_vec_compare_cc): Likewise.
17294         (s390_expand_vcond): Likewise.
17296 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17298         * machmode.h (opt_machine_mode): New type.
17299         (opt_mode<T>): Allow construction from anything that can be
17300         converted to a T.
17301         (is_a, as_a, dyn_cast): Add overloads for opt_mode.
17302         (mode_for_size): Return an opt_machine_mode.
17303         * stor-layout.c (mode_for_size): Likewise.
17304         (mode_for_size_tree): Update call accordingly.
17305         (bitwise_mode_for_mode): Likewise.
17306         (make_fract_type): Likewise.
17307         (make_accum_type): Likewise.
17308         * caller-save.c (replace_reg_with_saved_mem): Update call
17309         accordingly.
17310         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17311         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17312         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17313         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17314         * expmed.c (extract_bit_field_1): Likewise.
17315         * reload.c (get_secondary_mem): Likewise.
17316         * varasm.c (assemble_integer): Likewise.
17317         * lower-subreg.c (simplify_subreg_concatn): Likewise.  Move
17318         early-out.
17320 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17322         * machmode.h (decimal_float_mode_for_size): New function.
17323         * real.h (REAL_VALUE_TO_TARGET_LONG_DOUBLE): Use float_mode_for_size.
17324         (REAL_VALUE_TO_TARGET_DOUBLE): Likewise.
17325         (REAL_VALUE_TO_TARGET_SINGLE): Likewise.
17326         (REAL_VALUE_TO_TARGET_DECIMAL128): Use decimal_float_mode_for_size.
17327         (REAL_VALUE_TO_TARGET_DECIMAL64): Likewise.
17328         (REAL_VALUE_TO_TARGET_DECIMAL32): Likewise.
17330 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17332         * builtins.c (expand_builtin_powi): Use int_mode_for_size.
17333         (get_builtin_sync_mode): Likewise.
17334         (expand_ifn_atomic_compare_exchange): Likewise.
17335         (expand_builtin_atomic_clear): Likewise.
17336         (expand_builtin_atomic_test_and_set): Likewise.
17337         (fold_builtin_atomic_always_lock_free): Likewise.
17338         * calls.c (compute_argument_addresses): Likewise.
17339         (emit_library_call_value_1): Likewise.
17340         (store_one_arg): Likewise.
17341         * combine.c (combine_instructions): Likewise.
17342         * config/aarch64/aarch64.c (aarch64_function_value): Likewise.
17343         * config/arm/arm.c (arm_function_value): Likewise.
17344         (aapcs_allocate_return_reg): Likewise.
17345         * config/c6x/c6x.c (c6x_expand_movmem): Likewise.
17346         * config/i386/i386.c (construct_container): Likewise.
17347         (ix86_gimplify_va_arg): Likewise.
17348         (ix86_expand_sse_cmp): Likewise.
17349         (emit_memmov): Likewise.
17350         (emit_memset): Likewise.
17351         (expand_small_movmem_or_setmem): Likewise.
17352         (ix86_expand_pextr): Likewise.
17353         (ix86_expand_pinsr): Likewise.
17354         * config/lm32/lm32.c (lm32_block_move_inline): Likewise.
17355         * config/microblaze/microblaze.c (microblaze_block_move_straight):
17356         Likewise.
17357         * config/mips/mips.c (mips_function_value_1) Likewise.
17358         (mips_block_move_straight): Likewise.
17359         (mips_expand_ins_as_unaligned_store): Likewise.
17360         * config/powerpcspe/powerpcspe.c
17361         (rs6000_darwin64_record_arg_advance_flush): Likewise.
17362         (rs6000_darwin64_record_arg_flush): Likewise.
17363         * config/rs6000/rs6000.c
17364         (rs6000_darwin64_record_arg_advance_flush): Likewise.
17365         (rs6000_darwin64_record_arg_flush): Likewise.
17366         * config/sparc/sparc.c (sparc_function_arg_1): Likewise.
17367         (sparc_function_value_1): Likewise.
17368         * config/spu/spu.c (adjust_operand): Likewise.
17369         (spu_emit_branch_or_set): Likewise.
17370         (arith_immediate_p): Likewise.
17371         * emit-rtl.c (gen_lowpart_common): Likewise.
17372         * expr.c (expand_expr_real_1): Likewise.
17373         * function.c (assign_parm_setup_block): Likewise.
17374         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Likewise.
17375         * reload1.c (alter_reg): Likewise.
17376         * stor-layout.c (mode_for_vector): Likewise.
17377         (layout_type): Likewise.
17379 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17381         * config/spu/spu.c (exp2_immediate_p): Use int_mode_for_mode.
17382         (spu_convert_move): Likewise.
17383         * lower-subreg.c (resolve_simple_move): Likewise.
17385 2017-09-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
17387         PR target/81833
17388         * config/rs6000/altivec.md (altivec_vsum2sws): Convert from a
17389         define_insn to a define_expand.
17390         (altivec_vsum2sws_direct): New define_insn.
17391         (altivec_vsumsws): Convert from a define_insn to a define_expand.
17393 2017-09-05  Wilco Dijkstra  <wdijkstr@arm.com>
17395         * config/arm/arm.c (arm_option_params_internal): Improve setting of
17396         max_insns_skipped.
17398 2017-09-05  H.J. Lu  <hongjiu.lu@intel.com>
17400         PR target/59501
17401         PR target/81624
17402         PR target/81769
17403         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Don't
17404         realign stack if stack alignment needed is less than incoming
17405         stack boundary.
17407 2017-09-05  Marek Polacek  <polacek@redhat.com>
17409         PR sanitizer/82072
17410         * convert.c (convert_to_integer_1) <case NEGATE_EXPR>: Move the ubsan
17411         check earlier.
17413 2017-09-05  Wilco Dijkstra  <wdijkstr@arm.com>
17415         * explow.c (get_dynamic_stack_size): Improve dynamic alignment.
17417 2017-09-05  Richard Biener  <rguenther@suse.de>
17419         PR tree-optimization/82084
17420         * fold-const.c (can_native_encode_string_p): Handle wide characters.
17422 2017-09-05  Richard Biener  <rguenther@suse.de>
17424         PR tree-optimization/82102
17425         * tree-ssa-pre.c (fini_eliminate): Check if lhs is NULL.
17427 2017-09-05  Martin Liska  <mliska@suse.cz>
17429         PR tree-optimization/82032
17430         * tree-cfg.c (generate_range_test): New function.
17431         * tree-cfg.h (generate_range_test): Declared here.
17432         * tree-cfgcleanup.c (convert_single_case_switch): New function.
17433         (cleanup_control_expr_graph): Use it.
17434         * tree-switch-conversion.c (try_switch_expansion): Remove
17435         assert.
17436         (emit_case_nodes): Use generate_range_test.
17438 2017-09-04  Uros Bizjak  <ubizjak@gmail.com>
17440         PR target/82098
17441         * config/i386/i386.md (*<btsc><mode>_mask): Add
17442         TARGET_USE_BT to insn constraint.
17443         (*btr<mode>_mask): Ditto.
17445 2017-09-04  Wilco Dijkstra  <wdijkstr@arm.com>
17447         * config/arm/arm.c (arm_legitimate_index_p): Add comment.
17448         (thumb2_legitimate_index_p): Use correct range for DI/DF mode.
17450 2017-09-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
17452         PR target/77308
17453         * config/arm/arm.md (*arm_adddi3, *arm_subdi3): Split early except for
17454         TARGET_NEON and TARGET_IWMMXT.
17455         (anddi3, iordi3, xordi3, one_cmpldi2): Split while expanding except for
17456         TARGET_NEON and TARGET_IWMMXT.
17457         (*one_cmpldi2_insn): Moved the body of one_cmpldi2 here.
17459 2017-09-04  Uros Bizjak  <ubizjak@gmail.com>
17461         * config/i386/i386-protos.h (ix86_tls_address_pattern_p) New prototype.
17462         (ix86_rewrite_tls_address): Ditto.
17463         * config/i386/i386.c (ix86_tls_address_pattern_p) New function.
17464         (ix86_rewrite_tls_address_1): Ditto.
17465         (ix86_rewrite_tls_address): Ditto.
17466         * config/i386/predicates.md (tls_address_pattern): New predicate.
17467         * config/i386/i386.md (TLS address splitter): New splitter.
17469 2017-09-04  Richard Biener  <rguenther@suse.de>
17471         PR tree-optimization/82084
17472         * fold-const.h (can_native_encode_string_p): Declare.
17473         * fold-const.c (can_native_encode_string_p): Factor out from ...
17474         (native_encode_string): ... here.
17475         * tree-vect-stmts.c (vectorizable_store): Call it to avoid
17476         vectorizing stores from constants we later cannot handle.
17478 2017-09-04  Marek Polacek  <polacek@redhat.com>
17480         PR c/81783
17481         * doc/invoke.texi: Update -Wtautological-compare documentation.
17483 2017-09-04  Jeff Law  <law@redhat.com>
17485         PR tree-optimization/64910
17486         * tree-ssa-reassoc.c (reassociate_bb): For bitwise binary ops,
17487         swap the first and last operand if the last is a constant.
17489 2017-09-04  Marek Polacek  <polacek@redhat.com>
17491         PR sanitizer/82072
17492         * convert.c (do_narrow): When sanitizing signed integer overflows,
17493         bail out for signed types.
17494         (convert_to_integer_1) <case NEGATE_EXPR>: Likewise.
17496 2017-09-04  Richard Biener  <rguenther@suse.de>
17498         PR tree-optimization/82060
17499         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
17500         Move devirtualization after stmt folding and before EH/AB/noreturn
17501         cleanup to get the stmt refs canonicalized.  Use a bool instead
17502         of gimple_modified_p since that doesn't work for NOPs.  Schedule
17503         NOPs generated by folding for removal.
17505 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17506             Alan Hayward  <alan.hayward@arm.com>
17507             David Sherwood  <david.sherwood@arm.com>
17509         * coretypes.h (pad_direction): New enum.
17510         * defaults.h (DEFAULT_FUNCTION_ARG_PADDING): Delete.
17511         (FUNCTION_ARG_PADDING): Likewise.
17512         * target.def (function_arg_padding): New hook.
17513         * targhooks.h (default_function_arg_padding): Declare.
17514         * targhooks.c (default_function_arg_padding): New function.
17515         * doc/tm.texi.in (FUNCTION_ARG_PADDING): Replace with...
17516         (TARGET_FUNCTION_ARG_PADDING): ...this.
17517         * doc/tm.texi: Regenerate.
17518         * calls.c (store_unaligned_arguments_into_pseudos): Use pad_direction
17519         instead of direction.
17520         (compute_argument_addresses): Likewise.
17521         (load_register_parameters): Likewise.
17522         (emit_library_call_value_1): Likewise.
17523         (store_one_arg): Use targetm.calls.function_arg_padding instead
17524         of FUNCTION_ARG_PADDING.
17525         (must_pass_in_stack_var_size_or_pad): Likewise.
17526         * expr.c (emit_group_load_1): Use pad_direction instead of direction.
17527         (emit_group_store): Likewise.
17528         (emit_single_push_insn_1): Use targetm.calls.function_arg_padding
17529         instead of FUNCTION_ARG_PADDING.
17530         (emit_push_insn): Likewise, and propagate enum change throughout
17531         function.
17532         * function.h (direction): Delete.
17533         (locate_and_pad_arg_data::where_pad): Use pad_direction instead
17534         of direction.
17535         * function.c (assign_parm_find_stack_rtl): Likewise.
17536         (assign_parm_setup_block_p): Likewise.
17537         (assign_parm_setup_block): Likewise.
17538         (gimplify_parameters): Likewise.
17539         (locate_and_pad_parm): Use targetm.calls.function_arg_padding
17540         instead of FUNCTION_ARG_PADDING, and propagate enum change throughout
17541         function.
17542         * config/aarch64/aarch64.h (FUNCTION_ARG_PADDING): Delete.
17543         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17544         * config/aarch64/aarch64-protos.h (aarch64_pad_arg_upward): Delete.
17545         * config/aarch64/aarch64.c (aarch64_pad_arg_upward): Replace with...
17546         (aarch64_function_arg_padding): ...this new function.
17547         (aarch64_gimplify_va_arg_expr): Use pad_direction instead of direction.
17548         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17549         * config/arm/arm.h (FUNCTION_ARG_PADDING): Delete.
17550         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17551         * config/arm/arm-protos.h (arm_pad_arg_upward): Delete.
17552         * config/arm/arm.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17553         (arm_pad_arg_upward): Replace with...
17554         (arm_function_arg_padding): ...this new function.
17555         * config/c6x/c6x.h (BLOCK_REG_PADDING): Use pad_direction instead
17556         of direction.
17557         * config/ia64/hpux.h (FUNCTION_ARG_PADDING): Delete.
17558         * config/ia64/ia64-protos.h (ia64_hpux_function_arg_padding): Delete.
17559         * config/ia64/ia64.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17560         (ia64_hpux_function_arg_padding): Replace with...
17561         (ia64_function_arg_padding): ...this new function.  Use pad_direction
17562         instead of direction.  Check for TARGET_HPUX.
17563         * config/iq2000/iq2000.h (FUNCTION_ARG_PADDING): Delete.
17564         * config/iq2000/iq2000.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17565         (iq2000_function_arg_padding): New function.
17566         * config/mips/mips-protos.h (mips_pad_arg_upward): Delete.
17567         * config/mips/mips.c (mips_pad_arg_upward): Replace with...
17568         (mips_function_arg_padding): ...this new function.
17569         (mips_pad_reg_upward): Update accordingly.
17570         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17571         * config/mips/mips.h (PAD_VARARGS_DOWN): Use
17572         targetm.calls.function_arg_padding.
17573         (FUNCTION_ARG_PADDING): Delete.
17574         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17575         * config/nios2/nios2.h (FUNCTION_ARG_PADDING): Delete.
17576         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17577         * config/nios2/nios2-protos.h (nios2_function_arg_padding): Delete.
17578         (nios2_block_reg_padding): Return pad_direction instead of direction.
17579         * config/nios2/nios2.c (nios2_block_reg_padding): Return pad_direction
17580         instead of direction.
17581         (nios2_function_arg_padding): Likewise.  Make static.
17582         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17583         * config/pa/pa.h (FUNCTION_ARG_PADDING): Delete.
17584         (BLOCK_REG_PADDING): Use targetm.calls.function_arg_padding.
17585         * config/pa/pa-protos.h (pa_function_arg_padding): Delete.
17586         * config/pa/pa.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17587         (pa_function_arg_padding): Make static.  Return pad_direction instead
17588         of direction.
17589         * config/powerpcspe/powerpcspe.h (FUNCTION_ARG_PADDING): Delete.
17590         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17591         * config/powerpcspe/aix.h (BLOCK_REG_PADDING): Use pad_direction
17592         instead of direction.  Use targetm.calls.function_arg_padding.
17593         * config/powerpcspe/darwin.h (BLOCK_REG_PADDING): Likewise.
17594         * config/powerpcspe/freebsd64.h (BLOCK_REG_PADDING): Likewise.
17595         * config/powerpcspe/linux64.h (BLOCK_REG_PADDING): Likewise.
17596         * config/powerpcspe/powerpcspe-protos.h (function_arg_padding): Delete.
17597         * config/powerpcspe/powerpcspe.c (TARGET_FUNCTION_ARG_PADDING):
17598         Redefine.
17599         (function_arg_padding): Rename to...
17600         (rs6000_function_arg_padding): ...this.  Make static.  Return
17601         pad_direction instead of direction.
17602         (rs6000_return_in_msb): Use rs6000_function_arg_padding.
17603         * config/rs6000/rs6000.h (FUNCTION_ARG_PADDING): Delete.
17604         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17605         * config/rs6000/aix.h (BLOCK_REG_PADDING): Use pad_direction
17606         instead of direction.  Use targetm.calls.function_arg_padding.
17607         * config/rs6000/darwin.h (BLOCK_REG_PADDING): Likewise.
17608         * config/rs6000/freebsd64.h (BLOCK_REG_PADDING): Likewise.
17609         * config/rs6000/linux64.h (BLOCK_REG_PADDING): Likewise.
17610         * config/rs6000/rs6000-protos.h (function_arg_padding): Delete.
17611         * config/rs6000/rs6000.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17612         (function_arg_padding): Rename to...
17613         (rs6000_function_arg_padding): ...this.  Make static.  Return
17614         pad_direction instead of direction.
17615         (rs6000_return_in_msb): Use rs6000_function_arg_padding.
17616         * config/s390/s390.h (FUNCTION_ARG_PADDING): Delete.
17617         * config/s390/s390.c (s390_function_arg_padding): New function.
17618         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17619         * config/sparc/sparc.h (FUNCTION_ARG_PADDING): Delete.
17620         * config/sparc/sparc-protos.h (function_arg_padding): Delete.
17621         * config/sparc/sparc.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17622         (function_arg_padding): Rename to...
17623         (sparc_function_arg_padding): ...this.  Make static.  Return
17624         pad_direction instead of direction.
17625         * config/spu/spu.h (FUNCTION_ARG_PADDING): Delete.
17626         * config/spu/spu.c (spu_function_arg_padding): New function.
17627         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17628         * system.h (FUNCTION_ARG_PADDING): Poison.
17630 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17631             Alan Hayward  <alan.hayward@arm.com>
17632             David Sherwood  <david.sherwood@arm.com>
17634         * target.def (modes_tieable_p): New hook.
17635         * doc/tm.texi (MODES_TIEABLE_P): Replace with...
17636         (TARGET_MODES_TIEABLE_P): ...this.
17637         * doc/tm.texi.in: Regenerate.
17638         * hooks.h (hook_bool_mode_mode_true): Declare.
17639         * hooks.c (hook_bool_mode_mode_true): New function.
17640         * combine.c (subst): Use targetm.modes_tieable_p instead of
17641         MODES_TIEABLE_P.
17642         * dse.c (find_shift_sequence): Likewise.
17643         * expmed.c (extract_low_bits): Likewise.
17644         * lower-subreg.c: Include target.h.
17645         (find_decomposable_subregs): Use targetm.modes_tieable_p instead of
17646         MODES_TIEABLE_P.
17647         * rtlanal.c (rtx_cost): Likewise.
17648         * config/aarch64/aarch64.h (MODES_TIEABLE_P): Delete.
17649         * config/aarch64/aarch64-protos.h (aarch64_modes_tieable_p): Delete.
17650         * config/aarch64/aarch64.c (aarch64_modes_tieable_p): Make static.
17651         (TARGET_MODES_TIEABLE_P): Redefine.
17652         * config/alpha/alpha.h (MODES_TIEABLE_P): Delete.
17653         * config/alpha/alpha.c (alpha_modes_tieable_p): New function.
17654         (TARGET_MODES_TIEABLE_P): Redefine.
17655         * config/arc/arc.h (MODES_TIEABLE_P): Delete.
17656         * config/arc/arc.c (TARGET_MODES_TIEABLE_P): Redefine.
17657         (arc_modes_tieable_p): New function.
17658         * config/arm/arm.h (MODES_TIEABLE_P): Delete.
17659         * config/arm/arm-protos.h (arm_modes_tieable_p): Delete.
17660         * config/arm/arm.c (TARGET_MODES_TIEABLE_P): Redefine.
17661         (arm_modes_tieable_p): Make static.
17662         * config/avr/avr.h (MODES_TIEABLE_P): Delete.
17663         * config/bfin/bfin.h (MODES_TIEABLE_P): Delete.
17664         * config/bfin/bfin.c (bfin_modes_tieable_p): New function.
17665         (TARGET_MODES_TIEABLE_P): Redefine.
17666         * config/c6x/c6x.h (MODES_TIEABLE_P): Delete.
17667         * config/c6x/c6x.c (c6x_modes_tieable_p): New function.
17668         (TARGET_MODES_TIEABLE_P): Redefine.
17669         * config/cr16/cr16.h (MODES_TIEABLE_P): Delete.
17670         * config/cr16/cr16.c (TARGET_MODES_TIEABLE_P): Redefine.
17671         (cr16_modes_tieable_p): New function.
17672         * config/cris/cris.h (MODES_TIEABLE_P): Delete.
17673         * config/epiphany/epiphany.h (MODES_TIEABLE_P): Delete.
17674         * config/fr30/fr30.h (MODES_TIEABLE_P): Delete.
17675         (TRULY_NOOP_TRUNCATION): Update comment.
17676         * config/frv/frv.h (MODES_TIEABLE_P): Delete.
17677         (TRULY_NOOP_TRUNCATION): Update comment.
17678         * config/frv/frv.c (TARGET_MODES_TIEABLE_P): Redefine.
17679         (frv_modes_tieable_p): New function.
17680         * config/ft32/ft32.h (MODES_TIEABLE_P): Delete.
17681         * config/h8300/h8300.h (MODES_TIEABLE_P): Delete.
17682         * config/h8300/h8300.c (h8300_modes_tieable_p): New function.
17683         (TARGET_MODES_TIEABLE_P): Redefine.
17684         * config/i386/i386.h (MODES_TIEABLE_P): Delete.
17685         * config/i386/i386-protos.h (ix86_modes_tieable_p): Delete.
17686         * config/i386/i386.c (ix86_modes_tieable_p): Make static.
17687         (TARGET_MODES_TIEABLE_P): Redefine.
17688         * config/ia64/ia64.h (MODES_TIEABLE_P): Delete.
17689         * config/ia64/ia64.c (TARGET_MODES_TIEABLE_P): Redefine.
17690         (ia64_modes_tieable_p): New function.
17691         * config/iq2000/iq2000.h (MODES_TIEABLE_P): Delete.
17692         * config/iq2000/iq2000.c (TARGET_MODES_TIEABLE_P): Redefine.
17693         (iq2000_modes_tieable_p): New function.
17694         * config/lm32/lm32.h (MODES_TIEABLE_P): Delete.
17695         * config/lm32/lm32.c (TARGET_MODES_TIEABLE_P): Redefine.
17696         (lm32_modes_tieable_p): New function.
17697         * config/m32c/m32c.h (MODES_TIEABLE_P): Delete.
17698         * config/m32c/m32c-protos.h (m32c_modes_tieable_p): Delete.
17699         * config/m32c/m32c.c (m32c_modes_tieable_p): Make static.
17700         (TARGET_MODES_TIEABLE_P): Redefine.
17701         * config/m32r/m32r.h (MODES_TIEABLE_P): Delete.
17702         * config/m32r/m32r.c (TARGET_MODES_TIEABLE_P): Redefine.
17703         (m32r_modes_tieable_p): New function.
17704         * config/m68k/m68k.h (MODES_TIEABLE_P): Delete.
17705         * config/m68k/m68k.c (TARGET_MODES_TIEABLE_P): Redefine.
17706         (m68k_modes_tieable_p): New function.
17707         * config/mcore/mcore.h (MODES_TIEABLE_P): Delete.
17708         * config/mcore/mcore.c (TARGET_MODES_TIEABLE_P): Redefine.
17709         (mcore_modes_tieable_p): New function.
17710         * config/microblaze/microblaze.h (MODES_TIEABLE_P): Delete.
17711         * config/microblaze/microblaze.c (microblaze_modes_tieable_p): New
17712         function.
17713         (TARGET_MODES_TIEABLE_P): Redefine.
17714         * config/mips/mips.h (MODES_TIEABLE_P): Delete.
17715         * config/mips/mips-protos.h (mips_modes_tieable_p): Delete.
17716         * config/mips/mips.c (mips_modes_tieable_p): Make static.
17717         (TARGET_MODES_TIEABLE_P): Redefine.
17718         * config/mmix/mmix.h (MODES_TIEABLE_P): Delete.
17719         * config/mn10300/mn10300.h (MODES_TIEABLE_P): Delete.
17720         * config/mn10300/mn10300-protos.h (mn10300_modes_tieable): Delete.
17721         * config/mn10300/mn10300.c (mn10300_modes_tieable): Rename to...
17722         (mn10300_modes_tieable_p): ...this and make static.
17723         (TARGET_MODES_TIEABLE_P): Redefine.
17724         * config/moxie/moxie.h (MODES_TIEABLE_P): Delete.
17725         * config/msp430/msp430.h (MODES_TIEABLE_P): Delete.
17726         * config/msp430/msp430-protos.h (msp430_modes_tieable_p): Delete.
17727         * config/msp430/msp430.c (TARGET_MODES_TIEABLE_P): Redefine.
17728         (msp430_modes_tieable_p): Make static.
17729         * config/nds32/nds32.h (MODES_TIEABLE_P): Delete.
17730         * config/nds32/nds32.c (nds32_modes_tieable_p): New function.
17731         (TARGET_MODES_TIEABLE_P): Redefine.
17732         * config/nios2/nios2.h (MODES_TIEABLE_P): Delete.
17733         * config/nvptx/nvptx.h (MODES_TIEABLE_P): Delete.
17734         * config/nvptx/nvptx.c (nvptx_modes_tieable_p): New function.
17735         (TARGET_MODES_TIEABLE_P): Redefine.
17736         * config/pa/pa.h (MODES_TIEABLE_P): Delete.
17737         * config/pa/pa-protos.h (pa_modes_tieable_p): Delete.
17738         * config/pa/pa.c (pa_modes_tieable_p): Make static.
17739         (TARGET_MODES_TIEABLE_P): Redefine.
17740         * config/pdp11/pdp11.h (MODES_TIEABLE_P): Delete.
17741         * config/pdp11/pdp11.c (TARGET_MODES_TIEABLE_P): Redefine.
17742         (pdp11_modes_tieable_p): New function.
17743         * config/powerpcspe/powerpcspe.h (MODES_TIEABLE_P): Delete.
17744         * config/powerpcspe/powerpcspe.c (TARGET_MODES_TIEABLE_P): Redefine.
17745         (rs6000_modes_tieable_p): New function.
17746         (rs6000_debug_reg_global): Use it instead of MODES_TIEABLE_P.
17747         * config/powerpcspe/powerpcspe.md: Update comment.
17748         * config/riscv/riscv.h (MODES_TIEABLE_P): Delete.
17749         * config/riscv/riscv.c (riscv_modes_tieable_p): New function.
17750         (TARGET_MODES_TIEABLE_P): Redefine.
17751         * config/rl78/rl78.h (MODES_TIEABLE_P): Delete.
17752         * config/rl78/rl78.c (TARGET_MODES_TIEABLE_P): Redefine.
17753         (rl78_modes_tieable_p): New function.
17754         * config/rs6000/rs6000.h (MODES_TIEABLE_P): Delete.
17755         * config/rs6000/rs6000.c (TARGET_MODES_TIEABLE_P): Redefine.
17756         (rs6000_modes_tieable_p): New function.
17757         (rs6000_debug_reg_global): Use it instead of MODES_TIEABLE_P.
17758         * config/rs6000/rs6000.md: Update comment.
17759         * config/rx/rx.h (MODES_TIEABLE_P): Delete.
17760         * config/rx/rx.c (rx_modes_tieable_p): New function.
17761         (TARGET_MODES_TIEABLE_P): Redefine.
17762         * config/s390/s390.h (MODES_TIEABLE_P): Delete.
17763         * config/s390/s390.c (s390_modes_tieable_p): New function.
17764         (TARGET_MODES_TIEABLE_P): Redefine.
17765         * config/sh/sh.h (MODES_TIEABLE_P): Delete.
17766         * config/sh/sh.c (TARGET_MODES_TIEABLE_P): Redefine.
17767         (sh_modes_tieable_p): New function.
17768         * config/sparc/sparc.h (MODES_TIEABLE_P): Delete.
17769         * config/sparc/sparc-protos.h (sparc_modes_tieable_p): Delete.
17770         * config/sparc/sparc.c (TARGET_MODES_TIEABLE_P): Redefine.
17771         (sparc_modes_tieable_p): Make static.
17772         * config/spu/spu.h (MODES_TIEABLE_P): Delete.
17773         * config/spu/spu.c (spu_modes_tieable_p): New function.
17774         (TARGET_MODES_TIEABLE_P): Redefine.
17775         * config/stormy16/stormy16.h (MODES_TIEABLE_P): Delete.
17776         * config/stormy16/stormy16.c (xstormy16_modes_tieable_p): New function.
17777         (TARGET_MODES_TIEABLE_P): Redefine.
17778         * config/tilegx/tilegx.h (MODES_TIEABLE_P): Delete.
17779         * config/tilepro/tilepro.h (MODES_TIEABLE_P): Delete.
17780         * config/v850/v850.h (MODES_TIEABLE_P): Delete.
17781         * config/v850/v850.c (v850_modes_tieable_p): New function.
17782         (TARGET_MODES_TIEABLE_P): Redefine.
17783         * config/vax/vax.h (MODES_TIEABLE_P): Delete.
17784         * config/visium/visium.h (MODES_TIEABLE_P): Delete.
17785         * config/visium/visium.c (TARGET_MODES_TIEABLE_P): Redefine.
17786         (visium_modes_tieable_p): New function.
17787         * config/xtensa/xtensa.h (MODES_TIEABLE_P): Delete.
17788         * config/xtensa/xtensa.c (TARGET_MODES_TIEABLE_P): Redefine.
17789         (xtensa_modes_tieable_p): New function.
17790         * system.h (MODES_TIEABLE_P): Poison.
17792 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17793             Alan Hayward  <alan.hayward@arm.com>
17794             David Sherwood  <david.sherwood@arm.com>
17796         * target.def (hard_regno_mode_ok): New hook.
17797         * doc/tm.texi (HARD_REGNO_MODE_OK): Replace with...
17798         (TARGET_HARD_REGNO_MODE_OK): ...this.
17799         * doc/tm.texi.in: Regenerate.
17800         * hooks.h (hook_bool_uint_mode_true): Declare.
17801         * hooks.c (hook_bool_uint_mode_true): New function.
17802         * doc/md.texi: Refer to targetm.hard_regno_mode_ok instead of
17803         HARD_REGNO_MODE_OK.
17804         * genpreds.c (write_insn_preds_c): Add an include of target.h.
17805         * alias.c (init_alias_target): Use targetm.hard_regno_mode_ok
17806         instead of HARD_REGNO_MODE_OK.
17807         * caller-save.c: Include target.h.
17808         (reg_save_code): Use targetm.hard_regno_mode_ok instead of
17809         HARD_REGNO_MODE_OK.
17810         * combine.c (can_combine_p): Likewise.
17811         (combinable_i3pat): Likewise.
17812         (can_change_dest_mode): Likewise.
17813         * expr.c (init_expr_target): Likewise.
17814         (convert_move): Likewise.
17815         (convert_modes): Likewise.
17816         * ira.c (setup_prohibited_class_mode_regs): Likewise.
17817         (setup_prohibited_mode_move_regs): Likewise.
17818         * ira.h (target_ira): Likewise.
17819         * lra-assigns.c (find_hard_regno_for_1): Likewise.
17820         * lra-constraints.c (process_alt_operands): Likewise.
17821         (split_reg): Likewise.
17822         * recog.c (peep2_find_free_register): Likewise.
17823         * ree.c (combine_reaching_defs): Likewise.
17824         * regcprop.c (maybe_mode_change): Likewise.
17825         * reginfo.c (init_reg_sets_1): Likewise.
17826         (choose_hard_reg_mode): Likewise.
17827         (simplifiable_subregs): Likewise.
17828         * regrename.c (check_new_reg_p): Likewise.
17829         * reload.c (find_valid_class): Likewise.
17830         (find_valid_class_1): Likewise.
17831         (reload_inner_reg_of_subreg): Likewise.
17832         (push_reload): Likewise.
17833         (combine_reloads): Likewise.
17834         (find_dummy_reload): Likewise.
17835         (find_reloads): Likewise.
17836         * reload1.c (find_reg): Likewise.
17837         (set_reload_reg): Likewise.
17838         (allocate_reload_reg): Likewise.
17839         (choose_reload_regs): Likewise.
17840         (reload_adjust_reg_for_temp): Likewise.
17841         * rtlanal.c (subreg_size_offset_from_lsb): Likewise.
17842         (simplify_subreg_regno): Likewise.
17843         * sel-sched.c (init_regs_for_mode): Likewise.
17844         * varasm.c (make_decl_rtl): Likewise.
17845         * config/aarch64/aarch64.h (HARD_REGNO_MODE_OK): Delete.
17846         (MODES_TIEABLE_P): Use targetm.hard_regno_mode_ok instead of
17847         HARD_REGNO_MODE_OK.
17848         * config/aarch64/aarch64-protos.h (aarch64_hard_regno_mode_ok): Delete.
17849         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Make static.
17850         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17851         * config/alpha/alpha.h (HARD_REGNO_MODE_OK): Delete.
17852         * config/alpha/alpha.c (alpha_hard_regno_mode_ok): New function.
17853         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17854         * config/arc/arc.h (arc_hard_regno_mode_ok): Delete.
17855         (arc_mode_class): Delete.
17856         (HARD_REGNO_MODE_OK): Delete.
17857         * config/arc/arc.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17858         (arc_hard_regno_mode_ok): Rename old array to...
17859         (arc_hard_regno_mode_ok_modes): ...this.
17860         (arc_conditional_register_usage): Update accordingly.
17861         (arc_mode_class): Make static.
17862         (arc_hard_regno_mode_ok): New function.
17863         * config/arm/arm.h (HARD_REGNO_MODE_OK): Delete.
17864         * config/arm/arm-protos.h (arm_hard_regno_mode_ok): Delete.
17865         * config/arm/arm.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17866         (arm_hard_regno_mode_ok): Make static.
17867         * config/arm/arm.md (movdi): Use targetm.hard_regno_mode_ok instead of
17868         HARD_REGNO_MODE_OK.
17869         * config/avr/avr-protos.h (avr_hard_regno_mode_ok): Delete.
17870         * config/avr/avr.h (HARD_REGNO_MODE_OK): Delete.
17871         * config/avr/avr.c (avr_hard_regno_mode_ok): Make static and
17872         return a bool.
17873         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17874         * config/bfin/bfin-protos.h (hard_regno_mode_ok): Delete.
17875         * config/bfin/bfin.h (HARD_REGNO_MODE_OK): Delete.
17876         * config/bfin/bfin.c (hard_regno_mode_ok): Rename to...
17877         (bfin_hard_regno_mode_ok): ...this.  Make static and return a bool.
17878         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17879         * config/bfin/predicates.md (valid_reg_operand): Use
17880         targetm.hard_regno_mode_ok instead of HARD_REGNO_MODE_OK.
17881         * config/c6x/c6x.h (HARD_REGNO_MODE_OK): Delete.
17882         * config/c6x/c6x.c (c6x_hard_regno_mode_ok): New function.
17883         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17884         * config/cr16/cr16.h (HARD_REGNO_MODE_OK): Delete.
17885         * config/cr16/cr16-protos.h (cr16_hard_regno_mode_ok): Delete.
17886         * config/cr16/cr16.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17887         (cr16_hard_regno_mode_ok): Make static and return a bool.
17888         * config/cris/cris.h (HARD_REGNO_MODE_OK): Delete.
17889         * config/cris/cris.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17890         (cris_hard_regno_mode_ok): New function.
17891         * config/epiphany/epiphany.h (epiphany_hard_regno_mode_ok): Delete.
17892         (epiphany_mode_class): Delete.
17893         (HARD_REGNO_MODE_OK): Delete.
17894         * config/epiphany/epiphany-protos.h (hard_regno_mode_ok): Delete.
17895         * config/epiphany/epiphany.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17896         (hard_regno_mode_ok): Rename to...
17897         (epiphany_hard_regno_mode_ok): ...this.  Make static and return a bool.
17898         * config/fr30/fr30.h (HARD_REGNO_MODE_OK): Delete.
17899         * config/fr30/fr30.md: Refer to targetm.hard_regno_mode_ok instead of
17900         HARD_REGNO_MODE_OK.
17901         * config/frv/frv.h (HARD_REGNO_MODE_OK): Delete.
17902         * config/frv/frv-protos.h (frv_hard_regno_mode_ok): Delete.
17903         * config/frv/frv.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17904         (frv_hard_regno_mode_ok): Make static and return a bool.
17905         * config/frv/frv.md: Refer to targetm.hard_regno_mode_ok instead of
17906         HARD_REGNO_MODE_OK.
17907         * config/ft32/ft32.h (HARD_REGNO_MODE_OK): Delete.
17908         * config/h8300/h8300.h (HARD_REGNO_MODE_OK): Delete.
17909         * config/h8300/h8300-protos.h (h8300_hard_regno_mode_ok): Delete.
17910         * config/h8300/h8300.c (h8300_hard_regno_mode_ok): Make static
17911         and return a bool.
17912         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17913         * config/i386/i386.h (HARD_REGNO_MODE_OK): Delete.
17914         * config/i386/i386-protos.h (ix86_hard_regno_mode_ok): Delete.
17915         * config/i386/i386.c (ix86_hard_regno_mode_ok): Make static and
17916         return a bool.
17917         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17918         * config/ia64/ia64.h (HARD_REGNO_MODE_OK): Delete.
17919         * config/ia64/ia64.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17920         (ia64_hard_regno_mode_ok): New function.
17921         * config/iq2000/iq2000.h (HARD_REGNO_MODE_OK): Delete.
17922         * config/iq2000/iq2000.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17923         (iq2000_hard_regno_mode_ok): New function.
17924         * config/lm32/lm32.h (HARD_REGNO_MODE_OK): Delete.
17925         * config/lm32/lm32.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17926         (lm32_hard_regno_mode_ok): New function.
17927         * config/m32c/m32c.h (HARD_REGNO_MODE_OK): Delete.
17928         * config/m32c/m32c-protos.h (m32c_hard_regno_ok): Delete.
17929         * config/m32c/m32c.c (class_can_hold_mode): Use m32c_hard_regno_mode_ok
17930         instead of HARD_REGNO_MODE_OK.
17931         (m32c_hard_regno_ok): Rename to...
17932         (m32c_hard_regno_mode_ok): ...this.  Make static and return a bool.
17933         (m32c_cannot_change_mode_class): Update accordingly.
17934         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17935         * config/m32r/m32r.h (m32r_hard_regno_mode_ok): Delete.
17936         (m32r_mode_class): Delete.
17937         (HARD_REGNO_MODE_OK): Delete.
17938         * config/m32r/m32r.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17939         (m32r_hard_regno_mode_ok): Rename to...
17940         (m32r_hard_regno_modes): ...this.
17941         (m32r_mode_class): Make static.
17942         (m32r_hard_regno_mode_ok): New function.
17943         * config/m68k/m68k.h (HARD_REGNO_MODE_OK): Delete.
17944         * config/m68k/m68k-protos.h (m68k_regno_mode_ok): Delete.
17945         * config/m68k/m68k.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17946         (m68k_hard_regno_mode_ok): Make static.
17947         * config/mcore/mcore.h (HARD_REGNO_MODE_OK): Delete.
17948         * config/mcore/mcore.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17949         (mcore_hard_regno_mode_ok): New function.
17950         * config/microblaze/microblaze.h (microblaze_hard_regno_mode_ok)
17951         (HARD_REGNO_MODE_OK): Delete.
17952         * config/microblaze/microblaze.c (microblaze_hard_regno_mode_ok):
17953         Rename to...
17954         (microblaze_hard_regno_mode_ok_p): ...this and make static.
17955         (microblaze_hard_regno_mode_ok): New function.
17956         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17957         * config/mips/mips.h (HARD_REGNO_MODE_OK): Delete.
17958         (mips_hard_regno_mode_ok): Delete.
17959         * config/mips/mips.c (mips_hard_regno_mode_ok): Rename to...
17960         (mips_hard_regno_mode_ok_p): ...this and make static.
17961         (mips_hard_regno_mode_ok_p): Rename to...
17962         (mips_hard_regno_mode_ok_uncached): ...this.
17963         (mips_hard_regno_mode_ok): New function.
17964         (mips_class_max_nregs): Use mips_hard_regno_mode_ok instead
17965         of HARD_REGNO_MODE_OK.
17966         (mips_option_override): Update after above name changes.
17967         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17968         * config/mmix/mmix.h (HARD_REGNO_MODE_OK): Delete.
17969         * config/mn10300/mn10300.h (HARD_REGNO_MODE_OK): Delete.
17970         * config/mn10300/mn10300-protos.h (mn10300_hard_regno_mode_ok): Delete.
17971         * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Make static.
17972         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17973         * config/moxie/moxie.h (HARD_REGNO_MODE_OK): Delete.
17974         * config/msp430/msp430.h (HARD_REGNO_MODE_OK): Delete.
17975         * config/msp430/msp430-protos.h (msp430_hard_regno_mode_ok): Delete.
17976         * config/msp430/msp430.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17977         (msp430_hard_regno_mode_ok): Make static and return a bool.
17978         * config/nds32/nds32.h (HARD_REGNO_MODE_OK): Delete.
17979         * config/nds32/nds32-protos.h (nds32_hard_regno_mode_ok): Delete.
17980         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Make static
17981         and return a bool.
17982         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17983         * config/nios2/nios2.h (HARD_REGNO_MODE_OK): Delete.
17984         * config/nvptx/nvptx.h (HARD_REGNO_MODE_OK): Delete.
17985         * config/pa/pa.h (MODES_TIEABLE_P): Update commentary.
17986         * config/pa/pa32-regs.h (HARD_REGNO_MODE_OK): Rename to...
17987         (PA_HARD_REGNO_MODE_OK): ...this
17988         * config/pa/pa64-regs.h (HARD_REGNO_MODE_OK): Rename to...
17989         (PA_HARD_REGNO_MODE_OK): ...this.
17990         * config/pa/pa.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17991         (pa_hard_regno_mode_ok): New function.
17992         * config/pdp11/pdp11.h (HARD_REGNO_MODE_OK): Delete.
17993         * config/pdp11/pdp11.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17994         (pdp11_hard_regno_mode_ok): New function.
17995         * config/powerpcspe/powerpcspe.h (HARD_REGNO_MODE_OK): Delete.
17996         * config/powerpcspe/powerpcspe-protos.h (rs6000_hard_regno_mode_ok_p):
17997         Delete.
17998         * config/powerpcspe/powerpcspe.c (rs6000_hard_regno_mode_ok_p):
17999         Make static.
18000         (TARGET_HARD_REGNO_MODE_OK): Redefine.
18001         (rs6000_hard_regno_mode_ok): Rename to...
18002         (rs6000_hard_regno_mode_ok_uncached): ...this.
18003         (rs6000_init_hard_regno_mode_ok): Update accordingly.
18004         (rs6000_hard_regno_mode_ok): New function.
18005         * config/riscv/riscv.h (HARD_REGNO_MODE_OK): Delete.
18006         * config/riscv/riscv-protos.h (riscv_hard_regno_mode_ok_p): Delete.
18007         * config/riscv/riscv.c (riscv_hard_regno_mode_ok_p): Rename to...
18008         (riscv_hard_regno_mode_ok): ...this and make static.
18009         (TARGET_HARD_REGNO_MODE_OK): Redefine.
18010         * config/rl78/rl78.h (HARD_REGNO_MODE_OK): Delete.
18011         * config/rl78/rl78-protos.h (rl78_hard_regno_mode_ok): Delete.
18012         * config/rl78/rl78.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
18013         (rl78_hard_regno_mode_ok): Make static and return bool.
18014         * config/rs6000/rs6000.h (HARD_REGNO_MODE_OK): Delete.
18015         * config/rs6000/rs6000-protos.h (rs6000_hard_regno_mode_ok_p):
18016         Delete.
18017         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok_p): Make static.
18018         (TARGET_HARD_REGNO_MODE_OK): Redefine.
18019         (rs6000_hard_regno_mode_ok): Rename to...
18020         (rs6000_hard_regno_mode_ok_uncached): ...this.
18021         (rs6000_init_hard_regno_mode_ok): Update accordingly.
18022         (rs6000_hard_regno_mode_ok): New function.
18023         * config/rx/rx.h (HARD_REGNO_MODE_OK): Delete.
18024         * config/rx/rx.c (rx_hard_regno_mode_ok): New function.
18025         (TARGET_HARD_REGNO_MODE_OK): Redefine.
18026         * config/s390/s390.h (HARD_REGNO_MODE_OK): Delete.
18027         * config/s390/s390-protos.h (s390_hard_regno_mode_ok): Delete.
18028         * config/s390/s390.c (s390_hard_regno_mode_ok): Make static.
18029         (TARGET_HARD_REGNO_MODE_OK): Redefine.
18030         * config/sh/sh.h (HARD_REGNO_MODE_OK): Delete.
18031         * config/sh/sh-protos.h (sh_hard_regno_mode_ok): Delete.
18032         * config/sh/sh.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
18033         (sh_hard_regno_mode_ok): Make static.
18034         * config/sparc/constraints.md: Refer to targetm.hard_regno_mode_ok
18035         instead of HARD_REGNO_MODE_OK.
18036         * config/sparc/sparc.h (hard_regno_mode_classes): Delete.
18037         (sparc_mode_class): Delete.
18038         (HARD_REGNO_MODE_OK): Delete.
18039         * config/sparc/sparc.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
18040         (hard_regno_mode_classes): Make static.
18041         (sparc_mode_class): Likewise.
18042         (sparc_hard_regno_mode_ok): New function.
18043         * config/spu/spu.h (HARD_REGNO_MODE_OK): Delete.
18044         * config/stormy16/stormy16.h (HARD_REGNO_MODE_OK): Delete.
18045         * config/stormy16/stormy16.c (xstormy16_hard_regno_mode_ok): New
18046         function.
18047         (TARGET_HARD_REGNO_MODE_OK): Redefine.
18048         * config/tilegx/tilegx.h (HARD_REGNO_MODE_OK): Delete.
18049         * config/tilepro/tilepro.h (HARD_REGNO_MODE_OK): Delete.
18050         * config/v850/v850.h (HARD_REGNO_MODE_OK): Delete.
18051         * config/v850/v850.c (v850_hard_regno_mode_ok): New function.
18052         (TARGET_HARD_REGNO_MODE_OK): Redefine.
18053         * config/vax/vax.h (HARD_REGNO_MODE_OK): Delete.
18054         * config/visium/visium.h (HARD_REGNO_MODE_OK): Delete.
18055         * config/visium/visium.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
18056         (visium_hard_regno_mode_ok): New function.
18057         * config/visium/visium.md: Refer to targetm.hard_regno_mode_ok
18058         instead of HARD_REGNO_MODE_OK.
18059         * config/xtensa/xtensa.h (xtensa_hard_regno_mode_ok): Delete.
18060         (HARD_REGNO_MODE_OK): Delete.
18061         * config/xtensa/xtensa.c (xtensa_hard_regno_mode_ok): Rename to...
18062         (xtensa_hard_regno_mode_ok_p): ...this and make static.
18063         (xtensa_option_override): Update accordingly.
18064         (TARGET_HARD_REGNO_MODE_OK): Redefine.
18065         (xtensa_hard_regno_mode_ok): New function.
18066         * system.h (HARD_REGNO_MODE_OK): Poison.
18068 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
18069             Alan Hayward  <alan.hayward@arm.com>
18070             David Sherwood  <david.sherwood@arm.com>
18072         * target.def (hard_regno_call_part_clobbered): New hook.
18073         * doc/tm.texi.in (HARD_REGNO_CALL_PART_CLOBBERED): Replace with...
18074         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): ...this hook.
18075         * doc/tm.texi: Regenerate.
18076         * hooks.h (hook_bool_uint_mode_false): Declare.
18077         * hooks.c (hook_bool_uint_mode_false): New function.
18078         * regs.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18079         * cselib.c (cselib_process_insn): Use
18080         targetm.hard_regno_call_part_clobbered instead of
18081         HARD_REGNO_CALL_PART_CLOBBERED.
18082         * ira-conflicts.c (ira_build_conflicts): Likewise.
18083         * ira-costs.c (ira_tune_allocno_costs): Likewise.
18084         * lra-constraints.c (need_for_call_save_p): Likewise.
18085         * lra-lives.c: Include target.h.
18086         (check_pseudos_live_through_calls): Use
18087         targetm.hard_regno_call_part_clobbered instead of
18088         HARD_REGNO_CALL_PART_CLOBBERED.
18089         * regcprop.c: Include target.h.
18090         (copyprop_hardreg_forward_1): Use
18091         targetm.hard_regno_call_part_clobbered instead of
18092         HARD_REGNO_CALL_PART_CLOBBERED.
18093         * reginfo.c (choose_hard_reg_mode): Likewise.
18094         * regrename.c (check_new_reg_p): Likewise.
18095         * reload.c (find_equiv_reg): Likewise.
18096         * reload1.c (emit_reload_insns): Likewise.
18097         * sched-deps.c (deps_analyze_insn): Likewise.
18098         * sel-sched.c (init_regs_for_mode): Likewise.
18099         (mark_unavailable_hard_regs): Likewise.
18100         * targhooks.c (default_dwarf_frame_reg_mode): Likewise.
18101         * config/aarch64/aarch64.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18102         * config/aarch64/aarch64.c (aarch64_hard_regno_call_part_clobbered):
18103         New function.
18104         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18105         * config/avr/avr.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18106         * config/avr/avr-protos.h (avr_hard_regno_call_part_clobbered):
18107         Delete.
18108         * config/avr/avr.c (avr_hard_regno_call_part_clobbered): Make static
18109         and return a bool.
18110         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18111         * config/i386/i386.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18112         * config/i386/i386.c (ix86_hard_regno_call_part_clobbered): New
18113         function.
18114         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18115         * config/mips/mips.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18116         * config/mips/mips.c (mips_hard_regno_call_part_clobbered): New
18117         function.
18118         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18119         * config/powerpcspe/powerpcspe.h (HARD_REGNO_CALL_PART_CLOBBERED):
18120         Delete.
18121         * config/powerpcspe/powerpcspe.c
18122         (rs6000_hard_regno_call_part_clobbered): New function.
18123         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18124         * config/rs6000/rs6000.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18125         * config/rs6000/rs6000.c (rs6000_hard_regno_call_part_clobbered):
18126         New function.
18127         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18128         * config/s390/s390.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18129         * config/s390/s390.c (s390_hard_regno_call_part_clobbered): New
18130         function.
18131         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18132         * config/sh/sh.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18133         * system.h (HARD_REGNO_CALL_PART_CLOBBERED): Poison.
18135 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
18136             Alan Hayward  <alan.hayward@arm.com>
18137             David Sherwood  <david.sherwood@arm.com>
18139         * rtl.h (subreg_memory_offset): Declare.
18140         * emit-rtl.c (subreg_memory_offset): New function.
18141         * expmed.c (store_bit_field_1): Use it.
18142         * expr.c (undefined_operand_subword_p): Likewise.
18143         * simplify-rtx.c (simplify_subreg): Likewise.
18145 2017-09-04  Alexander Monakov  <amonakov@ispras.ru>
18147         PR rtl-optimization/57448
18148         PR target/67458
18149         PR target/81316
18150         * optabs.c (expand_atomic_load): Place compiler memory barriers if
18151         using atomic_load pattern.
18152         (expand_atomic_store): Likewise.
18154 2017-09-04  Jakub Jelinek  <jakub@redhat.com>
18156         PR sanitizer/81981
18157         * gimple-fold.c (gimple_fold_call): Optimize away useless UBSAN_PTR
18158         and UBSAN_BOUNDS internal calls.  Clean up IFN_UBSAN_OBJECT_SIZE
18159         handling.  Use replace_call_with_value with NULL instead of
18160         gsi_replace, unlink_stmt_vdef and release_defs.
18162         * gdbhooks.py (OptMachineModePrinter.to_string): Use 8 spaces
18163         instead of tab.
18165         * lra-remat.c (reg_overlap_for_remat_p): Fix a pasto.
18167 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
18169         PR bootstrap/82045
18170         * rtl.h (emit_library_call_value_1): Declare.
18171         (emit_library_call): Replace declaration with a series of overloads.
18172         Remove the parameter count argument.
18173         (emit_library_call_value): Likewise.
18174         * calls.c (emit_library_call_value_1): Make global.  Replace varargs
18175         with an "rtx_mode_t *".
18176         (emit_library_call_value): Delete.
18177         (emit_library_call): Likewise.
18178         * asan.c (asan_emit_stack_protection): Update calls accordingly.
18179         (asan_emit_allocas_unpoison): Likewise.
18180         * builtins.c (expand_builtin_powi): Likewise.
18181         (expand_asan_emit_allocas_unpoison): Likewise.
18182         * cfgexpand.c (expand_main_function): Likewise.
18183         * config/aarch64/aarch64.c (aarch64_trampoline_init): Likewise.
18184         * config/aarch64/aarch64.h (PROFILE_HOOK): Likewise.
18185         * config/alpha/alpha.c (alpha_trampoline_init): Likewise.
18186         * config/arm/arm.c (arm_trampoline_init): Likewise.
18187         (arm_call_tls_get_addr): Likewise.
18188         (arm_expand_divmod_libfunc): Likewise.
18189         * config/bfin/bfin.md (umulsi3_highpart): Likewise.
18190         (smulsi3_highpart): Likewise.
18191         * config/c6x/c6x.c (c6x_initialize_trampoline): Likewise.
18192         (c6x_expand_compare): Likewise.
18193         (c6x_expand_movmem): Likewise.
18194         * config/frv/frv.c (frv_trampoline_init): Likewise.
18195         * config/i386/i386.c (ix86_trampoline_init): Likewise.
18196         (ix86_expand_divmod_libfunc): Likewise.
18197         * config/ia64/ia64.c (ia64_expand_tls_address): Likewise.
18198         (ia64_expand_compare): Likewise.
18199         (ia64_profile_hook): Likewise.
18200         * config/ia64/ia64.md (save_stack_nonlocal): Likewise.
18201         (nonlocal_goto): Likewise.
18202         (restore_stack_nonlocal): Likewise.
18203         * config/m32r/m32r.c (block_move_call): Likewise.
18204         (m32r_trampoline_init): Likewise.
18205         * config/m68k/linux.h (FINALIZE_TRAMPOLINE): Likewise.
18206         * config/m68k/m68k.c (m68k_call_tls_get_addr): Likewise.
18207         (m68k_call_m68k_read_tp): Likewise.
18208         * config/microblaze/microblaze.c (microblaze_call_tls_get_addr)
18209         (microblaze_expand_divide): Likewise.
18210         * config/mips/mips.h (mips_args): Likewise.
18211         * config/mips/sdemtk.h (mips_sync_icache): Likewise.
18212         (MIPS_ICACHE_SYNC): Likewise.
18213         * config/nios2/nios2.c (nios2_emit_expensive_div): Likewise.
18214         (nios2_trampoline_init): Likewise.
18215         * config/pa/pa.c (hppa_tls_call): Likewise.
18216         (pa_trampoline_init): Likewise.
18217         * config/pa/pa.md (canonicalize_funcptr_for_compare): Likewise.
18218         * config/powerpcspe/powerpcspe.c (rs6000_legitimize_tls_address)
18219         (expand_strn_compare): Likewise.
18220         (rs6000_generate_compare): Likewise.
18221         (rs6000_expand_float128_convert): Likewise.
18222         (output_profile_hook): Likewise.
18223         (rs6000_trampoline_init): Likewise.
18224         * config/powerpcspe/powerpcspe.md (neg<mode>2): Likewise.
18225         * config/riscv/riscv.h (PROFILE_HOOK): Likewise.
18226         * config/rs6000/rs6000-string.c (expand_strn_compare): Likewise.
18227         * config/rs6000/rs6000.c (rs6000_legitimize_tls_address): Likewise.
18228         (rs6000_generate_compare): Likewise.
18229         (rs6000_expand_float128_convert): Likewise.
18230         (output_profile_hook): Likewise.
18231         (rs6000_trampoline_init): Likewise.
18232         * config/rs6000/rs6000.md (neg<mode>2): Likewise.
18233         * config/sh/sh.c (sh_trampoline_init): Likewise.
18234         * config/sparc/sparc.c (emit_soft_tfmode_libcall): Likewise.
18235         (sparc_emit_float_lib_cmp): Likewise.
18236         (sparc32_initialize_trampoline): Likewise.
18237         (sparc64_initialize_trampoline): Likewise.
18238         (sparc_profile_hook): Likewise.
18239         * config/spu/spu.c (ea_load_store): Likewise.
18240         * config/spu/spu.md (floatunssidf2): Likewise.
18241         * config/tilegx/tilegx.c (tilegx_trampoline_init): Likewise.
18242         * config/tilepro/tilepro.c (tilepro_trampoline_init): Likewise.
18243         * config/visium/visium.c (expand_block_move_4): Likewise.
18244         (expand_block_move_2): Likewise.
18245         (expand_block_move_1): Likewise.
18246         (expand_block_set_4): Likewise.
18247         (expand_block_set_2): Likewise.
18248         (expand_block_set_1): Likewise.
18249         (visium_trampoline_init): Likewise.
18250         (visium_profile_hook): Likewise.
18251         * config/xtensa/xtensa.c (xtensa_expand_nonlocal_goto): Likewise.
18252         (xtensa_setup_frame_addresses): Likewise.
18253         (xtensa_trampoline_init): Likewise.
18254         * except.c (sjlj_emit_function_enter): Likewise.
18255         (sjlj_emit_function_exit): Likewise.
18256         * explow.c (allocate_dynamic_stack_space): Likewise.
18257         (probe_stack_range): Likewise.
18258         * expr.c (convert_mode_scalar): Likewise.
18259         * optabs.c (expand_binop): Likewise.
18260         (expand_twoval_binop_libfunc): Likewise.
18261         (expand_unop): Likewise.
18262         (prepare_cmp_insn): Likewise.
18263         (prepare_float_lib_cmp): Likewise.
18264         (expand_float): Likewise.
18265         (expand_fix): Likewise.
18266         (expand_fixed_convert): Likewise.
18267         (maybe_emit_sync_lock_test_and_set): Likewise.
18268         (expand_atomic_compare_and_swap): Likewise.
18269         (expand_mem_thread_fence): Likewise.
18270         (expand_atomic_fetch_op): Likewise.
18272 2017-09-03  Gerald Pfeifer  <gerald@pfeifer.com>
18274         * doc/generic.texi (OpenACC): Adjust URL.
18275         * doc/invoke.texi (C Dialect Options): Ditto.
18277 2017-09-03  Uros Bizjak  <ubizjak@gmail.com>
18279         * config/i386/i386.md (*bt<mode>): Use nonimmediate_operand
18280         predicate for operand 1.  Add (m,<S>) constraint.
18281         (*jcc_bt<mode>): Use nonimmediate_operand predicate for operand 1.
18282         Prevent memory operand 1 with register operand 2.
18284 2017-09-01  Segher Boessenkool  <segher@kernel.crashing.org>
18286         PR rtl-optimization/82024
18287         * combine.c (try_combine): If the combination result is a PARALLEL,
18288         and we only need to retain the SET in there that would be placed
18289         at I2, check that we can place that at I3 instead, before doing so.
18291 2017-09-01  Jakub Jelinek  <jakub@redhat.com>
18293         PR target/81766
18294         * config/i386/i386.c (ix86_init_large_pic_reg): Return label
18295         instead of void.
18296         (ix86_init_pic_reg): Remember label from ix86_init_large_pic_reg,
18297         if non-NULL and preceded by NOTE_INSN_BASIC_BLOCK, swap the note
18298         and label.
18300 2017-09-01  Joerg Sonnenberger  <joerg@bec.de>
18301             Jeff Law  <law@redhat.com>
18303         * varasm.c (bss_initializer_p): Do not put constants into .bss
18304         (categorize_decl_for_section): Handle bss_initializer_p returning
18305         false when DECL_INITIAL is NULL.
18307 2017-09-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
18309         PR target/82012
18310         * config/s390/s390.c (s390_can_inline_p): New function.
18312 2017-09-01  Jeff Law  <law@redhat.com>
18314         PR tree-optimization/82052
18315         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
18316         Always initialize the returned slot after a hash table miss
18317         when INSERT is true.
18319 2017-09-01  Alexander Monakov  <amonakov@ispras.ru>
18321         * config/s390/s390.md (mem_signal_fence): Remove.
18322         * doc/md.texi (mem_signal_fence): Remove.
18323         * optabs.c (expand_mem_signal_fence): Remove uses of mem_signal_fence.
18324         Update comments.
18325         * target-insns.def (mem_signal_fence): Remove.
18327 2017-09-01  Jakub Jelinek  <jakub@redhat.com>
18329         PR sanitizer/81902
18330         * doc/invoke.texi: Document -fsanitize=pointer-overflow.
18332         PR sanitizer/81923
18333         * asan.c (create_odr_indicator): Strip name encoding from assembler
18334         name before appending it after __odr_asan_.
18336 2017-09-01  Martin Liska  <mliska@suse.cz>
18338         PR tree-optimization/82059
18339         * gimple-ssa-isolate-paths.c (isolate_path): Add profile and
18340         frequency only when an edge is redirected.
18342 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18344         * config/arc/arc-c.c (__ARC_LPC_WIDTH__): Add builtin define.
18345         * config/arc/arc.c (ARC_MAX_LOOP_LENGTH): Define.
18346         (arc_conditional_register_usage): Remove ARC600 lp_count
18347         exception.
18348         (arc_file_start): Emit Tag_ARC_CPU_variation.
18349         (arc_can_use_doloop_p): New conditions to use ZOLs.
18350         (hwloop_fail): New function.
18351         (hwloop_optimize): Likewise.
18352         (hwloop_pattern_reg): Likewise.
18353         (arc_doloop_hooks): New struct, to be used with reorg_loops.
18354         (arc_reorg_loops): New function, calls reorg_loops.
18355         (arc_reorg): Call arc_reorg_loops.  Remove old ZOL handling.
18356         (arc600_corereg_hazard): Remove ZOL checking, case handled by
18357         hwloop_optimize.
18358         (arc_loop_hazard): Remove function, functionality moved into
18359         hwloop_optimize.
18360         (arc_hazard): Remove arc_loop_hazard call.
18361         (arc_adjust_insn_length): Remove ZOL handling, functionality moved
18362         into hwloop_optimize.
18363         (arc_label_align): Remove ZOL handling.
18364         * config/arc/arc.h (LOOP_ALIGN): Changed to 0.
18365         * config/arc/arc.md (doloop_begin): Remove pattern.
18366         (doloop_begin_i): Likewise.
18367         (doloop_end_i): Likewise.
18368         (doloop_fallback): Likewise.
18369         (doloop_fallback_m): Likewise.
18370         (doloop_end): Reimplement expand.
18371         (arc_lp): New pattern for LP instruction.
18372         (loop_end): New pattern.
18373         (loop_fail): Likewise.
18374         (decrement_and_branch_until_zero): Likewise.
18375         * config/arc/arc.opt (mlpc-width): New option.
18376         * doc/invoke.texi (mlpc-width): Document option.
18378 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18380         * config/arc/arc.c (arc_ifcvt): Remove use of merge_blocks call.
18381         (arc_ccfsm_advance): Fix checking for delay slots.
18382         (arc_reorg): Add rtl dump after each call to arc_ifcvt.
18384 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18386         * config/arc/arc.md (movqi_insn): Add stores to save constant long
18387         immediates.
18388         (movhi_insn): Update store instruction constraint which are saving
18389         6-bit short immediates.
18390         (movsi_insn): Consider also short scaled load operations.
18391         (zero_extendhisi2_i): Use Usd constraint instead of T.
18392         (extendhisi2_i): Add q constraint.
18393         (arc_clzsi2): Add type and length attributes.
18394         (arc_ctzsi2): Likewise.
18395         * config/arc/constraints.md (Usc): Update constraint, the
18396         assembler can parse two relocations for a single instruction.
18398 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18400         * config/arc/arc.c (arc_use_anchors_for_symbol_p): New function.
18401         (TARGET_USE_ANCHORS_FOR_SYMBOL_P): Define.
18403 2017-08-31  Olivier Hainque  <hainque@adacore.com>
18405         * config.gcc (powerpc-wrs-vxworks|vxworksae|vxworksmils): Now
18406         match as powerpc-wrs-vxworks*.
18408 2017-08-31  James Greenhalgh  <james.greenhalgh@arm.com>
18410         * config/aarch64/aarch64-simd.md (aarch64_mla_elt_merge<mode>): Fix
18411         register constraint for by-element operand.
18412         (aarch64_mls_elt_merge<mode>): Likewise.
18414 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18416         * config/arc/arc.c (arc_can_follow_jump): Check for short
18417         branches.
18419 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18421         * config.gcc: Use g.opt for arc.
18422         * config/arc/arc.c (LEGITIMATE_SCALED_ADDRESS_P): Deleted,
18423         functionality moved to ...
18424         (legitimate_scaled_address_p): New function, ...here.
18425         (LEGITIMATE_SMALL_DATA_OFFSET_P): New define.
18426         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Use the above define.
18427         (legitimate_offset_address_p): Delete TARGET_NO_SDATA_SET
18428         condition.
18429         (arc_override_options): Handle G option.
18430         (arc_output_pic_addr_const): Correct function definition.
18431         (arc_legitimate_address_p): Use legitimate_scaled_address_p.
18432         (arc_decl_anon_ns_mem_p): Delete.
18433         (arc_in_small_data_p): Overhaul this function to take into
18434         consideration the value given via G option.
18435         (arc_rewrite_small_data_1): Renamed and corrected old
18436         arc_rewrite_small_data function.
18437         (arc_rewrite_small_data): New function.
18438         (small_data_pattern): Don't use pic_offset_table_rtx.
18439         * config/arc/arc.h (CC1_SPEC): Recognize G option.
18440         * config/arc/simdext.md (movmisalignv2hi): Use
18441         prepare_move_operands function.
18442         (mov*): Likewise.
18443         (movmisalign*): Likewise.
18444         * doc/invoke.texi (ARC options): Document -G option.
18446 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18448         * config/arc/arc-protos.h (compact_sda_memory_operand): Update
18449         prototype.
18450         * config/arc/arc.c (arc_print_operand): Output scalled address for
18451         sdata whenever is possible.
18452         (arc_in_small_data_p): Allow sdata for 64bit datum when double
18453         load/stores are available.
18454         (compact_sda_memory_operand): Check for the alignment required by
18455         code density instructions.
18456         * config/arc/arc.md (movsi_insn): Use newly introduced Us0
18457         constraint.
18458         * config/arc/constraints.md (Usd): Update constraint.
18459         (Us0): New constraint.
18460         (Usc): Update constraint.
18462 2017-08-31  Richard Biener  <rguenther@suse.de>
18464         PR middle-end/82054
18465         * dwarf2out.c (dwarf2out_early_global_decl): Process each
18466         function only once.
18468 2017-08-31  Tamar Christina  <tamar.christina@arm.com>
18470         * config/aarch64/aarch64-builtins.c (aarch64_init_simd_builtins):
18471         Resize type_signature.
18473 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18474             Alan Hayward  <alan.hayward@arm.com>
18475             David Sherwood  <david.sherwood@arm.com>
18477         * config/aarch64/aarch64.c (aarch64_base_register_rtx_p): Only allow
18478         subregs whose inner modes can be stored in GPRs.
18479         (aarch64_classify_index): Likewise.
18481 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18482             Alan Hayward  <alan.hayward@arm.com>
18483             David Sherwood  <david.sherwood@arm.com>
18485         * config/aarch64/iterators.md (V_cmp_result): Rename to...
18486         (V_INT_EQUIV): ...this.
18487         (v_cmp_result): Rename to...
18488         (v_int_equiv): ...this.
18489         * config/aarch64/aarch64.md (xorsign<mode>3): Update accordingly.
18490         * config/aarch64/aarch64-simd.md (xorsign<mode>3): Likewise.
18491         (copysign<mode>3): Likewise.
18492         (aarch64_simd_bsl<mode>_internal): Likewise.
18493         (aarch64_simd_bsl<mode>): Likewise.
18494         (vec_cmp<mode><mode>): Likewise.
18495         (vcond<mode><mode>): Likewise.
18496         (vcond<v_cmp_mixed><mode>): Likewise.
18497         (vcondu<mode><v_cmp_mixed>): Likewise.
18498         (aarch64_cm<optab><mode>): Likewise.
18499         (aarch64_cmtst<mode>): Likewise.
18500         (aarch64_fac<optab><mode>): Likewise.
18501         (vec_perm_const<mode>): Likewise.
18502         (vcond_mask_<mode><v_cmp_result>): Rename to...
18503         (vcond_mask_<mode><v_int_equiv>): ...this.
18504         (vec_cmp<mode><v_cmp_result>): Rename to...
18505         (vec_cmp<mode><v_int_equiv>): ...this.
18507 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18508             Alan Hayward  <alan.hayward@arm.com>
18509             David Sherwood  <david.sherwood@arm.com>
18511         * config/aarch64/aarch64-modes.def: Remove 32-, 48- and 64-byte
18512         vector modes.
18513         * config/aarch64/iterators.md (VRL2, VRL3, VRL4): Delete.
18514         * config/aarch64/aarch64.md (UNSPEC_LD2_DREG, UNSPEC_LD3_DREG)
18515         (UNSPEC_LD4_DREG): New unspecs.
18516         * config/aarch64/aarch64-simd.md (aarch64_ld2<mode>_dreg_le)
18517         (aarch64_ld2<mode>_dreg_be): Replace with...
18518         (aarch64_ld2<mode>_dreg): ...this pattern and use the new DREG
18519         unspec.
18520         (aarch64_ld3<mode>_dreg_le)
18521         (aarch64_ld3<mode>_dreg_be): Replace with...
18522         (aarch64_ld3<mode>_dreg): ...this pattern and use the new DREG
18523         unspec.
18524         (aarch64_ld4<mode>_dreg_le)
18525         (aarch64_ld4<mode>_dreg_be): Replace with...
18526         (aarch64_ld4<mode>_dreg): ...this pattern and use the new DREG
18527         unspec.
18529 2017-08-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
18531         PR tree-optimization/81987
18532         * gimple-ssa-strength-reduction.c (insert_initializers): Don't
18533         insert an initializer in a location not dominated by the stride
18534         definition.
18536 2017-08-30  Eric Botcazou  <ebotcazou@adacore.com>
18538         * tree-eh.c (lower_try_finally_switch): Set the location of the finally
18539         on the entire header of the finally block in the fallthru case.
18541 2017-08-30  Eric Botcazou  <ebotcazou@adacore.com>
18543         * varasm.c (decode_addr_const): Deal with INDIRECT_REF <INTEGER_CST>.
18545 2017-08-30  Pat Haugen  <pthaugen@us.ibm.com>
18547         * config/rs6000/rs6000.c (rs6000_emit_prologue_move_from_cr): Rename from
18548         rs6000_emit_move_from_cr and call renamed function.
18549         (rs6000_emit_prologue): Call renamed functions.
18550         * config/rs6000/rs6000.md (prologue_movesi_from_cr): Rename from
18551         movesi_from_cr, remove volatile CRs.
18553 2017-08-30  Jon Beniston  <jon@beniston.com>
18554             Richard Biener  <rguenther@suse.de>
18556         * tree-vect-patterns.c (vect_pattern_recog_1): Use VECTOR_TYPE_P instead
18557         of VECTOR_MODE_P check.
18558         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Allow single
18559         element vector types.
18561 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18563         * df.h (df_read_modify_subreg_p): Remove in favor of...
18564         * rtl.h (read_modify_subreg_p): ...this new function.  Take a
18565         const_rtx instead of an rtx.
18566         * cprop.c (local_cprop_find_used_regs): Update accordingly.
18567         * df-problems.c (df_word_lr_mark_ref): Likewise.
18568         * ira-lives.c (mark_pseudo_reg_live): Likewise.
18569         (mark_pseudo_reg_dead): Likewise.
18570         (mark_ref_dead): Likewise.
18571         * reginfo.c (init_subregs_of_mode): Likewise.
18572         * sched-deps.c (sched_analyze_1): Likewise.
18573         * df-scan.c (df_def_record_1): Likewise.
18574         (df_uses_record): Likewise.
18575         (df_read_modify_subreg_p): Remove in favor of...
18576         * rtlanal.c (read_modify_subreg_p): ...this new function.  Take a
18577         const_rtx instead of an rtx.
18579 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18580             Alan Hayward  <alan.hayward@arm.com>
18581             David Sherwood  <david.sherwood@arm.com>
18583         * rtl.h (partial_subreg_p): New function.
18584         * caller-save.c (save_call_clobbered_regs): Use it.
18585         * calls.c (expand_call): Likewise.
18586         * combine.c (combinable_i3pat): Likewise.
18587         (simplify_set): Likewise.
18588         (make_extraction): Likewise.
18589         (make_compound_operation_int): Likewise.
18590         (gen_lowpart_or_truncate): Likewise.
18591         (force_to_mode): Likewise.
18592         (make_field_assignment): Likewise.
18593         (reg_truncated_to_mode): Likewise.
18594         (record_truncated_value): Likewise.
18595         (move_deaths): Likewise.
18596         * cse.c (record_jump_cond): Likewise.
18597         (cse_insn): Likewise.
18598         * cselib.c (cselib_lookup_1): Likewise.
18599         * expmed.c (extract_bit_field_using_extv): Likewise.
18600         * function.c (assign_parm_setup_reg): Likewise.
18601         * ifcvt.c (noce_convert_multiple_sets): Likewise.
18602         * ira-build.c (create_insn_allocnos): Likewise.
18603         * lra-coalesce.c (merge_pseudos): Likewise.
18604         * lra-constraints.c (match_reload): Likewise.
18605         (simplify_operand_subreg): Likewise.
18606         (curr_insn_transform): Likewise.
18607         * lra-lives.c (process_bb_lives): Likewise.
18608         * lra.c (new_insn_reg): Likewise.
18609         (lra_substitute_pseudo): Likewise.
18610         * regcprop.c (mode_change_ok): Likewise.
18611         (maybe_mode_change): Likewise.
18612         (copyprop_hardreg_forward_1): Likewise.
18613         * reload.c (push_reload): Likewise.
18614         (find_reloads): Likewise.
18615         (find_reloads_subreg_address): Likewise.
18616         * reload1.c (alter_reg): Likewise.
18617         (eliminate_regs_1): Likewise.
18618         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
18620 2017-08-30  David Edelsohn  <dje.gcc@gmail.com>
18622         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Revert
18623         back to if statements, including unpack.
18625 2017-08-30  Martin Liska  <mliska@suse.cz>
18627         PR inline-asm/82001
18628         * ipa-icf-gimple.c (func_checker::compare_tree_list_operand):
18629         Rename to ...
18630         (func_checker::compare_asm_inputs_outputs): ... this function.
18631         (func_checker::compare_gimple_asm): Use the function to compare
18632         also ASM constrains.
18633         * ipa-icf-gimple.h: Rename the function.
18635 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18636             Alan Hayward  <alan.hayward@arm.com>
18637             David Sherwood  <david.sherwood@arm.com>
18639         * coretypes.h (complex_mode): New type.
18640         * gdbhooks.py (build_pretty_printer): Handle it.
18641         * machmode.h (complex_mode): New class.
18642         (complex_mode::includes_p): New function.
18643         (is_complex_int_mode): Likewise.
18644         (is_complex_float_mode): Likewise.
18645         * genmodes.c (get_mode_class): Handle complex mode classes.
18646         * function.c (expand_function_end): Use is_complex_int_mode.
18648 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18649             Alan Hayward  <alan.hayward@arm.com>
18650             David Sherwood  <david.sherwood@arm.com>
18652         * coretypes.h (scalar_mode_pod): New typedef.
18653         * gdbhooks.py (build_pretty_printer): Handle it.
18654         * machmode.h (gt_ggc_mx, gt_pch_nx): New functions.
18655         * fixed-value.h (fixed_value::mode): Change type to scalar_mode_pod.
18656         * fold-const.c (fold_convert_const_int_from_fixed): Use scalar_mode.
18657         * tree-streamer-in.c (unpack_ts_fixed_cst_value_fields): Use
18658         as_a <scalar_mode>.
18660 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18661             Alan Hayward  <alan.hayward@arm.com>
18662             David Sherwood  <david.sherwood@arm.com>
18664         * machmode.h (mode_for_vector): Take a scalar_mode instead
18665         of a machine_mode.
18666         * stor-layout.c (mode_for_vector): Likewise.
18667         * explow.c (promote_mode): Use as_a <scalar_mode>.
18668         * sdbout.c (sdbout_parms): Use is_a <scalar_mode>.
18670 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18671             Alan Hayward  <alan.hayward@arm.com>
18672             David Sherwood  <david.sherwood@arm.com>
18674         * target.def (preferred_simd_mode): Take a scalar_mode
18675         instead of a machine_mode.
18676         * targhooks.h (default_preferred_simd_mode): Likewise.
18677         * targhooks.c (default_preferred_simd_mode): Likewise.
18678         * config/arc/arc.c (arc_preferred_simd_mode): Likewise.
18679         * config/arm/arm.c (arm_preferred_simd_mode): Likewise.
18680         * config/c6x/c6x.c (c6x_preferred_simd_mode): Likewise.
18681         * config/epiphany/epiphany.c (epiphany_preferred_simd_mode): Likewise.
18682         * config/i386/i386.c (ix86_preferred_simd_mode): Likewise.
18683         * config/mips/mips.c (mips_preferred_simd_mode): Likewise.
18684         * config/nvptx/nvptx.c (nvptx_preferred_simd_mode): Likewise.
18685         * config/powerpcspe/powerpcspe.c (rs6000_preferred_simd_mode):
18686         Likewise.
18687         * config/rs6000/rs6000.c (rs6000_preferred_simd_mode): Likewise.
18688         * config/s390/s390.c (s390_preferred_simd_mode): Likewise.
18689         * config/sparc/sparc.c (sparc_preferred_simd_mode): Likewise.
18690         * config/aarch64/aarch64.c (aarch64_preferred_simd_mode): Likewise.
18691         (aarch64_simd_scalar_immediate_valid_for_move): Update accordingly.
18692         * doc/tm.texi: Regenerate.
18693         * optabs-query.c (can_vec_mask_load_store_p): Return false for
18694         non-scalar modes.
18696 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18697             Alan Hayward  <alan.hayward@arm.com>
18698             David Sherwood  <david.sherwood@arm.com>
18700         * target.def (scalar_mode_supported_p): Take a scalar_mode
18701         instead of a machine_mode.
18702         * targhooks.h (default_scalar_mode_supported_p): Likewise.
18703         * targhooks.c (default_scalar_mode_supported_p): Likewise.
18704         * config/aarch64/aarch64.c (aarch64_scalar_mode_supported_p): Likewise.
18705         * config/alpha/alpha.c (alpha_scalar_mode_supported_p): Likewise.
18706         * config/arm/arm.c (arm_scalar_mode_supported_p): Likewise.
18707         * config/avr/avr.c (avr_scalar_mode_supported_p): Likewise.
18708         * config/c6x/c6x.c (c6x_scalar_mode_supported_p): Likewise.
18709         * config/i386/i386.c (ix86_scalar_mode_supported_p): Likewise.
18710         * config/ia64/ia64.c (ia64_scalar_mode_supported_p): Likewise.
18711         * config/mips/mips.c (mips_scalar_mode_supported_p): Likewise.
18712         * config/msp430/msp430.c (msp430_scalar_mode_supported_p): Likewise.
18713         * config/pa/pa.c (pa_scalar_mode_supported_p): Likewise.
18714         * config/pdp11/pdp11.c (pdp11_scalar_mode_supported_p): Likewise.
18715         * config/powerpcspe/powerpcspe.c (rs6000_scalar_mode_supported_p):
18716         Likewise.
18717         * config/rs6000/rs6000.c (rs6000_scalar_mode_supported_p): Likewise.
18718         * config/s390/s390.c (s390_scalar_mode_supported_p): Likewise.
18719         * config/spu/spu.c (spu_scalar_mode_supported_p): Likewise.
18720         * config/tilegx/tilegx.c (tilegx_scalar_mode_supported_p): Likewise.
18721         * config/tilepro/tilepro.c (tilepro_scalar_mode_supported_p):
18722         Likewise.
18723         * doc/tm.texi: Regenerate.
18725 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18726             Alan Hayward  <alan.hayward@arm.com>
18727             David Sherwood  <david.sherwood@arm.com>
18729         * coretypes.h (opt_scalar_mode): New typedef.
18730         * gdbhooks.py (build_pretty_printers): Handle it.
18731         * machmode.h (mode_iterator::get_2xwider): Add overload for
18732         opt_mode<T>.
18733         * emit-rtl.c (init_emit_once): Use opt_scalar_mode when iterating
18734         over scalar modes.
18735         * expr.c (convert_mode_scalar): Likewise.
18736         * omp-low.c (omp_clause_aligned_alignment): Likewise.
18737         * optabs.c (expand_float): Likewise.
18738         (expand_fix): Likewise.
18739         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
18741 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18742             Alan Hayward  <alan.hayward@arm.com>
18743             David Sherwood  <david.sherwood@arm.com>
18745         * optabs.c (expand_float): Explicitly check for scalars before
18746         using a branching expansion.
18747         (expand_fix): Likewise.
18749 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18750             Alan Hayward  <alan.hayward@arm.com>
18751             David Sherwood  <david.sherwood@arm.com>
18753         * expr.c (convert_mode): Split scalar handling out into...
18754         (convert_mode_scalar): ...this new function.  Treat the modes
18755         as scalar_modes.
18757 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18758             Alan Hayward  <alan.hayward@arm.com>
18759             David Sherwood  <david.sherwood@arm.com>
18761         * omp-expand.c (expand_omp_atomic): Use is_int_mode, is_float_mode
18762         and scalar_mode.
18763         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Likewise.
18765 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18766             Alan Hayward  <alan.hayward@arm.com>
18767             David Sherwood  <david.sherwood@arm.com>
18769         * fixed-value.h (fixed_from_double_int): Take a scalar_mode
18770         rather than a machine_mode.
18771         (fixed_from_string): Likewise.
18772         (fixed_convert): Likewise.
18773         (fixed_convert_from_int): Likewise.
18774         (fixed_convert_from_real): Likewise.
18775         (real_convert_from_fixed): Likewise.
18776         * fixed-value.c (fixed_from_double_int): Likewise.
18777         (fixed_from_string): Likewise.
18778         (fixed_convert): Likewise.
18779         (fixed_convert_from_int): Likewise.
18780         (fixed_convert_from_real): Likewise.
18781         (real_convert_from_fixed): Likewise.
18782         * config/avr/avr.c (avr_out_round): Use as_a <scalar_mode>.
18784 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18785             Alan Hayward  <alan.hayward@arm.com>
18786             David Sherwood  <david.sherwood@arm.com>
18788         * emit-rtl.c (immed_double_const): Use is_a <scalar_mode> instead
18789         of separate mode class checks.  Do not allow vector modes here.
18790         (immed_wide_int_const): Use as_a <scalar_mode>.
18791         * explow.c (trunc_int_for_mode): Likewise.
18792         * rtl.h (wi::int_traits<rtx_mode_t>::get_precision): Likewise.
18793         (wi::shwi): Likewise.
18794         (wi::min_value): Likewise.
18795         (wi::max_value): Likewise.
18796         * dwarf2out.c (loc_descriptor): Likewise.
18797         * simplify-rtx.c (simplify_immed_subreg): Fix rtx_mode_t argument
18798         for CONST_WIDE_INT.
18800 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18801             Alan Hayward  <alan.hayward@arm.com>
18802             David Sherwood  <david.sherwood@arm.com>
18804         * tree.h (SCALAR_TYPE_MODE): New macro.
18805         * expr.c (expand_expr_addr_expr_1): Use it.
18806         (expand_expr_real_2): Likewise.
18807         * fold-const.c (fold_convert_const_fixed_from_fixed): Likeise.
18808         (fold_convert_const_fixed_from_int): Likewise.
18809         (fold_convert_const_fixed_from_real): Likewise.
18810         (native_encode_fixed): Likewise
18811         (native_encode_complex): Likewise
18812         (native_encode_vector): Likewise.
18813         (native_interpret_fixed): Likewise.
18814         (native_interpret_real): Likewise.
18815         (native_interpret_complex): Likewise.
18816         (native_interpret_vector): Likewise.
18817         * omp-simd-clone.c (simd_clone_adjust_return_type): Likewise.
18818         (simd_clone_adjust_argument_types): Likewise.
18819         (simd_clone_init_simd_arrays): Likewise.
18820         (simd_clone_adjust): Likewise.
18821         * stor-layout.c (layout_type): Likewise.
18822         * tree.c (build_minus_one_cst): Likewise.
18823         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
18824         * tree-inline.c (estimate_move_cost): Likewise.
18825         * tree-ssa-math-opts.c (convert_plusminus_to_widen): Likewise.
18826         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise.
18827         (vectorizable_reduction): Likewise.
18828         * tree-vect-patterns.c (vect_recog_widen_mult_pattern): Likewise.
18829         (vect_recog_mixed_size_cond_pattern): Likewise.
18830         (check_bool_pattern): Likewise.
18831         (adjust_bool_pattern): Likewise.
18832         (search_type_for_mask_1): Likewise.
18833         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
18834         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
18835         (vectorizable_load): Likewise.
18836         (vectorizable_store): Likewise.
18837         * ubsan.c (ubsan_encode_value): Likewise.
18838         * varasm.c (output_constant): Likewise.
18840 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18841             Alan Hayward  <alan.hayward@arm.com>
18842             David Sherwood  <david.sherwood@arm.com>
18844         * coretypes.h (scalar_mode): New class.
18845         * machmode.h (scalar_mode): Likewise.
18846         (scalar_mode::includes_p): New function.
18847         (mode_to_inner): Return a scalar_mode rather than a machine_mode.
18848         * gdbhooks.py (build_pretty_printers): Handle scalar_mode.
18849         * genmodes.c (get_mode_class): Handle remaining scalar modes.
18850         * cfgexpand.c (expand_debug_expr): Use scalar_mode.
18851         * expmed.c (store_bit_field_1): Likewise.
18852         (extract_bit_field_1): Likewise.
18853         * expr.c (write_complex_part): Likewise.
18854         (read_complex_part): Likewise.
18855         (emit_move_complex_push): Likewise.
18856         (expand_expr_real_2): Likewise.
18857         * function.c (assign_parm_setup_reg): Likewise.
18858         (assign_parms_unsplit_complex): Likewise.
18859         * optabs.c (expand_binop): Likewise.
18860         * rtlanal.c (subreg_get_info): Likewise.
18861         * simplify-rtx.c (simplify_immed_subreg): Likewise.
18862         * varasm.c (output_constant_pool_2): Likewise.
18864 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18865             Alan Hayward  <alan.hayward@arm.com>
18866             David Sherwood  <david.sherwood@arm.com>
18868         * expmed.c (extract_high_half): Use scalar_int_mode and remove
18869         assertion.
18870         (expmed_mult_highpart_optab): Likewise.
18871         (expmed_mult_highpart): Likewise.
18873 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18874             Alan Hayward  <alan.hayward@arm.com>
18875             David Sherwood  <david.sherwood@arm.com>
18877         * builtins.h (builtin_strncpy_read_str): Take a scalar_int_mode
18878         instead of a machine_mode.
18879         (builtin_memset_read_str): Likewise.
18880         * builtins.c (c_readstr): Likewise.
18881         (builtin_memcpy_read_str): Likewise.
18882         (builtin_strncpy_read_str): Likewise.
18883         (builtin_memset_read_str): Likewise.
18884         (builtin_memset_gen_str): Likewise.
18885         (expand_builtin_signbit): Use scalar_int_mode for local variables.
18886         * cfgexpand.c (convert_debug_memory_address): Take a scalar_int_mode
18887         instead of a machine_mode.
18888         * combine.c (simplify_if_then_else): Use scalar_int_mode for local
18889         variables.
18890         (make_extraction): Likewise.
18891         (try_widen_shift_mode): Take and return scalar_int_modes instead
18892         of machine_modes.
18893         * config/aarch64/aarch64.c (aarch64_libgcc_cmp_return_mode): Return
18894         a scalar_int_mode instead of a machine_mode.
18895         * config/avr/avr.c (avr_addr_space_address_mode): Likewise.
18896         (avr_addr_space_pointer_mode): Likewise.
18897         * config/cr16/cr16.c (cr16_unwind_word_mode): Likewise.
18898         * config/msp430/msp430.c (msp430_addr_space_pointer_mode): Likewise.
18899         (msp430_unwind_word_mode): Likewise.
18900         * config/spu/spu.c (spu_unwind_word_mode): Likewise.
18901         (spu_addr_space_pointer_mode): Likewise.
18902         (spu_addr_space_address_mode): Likewise.
18903         (spu_libgcc_cmp_return_mode): Likewise.
18904         (spu_libgcc_shift_count_mode): Likewise.
18905         * config/rl78/rl78.c (rl78_addr_space_address_mode): Likewise.
18906         (rl78_addr_space_pointer_mode): Likewise.
18907         (fl78_unwind_word_mode): Likewise.
18908         (rl78_valid_pointer_mode): Take a scalar_int_mode instead of a
18909         machine_mode.
18910         * config/alpha/alpha.c (vms_valid_pointer_mode): Likewise.
18911         * config/ia64/ia64.c (ia64_vms_valid_pointer_mode): Likewise.
18912         * config/mips/mips.c (mips_mode_rep_extended): Likewise.
18913         (mips_valid_pointer_mode): Likewise.
18914         * config/tilegx/tilegx.c (tilegx_mode_rep_extended): Likewise.
18915         * config/ft32/ft32.c (ft32_valid_pointer_mode): Likewise.
18916         (ft32_addr_space_pointer_mode): Return a scalar_int_mode instead
18917         of a machine_mode.
18918         (ft32_addr_space_address_mode): Likewise.
18919         * config/m32c/m32c.c (m32c_valid_pointer_mode): Take a
18920         scalar_int_mode instead of a machine_mode.
18921         (m32c_addr_space_pointer_mode): Return a scalar_int_mode instead
18922         of a machine_mode.
18923         (m32c_addr_space_address_mode): Likewise.
18924         * config/powerpcspe/powerpcspe.c (rs6000_abi_word_mode): Likewise.
18925         (rs6000_eh_return_filter_mode): Likewise.
18926         * config/rs6000/rs6000.c (rs6000_abi_word_mode): Likewise.
18927         (rs6000_eh_return_filter_mode): Likewise.
18928         * config/s390/s390.c (s390_libgcc_cmp_return_mode): Likewise.
18929         (s390_libgcc_shift_count_mode): Likewise.
18930         (s390_unwind_word_mode): Likewise.
18931         (s390_valid_pointer_mode): Take a scalar_int_mode rather than a
18932         machine_mode.
18933         * target.def (mode_rep_extended): Likewise.
18934         (valid_pointer_mode): Likewise.
18935         (addr_space.valid_pointer_mode): Likewise.
18936         (eh_return_filter_mode): Return a scalar_int_mode rather than
18937         a machine_mode.
18938         (libgcc_cmp_return_mode): Likewise.
18939         (libgcc_shift_count_mode): Likewise.
18940         (unwind_word_mode): Likewise.
18941         (addr_space.pointer_mode): Likewise.
18942         (addr_space.address_mode): Likewise.
18943         * doc/tm.texi: Regenerate.
18944         * dojump.c (prefer_and_bit_test): Take a scalar_int_mode rather than
18945         a machine_mode.
18946         (do_jump): Use scalar_int_mode for local variables.
18947         * dwarf2cfi.c (init_return_column_size): Take a scalar_int_mode
18948         rather than a machine_mode.
18949         * dwarf2out.c (convert_descriptor_to_mode): Likewise.
18950         (scompare_loc_descriptor_wide): Likewise.
18951         (scompare_loc_descriptor_narrow): Likewise.
18952         * emit-rtl.c (adjust_address_1): Use scalar_int_mode for local
18953         variables.
18954         * except.c (sjlj_emit_dispatch_table): Likewise.
18955         (expand_builtin_eh_copy_values): Likewise.
18956         * explow.c (convert_memory_address_addr_space_1): Likewise.
18957         Take a scalar_int_mode rather than a machine_mode.
18958         (convert_memory_address_addr_space): Take a scalar_int_mode rather
18959         than a machine_mode.
18960         (memory_address_addr_space): Use scalar_int_mode for local variables.
18961         * expmed.h (expand_mult_highpart_adjust): Take a scalar_int_mode
18962         rather than a machine_mode.
18963         * expmed.c (mask_rtx): Likewise.
18964         (init_expmed_one_conv): Likewise.
18965         (expand_mult_highpart_adjust): Likewise.
18966         (extract_high_half): Likewise.
18967         (expmed_mult_highpart_optab): Likewise.
18968         (expmed_mult_highpart): Likewise.
18969         (expand_smod_pow2): Likewise.
18970         (expand_sdiv_pow2): Likewise.
18971         (emit_store_flag_int): Likewise.
18972         (adjust_bit_field_mem_for_reg): Use scalar_int_mode for local
18973         variables.
18974         (extract_low_bits): Likewise.
18975         * expr.h (by_pieces_constfn): Take a scalar_int_mode rather than
18976         a machine_mode.
18977         * expr.c (pieces_addr::adjust):  Likewise.
18978         (can_store_by_pieces): Likewise.
18979         (store_by_pieces): Likewise.
18980         (clear_by_pieces_1): Likewise.
18981         (expand_expr_addr_expr_1): Likewise.
18982         (expand_expr_addr_expr): Use scalar_int_mode for local variables.
18983         (expand_expr_real_1): Likewise.
18984         (try_casesi): Likewise.
18985         * final.c (shorten_branches): Likewise.
18986         * fold-const.c (fold_convert_const_int_from_fixed): Change the
18987         type of "mode" to machine_mode.
18988         * internal-fn.c (expand_arith_overflow_result_store): Take a
18989         scalar_int_mode rather than a machine_mode.
18990         (expand_mul_overflow): Use scalar_int_mode for local variables.
18991         * loop-doloop.c (doloop_modify): Likewise.
18992         (doloop_optimize): Likewise.
18993         * optabs.c (expand_subword_shift): Take a scalar_int_mode rather
18994         than a machine_mode.
18995         (expand_doubleword_shift_condmove): Likewise.
18996         (expand_doubleword_shift): Likewise.
18997         (expand_doubleword_clz): Likewise.
18998         (expand_doubleword_popcount): Likewise.
18999         (expand_doubleword_parity): Likewise.
19000         (expand_absneg_bit): Use scalar_int_mode for local variables.
19001         (prepare_float_lib_cmp): Likewise.
19002         * rtl.h (convert_memory_address_addr_space_1): Take a scalar_int_mode
19003         rather than a machine_mode.
19004         (convert_memory_address_addr_space): Likewise.
19005         (get_mode_bounds): Likewise.
19006         (get_address_mode): Return a scalar_int_mode rather than a
19007         machine_mode.
19008         * rtlanal.c (get_address_mode): Likewise.
19009         * stor-layout.c (get_mode_bounds): Take a scalar_int_mode rather
19010         than a machine_mode.
19011         * targhooks.c (default_mode_rep_extended): Likewise.
19012         (default_valid_pointer_mode): Likewise.
19013         (default_addr_space_valid_pointer_mode): Likewise.
19014         (default_eh_return_filter_mode): Return a scalar_int_mode rather
19015         than a machine_mode.
19016         (default_libgcc_cmp_return_mode): Likewise.
19017         (default_libgcc_shift_count_mode): Likewise.
19018         (default_unwind_word_mode): Likewise.
19019         (default_addr_space_pointer_mode): Likewise.
19020         (default_addr_space_address_mode): Likewise.
19021         * targhooks.h (default_eh_return_filter_mode): Likewise.
19022         (default_libgcc_cmp_return_mode): Likewise.
19023         (default_libgcc_shift_count_mode): Likewise.
19024         (default_unwind_word_mode): Likewise.
19025         (default_addr_space_pointer_mode): Likewise.
19026         (default_addr_space_address_mode): Likewise.
19027         (default_mode_rep_extended): Take a scalar_int_mode rather than
19028         a machine_mode.
19029         (default_valid_pointer_mode): Likewise.
19030         (default_addr_space_valid_pointer_mode): Likewise.
19031         * tree-ssa-address.c (addr_for_mem_ref): Use scalar_int_mode for
19032         local variables.
19033         * tree-ssa-loop-ivopts.c (get_shiftadd_cost): Take a scalar_int_mode
19034         rather than a machine_mode.
19035         * tree-switch-conversion.c (array_value_type): Use scalar_int_mode
19036         for local variables.
19037         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19038         * var-tracking.c (use_narrower_mode): Take a scalar_int_mode rather
19039         than a machine_mode.
19041 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19042             Alan Hayward  <alan.hayward@arm.com>
19043             David Sherwood  <david.sherwood@arm.com>
19045         * dojump.c (do_jump_by_parts_greater_rtx): Change the type of
19046         the mode argument to scalar_int_mode.
19047         (do_jump_by_parts_zero_rtx): Likewise.
19048         (do_jump_by_parts_equality_rtx): Likewise.
19049         (do_jump_by_parts_greater): Take a mode argument.
19050         (do_jump_by_parts_equality): Likewise.
19051         (do_jump_1): Update calls accordingly.
19053 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19054             Alan Hayward  <alan.hayward@arm.com>
19055             David Sherwood  <david.sherwood@arm.com>
19057         * is-a.h (safe_dyn_cast): New function.
19058         * rtl.h (rtx_jump_table_data::get_data_mode): New function.
19059         (jump_table_for_label): Likewise.
19060         * final.c (final_addr_vec_align): Take an rtx_jump_table_data *
19061         instead of an rtx_insn *.
19062         (shorten_branches): Use dyn_cast instead of LABEL_P and
19063         JUMP_TABLE_DATA_P.  Use jump_table_for_label and
19064         rtx_jump_table_data::get_data_mode.
19065         (final_scan_insn): Likewise.
19067 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19068             Alan Hayward  <alan.hayward@arm.com>
19069             David Sherwood  <david.sherwood@arm.com>
19071         * combine.c (try_combine): Use is_a <scalar_int_mode> when
19072         trying to combine a full-register integer set with a subreg
19073         integer set.
19075 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19076             Alan Hayward  <alan.hayward@arm.com>
19077             David Sherwood  <david.sherwood@arm.com>
19079         * expr.c (expand_expr_addr_expr): Add a new_tmode local variable
19080         that is always either address_mode or pointer_mode.
19082 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19083             Alan Hayward  <alan.hayward@arm.com>
19084             David Sherwood  <david.sherwood@arm.com>
19086         * expr.c (expand_expr_real_2): Use word_mode instead of innermode
19087         when the two are known to be equal.
19089 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19090             Alan Hayward  <alan.hayward@arm.com>
19091             David Sherwood  <david.sherwood@arm.com>
19093         * simplify-rtx.c (simplify_const_unary_operation): Use
19094         is_a <scalar_int_mode> instead of checking for a nonzero
19095         precision.  Forcibly convert op_mode to a scalar_int_mode
19096         in that case.  More clearly differentiate the operand and
19097         result modes and use the former when deciding what the value
19098         of a count-bits operation should be.  Use is_int_mode instead
19099         of checking for a MODE_INT.  Remove redundant check for whether
19100         this mode has a zero precision.
19102 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19103             Alan Hayward  <alan.hayward@arm.com>
19104             David Sherwood  <david.sherwood@arm.com>
19106         * optabs.c (widen_leading): Change the type of the mode argument
19107         to scalar_int_mode.  Use opt_scalar_int_mode for the mode iterator.
19108         (widen_bswap): Likewise.
19109         (expand_parity): Likewise.
19110         (expand_ctz): Change the type of the mode argument to scalar_int_mode.
19111         (expand_ffs): Likewise.
19112         (epand_unop): Check for scalar integer modes before calling the
19113         above routines.
19115 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19116             Alan Hayward  <alan.hayward@arm.com>
19117             David Sherwood  <david.sherwood@arm.com>
19119         * expr.c (const_scalar_mask_from_tree): Add a mode argument.
19120         Expand commentary.
19121         (expand_expr_real_1): Update call accordingly.
19123 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19124             Alan Hayward  <alan.hayward@arm.com>
19125             David Sherwood  <david.sherwood@arm.com>
19127         * expmed.c (store_bit_field_using_insv): Add op0_mode and
19128         value_mode arguments.  Use scalar_int_mode internally.
19129         (store_bit_field_1): Rename the new integer mode from imode
19130         to op0_mode and use it instead of GET_MODE (op0).  Update calls
19131         to store_split_bit_field, store_bit_field_using_insv and
19132         store_fixed_bit_field.
19133         (store_fixed_bit_field): Add op0_mode and value_mode arguments.
19134         Use scalar_int_mode internally.  Use a bit count rather than a mode
19135         when calculating the largest bit size for get_best_mode.
19136         Update calls to store_split_bit_field and store_fixed_bit_field_1.
19137         (store_fixed_bit_field_1): Add mode and value_mode arguments.
19138         Remove assertion that OP0 has a scalar integer mode.
19139         (store_split_bit_field): Add op0_mode and value_mode arguments.
19140         Update calls to extract_fixed_bit_field.
19141         (extract_bit_field_using_extv): Add an op0_mode argument.
19142         Use scalar_int_mode internally.
19143         (extract_bit_field_1): Rename the new integer mode from imode to
19144         op0_mode and use it instead of GET_MODE (op0).  Update calls to
19145         extract_split_bit_field, extract_bit_field_using_extv and
19146         extract_fixed_bit_field.
19147         (extract_fixed_bit_field): Add an op0_mode argument.  Update calls
19148         to extract_split_bit_field and extract_fixed_bit_field_1.
19149         (extract_fixed_bit_field_1): Add a mode argument.  Remove assertion
19150         that OP0 has a scalar integer mode.  Use as_a <scalar_int_mode>
19151         on the target mode.
19152         (extract_split_bit_field): Add an op0_mode argument.  Update call
19153         to extract_fixed_bit_field.
19155 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19156             Alan Hayward  <alan.hayward@arm.com>
19157             David Sherwood  <david.sherwood@arm.com>
19159         * cse.c (cse_insn): Use opt_scalar_int_mode for the mode iterator.
19160         * explow.c (hard_function_value): Likewise.
19161         * expmed.c (extract_fixed_bit_field_1): Likewise.  Move the
19162         convert_to_mode call outside the loop.
19163         * expr.c (alignment_for_piecewise_move): Use opt_scalar_int_mode
19164         for the mode iterator.  Require the mode specified by max_pieces
19165         to exist.
19166         (emit_block_move_via_movmem): Use opt_scalar_int_mode for the
19167         mode iterator.
19168         (copy_blkmode_to_reg): Likewise.
19169         (set_storage_via_setmem): Likewise.
19170         * optabs.c (prepare_cmp_insn): Likewise.
19171         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
19172         * stor-layout.c (finish_bitfield_representative): Likewise.
19174 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19175             Alan Hayward  <alan.hayward@arm.com>
19176             David Sherwood  <david.sherwood@arm.com>
19178         * rtl.h (subreg_unpromoted_mode, subreg_promoted_mode): New functions.
19179         * expr.c (convert_move): Use them.
19180         (convert_modes): Likewise.
19181         (store_expr_with_bounds): Likewise.
19183 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19184             Alan Hayward  <alan.hayward@arm.com>
19185             David Sherwood  <david.sherwood@arm.com>
19187         * rtl.h (rtl_hooks::reg_nonzero_bits): Add a scalar_int_mode
19188         parameter for the mode of "x".  Remove the "known_x", "known_mode"
19189         and "known_ret" arguments.  Change the type of the mode argument
19190         to scalar_int_mode.
19191         (rtl_hooks:reg_num_sign_bit_copies): Likewise.
19192         * combine.c (reg_nonzero_bits_for_combine): Update accordingly.
19193         (reg_num_sign_bit_copies_for_combine): Likewise.
19194         * rtlanal.c (nonzero_bits1): Likewise.
19195         (num_sign_bit_copies1): Likewise.
19196         * rtlhooks-def.h (reg_nonzero_bits_general): Likewise.
19197         (reg_num_sign_bit_copies_general): Likewise.
19198         * rtlhooks.c (reg_num_sign_bit_copies_general): Likewise.
19199         (reg_nonzero_bits_general): Likewise.
19201 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19202             Alan Hayward  <alan.hayward@arm.com>
19203             David Sherwood  <david.sherwood@arm.com>
19205         * rtlanal.c (num_sign_bit_copies): Handle VOIDmode here rather
19206         than in subroutines.  Return 1 for non-integer modes.
19207         (cached_num_sign_bit_copies): Change the type of the mode parameter
19208         to scalar_int_mode.
19209         (num_sign_bit_copies1): Likewise.  Remove early exit for other mode
19210         classes.  Handle CONST_INT_P first and then check whether X also
19211         has a scalar integer mode.  Check the same thing for inner registers
19212         of a SUBREG and for values that are being extended or truncated.
19214 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19215             Alan Hayward  <alan.hayward@arm.com>
19216             David Sherwood  <david.sherwood@arm.com>
19218         * rtlanal.c (nonzero_bits): Handle VOIDmode here rather than
19219         in subroutines.  Return the mode mask for non-integer modes.
19220         (cached_nonzero_bits): Change the type of the mode parameter
19221         to scalar_int_mode.
19222         (nonzero_bits1): Likewise.  Remove early exit for other mode
19223         classes.  Handle CONST_INT_P first and then check whether X
19224         also has a scalar integer mode.
19226 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19227             Alan Hayward  <alan.hayward@arm.com>
19228             David Sherwood  <david.sherwood@arm.com>
19230         * expr.c (widest_int_mode_for_size): Make the comment match the code.
19231         Return a scalar_int_mode and assert that the size is greater than
19232         one byte.
19233         (by_pieces_ninsns): Update accordingly and remove VOIDmode handling.
19234         (op_by_pieces_d::op_by_pieces_d): Likewise.
19235         (op_by_pieces_d::run): Likewise.
19236         (can_store_by_pieces): Likewise.
19238 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19239             Alan Hayward  <alan.hayward@arm.com>
19240             David Sherwood  <david.sherwood@arm.com>
19242         * combine.c (extract_left_shift): Add a mode argument and update
19243         recursive calls.
19244         (make_compound_operation_int): Change the type of the mode parameter
19245         to scalar_int_mode and update the call to extract_left_shift.
19247 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19248             Alan Hayward  <alan.hayward@arm.com>
19249             David Sherwood  <david.sherwood@arm.com>
19251         * combine.c (simplify_and_const_int): Change the type of the mode
19252         parameter to scalar_int_mode.
19253         (simplify_and_const_int_1): Likewise.  Update recursive call.
19255 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19256             Alan Hayward  <alan.hayward@arm.com>
19257             David Sherwood  <david.sherwood@arm.com>
19259         * combine.c (simplify_compare_const): Check that the mode is a
19260         scalar_int_mode (rather than VOIDmode) before testing its
19261         precision.
19262         (simplify_comparison): Move COMPARISON_P handling out of the
19263         loop and restrict the latter part of the loop to scalar_int_modes.
19264         Check is_a <scalar_int_mode> before calling HWI_COMPUTABLE_MODE_P
19265         and when considering SUBREG_REGs.  Use is_int_mode instead of
19266         checking GET_MODE_CLASS against MODE_INT.
19268 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19269             Alan Hayward  <alan.hayward@arm.com>
19270             David Sherwood  <david.sherwood@arm.com>
19272         * combine.c (try_widen_shift_mode): Move check for equal modes to...
19273         (simplify_shift_const_1): ...here.  Use scalar_int_mode for
19274         shift_unit_mode and for modes involved in scalar shifts.
19276 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19277             Alan Hayward  <alan.hayward@arm.com>
19278             David Sherwood  <david.sherwood@arm.com>
19280         * combine.c (force_int_to_mode): New function, split out from...
19281         (force_to_mode): ...here.  Keep xmode up-to-date and use it
19282         instead of GET_MODE (x).
19284 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19285             Alan Hayward  <alan.hayward@arm.com>
19286             David Sherwood  <david.sherwood@arm.com>
19288         * optabs-query.h (extraction_insn::struct_mode): Change type to
19289         opt_scalar_int_mode and update comment.
19290         (extraction_insn::field_mode): Change type to scalar_int_mode.
19291         (extraction_insn::pos_mode): Likewise.
19292         * combine.c (make_extraction): Update accordingly.
19293         * optabs-query.c (get_traditional_extraction_insn): Likewise.
19294         (get_optab_extraction_insn): Likewise.
19295         * recog.c (simplify_while_replacing): Likewise.
19296         * expmed.c (narrow_bit_field_mem): Change the type of the mode
19297         parameter to opt_scalar_int_mode.
19299 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19300             Alan Hayward  <alan.hayward@arm.com>
19301             David Sherwood  <david.sherwood@arm.com>
19303         * machmode.h (bit_field_mode_iterator::next_mode): Take a pointer
19304         to a scalar_int_mode instead of a machine_mode.
19305         (bit_field_mode_iterator::m_mode): Change type to opt_scalar_int_mode.
19306         (get_best_mode): Return a boolean and use a pointer argument to store
19307         the selected mode.  Replace the limit mode parameter with a bit limit.
19308         * expmed.c (adjust_bit_field_mem_for_reg): Use scalar_int_mode
19309         for the values returned by bit_field_mode_iterator::next_mode.
19310         (store_bit_field): Update call to get_best_mode.
19311         (store_fixed_bit_field): Likewise.
19312         (extract_fixed_bit_field): Likewise.
19313         * expr.c (optimize_bitfield_assignment_op): Likewise.
19314         * fold-const.c (optimize_bit_field_compare): Likewise.
19315         (fold_truth_andor_1): Likewise.
19316         * stor-layout.c (bit_field_mode_iterator::next_mode): As above.
19317         Update for new type of m_mode.
19318         (get_best_mode): As above.
19320 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19321             Alan Hayward  <alan.hayward@arm.com>
19322             David Sherwood  <david.sherwood@arm.com>
19324         * expmed.c (strict_volatile_bitfield_p): Change the type of fieldmode
19325         to scalar_int_mode.  Remove check for SCALAR_INT_MODE_P.
19326         (store_bit_field): Check is_a <scalar_int_mode> before calling
19327         strict_volatile_bitfield_p.
19328         (extract_bit_field): Likewise.
19330 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19331             Alan Hayward  <alan.hayward@arm.com>
19332             David Sherwood  <david.sherwood@arm.com>
19334         * target.def (cstore_mode): Return a scalar_int_mode.
19335         * doc/tm.texi: Regenerate.
19336         * config/sparc/sparc.c (sparc_cstore_mode): Return a scalar_int_mode.
19337         * targhooks.h (default_cstore_mode): Likewise.
19338         * targhooks.c (default_cstore_mode): Likewise, using a forced
19339         conversion.
19340         * expmed.c (emit_cstore): Expect the target of the cstore to be
19341         a scalar_int_mode.
19343 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19344             Alan Hayward  <alan.hayward@arm.com>
19345             David Sherwood  <david.sherwood@arm.com>
19347         * cfgloop.h (rtx_iv): Change type of extend_mode and mode to
19348         scalar_int_mode.
19349         (niter_desc): Likewise mode.
19350         (iv_analyze): Add a mode parameter.
19351         (biv_p): Likewise.
19352         (iv_analyze_expr): Pass the mode paraeter before the rtx it describes
19353         and change its type to scalar_int_mode.
19354         * loop-iv.c: Update commentary at head of file.
19355         (iv_constant): Pass the mode paraeter before the rtx it describes
19356         and change its type to scalar_int_mode.  Remove VOIDmode handling.
19357         (iv_subreg): Change the type of the mode parameter to scalar_int_mode.
19358         (iv_extend): Likewise.
19359         (shorten_into_mode): Likewise.
19360         (iv_add): Use scalar_int_mode.
19361         (iv_mult): Likewise.
19362         (iv_shift): Likewise.
19363         (canonicalize_iv_subregs): Likewise.
19364         (get_biv_step_1): Pass the outer_mode parameter before the rtx
19365         it describes and change its mode to scalar_int_mode.   Also change
19366         the type of the returned inner_mode to scalar_int_mode.
19367         (get_biv_step): Likewise, turning outer_mode from a pointer
19368         into a direct parameter.  Update call to get_biv_step_1.
19369         (iv_analyze_biv): Add an outer_mode parameter.  Update calls to
19370         iv_constant and get_biv_step.
19371         (iv_analyze_expr): Pass the mode parameter before the rtx it describes
19372         and change its type to scalar_int_mode.  Don't initialise iv->mode
19373         to VOIDmode and remove later checks for its still being VOIDmode.
19374         Update calls to iv_analyze_op and iv_analyze_expr.  Check
19375         is_a <scalar_int_mode> when changing the mode under consideration.
19376         (iv_analyze_def): Ignore registers that don't have a scalar_int_mode.
19377         Update call to iv_analyze_expr.
19378         (iv_analyze_op): Add a mode parameter.  Reject subregs whose
19379         inner register is not also a scalar_int_mode.  Update call to
19380         iv_analyze_biv.
19381         (iv_analyze): Add a mode parameter.  Update call to iv_analyze_op.
19382         (biv_p): Add a mode parameter.  Update call to iv_analyze_biv.
19383         (iv_number_of_iterations): Use is_a <scalar_int_mode> instead of
19384         separate mode class checks.  Update calls to iv_analyze.  Remove
19385         fix-up of VOIDmodes after iv_analyze_biv.
19386         * loop-unroll.c (analyze_iv_to_split_insn): Reject registers that
19387         don't have a scalar_int_mode.  Update call to biv_p.
19389 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19390             Alan Hayward  <alan.hayward@arm.com>
19391             David Sherwood  <david.sherwood@arm.com>
19393         * cfgexpand.c (convert_debug_memory_address): Use
19394         as_a <scalar_int_mode>.
19395         * combine.c (expand_compound_operation): Likewise.
19396         (make_extraction): Likewise.
19397         (change_zero_ext): Likewise.
19398         (simplify_comparison): Likewise.
19399         * cse.c (cse_insn): Likewise.
19400         * dwarf2out.c (minmax_loc_descriptor): Likewise.
19401         (mem_loc_descriptor): Likewise.
19402         (loc_descriptor): Likewise.
19403         * expmed.c (init_expmed_one_mode): Likewise.
19404         (synth_mult): Likewise.
19405         (emit_store_flag_1): Likewise.
19406         (expand_divmod): Likewise.  Use HWI_COMPUTABLE_MODE_P instead
19407         of a comparison with size.
19408         * expr.c (expand_assignment): Use as_a <scalar_int_mode>.
19409         (reduce_to_bit_field_precision): Likewise.
19410         * function.c (expand_function_end): Likewise.
19411         * internal-fn.c (expand_arith_overflow_result_store): Likewise.
19412         * loop-doloop.c (doloop_modify): Likewise.
19413         * optabs.c (expand_binop): Likewise.
19414         (expand_unop): Likewise.
19415         (expand_copysign_absneg): Likewise.
19416         (prepare_cmp_insn): Likewise.
19417         (maybe_legitimize_operand): Likewise.
19418         * recog.c (const_scalar_int_operand): Likewise.
19419         * rtlanal.c (get_address_mode): Likewise.
19420         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19421         (simplify_cond_clz_ctz): Likewise.
19422         * tree-nested.c (get_nl_goto_field): Likewise.
19423         * tree.c (build_vector_type_for_mode): Likewise.
19424         * var-tracking.c (use_narrower_mode): Likewise.
19426 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19427             Alan Hayward  <alan.hayward@arm.com>
19428             David Sherwood  <david.sherwood@arm.com>
19430         * tree.h (SCALAR_INT_TYPE_MODE): New macro.
19431         * builtins.c (expand_builtin_signbit): Use it.
19432         * cfgexpand.c (expand_debug_expr): Likewise.
19433         * dojump.c (do_jump): Likewise.
19434         (do_compare_and_jump): Likewise.
19435         * dwarf2cfi.c (expand_builtin_init_dwarf_reg_sizes): Likewise.
19436         * expmed.c (make_tree): Likewise.
19437         * expr.c (expand_expr_real_2): Likewise.
19438         (expand_expr_real_1): Likewise.
19439         (try_casesi): Likewise.
19440         * fold-const-call.c (fold_const_call_ss): Likewise.
19441         * fold-const.c (unextend): Likewise.
19442         (extract_muldiv_1): Likewise.
19443         (fold_single_bit_test): Likewise.
19444         (native_encode_int): Likewise.
19445         (native_encode_string): Likewise.
19446         (native_interpret_int): Likewise.
19447         * gimple-fold.c (gimple_fold_builtin_memset): Likewise.
19448         * internal-fn.c (expand_addsub_overflow): Likewise.
19449         (expand_neg_overflow): Likewise.
19450         (expand_mul_overflow): Likewise.
19451         (expand_arith_overflow): Likewise.
19452         * match.pd: Likewise.
19453         * stor-layout.c (layout_type): Likewise.
19454         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
19455         * tree-ssa-math-opts.c (convert_mult_to_widen): Likewise.
19456         * tree-ssanames.c (get_range_info): Likewise.
19457         * tree-switch-conversion.c (array_value_type) Likewise.
19458         * tree-vect-patterns.c (vect_recog_rotate_pattern): Likewise.
19459         (vect_recog_divmod_pattern): Likewise.
19460         (vect_recog_mixed_size_cond_pattern): Likewise.
19461         * tree-vrp.c (extract_range_basic): Likewise.
19462         (simplify_float_conversion_using_ranges): Likewise.
19463         * tree.c (int_fits_type_p): Likewise.
19464         * ubsan.c (instrument_bool_enum_load): Likewise.
19465         * varasm.c (mergeable_string_section): Likewise.
19466         (narrowing_initializer_constant_valid_p): Likewise.
19467         (output_constant): Likewise.
19469 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19470             Alan Hayward  <alan.hayward@arm.com>
19471             David Sherwood  <david.sherwood@arm.com>
19473         * machmode.h (NARROWEST_INT_MODE): New macro.
19474         * expr.c (alignment_for_piecewise_move): Use it instead of
19475         GET_CLASS_NARROWEST_MODE (MODE_INT).
19476         (push_block): Likewise.
19477         * stor-layout.c (bit_field_mode_iterator::bit_field_mode_iterator):
19478         Likewise.
19479         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19481 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19482             Alan Hayward  <alan.hayward@arm.com>
19483             David Sherwood  <david.sherwood@arm.com>
19485         * postreload.c (move2add_valid_value_p): Change the type of the
19486         mode parameter to scalar_int_mode.
19487         (move2add_use_add2_insn): Add a mode parameter and use it instead
19488         of GET_MODE (reg).
19489         (move2add_use_add3_insn): Likewise.
19490         (reload_cse_move2add): Update accordingly.
19492 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19493             Alan Hayward  <alan.hayward@arm.com>
19494             David Sherwood  <david.sherwood@arm.com>
19496         * expr.c (expand_expr_real_2): Use scalar_int_mode for the
19497         double-word mode.
19498         * lower-subreg.c (resolve_shift_zext): Use is_a <scalar_int_mode>.
19499         * optabs.c (expand_unop): Likewise.
19501 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19502             Alan Hayward  <alan.hayward@arm.com>
19503             David Sherwood  <david.sherwood@arm.com>
19505         * dwarf2out.c (typed_binop): Change mode parameter to scalar_int_mode.
19506         (clz_loc_descriptor): Likewise.  Remove SCALAR_INT_MODE_P check.
19507         (popcount_loc_descriptor): Likewise.
19508         (bswap_loc_descriptor): Likewise.
19509         (rotate_loc_descriptor): Likewise.
19510         (mem_loc_descriptor): Add is_a <scalar_int_mode> checks before
19511         calling the functions above.
19513 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19514             Alan Hayward  <alan.hayward@arm.com>
19515             David Sherwood  <david.sherwood@arm.com>
19517         * combine.c (sign_extend_short_imm): Add is_a <scalar_int_mode>
19518         checks.
19519         (try_combine): Likewise.
19520         (simplify_if_then_else): Likewise.
19521         * cse.c (cse_insn): Likewise.
19522         * dwarf2out.c (mem_loc_descriptor): Likewise.
19523         * emit-rtl.c (gen_lowpart_common): Likewise.
19524         * simplify-rtx.c (simplify_truncation): Likewise.
19525         (simplify_binary_operation_1): Likewise.
19526         (simplify_const_relational_operation): Likewise.
19527         (simplify_ternary_operation): Likewise.
19528         * tree-ssa-loop-ivopts.c (force_expr_to_var_cost): Likewise.
19530 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19531             Alan Hayward  <alan.hayward@arm.com>
19532             David Sherwood  <david.sherwood@arm.com>
19534         * cse.c (cse_insn): Add is_a <scalar_int_mode> checks.
19535         * reload.c (push_reload): Likewise.
19536         (find_reloads): Likewise.
19538 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19539             Alan Hayward  <alan.hayward@arm.com>
19540             David Sherwood  <david.sherwood@arm.com>
19542         * combine.c (find_split_point): Add is_a <scalar_int_mode> checks.
19543         (make_compound_operation_int): Likewise.
19544         (change_zero_ext): Likewise.
19545         * expr.c (convert_move): Likewise.
19546         (convert_modes): Likewise.
19547         * fwprop.c (forward_propagate_subreg): Likewise.
19548         * loop-iv.c (get_biv_step_1): Likewise.
19549         * optabs.c (widen_operand): Likewise.
19550         * postreload.c (move2add_valid_value_p): Likewise.
19551         * recog.c (simplify_while_replacing): Likewise.
19552         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19553         (simplify_binary_operation_1): Likewise.  Remove redundant
19554         mode equality check.
19556 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19557             Alan Hayward  <alan.hayward@arm.com>
19558             David Sherwood  <david.sherwood@arm.com>
19560         * combine.c (combine_simplify_rtx): Add checks for
19561         is_a <scalar_int_mode>.
19562         (simplify_if_then_else): Likewise.
19563         (make_field_assignment): Likewise.
19564         (simplify_comparison): Likewise.
19565         * ifcvt.c (noce_try_bitop): Likewise.
19566         * loop-invariant.c (canonicalize_address_mult): Likewise.
19567         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19569 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19570             Alan Hayward  <alan.hayward@arm.com>
19571             David Sherwood  <david.sherwood@arm.com>
19573         * gimple-fold.c (gimple_fold_builtin_memory_op): Use
19574         is_a <scalar_int_mode> instead of != BLKmode.
19576 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19577             Alan Hayward  <alan.hayward@arm.com>
19578             David Sherwood  <david.sherwood@arm.com>
19580         * cfgexpand.c (expand_debug_expr): Use is_a <scalar_int_mode>
19581         instead of != VOIDmode.
19582         * combine.c (if_then_else_cond): Likewise.
19583         (change_zero_ext): Likewise.
19584         * dwarf2out.c (mem_loc_descriptor): Likewise.
19585         (loc_descriptor): Likewise.
19586         * rtlanal.c (canonicalize_condition): Likewise.
19587         * simplify-rtx.c (simplify_relational_operation_1): Likewise.
19589 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19590             Alan Hayward  <alan.hayward@arm.com>
19591             David Sherwood  <david.sherwood@arm.com>
19593         * simplify-rtx.c (simplify_binary_operation_1): Use
19594         is_a <scalar_int_mode> instead of !VECTOR_MODE_P.
19596 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19597             Alan Hayward  <alan.hayward@arm.com>
19598             David Sherwood  <david.sherwood@arm.com>
19600         * wide-int.h (int_traits<unsigned char>) New class.
19601         (int_traits<unsigned short>) Likewise.
19602         * cfgexpand.c (expand_debug_expr): Use is_a <scalar_int_mode>.
19603         Use GET_MODE_UNIT_PRECISION and remove redundant test for
19604         SCALAR_INT_MODE_P.
19605         * combine.c (set_nonzero_bits_and_sign_copies): Use
19606         is_a <scalar_int_mode>.
19607         (find_split_point): Likewise.
19608         (combine_simplify_rtx): Likewise.
19609         (simplify_logical): Likewise.
19610         (expand_compound_operation): Likewise.
19611         (expand_field_assignment): Likewise.
19612         (make_compound_operation): Likewise.
19613         (extended_count): Likewise.
19614         (change_zero_ext): Likewise.
19615         (simplify_comparison): Likewise.
19616         * dwarf2out.c (scompare_loc_descriptor): Likewise.
19617         (ucompare_loc_descriptor): Likewise.
19618         (minmax_loc_descriptor): Likewise.
19619         (mem_loc_descriptor): Likewise.
19620         (loc_descriptor): Likewise.
19621         * expmed.c (init_expmed_one_mode): Likewise.
19622         * lra-constraints.c (lra_constraint_offset): Likewise.
19623         * optabs.c (prepare_libcall_arg): Likewise.
19624         * postreload.c (move2add_note_store): Likewise.
19625         * reload.c (operands_match_p): Likewise.
19626         * rtl.h (load_extend_op): Likewise.
19627         * rtlhooks.c (gen_lowpart_general): Likewise.
19628         * simplify-rtx.c (simplify_truncation): Likewise.
19629         (simplify_unary_operation_1): Likewise.
19630         (simplify_binary_operation_1): Likewise.
19631         (simplify_const_binary_operation): Likewise.
19632         (simplify_const_relational_operation): Likewise.
19633         (simplify_subreg): Likewise.
19634         * stor-layout.c (bitwise_mode_for_mode): Likewise.
19635         * var-tracking.c (adjust_mems): Likewise.
19636         (prepare_call_arguments): Likewise.
19638 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19639             Alan Hayward  <alan.hayward@arm.com>
19640             David Sherwood  <david.sherwood@arm.com>
19642         * machmode.h (is_int_mode): New fuction.
19643         * combine.c (find_split_point): Use it.
19644         (combine_simplify_rtx): Likewise.
19645         (simplify_if_then_else): Likewise.
19646         (simplify_set): Likewise.
19647         (simplify_shift_const_1): Likewise.
19648         (simplify_comparison): Likewise.
19649         * config/aarch64/aarch64.c (aarch64_rtx_costs): Likewise.
19650         * cse.c (notreg_cost): Likewise.
19651         (cse_insn): Likewise.
19652         * cselib.c (cselib_lookup_1): Likewise.
19653         * dojump.c (do_jump_1): Likewise.
19654         (do_compare_rtx_and_jump): Likewise.
19655         * dse.c (get_call_args): Likewise.
19656         * dwarf2out.c (rtl_for_decl_init): Likewise.
19657         (native_encode_initializer): Likewise.
19658         * expmed.c (emit_store_flag_1): Likewise.
19659         (emit_store_flag): Likewise.
19660         * expr.c (convert_modes): Likewise.
19661         (store_field): Likewise.
19662         (expand_expr_real_1): Likewise.
19663         * fold-const.c (fold_read_from_constant_string): Likewise.
19664         * gimple-ssa-sprintf.c (get_format_string): Likewise.
19665         * optabs-libfuncs.c (gen_int_libfunc): Likewise.
19666         * optabs.c (expand_binop): Likewise.
19667         (expand_unop): Likewise.
19668         (expand_abs_nojump): Likewise.
19669         (expand_one_cmpl_abs_nojump): Likewise.
19670         * simplify-rtx.c (mode_signbit_p): Likewise.
19671         (val_signbit_p): Likewise.
19672         (val_signbit_known_set_p): Likewise.
19673         (val_signbit_known_clear_p): Likewise.
19674         (simplify_relational_operation_1): Likewise.
19675         * tree.c (vector_type_mode): Likewise.
19677 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19678             Alan Hayward  <alan.hayward@arm.com>
19679             David Sherwood  <david.sherwood@arm.com>
19681         * machmode.h (smallest_mode_for_size): Fix formatting.
19682         (smallest_int_mode_for_size): New function.
19683         * cfgexpand.c (expand_debug_expr): Use smallest_int_mode_for_size
19684         instead of smallest_mode_for_size.
19685         * combine.c (make_extraction): Likewise.
19686         * config/arc/arc.c (arc_expand_movmem): Likewise.
19687         * config/arm/arm.c (arm_expand_divmod_libfunc): Likewise.
19688         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
19689         * config/s390/s390.c (s390_expand_insv): Likewise.
19690         * config/sparc/sparc.c (assign_int_registers): Likewise.
19691         * config/spu/spu.c (spu_function_value): Likewise.
19692         (spu_function_arg): Likewise.
19693         * coverage.c (get_gcov_type): Likewise.
19694         (get_gcov_unsigned_t): Likewise.
19695         * dse.c (find_shift_sequence): Likewise.
19696         * expmed.c (store_bit_field_1): Likewise.
19697         * expr.c (convert_move): Likewise.
19698         (store_field): Likewise.
19699         * internal-fn.c (expand_arith_overflow): Likewise.
19700         * optabs-query.c (get_best_extraction_insn): Likewise.
19701         * optabs.c (expand_twoval_binop_libfunc): Likewise.
19702         * stor-layout.c (layout_type): Likewise.
19703         (initialize_sizetypes): Likewise.
19704         * targhooks.c (default_get_mask_mode): Likewise.
19705         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Likewise.
19707 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19708             Alan Hayward  <alan.hayward@arm.com>
19709             David Sherwood  <david.sherwood@arm.com>
19711         * machmode.h (opt_mode::else_blk): New function.
19712         (int_mode_for_mode): Declare.
19713         * stor-layout.c (int_mode_for_mode): Return an opt_scalar_int_mode.
19714         * builtins.c (expand_builtin_signbit): Adjust for new int_mode_for_mode
19715         return type.
19716         * cfgexpand.c (expand_debug_expr): Likewise.
19717         * combine.c (gen_lowpart_or_truncate): Likewise.
19718         (gen_lowpart_for_combine): Likewise.
19719         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Likewise.
19720         * config/avr/avr.c (avr_to_int_mode): Likewise.
19721         (avr_out_plus_1): Likewise.
19722         (avr_out_plus): Likewise.
19723         (avr_out_round): Likewise.
19724         * config/i386/i386.c (ix86_split_to_parts): Likewise.
19725         * config/s390/s390.c (s390_expand_vec_compare_cc): Likewise.
19726         (s390_expand_vcond): Likewise.
19727         * config/spu/spu.c (spu_split_immediate): Likewise.
19728         (spu_expand_mov): Likewise.
19729         * dse.c (get_stored_val): Likewise.
19730         * expmed.c (store_bit_field_1): Likewise.
19731         (convert_extracted_bit_field): Use int_mode_for_mode instead of
19732         int_mode_for_size.
19733         (extract_bit_field_1): Adjust for new int_mode_for_mode return type.
19734         (extract_low_bits): Likewise.
19735         * expr.c (emit_group_load_1): Likewise.  Separate out the BLKmode
19736         handling rather than repeating the check.
19737         (emit_group_store): Likewise.
19738         (emit_move_via_integer): Adjust for new int_mode_for_mode return type.
19739         * optabs.c (expand_absneg_bit): Likewise.
19740         (expand_copysign_absneg): Likewise.
19741         (expand_copysign_bit): Likewise.
19742         * tree-if-conv.c (ifcvt_can_use_mask_load_store): Likewise.
19743         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19744         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
19745         * var-tracking.c (prepare_call_arguments):  Likewise.
19746         * config/powerpcspe/powerpcspe.c (rs6000_do_expand_vec_perm): Use
19747         int_mode_for_mode instead of mode_for_size.
19748         * config/rs6000/rs6000.c (rs6000_do_expand_vec_perm): Likewise.
19750 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19751             Alan Hayward  <alan.hayward@arm.com>
19752             David Sherwood  <david.sherwood@arm.com>
19754         * machmode.h (int_mode_for_size): New function.
19755         * builtins.c (set_builtin_user_assembler_name): Use int_mode_for_size
19756         instead of mode_for_size.
19757         * calls.c (save_fixed_argument_area): Likewise.  Make use of BLKmode
19758         explicit.
19759         * combine.c (expand_field_assignment): Use int_mode_for_size
19760         instead of mode_for_size.
19761         (make_extraction): Likewise.
19762         (simplify_shift_const_1): Likewise.
19763         (simplify_comparison): Likewise.
19764         * dojump.c (do_jump): Likewise.
19765         * dwarf2out.c (mem_loc_descriptor): Likewise.
19766         * emit-rtl.c (init_derived_machine_modes): Likewise.
19767         * expmed.c (flip_storage_order): Likewise.
19768         (convert_extracted_bit_field): Likewise.
19769         * expr.c (copy_blkmode_from_reg): Likewise.
19770         * graphite-isl-ast-to-gimple.c (max_mode_int_precision): Likewise.
19771         * internal-fn.c (expand_mul_overflow): Likewise.
19772         * lower-subreg.c (simple_move): Likewise.
19773         * optabs-libfuncs.c (init_optabs): Likewise.
19774         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19775         * tree.c (vector_type_mode): Likewise.
19776         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
19777         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Likewise.
19778         * tree-vect-generic.c (expand_vector_parallel): Likewise.
19779         * tree-vect-stmts.c (vectorizable_load): Likewise.
19780         (vectorizable_store): Likewise.
19782 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19783             Alan Hayward  <alan.hayward@arm.com>
19784             David Sherwood  <david.sherwood@arm.com>
19786         * coretypes.h (pod_mode): New type.
19787         (scalar_int_mode_pod): New typedef.
19788         * machmode.h (pod_mode): New class.
19789         (int_n_data_t::m): Change type to scalar_int_mode_pod.
19790         * genmodes.c (emit_mode_int_n): Update accordingly.
19791         * lower-subreg.h (target_lower_subreg): Change type to
19792         scalar_int_mode_pod.
19793         * gdbhooks.py (build_pretty_printer): Handle pod_mode and
19794         scalar_int_mode_pod.
19796 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19797             Alan Hayward  <alan.hayward@arm.com>
19798             David Sherwood  <david.sherwood@arm.com>
19800         * config/powerpcspe/powerpcspe.h (rs6000_pmode): Change type from
19801         machine_mode to scalar_int_mode.
19802         * config/powerpcspe/powerpcspe.c (rs6000_pmode): Likewise.
19803         (rs6000_option_override_internal): Remove cast to int.
19804         * config/rs6000/rs6000.h (rs6000_pmode): Change type from
19805         machine_mode to scalar_int_mode.
19806         * config/rs6000/rs6000.c (rs6000_pmode): Likewise.
19807         (rs6000_option_override_internal): Remove cast to int.
19808         * config/s390/s390.h (Pmode): Remove cast to machine_mode.
19809         * config/epiphany/epiphany.h (RTX_OK_FOR_OFFSET_P): Add cast
19810         to machine_mode.
19811         * config/s390/s390.c (s390_expand_builtin): Likewise.
19812         * coretypes.h (scalar_int_mode): New type.
19813         (opt_scalar_int_mode): New typedef.
19814         * machmode.h (scalar_int_mode): New class.
19815         (scalar_int_mode::includes_p): New function.
19816         (byte_mode): Change type to scalar_int_mode.
19817         (word_mode): Likewise.
19818         (ptr_mode): Likewise.
19819         * emit-rtl.c (byte_mode): Likewise.
19820         (word_mode): Likewise.
19821         (ptr_mode): Likewise.
19822         (init_derived_machine_modes): Update accordingly.
19823         * genmodes.c (get_mode_class): Return scalar_int_mode for MODE_INT
19824         and MODE_PARTIAL_INT.
19825         * gdbhooks.py (build_pretty_printer): Handle scalar_int_mode and
19826         opt_scalar_int_mode.
19828 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19829             Alan Hayward  <alan.hayward@arm.com>
19830             David Sherwood  <david.sherwood@arm.com>
19832         * target.def (libgcc_floating_mode_supported_p): Take a
19833         scalar_float_mode.
19834         * doc/tm.texi: Regenerate.
19835         * targhooks.h (default_libgcc_floating_mode_supported_p): Take a
19836         scalar_float_mode.
19837         * targhooks.c (default_libgcc_floating_mode_supported_p): Likewise.
19838         * config/aarch64/aarch64.c (aarch64_libgcc_floating_mode_supported_p):
19839         Likewise.
19841 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19842             Alan Hayward  <alan.hayward@arm.com>
19843             David Sherwood  <david.sherwood@arm.com>
19845         * target.def (default_floatn_mode): Return an opt_scalar_float_mode.
19846         * doc/tm.texi: Regenerate.
19847         * config/arm/arm.c (arm_floatn_mode): Return an opt_scalar_float_mode.
19848         * config/powerpcspe/powerpcspe.c (rs6000_floatn_mode): Likewise.
19849         * config/rs6000/rs6000.c (rs6000_floatn_mode): Likewise.
19850         * targhooks.h (default_floatn_mode): Likewise.
19851         * targhooks.c (default_floatn_mode): Likewise.
19852         * tree.c (build_common_tree_nodes): Update accordingly.
19854 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19855             Alan Hayward  <alan.hayward@arm.com>
19856             David Sherwood  <david.sherwood@arm.com>
19858         * machmode.h (mode_iterator::start): Provide overload for opt_modes.
19859         (mode_iterator::iterate_p): Likewise.
19860         (mode_iterator::get_wider): Likewise.
19861         * expr.c (init_expr_target): Use opt_scalar_float_mode.
19863 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19864             Alan Hayward  <alan.hayward@arm.com>
19865             David Sherwood  <david.sherwood@arm.com>
19867         * coretypes.h (opt_scalar_float_mode): New typedef.
19868         * machmode.h (float_mode_for_size): New function.
19869         * emit-rtl.c (double_mode): Delete.
19870         (init_emit_once): Use float_mode_for_size.
19871         * stor-layout.c (layout_type): Likewise.
19872         * gdbhooks.py (build_pretty_printer): Handle opt_scalar_float_mode.
19874 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19875             Alan Hayward  <alan.hayward@arm.com>
19876             David Sherwood  <david.sherwood@arm.com>
19878         * output.h (assemble_real): Take a scalar_float_mode.
19879         * config/arm/arm.c (arm_assemble_integer): Update accordingly.
19880         * config/arm/arm.md (consttable_4): Likewise.
19881         (consttable_8): Likewise.
19882         (consttable_16): Likewise.
19883         * config/mips/mips.md (consttable_float): Likewise.
19884         * config/s390/s390.c (s390_output_pool_entry): Likewise.
19885         * varasm.c (assemble_real): Take a scalar_float_mode.
19886         (output_constant_pool_2): Update accordingly.
19887         (output_constant): Likewise.
19889 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19890             Alan Hayward  <alan.hayward@arm.com>
19891             David Sherwood  <david.sherwood@arm.com>
19893         * tree.h (SCALAR_FLOAT_TYPE_MODE): New macro.
19894         * builtins.c (expand_builtin_signbit): Use it instead of TYPE_MODE.
19895         * fold-const.c (fold_convert_const_real_from_fixed): Likewise.
19896         (native_encode_real): Likewise.
19897         (native_interpret_real): Likewise.
19898         * hsa-brig.c (emit_immediate_scalar_to_buffer): Likewise.
19899         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19901 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19902             Alan Hayward  <alan.hayward@arm.com>
19903             David Sherwood  <david.sherwood@arm.com>
19905         * optabs-libfuncs.c (gen_trunc_conv_libfunc): Use is_a
19906         <scalar_float_mode>.  Simplify.
19907         (gen_extend_conv_libfunc): Likewise.
19909 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19910             Alan Hayward  <alan.hayward@arm.com>
19911             David Sherwood  <david.sherwood@arm.com>
19913         * coretypes.h (scalar_float_mode): New type.
19914         * machmode.h (mode_traits::from_int): Use machine_mode if
19915         USE_ENUM_MODES is defined.
19916         (is_a): New function.
19917         (as_a): Likewise.
19918         (dyn_cast): Likewise.
19919         (scalar_float_mode): New class.
19920         (scalar_float_mode::includes_p): New function.
19921         (is_float_mode): Likewise.
19922         * gdbhooks.py (MachineModePrinter): New class.
19923         (build_pretty_printer): Use it for scalar_float_mode.
19924         * real.h (FLOAT_MODE_FORMAT): Use as_a <scalar_float_mode>.
19925         (format_helper::format_helper): Turn into a template.
19926         * genmodes.c (get_mode_class): New function.
19927         (emit_insn_modes_h): Give modes the class returned by get_mode_class,
19928         or machine_mode if none.
19929         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Use
19930         as_a <scalar_float_mode>.
19931         * dwarf2out.c (mem_loc_descriptor): Likewise.
19932         (insert_float): Likewise.
19933         (add_const_value_attribute): Likewise.
19934         * simplify-rtx.c (simplify_immed_subreg): Likewise.
19935         * optabs.c (expand_absneg_bit): Take a scalar_float_mode.
19936         (expand_unop): Update accordingly.
19937         (expand_abs_nojump): Likewise.
19938         (expand_copysign_absneg): Take a scalar_float_mode.
19939         (expand_copysign_bit): Likewise.
19940         (expand_copysign): Update accordingly.
19942 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19943             Alan Hayward  <alan.hayward@arm.com>
19944             David Sherwood  <david.sherwood@arm.com>
19946         * coretypes.h (opt_mode): New class.
19947         * machmode.h (opt_mode): Likewise.
19948         (opt_mode::else_void): New function.
19949         (opt_mode::require): Likewise.
19950         (opt_mode::exists): Likewise.
19951         (GET_MODE_WIDER_MODE): Turn into a function and return an opt_mode.
19952         (GET_MODE_2XWIDER_MODE): Likewise.
19953         (mode_iterator::get_wider): Update accordingly.
19954         (mode_iterator::get_2xwider): Likewise.
19955         (mode_iterator::get_known_wider): Likewise, turning into a template.
19956         * combine.c (make_extraction): Update use of GET_MODE_WIDER_MODE,
19957         forcing a wider mode to exist.
19958         * config/cr16/cr16.h (LONG_REG_P): Likewise.
19959         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
19960         * config/c6x/c6x.c (c6x_rtx_costs): Update use of
19961         GET_MODE_2XWIDER_MODE, forcing a wider mode to exist.
19962         * lower-subreg.c (init_lower_subreg): Likewise.
19963         * optabs-libfuncs.c (init_sync_libfuncs_1): Likewise, but not
19964         on the final iteration.
19965         * config/i386/i386.c (ix86_expand_set_or_movmem): Check whether
19966         a wider mode exists before asking for a move pattern.
19967         (get_mode_wider_vector): Update use of GET_MODE_WIDER_MODE,
19968         forcing a wider mode to exist.
19969         (expand_vselect_vconcat): Update use of GET_MODE_2XWIDER_MODE,
19970         returning false if no such mode exists.
19971         * config/ia64/ia64.c (expand_vselect_vconcat): Likewise.
19972         * config/mips/mips.c (mips_expand_vselect_vconcat): Likewise.
19973         * expmed.c (init_expmed_one_mode): Update use of GET_MODE_WIDER_MODE.
19974         Avoid checking for a MODE_INT if we already know the mode is not a
19975         SCALAR_INT_MODE_P.
19976         (extract_high_half): Update use of GET_MODE_WIDER_MODE,
19977         forcing a wider mode to exist.
19978         (expmed_mult_highpart_optab): Likewise.
19979         (expmed_mult_highpart): Likewise.
19980         * expr.c (expand_expr_real_2): Update use of GET_MODE_WIDER_MODE,
19981         using else_void.
19982         * lto-streamer-in.c (lto_input_mode_table): Likewise.
19983         * optabs-query.c (find_widening_optab_handler_and_mode): Likewise.
19984         * stor-layout.c (bit_field_mode_iterator::next_mode): Likewise.
19985         * internal-fn.c (expand_mul_overflow): Update use of
19986         GET_MODE_2XWIDER_MODE.
19987         * omp-low.c (omp_clause_aligned_alignment): Likewise.
19988         * tree-ssa-math-opts.c (convert_mult_to_widen): Update use of
19989         GET_MODE_WIDER_MODE.
19990         (convert_plusminus_to_widen): Likewise.
19991         * tree-switch-conversion.c (array_value_type): Likewise.
19992         * var-tracking.c (emit_note_insn_var_location): Likewise.
19993         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19994         Return false inside rather than outside the loop if no wider mode
19995         exists
19996         * optabs.c (expand_binop): Update use of GET_MODE_WIDER_MODE
19997         and GET_MODE_2XWIDER_MODE
19998         (can_compare_p): Use else_void.
19999         * gdbhooks.py (OptMachineModePrinter): New class.
20000         (build_pretty_printer): Use it for opt_mode.
20002 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20003             Alan Hayward  <alan.hayward@arm.com>
20004             David Sherwood  <david.sherwood@arm.com>
20006         * tree-switch-conversion.c (array_value_type): Only read TYPE_MODE
20007         once.  Use get_narrowest_mode instead of GET_CLASS_NARROWEST_MODE.
20009 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20010             Alan Hayward  <alan.hayward@arm.com>
20011             David Sherwood  <david.sherwood@arm.com>
20013         * machmode.h (mode_traits): New structure.
20014         (get_narrowest_mode): New function.
20015         (mode_iterator::start): Likewise.
20016         (mode_iterator::iterate_p): Likewise.
20017         (mode_iterator::get_wider): Likewise.
20018         (mode_iterator::get_known_wider): Likewise.
20019         (mode_iterator::get_2xwider): Likewise.
20020         (FOR_EACH_MODE_IN_CLASS): New mode iterator.
20021         (FOR_EACH_MODE): Likewise.
20022         (FOR_EACH_MODE_FROM): Likewise.
20023         (FOR_EACH_MODE_UNTIL): Likewise.
20024         (FOR_EACH_WIDER_MODE): Likewise.
20025         (FOR_EACH_2XWIDER_MODE): Likewise.
20026         * builtins.c (expand_builtin_strlen): Use new mode iterators.
20027         * combine.c (simplify_comparison): Likewise
20028         * config/i386/i386.c (type_natural_mode): Likewise.
20029         * cse.c (cse_insn): Likewise.
20030         * dse.c (find_shift_sequence): Likewise.
20031         * emit-rtl.c (init_derived_machine_modes): Likewise.
20032         (init_emit_once): Likewise.
20033         * explow.c (hard_function_value): Likewise.
20034         * expmed.c (extract_fixed_bit_field_1): Likewise.
20035         (extract_bit_field_1): Likewise.
20036         (expand_divmod): Likewise.
20037         (emit_store_flag_1): Likewise.
20038         * expr.c (init_expr_target): Likewise.
20039         (convert_move): Likewise.
20040         (alignment_for_piecewise_move): Likewise.
20041         (widest_int_mode_for_size): Likewise.
20042         (emit_block_move_via_movmem): Likewise.
20043         (copy_blkmode_to_reg): Likewise.
20044         (set_storage_via_setmem): Likewise.
20045         (compress_float_constant): Likewise.
20046         * omp-low.c (omp_clause_aligned_alignment): Likewise.
20047         * optabs-query.c (get_best_extraction_insn): Likewise.
20048         * optabs.c (expand_binop): Likewise.
20049         (expand_twoval_unop): Likewise.
20050         (expand_twoval_binop): Likewise.
20051         (widen_leading): Likewise.
20052         (widen_bswap): Likewise.
20053         (expand_parity): Likewise.
20054         (expand_unop): Likewise.
20055         (prepare_cmp_insn): Likewise.
20056         (prepare_float_lib_cmp): Likewise.
20057         (expand_float): Likewise.
20058         (expand_fix): Likewise.
20059         (expand_sfix_optab): Likewise.
20060         * postreload.c (move2add_use_add2_insn): Likewise.
20061         * reg-stack.c (reg_to_stack): Likewise.
20062         * reginfo.c (choose_hard_reg_mode): Likewise.
20063         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
20064         * stor-layout.c (mode_for_size): Likewise.
20065         (smallest_mode_for_size): Likewise.
20066         (mode_for_vector): Likewise.
20067         (finish_bitfield_representative): Likewise.
20068         * tree-ssa-math-opts.c (target_supports_divmod_p): Likewise.
20069         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
20070         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
20071         * var-tracking.c (prepare_call_arguments): Likewise.
20073 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20074             Alan Hayward  <alan.hayward@arm.com>
20075             David Sherwood  <david.sherwood@arm.com>
20077         * genconditions.c (write_header): Add a "#define USE_ENUM_MODES".
20078         * genmodes.c (emit_insn_modes_h): Define FOOmode to E_FOOmode if
20079         USE_ENUM_MODES is defined and to ((void) 0, E_FOOmode) otherwise.
20080         * machmode.h (mode_size): Move earlier in file.
20081         (mode_precision): Likewise.
20082         (mode_inner): Likewise.
20083         (mode_nunits): Likewise.
20084         (mode_unit_size): Likewise.
20085         (unit_unit_precision): Likewise.
20086         (mode_wider): Likewise.
20087         (mode_2xwider): Likewise.
20088         (machine_mode): New class.
20089         (mode_to_bytes): New function.
20090         (mode_to_bits): Likewise.
20091         (mode_to_precision): Likewise.
20092         (mode_to_inner): Likewise.
20093         (mode_to_unit_size): Likewise.
20094         (mode_to_unit_precision): Likewise.
20095         (mode_to_nunits): Likewise.
20096         (GET_MODE_SIZE): Use mode_to_bytes.
20097         (GET_MODE_BITSIZE): Use mode_to_bits.
20098         (GET_MODE_PRECISION): Use mode_to_precision.
20099         (GET_MODE_INNER): Use mode_to_inner.
20100         (GET_MODE_UNIT_SIZE): Use mode_to_unit_size.
20101         (GET_MODE_UNIT_PRECISION): Use mode_to_unit_precision.
20102         (GET_MODE_NUNITS): Use mode_to_nunits.
20103         * system.h (ALWAYS_INLINE): New macro.
20104         * config/powerpcspe/powerpcspe-c.c
20105         (altivec_resolve_overloaded_builtin): Use machine_mode instead of
20106         int for arg1_mode and arg2_mode.
20108 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20109             Alan Hayward  <alan.hayward@arm.com>
20110             David Sherwood  <david.sherwood@arm.com>
20112         * config/aarch64/aarch64-builtins.c (aarch64_simd_builtin_std_type):
20113         Prefix mode names with E_ in case statements.
20114         * config/aarch64/aarch64-elf.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20115         * config/aarch64/aarch64.c (aarch64_split_simd_combine): Likewise.
20116         (aarch64_split_simd_move): Likewise.
20117         (aarch64_gen_storewb_pair): Likewise.
20118         (aarch64_gen_loadwb_pair): Likewise.
20119         (aarch64_gen_store_pair): Likewise.
20120         (aarch64_gen_load_pair): Likewise.
20121         (aarch64_get_condition_code_1): Likewise.
20122         (aarch64_constant_pool_reload_icode): Likewise.
20123         (get_rsqrte_type): Likewise.
20124         (get_rsqrts_type): Likewise.
20125         (get_recpe_type): Likewise.
20126         (get_recps_type): Likewise.
20127         (aarch64_gimplify_va_arg_expr): Likewise.
20128         (aarch64_simd_container_mode): Likewise.
20129         (aarch64_emit_load_exclusive): Likewise.
20130         (aarch64_emit_store_exclusive): Likewise.
20131         (aarch64_expand_compare_and_swap): Likewise.
20132         (aarch64_gen_atomic_cas): Likewise.
20133         (aarch64_emit_bic): Likewise.
20134         (aarch64_emit_atomic_swap): Likewise.
20135         (aarch64_emit_atomic_load_op): Likewise.
20136         (aarch64_evpc_trn): Likewise.
20137         (aarch64_evpc_uzp): Likewise.
20138         (aarch64_evpc_zip): Likewise.
20139         (aarch64_evpc_ext): Likewise.
20140         (aarch64_evpc_rev): Likewise.
20141         (aarch64_evpc_dup): Likewise.
20142         (aarch64_gen_ccmp_first): Likewise.
20143         (aarch64_gen_ccmp_next): Likewise.
20144         * config/alpha/alpha.c (alpha_scalar_mode_supported_p): Likewise.
20145         (alpha_emit_xfloating_libcall): Likewise.
20146         (emit_insxl): Likewise.
20147         (alpha_arg_type): Likewise.
20148         * config/arc/arc.c (arc_vector_mode_supported_p): Likewise.
20149         (arc_preferred_simd_mode): Likewise.
20150         (arc_secondary_reload): Likewise.
20151         (get_arc_condition_code): Likewise.
20152         (arc_print_operand): Likewise.
20153         (arc_legitimate_constant_p): Likewise.
20154         * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20155         * config/arc/arc.md (casesi_load): Likewise.
20156         (casesi_compact_jump): Likewise.
20157         * config/arc/predicates.md (proper_comparison_operator): Likewise.
20158         (cc_use_register): Likewise.
20159         * config/arm/aout.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20160         * config/arm/arm-builtins.c (arm_simd_builtin_std_type): Likewise.
20161         (arm_init_iwmmxt_builtins): Likewise.
20162         * config/arm/arm.c (thumb1_size_rtx_costs): Likewise.
20163         (neon_expand_vector_init): Likewise.
20164         (arm_attr_length_move_neon): Likewise.
20165         (maybe_get_arm_condition_code): Likewise.
20166         (arm_emit_vector_const): Likewise.
20167         (arm_preferred_simd_mode): Likewise.
20168         (arm_output_iwmmxt_tinsr): Likewise.
20169         (thumb1_output_casesi): Likewise.
20170         (thumb2_output_casesi): Likewise.
20171         (arm_emit_load_exclusive): Likewise.
20172         (arm_emit_store_exclusive): Likewise.
20173         (arm_expand_compare_and_swap): Likewise.
20174         (arm_evpc_neon_vuzp): Likewise.
20175         (arm_evpc_neon_vzip): Likewise.
20176         (arm_evpc_neon_vrev): Likewise.
20177         (arm_evpc_neon_vtrn): Likewise.
20178         (arm_evpc_neon_vext): Likewise.
20179         (arm_validize_comparison): Likewise.
20180         * config/arm/neon.md (neon_vc<cmp_op><mode>): Likewise.
20181         * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Likewise.
20182         * config/avr/avr.c (avr_rtx_costs_1): Likewise.
20183         * config/c6x/c6x.c (c6x_vector_mode_supported_p): Likewise.
20184         (c6x_preferred_simd_mode): Likewise.
20185         * config/epiphany/epiphany.c (get_epiphany_condition_code): Likewise.
20186         (epiphany_rtx_costs): Likewise.
20187         * config/epiphany/predicates.md (proper_comparison_operator):
20188         Likewise.
20189         * config/frv/frv.c (condexec_memory_operand): Likewise.
20190         (frv_emit_move): Likewise.
20191         (output_move_single): Likewise.
20192         (output_condmove_single): Likewise.
20193         (frv_hard_regno_mode_ok): Likewise.
20194         (frv_matching_accg_mode): Likewise.
20195         * config/h8300/h8300.c (split_adds_subs): Likewise.
20196         (h8300_rtx_costs): Likewise.
20197         (h8300_print_operand): Likewise.
20198         (compute_mov_length): Likewise.
20199         (output_logical_op): Likewise.
20200         (compute_logical_op_length): Likewise.
20201         (compute_logical_op_cc): Likewise.
20202         (h8300_shift_needs_scratch_p): Likewise.
20203         (output_a_shift): Likewise.
20204         (compute_a_shift_length): Likewise.
20205         (compute_a_shift_cc): Likewise.
20206         (expand_a_rotate): Likewise.
20207         (output_a_rotate): Likewise.
20208         * config/i386/i386.c (classify_argument): Likewise.
20209         (function_arg_advance_32): Likewise.
20210         (function_arg_32): Likewise.
20211         (function_arg_64): Likewise.
20212         (function_value_64): Likewise.
20213         (ix86_gimplify_va_arg): Likewise.
20214         (ix86_legitimate_constant_p): Likewise.
20215         (put_condition_code): Likewise.
20216         (split_double_mode): Likewise.
20217         (ix86_avx256_split_vector_move_misalign): Likewise.
20218         (ix86_expand_vector_logical_operator): Likewise.
20219         (ix86_split_idivmod): Likewise.
20220         (ix86_expand_adjust_ufix_to_sfix_si): Likewise.
20221         (ix86_build_const_vector): Likewise.
20222         (ix86_build_signbit_mask): Likewise.
20223         (ix86_match_ccmode): Likewise.
20224         (ix86_cc_modes_compatible): Likewise.
20225         (ix86_expand_branch): Likewise.
20226         (ix86_expand_sse_cmp): Likewise.
20227         (ix86_expand_sse_movcc): Likewise.
20228         (ix86_expand_int_sse_cmp): Likewise.
20229         (ix86_expand_vec_perm_vpermi2): Likewise.
20230         (ix86_expand_vec_perm): Likewise.
20231         (ix86_expand_sse_unpack): Likewise.
20232         (ix86_expand_int_addcc): Likewise.
20233         (ix86_split_to_parts): Likewise.
20234         (ix86_vectorize_builtin_gather): Likewise.
20235         (ix86_vectorize_builtin_scatter): Likewise.
20236         (avx_vpermilp_parallel): Likewise.
20237         (inline_memory_move_cost): Likewise.
20238         (ix86_tieable_integer_mode_p): Likewise.
20239         (x86_maybe_negate_const_int): Likewise.
20240         (ix86_expand_vector_init_duplicate): Likewise.
20241         (ix86_expand_vector_init_one_nonzero): Likewise.
20242         (ix86_expand_vector_init_one_var): Likewise.
20243         (ix86_expand_vector_init_concat): Likewise.
20244         (ix86_expand_vector_init_interleave): Likewise.
20245         (ix86_expand_vector_init_general): Likewise.
20246         (ix86_expand_vector_set): Likewise.
20247         (ix86_expand_vector_extract): Likewise.
20248         (emit_reduc_half): Likewise.
20249         (ix86_emit_i387_round): Likewise.
20250         (ix86_mangle_type): Likewise.
20251         (ix86_expand_round_sse4): Likewise.
20252         (expand_vec_perm_blend): Likewise.
20253         (canonicalize_vector_int_perm): Likewise.
20254         (ix86_expand_vec_one_operand_perm_avx512): Likewise.
20255         (expand_vec_perm_1): Likewise.
20256         (expand_vec_perm_interleave3): Likewise.
20257         (expand_vec_perm_even_odd_pack): Likewise.
20258         (expand_vec_perm_even_odd_1): Likewise.
20259         (expand_vec_perm_broadcast_1): Likewise.
20260         (ix86_vectorize_vec_perm_const_ok): Likewise.
20261         (ix86_expand_vecop_qihi): Likewise.
20262         (ix86_expand_mul_widen_hilo): Likewise.
20263         (ix86_expand_sse2_abs): Likewise.
20264         (ix86_expand_pextr): Likewise.
20265         (ix86_expand_pinsr): Likewise.
20266         (ix86_preferred_simd_mode): Likewise.
20267         (ix86_simd_clone_compute_vecsize_and_simdlen): Likewise.
20268         * config/i386/sse.md (*andnot<mode>3): Likewise.
20269         (<mask_codefor><code><mode>3<mask_name>): Likewise.
20270         (*<code><mode>3): Likewise.
20271         * config/ia64/ia64.c (ia64_expand_vecint_compare): Likewise.
20272         (ia64_expand_atomic_op): Likewise.
20273         (ia64_arg_type): Likewise.
20274         (ia64_mode_to_int): Likewise.
20275         (ia64_scalar_mode_supported_p): Likewise.
20276         (ia64_vector_mode_supported_p): Likewise.
20277         (expand_vec_perm_broadcast): Likewise.
20278         * config/iq2000/iq2000.c (iq2000_move_1word): Likewise.
20279         (iq2000_function_arg_advance): Likewise.
20280         (iq2000_function_arg): Likewise.
20281         * config/m32c/m32c.c (m32c_preferred_reload_class): Likewise.
20282         * config/m68k/m68k.c (output_dbcc_and_branch): Likewise.
20283         (m68k_libcall_value): Likewise.
20284         (m68k_function_value): Likewise.
20285         (sched_attr_op_type): Likewise.
20286         * config/mcore/mcore.c (mcore_output_move): Likewise.
20287         * config/microblaze/microblaze.c (microblaze_function_arg_advance):
20288         Likewise.
20289         (microblaze_function_arg): Likewise.
20290         * config/mips/mips.c (mips16_build_call_stub): Likewise.
20291         (mips_print_operand): Likewise.
20292         (mips_mode_ok_for_mov_fmt_p): Likewise.
20293         (mips_vector_mode_supported_p): Likewise.
20294         (mips_preferred_simd_mode): Likewise.
20295         (mips_expand_vpc_loongson_even_odd): Likewise.
20296         (mips_expand_vec_unpack): Likewise.
20297         (mips_expand_vi_broadcast): Likewise.
20298         (mips_expand_vector_init): Likewise.
20299         (mips_expand_vec_reduc): Likewise.
20300         (mips_expand_msa_cmp): Likewise.
20301         * config/mips/mips.md (casesi_internal_mips16_<mode>): Likewise.
20302         * config/mn10300/mn10300.c (mn10300_print_operand): Likewise.
20303         (cc_flags_for_mode): Likewise.
20304         * config/msp430/msp430.c (msp430_print_operand): Likewise.
20305         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Likewise.
20306         (nds32_output_casesi_pc_relative): Likewise.
20307         * config/nds32/nds32.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20308         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Likewise.
20309         (nvptx_gen_unpack): Likewise.
20310         (nvptx_gen_pack): Likewise.
20311         (nvptx_gen_shuffle): Likewise.
20312         (nvptx_gen_wcast): Likewise.
20313         (nvptx_preferred_simd_mode): Likewise.
20314         * config/pa/pa.c (pa_secondary_reload): Likewise.
20315         * config/pa/predicates.md (base14_operand): Likewise.
20316         * config/powerpcspe/powerpcspe-c.c
20317         (altivec_resolve_overloaded_builtin): Likewise.
20318         * config/powerpcspe/powerpcspe.c (rs6000_setup_reg_addr_masks):
20319         Likewise.
20320         (rs6000_preferred_simd_mode): Likewise.
20321         (output_vec_const_move): Likewise.
20322         (rs6000_expand_vector_extract): Likewise.
20323         (rs6000_split_vec_extract_var): Likewise.
20324         (reg_offset_addressing_ok_p): Likewise.
20325         (rs6000_legitimate_offset_address_p): Likewise.
20326         (rs6000_legitimize_address): Likewise.
20327         (rs6000_emit_set_const): Likewise.
20328         (rs6000_const_vec): Likewise.
20329         (rs6000_emit_move): Likewise.
20330         (spe_build_register_parallel): Likewise.
20331         (rs6000_darwin64_record_arg_recurse): Likewise.
20332         (swap_selector_for_mode): Likewise.
20333         (spe_init_builtins): Likewise.
20334         (paired_init_builtins): Likewise.
20335         (altivec_init_builtins): Likewise.
20336         (do_load_for_compare): Likewise.
20337         (rs6000_generate_compare): Likewise.
20338         (rs6000_expand_float128_convert): Likewise.
20339         (emit_load_locked): Likewise.
20340         (emit_store_conditional): Likewise.
20341         (rs6000_output_function_epilogue): Likewise.
20342         (rs6000_handle_altivec_attribute): Likewise.
20343         (rs6000_function_value): Likewise.
20344         (emit_fusion_gpr_load): Likewise.
20345         (emit_fusion_p9_load): Likewise.
20346         (emit_fusion_p9_store): Likewise.
20347         * config/powerpcspe/predicates.md (easy_fp_constant): Likewise.
20348         (fusion_gpr_mem_load): Likewise.
20349         (fusion_addis_mem_combo_load): Likewise.
20350         (fusion_addis_mem_combo_store): Likewise.
20351         * config/rs6000/predicates.md (easy_fp_constant): Likewise.
20352         (fusion_gpr_mem_load): Likewise.
20353         (fusion_addis_mem_combo_load): Likewise.
20354         (fusion_addis_mem_combo_store): Likewise.
20355         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
20356         Likewise.
20357         * config/rs6000/rs6000-string.c (do_load_for_compare): Likewise.
20358         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Likewise.
20359         (rs6000_preferred_simd_mode): Likewise.
20360         (output_vec_const_move): Likewise.
20361         (rs6000_expand_vector_extract): Likewise.
20362         (rs6000_split_vec_extract_var): Likewise.
20363         (reg_offset_addressing_ok_p): Likewise.
20364         (rs6000_legitimate_offset_address_p): Likewise.
20365         (rs6000_legitimize_address): Likewise.
20366         (rs6000_emit_set_const): Likewise.
20367         (rs6000_const_vec): Likewise.
20368         (rs6000_emit_move): Likewise.
20369         (rs6000_darwin64_record_arg_recurse): Likewise.
20370         (swap_selector_for_mode): Likewise.
20371         (paired_init_builtins): Likewise.
20372         (altivec_init_builtins): Likewise.
20373         (rs6000_expand_float128_convert): Likewise.
20374         (emit_load_locked): Likewise.
20375         (emit_store_conditional): Likewise.
20376         (rs6000_output_function_epilogue): Likewise.
20377         (rs6000_handle_altivec_attribute): Likewise.
20378         (rs6000_function_value): Likewise.
20379         (emit_fusion_gpr_load): Likewise.
20380         (emit_fusion_p9_load): Likewise.
20381         (emit_fusion_p9_store): Likewise.
20382         * config/rx/rx.c (rx_gen_move_template): Likewise.
20383         (flags_from_mode): Likewise.
20384         * config/s390/predicates.md (s390_alc_comparison): Likewise.
20385         (s390_slb_comparison): Likewise.
20386         * config/s390/s390.c (s390_handle_vectorbool_attribute): Likewise.
20387         (s390_vector_mode_supported_p): Likewise.
20388         (s390_cc_modes_compatible): Likewise.
20389         (s390_match_ccmode_set): Likewise.
20390         (s390_canonicalize_comparison): Likewise.
20391         (s390_emit_compare_and_swap): Likewise.
20392         (s390_branch_condition_mask): Likewise.
20393         (s390_rtx_costs): Likewise.
20394         (s390_secondary_reload): Likewise.
20395         (__SECONDARY_RELOAD_CASE): Likewise.
20396         (s390_expand_cs): Likewise.
20397         (s390_preferred_simd_mode): Likewise.
20398         * config/s390/vx-builtins.md (vec_packsu_u<mode>): Likewise.
20399         * config/sh/sh.c (sh_print_operand): Likewise.
20400         (dump_table): Likewise.
20401         (sh_secondary_reload): Likewise.
20402         * config/sh/sh.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20403         * config/sh/sh.md (casesi_worker_1): Likewise.
20404         (casesi_worker_2): Likewise.
20405         * config/sparc/predicates.md (icc_comparison_operator): Likewise.
20406         (fcc_comparison_operator): Likewise.
20407         * config/sparc/sparc.c (sparc_expand_move): Likewise.
20408         (emit_soft_tfmode_cvt): Likewise.
20409         (sparc_preferred_simd_mode): Likewise.
20410         (output_cbranch): Likewise.
20411         (sparc_print_operand): Likewise.
20412         (sparc_expand_vec_perm_bmask): Likewise.
20413         (vector_init_bshuffle): Likewise.
20414         * config/spu/spu.c (spu_scalar_mode_supported_p): Likewise.
20415         (spu_vector_mode_supported_p): Likewise.
20416         (spu_expand_insv): Likewise.
20417         (spu_emit_branch_or_set): Likewise.
20418         (spu_handle_vector_attribute): Likewise.
20419         (spu_builtin_splats): Likewise.
20420         (spu_builtin_extract): Likewise.
20421         (spu_builtin_promote): Likewise.
20422         (spu_expand_sign_extend): Likewise.
20423         * config/tilegx/tilegx.c (tilegx_scalar_mode_supported_p): Likewise.
20424         (tilegx_simd_int): Likewise.
20425         * config/tilepro/tilepro.c (tilepro_scalar_mode_supported_p): Likewise.
20426         (tilepro_simd_int): Likewise.
20427         * config/v850/v850.c (const_double_split): Likewise.
20428         (v850_print_operand): Likewise.
20429         (ep_memory_offset): Likewise.
20430         * config/vax/vax.c (vax_rtx_costs): Likewise.
20431         (vax_output_int_move): Likewise.
20432         (vax_output_int_add): Likewise.
20433         (vax_output_int_subtract): Likewise.
20434         * config/visium/predicates.md (visium_branch_operator): Likewise.
20435         * config/visium/visium.c (rtx_ok_for_offset_p): Likewise.
20436         (visium_print_operand_address): Likewise.
20437         * config/visium/visium.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20438         * config/xtensa/xtensa.c (xtensa_mem_offset): Likewise.
20439         (xtensa_expand_conditional_branch): Likewise.
20440         (xtensa_copy_incoming_a7): Likewise.
20441         (xtensa_output_literal): Likewise.
20442         * dfp.c (decimal_real_maxval): Likewise.
20443         * targhooks.c (default_libgcc_floating_mode_supported_p): Likewise.
20445 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20446             Alan Hayward  <alan.hayward@arm.com>
20447             David Sherwood  <david.sherwood@arm.com>
20449         * genmodes.c (mode_size_inline): Add an E_ prefix to mode names.
20450         (mode_nunits_inline): Likewise.
20451         (mode_inner_inline): Likewise.
20452         (mode_unit_size_inline): Likewise.
20453         (mode_unit_precision_inline): Likewise.
20454         (emit_insn_modes_h): Likewise.  Also emit a #define of the
20455         unprefixed name.
20456         (emit_mode_wider): Add an E_ prefix to mode names.
20457         (emit_mode_complex): Likewise.
20458         (emit_mode_inner): Likewise.
20459         (emit_mode_adjustments): Likewise.
20460         (emit_mode_int_n): Likewise.
20461         * config/aarch64/aarch64-builtins.c (v8qi_UP, v4hi_UP, v4hf_UP)
20462         (v2si_UP, v2sf_UP, v1df_UP, di_UP, df_UP, v16qi_UP, v8hi_UP, v8hf_UP)
20463         (v4si_UP, v4sf_UP, v2di_UP, v2df_UP, ti_UP, oi_UP, ci_UP, xi_UP)
20464         (si_UP, sf_UP, hi_UP, hf_UP, qi_UP): Likewise.
20465         (CRC32_BUILTIN, ENTRY): Likewise.
20466         * config/aarch64/aarch64.c (aarch64_push_regs): Likewise.
20467         (aarch64_pop_regs): Likewise.
20468         (aarch64_process_components): Likewise.
20469         * config/alpha/alpha.c (alpha_emit_conditional_move): Likewise.
20470         * config/arm/arm-builtins.c (v8qi_UP, v4hi_UP, v4hf_UP, v2si_UP)
20471         (v2sf_UP, di_UP, v16qi_UP, v8hi_UP, v8hf_UP, v4si_UP, v4sf_UP)
20472         (v2di_UP, ti_UP, ei_UP, oi_UP, hf_UP, si_UP, void_UP): Likewise.
20473         * config/arm/arm.c (arm_init_libfuncs): Likewise.
20474         * config/i386/i386-builtin-types.awk (ix86_builtin_type_vect_mode):
20475         Likewise.
20476         * config/i386/i386-builtin.def (pcmpestr): Likewise.
20477         (pcmpistr): Likewise.
20478         * config/microblaze/microblaze.c (double_memory_operand): Likewise.
20479         * config/mmix/mmix.c (mmix_output_condition): Likewise.
20480         * config/powerpcspe/powerpcspe.c (rs6000_init_hard_regno_mode_ok):
20481         Likewise.
20482         * config/rl78/rl78.c (mduc_regs): Likewise.
20483         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
20484         (htm_expand_builtin): Likewise.
20485         * config/sh/sh.h (REGISTER_NATURAL_MODE): Likewise.
20486         * config/sparc/sparc.c (emit_save_or_restore_regs): Likewise.
20487         * config/xtensa/xtensa.c (print_operand): Likewise.
20488         * expmed.h (NUM_MODE_PARTIAL_INT): Likewise.
20489         (NUM_MODE_VECTOR_INT): Likewise.
20490         * genoutput.c (null_operand): Likewise.
20491         (output_operand_data): Likewise.
20492         * genrecog.c (print_parameter_value): Likewise.
20493         * lra.c (debug_operand_data): Likewise.
20495 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20497         * dwarf2out.c (scompare_loc_descriptor_wide)
20498         (scompare_loc_descriptor_narrow): New functions, split out from...
20499         (scompare_loc_descriptor): ...here.
20500         * expmed.c (emit_store_flag_int): New function, split out from...
20501         (emit_store_flag): ...here.
20503 2017-08-30  Richard Biener  <rguenther@suse.de>
20505         * dwarf2out.c (dwarf2out_finish): Remove setting AT_pubnames.
20506         (dwarf2out_early_finish): Move setting of AT_pubnames from
20507         early debug output to early finish.
20509 2017-08-30  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
20511         * gcc/config/msp430/msp430.h: Pass -mcode/data-region to the linker
20512         and -mdata-region to the assembler.
20514 2017-08-30  Richard Biener  <rguenther@suse.de>
20516         * dwarf2out.c (add_dwarf_attr): Check we don't add duplicate
20517         attributes.
20518         (gen_subprogram_die): Add DW_AT_object_pointer only early.
20519         (dwarf2out_early_global_decl): Only generate a DIE for the
20520         abstract origin if it doesn't already exist or is a declaration DIE.
20521         (resolve_addr): Do not add the linkage name twice when
20522         generating a stub DIE for the DW_TAG_GNU_call_site target.
20524 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20526         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
20527         Use machine_mode rather than int for arg1_mode.
20529 2017-08-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
20531         PR target/82015
20532         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Insure
20533         that the second argument of the built-in functions to unpack
20534         128-bit scalar types to 64-bit values is 0 or 1.  Change to use a
20535         switch statement instead a lot of if statements.
20536         * config/rs6000/rs6000.md (unpack<mode>, FMOVE128_VSX iterator):
20537         Allow 64-bit values to be in Altivec registers as well as
20538         traditional floating point registers.
20539         (pack<mode>, FMOVE128_VSX iterator): Likewise.
20541 2017-08-29  Alexander Monakov  <amonakov@ispras.ru>
20543         * ira-costs.c (record_address_regs): Handle both operands of PLUS for
20544         MAX_REGS_PER_ADDRESS == 1.
20546 2017-08-29  Uros Bizjak  <ubizjak@gmail.com>
20548         * config/i386/i386.opt (flag_fentry): Do not init to -1.
20549         * config/i386/i386.c (ix86_option_override_internal): Simplify
20550         setting of opts->x_flag_entry.
20552 2017-08-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20553             Jakub Jelinek  <jakub@redhat.com>
20554             Richard Biener  <rguenther@suse.de>
20556         PR tree-optimization/81503
20557         * gimple-ssa-strength-reduction.c (replace_mult_candidate): Ensure
20558         folded constant fits in the target type; reorder tests for clarity.
20560 2017-08-29  Martin Liska  <mliska@suse.cz>
20562         * passes.def: Include pass_lower_switch.
20563         * stmt.c (dump_case_nodes): Remove and move to
20564         tree-switch-conversion.
20565         (case_values_threshold): Likewise.
20566         (expand_switch_as_decision_tree_p): Likewise.
20567         (emit_case_decision_tree): Likewise.
20568         (expand_case): Likewise.
20569         (balance_case_nodes): Likewise.
20570         (node_has_low_bound): Likewise.
20571         (node_has_high_bound): Likewise.
20572         (node_is_bounded): Likewise.
20573         (emit_case_nodes): Likewise.
20574         (struct simple_case_node): New struct.
20575         (add_case_node): Remove.
20576         (emit_case_dispatch_table): Use vector instead of case_list.
20577         (reset_out_edges_aux): Remove.
20578         (compute_cases_per_edge): Likewise.
20579         (expand_case): Build list of simple_case_node.
20580         (expand_sjlj_dispatch_table): Use it.
20581         * tree-switch-conversion.c (struct case_node): Moved from
20582         stmt.c and adjusted.
20583         (emit_case_nodes): Likewise.
20584         (node_has_low_bound): Likewise.
20585         (node_has_high_bound): Likewise.
20586         (node_is_bounded): Likewise.
20587         (case_values_threshold): Likewise.
20588         (reset_out_edges_aux): Likewise.
20589         (compute_cases_per_edge): Likewise.
20590         (add_case_node): Likewise.
20591         (dump_case_nodes): Likewise.
20592         (balance_case_nodes): Likewise.
20593         (expand_switch_as_decision_tree_p): Likewise.
20594         (emit_jump): Likewise.
20595         (emit_case_decision_tree): Likewise.
20596         (try_switch_expansion): Likewise.
20597         (do_jump_if_equal): Likewise.
20598         (emit_cmp_and_jump_insns): Likewise.
20599         (fix_phi_operands_for_edge): New function.
20600         (record_phi_operand_mapping): Likewise.
20601         (class pass_lower_switch): New pass.
20602         (pass_lower_switch::execute): New function.
20603         (make_pass_lower_switch): Likewise.
20604         (conditional_probability):
20605         * timevar.def: Add TV_TREE_SWITCH_LOWERING.
20606         * tree-pass.h: Add make_pass_lower_switch.
20608 2017-08-29  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
20610         PR target/80993
20611         * gcc/config/msp430/msp430.c (msp430_attr): Mark interrupt
20612         handlers as used.
20614 2017-08-29  Richard Biener  <rguenther@suse.de>
20616         * dwarf2out.c (add_dwarf_attr): When checking is enabled verify
20617         we do not add a DW_AT_inline attribute twice.
20618         (gen_subprogram_die): Remove code setting DW_AT_inline on
20619         DECL_ABSTRACT_P nodes.
20621 2017-08-29  Richard Sandiford  <richard.sandiford@linaro.org>
20623         * gimplify.c (gimplify_call_expr): Copy the nothrow flag to
20624         calls to internal functions.
20625         (gimplify_modify_expr): Likewise.
20626         * tree-call-cdce.c (use_internal_fn): Likewise.
20627         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
20628         (convert_to_divmod): Set the nothrow flag.
20629         * tree-if-conv.c (predicate_mem_writes):  Likewise.
20630         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
20631         (vectorizable_call): Likewise.
20632         (vectorizable_store): Likewise.
20633         (vectorizable_load): Likewise.
20634         * tree-vect-patterns.c (vect_recog_pow_pattern): Likewise.
20635         (vect_recog_mask_conversion_pattern): Likewise.
20637 2017-08-29  Martin Liska  <mliska@suse.cz>
20639         PR other/39851
20640         * gcc.c (driver_handle_option): Add new argument.
20641         * opts-common.c (handle_option): Pass
20642         target_option_override_hook.
20643         * opts-global.c (lang_handle_option): Add new option.
20644         (set_default_handlers):  Add new argument.
20645         (decode_options): Likewise.
20646         * opts.c (target_handle_option): Likewise.
20647         (common_handle_option): Call target_option_override_hook.
20648         * opts.h (struct cl_option_handler_func): Add hook for
20649         target option override.
20650         (struct cl_option_handlers): Likewise.
20651         (set_default_handlers): Add new argument.
20652         (decode_options): Likewise.
20653         (common_handle_option): Likewise.
20654         (target_handle_option): Likewise.
20655         * toplev.c (toplev::main): Pass targetm.target_option.override
20656         hook.
20658 2017-08-29  Richard Biener  <rguenther@suse.de>
20659         Dominik Infuehr <dominik.infuehr@theobroma-systems.com>
20661         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Properly confine
20662         life to the active subtree.
20664 2017-08-28  Jeff Law  <law@redhat.com>
20666         * tree-ssa-dom.c (edge_info::record_simple_equiv): Call
20667         derive_equivalences.
20668         (derive_equivalences_from_bit_ior, record_temporary_equivalences):
20669         Code moved into....
20670         (edge_info::derive_equivalences): New private member function
20672         * tree-ssa-dom.c (class edge_info): Changed from a struct
20673         to a class.  Add ctor/dtor, methods and data members.
20674         (edge_info::edge_info): Renamed from allocate_edge_info.
20675         Initialize additional members.
20676         (edge_info::~edge_info): New.
20677         (free_dom_edge_info): Delete the edge info.
20678         (record_edge_info): Use new class & associated member functions.
20679         Tighten forms for testing for edge equivalences.
20680         (record_temporary_equivalences): Iterate over the simple
20681         equivalences rather than assuming there's only one per edge.
20682         (cprop_into_successor_phis): Iterate over the simple
20683         equivalences rather than assuming there's only one per edge.
20684         (optimize_stmt): Use operand_equal_p rather than pointer
20685         equality for mini-DSE code.
20687 2017-08-28  Nathan Sidwell  <nathan@acm.org>
20689         * gcc.c (execute): Fold SIGPIPE handling into switch
20690         statement.  Adjust internal error message.
20692 2017-08-28  Richard Biener  <rguenther@suse.de>
20694         PR debug/81993
20695         * dwarf2out.c (gen_remaining_tmpl_value_param_die_attributes):
20696         Do nothing for removed DIEs.
20698 2017-08-28  Richard Biener  <rguenther@suse.de>
20700         PR tree-optimization/81977
20701         * tree-ssa-sccvn.c (vn_reference_lookup_3): Fix look through
20702         memcpy.
20704 2017-08-28  Alexander Monakov  <amonakov@ispras.ru>
20706         PR target/80640
20707         * doc/md.texi (mem_thread_fence): Remove mention of mode.  Rewrite.
20708         * optabs.c (expand_mem_thread_fence): Emit a compiler barrier when
20709         using targetm.gen_mem_thread_fence.
20711 2017-08-27  Uros Bizjak  <ubizjak@gmail.com>
20713         PR target/81995
20714         * config/i386/i386.md (*<btsc><mode>): Change operand 2
20715         predicate to register_operand.  Reorder operands.
20716         (*btr<mode>): Ditto.
20717         (*<btsc><mode>_mask): Change operand 3 predicate to register_operand.
20718         (*btr<mode>_mask): Ditto.
20720 2017-08-25  Steven Munroe  <munroesj@gcc.gnu.org>
20722         * config.gcc (powerpc*-*-*): Add xmmintrin.h and mm_malloc.h.
20723         * config/rs6000/xmmintrin.h: New file.
20724         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include xmmintrin.h.
20726 2017-08-25  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20728         PR target/81504
20729         * config/rs6000/rs6000-p8swap.c (find_alignment_op): Add reference
20730         parameter and_insn and return it.
20731         (recombine_lvx_pattern): Insert a copy to ensure availability of
20732         the base register of the copied masking operation at the point of
20733         the instruction replacement.
20734         (recombine_stvx_pattern): Likewise.
20736 2017-08-25  Michael Meissner  <meissner@linux.vnet.ibm.com>
20738         * config/rs6000/rs6000.opt (-mpower9-dform-scalar): Delete
20739         undocumented switches.
20740         (-mpower9-dform-vector): Likewise.
20741         (-mpower9-dform): Likewise.
20742         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
20743         comments to delete references to -mpower9-dform* switches.
20744         * config/rs6000/predicates.md (vsx_quad_dform_memory_operand):
20745         Delete reference to -mpower9-dform* switches, test for
20746         -mpower9-vector instead.
20747         * config/rs6000/rs6000-cpus.def (ISA_3_0_MASKS_SERVER): Likewise.
20748         (OTHER_P9_VECTOR_MASKS): Likewise.
20749         (POWERPC_MASKS): Likewise.
20750         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Change
20751         tests against -mpower9-dform* to -mpower9-vector.  Delete code
20752         that checked for -mpower9-dform* consistancy with other options.
20753         Add test for -mpower9-misc to enable other power9 switches.
20754         (rs6000_init_hard_regno_mode_ok): Likewise.
20755         (rs6000_option_override_internal): Likewise.
20756         (rs6000_emit_prologue): Likewise.
20757         (rs6000_emit_epilogue): Likewise.
20758         (rs6000_opt_masks): Delete -mpower9-dform-{scalar,vector}.
20759         (rs6000_disable_incompatiable_switches): Delete -mpower9-dform.
20760         (emit_fusion_p9_load): Change tests for -mpower9-dform-scalar
20761         -mpower9-vector.
20762         (emit_fusion_p9_store): Likewise.
20763         * config/rs6000/rs6000.h (TARGET_P9_DFORM_SCALAR): Delete
20764         resetting these macros if the assembler does not support ISA 3.0
20765         instructions.
20766         (TARGET_P9_DFORM_VECTOR): Likewise.
20767         * config/rs6000/rs6000.md (peepholes to optimize altivec memory):
20768         Change to use -mpower9-vector instead of -mpower9-dform-scalar.
20770 2017-08-25  Alan Modra  <amodra@gmail.com>
20772         PR rtl-optimization/81747
20773         * cse.c (cse_extended_basic_block): Don't attempt to record
20774         equivalences for degenerate conditional branches that branch
20775         to their fall-through.
20777 2017-08-24  Martin Sebor  <msebor@redhat.com>
20779         PR middle-end/81908
20780         * gimple-fold.c (size_must_be_zero_p): New function.
20781         (gimple_fold_builtin_memory_op): Call it.
20783 2017-08-24  Steven Munroe  <munroesj@gcc.gnu.org>
20785         * config/rs6000/mm_malloc.h: New file.
20787 2017-08-24  Bin Cheng  <bin.cheng@arm.com>
20789         PR tree-optimization/81913
20790         * tree-ssa-loop-niter.c (number_of_iterations_cond): Skip niter
20791         analysis when either IVs in condition can wrap.
20793 2017-08-24  Uros Bizjak  <ubizjak@gmail.com>
20795         * dwarf2out.c (MAX_ARTIFICIAL_LABEL_BYTES): Increase to 40.
20796         * dwarf2cfi.c (MAX_ARTIFICIAL_LABEL_BYTES): Remove.
20798 2017-08-24  Richard Biener  <rguenther@suse.de>
20800         PR target/81921
20801         * targhooks.c (default_target_can_inline_p): Properly
20802         use target_option_default_node when no DECL_FUNCTION_SPECIFIC_TARGET
20803         is present and always compare.
20804         * config/i386/i386.c (ix86_valid_target_attribute_tree): Do not
20805         infer -mfpmath=sse from TARGET_SSE_P.
20806         (ix86_can_inline_p): Properly use target_option_default_node when
20807         no DECL_FUNCTION_SPECIFIC_TARGET is present and always compare.
20809 2017-08-24  Richard Biener  <rguenther@suse.de>
20811         PR debug/81936
20812         * dwarf2out.c (output_die): Handle flag_generate_offload like
20813         flag_generate_lto.
20814         (output_comp_unit): Likewise.
20815         (gen_array_type_die): Likewise.
20816         (dwarf2out_early_finish): Likewise.
20817         (note_variable_value_in_expr): Likewise.
20818         (dwarf2out_finish): Likewise.  Adjust assert.
20819         * cgraphunit.c (symbol_table::compile): Move setting of
20820         flag_generate_offload earlier ...
20821         (symbol_table::finalize_compilation_unit): ... here, before
20822         early debug finalization.
20824 2017-08-24  Richard Biener  <rguenther@suse.de>
20826         * config/i386/i386.c: Include symbol-summary.h, ipa-prop.h
20827         and ipa-fnsummary.h.
20828         (ix86_can_inline_p): When ix86_fpmath flags do not match
20829         check whether the callee uses FP math at all.
20831 2017-08-24  Aldy Hernandez  <aldyh@redhat.com>
20833         PR middle-end/81931
20834         * tree-ssanames.c (get_nonzero_bits): Use element_precision
20835         instead of TYPE_PRECISION.
20837 2017-08-24  Richard Sandiford  <richard.sandiford@linaro.org>
20838             Alan Hayward  <alan.hayward@arm.com>
20839             David Sherwood  <david.sherwood@arm.com>
20841         * combine.c (make_extraction): Use subreg_offset_from_lsb.
20843 2017-08-23  Daniel Santos  <daniel.santos@pobox.com>
20845         * config/i386/i386.h (ix86_frame::stack_realign_allocate_offset):
20846         Remove field.
20847         (ix86_frame::stack_realign_allocate): New field.
20848         (struct machine_frame_state): Modify comments.
20849         (machine_frame_state::sp_realigned_fp_end): New field.
20850         * config/i386/i386.c (ix86_compute_frame_layout): Rework stack frame
20851         layout calculation.
20852         (sp_valid_at): Add assertion to assure no attempt to access invalid
20853         offset of a realigned stack.
20854         (fp_valid_at): Likewise.
20855         (choose_baseaddr): Modify comments.
20856         (ix86_emit_outlined_ms2sysv_save): Adjust to changes in
20857         ix86_expand_prologue.
20858         (ix86_expand_prologue): Modify stack realignment and allocation.
20859         (ix86_expand_epilogue): Modify comments.
20860         * doc/sourcebuild.texi: Add documentation for target selectors avx2,
20861         avx2_runtime, avx512f, and avx512f_runtime.
20863 2017-08-23  Uros Bizjak  <ubizjak@gmail.com>
20865         * config/i386/i386.opt: Remove unneeded Init(0) initializations.
20866         (mstackrealign): Do not init to -1.
20867         * config/i386/i386.c (ix86_option_override_internal):
20868         Check opts_set, not opts when setting default value of
20869         opts->x_ix86_force_align_arg_pointer.
20871 2017-08-23  Richard Biener  <rguenther@suse.de>
20873         * function.c (fndecl_name): Use verbosity 1 (no arguments) for
20874         lang_hooks.decl_printable_name.
20875         * print-rtl-function.c (print_rtx_function): Likewise.
20876         * tree-pretty-print.c (dump_function_header): Likewise.
20878 2017-08-23  Richard Biener  <rguenther@suse.de>
20880         PR lto/81940
20881         * dwarf2out.c (dwarf2out_abstract_function): Handle LTO with
20882         -g0 at compile-time.
20884 2017-08-23  Tamar Christina  <tamar.christina@arm.com>
20886         PR middle-end/19706
20887         * doc/sourcebuild.texi (Other hardware attributes):
20888         Document xorsign.
20890 2017-08-23  Tamar Christina  <tamar.christina@arm.com>
20892         PR middle-end/19706
20893         * tree-ssa-math-opts.c (convert_expand_mult_copysign):
20894         Fix single-use check.
20896 2017-08-23  Thomas Preud'homme  <thomas.preudhomme@arm.com>
20898         * gcc.c (execute): Only test for SIGKILL and SIGQUIT if available.
20900 2017-08-22  Daniel Santos  <daniel.santos@pobox.com>
20902         * doc/install.texi: Modify to add more details on running selected
20903         tests.
20905 2017-08-22  Daniel Santos  <daniel.santos@pobox.com>
20907         * config/i386/i386.c (ix86_option_override_internal): Error when -mx32
20908         is combined with -mabi=ms.
20909         (ix86_function_type_abi): Limit errors for mixing -mx32 with attribute
20910         ms_abi.
20912 2017-08-22  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20914         PR tree-optimization/81488
20915         * gimple-ssa-strength-reduction (struct slsr_cand_d): Add visited
20916         and cached_basis fields.
20917         (MAX_SPREAD): New constant.
20918         (alloc_cand_and_find_basis): Initialize new fields.
20919         (clear_visited): New function.
20920         (create_phi_basis_1): Rename from create_phi_basis, set visited
20921         and cached_basis fields.
20922         (create_phi_basis): New wrapper function.
20923         (phi_add_costs_1): Rename from phi_add_costs, add spread
20924         parameter, set visited field, short-circuit when limits reached.
20925         (phi_add_costs): New wrapper function.
20926         (record_phi_increments_1): Rename from record_phi_increments, set
20927         visited field.
20928         (record_phi_increments): New wrapper function.
20929         (phi_incr_cost_1): Rename from phi_incr_cost, set visited field.
20930         (phi_incr_cost): New wrapper function.
20931         (all_phi_incrs_profitable_1): Rename from
20932         all_phi_incrs_profitable, set visited field.
20933         (all_phi_incrs_profitable): New wrapper function.
20935 2017-08-22  Richard Sandiford  <richard.sandiford@linaro.org>
20936             Alan Hayward  <alan.hayward@arm.com>
20937             David Sherwood  <david.sherwood@arm.com>
20939         * rtl.h (paradoxical_subreg_p): Define inline, and add a version
20940         that takes the outer and inner modes.
20941         * doc/rtl.texi: Use paradoxical_subreg_p instead of a GET_MODE_SIZE
20942         comparison as the canonical test for a paradoxical subreg.
20943         * combine.c (simplify_set): Use paradoxical_subreg_p.
20944         (make_extraction): Likewise.
20945         (force_to_mode): Likewise.
20946         (rtx_equal_for_field_assignment_p): Likewise.
20947         (gen_lowpart_for_combine): Likewise.
20948         (simplify_comparison): Likewise.
20949         * cse.c (equiv_constant): Likewise.
20950         * expmed.c (store_bit_field_1): Likewise.
20951         * final.c (alter_subreg): Likewise.
20952         * fwprop.c (propagate_rtx): Likewise.
20953         (forward_propagate_subreg): Likewise.
20954         * ira-conflicts.c (ira_build_conflicts): Likewise.
20955         * lower-subreg.c (simplify_gen_subreg_concatn): Likewise.
20956         * lra-constraints.c (curr_insn_transform): Likewise.
20957         (split_reg): Likewise.
20958         * lra-eliminations.c (move_plus_up): Likewise.
20959         (lra_eliminate_regs_1): Likewise.
20960         * recog.c (general_operand): Likewise.
20961         * ree.c (combine_reaching_defs): Likewise.
20962         * reload.c (push_reload): Likewise.
20963         (find_reloads): Likewise.
20964         * reload1.c (elimination_effects): Likewise.
20965         (compute_reload_subreg_offset): Likewise.
20966         (choose_reload_regs): Likewise.
20967         * rtlanal.c (subreg_lsb_1): Likewise.
20968         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
20969         (simplify_subreg): Likewise.
20970         * var-tracking.c (track_loc_p): Likewise.
20971         * emit-rtl.c (byte_lowpart_offset): Likewise.
20972         (paradoxical_subreg_p): Delete out-of-line definition.
20974 2017-08-22  Jeff Law  <law@redhat.com>
20976         PR tree-optimization/81741
20977         PR tree-optimization/71947
20978         * tree-ssa-dom.c: Include tree-inline.h.
20979         (record_temporary_equivalences): Only record SSA_NAME = SSA_NAME
20980         equivalences if one is more expensive to compute than the other.
20981         * tree-ssa-scopedtables.h (class const_or_copies): Make
20982         record_const_or_copy_raw method private.
20983         (class avail_exprs_stack): New method simplify_binary_operation.
20984         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr): Call
20985         avail_exprs_stack::simplify_binary_operation as needed.
20986         (avail_exprs_stack::simplify_binary_operation): New function.
20988 2017-08-22  Sebastian Huber  <sebastian.huber@embedded-brains.de>
20990         * config.gcc (powerpc-*-rtems*): Add rs6000/linux64.opt.
20991         * config/rs6000/rtems.h (ASM_PREFERRED_EH_DATA_FORMAT): New define.
20992         (DOT_SYMBOLS): Likewise.
20993         (MINIMAL_TOC_SECTION_ASM_OP): Likewise.
20994         (RELOCATABLE_NEEDS_FIXUP): Likewise.
20995         (RS6000_ABI_NAME): Likewise.
20996         (TARGET_CMODEL): Likewise.
20997         (TOC_SECTION_ASM_OP): Likewise.
20998         (SET_CMODEL): New macro.
20999         (SUBSUBTARGET_OVERRIDE_OPTIONS): Evaluate cmodel options.
21001 2017-08-22  Richard Biener  <rguenther@suse.de>
21003         * tree-inline.c (remap_type_1): Change asserts on TYPE_SIZE[_UNIT]
21004         to allow for free-lang-data replacements similar to verify_type_variant.
21006 2017-08-22  Richard Sandiford  <richard.sandiford@linaro.org>
21007             Alan Hayward  <alan.hayward@arm.com>
21008             David Sherwood  <david.sherwood@arm.com>
21010         * config/aarch64/aarch64.md (casesi): Use DImode rather than
21011         VOIDmode for the LABEL_REF.
21013 2017-08-22  Richard Biener  <rguenther@suse.de>
21015         * tree-cfg.c (gimple_split_edge): Avoid reallocating target PHI nodes.
21017 2017-08-22  Richard Biener  <rguenther@suse.de>
21019         * common.opt (feliminate-dwarf2-dups): Ignore.
21020         * doc/invoke.texi (feliminate-dwarf2-dups): Remove documentation.
21021         * dwarf2out.c (push_new_compile_unit, pop_compile_unit,
21022         same_die_p_wrap, compute_section_prefix,
21023         is_symbol_die, assign_symbol_names, break_out_includes): Remove.
21024         (comdat_symbol_id, comdat_symbol_number): Likewise.
21025         (cu_hash_table_entry, cu_hash_table_entry_hasher, cu_hash_type):
21026         Likewise.
21027         (check_duplicate_cu, record_comdat_symbol_number): Likewise.
21028         (output_die): Mark unreachable path unreachable.
21029         (dwarf2out_start_source_file): Do not create DW_TAG_GNU_BINCL.
21030         (dwarf2out_end_source_file): Do not create DW_TAG_GNU_EINCL.
21031         (dwarf2out_init): Remove code handling flag_eliminate_dwarf2_dups.
21032         (dwarf2out_early_finish): Likewise.
21034 2017-08-22  Aldy Hernandez  <aldyh@redhat.com>
21036         * wide-int.h (hwi_with_prec::hwi_with_prec): Sign extend.
21038 2017-08-22  Georg-Johann Lay  <avr@gjlay.de>
21040         PR target/81910
21041         * config/avr/avr.c (avr_handle_addr_attribute): Early return if
21042         not VAR_P. Filter attribute warnings with OPT_Wattributes.
21043         (avr_attribute_table) <io, io_low, address>: Initialize
21044         .decl_required with true.
21046 2017-08-21  Michael Meissner  <meissner@linux.vnet.ibm.com>
21048         * config/rs6000/rs6000-cpus.def (-mvsx-scalar-float): Delete
21049         undocumented debugging options.
21050         (-mvsx-scalar-double): Likewise.
21051         (-mallow-df-permute): Likewise.
21052         (-mvectorize-builtins): Likewise.
21053         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
21054         (rs6000_builtin_vectorized_function): Likewise.
21055         (rs6000_builtin_md_vectorized_function): Likewise.
21056         (rs6000_opt_vars): Likewise.
21058 2017-08-21  Uros Bizjak  <ubizjak@gmail.com>
21060         PR target/46091
21061         * config/i386/i386.md (*btsq_imm): Rename from *btsq.
21062         (*btrq_imm): Rename from *btrq.
21063         (*btcq_imm): Rename from *btcq.
21064         (btsc): New code attribute.
21065         (*<btsc><mode>): New insn pattern.
21066         (*btr<mode>): Ditto.
21067         (*<btsc><mode>_mask): New insn_and_split pattern.
21068         (*btr<mode>_mask): Ditto.
21070 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21071             Alan Hayward  <alan.hayward@arm.com>
21072             David Sherwood  <david.sherwood@arm.com>
21074         * function.c (pad_below): Simplify padding calculation.
21076 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21077             Alan Hayward  <alan.hayward@arm.com>
21078             David Sherwood  <david.sherwood@arm.com>
21080         * target.def (function_prologue): Remove frame size argument.
21081         (function_epilogue): Likewise.
21082         * doc/tm.texi: Regenerate.
21083         * final.c (final_start_function): Update call to function_prologue.
21084         (final_end_function): Update call to function_epilogue.
21085         (default_function_pro_epilogue): Remove frame size argument.
21086         * output.h (default_function_pro_epilogue): Likewise.
21087         * config/arm/arm.c (arm_output_function_epilogue): Likewise.
21088         (arm_output_function_prologue): Likewise.
21089         * config/frv/frv.c (frv_function_prologue): Likewise.
21090         (frv_function_epilogue): Likewise.
21091         * config/i386/i386.c (ix86_output_function_epilogue): Likewise.
21092         * config/ia64/ia64.c (ia64_output_function_prologue): Likewise.
21093         (ia64_output_function_epilogue): Likewise.
21094         * config/m32r/m32r.c (m32r_output_function_prologue): Likewise.
21095         (m32r_output_function_epilogue): Likewise.
21096         * config/microblaze/microblaze.c (microblaze_function_prologue)
21097         (microblaze_function_epilogue): Likewise.
21098         * config/mips/mips.c (mips_output_function_prologue): Likewise.
21099         (mips_output_function_epilogue): Likewise.
21100         * config/mmix/mmix.c (mmix_target_asm_function_prologue): Likewise.
21101         (mmix_target_asm_function_epilogue): Likewise.
21102         * config/msp430/msp430.c (msp430_start_function): Likewise.
21103         * config/nds32/nds32.c (nds32_asm_function_prologue): Likewise.
21104         (nds32_asm_function_epilogue): Likewise.
21105         * config/nios2/nios2.c (nios2_asm_function_prologue): Likewise.
21106         * config/pa/pa.c (pa_output_function_prologue): Likewise.
21107         (pa_output_function_epilogue): Likewise.
21108         * config/powerpcspe/powerpcspe.c (rs6000_output_function_prologue)
21109         (rs6000_output_function_epilogue): Likewise.
21110         * config/rl78/rl78.c (rl78_start_function): Likewise.
21111         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Likewise.
21112         (rs6000_output_function_epilogue): Likewise.
21113         * config/rx/rx.c (rx_output_function_prologue): Likewise.
21114         * config/sh/sh.c (sh_output_function_epilogue): Likewise.
21115         * config/sparc/sparc.c (sparc_asm_function_prologue): Likewise.
21116         (sparc_asm_function_epilogue): Likewise.
21118 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21120         * tree.h (type_has_mode_precision_p): New function.
21121         * convert.c (convert_to_integer_1): Use it.
21122         * expr.c (expand_expr_real_2): Likewise.
21123         (expand_expr_real_1): Likewise.
21124         * fold-const.c (fold_single_bit_test_into_sign_test): Likewise.
21125         * match.pd: Likewise.
21126         * tree-ssa-forwprop.c (simplify_rotate): Likewise.
21127         * tree-ssa-math-opts.c (convert_mult_to_fma): Likewise.
21128         * tree-tailcall.c (process_assignment): Likewise.
21129         * tree-vect-loop.c (vectorizable_reduction): Likewise.
21130         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern)
21131         (vect_recog_mult_pattern, vect_recog_divmod_pattern): Likewise.
21132         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
21133         (vectorizable_assignment): Likewise.
21134         (vectorizable_shift): Likewise.
21135         (vectorizable_operation): Likewise.
21136         * tree-vrp.c (register_edge_assert_for_2): Likewise.
21138 2017-08-21  Wilco Dijkstra  <wdijkstr@arm.com>
21140         * match.pd: Add pow (C, x) simplification.
21142 2017-08-21  Richard Biener  <rguenther@suse.de>
21144         PR tree-optimization/81900
21145         * tree-ssa-pre.c (compute_antic_aux): Properly compute changed
21146         for blocks with abnormal predecessors.
21147         (compute_antic): Do not set visited flag prematurely.
21149 2017-08-21  Georg-Johann Lay  <avr@gjlay.de>
21151         PR target/79883
21152         * config/avr/avr.c (avr_set_current_function): Typo in diagnostic.
21154 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21156         * stor-layout.h (vector_type_mode): Move to...
21157         * tree.h (vector_type_mode): ...here.
21158         * stor-layout.c (vector_type_mode): Move to...
21159         * tree.c (vector_type_mode): ...here.  Include rtl.h and regs.h.
21161 2017-08-21  Richard Biener  <rguenther@suse.de>
21163         * debug.h (struct gcc_debug_hooks): Add die_ref_for_decl and
21164         register_external_die hooks.
21165         (debug_false_tree_charstarstar_uhwistar): Declare.
21166         (debug_nothing_tree_charstar_uhwi): Likewise.
21167         * debug.c (do_nothing_debug_hooks): Adjust.
21168         (debug_false_tree_charstarstar_uhwistar): New do nothing.
21169         (debug_nothing_tree_charstar_uhwi): Likewise.
21170         * dbxout.c (dbx_debug_hooks): Adjust.
21171         (xcoff_debug_hooks): Likewise.
21172         * sdbout.c (sdb_debug_hooks): Likewise.
21173         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
21174         * dwarf2out.c (macinfo_label_base): New global.
21175         (dwarf2out_register_external_die): New function for the
21176         register_external_die hook.
21177         (dwarf2out_die_ref_for_decl): Likewise for die_ref_for_decl.
21178         (dwarf2_debug_hooks): Use them.
21179         (dwarf2_lineno_debug_hooks): Adjust.
21180         (struct die_struct): Add with_offset flag.
21181         (DEBUG_LTO_DWO_INFO_SECTION, DEBUG_LTO_INFO_SECTION,
21182         DEBUG_LTO_DWO_ABBREV_SECTION, DEBUG_LTO_ABBREV_SECTION,
21183         DEBUG_LTO_DWO_MACINFO_SECTION, DEBUG_LTO_MACINFO_SECTION,
21184         DEBUG_LTO_DWO_MACRO_SECTION, DEBUG_LTO_MACRO_SECTION,
21185         DEBUG_LTO_LINE_SECTION, DEBUG_LTO_DWO_STR_OFFSETS_SECTION,
21186         DEBUG_LTO_STR_DWO_SECTION, DEBUG_STR_LTO_SECTION): New macros
21187         defining section names for the early LTO debug variants.
21188         (reset_indirect_string): New helper.
21189         (add_AT_external_die_ref): Helper for dwarf2out_register_external_die.
21190         (print_dw_val): Add support for offsetted symbol references.
21191         (get_ultimate_context): Split out from is_cxx.
21192         (is_cxx): Use get_ultimate_context.
21193         (is_fortran): Add decl overload.
21194         (compute_comp_unit_symbol): Split out worker from
21195         compute_section_prefix.
21196         (compute_section_prefix): Call compute_comp_unit_symbol and
21197         set comdat_type_p here.
21198         (output_die): Skip DIE symbol output for the LTO added one.
21199         Handle DIE symbol references with offset.
21200         (output_comp_unit): Guard section name mangling properly.
21201         For LTO debug sections emit a symbol at the section beginning
21202         which we use to refer to its DIEs.
21203         (add_abstract_origin_attribute): For DIEs registered via
21204         dwarf2out_register_external_die directly refer to the early
21205         DIE rather than indirectly through the shadow one we created.
21206         Remove obsolete call to dwarf2out_abstract_function for
21207         non-function/block origins.
21208         (gen_array_type_die): When generating early LTO debug do
21209         not emit DW_AT_string_length.
21210         (gen_formal_parameter_die): Do not re-create DIEs for PARM_DECLs
21211         late when in LTO.  As suggested place a gcc_unreachable for
21212         the DECL_ABSTRACT_P case.
21213         (gen_subprogram_die): Avoid another specification DIE
21214         for early built declarations/definitions for the late LTO case.
21215         (gen_variable_die): Add type references for late duplicated VLA dies
21216         when in late LTO.
21217         (gen_inlined_subroutine_die): Do not call dwarf2out_abstract_function,
21218         we have the abstract instance already.
21219         (process_scope_var): Adjust decl DIE contexts in LTO which
21220         first puts them in limbo.
21221         (gen_decl_die): Do not generate type DIEs late apart from
21222         types for VLAs or for decls we do not yet have a DIE.  Do not
21223         call dwarf2out_abstract_function late.
21224         (dwarf2out_early_global_decl): Make sure to create DIEs
21225         for abstract instances of a decl first.
21226         (dwarf2out_late_global_decl): Adjust comment.
21227         (output_macinfo_op): With multiple macro sections use
21228         macinfo_label_base to distinguish labels.
21229         (output_macinfo): Likewise.  Update macinfo_label_base.
21230         Pass in the line info label.
21231         (note_variable_value_in_expr): When generating LTO resolve
21232         all variable values here by generating DIEs as needed.
21233         (init_sections_and_labels): Add early LTO debug flag parameter
21234         and generate different sections and names if set.  Add generation
21235         counter for the labels so we can have multiple of them.
21236         (reset_dies): Helper to allow DIEs to be output multiple times.
21237         (dwarf2out_finish): When outputting DIEs to the fat part of an
21238         LTO object first reset DIEs.
21239         (dwarf2out_early_finish): Output early DIEs when generating LTO.
21240         (modified_type_die): Check for decl_ultimate_origin being self
21241         before recursing.
21242         (gen_type_die_with_usage): Likewise.
21243         (gen_typedef_die): Allow decl_ultimate_origin being self.
21244         (set_decl_abstract_flags): Remove.
21245         (set_block_abstract_flags): Likewise.
21246         (dwarf2out_abstract_function): Treat the early generated DIEs
21247         as the abstract copy and only add DW_AT_inline and
21248         DW_AT_artificial here and call set_decl_origin_self.
21249         If the DIE has an abstract origin don't do anything.
21250         * tree.c (free_lang_data): Build a dummy TRANSLATION_UNIT_DECL
21251         if we have none yet (Go fails to build one, PR78628).
21252         (variably_modified_type_p): Prevent endless recursion for Ada
21253         cyclic pointer types.
21254         * lto-streamer-in.c: Include debug.h.
21255         (dref_queue): New global.
21256         (lto_read_tree_1): Stream in DIE references.
21257         (lto_input_tree): Register DIE references.
21258         (input_function): Stream DECL_DEBUG_ARGS.
21259         * lto-streamer-out.c: Include debug.h.
21260         (lto_write_tree_1): Output DIE references.
21261         (DFS::DFS_write_tree_body): Follow DECL_ABSTRACT_ORIGIN.
21262         Force a TRANSLATION_UNIT_DECL DECL_CONTEXT for file-scope decls.
21263         (output_function): Stream DECL_DEBUG_ARGS.
21264         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
21265         Stream DECL_ABSTRACT_ORIGIN.
21266         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
21267         (write_ts_decl_minimal_tree_pointers): Force a TRANSLATION_UNIT_DECL
21268         DECL_CONTEXT for file-scope decls.
21269         * lto-streamer.h (struct dref_entry): Declare.
21270         (dref_queue): Likewise.
21271         * cfgexpand.c (pass_expand::execute): Do not call the
21272         outlining_inline_function hook here.
21273         * lto-wrapper.c (debug_obj): New global.
21274         (tool_cleanup): Unlink it if required.
21275         (debug_objcopy): New function.
21276         (run_gcc): Handle early debug sections in the IL files by
21277         extracting them to separate files, partially linkin them and
21278         feeding the result back as result to the linker.
21279         * config/darwin.h (DEBUG_LTO_INFO_SECTION, DEBUG_LTO_ABBREV_SECTION,
21280         DEBUG_LTO_MACINFO_SECTION, DEBUG_LTO_LINE_SECTION,
21281         DEBUG_STR_LTO_SECTION, DEBUG_LTO_MACRO_SECTION): Put early debug
21282         sections into a separate segment.
21283         * config/darwin.c (darwin_asm_named_section): Handle __GNU_DWARF_LTO
21284         segments.
21285         (darwin_asm_dwarf_section): Likewise.
21286         (darwin_asm_output_dwarf_offset): Likewise.
21287         * config/i386/i386.c (make_resolver_func): Set DECL_IGNORED_P.
21289 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21290             Alan Hayward  <alan.hayward@arm.com>
21291             David Sherwood  <david.sherwood@arm.com>
21293         * read-md.h (md_reader::record_potential_iterator_use): Replace
21294         pointer argument with an rtx and an index.
21295         * read-rtl.c (iterator_group::apply_iterator): Likewise.
21296         (apply_mode_iterator): Likewise.
21297         (apply_code_iterator): Likewise.
21298         (apply_int_iterator): Likewise.
21299         (apply_subst_iterator): Likewise.
21300         (record_iterator_use): Likewise.
21301         (record_attribute_use): Likewise.
21302         (md_reader::record_potential_iterator_use): Likewise.  Update calls
21303         to record_iterator_use and apply_iterator.
21304         (iterator_use): Replace ptr with x and index.
21305         (attribute_use): Likewise.
21306         (apply_attribute_uses): Update calls to apply_iterator.
21307         (apply_iterators): Likewise.  Update initialization of iterator_use.
21308         (rtx_reader::read_rtx_code): Update calls to record_iterator_use
21309         and record_potential_iterator_use.
21310         (rtx_reader::read_rtx_operand): Likewise.
21312 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21313             Alan Hayward  <alan.hayward@arm.com>
21314             David Sherwood  <david.sherwood@arm.com>
21316         * varasm.c (const_rtx_hash_1): Don't hash in the mode of a
21317         CONST_WIDE_INT.
21319 2017-08-21  Richard Biener  <rguenther@suse.de>
21321         PR middle-end/81884
21322         * tree-ssa-alias.c (stmt_kills_ref_p): Handle array accesses
21323         at struct end conservatively when comparing common bases.
21325 2017-08-21  Richard Biener  <rguenther@suse.de>
21327         * tree-ssa-loop-im.c (struct lim_aux_data): Add ref index member.
21328         (mem_ref_in_stmt): Remove.
21329         (determine_max_movement): Use ref index to get at the reference.
21330         (invariantness_dom_walker::before_dom_children): Deal with
21331         lim data already initialized.
21332         (gather_mem_refs_stmt): Initialize lim data and record ref index.
21334 2017-08-19  Uros Bizjak  <ubizjak@gmail.com>
21336         * config/i386/i386.h (OPTION_MASK_ISA_ROUND): Remove.
21337         (TARGET_ISA_ROUND): Ditto.
21338         (TARGET_ROUND): Ditto.
21339         * config/i386/i386.c: Substitute TARGET_ROUND with TARGET_SSE4_1.
21340         * config/i386/i386.md: Ditto.
21341         * config/i386/sse.md: Ditto.
21342         * config/i386/i386-builtin.def: Substitute OPTION_MASK_ISA_ROUND
21343         with OPTION_MASK_ISA_SSE4_1.
21345 2017-08-19  Uros Bizjak  <ubizjak@gmail.com>
21347         PR target/81894
21348         * doc/extend.texi (x86 Built-in Functions): Correct the name of
21349         __builtin_ia32_lzcnt_u16.
21351 2017-08-18  Peter Bergner  <bergner@vnet.ibm.com>
21353         PR target/80210
21354         * config/rs6000/rs6000.c (rs6000_activate_target_options): New function.
21355         (rs6000_set_current_function): Rewrite function to use it.
21357 2017-08-18  H.J. Lu  <hongjiu.lu@intel.com>
21359         PR c/53037
21360         * print-tree.c (print_node): Support DECL_WARN_IF_NOT_ALIGN
21361         and TYPE_WARN_IF_NOT_ALIGN.
21362         * stor-layout.c (do_type_align): Merge DECL_WARN_IF_NOT_ALIGN.
21363         (handle_warn_if_not_align): New.
21364         (place_union_field): Call handle_warn_if_not_align.
21365         (place_field): Call handle_warn_if_not_align.
21366         Copy TYPE_WARN_IF_NOT_ALIGN.
21367         (finish_builtin_struct): Copy TYPE_WARN_IF_NOT_ALIGN.
21368         (layout_type): Likewise.
21369         * tree-core.h (tree_type_common): Add warn_if_not_align.  Set
21370         spare to 18.
21371         (tree_decl_common): Add warn_if_not_align.
21372         * tree.c (build_range_type_1): Copy TYPE_WARN_IF_NOT_ALIGN.
21373         * tree.h (TYPE_WARN_IF_NOT_ALIGN): New.
21374         (SET_TYPE_WARN_IF_NOT_ALIGN): Likewise.
21375         (DECL_WARN_IF_NOT_ALIGN): Likewise.
21376         (SET_DECL_WARN_IF_NOT_ALIGN): Likewise.
21377         * doc/extend.texi: Document warn_if_not_aligned attribute.
21378         * doc/invoke.texi: Document -Wif-not-aligned and -Wpacked-not-aligned.
21380 2017-08-17  Martin Liska  <mliska@suse.cz>
21382         PR bootstrap/81864
21383         * tree-loop-distribution.c (ddrs_table): Change type to pointer type.
21384         (get_data_dependence): Use it as pointer type.
21385         (distribute_loop): Likewise.
21387 2017-08-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21389         * config/rs6000/altivec.md (UNSPEC_VMRGOW_DIRECT): New constant.
21390         (p8_vmrgew_v4sf_direct): Generalize to p8_vmrgew_<mode>_direct.
21391         (p8_vmrgow_<mode>_direct): New define_insn.
21392         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Properly
21393         handle endianness for vmrgew and vmrgow permute patterns.
21395 2017-08-17  Peter Bergner  <bergner@vnet.ibm.com>
21397         * config/rs6000/altivec.md (VParity): Remove TARGET_VSX_TIMODE.
21398         * config/rs6000/rs6000-cpus.def: Remove comment.
21399         (ISA_2_7_MASKS_SERVER): Delete OPTION_MASK_VSX_TIMODE;
21400         (POWERPC_MASKS): Likewise.
21401         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Remove unneeded
21402         use of TARGET_VSX_TIMODE.
21403         (rs6000_setup_reg_addr_masks): Change TARGET_VSX_TIMODE to TARGET_VSX.
21404         (rs6000_init_hard_regno_mode_ok): Remove unneeded uses of
21405         TARGET_VSX_TIMODE.  Change use of TARGET_VSX_TIMODE to TARGET_VSX.
21406         (rs6000_option_override_internal): Remove dead code.
21407         (rs6000_legitimize_address): Change TARGET_VSX_TIMODE to TARGET_VSX.
21408         (rs6000_legitimize_reload_address): Likewise.
21409         (rs6000_legitimate_address_p): Likewise.
21410         (rs6000_opt_masks): Delete "vsx-timode".
21411         (rs6000_disable_incompatible_switches): Remove mention of -mvsx-timode
21412         from function comment.
21413         * config/rs6000/rs6000.h (MASK_VSX_TIMODE): Delete.
21414         * config/rs6000/rs6000.md (FMOVE128_GPR): Remove TARGET_VSX_TIMODE.
21415         (V16QI, V8HI, V4SI, V4SF, V2DI, V2DF, V1TI): Remove useless empty
21416         condition.
21417         * config/rs6000/rs6000.opt (mvsx-timode): Replace with stub.
21418         * config/rs6000/vector.md (VEC_IP): Remove TARGET_VSX_TIMODE.
21419         * config/rs6000/vsx.md (VSX_LE_128): Likewise.
21420         (VSX_TI): Likewise.
21421         (VSX_M): Likewise.
21422         (define_peephole2): Likewise.
21424 2017-08-17  Martin Sebor  <msebor@redhat.com>
21426         PR c/81859
21427         * pretty-print.c (pp_format): Use strnlen in %.*s to avoid reading
21428         past the end of an array.
21429         (test_pp_format): Add test cases.
21431 2017-08-17  Richard Sandiford  <richard.sandiford@linaro.org>
21433         * internal-fn.def (CLRSB, CLZ, CTZ, FFS, PARITY, POPCOUNT): Add
21434         missing ECF_NOTHROW flags.
21436 2017-08-17  Peter Bergner  <bergner@vnet.ibm.com>
21438         PR target/72804
21439         * config/rs6000/vsx.md (*vsx_le_permute_<mode>): Add support for
21440         operands residing in integer registers.
21441         (*vsx_le_perm_load_<mode>): Likewise.
21442         (*vsx_le_perm_store_<mode>): Likewise.
21443         (define_peephole2): Add peepholes to optimize the above.
21445 2017-08-17  Marek Polacek  <polacek@redhat.com>
21447         PR middle-end/81814
21448         * fold-const.c (operand_equal_for_comparison_p): Remove code that used
21449         to mimic what shorten_compare did.  Change the return type to bool.
21450         (fold_cond_expr_with_comparison): Update call to
21451         operand_equal_for_comparison_p.
21452         (fold_ternary_loc): Likewise.
21454 2017-08-17  Jackson Woodruff  <jackson.woodruff@arm.com>
21456         * aarch64-simd.md (mov<mode>): No longer force zero immediate into
21457         register.
21458         (*aarch64_simd_mov<mode>): Add new case for stp using zero immediate.
21460 2017-08-17  Richard Biener  <rguenther@suse.de>
21462         * tree-ssa-structalias.c (solve_graph): When propagating
21463         to successors update the graphs succ edges and avoid duplicate work.
21465 2017-08-17  Maxim Ostapenko  <m.ostapenko@samsung.com>
21467         PR target/81861
21468         * config/i386/i386.c (ix86_option_override_internal): Save target
21469         specific options after ix86_stack_protector_guard_reg was changed.
21471 2017-08-17  Richard Biener  <rguenther@suse.de>
21473         PR tree-optimization/81827
21474         * tree-ssa-structalias.c (struct variable_info): Add is_reg_var flag.
21475         (new_var_info): Initialize it conservatively.
21476         (get_call_vi): Mark register vars.
21477         (new_scalar_tmp_constraint_exp): Likewise.
21478         (handle_rhs_call): Likewise.
21479         (handle_const_call): Likewise.
21480         (create_function_info_for): Likewise.
21481         (solve_constraints): Sort varinfos to separate register from
21482         non-register vars to pack points-to solution bitmaps during
21483         iteration.
21485 2017-08-17  Marek Polacek  <polacek@redhat.com>
21487         * gimplify.c (gimplify_adjust_omp_clauses): Compare with 0 instead of 1.
21489 2017-08-17  Richard Biener  <rguenther@suse.de>
21491         * tree-vrp.c (vrp_int_const_binop): Do not set *overflow_p
21492         to true when overflow is undefined and we saturated the result.
21494 2017-08-17  Alan Modra  <amodra@gmail.com>
21496         PR target/80938
21497         * config/rs6000/rs6000.c (rs6000_savres_strategy): Revert 2017-08-09.
21498         Don't use store multiple if only one reg needs saving.
21499         (interesting_frame_related_regno): New function.
21500         (rs6000_frame_related): Don't emit frame info for regs that
21501         don't need saving.
21502         (rs6000_emit_epilogue): Likewise.
21504 2017-08-16  Nathan Sidwell  <nathan@acm.org>
21506         * tree-core.h (tree_type_non_common): Rename binfo to lang_1.
21507         * tree.h (TYPE_BINFO): Use type_non_common.maxval.
21508         (TYPE_LANG_SLOT_1): Use type_non_common.lang_1, for any type.
21509         * tree.c (free_lang_data_in_type): Use else-if chain.  Always
21510         clear TYPE_LANG_1.  Remove obsolete member-function stripping.
21511         (find_decls_types_r): Comment about TYPE_MAX_VALUES_RAW.
21512         (verify_type): Adjust for TYPE_BINFO move.
21513         * lto-streamer-out.c (DFS::DFS_write_tree_body): No need to
21514         process TYPE_BINFO directly.
21515         (hash_tree): Likewise.
21516         * tree-streamer-in.c (lto_input_ts_type_non_common_tree_pointers):
21517         Likewise.
21518         * tree-streamer-out.c (write_ts_type_non_common_tree_pointers):
21519         Likewise.
21521 2017-08-16  David Malcolm  <dmalcolm@redhat.com>
21523         * diagnostic-show-locus.c (colorizer::m_caret): Remove unused field.
21525 2017-08-16  Uros Bizjak  <ubizjak@gmail.com>
21527         PR target/46091
21528         * config/i386/i386.md (*anddi_1_btr): Change predicates of
21529         operand 0 and operand 1 to nomimmediate_operand. Add "m" constraint.
21530         Add ix86_binary_operator_ok to insn constraint.
21531         (*iordi_1_bts): Ditto.
21532         (*xordi_1_btc): Ditto.
21533         (*btsq): Change predicate of operand 0 to nonimmediate_operand.
21534         Update corresponding peephole2 pattern.
21535         (*btrq): Ditto.
21536         (*btcq): Ditto.
21538 2017-08-16  Bin Cheng  <bin.cheng@arm.com>
21540         PR tree-optimization/81832
21541         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Don't
21542         copy loop header which has IFN_LOOP_DIST_ALIAS call.
21544 2017-08-16  Marek Polacek  <polacek@redhat.com>
21546         PR middle/81695
21547         * fold-const.c (fold_indirect_ref_1): Restore original behavior
21548         regarding size_zero_node.
21550 2017-08-16  Martin Liska  <mliska@suse.cz>
21552         PR target/81753
21553         * config.gcc: Respect previously set extra_objs in case
21554         of darwin target.
21556 2017-08-16  Richard Sandiford  <richard.sandiford@linaro.org>
21558         PR tree-optimization/81835
21559         * tree-vect-loop.c (vect_is_simple_reduction): Simply checks for
21560         the phi SSA_NAME.  Check that the condition in a COND_EXPR does
21561         not depend on the phi.
21563 2017-08-16  Alan Modra  <amodra@gmail.com>
21565         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Delete
21566         dead code.
21568 2017-08-16  Alan Modra  <amodra@gmail.com>
21570         * config/rs6000/rs6000.c (rs6000_reg_live_or_pic_offset_p): Merge..
21571         (save_reg_p): ..into this.  Update all callers.
21572         (first_reg_to_save): Simplify.
21574 2017-08-16  Alan Modra  <amodra@gmail.com>
21576         * config/rs6000/rs6000.c (rs6000_savres_strategy): Don't restore
21577         fixed regs.
21579 2017-08-15  Joseph Myers  <joseph@codesourcery.com>
21581         PR target/78460
21582         PR target/67712
21583         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Only unroll for
21584         constant count if that count is less than 32.
21586 2017-08-15  Nathan Sidwell  <nathan@acm.org>
21588         * gcc.c (execute): Emit friendlier message if inferior is killed
21589         by an external cause.
21591 2017-08-15  Richard Biener  <rguenther@suse.de>
21593         PR tree-optimization/81790
21594         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Handle both
21595         CONSTRUCTORs from simplifying and VN.
21597 2017-08-14  Martin Sebor  <msebor@redhat.com>
21599         * builtin-attrs.def: Add comments.
21601 2017-08-14  Martin Sebor  <msebor@redhat.com>
21603         PR c/81117
21604         * doc/extend.texi (attribute nonstring): Document new attribute.
21606 2017-08-14  Martin Sebor  <msebor@redhat.com>
21608         PR c/81117
21609         * tree-diagnostic.c (default_tree_printer): Handle %G.
21610         * gimple-pretty-print.h (percent_G_format): Declare new function.
21611         * gimple-pretty-print.c (percent_G_format): Define.
21612         * tree-pretty-print.c (percent_K_format): Add argument.
21614 2017-08-14  Martin Sebor  <msebor@redhat.com>
21616         PR translation/79998
21617         * gimple-ssa-sprintf.c (pass_sprintf_length::handle_gimple_call):
21618         Remove a stray space.
21620 2017-08-14  Uros Bizjak  <ubizjak@gmail.com>
21622         PR target/46091
21623         * config/i386/i386.md (*anddi_1_btr): New insn_and_split pattern.
21624         (*iordi_1_bts): Ditto.
21625         (*xordi_1_btc): Ditto.
21627 2017-08-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21629         PR target/79845
21630         * config/rs6000/linux64.h (INVALID_64BIT): Use quoted strings.
21631         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
21632         Likewise.
21633         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Use
21634         quoted strings, and make more translator-friendly.
21635         (darwin_rs6000_override_options): Likewise.
21636         (rs6000_option_override_internal): Likewise.
21637         (rs6000_return_in_memory): Fix overlong line.
21638         (init_cmulative_args): Use quoted strings, and make more
21639         translator-friendly.
21640         (rs6000_pass_by_reference): Fix overlong line.
21641         (def_builtin): Use quoted strings.
21642         (altivec_expand_predicate_builtin): Use quoted strings, and make
21643         more translator-friendly.
21644         (htm_expand_builtin): Use quoted strings.
21645         (cpu_expand_builtin): Use quoted strings, and make more
21646         translator-friendly.
21647         (altivec_expand_builtin): Likewise.
21648         (paired_expand_predicate_builtin): Likewise.
21649         (rs6000_invalid_builtin): Likewise.
21650         (builtin_function_type): Use quoted strings.
21651         (rs6000_expand_split_stack_prologue): Use quoted strings, and make
21652         more translator-friendly.
21653         (rs6000_trampoline_init): Likewise.
21654         (rs6000_handle_altivec_attribute): Likewise.
21655         (rs6000_inner_target_options): Use quoted strings.
21656         (rs6000_disable_incompatible_switches): Likewise.
21657         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Use quoted
21658         strings, and make more translator-friendly.
21659         (SUBSUBTARGET_OVERRIDE_OPTIONS): Use quoted strings.
21661 2017-08-14  Bin Cheng  <bin.cheng@arm.com>
21663         PR tree-optimization/81799
21664         * tree-loop-distribution.c (version_loop_by_alias_check): Force
21665         cond_expr to simple gimple operand.
21667 2017-08-14  Wilco Dijkstra  <wdijkstr@arm.com>
21669         PR middle-end/46932
21670         * doc/sourcebuild.texi (autoincdec): Add autoincdec description.
21672 2017-08-14  Georg-Johann Lay  <avr@gjlay.de>
21674         PR target/81754
21675         PR target/81268
21676         * config/avr/avr.opt (mgas-isr-prologues): New Var avr_gasisr_prologues.
21677         * config/avr/avr.md (gasisr, *gasisr): Use it instead of
21678         TARGET_GASISR_PROLOGUES.
21679         * config/avr/avr.c (avr_option_override): Same.
21680         (avr_pass_pre_proep::execute): Same.
21682 2017-08-13  H.J. Lu  <hongjiu.lu@intel.com>
21684         PR target/81820
21685         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Replace
21686         frame pointer with stack pointer - UNITS_PER_WORD in debug insns.
21688 2017-08-13  Uros Bizjak  <ubizjak@gmail.com>
21690         * config/i386/i386.md (*load_tp_<mode>): Redefine as
21691         define_insn_and_split.  Split to a memory load from 0 in
21692         DEFAULT_TLS_SEG_REG address space.  Merge with *load_tp_x32
21693         using PTR mode iterator.
21694         (*load_tp_x32_zext"): Redefine as define_insn_and_split.
21695         Split to a memory load from 0 in DEFAULT_TLS_SEG_REG address space.
21696         (*add_tp_<mode>): Redefine as define_insn_and_split.
21697         Split to an add with a memory load from 0 in DEFAULT_TLS_SEG_REG
21698         address space.  Merge with *add_tp_x32 using PTR mode iterator.
21699         (*add_tp_x32_zext"): Redefine as define_insn_and_split.
21700         Split to an add with a  memory load from 0 in
21701         DEFAULT_TLS_SEG_REG address space.
21703 2017-08-12  Andrew Pinski  <apinski@cavium.com>
21705         * config/aarch64/aarch64-option-extensions.def (rdma):
21706         Fix feature string to what Linux prints out in /proc/cpuinfo.
21708 2017-08-12  Pierre-Marie de Rodat  <derodat@adacore.com>
21710         PR ada/79542
21711         * dwarf2out.c (modified_type_die): For C typedef types that have
21712         an ultimate origin, process the ultimate origin instead of the
21713         input type.
21714         (gen_typedef_die): Assert that input DECLs have no ultimate
21715         origin.
21716         (gen_type_die_with_usage): For typedef variants that have an
21717         ultimate origin, just call gen_decl_die on the original DECL.
21718         (process_scope_var): Avoid creating DIEs for local typedefs and
21719         concrete static variables.
21721 2017-08-12  Alan Modra  <amodra@gmail.com>
21723         PR target/81170
21724         PR target/81295
21725         * config/rs6000/sysv4.h (STARTFILE_LINUX_SPEC): Upgrade to
21726         match gnu-user.h startfile.
21727         (ENDFILE_LINUX_SPEC): Similarly.
21729 2017-08-11  Thomas Schwinge  <thomas@codesourcery.com>
21731         PR lto/81430
21732         * config/nvptx/nvptx.c (nvptx_override_options_after_change):
21733         Remove function.
21734         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Remove definition.
21736 2017-08-11  Tamar Christina  <tamar.christina@arm.com>
21737         * config/aarch64/aarch64.md (mov<mode>): Change.
21738         (*movhf_aarch64, *movsf_aarch64, *movdf_aarch64):
21739         aarch64_reg_or_fp_float into aarch64_reg_or_fp_zero.
21740         * config/aarch64/predicates.md (aarch64_reg_or_fp_float): Removed.
21742 2017-08-11  Eric Botcazou  <ebotcazou@adacore.com>
21744         * tree-sra.c (build_access_from_expr_1): Use more precise diagnostics
21745         for storage order barriers.
21747 2017-08-11  Martin Liska  <mliska@suse.cz>
21749         PR tree-opt/79987
21750         * tree-chkp.c (chkp_get_bounds_for_decl_addr): Do not instrument
21751         variables of void type.
21753 2017-08-11  Martin Liska  <mliska@suse.cz>
21755         * asan.c (asan_protect_global): Replace ASM_OUTPUT_DEF with
21756         TARGET_SUPPORTS_ALIASES.
21757         * cgraph.c (cgraph_node::create_same_body_alias): Likewise.
21758         * ipa-visibility.c (can_replace_by_local_alias): Likewise.
21759         (optimize_weakref): Likewise.
21760         * symtab.c (symtab_node::noninterposable_alias): Likewise.
21761         * varpool.c (varpool_node::create_extra_name_alias): Likewise.
21762         * defaults.h: Introduce TARGET_SUPPORTS_ALIASES.
21764 2017-08-11  Martin Liska  <mliska@suse.cz>
21766         PR ipa/81213
21767         * config/i386/i386.c (make_resolver_func): Do complete
21768         refactoring of the function.
21770 2017-08-10  Uros Bizjak  <ubizjak@gmail.com>
21772         PR target/81708
21773         * config/i386/i386.opt (mstack-protector-guard-symbol=): New option
21774         * config/i386/i386.c (ix86_stack_protect_guard): Use
21775         ix86_stack_protect_guard_symbol_str to generate varible declaration.
21776         * doc/invoke.texi (x86 Options): Document
21777         -mstack-protector-guard-symbol= option.
21779 2017-08-10  Uros Bizjak  <ubizjak@gmail.com>
21781         * config/i386/i386-protos.h (ix86_split_stack_guard): New prototype.
21782         * config/i386/i386.c (ix86_split_stack_guard): New function.
21783         (ix86_xpand_split_stack_prologue): Call ix86_split_stack_guard.
21784         (ix86_legitimate_address_p) <case UNSPEC_STACK_CHECK>: Remove.
21785         (i386_asm_output_addr_const_extra) <case UNSPEC_STACK_CHECK>: Ditto.
21786         (optput_pic_addr_const): Remove UNSPEC_STACK_CHECK handling.
21787         * config/i386/i386.md (unspec): Remove UNSPEC_STACK_CHECK.
21788         (split_stack_space_check): Call ix86_split_stack_guard.
21790 2017-08-10  Martin Sebor  <msebor@redhat.com>
21792         * print-tree.c (print_node): Print location using the established
21793         format %s:%i%i.
21794         Replace spaces with colons.
21795         (debug_raw, debug): Ditto.
21797 2017-08-10  Martin Sebor  <msebor@redhat.com>
21799         PR c++/81586
21800         * pretty-print.c (pp_format): Correct the handling of %s precision.
21802 2017-08-10  H.J. Lu  <hongjiu.lu@intel.com>
21804         PR target/81736
21805         * config/i386/i386.c (ix86_finalize_stack_realign_flags): Renamed
21806         to ...
21807         (ix86_finalize_stack_frame_flags): This.  Also clear
21808         frame_pointer_needed if -fno-omit-frame-pointer is used without
21809         stack access.
21810         (ix86_expand_prologue): Replace ix86_finalize_stack_realign_flags
21811         with ix86_finalize_stack_frame_flags.
21812         (ix86_expand_epilogue): Likewise.
21813         (ix86_expand_split_stack_prologue): Likewise.
21814         * doc/invoke.texi: Add a note for -fno-omit-frame-pointer.
21816 2017-08-10  Martin Liska  <mliska@suse.cz>
21818         PR c++/81355
21819         * c-attribs.c (handle_target_attribute):
21820         Report warning for an empty string argument of target attribute.
21822 2017-08-09  Jakub Jelinek  <jakub@redhat.com>
21824         PR c/81687
21825         * omp-low.c (omp_copy_decl): Don't remap FORCED_LABEL or DECL_NONLOCAL
21826         LABEL_DECLs.
21827         * tree-cfg.c (move_stmt_op): Don't adjust DECL_CONTEXT of FORCED_LABEL
21828         or DECL_NONLOCAL labels.
21829         (move_stmt_r) <case GIMPLE_LABEL>: Adjust DECL_CONTEXT of FORCED_LABEL
21830         or DECL_NONLOCAL labels here.
21832 2017-08-09  Will Schmidt  <will_schmidt@vnet.ibm.com>
21834         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add blurb
21835         to indicate when early gimple folding has been disabled.
21836         (rs6000_gimple_fold_builtin): Add debug content.
21837         (rs6000_invalid_builtin): Fix whitespace.
21838         (rs6000_expand_builtin): Fix whitespace.
21839         * config/rs6000/rs6000.opt: Add option for -mfold-gimple.
21841 2017-08-09  Segher Boessenkool  <segher@kernel.crashing.org>
21843         PR target/80938
21844         * config/rs6000/rs6000.c (rs6000_savres_strategy): Don't use
21845         SAVE_MULTIPLE if not all the registers that saves, should be saved.
21847 2017-08-09  Jim Wilson  <jim.wilson@linaro.org>
21849         * config/aarch64/aarch64-cores.def (falkor): Use falkor pipeline.
21850         (qdf24xx): Likewise.
21851         * config/aarch64/aarch64.md: Include falkor.md.
21852         * config/aarch64/falkor.md: New.
21854 2017-08-09  Marek Polacek  <polacek@redhat.com>
21856         PR c/81233
21857         * diagnostic-core.h (emit_diagnostic_valist): Add declaration.
21858         * diagnostic.c (emit_diagnostic): Add a comment.
21859         (emit_diagnostic_valist): New function.
21861 2017-08-09  Marek Polacek  <polacek@redhat.com>
21863         PR c/81417
21864         * input.c (make_location): New overload.
21865         * input.h (make_location): Declare.
21867 2017-08-08  Alan Modra  <amodra@gmail.com>
21868             H.J. Lu  <hongjiu.lu@intel.com>
21870         PR driver/81523
21871         * gcc.c (NO_PIE_SPEC): Delete.
21872         (PIE_SPEC): Define as !no-pie/pie.  Move static|shared|r
21873         exclusion..
21874         (LINK_PIE_SPEC): ..to here.
21875         (LINK_COMMAND_SPEC): Support -no-pie.
21876         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Correct
21877         chain of crtbegin*.o selection, update for PIE_SPEC changes and
21878         format.
21879         (GNU_USER_TARGET_ENDFILE_SPEC): Similarly.
21880         * config/sol2.h (STARTFILE_CRTBEGIN_SPEC): Similarly.
21881         (ENDFILE_CRTEND_SPEC): Similarly.
21883 2017-08-08  Uros Bizjak  <ubizjak@gmail.com>
21885         PR target/81708
21886         * config/i386/i386.opt (mstack-protector-guard-reg=): New option
21887         (mstack-protector-guard-offset=): Ditto.
21888         * config/i386/i386.c (ix86_option_override): Handle
21889         -mstack-protector-guard-reg= and -mstack-protector-guard-offset=
21890         options.
21891         (ix86_stack_protect_guard): Use ix86_stack_protect_guard_reg and
21892         ix86_stack_protect_guard_offset variables.
21893         (TARGET_STACK_PROTECT_GUARD): Always define.
21894         * doc/invoke.texi (x86 Options): Document -mstack-protector-guard-reg=
21895         and -mstack-protector-guard-offset= options.
21897 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21899         * tree-ssa-loop-ivopts.c (relate_compare_use_with_all_cands): Handle
21900         boundary case for the last candidate.
21902 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21904         * doc/invoke.texi: Document -ftree-loop-distribution for O3.
21905         * opts.c (default_options_table): Add OPT_ftree_loop_distribution.
21907 2017-08-08  Tamar Christina  <tamar.christina@arm.com>
21909         PR middle-end/19706
21910         * config/aarch64/aarch64.md (xorsign<mode>3): New optabs.
21911         * config/aarch64/aarch64-builtins.c
21912         (aarch64_builtin_vectorized_function): Added CASE_CFN_XORSIGN.
21913         * config/aarch64/aarch64-simd-builtins.def: Added xorsign BINOP.
21914         * config/aarch64/aarch64-simd.md: Added xorsign<mode>3
21916 2017-08-08  Tamar Christina  <tamar.christina@arm.com>
21917             Andrew Pinski <pinskia@gmail.com>
21919         PR middle-end/19706
21920         * internal-fn.def (XORSIGN): New.
21921         * optabs.def (xorsign_optab): New.
21922         * tree-ssa-math-opts.c (is_copysign_call_with_1): New.
21923         (convert_expand_mult_copysign): New.
21924         (pass_optimize_widening_mul::execute): Call
21925         convert_expand_mult_copysign.
21927 2017-08-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21929         PR tree-optimization/81354
21930         * gimple-ssa-strength-reduction.c (create_add_on_incoming_edge):
21931         Insert on edges rather than explicitly creating landing pads.
21932         (analyze_candidates_and_replace): Commit edge inserts.
21934 2017-08-08  Richard Biener  <rguenther@suse.de>
21936         PR middle-end/81719
21937         * tree-ssa-loop-niter.c: Include tree-dfa.h.
21938         (expand_simple_operations): Also look through ADDR_EXPRs with
21939         MEM_REF bases treating them as POINTER_PLUS_EXPR.
21941 2017-08-08  Richard Biener  <rguenther@suse.de>
21943         PR tree-optimization/81723
21944         * tree-vect-slp.c (struct bst_traits): New hash traits.
21945         (bst_fail): New global.
21946         (vect_build_slp_tree_2): New worker, split out from ...
21947         (vect_build_slp_tree): ... this now wrapping it with using
21948         bst_fail set to cache SLP tree build fails.  Properly handle
21949         max_tree_size.
21950         (vect_analyze_slp_instance): Allocate and free bst_fail.
21952 2017-08-08  Martin Liska  <mliska@suse.cz>
21954         PR tree-opt/81696
21955         * ipa-icf-gimple.c (func_checker::compare_cst_or_decl): Consider
21956         LABEL_DECLs that can be from a different function.
21958 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21960         PR tree-optimization/81744
21961         * tree-predcom.c (prepare_finalizers_chain): Deep copy expr of
21962         loop's number of iterations.
21964 2017-08-08  Martin Liska  <mliska@suse.cz>
21966         * asan.c: Include header files.
21967         * attribs.c (build_decl_attribute_variant): New function moved
21968         from tree.[ch].
21969         (build_type_attribute_qual_variant): Likewise.
21970         (cmp_attrib_identifiers): Likewise.
21971         (simple_cst_list_equal): Likewise.
21972         (omp_declare_simd_clauses_equal): Likewise.
21973         (attribute_value_equal): Likewise.
21974         (comp_type_attributes): Likewise.
21975         (build_type_attribute_variant): Likewise.
21976         (lookup_ident_attribute): Likewise.
21977         (remove_attribute): Likewise.
21978         (merge_attributes): Likewise.
21979         (merge_type_attributes): Likewise.
21980         (merge_decl_attributes): Likewise.
21981         (merge_dllimport_decl_attributes): Likewise.
21982         (handle_dll_attribute): Likewise.
21983         (attribute_list_equal): Likewise.
21984         (attribute_list_contained): Likewise.
21985         * attribs.h (lookup_attribute): New function moved from tree.[ch].
21986         (lookup_attribute_by_prefix): Likewise.
21987         * bb-reorder.c: Include header files.
21988         * builtins.c: Likewise.
21989         * calls.c: Likewise.
21990         * cfgexpand.c: Likewise.
21991         * cgraph.c: Likewise.
21992         * cgraphunit.c: Likewise.
21993         * convert.c: Likewise.
21994         * dwarf2out.c: Likewise.
21995         * final.c: Likewise.
21996         * fold-const.c: Likewise.
21997         * function.c: Likewise.
21998         * gimple-expr.c: Likewise.
21999         * gimple-fold.c: Likewise.
22000         * gimple-pretty-print.c: Likewise.
22001         * gimple.c: Likewise.
22002         * gimplify.c: Likewise.
22003         * hsa-common.c: Likewise.
22004         * hsa-gen.c: Likewise.
22005         * internal-fn.c: Likewise.
22006         * ipa-chkp.c: Likewise.
22007         * ipa-cp.c: Likewise.
22008         * ipa-devirt.c: Likewise.
22009         * ipa-fnsummary.c: Likewise.
22010         * ipa-inline.c: Likewise.
22011         * ipa-visibility.c: Likewise.
22012         * ipa.c: Likewise.
22013         * lto-cgraph.c: Likewise.
22014         * omp-expand.c: Likewise.
22015         * omp-general.c: Likewise.
22016         * omp-low.c: Likewise.
22017         * omp-offload.c: Likewise.
22018         * omp-simd-clone.c: Likewise.
22019         * opts-global.c: Likewise.
22020         * passes.c: Likewise.
22021         * predict.c: Likewise.
22022         * sancov.c: Likewise.
22023         * sanopt.c: Likewise.
22024         * symtab.c: Likewise.
22025         * toplev.c: Likewise.
22026         * trans-mem.c: Likewise.
22027         * tree-chkp.c: Likewise.
22028         * tree-eh.c: Likewise.
22029         * tree-into-ssa.c: Likewise.
22030         * tree-object-size.c: Likewise.
22031         * tree-parloops.c: Likewise.
22032         * tree-profile.c: Likewise.
22033         * tree-ssa-ccp.c: Likewise.
22034         * tree-ssa-live.c: Likewise.
22035         * tree-ssa-loop.c: Likewise.
22036         * tree-ssa-sccvn.c: Likewise.
22037         * tree-ssa-structalias.c: Likewise.
22038         * tree-ssa.c: Likewise.
22039         * tree-streamer-in.c: Likewise.
22040         * tree-vectorizer.c: Likewise.
22041         * tree-vrp.c: Likewise.
22042         * tsan.c: Likewise.
22043         * ubsan.c: Likewise.
22044         * varasm.c: Likewise.
22045         * varpool.c: Likewise.
22046         * tree.c: Remove functions moved to attribs.[ch].
22047         * tree.h: Likewise.
22048         * config/aarch64/aarch64.c: Add attrs.h header file.
22049         * config/alpha/alpha.c: Likewise.
22050         * config/arc/arc.c: Likewise.
22051         * config/arm/arm.c: Likewise.
22052         * config/avr/avr.c: Likewise.
22053         * config/bfin/bfin.c: Likewise.
22054         * config/c6x/c6x.c: Likewise.
22055         * config/cr16/cr16.c: Likewise.
22056         * config/cris/cris.c: Likewise.
22057         * config/darwin.c: Likewise.
22058         * config/epiphany/epiphany.c: Likewise.
22059         * config/fr30/fr30.c: Likewise.
22060         * config/frv/frv.c: Likewise.
22061         * config/ft32/ft32.c: Likewise.
22062         * config/h8300/h8300.c: Likewise.
22063         * config/i386/winnt.c: Likewise.
22064         * config/ia64/ia64.c: Likewise.
22065         * config/iq2000/iq2000.c: Likewise.
22066         * config/lm32/lm32.c: Likewise.
22067         * config/m32c/m32c.c: Likewise.
22068         * config/m32r/m32r.c: Likewise.
22069         * config/m68k/m68k.c: Likewise.
22070         * config/mcore/mcore.c: Likewise.
22071         * config/microblaze/microblaze.c: Likewise.
22072         * config/mips/mips.c: Likewise.
22073         * config/mmix/mmix.c: Likewise.
22074         * config/mn10300/mn10300.c: Likewise.
22075         * config/moxie/moxie.c: Likewise.
22076         * config/msp430/msp430.c: Likewise.
22077         * config/nds32/nds32-isr.c: Likewise.
22078         * config/nds32/nds32.c: Likewise.
22079         * config/nios2/nios2.c: Likewise.
22080         * config/nvptx/nvptx.c: Likewise.
22081         * config/pa/pa.c: Likewise.
22082         * config/pdp11/pdp11.c: Likewise.
22083         * config/powerpcspe/powerpcspe.c: Likewise.
22084         * config/riscv/riscv.c: Likewise.
22085         * config/rl78/rl78.c: Likewise.
22086         * config/rx/rx.c: Likewise.
22087         * config/s390/s390.c: Likewise.
22088         * config/sh/sh.c: Likewise.
22089         * config/sol2.c: Likewise.
22090         * config/sparc/sparc.c: Likewise.
22091         * config/spu/spu.c: Likewise.
22092         * config/stormy16/stormy16.c: Likewise.
22093         * config/tilegx/tilegx.c: Likewise.
22094         * config/tilepro/tilepro.c: Likewise.
22095         * config/v850/v850.c: Likewise.
22096         * config/vax/vax.c: Likewise.
22097         * config/visium/visium.c: Likewise.
22098         * config/xtensa/xtensa.c: Likewise.
22100 2017-08-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
22102         PR target/81593
22103         * config/rs6000/vsx.md (vsx_concat_<mode>, VSX_D): Cleanup
22104         constraints since the -mupper-regs-* switches have been
22105         eliminated.
22106         (vsx_concat_<mode>_1): New combiner insns to recognize inserting
22107         into a vector from a double word element that was extracted from
22108         another vector, and eliminate extra XXPERMDI instructions.
22109         (vsx_concat_<mode>_2): Likewise.
22110         (vsx_concat_<mode>_3): Likewise.
22111         (vsx_set_<mode>, VSX_D): Rewrite vector set in terms of vector
22112         concat to allow optimizing inserts from previous extracts.
22114 2017-08-07  Uros Bizjak  <ubizjak@gmail.com>
22116         * config/i386/i386.c (ix86_stack_protect_guard): Generate
22117         memory reference to a SSP offset in TLS address space.
22118         (ix86_print_operand) <case '@'>: Remove.
22119         (ix86_print_operand_punct_valid_p): Remove '@' code.
22120         * config/i386/i386.md (unspec): Remove UNSPEC_SP_TLS_SET and
22121         UNSPEC_SP_TLS_TEST.
22122         (stack_tls_protect_set_<mode>): Remove.
22123         (stack_protect_set): Do not call gen_stack_tls_protect_set_<mode>.
22124         (stack_tls_protect_test_<mode>): Remove.
22125         (stack_protect_test): Do not call gen_stack_tls_protect_test_<mode>.
22127 2017-08-07  Olivier Hainque  <hainque@adacore.com>
22129         PR target/81755
22130         * config/vxworksae.h (VXWORKS_HAVE_TLS): Define.
22132 2017-08-07  Douglas Rupp  <rupp@adacore.com>
22134         * Makefile.in (install-mkheaders): Fix typo, where the multi_dir
22135         variable was referenced as multidir in command.
22137 2017-08-07  Jakub Jelinek  <jakub@redhat.com>
22139         PR c/69389
22140         * gimplify.c (goa_stabilize_expr): Handle BIT_INSERT_EXPR and
22141         BIT_FIELD_REF.
22143 2017-08-07  Martin Liska  <mliska@suse.cz>
22145         * config/m32c/m32c.c: Add include of stringpool.h and attribs.h.
22146         * config/rl78/rl78.c: Add include of attribs.h.
22147         * config/sh/sh.c: Likewise.
22148         * config/v850/v850.c: Likewise.
22150 2017-08-07  Tom de Vries  <tom@codesourcery.com>
22152         PR middle-end/78266
22153         * omp-expand.c (expand_oacc_for): Ensure diff_type is large enough.
22155 2017-08-07  Martin Liska  <mliska@suse.cz>
22157         * config/mips/mips.c: Include attribs.h.
22159 2017-08-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
22161         PR fortran/68829
22162         * doc/invoke.texi: Document change in behvaior for -Ofast for
22163         Fortran.
22165 2017-08-07  Wilco Dijkstra  <wdijkstr@arm.com>
22167         * config/aarch64/aarch64.c (aarch64_pushwb_single_reg):
22168         Use gen_frame_mem.
22169         (aarch64_pop_regs): Likewise.
22170         (aarch64_gen_load_pair): Likewise.
22171         (aarch64_save_callee_saves): Likewise.
22172         (aarch64_restore_callee_saves): Likewise.
22174 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
22176         * config/i386/i386.c: Revert the last change.
22178 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
22180         PR target/81736
22181         * config/i386/i386.c (ix86_finalize_stack_realign_flags): Renamed
22182         to ...
22183         (ix86_finalize_stack_frame_flags): This.  Also clear
22184         frame_pointer_needed if -fno-omit-frame-pointer is used without
22185         stack access.
22186         (ix86_expand_prologue): Replace ix86_finalize_stack_realign_flags
22187         with ix86_finalize_stack_frame_flags.
22188         (ix86_expand_epilogue): Likewise.
22189         (ix86_expand_split_stack_prologue): Likewise.
22191 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
22193         PR target/81743
22194         * config/i386/i386.c (get_builtin_code_for_version): Set priority
22195         to P_AES for Westmere.
22197 2017-08-07  Jonathan Yong  <10walls@gmail.com>
22199         * config/i386/mingw.opt (fset-stack-executable): Removed.
22200         * config/i386/cygming.opt (fset-stack-executable): Moved
22201         from mingw.opt.
22202         * config/i386/cygwin.h: Define CHECK_EXECUTE_STACK_ENABLED.
22204 2017-08-07  Segher Boessenkool  <segher@kernel.crashing.org>
22206         * print-rtl.c (print_exp): Print NOT as "~" instead of as "!".
22208 2017-08-07  Marek Polacek  <polacek@redhat.com>
22210         PR middle-end/81737
22211         * fold-const.c (fold_indirect_ref_1): Check type_domain.
22213 2017-08-07  Martin Liska  <mliska@suse.cz>
22215         * attribs.h (canonicalize_attr_name): New function.
22216         (cmp_attribs): Move from c-format.c and adjusted.
22217         (is_attribute_p): Moved from tree.h.
22218         * tree-inline.c: Add new includes.
22219         * tree.c (cmp_attrib_identifiers): Use cmp_attribs.
22220         (private_is_attribute_p): Remove.
22221         (private_lookup_attribute): Likewise.
22222         (private_lookup_attribute_by_prefix): Simplify.
22223         (remove_attribute): Use is_attribute_p.
22224         * tree.h: Remove removed declarations.
22226 2017-08-07  Jakub Jelinek  <jakub@redhat.com>
22228         PR middle-end/81698
22229         * stmt.c (emit_case_dispatch_table): Add DEFAULT_EDGE argument,
22230         instead of computing it in the function.  Formatting fix.
22231         (expand_case): Don't rely on default_edge being the first edge,
22232         clear it if removing it, pass default_edge to
22233         emit_case_dispatch_table.
22234         (expand_sjlj_dispatch_table): Pass NULL as DEFAULT_EDGE, formatting
22235         fix.
22237 2017-08-06  Uros Bizjak  <ubizjak@gmail.com>
22239         * config/alpha/alpha.c (alpha_reorg): If trap is the last active
22240         insn in the function, emit NOP after the insn.
22242 2017-08-06  Tom de Vries  <tom@codesourcery.com>
22244         * omp-expand.c (expand_oacc_for): Add missing edge probability for tile
22245         and element loops.
22247 2017-08-06  Tom de Vries  <tom@codesourcery.com>
22249         * omp-expand.c (expand_oacc_for): Add missing edge probability for chunk
22250         loop.
22252 2017-08-04  Yury Gribov  <tetra2005@gmail.com>
22254         PR tree-optimization/57371
22255         * match.pd: New pattern.
22257 2017-08-04  Marek Polacek  <polacek@redhat.com>
22259         PR middle-end/81695
22260         * fold-const.c (fold_indirect_ref_1): For ((int *)&a + 4 -> a[1],
22261         perform the computation in offset_int.
22263 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22265         PR tree-optimization/81136
22266         * tree-vectorizer.h: Include tree-hash-traits.h.
22267         (vec_base_alignments): New typedef.
22268         (vec_info): Add a base_alignments field.
22269         (vect_record_base_alignments): Declare.
22270         * tree-data-ref.h (data_reference): Add an is_conditional_in_stmt
22271         field.
22272         (DR_IS_CONDITIONAL_IN_STMT): New macro.
22273         (create_data_ref): Add an is_conditional_in_stmt argument.
22274         * tree-data-ref.c (create_data_ref): Likewise.  Use it to initialize
22275         the is_conditional_in_stmt field.
22276         (data_ref_loc): Add an is_conditional_in_stmt field.
22277         (get_references_in_stmt): Set the is_conditional_in_stmt field.
22278         (find_data_references_in_stmt): Update call to create_data_ref.
22279         (graphite_find_data_references_in_stmt): Likewise.
22280         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Likewise.
22281         * tree-vect-data-refs.c (vect_analyze_data_refs): Likewise.
22282         (vect_record_base_alignment): New function.
22283         (vect_record_base_alignments): Likewise.
22284         (vect_compute_data_ref_alignment): Adjust base_addr and aligned_to
22285         for nested statements even if we fail to compute a misalignment.
22286         Use pooled base alignments for unconditional references.
22287         (vect_find_same_alignment_drs): Compare base addresses instead
22288         of base objects.
22289         (vect_analyze_data_refs_alignment): Call vect_record_base_alignments.
22290         * tree-vect-slp.c (vect_slp_analyze_bb_1): Likewise.
22292 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22294         * tree-vectorizer.h (vec_info): Add a constructor and destructor.
22295         Add an explicit name for the enum.  Use auto_vec for slp_instances
22296         and grouped_stores.
22297         (_loop_vec_info): Add a constructor and destructor.  Use auto_vec
22298         for all vectors.
22299         (_bb_vec_info): Add a constructor and destructor.
22300         (vinfo_for_stmt): Return NULL for uids of -1 as well.
22301         (destroy_loop_vec_info): Delete.
22302         (vect_destroy_datarefs): Likewise.
22303         * tree-vectorizer.c (vect_destroy_datarefs): Delete.
22304         (vec_info::vec_info): New function.
22305         (vec_info::~vec_info): Likewise.
22306         (vectorize_loops): Use delete instead of destroy_loop_vec_info.
22307         * tree-parloops.c (gather_scalar_reductions): Use delete instead of
22308         destroy_loop_vec_info.
22309         * tree-vect-loop.c (new_loop_vec_info): Replace with...
22310         (_loop_vec_info::_loop_vec_info): ...this.
22311         (destroy_loop_vec_info): Replace with...
22312         (_loop_vec_info::~_loop_vec_info): ...this.  Unconditionally delete
22313         the stmt_vec_infos.  Leave handling of vec_info information to its
22314         destructor.  Remove explicit vector releases.
22315         (vect_analyze_loop_form): Use new instead of new_loop_vec_info.
22316         (vect_analyze_loop): Use delete instead of destroy_loop_vec_info.
22317         * tree-vect-slp.c (new_bb_vec_info): Replace with...
22318         (_bb_vec_info::_bb_vec_info): ...this.  Don't reserve space in
22319         BB_VINFO_GROUPED_STORES or BB_VINFO_SLP_INSTANCES.
22320         (destroy_bb_vec_info): Replace with...
22321         (_bb_vec_info::~_bb_vec_info): ...this.  Leave handling of vec_info
22322         information to its destructor.
22323         (vect_slp_analyze_bb_1): Use new and delete instead of
22324         new_bb_vec_info and destroy_bb_vec_info.
22325         (vect_slp_bb): Replace 2 calls to destroy_bb_vec_info with a
22326         single delete.
22328 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22330         * tree-data-ref.h (subscript): Add access_fn field.
22331         (data_dependence_relation): Add could_be_independent_p.
22332         (SUB_ACCESS_FN, DDR_COULD_BE_INDEPENDENT_P): New macros.
22333         (same_access_functions): Move to tree-data-ref.c.
22334         * tree-data-ref.c (ref_contains_union_access_p): New function.
22335         (access_fn_component_p): Likewise.
22336         (access_fn_components_comparable_p): Likewise.
22337         (dr_analyze_indices): Add a reference to access_fn_component_p.
22338         (dump_data_dependence_relation): Use SUB_ACCESS_FN instead of
22339         DR_ACCESS_FN.
22340         (constant_access_functions): Likewise.
22341         (add_other_self_distances): Likewise.
22342         (same_access_functions): Likewise.  (Moved from tree-data-ref.h.)
22343         (initialize_data_dependence_relation): Use XCNEW and remove
22344         explicit zeroing of DDR_REVERSED_P.  Look for a subsequence
22345         of access functions that have the same type.  Allow the
22346         subsequence to end with different bases in some circumstances.
22347         Record the chosen access functions in SUB_ACCESS_FN.
22348         (build_classic_dist_vector_1): Replace ddr_a and ddr_b with
22349         a_index and b_index.  Use SUB_ACCESS_FN instead of DR_ACCESS_FN.
22350         (subscript_dependence_tester_1): Likewise dra and drb.
22351         (build_classic_dist_vector): Update calls accordingly.
22352         (subscript_dependence_tester): Likewise.
22353         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Check
22354         DDR_COULD_BE_INDEPENDENT_P.
22355         * tree-vectorizer.h (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Test
22356         comp_alias_ddrs instead of may_alias_ddrs.
22357         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
22358         New function.
22359         (vect_analyze_data_ref_dependence): Use it if
22360         DDR_COULD_BE_INDEPENDENT_P, but fall back to using the recorded
22361         distance vectors if that fails.
22362         (dependence_distance_ge_vf): New function.
22363         (vect_prune_runtime_alias_test_list): Use it.  Don't clear
22364         LOOP_VINFO_MAY_ALIAS_DDRS.
22366 2017-08-04  Richard Biener  <rguenther@suse.de>
22368         PR middle-end/81705
22369         * fold-const.c (fold_binary_loc): Properly restrict
22370         minus_var0 && minus_var1 case when associating undefined overflow
22371         entities.
22373 2017-08-04  Tom de Vries  <tom@codesourcery.com>
22375         * omp-simd-clone.c (simd_clone_adjust): Add missing edge probability.
22377 2017-08-03  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
22379         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
22380         Don't start diagnostic messages with a capital letter.
22381         * config/rs6000/rs6000.c (rs6000_option_override_internal):
22382         Likewise.
22383         (rs6000_invalid_builtin): Likewise.
22384         (rs6000_trampoline_init): Likewise.
22386 2017-08-03  Jakub Jelinek  <jakub@redhat.com>
22388         PR target/81621
22389         * bb-reorder.c (pass_partition_blocks::execute): Return TODO_df_finish
22390         after setting changeable df flags.
22392 2017-08-03  Richard Biener  <rguenther@suse.de>
22394         * tree-ssa-reassoc.c (should_break_up_subtract): Also break
22395         up if the use is in USE - X.
22397 2017-08-03  Alexander Monakov  <amonakov@ispras.ru>
22399         * toplev.c (dumpfile.h): New include.
22400         (internal_error_reentered): New static function.  Use it...
22401         (internal_error_function): ...here to handle reentered internal_error.
22403 2017-08-03  Richard Biener  <rguenther@suse.de>
22405         PR middle-end/81148
22406         * fold-const.c (split_tree): Add minus_var and minus_con
22407         arguments, remove unused loc arg.  Never generate NEGATE_EXPRs
22408         here but always use minus_*.
22409         (associate_trees): Assert we never associate with MINUS_EXPR
22410         and NULL first operand.  Do not recurse for PLUS_EXPR operands
22411         when associating as MINUS_EXPR either.
22412         (fold_binary_loc): Track minus_var and minus_con.
22414 2017-08-03  Tom de Vries  <tom@codesourcery.com>
22416         PR lto/81430
22417         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers): If
22418         ACCEL_COMPILER, apply finish_options on
22419         DECL_FUNCTION_SPECIFIC_OPTIMIZATION.
22421 2017-08-03  Tom de Vries  <tom@codesourcery.com>
22423         PR target/81662
22424         * config/nvptx/nvptx.c (nvptx_option_override): Emit sorry if
22425         function_entry_patch_area_size > 0.
22427 2017-08-03  Jakub Jelinek  <jakub@redhat.com>
22429         PR driver/81650
22430         * calls.c (alloc_max_size): Use HOST_WIDE_INT_UC (10??)
22431         instead of 10??LU, perform unit multiplication in wide_int,
22432         don't change alloc_object_size_limit if the limit is larger
22433         than SSIZE_MAX.
22435         PR tree-optimization/81655
22436         PR tree-optimization/81588
22437         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Handle also
22438         the case when ranges[i].low and high are 1 for unsigned type with
22439         precision 1.
22441         PR middle-end/81052
22442         * omp-low.c (diagnose_sb_0): Handle flag_openmp_simd like flag_openmp.
22443         (pass_diagnose_omp_blocks::gate): Enable also for flag_openmp_simd.
22445 2017-08-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
22447         * tree-vrp.h: Add include guard.
22449 2017-08-02  Uros Bizjak  <ubizjak@gmail.com>
22451         PR target/81644
22452         * config/i386/i386.md (unspecv): Add UNSPECV_UD2.
22453         (ud2): New insn pattern.
22454         * config/i386/i386.c (ix86_expand_epilogue):
22455         For naked functions, generate ud2 instead of trap insn.
22457 2017-08-02  Marek Polacek  <polacek@redhat.com>
22459         PR other/81667
22460         * alloc-pool.h (base_pool_allocator): Initialize m_elt_size.
22462 2017-08-02  Tom de Vries  <tom@codesourcery.com>
22463             Cesar Philippidis  <cesar@codesourcery.com>
22465         * config/nvptx/nvptx.c (nvptx_lockless_update, nvptx_lockfull_update):
22466         Add missing edge probabilities.
22468 2017-08-02  Tamar Christina  <tamar.christina@arm.com>
22470         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
22471         Correct endianness.
22473 2017-08-02  Jakub Jelinek  <jakub@redhat.com>
22475         PR middle-end/79499
22476         * function.c (thread_prologue_and_epilogue_insns): Determine blocks
22477         for find_many_sub_basic_blocks bitmap by looking up BLOCK_FOR_INSN
22478         of first NONDEBUG_INSN_P in each of the split_prologue_seq and
22479         prologue_seq sequences - if any.
22481 2017-08-02  Richard Biener  <rguenther@suse.de>
22483         * tree-vect-stmts.c (vectorizable_store): Perform vector extracts
22484         via vectors if supported, integer extracts via punning if supported
22485         or otherwise vector extracts.
22487 2017-08-02  Richard Biener  <rguenther@suse.de>
22489         * tree-ssa-pre.c (bitmap_insert_into_set_1): Remove and inline
22490         into ...
22491         (bitmap_insert_into_set): ... this.
22493 2017-08-02  Richard Biener  <rguenther@suse.de>
22495         PR tree-optimization/81633
22496         Revert
22497         2015-08-17  Alan Hayward  <alan.hayward@arm.com>
22499         PR tree-optimization/71752
22500         * tree-vect-slp.c (vect_get_slp_defs): Handle null operands.
22502 2017-08-01  Daniel Santos  <daniel.santos@pobox.com>
22504         * config/i386/i386.h (ix86_frame::outlined_save_offset): Remove field.
22505         (machine_function::call_ms2sysv_pad_out): Remove field.
22506         * config/i386/i386.c (xlogue_layout::get_stack_space_used): Modify.
22507         (ix86_compute_frame_layout): Likewise.
22509 2017-08-01  H.J. Lu  <hongjiu.lu@intel.com>
22511         PR target/81654
22512         * config/i386/i386.c (ix86_set_func_type): Disallow naked
22513         attribute with interrupt attribute.
22515 2017-08-01  Andrew Pinski  <apinski@cavium.com>
22517         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Check
22518         BIT_INSERT_EXPR's operand 1
22519         to see if the types precision matches.
22521 2017-08-01  Martin Liska  <mliska@suse.cz>
22523         PR middle-end/70140
22524         * builtins.c (expand_builtin_memcpy_args): Remove.
22525         (expand_builtin_memcpy): Call newly added function
22526         expand_builtin_memory_copy_args.
22527         (expand_builtin_memcpy_with_bounds): Likewise.
22528         (expand_builtin_mempcpy): Remove last argument.
22529         (expand_builtin_mempcpy_with_bounds): Likewise.
22530         (expand_builtin_memory_copy_args): New function created from
22531         expand_builtin_mempcpy_args with small modifications.
22532         (expand_builtin_mempcpy_args): Remove.
22533         (expand_builtin_stpcpy): Remove unused argument.
22534         (expand_builtin): Likewise.
22535         (expand_builtin_with_bounds): Likewise.
22537 2017-08-01  Martin Liska  <mliska@suse.cz>
22539         Revert r250771
22540         Make mempcpy more optimal (PR middle-end/70140).
22542 2017-08-01  Jakub Jelinek  <jakub@redhat.com>
22544         PR target/81622
22545         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): For
22546         __builtin_vec_cmpne verify both arguments are compatible vectors
22547         before looking at TYPE_MODE on the element type.  For __builtin_vec_ld
22548         verify arg1_type is a pointer or array type.  For __builtin_vec_st,
22549         move computation of aligned to after checking the argument types.
22550         Formatting fixes.
22552         PR target/80846
22553         * config/rs6000/vsx.md (vextract_fp_from_shorth,
22554         vextract_fp_from_shortl): Add element mode after mode in gen_vec_init*
22555         calls.
22557 2017-08-01  Jerome Lambourg  <lambourg@adacore.com>
22558             Doug Rupp  <rupp@adacore.com>
22559             Olivier Hainque  <hainque@adacore.com>
22561         * config.gcc (arm-wrs-vxworks*): Rework to handle arm-wrs-vxworks7 as
22562         well as arm-wrs-vxworks. Update target_cpu_name from arm6 (arch v3) to
22563         arm8 (arch v4).
22564         * config/arm/vxworks.h (MAYBE_TARGET_BPABI_CPP_BUILTINS): New, helper
22565         for TARGET_OS_CPP_BUILTIN.
22566         (TARGET_OS_CPP_BUILTIN): Invoke MAYBE_TARGET_BPABI_CPP_BUILTINS(),
22567         refine CPU definitions for arm_arch5 and add those for arm_arch6 and
22568         arm_arch7.
22569         (MAYBE_ASM_ABI_SPEC): New, helper for SUBTARGET_EXTRA_ASM_SPEC,
22570         passing required abi options to the assembler for EABI configurations.
22571         (EXTRA_CC1_SPEC): New macro, to help prevent the implicit production
22572         of .text.hot and .text.unlikely sections for kernel modules when
22573         using ARM style exceptions.
22574         (CC1_SPEC): Remove obsolete attempt at mimicking Diab toolchain
22575         options. Add EXTRA_CC1_SPEC.
22576         (VXWORKS_ENDIAN_SPEC): Adjust comment and remove handling of Diab
22577         toolchain options.
22578         (DWARF2_UNWIND_INFO): Redefine to handle the pre/post VxWorks 7
22579         transition.
22580         (ARM_TARGET2_DWARF_FORMAT): Define.
22581         * config/arm/t-vxworks: Adjust multilib control to removal of the
22582         Diab command line options.
22584 2017-08-01  Martin Liska  <mliska@suse.cz>
22586         PR gcov-profile/81561
22587         * gcov.c (unblock): Make unblocking safe as we need to preserve
22588         index correspondence of blocks and block_lists.
22590 2017-08-01  Richard Biener  <rguenther@suse.de>
22592         PR tree-optimization/81181
22593         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
22594         (compute_antic): ... end of iteration here.
22596 2017-08-01  James Greenhalgh  <james.greenhalgh@arm.com>
22598         * common.opt (ftree-vectorize): No longer set flag_tree_vectorize.
22599         (ftree-loop-vectorize): Set as EnabledBy ftree-vectorize.
22600         (ftree-slp-vectorize): Likewise.
22601         * omp-expand (expand_omp_simd): Remove flag_tree_vectorize, as it
22602         can no longer be set independent of flag_tree_loop_vectorize.
22603         * omp-general.c (emp_max_vf): Likewise.
22604         * opts.c (enable_fdo_optimizations): Remove references to
22605         flag_tree_vectorize, these are now implicit.
22606         (common_handle_option): Remove handling for OPT_ftree_vectorize,
22607         and leave it for the options machinery.
22609 2017-08-01  Martin Liska  <mliska@suse.cz>
22611         PR middle-end/70140
22612         * builtins.c (expand_builtin_memcpy_args): Remove.
22613         (expand_builtin_memcpy): Call newly added function
22614         expand_builtin_memory_copy_args.
22615         (expand_builtin_memcpy_with_bounds): Likewise.
22616         (expand_builtin_mempcpy): Remove last argument.
22617         (expand_builtin_mempcpy_with_bounds): Likewise.
22618         (expand_builtin_memory_copy_args): New function created from
22619         expand_builtin_mempcpy_args with small modifications.
22620         (expand_builtin_mempcpy_args): Remove.
22621         (expand_builtin_stpcpy): Remove unused argument.
22622         (expand_builtin): Likewise.
22623         (expand_builtin_with_bounds): Likewise.
22625 2017-08-01  Uros Bizjak  <ubizjak@gmail.com>
22627         PR target/81641
22628         * config/i386/i386.c (ix86_print_operand_address_as): For -masm=intel
22629         print "ds:" only for immediates in generic address space.
22631 2017-08-01  Uros Bizjak  <ubizjak@gmail.com>
22633         PR target/81639
22634         * config/i386/i386.c (ix86_funciton_naked): New prototype.
22635         (ix86_function_ok_for_sibcall): Return false for naked functions.
22637 2017-08-01  Richard Biener  <rguenther@suse.de>
22639         * tree-ssa-pre.c (print_pre_expr): Handle NULL expr.
22640         (compute_antic): Seed worklist with exit block predecessors.
22641         * cfganal.c (dfs_find_deadend): For a cycle return the source
22642         of the edge closing it.
22644 2017-08-01  Tamar Christina  <tamar.christina@arm.com>
22646         * config/aarch64/aarch64.c
22647         (aarch64_can_const_movi_rtx_p): Move 0 check.
22649 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22651         * tree.h (POINTER_TYPE_OVERFLOW_UNDEFINED): Delete.
22652         * fold-const.c (fold_comparison, fold_binary_loc): Delete use of
22653         above macro.
22654         * match.pd: Ditto in address comparison pattern.
22656 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22658         PR tree-optimization/81627
22659         * tree-predcom.c (prepare_finalizers): Always rewrite into loop
22660         closed ssa form for store-store chain.
22662 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22664         PR tree-optimization/81620
22665         * tree-predcom.c (add_ref_to_chain): Don't set has_max_use_after
22666         for store-store chain.
22668 2017-08-01  Jakub Jelinek  <jakub@redhat.com>
22670         PR tree-optimization/81588
22671         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If
22672         ranges[i].in_p, invert comparison code ccode.  For >/>=,
22673         swap rhs1 and rhs2 and comparison code unconditionally,
22674         for </<= don't do that.  Don't swap rhs1/rhs2 again if
22675         ranges[i].in_p, instead invert comparison code ccode if
22676         opcode or oe->rank is BIT_IOR_EXPR.
22678         PR target/80846
22679         * optabs.def (vec_extract_optab, vec_init_optab): Change from
22680         a direct optab to conversion optab.
22681         * optabs.c (expand_vector_broadcast): Use convert_optab_handler
22682         with GET_MODE_INNER as last argument instead of optab_handler.
22683         * expmed.c (extract_bit_field_1): Likewise.  Use vector from
22684         vector extraction if possible and optab is available.
22685         * expr.c (store_constructor): Use convert_optab_handler instead
22686         of optab_handler.  Use vector initialization from smaller
22687         vectors if possible and optab is available.
22688         * tree-vect-stmts.c (vectorizable_load): Likewise.
22689         * doc/md.texi (vec_extract, vec_init): Document that the optabs
22690         now have two modes.
22691         * config/i386/i386.c (ix86_expand_vector_init): Handle expansion
22692         of vec_init from half-sized vectors with the same element mode.
22693         * config/i386/sse.md (ssehalfvecmode): Add V4TI case.
22694         (ssehalfvecmodelower, ssescalarmodelower): New mode attributes.
22695         (reduc_plus_scal_v8df, reduc_plus_scal_v4df, reduc_plus_scal_v2df,
22696         reduc_plus_scal_v16sf, reduc_plus_scal_v8sf, reduc_plus_scal_v4sf,
22697         reduc_<code>_scal_<mode>, reduc_umin_scal_v8hi): Add element mode
22698         after mode in gen_vec_extract* calls.
22699         (vec_extract<mode>): Renamed to ...
22700         (vec_extract<mode><ssescalarmodelower>): ... this.
22701         (vec_extract<mode><ssehalfvecmodelower>): New expander.
22702         (rotl<mode>3, rotr<mode>3, <shift_insn><mode>3, ashrv2di3): Add
22703         element mode after mode in gen_vec_init* calls.
22704         (VEC_INIT_HALF_MODE): New mode iterator.
22705         (vec_init<mode>): Renamed to ...
22706         (vec_init<mode><ssescalarmodelower>): ... this.
22707         (vec_init<mode><ssehalfvecmodelower>): New expander.
22708         * config/i386/mmx.md (vec_extractv2sf): Renamed to ...
22709         (vec_extractv2sfsf): ... this.
22710         (vec_initv2sf): Renamed to ...
22711         (vec_initv2sfsf): ... this.
22712         (vec_extractv2si): Renamed to ...
22713         (vec_extractv2sisi): ... this.
22714         (vec_initv2si): Renamed to ...
22715         (vec_initv2sisi): ... this.
22716         (vec_extractv4hi): Renamed to ...
22717         (vec_extractv4hihi): ... this.
22718         (vec_initv4hi): Renamed to ...
22719         (vec_initv4hihi): ... this.
22720         (vec_extractv8qi): Renamed to ...
22721         (vec_extractv8qiqi): ... this.
22722         (vec_initv8qi): Renamed to ...
22723         (vec_initv8qiqi): ... this.
22724         * config/rs6000/vector.md (VEC_base_l): New mode attribute.
22725         (vec_init<mode>): Renamed to ...
22726         (vec_init<mode><VEC_base_l>): ... this.
22727         (vec_extract<mode>): Renamed to ...
22728         (vec_extract<mode><VEC_base_l>): ... this.
22729         * config/rs6000/paired.md (vec_initv2sf): Renamed to ...
22730         (vec_initv2sfsf): ... this.
22731         * config/rs6000/altivec.md (splitter, altivec_copysign_v4sf3,
22732         vec_unpacku_hi_v16qi, vec_unpacku_hi_v8hi, vec_unpacku_lo_v16qi,
22733         vec_unpacku_lo_v8hi, mulv16qi3, altivec_vreve<mode>2): Add
22734         element mode after mode in gen_vec_init* calls.
22735         * config/aarch64/aarch64-simd.md (vec_init<mode>): Renamed to ...
22736         (vec_init<mode><Vel>): ... this.
22737         (vec_extract<mode>): Renamed to ...
22738         (vec_extract<mode><Vel>): ... this.
22739         * config/aarch64/iterators.md (Vel): New mode attribute.
22740         * config/s390/s390.c (s390_expand_vec_strlen, s390_expand_vec_movstr):
22741         Add element mode after mode in gen_vec_extract* calls.
22742         * config/s390/vector.md (non_vec_l): New mode attribute.
22743         (vec_extract<mode>): Renamed to ...
22744         (vec_extract<mode><non_vec_l>): ... this.
22745         (vec_init<mode>): Renamed to ...
22746         (vec_init<mode><non_vec_l>): ... this.
22747         * config/s390/s390-builtins.def (s390_vlgvb, s390_vlgvh, s390_vlgvf,
22748         s390_vlgvf_flt, s390_vlgvg, s390_vlgvg_dbl): Add element mode after
22749         vec_extract mode.
22750         * config/arm/iterators.md (V_elem_l): New mode attribute.
22751         * config/arm/neon.md (vec_extract<mode>): Renamed to ...
22752         (vec_extract<mode><V_elem_l>): ... this.
22753         (vec_extractv2di): Renamed to ...
22754         (vec_extractv2didi): ... this.
22755         (vec_init<mode>): Renamed to ...
22756         (vec_init<mode><V_elem_l>): ... this.
22757         (reduc_plus_scal_<mode>, reduc_plus_scal_v2di, reduc_smin_scal_<mode>,
22758         reduc_smax_scal_<mode>, reduc_umin_scal_<mode>,
22759         reduc_umax_scal_<mode>, neon_vget_lane<mode>, neon_vget_laneu<mode>):
22760         Add element mode after gen_vec_extract* calls.
22761         * config/mips/mips-msa.md (vec_init<mode>): Renamed to ...
22762         (vec_init<mode><unitmode>): ... this.
22763         (vec_extract<mode>): Renamed to ...
22764         (vec_extract<mode><unitmode>): ... this.
22765         * config/mips/loongson.md (vec_init<mode>): Renamed to ...
22766         (vec_init<mode><unitmode>): ... this.
22767         * config/mips/mips-ps-3d.md (vec_initv2sf): Renamed to ...
22768         (vec_initv2sfsf): ... this.
22769         (vec_extractv2sf): Renamed to ...
22770         (vec_extractv2sfsf): ... this.
22771         (reduc_plus_scal_v2sf, reduc_smin_scal_v2sf, reduc_smax_scal_v2sf):
22772         Add element mode after gen_vec_extract* calls.
22773         * config/mips/mips.md (unitmode): New mode iterator.
22774         * config/spu/spu.c (spu_expand_prologue, spu_allocate_stack,
22775         spu_builtin_extract): Add element mode after gen_vec_extract* calls.
22776         * config/spu/spu.md (inner_l): New mode attribute.
22777         (vec_init<mode>): Renamed to ...
22778         (vec_init<mode><inner_l>): ... this.
22779         (vec_extract<mode>): Renamed to ...
22780         (vec_extract<mode><inner_l>): ... this.
22781         * config/sparc/sparc.md (veltmode): New mode iterator.
22782         (vec_init<VMALL:mode>): Renamed to ...
22783         (vec_init<VMALL:mode><VMALL:veltmode>): ... this.
22784         * config/ia64/vect.md (vec_initv2si): Renamed to ...
22785         (vec_initv2sisi): ... this.
22786         (vec_initv2sf): Renamed to ...
22787         (vec_initv2sfsf): ... this.
22788         (vec_extractv2sf): Renamed to ...
22789         (vec_extractv2sfsf): ... this.
22790         * config/powerpcspe/vector.md (VEC_base_l): New mode attribute.
22791         (vec_init<mode>): Renamed to ...
22792         (vec_init<mode><VEC_base_l>): ... this.
22793         (vec_extract<mode>): Renamed to ...
22794         (vec_extract<mode><VEC_base_l>): ... this.
22795         * config/powerpcspe/paired.md (vec_initv2sf): Renamed to ...
22796         (vec_initv2sfsf): ... this.
22797         * config/powerpcspe/altivec.md (splitter, altivec_copysign_v4sf3,
22798         vec_unpacku_hi_v16qi, vec_unpacku_hi_v8hi, vec_unpacku_lo_v16qi,
22799         vec_unpacku_lo_v8hi, mulv16qi3): Add element mode after mode in
22800         gen_vec_init* calls.
22802 2017-08-01  Richard Biener  <rguenther@suse.de>
22804         PR tree-optimization/81297
22805         * tree-vrp.c (get_single_symbol): Remove assert, instead drop
22806         TREE_OVERFLOW from INTEGER_CSTs.
22808 2017-07-31  Segher Boessenkool  <segher@kernel.crashing.org>
22810         * config/rs6000/rs6000.c (enum rs6000_reg_type): Delete trailing comma.
22812 2017-07-31  Carl Love  <cel@us.ibm.com>
22814         * config/rs6000/rs6000-c: Add support for built-in functions
22815         vector signed char vec_xl_be (signed long long, signed char *);
22816         vector unsigned char vec_xl_be (signed long long, unsigned char *);
22817         vector signed int vec_xl_be (signed long long, signed int *);
22818         vector unsigned int vec_xl_be (signed long long, unsigned int *);
22819         vector signed long long vec_xl_be (signed long long, signed long long *);
22820         vector unsigned long long vec_xl_be (signed long long, unsigned long long *);
22821         vector signed short vec_xl_be (signed long long, signed short *);
22822         vector unsigned short vec_xl_be (signed long long, unsigned short *);
22823         vector double vec_xl_be (signed long long, double *);
22824         vector float vec_xl_be (signed long long, float *);
22825         * config/rs6000/altivec.h (vec_xl_be): Add #define.
22826         * config/rs6000/rs6000-builtin.def (XL_BE_V16QI, XL_BE_V8HI, XL_BE_V4SI)
22827         XL_BE_V2DI, XL_BE_V4SF, XL_BE_V2DF, XL_BE): Add definitions
22828         for the builtins.
22829         * config/rs6000/rs6000.c (altivec_expand_xl_be_builtin): Add function.
22830         (altivec_expand_builtin): Add switch statement to call
22831         altivec_expand_xl_be for each builtin.
22832         (altivec_init_builtins): Add def_builtin for _builtin_vsx_le_be_v8hi,
22833         __builtin_vsx_le_be_v4si, __builtin_vsx_le_be_v2di,
22834         __builtin_vsx_le_be_v4sf, __builtin_vsx_le_be_v2df,
22835         __builtin_vsx_le_be_v16qi.
22836         * doc/extend.texi: Update the built-in documentation file for the
22837         new built-in functions.
22839 2017-07-31  Uros Bizjak  <ubizjak@gmail.com>
22841         PR target/25967
22842         * config/i386/i386.c (ix86_allocate_stack_slots_for_args):
22843         New function.
22844         (TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS): Define.
22846 2017-07-31  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
22848         * config.gcc: Add z14.
22849         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
22850         CPU model numbers for z13s and z14.
22851         * config/s390/s390-c.c (s390_resolve_overloaded_builtin): Replace
22852         arch12 with z14.
22853         * config/s390/s390-opts.h (enum processor_type): Rename
22854         PROCESSOR_ARCH12 to PROCESSOR_3906_Z14.
22855         * config/s390/s390.c (processor_table): Add field for CPU name to
22856         be passed to Binutils.
22857         (s390_asm_output_machine_for_arch): Use the new field in
22858         processor_table for Binutils.
22859         (s390_expand_builtin): Replace arch12 with z14.
22860         (s390_issue_rate): Rename PROCESSOR_ARCH12 to PROCESSOR_3906_Z14.
22861         (s390_get_sched_attrmask): Likewise.
22862         (s390_get_unit_mask): Likewise.
22863         * config/s390/s390.opt: Add z14 to processor_type enum.
22865 2017-07-31  Martin Jambor  <mjambor@suse.cz>
22867         PR hsa/81477
22868         * ipa-fnsummary.c (ipa_fn_summary_generate): Set versionable
22869         regardless of optimization level.
22871 2017-07-31  Jan Hubicka <hubicka@ucw.cz>
22872             Martin Liska  <mliska@suse.cz>
22874         * predict.def: Remove old comment and adjust probability.
22875         * gimplify.c (should_warn_for_implicit_fallthrough): Ignore
22876         PREDICT statements.
22878 2017-07-31  Uros Bizjak  <ubizjak@gmail.com>
22880         PR target/25967
22881         * config/i386/i386.c (ix86_function_naked): New function.
22882         (ix86_can_use_return_insn_p): Return false for naked functions.
22883         (ix86_expand_prologue): Skip prologue for naked functions.
22884         (ix86_expand_epilogue): Skip epilogue for naked functions
22885         and emit trap instruction.
22886         (ix86_warn_func_return): New function.
22887         (ix86_attribute_table): Add "naked" attribute specification.
22888         (TARGET_WARN_FUNC_RETURN): Define.
22889         * doc/extend.texi (x86 Function Attributes) <naked>: Document it.
22891 2017-07-31  Martin Liska  <mliska@suse.cz>
22893         * gimple-pretty-print.c (dump_gimple_label): Never dump BB info.
22894         (dump_gimple_bb_header): Always dump BB info.
22895         (pp_cfg_jump): Do not append info about BB when dumping a jump.
22897 2017-07-31  Martin Liska  <mliska@suse.cz>
22899         PR sanitize/81530
22900         * convert.c (convert_to_integer_1): Guard condition with flag_sanitize_p
22901         also with current_function_decl non-null equality.
22903 2017-07-31  Jakub Jelinek  <jakub@redhat.com>
22905         PR sanitizer/81604
22906         * ubsan.c (ubsan_type_descriptor): For UBSAN_PRINT_ARRAY don't
22907         change type to the element type, instead add eltype variable and
22908         use it where we are interested in the element type.
22910         PR tree-optimization/81603
22911         * ipa-polymorphic-call.c
22912         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Perform
22913         offset arithmetic in offset_int, bail out if the resulting bit offset
22914         doesn't fit into shwi.
22916 2017-07-31  Martin Liska  <mliska@suse.cz>
22918         * gimplify.c (mostly_copy_tree_r): Remove Java specific hunk.
22919         (gimplify_save_expr): Fix comment.
22921 2017-07-30  H.J. Lu  <hongjiu.lu@intel.com>
22923         PR target/79793
22924         * config/i386/i386.c (ix86_function_arg): Update arguments for
22925         exception handler.
22926         (ix86_compute_frame_layout): Set the initial stack offset to
22927         INCOMING_FRAME_SP_OFFSET.  Update red-zone offset with
22928         INCOMING_FRAME_SP_OFFSET.
22929         (ix86_expand_epilogue): Don't pop the 'ERROR_CODE' off the
22930         stack before exception handler returns.
22931         * config/i386/i386.h (INCOMING_FRAME_SP_OFFSET): Add the
22932         the 'ERROR_CODE' for exception handler.
22934 2017-07-30  Uros Bizjak  <ubizjak@gmail.com>
22936         * config/i386/i386.h (ASM_PRINTF_EXTENSIONS): New macro.
22937         (ASM_OUTPUT_REG_PUSH): Rewrite with new operand modifiers.
22938         (ASM_OUTPUT_REG_POP): Ditto.
22939         * config/i386/i386.c (ix86_asm_output_function_label): Use fputs
22940         instead of asm_fprintf to output pure string.
22942 2017-07-29  Jakub Jelinek  <jakub@redhat.com>
22944         * debug.h (struct gcc_debug_hooks): Add IMPLICIT argument
22945         to imported_module_or_decl hook.
22946         (debug_nothing_tree_tree_tree_bool): Remove.
22947         (debug_nothing_tree_tree_tree_bool_bool): New declaration.
22948         * debug.c (do_nothing_debug_hooks): Use
22949         debug_nothing_tree_tree_tree_bool_bool instead of
22950         debug_nothing_tree_tree_tree_bool.
22951         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
22952         * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Likewise.
22953         * sdbout.c (sdb_debug_hooks): Likewise.
22954         * dwarf2out.c (dwarf2_lineno_debug_hooks): Likewise.
22955         (gen_namespace_die): Add DW_AT_export_symbols attribute if
22956         langhook wants it.
22957         (dwarf2out_imported_module_or_decl): Add IMPLICIT argument,
22958         if true, -gdwarf-5 and decl will have DW_AT_export_symbols
22959         attribute, don't add anything.
22961 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22963         * fold-const.c (fold_build1_stat_loc): Remove _stat from name.
22964         (fold_build2_stat_loc): Likewise.
22965         (fold_build3_stat_loc): Likewise.
22966         * fold-const.h (fold_build1, fold_build2, fold_build3): Adjust.
22967         (fold_build1_loc): Remove macro.
22968         (fold_build2_loc): Likewise.
22969         (fold_build3_loc): Likewise.
22971 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22973         * gimple.c (gimple_build_debug_bind_stat): Remove _stat from name.
22974         (gimple_build_debug_bind_source_stat): Likewise.
22975         * gimple.h (gimple_build_debug_bind): Remove macro.
22976         (gimple_build_debug_bind_source): Likewise.
22978 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22980         * bitmap.c (bitmap_alloc): Adjust.
22981         (bitmap_gc_alloc): Likewise.
22982         * bitmap.h (bitmap_initialize_stat): Remove _stat from name.
22984 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22986         * bitmap.c (bitmap_obstack_alloc_stat): Rename to bitmap_alloc.
22987         (bitmap_gc_alloc_stat): Rename to bitmap_gc_alloc.
22988         * bitmap.h (bitmap_obstack_alloc_stat): Adjust prototype.
22989         (bitmap_gc_alloc_stat): Likewise.
22990         (BITMAP_ALLOC, BITMAP_GGC_ALLOC): Adjust.
22992 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22994         * rtl.c (shallow_copy_rtx_stat): Remove _stat from name.
22995         * rtl.h (shallow_copy_rtx): Remove macro.
22997 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22999         * emit-rtl.c (gen_raw_REG): Adjust.
23000         * gengenrtl.c (gendef): Likewise.
23001         * rtl.c (rtx_alloc_stat): Remove _stat from name.
23002         * rtl.h (rtx_alloc): Remove macro.
23004 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
23006         * tree.c (build_tree_list_vec_stat): Remove _stat from name.
23007         (build_tree_list_stat): Likewise.
23008         * tree.h (build_tree_list): Remove macro.
23009         (build_tree_list_vec): Likewise.
23011 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
23013         * tree.c (make_vector_stat): Remove _stat from name.
23014         (build_vector_stat): Likewise.
23015         * tree.h (make_vector_stat): Remove macro.
23016         (build_vector_stat): Likewise.
23018 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
23020         * tree.h (build_var_debug_value): Remove prototype.
23022 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
23024         * tree.c (tree_cons_stat): Remove _stat from name.
23025         * tree.h (tree_cons): Remove macro.
23027 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
23029         * tree.c (build_vl_exp_stat): Remove _stat from name.
23030         * tree.h (build_vl_exp): Remove macro.
23032 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
23034         * tree.c (build_decl_stat): Remove _stat from name.
23035         * tree.h (build_decl): Remove macro.
23037 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
23039         * gimple.c (gimple_build_with_ops_stat): Adjust.
23040         (gimple_alloc_stat): Remove _stat from name.
23041         * gimple.h (gimple_alloc): Remove macro.
23043 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
23045         * tree.c (make_tree_vec_stat): Remove _stat from name.
23046         (grow_tree_vec_stat): Likewise.
23047         * tree.h (make_tree_vec_stat): Adjust prototype.
23048         (grow_tree_vec_stat): Likewise.
23049         (make_tree_vec): Remove macro.
23050         (grow_tree_vec): Likewise.
23052 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
23054         * fold-const.c (fold_build1_stat_loc): Adjust.
23055         (fold_build2_stat_loc): Likewise.
23056         (fold_build3_stat_loc): Likewise.
23057         * tree.c (build0_stat): Remove _stat from name.
23058         (build1_stat): Likewise.
23059         (build2_stat): Likewise.
23060         (build3_stat): Likewise.
23061         (build4_stat): Likewise.
23062         (build5_stat): Likewise.
23063         * tree.h (build1_loc): Remove macro, and rename _stat function
23064         to this.
23065         (build2_loc): Likewise.
23066         (build3_loc): Likewise.
23067         (build4_loc): Likewise.
23068         (build5_loc): Likewise.
23070 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
23072         * tree.c (make_int_cst_stat): Remove _stat from name.
23073         * tree.h (make_int_cst_stat): Adjust prototype.
23074         (make_int_cst): Remove macro.
23076 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
23078         * tree.c (make_tre_binfo_stat): Remove _stat from name.
23079         * tree.h (make_tree_binfo_stat): Adjust prototype.
23080         (make_tree_binfo): Remove.
23082 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
23084         * tree.c (copy_node_stat): Rename to copy_node.
23085         (build_distinct_type_copy): Adjust.
23086         * tree.h (copy_node_stat): Adjust prototype.
23087         (copy_node): Remove macro.
23089 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
23091         * tree.c (make_node_stat): rename to make_node.
23092         (build_tree_list_stat): Adjust.
23093         (build0_stat): Likewise.
23094         (build2_stat): Likewise.
23095         (build3_stat): Likewise.
23096         (build4_stat): Likewise.
23097         (build5_stat): Likewise.
23098         (build_decl_stat): Likewise.
23099         * tree.h (make_node_stat): Adjust prototype.
23100         (make_node): remove macro.
23102 2017-07-28  Peter Bergner  <bergner@vnet.ibm.com>
23104         * config/rs6000/ppc-auxv.h (PPC_FEATURE2_DARN): New define.
23105         (PPC_FEATURE2_SCV): Likewise.
23106         * config/rs6000/rs6000.c (cpu_supports_info): Use them.
23108 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
23110         * config/aarch64/aarch64.c
23111         (aarch64_internal_mov_immediate): Add new special pattern.
23112         * config/aarch64/aarch64.md (*movdi_aarch64):
23113         Add reg/32bit const mov case.
23115 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
23116             Richard Sandiford <richard.sandiford@linaro.org>
23118         * config/aarch64/aarch64.md (mov<mode>): Generalize.
23119         (*movhf_aarch64, *movsf_aarch64, *movdf_aarch64):
23120         Add integer and movi cases.
23121         (movi-split-hf-df-sf split, fp16): New.
23122         (enabled): Added TARGET_FP_F16INST.
23123         * config/aarch64/iterators.md (GPF_HF): New.
23124         * config/aarch64/predicates.md (aarch64_reg_or_fp_float): New.
23126 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
23128         * config/aarch64/aarch64.c
23129         (aarch64_simd_container_mode): Add prototype.
23130         (aarch64_expand_mov_immediate): Add HI support.
23131         (aarch64_reinterpret_float_as_int, aarch64_float_const_rtx_p: New.
23132         (aarch64_can_const_movi_rtx_p): New.
23133         (aarch64_preferred_reload_class):
23134         Remove restrictions of using FP registers for certain SIMD operations.
23135         (aarch64_rtx_costs): Added new cost for CONST_DOUBLE moves.
23136         (aarch64_valid_floating_const): Add integer move validation.
23137         (aarch64_simd_imm_scalar_p): Remove.
23138         (aarch64_output_scalar_simd_mov_immediate): Generalize function.
23139         (aarch64_legitimate_constant_p): Expand list of supported cases.
23140         * config/aarch64/aarch64-protos.h
23141         (aarch64_float_const_rtx_p, aarch64_can_const_movi_rtx_p): New.
23142         (aarch64_reinterpret_float_as_int): New.
23143         (aarch64_simd_imm_scalar_p): Remove.
23144         * config/aarch64/constraints.md (Uvi): New.
23145         (Dd): Split into Ds and new Dd.
23146         * config/aarch64/aarch64.md (*movsi_aarch64):
23147         Add SIMD mov case.
23148         (*movdi_aarch64): Add SIMD mov case.
23150 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23152         * tree-predcom.c: (struct chain): Handle store-store chain in which
23153         stores for elimination only store loop invariant values.
23154         (execute_pred_commoning_chain): Ditto.
23155         (prepare_initializers_chain_store_elim): Ditto.
23156         (prepare_finalizers): Ditto.
23157         (is_inv_store_elimination_chain): New function.
23158         (initialize_root_vars_store_elim_1): New function.
23160 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23162         * tree-predcom.c: Revise general description of the pass.
23163         (enum chain_type): New enum type for store elimination.
23164         (struct chain): New field supporting store elimination.
23165         (struct component): Ditto.
23166         (dump_chain): Dump store-stores chain.
23167         (release_chain): Release resources.
23168         (split_data_refs_to_components): Compute and create component
23169         contains only stores for elimination.
23170         (get_chain_last_ref_at): New function.
23171         (make_invariant_chain): Initialization.
23172         (make_rooted_chain): Specify chain type in parameter and record it.
23173         (add_looparound_copies): Skip for store-stores chain.
23174         (determine_roots_comp): Compute type of chain and pass it to
23175         make_rooted_chain.
23176         (initialize_root_vars_store_elim_2): New function.
23177         (finalize_eliminated_stores): New function.
23178         (remove_stmt): Handle store for elimination.
23179         (execute_pred_commoning_chain): Execute predictive commoning on
23180         store-store chains.
23181         (determine_unroll_factor): Skip unroll for store-stores chain.
23182         (prepare_initializers_chain_store_elim): New function.
23183         (prepare_initializers_chain): Hanlde store-store chain.
23184         (prepare_finalizers_chain, prepare_finalizers): New function.
23185         (tree_predictive_commoning_loop): Return integer value indicating
23186         if loop is unrolled or lcssa form is corrupted.
23187         (tree_predictive_commoning): Rewrite for lcssa form if necessary.
23189 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23191         * tree-predcom.c (initialize_root): Delete.
23192         (execute_pred_commoning_chain): Initialize root vars and replace
23193         reference of non-combined chain directly, rather than call above
23194         function.
23196 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23198         * tree-predcom.c (ref_at_iteration): Add parameter NITERS.  Compute
23199         memory reference to DR at (NITERS + ITERS)-th iteration of loop.
23201 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23203         * tree-predcom.c (struct chain): New field init_seq.
23204         (release_chain): Release init_seq.
23205         (prepare_initializers_chain): Record intialization stmts in above
23206         field.
23207         (insert_init_seqs): New function.
23208         (tree_predictive_commoning_loop): Call insert_init_seqs.
23210 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23212         * tree-predcom.c (determine_roots_comp): Skip trivial components.
23214 2017-07-28  Richard Biener  <rguenther@suse.de>
23216         * match.pd: Remove superfluous :c.
23217         * genmatch.c (simplify::id): Add member.
23218         (lower_commutative, lower_opt_convert, lower_cond, lower_for):
23219         Copy id.
23220         (current_id): New global.
23221         (dt_node::parent): Move from ...
23222         (dt_operand::parent): ... here.  Add for_id member.
23223         (is_a_helper <dt_operand *>::test): DT_TRUE is also a dt_operand.
23224         (decision_tree::find_node): Relax order requirement when
23225         merging DT_TRUE nodes to ones inbetween the current simplify
23226         and the one we try to merge with.  Add diagnostic whenever
23227         we need to enforce pattern order by not merging.
23228         (decision_tree::insert): Set current_id.
23229         (decision_tree::print_node): Dump parent node and for_id.
23230         (parser::last_id): Add member.
23231         (parser::push_simplify): Assign unique id.
23232         (parser::parser): Initialize last_id.
23234 2017-07-28  Martin Liska  <mliska@suse.cz>
23236         PR sanitizer/81340
23237         * sanopt.c (sanitize_rewrite_addressable_params): Set VALUE_EXPR after
23238         gimple_build_debug_bind.
23240 2017-07-28  Richard Biener  <rguenther@suse.de>
23242         PR tree-optimization/81502
23243         * match.pd: Add pattern combining BIT_INSERT_EXPR with
23244         BIT_FIELD_REF.
23245         * tree-cfg.c (verify_expr): Verify types of BIT_FIELD_REF
23246         size/pos operands.
23247         (verify_gimple_assign_ternary): Likewise for BIT_INSERT_EXPR pos.
23248         * gimple-fold.c (maybe_canonicalize_mem_ref_addr): Use bitsizetype
23249         for BIT_FIELD_REF args.
23250         * fold-const.c (make_bit_field_ref): Likewise.
23251         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
23253 2017-07-28  Jakub Jelinek  <jakub@redhat.com>
23255         PR sanitizer/80998
23256         * sanopt.c (pass_sanopt::execute): Handle IFN_UBSAN_PTR.
23257         * tree-ssa-alias.c (call_may_clobber_ref_p_1): Likewise.
23258         * flag-types.h (enum sanitize_code): Add SANITIZER_POINTER_OVERFLOW.
23259         Or it into SANITIZER_UNDEFINED.
23260         * ubsan.c: Include gimple-fold.h and varasm.h.
23261         (ubsan_expand_ptr_ifn): New function.
23262         (instrument_pointer_overflow): New function.
23263         (maybe_instrument_pointer_overflow): New function.
23264         (instrument_object_size): Formatting fix.
23265         (pass_ubsan::execute): Call instrument_pointer_overflow
23266         and maybe_instrument_pointer_overflow.
23267         * internal-fn.c (expand_UBSAN_PTR): New function.
23268         * ubsan.h (ubsan_expand_ptr_ifn): Declare.
23269         * sanitizer.def (__ubsan_handle_pointer_overflow,
23270         __ubsan_handle_pointer_overflow_abort): New builtins.
23271         * tree-ssa-tail-merge.c (merge_stmts_p): Handle IFN_UBSAN_PTR.
23272         * internal-fn.def (UBSAN_PTR): New internal function.
23273         * opts.c (sanitizer_opts): Add pointer-overflow.
23274         * lto-streamer-in.c (input_function): Handle IFN_UBSAN_PTR.
23275         * fold-const.c (build_range_check): Compute pointer range check in
23276         integral type if pointer arithmetics would be needed.  Formatting
23277         fixes.
23279 2017-07-28  Martin Liska  <mliska@suse.cz>
23281         PR sanitizer/81460
23282         * sanopt.c (sanitize_rewrite_addressable_params): Do not rewrite
23283         parameters that are of a variable-length.
23285 2017-07-28  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23287         * config.gcc (powerpc-*-rtems*): Remove rs6000/eabi.h.  Add
23288         rs6000/biarch64.h.
23289         * config/rs6000/rtems.h (ASM_DECLARE_FUNCTION_SIZE): New macro.
23290         (ASM_OUTPUT_SPECIAL_POOL_ENTRY_P): Likewise.
23291         (CRT_CALL_STATIC_FUNCTION): Likewise.
23292         (ASM_DEFAULT_SPEC): New define.
23293         (ASM_SPEC32): Likewise.
23294         (ASM_SPEC64): Likewise.
23295         (ASM_SPEC_COMMON): Likewise.
23296         (ASM_SPEC): Likewise.
23297         (INVALID_64BIT): Likewise.
23298         (LINK_OS_DEFAULT_SPEC): Likewise.
23299         (LINK_OS_SPEC32): Likewise.
23300         (LINK_OS_SPEC64): Likewise.
23301         (POWERPC_LINUX): Likewise.
23302         (PTRDIFF_TYPE): Likewise.
23303         (RESTORE_FP_PREFIX): Likewise.
23304         (RESTORE_FP_SUFFIX): Likewise.
23305         (SAVE_FP_PREFIX): Likewise.
23306         (SAVE_FP_SUFFIX): Likewise.
23307         (SIZE_TYPE): Likewise.
23308         (SUBSUBTARGET_OVERRIDE_OPTIONS): Likewise.
23309         (TARGET_64BIT): Likewise.
23310         (TARGET_64BIT): Likewise.
23311         (TARGET_AIX): Likewise.
23312         (WCHAR_TYPE_SIZE): Likewise.
23313         (WCHAR_TYPE): Undefine.
23314         (TARGET_OS_CPP_BUILTINS): Add 64-bit PowerPC defines.
23315         (CPP_OS_DEFAULT_SPEC): Use previous CPP_OS_RTEMS_SPEC.
23316         (CPP_OS_RTEMS_SPEC): Delete.
23317         (SUBSUBTARGET_EXTRA_SPECS): Remove cpp_os_rtems.  Add
23318         asm_spec_common, asm_spec32, asm_spec64, link_os_spec32, and
23319         link_os_spec64.
23320         * config/rs6000/t-rtems: Add mcpu=e6500/m64 multilibs.
23322 2017-07-28  Jakub Jelinek  <jakub@redhat.com>
23324         PR tree-optimization/81578
23325         * tree-parloops.c (build_new_reduction): Bail out if
23326         reduction_code isn't one of the standard OpenMP reductions.
23327         Move the details printing after that decision.
23329 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23331         * config/rs6000/predicates.md (volatile_mem_operand): Remove code
23332         related to reload_in_progress.
23333         (splat_input_operand): Likewise.
23334         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_rtx):
23335         Delete prototype.
23336         * config/rs6000/rs6000.c (machine_function): Remove sdmode_stack_slot
23337         field.
23338         (TARGET_EXPAND_TO_RTL_HOOK): Delete.
23339         (TARGET_INSTANTIATE_DECLS): Likewise.
23340         (legitimate_indexed_address_p): Delete reload_in_progress code.
23341         (rs6000_debug_legitimate_address_p): Likewise.
23342         (rs6000_eliminate_indexed_memrefs): Likewise.
23343         (rs6000_emit_le_vsx_store): Likewise.
23344         (rs6000_emit_move_si_sf_subreg): Likewise.
23345         (rs6000_emit_move): Likewise.
23346         (register_to_reg_type): Likewise.
23347         (rs6000_pre_atomic_barrier): Likewise.
23348         (rs6000_machopic_legitimize_pic_address): Likewise.
23349         (rs6000_allocate_stack_temp): Likewise.
23350         (rs6000_address_for_fpconvert): Likewise.
23351         (rs6000_address_for_altivec): Likewise.
23352         (rs6000_secondary_memory_needed_rtx): Delete function.
23353         (rs6000_check_sdmode): Likewise.
23354         (rs6000_alloc_sdmode_stack_slot): Likewise.
23355         (rs6000_instantiate_decls): Likewise.
23356         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED_RTX): Delete.
23357         * config/rs6000/rs6000.md (splitter for *movsi_got_internal):
23358         Delete reload_in_progress.
23359         (*vec_reload_and_plus_<mptrsize>): Likewise.
23360         * config/rs6000/vsx.md (vsx_mul_v2di): Likewise.
23361         (vsx_div_v2di): Likewise.
23362         (vsx_udiv_v2di): Likewise.
23364 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23366         * config/rs6000/rs6000.opt (mlra): Replace with stub.
23367         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete OPTION_MASK_LRA.
23368         * config/rs6000/rs6000.c (TARGET_LRA_P): Delete.
23369         (rs6000_debug_reg_global): Delete print of LRA status.
23370         (rs6000_option_override_internal): Delete dead LRA related code.
23371         (rs6000_lra_p): Delete function.
23372         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mlra.
23374 2017-07-27  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23376         * config.gcc (riscv*-*-elf*): Add (riscv*-*-rtems*).
23377         * config/riscv/rtems.h: New file.
23379 2017-07-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23380             Sudakshina Das  <sudi.das@arm.com>
23382         * config/aarch64/aarch64.md
23383         (define_split for and<mode>3nr_compare): Move
23384         non aarch64_logical_operand to a register.
23385         (define_split for and_<SHIFT:optab><mode>3nr_compare0): Move non
23386         register immediate operand to a register.
23387         * config/aarch64/predicates.md (aarch64_mov_imm_operand): New.
23389 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23391         PR middle-end/81564
23392         * tree-cfg.c (group_case_labels_stmt): Handle already deleted blocks.
23394 2017-07-27  Richard Biener  <rguenther@suse.de>
23396         PR tree-optimization/81573
23397         PR tree-optimization/81494
23398         * tree-vect-loop.c (vect_create_epilog_for_reduction): Handle
23399         multi defuse cycle case.
23401 2017-07-27  Richard Biener  <rguenther@suse.de>
23403         PR tree-optimization/81571
23404         * tree-vect-slp.c (vect_build_slp_tree): Properly verify reduction
23405         PHIs.
23407 2017-07-27  Eric Botcazou  <ebotcazou@adacore.com>
23409         * config/sparc/sparc.c (sparc_option_override): Set MASK_FSMULD flag
23410         earlier and only if MASK_FPU is set.  Adjust formatting.
23412 2017-07-27  Martin Liska  <mliska@suse.cz>
23414         * opt-functions.awk: Add validation of value of Init.
23415         * optc-gen.awk: Pass new argument.
23417 2017-07-27  Martin Liska  <mliska@suse.cz>
23419         * auto-profile.c (autofdo_source_profile::update_inlined_ind_target):
23420         Fix wrong condition.
23422 2017-07-27  Martin Liska  <mliska@suse.cz>
23424         * auto-profile.c (afdo_annotate_cfg): Assign zero counts to
23425         BBs and edges seen by autoFDO.
23427 2017-07-27  Richard Biener  <rguenther@suse.de>
23429         PR tree-optimization/81502
23430         * tree-ssa.c (non_rewritable_lvalue_p): Handle BIT_INSERT_EXPR
23431         with incompatible but same sized type.
23432         (execute_update_addresses_taken): Likewise.
23434 2017-07-27  James Greenhalgh  <james.greenhalgh@arm.com>
23436         * tree-ssa-loop-ch.c (pass_ch::process_loop_p): Guard on
23437         flag_tree_loop_vectorize rather than flag_tree_vectorize.
23439 2017-07-27  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
23441         PR target/81534
23442         * config/s390/s390.md ("*atomic_compare_and_swap<mode>_1")
23443         ("*atomic_compare_and_swapdi_2", "*atomic_compare_and_swapsi_3"):
23444         Change s_operand to memory_operand.
23446 2017-07-27  Richard Sandiford  <richard.sandiford@linaro.org>
23448         * config/rs6000/rs6000-protos.h (rs6000_emit_le_vsx_permute): Declare.
23449         * config/rs6000/rs6000.c (rs6000_gen_le_vsx_permute): Replace with...
23450         (rs6000_emit_le_vsx_permute): ...this.  Take the destination as input.
23451         Emit instructions rather than returning an expression.  Handle TFmode
23452         and KFmode by casting to TImode.
23453         (rs6000_emit_le_vsx_load): Update to use rs6000_emit_le_vsx_permute.
23454         (rs6000_emit_le_vsx_store): Likewise.
23455         * config/rs6000/vsx.md (VSX_TI): New iterator.
23456         (*vsx_le_permute_<mode>): Use it instead of VSX_LE_128.
23457         (*vsx_le_undo_permute_<mode>): Likewise.
23458         (*vsx_le_perm_load_<mode>): Use rs6000_emit_le_vsx_permute to
23459         emit the split sequence.
23460         (*vsx_le_perm_store_<mode>): Likewise.
23462 2017-07-27  Jakub Jelinek  <jakub@redhat.com>
23464         PR tree-optimization/81555
23465         PR tree-optimization/81556
23466         * tree-ssa-reassoc.c (rewrite_expr_tree): Add NEXT_CHANGED argument,
23467         if true, force CHANGED for the recursive invocation.
23468         (reassociate_bb): Remember original length of ops array, pass
23469         len != orig_len as NEXT_CHANGED in rewrite_expr_tree call.
23471         * attribs.c (decl_attributes): Imply noinline, noclone and no_icf
23472         attributes for noipa attribute.  For naked attribute use
23473         lookup_attribute first before lookup_attribute_spec.
23474         * final.c (rest_of_handle_final): Disable IPA RA for functions with
23475         noipa attribute.
23476         * ipa-visibility.c (non_local_p): Fix comment typos.  Return true
23477         for functions with noipa attribute.
23478         (cgraph_externally_visible_p): Return true for functions with noipa
23479         attribute.
23480         * cgraph.c (cgraph_node::get_availability): Return AVAIL_INTERPOSABLE
23481         for functions with noipa attribute.
23482         * doc/extend.texi: Document noipa function attribute.
23483         * tree-ssa-structalias.c (refered_from_nonlocal_fn): Set *nonlocal_p
23484         also for functions with noipa attribute.
23485         (ipa_pta_execute): Set nonlocal_p also for nodes with noipa attribute.
23487 2017-07-26  Andrew Pinski  <apinski@cavium.com>
23489         * config/aarch64/aarch64.c (thunderx_vector_cost): Decrease cost of
23490         vec_unalign_load_cost and vec_unalign_store_cost.
23492 2017-07-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
23494         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Delete
23495         -mvsx-small-integer option.
23496         (ISA_3_0_MASKS_IEEE): Likewise.
23497         (OTHER_VSX_VECTOR_MASKS): Likewise.
23498         (POWERPC_MASKS): Likewise.
23499         * config/rs6000/rs6000.opt (-mvsx-small-integer): Likewise.
23500         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Simplify
23501         code, only testing for DImode being allowed in non-VSX floating
23502         point registers.
23503         (rs6000_init_hard_regno_mode_ok): Change TARGET_VSX_SMALL_INTEGER
23504         to TARGET_P8_VECTOR test.  Remove redundant VSX test inside of
23505         another VSX test.
23506         (rs6000_option_override_internal): Delete -mvsx-small-integer.
23507         (rs6000_expand_vector_set): Change TARGET_VSX_SMALL_INTEGER to
23508         TARGET_P8_VECTOR test.
23509         (rs6000_secondary_reload_simple_move): Likewise.
23510         (rs6000_preferred_reload_class): Delete TARGET_VSX_SMALL_INTEGER,
23511         since TARGET_P9_VECTOR was already tested.
23512         (rs6000_opt_masks): Remove -mvsx-small-integer.
23513         * config/rs6000/vsx.md (vsx_extract_<mode>): Delete
23514         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23515         used.
23516         (vsx_extract_<mode>_p9): Delete TARGET_VSX_SMALL_INTEGER, since a
23517         test for TARGET_VEXTRACTUB was used, and that uses
23518         TARGET_P9_VECTOR.
23519         (p9 extract splitter): Likewise.
23520         (vsx_extract_<mode>_di_p9): Likewise.
23521         (vsx_extract_<mode>_store_p9): Likewise.
23522         (vsx_extract_si): Delete TARGET_VSX_SMALL_INTEGER, since a test
23523         for TARGET_P9_VECTOR was used.  Delete code that is now dead with
23524         the elimination of TARGET_VSX_SMALL_INTEGER.
23525         (vsx_extract_<mode>_p8): Likewise.
23526         (vsx_ext_<VSX_EXTRACT_I:VS_scalar>_fl_<FL_CONV:mode>): Likewise.
23527         (vsx_ext_<VSX_EXTRACT_I:VS_scalar>_ufl_<FL_CONV:mode>): Likewise.
23528         (vsx_set_<mode>_p9): Likewise.
23529         (vsx_set_v4sf_p9): Likewise.
23530         (vsx_set_v4sf_p9_zero): Likewise.
23531         (vsx_insert_extract_v4sf_p9): Likewise.
23532         (vsx_insert_extract_v4sf_p9_2): Likewise.
23533         * config/rs6000/rs6000.md (sign extend splitter): Change
23534         TARGET_VSX_SMALL_INTEGER to TARGET_P8_VECTOR test.
23535         (floatsi<mode>2_lfiwax_mem): Likewise.
23536         (floatunssi<mode>2_lfiwzx_mem): Likewise.
23537         (float<QHI:mode><FP_ISA3:mode>2): Delete TARGET_VSX_SMALL_INTEGER,
23538         since a test for TARGET_P9_VECTOR was used.
23539         (float<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23540         (floatuns<QHI:mode><FP_ISA3:mode>2): Likewise.
23541         (floatuns<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23542         (fix_trunc<mode>si2): Change TARGET_VSX_SMALL_INTEGER to
23543         TARGET_P8_VECTOR test.
23544         (fix_trunc<mode>si2_stfiwx): Likewise.
23545         (fix_trunc<mode>si2_internal): Likewise.
23546         (fix_trunc<SFDF:mode><QHI:mode>2): Delete
23547         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23548         used.
23549         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
23550         (fixuns_trunc<mode>si2): Change TARGET_VSX_SMALL_INTEGER to
23551         TARGET_P8_VECTOR test.
23552         (fixuns_trunc<mode>si2_stfiwx): Likewise.
23553         (fixuns_trunc<SFDF:mode><QHI:mode>2): Delete
23554         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23555         used.
23556         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
23557         (fctiw<u>z_<mode>_smallint): Delete TARGET_VSX_SMALL_INTEGER,
23558         since a test for TARGET_P9_VECTOR was used.
23559         (splitter for loading small constants): Likewise.
23561 2017-07-26  Andrew Pinski  <apinski@cavium.com>
23563         * config/aarch64/aarch64.c (thunderx_vector_cost): Fix
23564         vec_fp_stmt_cost.
23566 2017-07-26  H.J. Lu  <hongjiu.lu@intel.com>
23568         PR target/81563
23569         * config/i386/i386.c (sp_valid_at): Properly check CFA offset.
23570         (fp_valid_at): Likewise.
23572 2017-07-26  James Greenhalgh  <james.greenhalgh@arm.com>
23574         * config/aarch64/aarch64.c (cortexa57_addrcost_table): Remove.
23575         (qdf24xx_addrcost_table): Likewise.
23576         (cortexa57_tunings): Update to use generic_branch_cost.
23577         (cortexa72_tunings): Likewise.
23578         (cortexa73_tunings): Likewise.
23579         (qdf24xx_tunings): Likewise.
23581 2017-07-26  James Greenhalgh  <james.greenhalgh@arm.com>
23583         * config/aarch64/aarch64.c (cortexa57_branch_cost): Remove.
23584         (thunderx2t99_branch_cost): Likewise.
23585         (cortexa35_tunings): Update to use generic_branch_cost.
23586         (cortexa53_tunings): Likewise.
23587         (cortexa57_tunings): Likewise.
23588         (cortexa72_tunings): Likewise.
23589         (cortexa73_tunings): Likewise.
23590         (thunderx2t99_tunings): Likewise.
23592 2017-07-26  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23594         * config/sparc/sparc.c (dump_target_flag_bits): Dump MASK_FSMULD.
23595         (sparc_option_override): Honour MASK_FSMULD.
23596         * config/sparc/sparc.h (MASK_FEATURES): Add MASK_FSMULD.
23597         * config/sparc/sparc.md (muldf3_extend): Use TARGET_FSMULD.
23598         * config/sparc/sparc.opt (mfsmuld): New option.
23599         * doc/invoke.texi (mfsmuld): Document option.
23601 2017-07-26  Marek Polacek  <polacek@redhat.com>
23603         PR middle-end/70992
23604         * tree.c (build2_stat): Don't set TREE_CONSTANT on divisions by zero.
23606 2017-07-26  Richard Biener  <rguenther@suse.de>
23608         * gimple-match-head.c (do_valueize): Return OP if valueize
23609         returns NULL_TREE.
23610         (get_def): New helper to get at the def stmt of a SSA name
23611         if valueize allows.
23612         * genmatch.c (dt_node::gen_kids_1): Use get_def instead of
23613         do_valueize to get at the def stmt.
23614         (dt_operand::gen_gimple_expr): Simplify do_valueize calls.
23616 2017-07-26  Wilco Dijkstra  <wdijkstr@arm.com>
23618         PR middle-end/46932
23619         * auto-inc-dec.c (parse_add_or_inc): Block autoinc on sfp.
23621 2017-07-26  Martin Liska  <mliska@suse.cz>
23623         PR sanitize/81186
23624         * function.c (expand_function_start): Make expansion of
23625         nonlocal_goto_save_area after parm_birth_insn.
23627 2017-07-26  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23629         * config/sparc/sparc.c (sparc_option_override): Remove MASK_FPU
23630         from all CPU target flags enable members.
23632 2017-07-26  Richard Biener  <rguenther@suse.de>
23634         * genmatch.c (dt_simplify::gen): Make iterator vars const.
23635         (decision_tree::gen): Make 'type' const.
23636         (write_predicate): Likewise.
23638 2017-07-24  Michael Meissner  <meissner@linux.vnet.ibm.com>
23640         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok):
23641         Eliminate TARGET_UPPER_REGS_{DF,DI,SF} usage.
23642         (rs6000_option_override_internal): Likewise.
23643         (rs6000_expand_vector_set): Likewise.
23644         * config/rs6000/rs6000.h (TARGET_UPPER_REGS_DF): Delete.
23645         (TARGET_UPPER_REGS_SF): Likewise.
23646         (TARGET_UPPER_REGS_DI): Likewise.
23647         (TARGET_VEXTRACTUB): Eliminate TARGET_UPPER_REGS_{DF,DI,SF}.
23648         (TARGET_DIRECT_MOVE_64BIT): Likewise.
23649         * config/rs6000/rs6000.md (ALTIVEC_DFORM): Likewise.
23650         (float<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23651         (Splitters for DI constants in Altivec registers): Likewise.
23652         * config/rs6000/vsx.md (vsx_set_<mode>_p9): Likewise.
23653         (vsx_set_v4sf_p9): Likewise.
23654         (vsx_set_v4sf_p9_zero): Likewise.
23655         (vsx_insert_extract_v4sf_p9): Likewise.
23656         (vsx_insert_extract_v4sf_p9_2): Likewise.
23658 2017-07-25  Carl Love  <cel@us.ibm.com>
23660         * doc/extend.texi: Update the built-in documentation file for the
23661         existing built-in functions
23662         vector signed char vec_cnttz (vector signed char);
23663         vector unsigned char vec_cnttz (vector unsigned char);
23664         vector signed short vec_cnttz (vector signed short);
23665         vector unsigned short vec_cnttz (vector unsigned short);
23666         vector signed int vec_cnttz (vector signed int);
23667         vector unsigned int vec_cnttz (vector unsigned int);
23668         vector signed long long vec_cnttz (vector signed long long);
23669         vector unsigned long long vec_cnttz (vector unsigned long long);
23671 2017-07-25  Andrew Pinski  <apinski@cavium.com>
23673         * tree-ssa-uninit.c (warn_uninitialized_vars): Don't warn about memory
23674         accesses where the use is for the first operand of a BIT_INSERT.
23676 2017-07-25  Jim Wilson  <jim.wilson@linaro.org>
23678         PR bootstrap/81521
23679         * config/i386/winnt-cxx.c (i386_pe_adjust_class_at_definition): Look
23680         for FUNCTION_DECLs in TYPE_FIELDS rather than TYPE_METHODS.
23682 2017-07-25  Jim Wilson  <jim.wilson@linaro.org>
23684         * config/i386/gstabs.h: Delete.
23685         * config/i386/openbsd.h, config/i386/t-openbsd: Likewise.
23687 2017-07-25  Uros Bizjak  <ubizjak@gmail.com>
23689         * config/i386/i386.c (ix86_decompose_address): Do not check for
23690         register RTX when looking at index_reg or base_reg.
23691         * config/i386/i386.h (INCOMING_RETURN_ADDR_RTX): Use stack_pointer_rtx.
23693 2017-07-25  Eric Botcazou  <ebotcazou@adacore.com>
23695         * gimple.c (gimple_assign_set_rhs_with_ops): Do not ask gsi_replace
23696         to update EH info here.
23698 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23700         * match.pd ((X * CST1) * CST2): Simplify to X * (CST1 * CST2).
23702 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23704         * match.pd ((X * CST) * Y): Reassociate to (X * Y) * CST.
23706 2017-07-25  Torsten Duwe  <duwe@suse.de>
23708         * common.opt: Introduce -fpatchable-function-entry
23709         command line option, and its variables function_entry_patch_area_size
23710         and function_entry_patch_area_start.
23711         * opts.c (common_handle_option): Add -fpatchable_function_entry_ case,
23712         including a two-value parser.
23713         * target.def (print_patchable_function_entry): New target hook.
23714         * targhooks.h (default_print_patchable_function_entry): New function.
23715         * targhooks.c (default_print_patchable_function_entry): Likewise.
23716         * toplev.c (process_options): Switch off IPA-RA if
23717         patchable function entries are being generated.
23718         * varasm.c (assemble_start_function): Look at the
23719         patchable-function-entry command line switch and current
23720         function attributes and maybe generate NOP instructions by
23721         calling the print_patchable_function_entry hook.
23722         * doc/extend.texi: Document patchable_function_entry attribute.
23723         * doc/invoke.texi: Document -fpatchable_function_entry
23724         command line option.
23725         * doc/tm.texi.in (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY):
23726         New target hook.
23727         * doc/tm.texi: Re-generate.
23729 2017-07-25  Jakub Jelinek  <jakub@redhat.com>
23731         PR target/81532
23732         * config/i386/constraints.md (Yd, Ye): Use ALL_SSE_REGS for
23733         TARGET_AVX512DQ rather than TARGET_AVX512BW.
23735 2017-07-25  Tamar Christina  <tamar.christina@arm.com>
23737         * config/arm/parsecpu.awk (all_cores): Remove duplicates.
23739 2017-07-25  Richard Biener  <rguenther@suse.de>
23741         PR tree-optimization/81455
23742         * tree-ssa-loop-unswitch.c (find_loop_guard): Make sure to
23743         not walk in cycles when looking for guards.
23745 2017-07-25  Richard Biener  <rguenther@suse.de>
23747         PR tree-optimization/81529
23748         * tree-vect-stmts.c (process_use): Disregard live induction PHIs
23749         when optimizing backedge uses.
23751 2017-07-25  David Edelsohn  <dje.gcc@gmail.com>
23753         * dwarf2asm.c (dw2_asm_output_nstring): Encode double quote
23754         character for AIX.
23755         * dwarf2out.c (output_macinfo): Copy debug_line_section_label
23756         to dl_section_ref.  On AIX, append an expression to subtract
23757         the size of the section length to dl_section_ref.
23759 2017-07-25  Segher Boessenkool  <segher@kernel.crashing.org>
23761         * configure.ac: If any of the config.* scripts fail, exit 1.
23762         * configure: Regenerate.
23764 2017-07-25  Richard Biener  <rguenther@suse.de>
23766         PR middle-end/81546
23767         * tree-ssa-operands.c (verify_imm_links): Remove cap on number
23768         of immediate uses, be more verbose on errors.
23770 2017-07-25  Richard Biener  <rguenther@suse.de>
23772         PR tree-optimization/81510
23773         * tree-vect-loop.c (vect_is_simple_reduction): When the
23774         reduction stmt is not inside the loop bail out.
23776 2017-07-25  Richard Biener  <rguenther@suse.de>
23778         PR tree-optimization/81303
23779         * tree-vect-loop-manip.c (vect_loop_versioning): Build
23780         profitability check against LOOP_VINFO_NITERSM1.
23782 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23784         * domwalk.c (cmp_bb_postorder): Simplify.
23785         (sort_bbs_postorder): New function.  Use it...
23786         (dom_walker::walk): ...here to optimize common cases.
23788 2017-07-25  Martin Liska  <mliska@suse.cz>
23790         PR ipa/81520
23791         * ipa-visibility.c (function_and_variable_visibility): Make the
23792         redirection just on target that supports aliasing.
23793         Fix GNU coding style.
23795 2017-07-25  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23797         PR libgcc/61152
23798         * config/aarch64/rtems.h: Add GCC Runtime Library Exception.
23799         Format changes.
23800         * config/arm/rtems.h: Likewise.
23801         * config/bfin/rtems.h: Likewise.
23802         * config/i386/rtemself.h: Likewise.
23803         * config/lm32/rtems.h: Likewise.
23804         * config/m32c/rtems.h: Likewise.
23805         * config/m68k/rtemself.h: Likewise.
23806         * config/microblaze/rtems.h: Likewise.
23807         * config/mips/rtems.h: Likewise.
23808         * config/moxie/rtems.h: Likewise.
23809         * config/nios2/rtems.h: Likewise.
23810         * config/powerpcspe/rtems.h: Likewise.
23811         * config/rs6000/rtems.h: Likewise.
23812         * config/rtems.h: Likewise.
23813         * config/sh/rtems.h: Likewise.
23814         * config/sh/rtemself.h: Likewise.
23815         * config/sparc/rtemself.h: Likewise.
23817 2017-07-25  Georg-Johann Lay  <avr@gjlay.de>
23819         PR 81487
23820         * hsa-brig.c (brig_init): Use xasprintf instead of asprintf.
23821         * gimple-pretty-print.c (dump_profile, dump_probability): Same.
23822         * tree-ssa-structalias.c (alias_get_name): Same.
23824 2017-07-25  Bin Cheng  <bin.cheng@arm.com>
23826         PR target/81414
23827         * config/aarch64/cortex-a57-fma-steering.c (analyze): Skip fmul/fmac
23828         instructions if no du chain is found.
23830 2017-07-25  Georg-Johann Lay  <avr@gjlay.de>
23832         * config/avr/avr-log.c (avr_log_vadump) ['T']: Print NULL-TREE.
23834 2017-07-25  Richard Biener  <rguenther@suse.de>
23836         PR middle-end/81505
23837         * fold-const.c (fold_negate_const): TREE_OVERFLOW should be
23838         sticky.
23840 2017-07-24  Michael Meissner  <meissner@linux.vnet.ibm.com>
23842         * config/rs6000/rs6000-cpus.def (ISA_2_6_MASKS_SERVER): Delete
23843         upper-regs options.
23844         (ISA_2_7_MASKS_SERVER): Likewise.
23845         (ISA_3_0_MASKS_IEEE): Likewise.
23846         (OTHER_P8_VECTOR_MASKS): Likewise.
23847         (OTHER_VSX_VECTOR_MASKS): Likewise.
23848         (POWERPC_MASKS): Likewise.
23849         (power7 cpu): Use ISA_2_6_MASKS_SERVER instead of using a
23850         duplicate list of options.
23851         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove
23852         explicit -mupper-regs options.
23853         * config/rs6000/rs6000.opt (-mvsx-scalar-memory): Delete
23854         -mupper-regs* options.  Delete -mvsx-scalar-memory, which was an
23855         alias for -mupper-regs-df.
23856         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Likewise.
23857         (rs6000_init_hard_regno_mode_ok): Likewise.
23858         (rs6000_option_override_internal): Likewise.
23859         (rs6000_opt_masks): Likewise.
23860         * config/rs6000/rs6000.h (TARGET_UPPER_REGS_DF): Define upper regs
23861         options in terms of whether -mvsx or -mpower8-vector was used.
23862         (TARGET_UPPER_REGS_DI): Likewise.
23863         (TARGET_UPPER_REGS_SF): Likewise.
23864         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete the
23865         -mupper-regs-* options.
23867 2017-07-24  Segher Boessenkool  <segher@kernel.crashing.org>
23869         * passes.c (emergency_dump_function): Print some empty lines and a
23870         header before the RTL dump.
23872 2017-07-24  Segher Boessenkool  <segher@kernel.crashing.org>
23874         * cfgrtl.c (rtl_dump_bb): Don't call NEXT_INSN on NULL.
23876 2017-07-24  Wilco Dijkstra  <wdijkstr@arm.com>
23878         PR target/79041
23879         * config/aarch64/aarch64.c (aarch64_classify_symbol):
23880         Avoid SYMBOL_SMALL_ABSOLUTE for literals with pc-relative literals.
23882 2017-07-24  Carl Love  <cel@us.ibm.com>
23884         * config/rs6000/rs6000-c.c: Add support for built-in functions
23885         vector float vec_extract_fp32_from_shorth (vector unsigned short);
23886         vector float vec_extract_fp32_from_shortl (vector unsigned short);
23887         * config/rs6000/altivec.h (vec_extract_fp_from_shorth,
23888         vec_extract_fp_from_shortl): Add defines for the two builtins.
23889         * config/rs6000/rs6000-builtin.def (VEXTRACT_FP_FROM_SHORTH,
23890         VEXTRACT_FP_FROM_SHORTL): Add BU_P9V_OVERLOAD_1 and BU_P9V_VSX_1
23891         new builtins.
23892         * config/rs6000/vsx.md vsx_xvcvhpsp): Add define_insn.
23893         (vextract_fp_from_shorth, vextract_fp_from_shortl): Add define_expands.
23894         * doc/extend.texi: Update the built-in documentation file for the
23895         new built-in function.
23897 2017-07-24  Jakub Jelinek  <jakub@redhat.com>
23899         PR bootstrap/81521
23900         * tree.def: Remove TYPE_METHODS documentation, adjust TYPE_FIELDS
23901         documentation.
23902         * doc/generic.texi: Likewise.
23903         * config/i386/winnt-cxx.c (i386_pe_adjust_class_at_definition): Look
23904         for FUNCTION_DECLs in TYPE_FIELDS rather than TYPE_METHODS.
23906 2017-07-24  Jackson Woodruff  <jackson.woodruff@arm.com>
23908         * config/aarch64/aarch64-simd.md (aarch64_mla_elt_merge<mode>): New.
23909         (aarch64_mls_elt_merge<mode>): Likewise.
23911 2017-07-23  Krister Walfridsson  <krister.walfridsson@gmail.com>
23913         * config.gcc (*-*-netbsd*): Remove check for NetBSD versions not
23914         having __cxa_atexit.
23916 2017-07-23  Michael Collison  <michael.collison@arm.com>
23918         * config/arm/arm.c (arm_option_override): Deprecate
23919         use of -mstructure-size-boundary.
23920         * config/arm/arm.opt: Deprecate -mstructure-size-boundary.
23921         * doc/invoke.texi: Deprecate -mstructure-size-boundary.
23923 2017-07-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23925         PR target/80695
23926         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost):
23927         Reduce cost estimate for direct moves.
23929 2017-07-23  Uros Bizjak  <ubizjak@gmail.com>
23931         PR target/80569
23932         * config/i386/i386.c (ix86_option_override_internal): Disable
23933         BMI, BMI2 and TBM instructions for -m16.
23935 2017-07-21  Carl Love  <cel@us.ibm.com>
23937         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
23938         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
23939         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
23940         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
23941         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
23942         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
23943         VMULOSW): New enum "unspec" values.
23944         (altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
23945         altivec_vmulosw): New patterns.
23946         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
23947         VMULOSW): Add definitions.
23949 2017-07-21  Jim Wilson  <jim.wilson@linaro.org>
23951         * config/aarch64/aarch64-cores.def (falkor): Add AARCH64_FL_RDMA.
23952         (qdf24xx): Likewise.
23953         * config/aarch64/aarch64-options-extensions.def (rdma); New.
23954         * config/aarch64/aarch64.h (AARCH64_FL_RDMA): New.
23955         (AARCH64_FL_V8_1): Renumber.
23956         (AARCH64_FL_FOR_ARCH8_1): Add AARCH64_FL_RDMA.
23957         (AARCH64_ISA_RDMA): Use AARCH64_FL_RDMA.
23958         * config/aarch64/arm_neon.h: Use +rdma instead of arch=armv8.1-a.
23959         * doc/invoke.texi (AArch64 Options): Mention +rmda in -march docs.  Add
23960         rdma to feature modifiers list.
23962 2017-07-21  Yury Gribov  <tetra2005@gmail.com>
23964         PR middle-end/56727
23965         * ipa-visibility (function_and_variable_visibility): Convert
23966         recursive PLT call to direct call if appropriate.
23968 2017-07-21  Andrew Pinski  <apinski@cavium.com>
23970         * tree-ssa-sccvn.c (vn_nary_op_eq): Check BIT_INSERT_EXPR's
23971         operand 1 to see if the types precision matches.
23972         * fold-const.c (operand_equal_p): Likewise.
23974 2017-07-21  Richard Biener  <rguenther@suse.de>
23976         PR tree-optimization/81303
23977         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs): Pass
23978         in datarefs vector.  Allow NULL dr0 for no peeling cost estimate.
23979         (vect_peeling_hash_get_lowest_cost): Adjust.
23980         (vect_enhance_data_refs_alignment): Likewise.  Use
23981         vect_get_peeling_costs_all_drs to compute the penalty for no
23982         peeling to match up costs.
23984 2017-07-21  Richard Biener  <rguenther@suse.de>
23986         PR tree-optimization/81500
23987         * tree-vect-loop.c (vect_is_simple_reduction): Properly fail if
23988         we didn't identify a reduction path.
23990 2017-07-21  Tom de Vries  <tom@codesourcery.com>
23991             Cesar Philippidis  <cesar@codesourcery.com>
23993         PR gcov-profile/81442
23994         * config/nvptx/nvptx.c (nvptx_goacc_reduction_init): Add missing edge
23995         probabilities.
23997 2017-07-21  Tom de Vries  <tom@codesourcery.com>
23999         PR lto/81430
24000         * config/nvptx/nvptx.c (nvptx_override_options_after_change): New
24001         function.
24002         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define to
24003         nvptx_override_options_after_change.
24005 2017-07-21  Ulrich Drepper  <drepper@redhat.com>
24007         * dwarf2out.c (output_file_names): Avoid double testing for
24008         dwarf_version >= 5.
24010 2017-07-21  Georg-Johann Lay  <avr@gjlay.de>
24012         * doc/invoke.texi (AVR Built-in Functions): Re-layout section.
24014 2016-07-21  Jan Hubicka  <hubicka@ucw.cz>
24016         * cfgcleanup.c (flow_find_cross_jump): Do not crossjump across
24017         hot/cold regions.
24018         (try_crossjump_to_edge): Do not punt on partitioned functions.
24020 2016-07-21  Jan Hubicka  <hubicka@ucw.cz>
24022         * bb-reorder.c (find_rarely_executed_basic_blocks_and_crossing_edges):
24023         Put all BBs reachable only via paths crossing cold region to cold
24024         region.
24025         * cfgrtl.c (find_bbs_reachable_by_hot_paths): New function.
24027 2016-07-21  Richard Biener  <rguenther@suse.de>
24029         PR tree-optimization/81303
24030         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Take
24031         into account prologue and epilogue iterations when raising
24032         min_profitable_iters to sth at least covering one vector iteration.
24034 2017-07-21  Tamar Christina  <tamar.christina@arm.com>
24036         * config/arm/arm.c (arm_test_cpu_arch_dat):
24037         Check for overlap.
24039 2017-07-20  Nathan Sidwell  <nathan@acm.org>
24041         Remove TYPE_METHODS.
24042         * tree.h (TYPE_METHODS): Delete.
24043         * dwarf2out.c (gen_member_die): Member fns are on TYPE_FIELDS.
24044         * dbxout.c (dbxout_type_fields): Ignore FUNCTION_DECLs.
24045         (dbxout_type_methods): Scan TYPE_FIELDS.
24046         (dbxout_type): Don't check TYPE_METHODS here.
24047         * function.c (use_register_for_decl): Always ignore register for
24048         class types when not optimizing.
24049         * ipa-devirt.c (odr_types_equivalent_p): Delete TYPE_METHODS scan.
24050         * tree.c (free_lang_data_in_type): Stitch out member functions and
24051         templates from TYPE_FIELDS.
24052         (build_distinct_type_copy, verify_type_variant,
24053         verify_type): Member fns are on TYPE_FIELDS.
24054         * tree-dump.c (dequeue_and_dump): No TYPE_METHODS.
24055         * tree-pretty-print.c (dump_generic_node): Likewise.
24057 2017-07-20  Jakub Jelinek  <jakub@redhat.com>
24059         PR target/80846
24060         * config/i386/i386.c (ix86_expand_vector_init_general): Handle
24061         V2TImode and V4TImode.
24062         (ix86_expand_vector_extract): Likewise.
24063         * config/i386/sse.md (VMOVE): Enable V4TImode even for just
24064         TARGET_AVX512F, instead of only for TARGET_AVX512BW.
24065         (ssescalarmode): Handle V4TImode and V2TImode.
24066         (VEC_EXTRACT_MODE): Add V4TImode and V2TImode.
24067         (*vec_extractv2ti, *vec_extractv4ti): New insns.
24068         (VEXTRACTI128_MODE): New mode iterator.
24069         (splitter for *vec_extractv?ti first element): New.
24070         (VEC_INIT_MODE): New mode iterator.
24071         (vec_init<mode>): Consolidate 3 expanders into one using
24072         VEC_INIT_MODE mode iterator.
24074 2017-07-20  Alexander Monakov  <amonakov@ispras.ru>
24076         * lra-assigns.c (pseudo_compare_func): Fix comparison step based on
24077         non_spilled_static_chain_regno_p.
24079 2017-07-20  Alexander Monakov  <amonakov@ispras.ru>
24081         * gimple-ssa-store-merging.c (sort_by_bitpos): Return 0 on equal bitpos.
24083 2017-07-20  Jan Hubicka  <hubicka@ucw.cz>
24085         * bb-reorder.c (connect_traces): Allow copying of blocks within
24086         single partition.
24088 2017-07-20  Richard Biener  <rguenther@suse.de>
24090         * gimple.h (gimple_phi_result): Add gphi * overload.
24091         (gimple_phi_result_ptr): Likewise.
24092         (gimple_phi_arg): Likewise.  Adjust index assert to only
24093         allow actual argument accesses rather than all slots available
24094         by capacity.
24095         (gimple_phi_arg_def): Add gphi * overload.
24096         * tree-phinodes.c (make_phi_node): Initialize only actual
24097         arguments.
24098         (resize_phi_node): Clear memory not covered by old node,
24099         do not initialize excess argument slots.
24100         (reserve_phi_args_for_new_edge): Initialize new argument slot
24101         completely.
24103 2017-07-20  Bin Cheng  <bin.cheng@arm.com>
24105         PR tree-optimization/81388
24106         Revert r238585:
24107         2016-07-21  Bin Cheng  <bin.cheng@arm.com>
24109         * tree-ssa-loop-niter.c (number_of_iterations_lt_to_ne): Clean up
24110         by removing computation of may_be_zero.
24112 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
24113             Tom de Vries  <tom@codesourcery.com>
24115         PR middle-end/81030
24116         * cfgbuild.c (find_many_sub_basic_blocks): Update REG_BR_PROB note
24117         when gimple level profile disagrees with what RTL expander did.
24119 2017-07-20  Richard Biener  <rguenther@suse.de>
24121         PR tree-optimization/61171
24122         * tree-vectorizer.h (slp_instance): Add reduc_phis member.
24123         (vect_analyze_stmt): Add slp instance parameter.
24124         (vectorizable_reduction): Likewise.
24125         * tree-vect-loop.c (vect_analyze_loop_operations): Adjust.
24126         (vect_is_simple_reduction): Deal with chains not detected
24127         as SLP reduction chain, specifically not properly associated
24128         chains containing a mix of plus/minus.
24129         (get_reduction_op): Remove.
24130         (get_initial_defs_for_reduction): Simplify, pass in whether
24131         this is a reduction chain, pass in the SLP node for the PHIs.
24132         (vect_create_epilog_for_reduction): Get the SLP instance as
24133         arg and adjust.
24134         (vectorizable_reduction): Get the SLP instance as arg.
24135         During analysis remember the SLP node with the PHIs in the
24136         instance.  Simplify getting at the vectorized reduction PHIs.
24137         * tree-vect-slp.c (vect_slp_analyze_node_operations): Pass
24138         through SLP instance.
24139         (vect_slp_analyze_operations): Likewise.
24140         * tree-vect-stms.c (vect_analyze_stmt): Likewise.
24141         (vect_transform_stmt): Likewise.
24143 2017-07-20  Tom de Vries  <tom@codesourcery.com>
24145         PR tree-optimization/81489
24146         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior): Move
24147         read of phi arg location to before loop that modifies phi.
24149 2017-07-20  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
24151         * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0):
24152         New pattern.
24154 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
24156         PR middle-end/81331
24157         * except.c (execute): Fix ordering issue.
24159 2018-07-19  Segher Boessenkool  <segher@kernel.crashing.org>
24161         PR rtl-optimization/81423
24162         * combine.c (make_compound_operation_int): Don't try to optimize
24163         the AND of a SUBREG of an LSHIFTRT if that SUBREG is paradoxical.
24165 2017-07-19  Segher Boessenkool  <segher@kernel.crashing.org>
24167         PR rtl-optimization/81423
24168         * simplify-rtx.c (simplify_truncation): Handle truncating an IOR
24169         with a constant that is -1 in the truncated to mode.
24171 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
24173         * predict.c (propagate_unlikely_bbs_forward): Break out from ...
24174         (determine_unlikely_bbs): ... here.
24175         * predict.h (propagate_unlikely_bbs_forward): Declare.
24176         * cfgexpand.c (pass_expand::execute): Use it.
24177         * bb-reorder.c (sanitize_hot_paths): Do not consider known to be
24178         unlikely edges.
24179         (find_rarely_executed_basic_blocks_and_crossing_edges): Use
24180         propagate_unlikely_bbs_forward.
24182 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
24184         PR middle-end/81331
24185         * except.c (maybe_add_nop_after_section_switch): New function.
24186         (execute): Use it.
24188 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24190         * gimple.h (gimple_phi_set_arg): Make assert more strict.
24192 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24194         * gimple.h (gimple_phi_arg): Make assert more strict.
24196 2017-07-19  Steven Munroe  <munroesj@gcc.gnu.org>
24198         * config.gcc (powerpc*-*-*): Add mmintrin.h.
24199         * config/rs6000/mmintrin.h: New file.
24200         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include mmintrin.h.
24202 2017-07-19  Jakub Jelinek  <jakub@redhat.com>
24204         PR tree-optimization/81346
24205         * match.pd: Optimize (X - 1U) <= INT_MAX-1U into (int) X > 0.
24207 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24209         * config/nvptx/nvptx.md (VECIM): Add V2DI.
24211 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24213         * config/nvptx/nvptx-modes.def: Add V2DImode.
24214         * config/nvptx/nvptx-protos.h (nvptx_data_alignment): Declare.
24215         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Handle V2DImode.
24216         (nvptx_output_mov_insn): Handle lack of mov.b128.
24217         (nvptx_print_operand): Handle 'H' and 'L' codes.
24218         (nvptx_vector_mode_supported): Allow V2DImode.
24219         (nvptx_preferred_simd_mode): New function.
24220         (nvptx_data_alignment): New function.
24221         (TARGET_VECTORIZE_PREFERRED_SIMD_MODE): Redefine to
24222         nvptx_preferred_simd_mode.
24223         * config/nvptx/nvptx.h (STACK_BOUNDARY, BIGGEST_ALIGNMENT): Change from
24224         64 to 128 bits.
24225         (DATA_ALIGNMENT): Define.  Set to nvptx_data_alignment.
24227 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24229         * config/nvptx/nvptx-modes.def: New file.  Add V2SImode.
24230         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Handle V2SImode.
24231         (nvptx_vector_mode_supported): New function.  Allow V2SImode.
24232         (TARGET_VECTOR_MODE_SUPPORTED_P): Redefine to nvptx_vector_mode_supported.
24233         * config/nvptx/nvptx.md (VECIM): New mode iterator. Add V2SI.
24234         (mov<VECIM>_insn): New define_insn.
24235         (define_expand "mov<VECIM>): New define_expand.
24237 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24239         * config/nvptx/nvptx.c (nvptx_print_operand): Handle v2 vector mode.
24241 2017-07-19  Jakub Jelinek  <jakub@redhat.com>
24243         PR tree-optimization/81346
24244         * fold-const.h (fold_div_compare, range_check_type): Declare.
24245         * fold-const.c (range_check_type): New function.
24246         (build_range_check): Use range_check_type.
24247         (fold_div_compare): No longer static, rewritten into
24248         a match.pd helper function.
24249         (fold_comparison): Don't call fold_div_compare here.
24250         * match.pd (X / C1 op C2): New optimization using fold_div_compare
24251         as helper function.
24253 2017-07-19  Nathan Sidwell  <nathan@acm.org>
24255         * tree.h (TYPE_MINVAL, TYPE_MAXVAL): Rename to ...
24256         (TYPE_MIN_VALUE_RAW, TYPE_MAX_VALUE_RAW): ... these.
24257         * tree.c (find_decls_types_r, verify_type): Use
24258         TYPE_{MIN,MAX}_VALUE_RAW.
24259         * lto-streamer-out.c (DFS::DFS_write_tree_body): Likewise.
24260         (hash_tree): Likewise.
24261         * tree-streamer-in.c (lto_input_ts_type_non_common_tree_pointers):
24262         Likewise.
24263         * tree-streamer-out.c (write_ts_type_non_common_tree_pointers):
24264         Likewise.
24266 2017-07-18  Tom de Vries  <tom@codesourcery.com>
24268         PR middle-end/81464
24269         * omp-expand.c (expand_omp_for_static_chunk): Handle
24270         equal-argument loop exit phi.
24272 2017-07-18  Uros Bizjak  <ubizjak@gmail.com>
24274         PR target/81471
24275         * config/i386/i386.md (rorx_immediate_operand): New mode attribute.
24276         (*bmi2_rorx<mode>3_1): Use rorx_immediate_operand as
24277         operand 2 predicate.
24278         (*bmi2_rorxsi3_1_zext): Use const_0_to_31_operand as
24279         operand 2 predicate.
24280         (ror,rol -> rorx splitters): Use const_int_operand as
24281         operand 2 predicate.
24283 2017-06-18  Richard Biener  <rguenther@suse.de>
24285         PR tree-optimization/81410
24286         * tree-vect-stmts.c (vectorizable_load): Properly adjust for
24287         the gap in the ! slp_perm SLP case after each group.
24289 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
24291         PR middle-end/81463
24292         * cfgloopmanip.c (scale_loop_profile): Watch out for zero frequency
24293         again.
24295 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
24297         PR middle-end/81462
24298         * predict.c (set_even_probabilities): Cleanup; do not affect
24299         probabilities that are already known.
24300         (combine_predictions_for_bb): Call even when count is set.
24302 2017-07-18  Nathan Sidwell  <nathan@acm.org>
24304         * tree-parloops.c (try_transform_to_exit_first_loop_alt): Use
24305         TYPE_MAX_VALUE.
24307 2017-07-18  Bin Cheng  <bin.cheng@arm.com>
24309         PR target/81408
24310         * tree-ssa-loop-niter.c (number_of_iterations_exit): Dump missed
24311         optimization for loop niter analysis.
24313 2017-07-18  Georg-Johann Lay  <avr@gjlay.de>
24315         PR target/81473
24316         * config/avr/avr.c (avr_optimize_casesi): Don't use
24317         INT8_MIN, INT8_MAX, UINT8_MAX, INT16_MIN, INT16_MAX, UINT16_MAX.
24319 2017-07-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
24321         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Remove
24322         body_cost_vec from _vect_peel_extended_info.
24323         (vect_peeling_hash_get_lowest_cost): Do not set body_cost_vec.
24324         (vect_peeling_hash_choose_best_peeling): Remove body_cost_vec and
24325         npeel.
24327 2017-07-18  Bin Cheng  <bin.cheng@arm.com>
24329         * config/arm/arm.c (emit_unlikely_jump): Remove unused var.
24331 2017-07-18  Richard Biener  <rguenther@suse.de>
24333         PR tree-optimization/80620
24334         PR tree-optimization/81403
24335         * tree-ssa-pre.c (phi_translate_1): Clear range and points-to
24336         info when re-using a VN table entry.
24338 2017-07-18  Richard Biener  <rguenther@suse.de>
24340         PR tree-optimization/81418
24341         * tree-vect-loop.c (vectorizable_reduction): Properly compute
24342         vectype_in.  Verify that with lane-reducing reduction operations
24343         we have a single def-use cycle.
24345 2017-07-17  Carl Love  <cel@us.ibm.com>
24347         Revert commit r249424   2017-06-20  Carl Love  <cel@us.ibm.com>
24349         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
24350         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
24351         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
24352         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
24353         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
24354         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
24355         VMULOSW): New enum "unspec" values.
24356         (vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
24357         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si,
24358         altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
24359         altivec_vmulosw): New patterns.
24360         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
24361         VMULOSW): Add definitions.
24363 2017-07-17  Uros Bizjak  <ubizjak@gmail.com>
24365         * config/alpha/alpha.c: Include predict.h.
24367 2017-07-17  Yury Gribov  <tetra2005@gmail.com>
24369         * tree-vrp.c (compare_assert_loc): Fix comparison function
24370         to return predictable results.
24372 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24374         * config/arc/arc.md (adddi3): Remove support for mexpand-adddi
24375         option.
24376         (subdi3): Likewise.
24377         * config/arc/arc.opt (mexpand-adddi): Deprecate it.
24378         * doc/invoke.texi (mexpand-adddi): Update text.
24380 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24382         * config/arc/arc.md (clzsi2): Expand to an arc_clzsi2 instruction
24383         that also clobbers the CC register. The old expand code is moved
24384         to ...
24385         (*arc_clzsi2): ... here.
24386         (ctzsi2): Expand to an arc_ctzsi2 instruction that also clobbers
24387         the CC register. The old expand code is moved to ...
24388         (arc_ctzsi2): ... here.
24390 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24392         * config/arc/arc.opt (mindexed-loads): Use initial value
24393         TARGET_INDEXED_LOADS_DEFAULT.
24394         (mauto-modify-reg): Use initial value
24395         TARGET_AUTO_MODIFY_REG_DEFAULT.
24396         * config/arc/elf.h (TARGET_INDEXED_LOADS_DEFAULT): Define.
24397         (TARGET_AUTO_MODIFY_REG_DEFAULT): Likewise.
24398         * config/arc/linux.h (TARGET_INDEXED_LOADS_DEFAULT): Define.
24399         (TARGET_AUTO_MODIFY_REG_DEFAULT): Likewise.
24401 2017-07-17  Martin Liska  <mliska@suse.cz>
24403         PR sanitizer/81302
24404         * opts.c (finish_options): Do not allow -fgnu-tm
24405         w/ -fsanitize={kernel-,}address.  Say sorry.
24407 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24409         PR target/81369
24410         * tree-loop-distribution.c (classify_partition): Only assert on
24411         numer of iterations.
24412         (merge_dep_scc_partitions): Delete prameter.  Update function call.
24413         (distribute_loop): Remove code handling loop with unknown niters.
24414         (pass_loop_distribution::execute): Skip loop with unknown niters.
24416 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24418         PR target/81369
24419         * tree-loop-distribution.c (merge_dep_scc_partitions): Sink call to
24420         function sort_partitions_by_post_order.
24422 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24424         PR tree-optimization/81374
24425         * tree-loop-distribution.c (pass_loop_distribution::execute): Record
24426         the max index of basic blocks, rather than number of basic blocks.
24428 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24430         * config/arc/arc-protos.h (arc_legitimate_pc_offset_p): Remove
24431         proto.
24432         (arc_legitimate_pic_operand_p): Likewise.
24433         * config/arc/arc.c (arc_legitimate_pic_operand_p): Remove
24434         function.
24435         (arc_needs_pcl_p): Likewise.
24436         (arc_legitimate_pc_offset_p): Likewise.
24437         (arc_legitimate_pic_addr_p): Remove LABEL_REF case, as this
24438         function is also used in constrains.md.
24439         (arc_legitimate_constant_p): Use arc_legitimate_pic_addr_p to
24440         validate pic constants. Handle CONST_INT, CONST_DOUBLE, MINUS and
24441         PLUS.  Only return true/false in known cases, otherwise assert.
24442         (arc_legitimate_address_p): Remove arc_legitimate_pic_addr_p as it
24443         is already called in arc_legitimate_constant_p.
24444         * config/arc/arc.h (CONSTANT_ADDRESS_P): Consider also LABEL for
24445         pic addresses.
24446         (LEGITIMATE_PIC_OPERAND_P): Use
24447         arc_raw_symbolic_reference_mentioned_p function.
24448         * config/arc/constraints.md (Cpc): Use arc_legitimate_pic_addr_p
24449         function.
24450         (Cal): Likewise.
24451         (C32): Likewise.
24453 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24454         Andrew Burgess  <andrew.burgess@embecosm.com>
24456         * config/arc/arc-protos.h (arc_compute_function_type): Change prototype.
24457         (arc_return_address_register): New function.
24458         * config/arc/arc.c (arc_handle_fndecl_attribute): New function.
24459         (arc_handle_fndecl_attribute): Add naked attribute.
24460         (TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS): Define.
24461         (TARGET_WARN_FUNC_RETURN): Likewise.
24462         (arc_allocate_stack_slots_for_args): New function.
24463         (arc_warn_func_return): Likewise.
24464         (machine_function): Change type fn_type.
24465         (arc_compute_function_type): Consider new naked function type,
24466         change function return type.
24467         (arc_must_save_register): Adapt to handle new
24468         arc_compute_function_type's return type.
24469         (arc_expand_prologue): Likewise.
24470         (arc_expand_epilogue): Likewise.
24471         (arc_return_address_regs): Delete.
24472         (arc_return_address_register): New function.
24473         (arc_epilogue_uses): Use above function.
24474         * config/arc/arc.h (arc_return_address_regs): Delete prototype.
24475         (arc_function_type): Change encoding, add naked type.
24476         (ARC_INTERRUPT_P): Change to handle the new encoding.
24477         (ARC_FAST_INTERRUPT_P): Likewise.
24478         (ARC_NORMAL_P): Define.
24479         (ARC_NAKED_P): Likewise.
24480         (arc_compute_function_type): Delete prototype.
24481         * config/arc/arc.md (in_ret_delay_slot): Use
24482         arc_return_address_register function.
24483         (simple_return): Likewise.
24484         (p_return_i): Likewise.
24486 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24488         PR tree-optimization/81428
24489         * match.pd (X / X -> one): Don't optimize _Fract divisions, as 1
24490         can't be built for those types.
24492 2017-07-17  Georg-Johann Lay  <avr@gjlay.de>
24494         Remove stuff dead since r239246.
24496         * config/avr/avr-arch.h (avr_inform_devices): Remove dead proto.
24497         * config/avr/avr-devices.c (mcu_name, comparator, avr_mcus_str)
24498         (avr_inform_devices): Remove dead stuff.
24500 2017-07-17  Tamar Christina  <tamar.christina@arm.com>
24502         * config/arm/arm_neon.h: Fix softp typo.
24504 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24506         PR tree-optimization/81365
24507         * tree-ssa-phiprop.c (propagate_with_phi): When considering hoisting
24508         aggregate moves onto bb predecessor edges, make sure there are no
24509         loads that could alias the lhs in between the start of bb and the
24510         loads from *phi.
24512 2017-07-17  Georg-Johann Lay  <avr@gjlay.de>
24514         PR 80929
24515         * config/avr/avr.c (avr_mul_highpart_cost): New static function.
24516         (avr_rtx_costs_1) [TRUNCATE]: Use it to compute mul_highpart cost.
24517         [LSHIFTRT, outer_code = TRUNCATE]: Same.
24519 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24521         PR tree-optimization/81396
24522         * tree-ssa-math-opts.c (struct symbolic_number): Add n_ops field.
24523         (init_symbolic_number): Initialize it to 1.
24524         (perform_symbolic_merge): Add n_ops from both operands into the new
24525         n_ops.
24526         (find_bswap_or_nop): Don't consider n->n == cmpnop computations
24527         without base_addr as useless if they need more than one operation.
24528         (bswap_replace): Handle !bswap case for NULL base_addr.
24530 2017-07-17  Tom de Vries  <tom@codesourcery.com>
24532         PR target/81069
24533         * config/nvptx/nvptx.c (nvptx_single): Insert diverging branch as late
24534         as possible.
24536 2017-07-17  Sebastian Huber  <sebastian.huber@embedded-brains.de>
24538         * config/sparc/rtemself.h (TARGET_OS_CPP_BUILTINS): Add
24539         conditional builtin define __FIX_LEON3FT_B2BST.
24541 2017-07-17  Daniel Cederman  <cederman@gaisler.com>
24543         * config/sparc/t-rtems: Add mfix-gr712rc multilibs. Replace
24544         MULTILIB_EXCEPTIONS with MULTILIB_REQUIRED. Match -mfix-gr712rc
24545         with -mfix-ut700.
24547 2017-07-16  Eric Botcazou  <ebotcazou@adacore.com>
24549         PR rtl-optimization/81424
24550         * optabs.c (prepare_cmp_insn): Use copy_to_reg instead of force_reg
24551         to remove potential trapping from operands if -fnon-call-exceptions.
24553 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24555         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Use
24556         profile_proability for scalling.
24557         * scale_profile_for_vect_loop.c (scale_profile_for_vect_loop): Likewise.
24559 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24561         * cgraph.c (cgraph_edge::redirect_call_stmt_to_caller): Cleanup.
24563 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24565         * cfgloopmanip.c (scale_loop_profile): Avoid use of REG_BR_PROB_BASE
24566         fixpoint arithmetics.
24568 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24570         * tree-ssa-loop-unswitch.c (hoist_guard): Avoid use of REG_BR_PROB_BASE
24571         fixpoint arithmetics.
24573 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24575         * asan.c (create_cond_insert_point): Avoid use of REG_BR_PROB_BASE
24576         fixpoint arithmetics.
24578 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24580         * profile-count.h (profile_probability::from_reg_br_prob_note,
24581         profile_probability::to_reg_br_prob_note): New functions.
24582         * doc/rtl.texi (REG_BR_PROB_NOTE): Update documentation.
24583         * reg-notes.h (REG_BR_PROB, REG_BR_PRED): Update docs.
24584         * predict.c (probability_reliable_p): Update.
24585         (edge_probability_reliable_p): Update.
24586         (br_prob_note_reliable_p): Update.
24587         (invert_br_probabilities): Update.
24588         (add_reg_br_prob_note): New function.
24589         (combine_predictions_for_insn): Update.
24590         * asan.c (asan_clear_shadow): Update.
24591         * cfgbuild.c (compute_outgoing_frequencies): Update.
24592         * cfgrtl.c (force_nonfallthru_and_redirect): Update.
24593         (update_br_prob_note): Update.
24594         (rtl_verify_edges): Update.
24595         (purge_dead_edges): Update.
24596         (fixup_reorder_chain): Update.
24597         * emit-rtl.c (try_split): Update.
24598         * ifcvt.c (cond_exec_process_insns): Update.
24599         (cond_exec_process_if_block): Update.
24600         (dead_or_predicable): Update.
24601         * internal-fn.c (expand_addsub_overflow): Update.
24602         (expand_neg_overflow): Update.
24603         (expand_mul_overflow): Update.
24604         * loop-doloop.c (doloop_modify): Update.
24605         * loop-unroll.c (compare_and_jump_seq): Update.
24606         * optabs.c (emit_cmp_and_jump_insn_1): Update.
24607         * predict.h: Update.
24608         * reorg.c (mostly_true_jump): Update.
24609         * rtl.h: Update.
24610         * config/aarch64/aarch64.c (aarch64_emit_unlikely_jump): Update.
24611         * config/alpha/alpha.c (emit_unlikely_jump): Update.
24612         * config/arc/arc.c: (emit_unlikely_jump): Update.
24613         * config/arm/arm.c: (emit_unlikely_jump): Update.
24614         * config/bfin/bfin.c (cbranch_predicted_taken_p): Update.
24615         * config/frv/frv.c (frv_print_operand_jump_hint): Update.
24616         * config/i386/i386.c (ix86_expand_split_stack_prologue): Update.
24617         (ix86_print_operand): Update.
24618         (ix86_split_fp_branch): Update.
24619         (predict_jump): Update.
24620         * config/ia64/ia64.c (ia64_print_operand): Update.
24621         * config/mmix/mmix.c (mmix_print_operand): Update.
24622         * config/powerpcspe/powerpcspe.c (output_cbranch): Update.
24623         (rs6000_expand_split_stack_prologue): Update.
24624         * config/rs6000/rs6000.c: Update.
24625         * config/s390/s390.c (s390_expand_vec_strlen): Update.
24626         (s390_expand_vec_movstr): Update.
24627         (s390_expand_cs_tdsi): Update.
24628         (s390_expand_split_stack_prologue): Update.
24629         * config/sh/sh.c (sh_print_operand): Update.
24630         (expand_cbranchsi4): Update.
24631         (expand_cbranchdi4): Update.
24632         * config/sparc/sparc.c (output_v9branch): Update.
24633         * config/spu/spu.c (get_branch_target): Update.
24634         (ea_load_store_inline): Update.
24635         * config/tilegx/tilegx.c (cbranch_predicted_p): Update.
24636         * config/tilepro/tilepro.c: Update.
24638 2017-07-16  Eric Botcazou  <ebotcazou@adacore.com>
24640         * gimplify.c (mostly_copy_tree_r): Revert latest change.
24641         (gimplify_save_expr): Likewise.
24643 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24645         * ipa-visibility.c (function_and_variable_visibility): Fix pasto.
24647 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24649         * ipa-fnsummary.c (pass_data_ipa_fn_summary): Use
24650         TV_IPA_FNSUMMARY.
24651         * timevar.def (TV_IPA_FNSUMMARY): Define.
24653 2017-07-16  Daniel Cederman  <cederman@gaisler.com>
24655         * config/sparc/sparc.md (divdf3_fix): Add NOP to prevent back
24656         to back store errata sensitive sequence from being generated.
24657         (sqrtdf2_fix): Likewise.
24659 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24661         * tree-ssa-threadupdate.c (compute_path_counts,
24662         update_joiner_offpath_counts): Use profile_probability.
24664 2017-07-15  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24666         Revert:
24667         2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24669         * config/arm/arm-c.c (arm_cpu_builtins): Define
24670         __ARM_FEATURE_NUMERIC_MAXMIN solely based on TARGET_VFP5.
24672 2017-07-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
24674         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
24675         array entries to represent __ieee128 versions of the
24676         scalar_test_data_class, scalar_test_neg, scalar_extract_exp,
24677         scalar_extract_sig, and scalar_insert_exp built-in functions.
24678         (altivec_resolve_overloaded_builtin): Add special case handling
24679         for the __builtin_scalar_insert_exp function, as represented by
24680         the P9V_BUILTIN_VEC_VSIEDP constant.
24681         * config/rs6000/rs6000-builtin.def (VSEEQP): Add scalar extract
24682         exponent support for __ieee128 argument.
24683         (VSESQP): Add scalar extract signature support for __ieee128
24684         argument.
24685         (VSTDCNQP): Add scalar test negative support for __ieee128
24686         argument.
24687         (VSIEQP): Add scalar insert exponent support for __int128 argument
24688         with __ieee128 result.
24689         (VSIEQPF): Add scalar insert exponent support for __ieee128
24690         argument with __ieee128 result.
24691         (VSTDCQP): Add scalar test data class support for __ieee128
24692         argument.
24693         (VSTDCNQP): Add overload support for scalar test negative with
24694         __ieee128 argument.
24695         (VSTDCQP): Add overload support for scalar test data class
24696         __ieee128 argument.
24697         * config/rs6000/vsx.md (UNSPEC_VSX_SXSIG) Replace
24698         UNSPEC_VSX_SXSIGDP.
24699         (UNSPEC_VSX_SIEXPQP): New constant.
24700         (xsxexpqp): New insn for VSX scalar extract exponent quad
24701         precision.
24702         (xsxsigqp): New insn for VSX scalar extract significand quad
24703         precision.
24704         (xsiexpqpf): New insn for VSX scalar insert exponent quad
24705         precision with floating point argument.
24706         (xststdcqp): New expand for VSX scalar test data class quad
24707         precision.
24708         (xststdcnegqp): New expand for VSX scalar test negative quad
24709         precision.
24710         (xststdcqp): New insn to match expansions for VSX scalar test data
24711         class quad precision and VSX scalar test negative quad precision.
24712         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Add
24713         special case operand checking to enforce that second operand of
24714         VSX scalar test data class with quad precision argument is a 7-bit
24715         unsigned literal.
24716         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
24717         prototypes and descriptions of __ieee128 versions of
24718         scalar_extract_exp, scalar_extract_sig, scalar_insert_exp,
24719         scalar_test_data_class, and scalar_test_neg built-in functions.
24721 2016-07-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24723         PR tree-optimization/81162
24724         * gimple-ssa-strength-reduction.c (replace_mult_candidate): Don't
24725         replace a negate with an add.
24727 2017-07-14  James Greenhalgh  <james.greenhalgh@arm.com>
24729         * doc/invoke.texi (arm/-mcpu): Document +crypto.
24731 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24733         * config/arm/arm-c.c (arm_cpu_builtins): Define
24734         __ARM_FEATURE_NUMERIC_MAXMIN solely based on TARGET_VFP5.
24736 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24738         * config/arm/arm-cpus.in (cortex-r52): Add new entry.
24739         (armv8-r): Set ARM Cortex-R52 as default CPU.
24740         * config/arm/arm-tables.opt: Regenerate.
24741         * config/arm/arm-tune.md: Regenerate.
24742         * config/arm/driver-arm.c (arm_cpu_table): Add entry for ARM
24743         Cortex-R52.
24744         * doc/invoke.texi: Mention -mtune=cortex-r52 and availability of fp.dp
24745         extension for -mcpu=cortex-r52.
24747 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24749         * config/arm/arm-isa.h (isa_bit_FP_ARMv8): Delete enumerator.
24750         (ISA_FP_ARMv8): Define as ISA_FPv5 and ISA_FP_D32.
24751         * config/arm/arm-cpus.in (armv8-r): Define fp.sp as enabling FPv5.
24752         (fp-armv8): Define it as FP_ARMv8 only.
24753         config/arm/arm.h (TARGET_FPU_ARMV8): Delete.
24754         (TARGET_VFP_FP16INST): Define using TARGET_VFP5 rather than
24755         TARGET_FPU_ARMV8.
24756         config/arm/arm.c (arm_rtx_costs_internal): Replace checks against
24757         TARGET_FPU_ARMV8 by checks against TARGET_VFP5.
24758         * config/arm/arm-builtins.c (arm_builtin_vectorized_function): Define
24759         first ARM_CHECK_BUILTIN_MODE definition using TARGET_VFP5 rather
24760         than TARGET_FPU_ARMV8.
24761         * config/arm/arm-c.c (arm_cpu_builtins): Likewise for
24762         __ARM_FEATURE_NUMERIC_MAXMIN macro definition.
24763         * config/arm/arm.md (cmov<mode>): Condition on TARGET_VFP5 rather than
24764         TARGET_FPU_ARMV8.
24765         * config/arm/neon.md (neon_vrint): Likewise.
24766         (neon_vcvt): Likewise.
24767         (neon_<fmaxmin_op><mode>): Likewise.
24768         (<fmaxmin><mode>3): Likewise.
24769         * config/arm/vfp.md (l<vrint_pattern><su_optab><mode>si2): Likewise.
24770         * config/arm/predicates.md (arm_cond_move_operator): Check against
24771         TARGET_VFP5 rather than TARGET_FPU_ARMV8 and fix spacing.
24773 2017-07-14  Jackson Woodruff  <jackson.woodruff@arm.com>
24775         * config/aarch64/aarch64.c (aarch64_print_operand): Move comments
24776         to top of function.
24778 2017-07-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24780         * gimple-ssa-store-merging.c (clear_bit_region): Replace reference to
24781         loop in comment with memset.
24783 2017-07-14  Martin Liska  <mliska@suse.cz>
24785         * cfgexpand.c (expand_gimple_basic_block): Remove dead comment.
24786         * dwarf2out.c (is_java): Remove the function.
24787         (output_pubname): Remove usage of the function.
24788         (lower_bound_default): Remove usage of DW_LANG_Java.
24789         (gen_compile_unit_die): Likewise.
24790         * gcc.c: Remove compiler defaults for .java and .zip files.
24791         * gimple-expr.c (remove_suffix): Change as there's no longer
24792         extension than 4-letter one.
24793         * gimplify.c (mostly_copy_tree_r): Remove Java-special part.
24794         (gimplify_save_expr): Likewise.
24795         * ipa-utils.h (polymorphic_type_binfo_p): Remove the comment
24796         as it's possible even for other languages than Java.
24797         * langhooks.h (struct lang_hooks): Remove Java from a comment.
24798         * lto-opts.c (lto_write_options): Remove reference to Java.
24799         * opts.c (strip_off_ending): Update file extension handling.
24800         * tree-cfg.c (verify_gimple_call): Remove comment with Java.
24801         * tree-eh.c (lower_resx): Likewise.
24802         * tree.c (free_lang_data_in_type): Remove dead code.
24803         (find_decls_types_r): Likewise.
24804         (build_common_builtin_nodes): Remove Java from a comment.
24805         (verify_type): Remove dead code.
24806         * varasm.c (assemble_external): Remove Java from a comment.
24808 2017-07-14  Martin Liska  <mliska@suse.cz>
24810         * opts.c (finish_options): Add quotes.
24811         (common_handle_option): Likewise.
24813 2017-07-14  Martin Liska  <mliska@suse.cz>
24815         * dbxout.c (get_lang_number): Do not handle GNU Pascal.
24816         * dbxout.h (extern void dbxout_stab_value_internal_label_diff):
24817         Remove N_SO_PASCAL.
24818         * dwarf2out.c (lower_bound_default): Do not handle
24819         DW_LANG_Pascal83.
24820         (gen_compile_unit_die): Likewise.
24821         * gcc.c: Remove default extension binding for GNU Pascal.
24822         * stmt.c: Remove Pascal language from a comment.
24823         * xcoffout.c: Likewise.
24825 2017-07-13  David Malcolm  <dmalcolm@redhat.com>
24827         PR c/81405
24828         * diagnostic-show-locus.c (fixit_cmp): New function.
24829         (layout::layout): Sort m_fixit_hints.
24830         (column_range::column_range): Assert that the values are valid.
24831         (struct char_span): New struct.
24832         (correction::overwrite): New method.
24833         (struct source_line): New struct.
24834         (line_corrections::add_hint): Add assertions.  Reimplement memcpy
24835         calls in terms of classes source_line and char_span, and
24836         correction::overwrite.
24837         (selftest::test_overlapped_fixit_printing_2): New function.
24838         (selftest::diagnostic_show_locus_c_tests): Call it.
24840 2017-07-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
24842         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Return
24843         early if there is no lhs.
24845 2017-07-13  Martin Liska  <mliska@suse.cz>
24847         * dwarf2out.c (gen_pointer_type_die): Remove dead code.
24848         (gen_reference_type_die): Likewise.
24849         * stor-layout.c: Remove Pascal-related comment.
24851 2017-07-13  Martin Liska  <mliska@suse.cz>
24853         * opts.c (finish_options): Add quotes to error messages.
24854         (parse_sanitizer_options): Likewise.
24856 2017-07-13  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24858         * doc/invoke.texi (armv8-r): Document +fp.sp ARMv8-R extension.
24860 2017-07-13  Richard Earnshaw  <rearnsha@arm.com>
24862         * config/arm/vxworks.h (TARGET_ENDIAN_DEFAULT): Define.
24864 2017-07-13  Maxim Ostapenko  <m.ostapenko@samsung.com>
24866         * asan.c (asan_emit_allocas_unpoison): Use ptr_mode for arguments
24867         during expansion.
24868         * builtins.c (expand_asan_emit_allocas_unpoison): Likewise.
24870 2017-07-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
24872         PR target/81193
24873         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If GLIBC
24874         provides the hardware capability bits, define the macro
24875         __BUILTIN_CPU_SUPPORTS__.
24876         * config/rs6000/rs6000.c (cpu_expand_builtin): Generate a warning
24877         if GLIBC does not provide the hardware capability bits.  Add a
24878         gcc_unreachable call if the built-in cpu function is neither
24879         __builtin_cpu_is nor __builtin_cpu_supports.
24880         (rs6000_get_function_versions_dispatcher): Change the warning
24881         that an old GLIBC is used which does not export the capability
24882         bits to be an error.
24883         * doc/extend.texi (target_clones attribute): Document the
24884         restriction that GLIBC 2.23 or newer is needed on the PowerPC.
24885         (PowerPC built-in functions): Document that GLIBC 2.23 or newer is
24886         needed by __builtin_cpu_is and __builtin_cpu_supports.  Document
24887         the macros defined by GCC if the newer GLIBC is available.
24889 2017-07-12  Jeff Law  <law@redhat.com>
24891         * config/riscv/riscv.c: Remove unnecessary includes.  Reorder
24892         remaining includes slightly.
24893         * config/riscv/riscv-builtins.c: Include profile-count.h.
24895 2017-07-12  Georg-Johann Lay  <avr@gjlay.de>
24897         PR target/79883
24898         * config/avr/avr.c (avr_set_current_function): In diagnostic
24899         messages: Quote keywords and (parts of) identifiers.
24900         [WITH_AVRLIBC]: Warn for functions named "ISR", "SIGNAL" or
24901         "INTERUPT".
24903 2017-07-12  Carl Love  <cel@us.ibm.com>
24905         * config/rs6000/rs6000-c.c: Add support for built-in functions
24906         vector bool char vec_revb (vector bool char);
24907         vector bool short vec_revb (vector short char);
24908         vector bool int vec_revb (vector bool int);
24909         vector bool long long vec_revb (vector bool long long);
24910         * doc/extend.texi: Update the built-in documentation file for the
24911         new built-in functions.
24913 2017-07-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24915         * config/s390/s390.md: Remove movcc splitter.
24917 2017-07-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24919         * config/s390/s390.c (s390_rtx_costs): Return proper costs for
24920         load/store on condition.
24922 2017-07-12  Georg-Johann Lay  <avr@gjlay.de>
24924         PR target/81407
24925         * config/avr/avr.c (avr_encode_section_info)
24926         [progmem && !TREE_READONLY]: Error if progmem object needs
24927         constructing.
24929 2017-07-11  Michael Collison  <michael.collison@arm.com>
24931         * config/aarch64/aarch64-simd.md (aarch64_sub<mode>_compare0):
24932         New pattern.
24934 2017-07-11  Carl Love  <cel@us.ibm.com>
24936         * config/rs6000/rs6000-c.c: Add support for builtins
24937         vector unsigned int vec_parity_lsbb (vector signed int);
24938         vector unsigned int vec_parity_lsbb (vector unsigned int);
24939         vector unsigned __int128 vec_parity_lsbb (vector signed __int128);
24940         vector unsigned __int128 vec_parity_lsbb (vector unsigned __int128);
24941         vector unsigned long long vec_parity_lsbb (vector signed long long);
24942         vector unsigned long long vec_parity_lsbb (vector unsigned long long);
24943         * config/rs6000/rs6000-builtin.def (VPARITY_LSBB): Add BU_P9V_OVERLOAD1.
24944         * config/rs6000/altivec.h (vec_parity_lsbb): Add define.
24945         * doc/extend.texi: Update the built-in documentation file for the
24946         new built-in functions.
24948 2017-07-11  David Malcolm  <dmalcolm@redhat.com>
24950         * diagnostic-show-locus.c: Include "gcc-rich-location.h".
24951         (layout::m_primary_loc): New field.
24952         (layout::layout): Initialize new field.  Move location filtering
24953         logic from here to...
24954         (layout::maybe_add_location_range): ...this new method.  Add
24955         support for filtering to just the lines already specified by other
24956         locations.
24957         (layout::will_show_line_p): New method.
24958         (gcc_rich_location::add_location_if_nearby): New method.
24959         (selftest::test_add_location_if_nearby): New test function.
24960         (selftest::diagnostic_show_locus_c_tests): Call it.
24961         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
24962         New method.
24964 2017-07-11  Tom de Vries  <tom@codesourcery.com>
24966         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG): New macro.
24967         (bb_first_real_insn): New function.
24968         (nvptx_single): Add extra initialization of broadcasted condition
24969         variables.
24971 2017-07-11  Nathan Sidwell  <nathan@acm.org>
24973         * dwarf2out.c (gen_member_die): Remove useless check for anon ctors.
24975 2017-07-11  Georg-Johann Lay  <avr@gjlay.de>
24977         * doc/extend.texi (AVR Function Attributes): Remove weblink to
24978         Binutils doc as TEXI will mess them up.
24979         * doc/invoke.texi (AVR Options): Same here.
24981 2017-07-11  Daniel Cederman  <cederman@gaisler.com>
24983         * config/sparc/sparc.opt (mfix-ut700): New option.
24984         (mfix-gr712rc): Likewise.
24985         (sparc_fix_b2bst): New variable.
24986         * doc/invoke.texi (SPARC options): Document them.
24987         (ARM options): Fix warnings.
24988         * config/sparc/sparc.c (sparc_do_work_around_errata): Insert NOP
24989         instructions to prevent sequences that can trigger the store-store
24990         errata for certain LEON3FT processors.
24991         (pass_work_around_errata::gate): Also test sparc_fix_b2bst.
24992         (sparc_option_override): Set sparc_fix_b2bst appropriately.
24993         * config/sparc/sparc.md (fix_b2bst): New attribute.
24994         (in_branch_delay): Prevent stores in delay slot if fix_b2bst.
24996 2017-07-10  Uros Bizjak  <ubizjak@gmail.com>
24998         PR target/81375
24999         * config/i386/i386.md (divsf3): Add TARGET_SSE to TARGET_SSE_MATH.
25000         (rcpps): Ditto.
25001         (*rsqrtsf2_sse): Ditto.
25002         (rsqrtsf2): Ditto.
25003         (div<mode>3): Macroize insn from divdf3 and divsf3
25004         using MODEF mode iterator.
25006 2017-07-10  Martin Sebor  <msebor@redhat.com>
25008         PR tree-optimization/80397
25009         * gimple-ssa-sprintf.c (format_integer): Use INTEGRAL_TYPE_P()
25010         instead of testing for equality to INTEGER_TYPE.
25012 2017-07-10  Vineet Gupta <vgupta@synopsys.com>
25014         * config.gcc: Remove uclibc from arc target spec.
25016 2017-07-10  Claudiu Zissulescu  <claziss@synopsys.com>
25018         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Define.
25020 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
25022         PR lto/80838
25023         * lto-wrapper.c (remove_option): New function.
25024         (merge_and_complain): Merge PIC/PIE options more realistically.
25026 2017-07-10  Georg-Johann Lay  <avr@gjlay.de>
25028         Better ISR prologues by supporting GASes __gcc_isr pseudo insn.
25030         PR target/20296
25031         PR target/81268
25032         * configure.ac [target=avr]: Add GAS check for -mgcc-isr.
25033         (HAVE_AS_AVR_MGCCISR_OPTION):  If so, AC_DEFINE it.
25034         * config.in: Regenerate.
25035         * configure: Regenerate.
25036         * doc/extend.texi (AVR Function Attributes) <no_gccisr>: Document it.
25037         * doc/invoke.texi (AVR Options) <-mgas-isr-prologues>: Document it.
25038         * config/avr/avr.opt (-mgas-isr-prologues): New option and...
25039         (TARGET_GASISR_PROLOGUES): ...target mask.
25040         * common/config/avr/avr-common.c
25041         (avr_option_optimization_table) [OPT_LEVELS_1_PLUS_NOT_DEBUG]:
25042         Set -mgas-isr-prologues.
25043         * config/avr/avr-passes.def (avr_pass_pre_proep): Add
25044         INSERT_PASS_BEFORE for it.
25045         * config/avr/avr-protos.h (make_avr_pass_pre_proep): New proto.
25046         * config/avr/avr.c (avr_option_override)
25047         [!HAVE_AS_AVR_MGCCISR_OPTION]: Unset TARGET_GASISR_PROLOGUES.
25048         (avr_no_gccisr_function_p, avr_hregs_split_reg): New static functions.
25049         (avr_attribute_table) <no_gccisr>: Add new function attribute.
25050         (avr_set_current_function) <is_no_gccisr>: Init machine field.
25051         (avr_pass_data_pre_proep, avr_pass_pre_proep): New pass data
25052         and rtl_opt_pass.
25053         (make_avr_pass_pre_proep): New function.
25054         (emit_push_sfr) <treg>: Add argument to function and use it
25055         instead of TMP_REG.
25056         (avr_expand_prologue) [machine->gasisr.maybe]: Emit gasisr insn
25057         and set machine->gasisr.yes.
25058         (avr_expand_epilogue) [machine->gasisr.yes]: Similar.
25059         (avr_asm_function_end_prologue) [machine->gasisr.yes]: Add
25060         __gcc_isr.n_pushed to .L__stack_usage.
25061         (TARGET_ASM_FINAL_POSTSCAN_INSN): Define to...
25062         (avr_asm_final_postscan_insn): ...this new static function.
25063         * config/avr/avr.h (machine_function)
25064         <is_no_gccisr, use_L__stack_usage>: New fields.
25065         <gasisr, gasisr.yes, gasisr.maybe, gasisr.regno>: New fields.
25066         * config/avr/avr.md (UNSPECV_GASISR): Add unspecv enum.
25067         (GASISR_Prologue, GASISR_Epilogue, GASISR_Done): New define_constants.
25068         (gasisr, *gasisr): New expander and insn.
25069         * config/avr/gen-avr-mmcu-specs.c (print_mcu)
25070         [HAVE_AS_AVR_MGCCISR_OPTION]: Print asm_gccisr spec.
25071         * config/avr/specs.h (ASM_SPEC) <asm_gccisr>: Add sub spec.
25073 2017-07-10  Richard Earnshaw  <rearnsha@arm.com>
25075         * config/arm/parsecpu.awk (gen_comm_data): Do not escape single quotes
25076         in quoted strings.
25078 2017-07-10  Georg-Johann Lay  <avr@gjlay.de>
25080         Move jump-tables out of .text again.
25082         PR target/81075
25083         * config/avr/avr.c (ASM_OUTPUT_ADDR_VEC_ELT): Remove function.
25084         (ASM_OUTPUT_ADDR_VEC): New function.
25085         (avr_adjust_insn_length) [JUMP_TABLE_DATA_P]: Return 0.
25086         (avr_final_prescan_insn) [avr_log.insn_addresses]: Dump
25087         INSN_ADDRESSes as asm comment.
25088         * config/avr/avr.h (JUMP_TABLES_IN_TEXT_SECTION): Adjust comment.
25089         (ASM_OUTPUT_ADDR_VEC_ELT): Remove define.
25090         (ASM_OUTPUT_ADDR_VEC): Define to avr_output_addr_vec.
25091         * config/avr/avr.md (*tablejump): Adjust comment.
25092         * config/avr/elf.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove.
25093         * config/avr/avr-log.c (avr_log_set_avr_log) <insn_addresses>:
25094         New detail.
25095         * config/avr/avr-protos.h (avr_output_addr_vec_elt): Remove proto.
25096         (avr_output_addr_vec): New proto.
25097         (avr_log_t) <insn_addresses>: New field.
25099 2017-07-09  H.J. Lu  <hongjiu.lu@intel.com>
25101         PR target/81313
25102         * config/i386/i386.c (ix86_function_arg_advance): Set
25103         outgoing_args_on_stack to true if there are outgoing arguments
25104         on stack.
25105         (ix86_function_arg): Likewise.
25106         (ix86_get_drap_rtx): Use DRAP only if there are outgoing
25107         arguments on stack and ACCUMULATE_OUTGOING_ARGS is false.
25108         * config/i386/i386.h (machine_function): Add
25109         outgoing_args_on_stack.
25111 2017-07-09  Krister Walfridsson  <krister.walfridsson@gmail.com>
25113         * config.gcc (*-*-netbsd*): Remove check for NetBSD versions not
25114         supporting pthreds.
25115         * config/netbsd.h (NETBSD_LIBGCC_SPEC): Always enable pthreads.
25117 2017-07-08  Richard Sandiford  <richard.sandiford@linaro.org>
25119         * Makefile.in (HOOKS_H, RTL_BASE_H, FUNCTION_H, EXPR_H, REGS_H)
25120         (REAL_H): Remove $(MACHMODE_H).
25121         (FIXED_VALUE_H, TREE_CORE_H, CFGLOOP_H): Remove $(MACHMODE_H) and
25122         double-int.h.
25123         (CORETYPES_H): Add signop.h, wide-int.h, wide-int-print.h,
25124         $(MACHMODE_H) and double-int.h.
25125         (build/min-insn-modes.o): Depend on $(CORETYPES_H) rather than
25126         $(MACHMODE_H).
25127         (gengtype-state.o, gengtype.o, build/gengtype.o): Don't depend on
25128         double-int.h.
25130 2017-07-07  Andrew Pinski  <apinski@cavium.com>
25132         * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Check
25133         prev_set and curr_set for AARCH64_FUSE_ALU_BRANCH.
25135 2017-07-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
25137         * config/rs6000/rs6000.c (rs6000_get_function_versions_dispatcher):
25138         Add warning if GCC was not configured to link against a GLIBC that
25139         exports the hardware capability bits.
25140         (make_resolver_func): Make resolver function private and not a
25141         COMDAT function.  Create the name with clone_function_name instead
25142         of make_unique_name.
25144         PR target/81348
25145         * config/rs6000/rs6000.md (HI sign_extend splitter): Use the
25146         correct operand in doing the split.
25148 2017-07-07 Carl Love  <cel@us.ibm.com>
25150         * config/rs6000/rs6000-c: Add support for built-in function
25151         vector unsigned short vec_pack_to_short_fp32 (vector float,
25152                                                       vector float).
25153         * config/rs6000/rs6000-builtin.def (CONVERT_4F32_8I16): Add
25154         BU_P9V_AV_2 and BU_P9V_OVERLOAD_2 definitions.
25155         * config/rs6000/altivec.h (vec_pack_to_short_fp32): Add define.
25156         * config/rs6000/altivec.md(UNSPEC_CONVERT_4F32_8I16): Add UNSPEC.
25157         (convert_4f32_8i16): Add define_expand.
25158         * doc/extend.texi: Update the built-in documentation file for the
25159         new built-in function.
25161 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25163         * config/sparc/m8.md: New file.
25164         * config/sparc/sparc.md: Include m8.md.
25166 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25168         * config/sparc/sparc.opt: New option -mvis4b.
25169         * config/sparc/sparc.c (dump_target_flag_bits): Handle MASK_VIS4B.
25170         (sparc_option_override): Handle VIS4B.
25171         (enum sparc_builtins): Define
25172         SPARC_BUILTIN_DICTUNPACK{8,16,32},
25173         SPARC_BUILTIN_FPCMP{LE,GT,EQ,NE}{8,16,32}SHL,
25174         SPARC_BUILTIN_FPCMPU{LE,GT}{8,16,32}SHL,
25175         SPARC_BUILTIN_FPCMPDE{8,16,32}SHL and
25176         SPARC_BUILTIN_FPCMPUR{8,16,32}SHL.
25177         (check_constant_argument): New function.
25178         (sparc_vis_init_builtins): Define builtins
25179         __builtin_vis_dictunpack{8,16,32},
25180         __builtin_vis_fpcmp{le,gt,eq,ne}{8,16,32}shl,
25181         __builtin_vis_fpcmpu{le,gt}{8,16,32}shl,
25182         __builtin_vis_fpcmpde{8,16,32}shl and
25183         __builtin_vis_fpcmpur{8,16,32}shl.
25184         (sparc_expand_builtin): Check that the constant operands to
25185         __builtin_vis_fpcmp*shl and _builtin_vis_dictunpack* are indeed
25186         constant and in range.
25187         * config/sparc/sparc-c.c (sparc_target_macros): Handle
25188         TARGET_VIS4B.
25189         * config/sparc/sparc.h (SPARC_IMM2_P): Define.
25190         (SPARC_IMM5_P): Likewise.
25191         * config/sparc/sparc.md (cpu_feature): Add new feagure "vis4b".
25192         (enabled): Handle vis4b.
25193         (UNSPEC_DICTUNPACK): New unspec.
25194         (UNSPEC_FPCMPSHL): Likewise.
25195         (UNSPEC_FPUCMPSHL): Likewise.
25196         (UNSPEC_FPCMPDESHL): Likewise.
25197         (UNSPEC_FPCMPURSHL): Likewise.
25198         (cpu_feature): New CPU feature `vis4b'.
25199         (dictunpack{8,16,32}): New insns.
25200         (FPCSMODE): New mode iterator.
25201         (fpcscond): New code iterator.
25202         (fpcsucond): Likewise.
25203         (fpcmp{le,gt,eq,ne}{8,16,32}{si,di}shl): New insns.
25204         (fpcmpu{le,gt}{8,16,32}{si,di}shl): Likewise.
25205         (fpcmpde{8,16,32}{si,di}shl): Likewise.
25206         (fpcmpur{8,16,32}{si,di}shl): Likewise.
25207         * config/sparc/constraints.md: Define constraints `q' for unsigned
25208         2-bit integer constants and `t' for unsigned 5-bit integer
25209         constants.
25210         * config/sparc/predicates.md (imm5_operand_dictunpack8): New
25211         predicate.
25212         (imm5_operand_dictunpack16): Likewise.
25213         (imm5_operand_dictunpack32): Likewise.
25214         (imm2_operand): Likewise.
25215         * doc/invoke.texi (SPARC Options): Document -mvis4b.
25216         * doc/extend.texi (SPARC VIS Built-in Functions): Document the
25217         ditunpack* and fpcmp*shl builtins.
25219 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25221         * config.gcc: Handle m8 in --with-{cpu,tune} options.
25222         * config.in: Add HAVE_AS_SPARC6 define.
25223         * config/sparc/driver-sparc.c (cpu_names): Add entry for the SPARC
25224         M8.
25225         * config/sparc/sol2.h (CPP_CPU64_DEFAULT_SPEC): Define for
25226         TARGET_CPU_m8.
25227         (ASM_CPU32_DEFAUILT_SPEC): Likewise.
25228         (CPP_CPU_SPEC): Handle m8.
25229         (ASM_CPU_SPEC): Likewise.
25230         * config/sparc/sparc-opts.h (enum processor_type): Add
25231         PROCESSOR_M8.
25232         * config/sparc/sparc.c (m8_costs): New struct.
25233         (sparc_option_override): Handle TARGET_CPU_m8.
25234         (sparc32_initialize_trampoline): Likewise.
25235         (sparc64_initialize_trampoline): Likewise.
25236         (sparc_issue_rate): Likewise.
25237         (sparc_register_move_cost): Likewise.
25238         * config/sparc/sparc.h (TARGET_CPU_m8): Define.
25239         (CPP_CPU64_DEFAULT_SPEC): Define for M8.
25240         (ASM_CPU64_DEFAULT_SPEC): Likewise.
25241         (CPP_CPU_SPEC): Handle M8.
25242         (ASM_CPU_SPEC): Likewise.
25243         (AS_M8_FLAG): Define.
25244         * config/sparc/sparc.md: Add m8 to the cpu attribute.
25245         * config/sparc/sparc.opt: New option -mcpu=m8 for sparc targets.
25246         * configure.ac (HAVE_AS_SPARC6): Check for assembler support for
25247         M8 instructions.
25248         * configure: Regenerate.
25249         * doc/invoke.texi (SPARC Options): Document -mcpu=m8 and
25250         -mtune=m8.
25252 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25254         * config/sparc/niagara7.md: Rework the DFA scheduler to use insn
25255         subtypes.
25256         * config/sparc/sparc.md: Remove the `v3pipe' insn attribute.
25257         ("*movdi_insn_sp32"): Do not set v3pipe.
25258         ("*movsi_insn"): Likewise.
25259         ("*movdi_insn_sp64"): Likewise.
25260         ("*movsf_insn"): Likewise.
25261         ("*movdf_insn_sp32"): Likewise.
25262         ("*movdf_insn_sp64"): Likewise.
25263         ("*zero_extendsidi2_insn_sp64"): Likewise.
25264         ("*sign_extendsidi2_insn"): Likewise.
25265         ("*mov<VM32:mode>_insn"): Likewise.
25266         ("*mov<VM64:mode>_insn_sp64"): Likewise.
25267         ("*mov<VM64:mode>_insn_sp32"): Likewise.
25268         ("<plusminus_insn><VADDSUB:mode>3"): Likewise.
25269         ("<vlop:code><VL:mode>3"): Likewise.
25270         ("*not_<vlop:code><VL:mode>3"): Likewise.
25271         ("*nand<VL:mode>_vis"): Likewise.
25272         ("*<vlnotop:code>_not1<VL:mode>_vis"): Likewise.
25273         ("*<vlnotop:code>_not2<VL:mode>_vis"): Likewise.
25274         ("one_cmpl<VL:mode>2"): Likewise.
25275         ("faligndata<VM64:mode>_vis"): Likewise.
25276         ("alignaddrsi_vis"): Likewise.
25277         ("alignaddrdi_vis"): Likweise.
25278         ("alignaddrlsi_vis"): Likewise.
25279         ("alignaddrldi_vis"): Likewise.
25280         ("fcmp<gcond:code><GCM:gcm_name><P:mode>_vis"): Likewise.
25281         ("bmaskdi_vis"): Likewise.
25282         ("bmasksi_vis"): Likewise.
25283         ("bshuffle<VM64:mode>_vis"): Likewise.
25284         ("cmask8<P:mode>_vis"): Likewise.
25285         ("cmask16<P:mode>_vis"): Likewise.
25286         ("cmask32<P:mode>_vis"): Likewise.
25287         ("pdistn<P:mode>_vis"): Likewise.
25288         ("<vis3_addsub_ss_patname><VASS:mode>3"): Likewise.
25290 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25292         * config/sparc/sparc.md ("subtype"): New insn attribute.
25293         ("*wrgsr_sp64"): Set insn subtype.
25294         ("*rdgsr_sp64"): Likewise.
25295         ("alignaddrsi_vis"): Likewise.
25296         ("alignaddrdi_vis"): Likewise.
25297         ("alignaddrlsi_vis"): Likewise.
25298         ("alignaddrldi_vis"): Likewise.
25299         ("<plusminus_insn><VADDSUB:mode>3"): Likewise.
25300         ("fexpand_vis"): Likewise.
25301         ("fpmerge_vis"): Likewise.
25302         ("faligndata<VM64:mode>_vis"): Likewise.
25303         ("bshuffle<VM64:mode>_vis"): Likewise.
25304         ("cmask8<P:mode>_vis"): Likewise.
25305         ("cmask16<P:mode>_vis"): Likewise.
25306         ("cmask32<P:mode>_vis"): Likewise.
25307         ("fchksm16_vis"): Likewise.
25308         ("v<vis3_shift_patname><GCM:mode>3"): Likewise.
25309         ("fmean16_vis"): Likewise.
25310         ("fp<plusminus_insn>64_vis"): Likewise.
25311         ("<plusminus_insn>v8qi3"): Likewise.
25312         ("<vis3_addsub_ss_patname><VASS:mode>3"): Likewise.
25313         ("<vis4_minmax_patname><VMMAX:mode>3"): Likewise.
25314         ("<vis4_uminmax_patname><VMMAX:mode>3"): Likewise.
25315         ("<vis3_addsub_ss_patname>v8qi3"): Likewise.
25316         ("<vis4_addsub_us_patname><VAUS:mode>3"): Likewise.
25317         ("*movqi_insn"): Likewise.
25318         ("*movhi_insn"): Likewise.
25319         ("*movsi_insn"): Likewise.
25320         ("movsi_pic_gotdata_op"): Likewise.
25321         ("*movdi_insn_sp32"): Likewise.
25322         ("*movdi_insn_sp64"): Likewise.
25323         ("movdi_pic_gotdata_op"): Likewise.
25324         ("*movsf_insn"): Likewise.
25325         ("*movdf_insn_sp32"): Likewise.
25326         ("*movdf_insn_sp64"): Likewise.
25327         ("*zero_extendhisi2_insn"): Likewise.
25328         ("*zero_extendqihi2_insn"): Likewise.
25329         ("*zero_extendqisi2_insn"): Likewise.
25330         ("*zero_extendqidi2_insn"): Likewise.
25331         ("*zero_extendhidi2_insn"): Likewise.
25332         ("*zero_extendsidi2_insn_sp64"): Likewise.
25333         ("ldfsr"): Likewise.
25334         ("prefetch_64"): Likewise.
25335         ("prefetch_32"): Likewise.
25336         ("tie_ld32"): Likewise.
25337         ("tie_ld64"): Likewise.
25338         ("*tldo_ldub_sp32"): Likewise.
25339         ("*tldo_ldub1_sp32"): Likewise.
25340         ("*tldo_ldub2_sp32"): Likewise.
25341         ("*tldo_ldub_sp64"): Likewise.
25342         ("*tldo_ldub1_sp64"): Likewise.
25343         ("*tldo_ldub2_sp64"): Likewise.
25344         ("*tldo_ldub3_sp64"): Likewise.
25345         ("*tldo_lduh_sp32"): Likewise.
25346         ("*tldo_lduh1_sp32"): Likewise.
25347         ("*tldo_lduh_sp64"): Likewise.
25348         ("*tldo_lduh1_sp64"): Likewise.
25349         ("*tldo_lduh2_sp64"): Likewise.
25350         ("*tldo_lduw_sp32"): Likewise.
25351         ("*tldo_lduw_sp64"): Likewise.
25352         ("*tldo_lduw1_sp64"): Likewise.
25353         ("*tldo_ldx_sp64"): Likewise.
25354         ("*mov<VM32:mode>_insn"): Likewise.
25355         ("*mov<VM64:mode>_insn_sp64"): Likewise.
25356         ("*mov<VM64:mode>_insn_sp32"): Likewise.
25358 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25360         * config/sparc/sparc.md ("type"): New insn type viscmp.
25361         ("fcmp<gcond:code><GCM:gcm_name><P:mode>_vis"): Set insn type to
25362         viscmp.
25363         ("fpcmp<gcond:code>8<P:mode>_vis"): Likewise.
25364         ("fucmp<gcond:code>8<P:mode>_vis"): Likewise.
25365         ("fpcmpu<gcond:code><GCM:gcm_name><P:mode>_vis"): Likewise.
25366         * config/sparc/niagara7.md ("n7_vis_logical_v3pipe"): Handle
25367         viscmp.
25368         ("n7_vis_logical_11cycle"): Likewise.
25369         * config/sparc/niagara4.md ("n4_vis_logical"): Likewise.
25370         * config/sparc/niagara2.md ("niag3_vis": Likewise.
25371         * config/sparc/niagara.md ("niag_vis"): Likewise.
25372         * config/sparc/ultra3.md ("us3_fga"): Likewise.
25373         * config/sparc/ultra1_2.md ("us1_fga_double"): Likewise.
25375 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25377         * config/sparc/sparc.md: New instruction type `bmask'.
25378         (bmaskdi_vis): Use the `bmask' type.
25379         (bmasksi_vis): Likewise.
25380         * config/sparc/ultra3.md (us3_array): Likewise.
25381         * config/sparc/niagara7.md (n7_array): Likewise.
25382         * config/sparc/niagara4.md (n4_array): Likewise.
25383         * config/sparc/niagara2.md (niag2_vis): Likewise.
25384         (niag3_vis): Likewise.
25385         * config/sparc/niagara.md (niag_vis): Likewise.
25387 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25389         * ipa-comdats.c: Remove optimize check from gate.
25390         * ipa-fnsummary.c (ipa_fn_summary_generate): do not generate summary
25391         for functions not optimized.
25392         (ipa_fn_summary_read): Skip optimize check.
25393         (ipa_fn_summary_write): Likewise.
25394         * ipa-inline-analysis.c (do_estimate_growth_1): Check that caller
25395         is optimized.
25396         * ipa-inline.c (can_inline_edge_p): Not optimized functions are
25397         uninlinable.
25398         (can_inline_edge_p): Check flag_pcc_struct_return for match.
25399         (check_callers): Give up on caller which is not optimized.
25400         (inline_small_functions): Likewise.
25401         (ipa_inline): Do not give up when not optimizing.
25402         * ipa-visbility.c (function_and_variable_visibility): Do not optimize
25403         away unoptimizes cdtors.
25404         (whole_program_function_and_variable_visibility): Do
25405         ipa_discover_readonly_nonaddressable_vars in LTO mode.
25406         * ipa.c (process_references): Do not check optimize.
25407         (symbol_table::remove_unreachable_nodes): Update optimize check.
25408         (set_writeonly_bit): Update optimize check.
25409         (pass_ipa_cdtor_merge::gate): Do not check optimize.
25410         (pass_ipa_single_use::gate): Remove.
25412 2017-07-06  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
25414         * config/rs6000/rs6000.c (union_defs, union_uses, insn_is_load_p,
25415         insn_is_store_p, insn_is_swap_p, const_load_sequence_p, v2df_reduction_p,
25416         rtx_is_swappable_p, insn_is_swappable_p, chain_contains_only_swaps,
25417         mark_swaps_for_removal, swap_const_vector_halves, adjust_subreg_index,
25418         permute_load, permute_store, adjust_extract, adjust_splat,
25419         adjust_xxpermdi, adjust_concat, adjust_vperm, handle_special_swappables,
25420         replace_swap_with_copy, dump_swap_insn_table,
25421         alignment_with_canonical_addr, alignment_mask, find_alignment_op,
25422         recombine_lvx_pattern, recombine_stvx_pattern,
25423         recombine_lvx_stvx_patterns, rs6000_analyze_swaps,
25424         make_pass_analyze_swaps): Move all code related to p8 swap optimizations
25425         to file rs6000-p8swap.c.
25426         * config/rs6000/rs6000-p8swap.c: New file.
25427         * config/rs6000/t-rs6000: Add rule to build rs6000-p8swap.o.
25428         * config.gcc: Add rs6000-p8swap.o to extra_objs for powerpc*-*-*
25429         and rs6000*-*-* targets.
25431 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25433         * Makefile.in (selftest): Remove dependency on s-selftest-c++.
25435 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25437         * lto-wrapper.c (merge_and_complain): Do not merge
25438         fexceptions, fnon_call_exceptions, ftrapv, ffp_contract_, fmath_errno,
25439         fsigned_zeros, ftrapping_math, fwrapv.
25440         (append_compiler_options): Do not track these options.
25441         (append_linker_options): Likewie
25443 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25445         * cgraphunit.c (cgraph_node::finalize_function): When
25446         !flag_toplevel_reorde set no_reorder flag.
25447         (varpool_node::finalize_decl): Likewise.
25448         (symbol_table::compile): Drop no toplevel reorder path.
25450 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25452         * bb-reorder.c (better_edge_p): Do not build traces across abnormal/eh
25453         edges; zero probability is not better than uninitialized.
25455 2017-07-06  Maxim Ostapenko  <m.ostapenko@samsung.com>
25457         * asan.h (asan_sanitize_allocas_p): Declare.
25458         * asan.c (asan_sanitize_allocas_p): New function.
25459         (handle_builtin_stack_restore): Bail out if !asan_sanitize_allocas_p.
25460         (handle_builtin_alloca): Likewise.
25461         * cfgexpand.c (expand_used_vars): Do not add allocas unpoisoning stuff
25462         if !asan_sanitize_allocas_p.
25463         * params.def (asan-instrument-allocas): Add new option.
25464         * params.h (ASAN_PROTECT_ALLOCAS): Define.
25465         * opts.c (common_handle_option): Disable allocas sanitization for
25466         KASan by default.
25468 2017-07-06  Maxim Ostapenko  <m.ostapenko@samsung.com>
25470         * asan.c: Include gimple-fold.h.
25471         (get_last_alloca_addr): New function.
25472         (handle_builtin_stackrestore): Likewise.
25473         (handle_builtin_alloca): Likewise.
25474         (asan_emit_allocas_unpoison): Likewise.
25475         (get_mem_refs_of_builtin_call): Add new parameter, remove const
25476         quallifier from first paramerer. Handle BUILT_IN_ALLOCA,
25477         BUILT_IN_ALLOCA_WITH_ALIGN and BUILT_IN_STACK_RESTORE builtins.
25478         (instrument_builtin_call): Pass gimple iterator to
25479         get_mem_refs_of_builtin_call.
25480         (last_alloca_addr): New global.
25481         * asan.h (asan_emit_allocas_unpoison): Declare.
25482         * builtins.c (expand_asan_emit_allocas_unpoison): New function.
25483         (expand_builtin): Handle BUILT_IN_ASAN_ALLOCAS_UNPOISON.
25484         * cfgexpand.c (expand_used_vars): Call asan_emit_allocas_unpoison
25485         if function calls alloca.
25486         * gimple-fold.c (replace_call_with_value): Remove static keyword.
25487         * gimple-fold.h (replace_call_with_value): Declare.
25488         * internal-fn.c: Include asan.h.
25489         * sanitizer.def (BUILT_IN_ASAN_ALLOCA_POISON,
25490         BUILT_IN_ASAN_ALLOCAS_UNPOISON): New builtins.
25492 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25494         * Makefile.in (SELFTEST_FLAGS): Drop "-x c", moving it to...
25495         (C_SELFTEST_FLAGS): New.
25496         (CPP_SELFTEST_FLAGS): New.
25497         (SELFTEST_DEPS): New, from deps of s-selftest.
25498         (C_SELFTEST_DEPS): New, from deps of s-selftest.
25499         (CPP_SELFTEST_DEPS): New.
25500         (selftest): Add dependency on s-selftest-c++.
25501         (s-selftest): Rename to...
25502         (s-selftest-c): ...this, moving deps to SELFTEST_DEPS
25503         and C_SELFTEST_DEPS, and using C_SELFTEST_FLAGS rather
25504         than SELFTEST_FLAGS.
25505         (selftest-gdb): Rename to...
25506         (selftest-c-gdb): ...this, using C_SELFTEST_DEPS and
25507         C_SELFTEST_FLAGS.
25508         (selftest-gdb): Reintroduce as an alias for selftest-c-gdb.
25509         (selftest-valgrind): Rename to...
25510         (selftest-c-valgrind): ...this, using C_SELFTEST_DEPS and
25511         C_SELFTEST_FLAGS.
25512         (selftest-valgrind): Reintroduce as an alias for
25513         selftest-c-valgrind.
25514         (s-selftest-c++): New.
25515         (selftest-c++-gdb): New.
25516         (selftest-c++-valgrind): New.
25518 2017-07-06  Olivier Hainque  <hainque@adacore.com>
25520         * gcc.c (process_command): When deciding if undefined variables
25521         should be ignored when processing specs, accept "gcc -v" as well.
25523 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25525         * auto-profile.c (afdo_set_bb_count, afdo_propagate_edge,
25526         afdo_annotate_cfg): Set counts/probabilities as determined by afdo.
25528 2017-07-06  Thomas Preud'homme  <thomas.preudhomme@arm.com>
25530         * config/arm/arm-cpus.in (armv8-r): Add new entry.
25531         * config/arm/arm-isa.h (ISA_ARMv8r): Define macro.
25532         * config/arm/arm-tables.opt: Regenerate.
25533         * config/arm/arm.h (enum base_architecture): Add BASE_ARCH_8R
25534         enumerator.
25535         * doc/invoke.texi: Mention -march=armv8-r and its extensions.
25537 2017-07-06  Carl Love  <cel@us.ibm.com>
25539         * ChangeLog: Clean up from mid air collision
25541 2017-07-06  Carl Love  <cel@us.ibm.com>
25543         * config/rs6000/rs6000-c.c: Add support for built-in functions
25544         vector signed int vec_subc (vector signed int, vector signed int);
25545         vector signed __int128 vec_subc (vector signed __int128,
25546                                          vector signed __int128);
25547         vector unsigned __int128 vec_subc (vector unsigned __int128,
25548                                            vector unsigned __int128);
25549         vector signed int vec_sube (vector signed int, vector signed int,
25550                                     vector signed int);
25551         vector unsigned int vec_sube (vector unsigned int,
25552                                       vector unsigned int,
25553                                       vector unsigned int);
25554         vector signed __int128 vec_sube (vector signed __int128,
25555                                          vector signed __int128,
25556                                          vector signed__int128);
25557         vector unsigned __int128 vec_sube (vector unsigned __int128,
25558                                            vector unsigned __int128,
25559                                            vector unsigned __int128);
25560         vector signed int vec_subec (vector signed int, vector signed int,
25561                                      vector signed int);
25562         vector unsigned int vec_subec (vector unsigned int,
25563                                        vector unsigned int,
25564                                        vector unsigned int);
25565         vector signed __int128 vec_subec (vector signed __int128,
25566                                           vector signed __int128,
25567                                           vector signed__int128);
25568         vector unsigned __int128 vec_subec (vector unsigned __int128,
25569                                             vector unsigned __int128,
25570                                             vector unsigned __int128);
25571         * config/rs6000/rs6000.c (ALTIVEC_BUILTIN_VEC_SUBE,
25572         ALTIVEC_BUILTIN_VEC_SUBEC): Add ef_builtins.
25573         * config/rs6000/rs6000-builtin.def (SUBE, SUBEC): Add
25574         BU_ALTIVEC_OVERLOAD_X definitions.
25575         * config/rs6000/altivec.h (vec_sube, vec_subec): Add builtin defines.
25576         * doc/extend.texi: Update the built-in documentation file for the new
25577         built-in functions.
25579 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25581         PR c++/79300
25582         * diagnostic-show-locus.c (layout::layout): Use start and finish
25583         spelling location for the start and finish of each range.
25584         * genmatch.c (linemap_client_expand_location_to_spelling_point):
25585         Add unused aspect param.
25586         * input.c (expand_location_1): Add "aspect" param, and use it
25587         to access the correct part of the location.
25588         (expand_location): Pass LOCATION_ASPECT_CARET to new param of
25589         expand_location_1.
25590         (expand_location_to_spelling_point): Likewise.
25591         (linemap_client_expand_location_to_spelling_point): Add "aspect"
25592         param, and pass it to expand_location_1.
25594 2017-07-06  Sebastian Peryt  <sebastian.peryt@intel.com>
25596         * config/i386/avx512fintrin.h (_mm_mask_getexp_round_ss,
25597         _mm_maskz_getexp_round_ss, _mm_mask_getexp_round_sd,
25598         _mm_maskz_getexp_round_sd, _mm_mask_getmant_round_sd,
25599         _mm_maskz_getmant_round_sd, _mm_mask_getmant_round_ss,
25600         _mm_maskz_getmant_round_ss, _mm_mask_getexp_ss, _mm_maskz_getexp_ss,
25601         _mm_mask_getexp_sd, _mm_maskz_getexp_sd, _mm_mask_getmant_sd,
25602         _mm_maskz_getmant_sd, _mm_mask_getmant_ss,
25603         _mm_maskz_getmant_ss): New intrinsics.
25604         (__builtin_ia32_getexpss128_mask): Changed to ...
25605         __builtin_ia32_getexpss128_round ... this.
25606         (__builtin_ia32_getexpsd128_mask): Changed to ...
25607         __builtin_ia32_getexpsd128_round ... this.
25608         * config/i386/i386-builtin-types.def
25609         ((V2DF, V2DF, V2DF, INT, V2DF, UQI, INT),
25610         (V4SF, V4SF, V4SF, INT, V4SF, UQI, INT)): New function type aliases.
25611         * config/i386/i386-builtin.def (__builtin_ia32_getexpsd_mask_round,
25612         __builtin_ia32_getexpss_mask_round,     __builtin_ia32_getmantsd_mask_round,
25613         __builtin_ia32_getmantss_mask_round): New builtins.
25614         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_INT_V2DF_UQI_INT,
25615         V4SF_FTYPE_V4SF_V4SF_INT_V4SF_UQI_INT): Handle new types.
25616         (CODE_FOR_avx512f_vgetmantv2df_mask_round,
25617         CODE_FOR_avx512f_vgetmantv4sf_mask_round): New cases.
25618         * config/i386/sse.md
25619         (avx512f_sgetexp<mode><round_saeonly_name>): Changed to ...
25620         avx512f_sgetexp<mode><mask_scalar_name>
25621         <round_saeonly_scalar_name> ... this.
25622         (vgetexp<ssescalarmodesuffix>\t{<round_saeonly_op3>%2, %1, %0|
25623         %0, %1, %2<round_saeonly_op3>}): Changed to ...
25624         vgetexp<ssescalarmodesuffix>
25625         \t{<round_saeonly_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25626         %0<mask_scalar_operand3>, %1, %2<round_saeonly_scalar_mask_op3>} ... this.
25627         (avx512f_vgetmant<mode><round_saeonly_name>): Changed to ...
25628         avx512f_vgetmant<mode><mask_scalar_name>
25629         <round_saeonly_scalar_name> ... this.
25630         (vgetmant<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
25631         %0, %1, %2<round_saeonly_op4>, %3}): Changed to ...
25632         vgetmant<ssescalarmodesuffix>
25633         \t{%3, <round_saeonly_scalar_mask_op4>%2, %1, %0<mask_scalar_operand4>|
25634         %0<mask_scalar_operand4>, %1, %2
25635         <round_saeonly_scalar_mask_op4>, %3} ... this.
25636         * config/i386/subst.md (mask_scalar_operand4,
25637         round_saeonly_scalar_mask_operand4,     round_saeonly_scalar_mask_op4,
25638         round_saeonly_scalar_nimm_predicate): New subst attributes.
25640 2017-07-06  Julia Koval  <julia.koval@intel.com>
25642         * config/i386/i386.c (ix86_erase_embedded_rounding):
25643         Remove code for old rounding pattern.
25645 2017-07-06  Richard Earnshaw  <rearnsha@arm.com>
25647         * config/arm/t-arm (GTM_H): Add arm-cpu.h.
25649 2017-07-06  Christophe Lyon  <christophe.lyon@linaro.org>
25651         * doc/sourcebuild.texi (Test Directives, Variants of
25652         dg-require-support): Add documentation for dg-require-stack-check.
25654 2017-07-05  Sebastian Peryt  <sebastian.peryt@intel.com>
25656         * config/i386/subst.md (mask_scalar, round_scalar,
25657         round_saeonly_scalar): New meta-templates.
25658         (mask_scalar_name, mask_scalar_operand3, round_scalar_name,
25659         round_scalar_mask_operand3, round_scalar_mask_op3,
25660         round_scalar_constraint, round_scalar_prefix, round_saeonly_scalar_name,
25661         round_saeonly_scalar_mask_operand3, round_saeonly_scalar_mask_op3,
25662         round_saeonly_scalar_constraint,
25663         round_saeonly_scalar_prefix): New subst attribute.
25664         * config/i386/sse.md
25665         (<sse>_vm<plusminus_insn><mode>3<mask_name><round_name>): Renamed to ...
25666         <sse>_vm<plusminus_insn><mode>3<mask_scalar_name>
25667         <round_scalar_name> ... this.
25668         (<sse>_vm<multdiv_mnemonic><mode>3<mask_name><round_name>): Renamed to ...
25669         <sse>_vm<multdiv_mnemonic><mode>3<mask_scalar_name>
25670         <round_scalar_name> ... this.
25671         (<sse>_vm<code><mode>3<mask_name><round_saeonly_name>): Renamed to ...
25672         <sse>_vm<code><mode>3<mask_scalar_name>
25673         <round_saeonly_scalar_name> ... this.
25674         (v<plusminus_mnemonic><ssescalarmodesuffix>
25675         \t{<round_mask_op3>%2, %1, %0<mask_operand3>|
25676         %0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}): Changed to ...
25677         v<plusminus_mnemonic><ssescalarmodesuffix>
25678         \t{<round_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25679         %0<mask_scalar_operand3>, %1, %<iptr>2<round_scalar_mask_op3>} ... this.
25680         (v<multdiv_mnemonic><ssescalarmodesuffix>
25681         \t{<round_mask_op3>%2, %1, %0<mask_operand3>|
25682         %0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}): Changed to ...
25683         v<multdiv_mnemonic><ssescalarmodesuffix>
25684         \t{<round_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25685         %0<mask_scalar_operand3>, %1, %<iptr>2<round_scalar_mask_op3>} ... this.
25686         (v<maxmin_float><ssescalarmodesuffix>
25687         \t{<round_saeonly_mask_op3>%2, %1, %0<mask_operand3>|
25688         %0<mask_operand3>, %1, %<iptr>2<round_saeonly_mask_op3>}): Changed to ...
25689         v<maxmin_float><ssescalarmodesuffix>
25690         \t{<round_saeonly_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25691         %0<mask_scalar_operand3>, %1, %<iptr>2
25692         <round_saeonly_scalar_mask_op3>} ... this.
25694 2017-07-05  Richard Earnshaw  <rearnsha@arm.com>
25696         * config/arm/arm.c (arm_fixed_condition_code_regs): New function.
25697         (TARGET_FIXED_CONDITION_CODE_REGS): Redefine.
25699 2017-07-05  Richard Sandiford  <richard.sandiford@linaro.org>
25700             Alan Hayward  <alan.hayward@arm.com>
25701             David Sherwood  <david.sherwood@arm.com>
25703         * combine.c (simplify_if_then_else): Remove "enum" before
25704         "machine_mode".
25705         * compare-elim.c (can_eliminate_compare): Likewise.
25706         * config/aarch64/aarch64-builtins.c (aarch64_simd_builtin_std_type):
25707         Likewise.
25708         (aarch64_lookup_simd_builtin_type): Likewise.
25709         (aarch64_simd_builtin_type): Likewise.
25710         (aarch64_init_simd_builtin_types): Likewise.
25711         (aarch64_simd_expand_args): Likewise.
25712         * config/aarch64/aarch64-protos.h (aarch64_simd_attr_length_rglist):
25713         Likewise.
25714         (aarch64_reverse_mask): Likewise.
25715         (aarch64_simd_emit_reg_reg_move): Likewise.
25716         (aarch64_gen_adjusted_ldpstp): Likewise.
25717         (aarch64_ccmp_mode_to_code): Likewise.
25718         (aarch64_operands_ok_for_ldpstp): Likewise.
25719         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
25720         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
25721         Likewise.
25722         (aarch64_min_divisions_for_recip_mul): Likewise.
25723         (aarch64_reassociation_width): Likewise.
25724         (aarch64_get_condition_code_1): Likewise.
25725         (aarch64_simd_emit_reg_reg_move): Likewise.
25726         (aarch64_simd_attr_length_rglist): Likewise.
25727         (aarch64_reverse_mask): Likewise.
25728         (aarch64_operands_ok_for_ldpstp): Likewise.
25729         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
25730         (aarch64_gen_adjusted_ldpstp): Likewise.
25731         * config/aarch64/cortex-a57-fma-steering.c (fma_node::rename):
25732         Likewise.
25733         * config/arc/arc.c (legitimate_offset_address_p): Likewise.
25734         * config/arm/arm-builtins.c (arm_simd_builtin_std_type): Likewise.
25735         (arm_lookup_simd_builtin_type): Likewise.
25736         (arm_simd_builtin_type): Likewise.
25737         (arm_init_simd_builtin_types): Likewise.
25738         (arm_expand_builtin_args): Likewise.
25739         * config/arm/arm-protos.h (arm_expand_builtin): Likewise.
25740         * config/ft32/ft32.c (ft32_libcall_value): Likewise.
25741         (ft32_setup_incoming_varargs): Likewise.
25742         (ft32_function_arg): Likewise.
25743         (ft32_function_arg_advance): Likewise.
25744         (ft32_pass_by_reference): Likewise.
25745         (ft32_arg_partial_bytes): Likewise.
25746         (ft32_valid_pointer_mode): Likewise.
25747         (ft32_addr_space_pointer_mode): Likewise.
25748         (ft32_addr_space_legitimate_address_p): Likewise.
25749         * config/i386/i386-protos.h (ix86_operands_ok_for_move_multiple):
25750         Likewise.
25751         * config/i386/i386.c (ix86_setup_incoming_vararg_bounds): Likewise.
25752         (ix86_emit_outlined_ms2sysv_restore): Likewise.
25753         (iamcu_alignment): Likewise.
25754         (canonicalize_vector_int_perm): Likewise.
25755         (ix86_noce_conversion_profitable_p): Likewise.
25756         (ix86_mpx_bound_mode): Likewise.
25757         (ix86_operands_ok_for_move_multiple): Likewise.
25758         * config/microblaze/microblaze-protos.h
25759         (microblaze_expand_conditional_branch_reg): Likewise.
25760         * config/microblaze/microblaze.c
25761         (microblaze_expand_conditional_branch_reg): Likewise.
25762         * config/powerpcspe/powerpcspe.c (rs6000_init_hard_regno_mode_ok):
25763         Likewise.
25764         (rs6000_reassociation_width): Likewise.
25765         (rs6000_invalid_binary_op): Likewise.
25766         (fusion_p9_p): Likewise.
25767         (emit_fusion_p9_load): Likewise.
25768         (emit_fusion_p9_store): Likewise.
25769         * config/riscv/riscv-protos.h (riscv_regno_mode_ok_for_base_p):
25770         Likewise.
25771         (riscv_hard_regno_mode_ok_p): Likewise.
25772         (riscv_address_insns): Likewise.
25773         (riscv_split_symbol): Likewise.
25774         (riscv_legitimize_move): Likewise.
25775         (riscv_function_value): Likewise.
25776         (riscv_hard_regno_nregs): Likewise.
25777         (riscv_expand_builtin): Likewise.
25778         * config/riscv/riscv.c (riscv_build_integer_1): Likewise.
25779         (riscv_build_integer): Likewise.
25780         (riscv_split_integer): Likewise.
25781         (riscv_legitimate_constant_p): Likewise.
25782         (riscv_cannot_force_const_mem): Likewise.
25783         (riscv_regno_mode_ok_for_base_p): Likewise.
25784         (riscv_valid_base_register_p): Likewise.
25785         (riscv_valid_offset_p): Likewise.
25786         (riscv_valid_lo_sum_p): Likewise.
25787         (riscv_classify_address): Likewise.
25788         (riscv_legitimate_address_p): Likewise.
25789         (riscv_address_insns): Likewise.
25790         (riscv_load_store_insns): Likewise.
25791         (riscv_force_binary): Likewise.
25792         (riscv_split_symbol): Likewise.
25793         (riscv_force_address): Likewise.
25794         (riscv_legitimize_address): Likewise.
25795         (riscv_move_integer): Likewise.
25796         (riscv_legitimize_const_move): Likewise.
25797         (riscv_legitimize_move): Likewise.
25798         (riscv_address_cost): Likewise.
25799         (riscv_subword): Likewise.
25800         (riscv_output_move): Likewise.
25801         (riscv_canonicalize_int_order_test): Likewise.
25802         (riscv_emit_int_order_test): Likewise.
25803         (riscv_function_arg_boundary): Likewise.
25804         (riscv_pass_mode_in_fpr_p): Likewise.
25805         (riscv_pass_fpr_single): Likewise.
25806         (riscv_pass_fpr_pair): Likewise.
25807         (riscv_get_arg_info): Likewise.
25808         (riscv_function_arg): Likewise.
25809         (riscv_function_arg_advance): Likewise.
25810         (riscv_arg_partial_bytes): Likewise.
25811         (riscv_function_value): Likewise.
25812         (riscv_pass_by_reference): Likewise.
25813         (riscv_setup_incoming_varargs): Likewise.
25814         (riscv_print_operand): Likewise.
25815         (riscv_elf_select_rtx_section): Likewise.
25816         (riscv_save_restore_reg): Likewise.
25817         (riscv_for_each_saved_reg): Likewise.
25818         (riscv_register_move_cost): Likewise.
25819         (riscv_hard_regno_mode_ok_p): Likewise.
25820         (riscv_hard_regno_nregs): Likewise.
25821         (riscv_class_max_nregs): Likewise.
25822         (riscv_memory_move_cost): Likewise.
25823         * config/rl78/rl78-protos.h (rl78_split_movsi): Likewise.
25824         * config/rl78/rl78.c (rl78_split_movsi): Likewise.
25825         (rl78_addr_space_address_mode): Likewise.
25826         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
25827         Likewise.
25828         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
25829         (rs6000_reassociation_width): Likewise.
25830         (rs6000_invalid_binary_op): Likewise.
25831         (fusion_p9_p): Likewise.
25832         (emit_fusion_p9_load): Likewise.
25833         (emit_fusion_p9_store): Likewise.
25834         * config/visium/visium-protos.h (prepare_move_operands): Likewise.
25835         (ok_for_simple_move_operands): Likewise.
25836         (ok_for_simple_move_strict_operands): Likewise.
25837         (ok_for_simple_arith_logic_operands): Likewise.
25838         (visium_legitimize_reload_address): Likewise.
25839         (visium_select_cc_mode): Likewise.
25840         (output_cbranch): Likewise.
25841         (visium_split_double_move): Likewise.
25842         (visium_expand_copysign): Likewise.
25843         (visium_expand_int_cstore): Likewise.
25844         (visium_expand_fp_cstore): Likewise.
25845         * config/visium/visium.c (visium_pass_by_reference): Likewise.
25846         (visium_function_arg): Likewise.
25847         (visium_function_arg_advance): Likewise.
25848         (visium_libcall_value): Likewise.
25849         (visium_setup_incoming_varargs): Likewise.
25850         (visium_legitimate_constant_p): Likewise.
25851         (visium_legitimate_address_p): Likewise.
25852         (visium_legitimize_address): Likewise.
25853         (visium_secondary_reload): Likewise.
25854         (visium_register_move_cost): Likewise.
25855         (visium_memory_move_cost): Likewise.
25856         (prepare_move_operands): Likewise.
25857         (ok_for_simple_move_operands): Likewise.
25858         (ok_for_simple_move_strict_operands): Likewise.
25859         (ok_for_simple_arith_logic_operands): Likewise.
25860         (visium_function_value_1): Likewise.
25861         (rtx_ok_for_offset_p): Likewise.
25862         (visium_legitimize_reload_address): Likewise.
25863         (visium_split_double_move): Likewise.
25864         (visium_expand_copysign): Likewise.
25865         (visium_expand_int_cstore): Likewise.
25866         (visium_expand_fp_cstore): Likewise.
25867         (visium_split_cstore): Likewise.
25868         (visium_select_cc_mode): Likewise.
25869         (visium_split_cbranch): Likewise.
25870         (output_cbranch): Likewise.
25871         (visium_print_operand_address): Likewise.
25872         * expmed.c (flip_storage_order): Likewise.
25873         * expmed.h (emit_cstore): Likewise.
25874         (flip_storage_order): Likewise.
25875         * genrecog.c (validate_pattern): Likewise.
25876         * hsa-gen.c (gen_hsa_addr): Likewise.
25877         * internal-fn.c (expand_arith_overflow): Likewise.
25878         * ira-color.c (allocno_copy_cost_saving): Likewise.
25879         * lra-assigns.c (find_hard_regno_for_1): Likewise.
25880         * lra-constraints.c (prohibited_class_reg_set_mode_p): Likewise.
25881         (process_invariant_for_inheritance): Likewise.
25882         * lra-eliminations.c (move_plus_up): Likewise.
25883         * omp-low.c (lower_oacc_reductions): Likewise.
25884         * simplify-rtx.c (simplify_subreg): Likewise.
25885         * target.def (TARGET_SETUP_INCOMING_VARARG_BOUNDS): Likewise.
25886         (TARGET_CHKP_BOUND_MODE): Likewise..
25887         * targhooks.c (default_chkp_bound_mode): Likewise.
25888         (default_setup_incoming_vararg_bounds): Likewise.
25889         * targhooks.h (default_chkp_bound_mode): Likewise.
25890         (default_setup_incoming_vararg_bounds): Likewise.
25891         * tree-ssa-math-opts.c (divmod_candidate_p): Likewise.
25892         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
25893         (have_whole_vector_shift): Likewise.
25894         * tree-vect-stmts.c (vectorizable_load): Likewise.
25895         * doc/tm.texi: Regenerate.
25897 2017-07-05  Georg-Johann Lay  <avr@gjlay.de>
25899         Graceful degrade if Binutils PR21472 is not available.
25901         PR target/81072
25902         * configure.ac [target=avr]: WARN instead of ERROR if avrxmega3
25903         .rodata in flash test fails.
25904         (HAVE_LD_AVR_AVRXMEGA3_RODATA_IN_FLASH): Define it if test passes.
25905         * confgure: Regenerate.
25906         * config.in: Regenerate.
25907         * config/avr/avr.c (avr_asm_named_section)
25908         [HAVE_LD_AVR_AVRXMEGA3_RODATA_IN_FLASH]: Only trigger
25909         __do_copy_data for stuff in .rodata if flash_pm_offset = 0.
25910         (avr_asm_init_sections): Same.
25912 2017-07-05  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
25914         * config/arm/neon.md (fma<VCVTF:mode>4): Remove %?.
25915         (fma<VH:mode>4_intrinsic): Likewise.
25916         (*fmsub<VCVTF:mode>4): Likewise.
25917         (*fmsub<VH:mode>4_intrinsic): Likewise.
25919 2017-07-05  Georg-Johann Lay  <avr@gjlay.de>
25921         PR target/81305
25922         * config/avr/avr.c (avr_out_movhi_mr_r_xmega) [CONSTANT_ADDRESS_P]:
25923         Don't depend on "optimize > 0".
25924         (out_movhi_r_mr, out_movqi_mr_r): Same.
25925         (out_movhi_mr_r, out_movqi_r_mr): Same.
25926         (avr_address_cost) [CONSTANT_ADDRESS_P]: Don't depend cost for
25927         io_address_operand on "optimize > 0".
25929 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25931         * tree-loop-distribution.c: Add general explanantion on the pass.
25932         (generate_loops_for_partition): Mark distributed loop.
25933         (pg_add_dependence_edges): New parameter.  Handle alias data
25934         dependence specially and record it in the parameter if asked.
25935         (struct pg_vdata, pg_edata, pg_edge_callback_data): New structs.
25936         (init_partition_graph_vertices, add_partition_graph_edge): New.
25937         (pg_skip_alias_edge, free_partition_graph_edata_cb): New.
25938         (free_partition_graph_vdata, build_partition_graph): New.
25939         (sort_partitions_by_post_order, merge_dep_scc_partitions): New.
25940         (pg_collect_alias_ddrs, break_alias_scc_partitions): New.
25941         (data_ref_segment_size, latch_dominated_by_data_ref): New.
25942         (compute_alias_check_pairs, version_loop_by_alias_check): New.
25943         (version_for_distribution_p, finalize_partitions): New.
25944         (distribute_loop): Handle alias data dependence specially.  Factor
25945         out loop fusion code as functions and call these functions.
25947 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25949         * tree-loop-distribution.c (classify_partition): New parameter and
25950         better handle reduction statement.
25951         (rdg_build_partitions): Revise comment.
25952         (distribute_loop): Compute statements in all partitions and pass it
25953         to classify_partition.
25955 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25957         * tree-loop-distribution.c (enum partition_type): New.
25958         (struct partition): New field type.
25959         (partition_merge_into): Add parameter.  Update partition type.
25960         (data_dep_in_cycle_p, update_type_for_merge): New functions.
25961         (build_rdg_partition_for_vertex): Compute partition type.
25962         (rdg_build_partitions): Dump partition type.
25963         (distribute_loop): Update calls to partition_merge_into.
25965 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25967         * tree-loop-distribution.c (struct ddr_hasher): New.
25968         (ddr_hasher::hash, ::equal, get_data_dependence): New function.
25969         (ddrs_table): New.
25970         (classify_partition): Call get_data_dependence.
25971         (pg_add_dependence_edges): Ditto.
25972         (distribute_loop): Release data dependence hash table.
25974 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25976         * tree-loop-distribution.c (ref_base_address): Delete.
25977         (similar_memory_accesses): Rename ...
25978         (share_memory_accesses): ... to this.  Check if partitions access
25979         the same memory reference.
25980         (distribute_loop): Call share_memory_accesses.
25982 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25984         * tree-loop-distribution.c (struct partition): New field recording
25985         its data reference.
25986         (partition_alloc, partition_free): Init and release data refs.
25987         (partition_merge_into): Merge data refs.
25988         (build_rdg_partition_for_vertex): Collect data refs for partition.
25989         (pg_add_dependence_edges): Change parameters from vector to bitmap.
25990         Update uses.
25991         (distribute_loop): Remve data refs from vertice data of partition
25992         graph.
25994 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25996         * tree-loop-distribution.c (params.h): Include header file.
25997         (MAX_DATAREFS_NUM, DR_INDEX): New macro.
25998         (datarefs_vec): New global var.
25999         (create_rdg_vertices): Use datarefs_vec directly.
26000         (free_rdg): Don't free data references.
26001         (build_rdg): Update use.  Don't free data references.
26002         (distribute_loop): Compute global variable for data references.
26003         Bail out if there are too many data references.
26005 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
26007         * tree-loop-distribution.c (loop_nest): New global var.
26008         (build_rdg): Use loop directly, rather than loop nest.
26009         (pg_add_dependence_edges): Remove loop nest parameter.  Use global
26010         variable directly.
26011         (distribute_loop): Compute global variable loop nest.  Update use.
26013 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
26015         * tree-loop-distribution.c (enum fuse_type, fuse_message): New.
26016         (partition_merge_into): New parameter.  Dump reason for fusion.
26017         (distribute_loop): Update use of partition_merge_into.
26019 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
26021         * tree-loop-distribution.c (bb_top_order_index): New.
26022         (bb_top_order_index_size, bb_top_order_cmp): New.
26023         (stmts_from_loop): Use topological order.
26024         (pass_loop_distribution::execute): Compute and release topological
26025         order for basic blocks.
26027 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
26029         * tree-loop-distribution.c (pass_loop_distribution::execute): Skip
26030         if no loops.
26032 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
26034         * cfgloop.h (struct loop): Add comment.  New field orig_loop_num.
26035         * cfgloopmanip.c (lv_adjust_loop_entry_edge): Comment change.
26036         * internal-fn.c (expand_LOOP_DIST_ALIAS): New function.
26037         * internal-fn.def (LOOP_DIST_ALIAS): New.
26038         * tree-vectorizer.c (fold_loop_vectorized_call): Rename to ...
26039         (fold_loop_internal_call): ... this.
26040         (vect_loop_dist_alias_call): New function.
26041         (set_uid_loop_bbs): Call fold_loop_internal_call.
26042         (vectorize_loops): Fold IFN_LOOP_VECTORIZED and IFN_LOOP_DIST_ALIAS
26043         internal calls.
26045 2017-07-04  Uros Bizjak  <ubizjak@gmail.com>
26047         PR target/81300
26048         * config/i386/i386.md (setcc + movzbl/and to xor + setcc peepholes):
26049         Require dead FLAGS_REG at the beginning of a peephole.
26051 2017-07-04  Uros Bizjak  <ubizjak@gmail.com>
26053         PR target/81294
26054         * config/i386/adxintrin.h (_subborrow_u32): Swap _X and _Y
26055         arguments in the call to __builtin_ia32_sbb_u32.
26056         (_subborrow_u64): Swap _X and _Y arguments in the call to
26057         __builtin_ia32_sbb_u64.
26059 2017-07-04  Jakub Jelinek  <jakub@redhat.com>
26061         PR debug/81278
26062         * tree-vrp.c (compare_assert_loc): Turn into a function template
26063         with stable template parameter.  Only test if a->e is NULL,
26064         !a->e == !b->e has been verified already.  Use e == NULL or
26065         e != NULL instead of e or ! e tests.  If stable is true, don't use
26066         iterative_hash_expr, on the other side allow a or b or both NULL
26067         and sort the NULLs last.
26068         (process_assert_insertions): Sort using compare_assert_loc<false>
26069         instead of compare_assert_loc, later sort using
26070         compare_assert_loc<true> before calling process_assert_insertions_for
26071         in a loop.  Use break instead of continue once seen NULL pointer.
26073 2017-07-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
26075         * config/arm/driver-arm.c (arm_cpu_table): Add entry for ARM
26076         Cortex-R7 and Cortex-R8 processors.
26078 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
26080         * ipa-utils.c (ipa_merge_profiles): Fix merging when dst is
26081         uninitialized while src is not.
26083 2017-07-04  Richard Earnshaw  <rearnsha@arm.com>
26085         * common/config/arm/arm-common.c: Adjust include path for
26086         arm-cpu-cdata.h
26087         * config/arm/t-arm (TM_H): Adjust path for arm-cpu.h.
26088         (arm-cpu.h): Create in build directory.  Adjust dependency rules.
26089         (arm-cpu-data.h): Likewise.
26090         (arm-cpu-cdata.h): Likewise.
26091         * config/arm/arm-cpu.h: Delete.
26092         * config/arm/arm-cpu-cdata.h: Delete.
26093         * config/arm/arm-cpu-data.h: Delete.
26095 2017-07-04  James Greenhalgh  <james.greenhalgh@arm.com>
26097         * config/arm/arm-cpus.in (cortex-a55): New.
26098         (cortex-a75): Likewise.
26099         (cortex-a75.cortex-a55): Likewise.
26100         * config/arm/driver-arm.c (arm_cpu_table): Add cortex-a55 and
26101         cortex-a75.
26102         * doc/invoke.texi (-mcpu): Document cortex-a55 and cortex-a75.
26103         * config/arm/arm-cpu-cdata.h: Regenerate.
26104         * config/arm/arm-cpu-data.h: Regenerate.
26105         * config/arm/arm-cpu.h: Regenerate.
26106         * config/arm/arm-tables.opt: Regenerate.
26107         * config/arm/arm-tune.md: Regenerate.
26109 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
26111         * haifa-sched.c (sched_create_recovery_edges): Update profile.
26113 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
26115         * bb-reorder.c (better_edge_p): Fix handling of uninitialized
26116         probability.
26118 2017-07-04  Richard Sandiford  <richard.sandiford@linaro.org>
26120         PR tree-optimization/81292
26121         * tree-ssa-strlen.c (handle_builtin_strlen): When setting
26122         full_string_p, also call adjust_related_strinfos if the adjustment
26123         is simple, otherwise invalidate related strinfos.
26125 2017-07-04  Martin Liska  <mliska@suse.cz>
26127         PR sanitizer/81040
26128         * sanopt.c (sanitize_rewrite_addressable_params): Mark the
26129         newly created variable as DECL_IGNORED_P.
26131 2017-07-04  Martin Liska  <mliska@suse.cz>
26133         PR ipa/81293
26134         * ipa-inline.c (inline_small_functions):
26135         Use xstrdup_for_dump.
26137 2017-07-04  Tom de Vries  <tom@codesourcery.com>
26139         * graph.c (draw_cfg_edges): Save and restore EDGE_DFS_BACK.
26141 2017-07-03  Dominique d'Humieres  <dominiq@lps.ens.fr>
26143         PR target/81033
26144         * config/darwin.c (darwin_function_switched_text_sections):
26145         Fix spaces.
26147 2017-07-03  Jan Hubicka  <hubicka@ucw.cz>
26149         * tree-vect-loop-manip.c (vect_do_peeling): Fix scaling up.
26151 2017-07-03  Richard Earnshaw  <rearnsha@arm.com>
26153         * doc/invoke.texi (ARM Options): Add -mbe8 and -mbe32 to option summary.
26155 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26157         * tree-vect-loop.c (vect_analyze_loop_2): Treat min_scalar_loop_bound,
26158         min_profitable_iters, and th as inclusive lower bounds.
26159         Fix LOOP_VINFO_PEELING_FOR_GAPS condition.
26160         (vect_estimate_min_profitable_iters): Return inclusive lower bounds
26161         for min_profitable_iters and min_profitable_estimate.
26162         (vect_transform_loop): Treat th as an inclusive lower bound.
26163         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
26165 2017-07-03  Dominique d'Humieres  <dominiq@lps.ens.fr>
26167         PR target/81033
26168         * config/darwin.c (darwin_function_switched_text_sections):
26169         Replace DECL_NAME with DECL_ASSEMBLER_NAME, split assemble_name_raw
26170         in two pieces, and suppress the use of buf.
26172 2017-07-03  Nathan Sidwell  <nathan@acm.org>
26174         * hash-table.h (hash_table_mod1): Fix indentation.
26176 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26178         PR middle-end/81290
26179         * predict.c (force_edge_cold): Be more careful about propagation
26180         backward.
26181         * profile-count.h (profile_probability::guessed,
26182         profile_probability::fdo, profile_count::guessed, profile_count::fdo):
26183         New.
26184         * tree-ssa-threadupdate.c (recompute_probabilities): Result is guessed.
26186 2017-07-03  James Greenhalgh  <james.greenhalgh@arm.com>
26188         * doc/invoke.texi (rcpc architecture extension): Document it.
26190 2017-07-03  Richard Biener  <rguenther@suse.de>
26192         PR tree-optimization/60510
26193         * tree-vect-loop.c (vect_create_epilog_for_reduction): Pass in
26194         the scalar reduction PHI and use it.
26195         (vectorizable_reduction): Properly guard the single_defuse_cycle
26196         path for non-SLP reduction chains where we cannot use it.
26197         Rework reduc_def/index and vector type deduction.  Rework
26198         vector operand gathering during reduction op code-gen.
26199         * tree-vect-slp.c (vect_analyze_slp): For failed SLP reduction
26200         chains dissolve the chain and leave it to non-SLP reduction
26201         handling.
26203 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26205         * tree-data-ref.h (dr_alignment): Declare.
26206         * tree-data-ref.c (dr_alignment): New function.
26207         * tree-vectorizer.h (dataref_aux): Remove base_element_aligned.
26208         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
26209         set it.
26210         * tree-vect-stmts.c (vectorizable_store): Use dr_alignment.
26212 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26214         * tree-data-ref.h (innermost_loop_behavior): Add base_alignment
26215         and base_misalignment fields.
26216         (DR_BASE_ALIGNMENT, DR_BASE_MISALIGNMENT): New macros.
26217         * tree-data-ref.c: Include builtins.h.
26218         (dr_analyze_innermost): Set up the new innmost_loop_behavior fields.
26219         * tree-vectorizer.h (STMT_VINFO_DR_BASE_ALIGNMENT): New macro.
26220         (STMT_VINFO_DR_BASE_MISALIGNMENT): Likewise.
26221         * tree-vect-data-refs.c: Include tree-cfg.h.
26222         (vect_compute_data_ref_alignment): Use the new innermost_loop_behavior
26223         fields instead of calculating an alignment here.
26224         (vect_analyze_data_refs): Use dr_analyze_innermost.  Dump the new
26225         innermost_loop_behavior fields.
26227 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26229         * tree-data-ref.h (innermost_loop_behavior): Add a step_alignment
26230         field.
26231         (DR_STEP_ALIGNMENT): New macro.
26232         * tree-vectorizer.h (STMT_VINFO_DR_STEP_ALIGNMENT): Likewise.
26233         * tree-data-ref.c (dr_analyze_innermost): Initalize step_alignment.
26234         (create_data_ref): Print it.
26235         * tree-vect-stmts.c (vectorizable_load): Use the step alignment
26236         to tell whether the step preserves vector (mis)alignment.
26237         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Likewise.
26238         Move the check for an integer step and generalise to all INTEGER_CST.
26239         (vect_analyze_data_refs): Set DR_STEP_ALIGNMENT when setting DR_STEP.
26240         Print the outer step alignment.
26242 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26244         * tree-data-ref.h (innermost_loop_behavior): Replace aligned_to
26245         with offset_alignment.
26246         (DR_ALIGNED_TO): Delete.
26247         (DR_OFFSET_ALIGNMENT): New macro.
26248         * tree-vectorizer.h (STMT_VINFO_DR_ALIGNED_TO): Delete.
26249         (STMT_VINFO_DR_OFFSET_ALIGNMENT): New macro.
26250         * tree-data-ref.c (dr_analyze_innermost): Update after above changes.
26251         (create_data_ref): Likewise.
26252         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Likewise.
26253         (vect_analyze_data_refs): Likewise.
26254         * tree-if-conv.c (if_convertible_loop_p_1): Use memset before
26255         creating dummy innermost behavior.
26257 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26259         * tree-data-ref.h (dr_analyze_innermost): Replace the dr argument
26260         with a "innermost_loop_behavior *" and refeence tree.
26261         * tree-data-ref.c (dr_analyze_innermost): Likewise.
26262         (create_data_ref): Update call accordingly.
26263         * tree-predcom.c (find_looparound_phi): Likewise.
26265 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26267         * tree-vectorizer.h (_stmt_vec_info): Replace individual dr_*
26268         fields with dr_wrt_vec_loop.
26269         (STMT_VINFO_DR_BASE_ADDRESS, STMT_VINFO_DR_INIT, STMT_VINFO_DR_OFFSET)
26270         (STMT_VINFO_DR_STEP, STMT_VINFO_DR_ALIGNED_TO): Update accordingly.
26271         (STMT_VINFO_DR_WRT_VEC_LOOP): New macro.
26272         (vect_dr_behavior): New function.
26273         (vect_create_addr_base_for_vector_ref): Remove loop parameter.
26274         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
26275         vect_dr_behavior.  Use a step_preserves_misalignment_p boolean to
26276         track whether the step preserves the misalignment.
26277         (vect_create_addr_base_for_vector_ref): Remove loop parameter.
26278         Use vect_dr_behavior.
26279         (vect_setup_realignment): Update call accordingly.
26280         (vect_create_data_ref_ptr): Likewise.  Use vect_dr_behavior.
26281         * tree-vect-loop-manip.c (vect_gen_prolog_loop_niters): Update
26282         call to vect_create_addr_base_for_vector_ref.
26283         (vect_create_cond_for_align_checks): Likewise.
26284         * tree-vect-patterns.c (vect_recog_bool_pattern): Copy
26285         STMT_VINFO_DR_WRT_VEC_LOOP as a block.
26286         (vect_recog_mask_conversion_pattern): Likewise.
26287         * tree-vect-stmts.c (compare_step_with_zero): Use vect_dr_behavior.
26288         (new_stmt_vec_info): Remove redundant zeroing.
26290 2017-07-03  Richard Earnshaw  <rearnsha@arm.com>
26292         * common/config/arm/arm-common.c (arm_be8_option): New function.
26293         * config/arm/arm-isa.h (isa_feature): Add new feature bit isa_bit_be8.
26294         (ISA_ARMv6): Add isa_bit_be8.
26295         * config/arm/arm.h (arm_be8_option): Add prototype.
26296         (BE8_SPEC_FUNCTION): New define.
26297         (EXTRA_SPEC_FUNCTIONS): Add BE8_SPEC_FUNCTION.
26298         * config/arm/arm.opt (mbig-endian): Mark as Negative of mlittle-endian.
26299         (mlittle-endian): Similarly.
26300         (mbe8, mbe32): New options.
26301         * config/arm/bpabi.h (BE8_LINK_SPEC): Call arm_be8_option.
26302         * doc/invoke.texi (ARM Options): Document -mbe8 and -mbe32.
26304 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26306         * tree-cfg.c (gimple_find_sub_bbs): Fix profile updating.
26308 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26310         * tree-cfgcleanup.c (want_merge_blocks_p): New function.
26311         (cleanup_tree_cfg_bb): Use it.
26312         * profile-count.h (profile_count::of_for_merging, profile_count::merge):
26313         New functions.
26314         * tree-cfg.c (gimple_merge_blocks): Use profile_count::merge.
26316 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26318         PR bootstrap/81285
26319         * loop-doloop.c (add_test): Update profile.
26321 2017-07-03  Martin Liska  <mliska@suse.cz>
26323         PR sanitize/81040
26324         * sanopt.c (rewrite_usage_of_param): New function.
26325         (sanitize_rewrite_addressable_params): Likewise.
26326         (pass_sanopt::execute): Call rewrite_usage_of_param.
26328 2017-07-03  Richard Biener  <rguenther@suse.de>
26330         * tree-vect-loop.c (vect_create_epilog_for_reduction): Revert
26331         back to using VIEW_CONVERT_EXPR.
26333 2017-07-03  Martin Liska  <mliska@suse.cz>
26335         PR other/78366
26336         * doc/extend.texi: Document when a resolver function is
26337         generated for target_clones.
26339 2017-07-03  Martin Liska  <mliska@suse.cz>
26341         * asan.c (asan_emit_stack_protection): Unpoison just red zones
26342         and shadow memory of auto variables which are subject of
26343         use-after-scope sanitization.
26344         (asan_expand_mark_ifn): Add do set only when is_poison.
26346 2016-07-03  Richard Biener  <rguenther@suse.de>
26348         * tree-vect-loop.c (vect_analyze_loop_operations): Also analyze
26349         reduction PHIs.
26350         (vect_force_simple_reduction): Record reduction def -> phi mapping.
26351         (vectorizable_reduction): Perform reduction PHI creation when
26352         visiting a reduction PHI and adjust and simplify code generation
26353         phase of the reduction op.  Cache dts, use fold_binary, not fold_build2.
26354         (vect_transform_loop): Visit reduction PHIs.
26355         * tree-vect-slp.c (vect_get_and_check_slp_defs): Record reduction
26356         defs into the SLP tree.
26357         (vect_build_slp_tree): Reduction defs terminate the recursion.
26358         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Allow lookup
26359         of reduction defs.
26360         (vect_get_vec_defs_for_stmt_copy): Export.
26361         (vect_get_vec_defs): Likewise.
26362         * tree-vectorizer.h (struct _stmt_vec_info): Amend reduc_def
26363         purpose.
26364         (vect_get_vec_defs_for_stmt_copy): Declare.
26365         (vect_get_vec_defs): Likewise.
26367 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26369         * tree-data-ref.c (dr_analyze_innermost): Replace the "nest"
26370         parameter with a "loop" parameter and use it instead of the
26371         loop containing DR_STMT.  Don't check simple_iv when doing
26372         BB analysis.  Describe the two analysis modes in the comment.
26374 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26376         PR tree-optimization/69468
26377         * tree-ssa-tail-merge.c (ignore_edge_flags): New constant.
26378         (find_same_succ_bb): Handle ignore_edge_flags.
26380 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26382         PR tree-optimization/81192
26383         * tree-ssa-tail-merge.c (same_succ_hash): Use bb->loop_father->num in
26384         hash.
26385         (same_succ::equal): Don't find bbs to be equal if bb->loop_father
26386         differs.
26387         (find_same_succ_bb): Remove obsolete test on bb->loop_father->latch.
26389 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26391         PR tree-optimization/81192
26392         * tree-ssa-tail-merge.c (same_succ_flush_bb): Handle
26393         BB_SAME_SUCC (bb) == NULL.
26395 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26397         * cfgrtl.c (rtl_verify_edges): Enable checking of profile_probability
26398         consistency.
26400 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26402         * dumpfile.c: Include profile-count.h
26403         * tree-cfg.c (gimple_duplicate_sese_tail): Drop UNUSED attributes;
26404         update profile.
26405         (insert_cond_bb): Update profile.
26406         * tree-cfg.h (insert_cond_bb): Update prototype.
26407         * tree-chkp-opt.c (chkp_optimize_string_function_calls): Update.
26408         * tree-dump.c: Do not include tree-cfg.
26410 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26412         * bb-reorder.c (fix_up_crossing_landing_pad): Update profile.
26414 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26416         * expect.c (dw2_build_landing_pads): Update profile of the landing pad
26417         bb.
26419 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26421         * tree-complex.c (expand_complex_div_wide): update profile.
26423 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26424             Alan Hayward  <alan.hayward@arm.com>
26425             David Sherwood  <david.sherwood@arm.com>
26427         * Makefile.in (MACHMODE_H): Remove insn-modes.h
26428         (CORETYPES_H): New define.
26429         (MOSTLYCLEANFILES): Add insn-modes-inline.h.
26430         (insn-modes-inline.h, s-modes-inline-h): New rules.
26431         (generated_files): Add insn-modes-inline.h.
26432         (RTL_BASE_H, TREE_CORE_H): Use CORETYPES_H instead of coretypes.h.
26433         (build/gensupport.o, build/ggc-none.o, build/print-rtl.o): Likewise.
26434         (build/read-md.o, build/read-rtl.o, build/rtl.o): Likewise.
26435         (build/vec.o, build/hash-table.o, build/inchash.o): Likewise.
26436         (build/gencondmd.o, build/genattr.o, build/genattr-common.o): Likewise.
26437         (build/genattrtab.o, build/genautomata.o, build/gencheck.o): Likewise.
26438         (build/gencodes.o, build/genconditions.o): Likewise.
26439         (build/genconfig.o, build/genconstants.o, build/genemit.o): Likewise.
26440         (build/genenums.o, build/genextract.o, build/genflags.o): Likewise.
26441         (build/gentarget-def.o, build/genmddeps.o, build/genopinit.o)
26442         (build/genoutput.o, build/genpeep.o, build/genpreds.o): Likewise.
26443         (build/genrecog.o, build/genmddump.o, build/genmatch.o): Likewise.
26444         (build/gencfn-macros.o, build/gcov-iov.o): Likewise.
26445         * coretypes.h: Include everything up to real.h for generators.
26446         Include insn-modes.h first.  Include wide-int-print.h after
26447         wide-int.h.  Include insn-modes-inline.h and then machmode.h.
26448         * machmode.h: Don't include insn-modes.h here.
26449         * function-tests.c: Remove includes of signop.h, machmode.h,
26450         double-int.h and wide-int.h.
26451         * rtl.h: Likewise.
26452         * gcc-rich-location.c: Remove includes of machmode.h, double-int.h
26453         and wide-int.h.
26454         * optc-save-gen.awk: Likewise.
26455         * gencheck.c (BITS_PER_UNIT): Delete dummy definition.
26456         * godump.c: Remove include of wide-int-print.h.
26457         * pretty-print.h: Likewise.
26458         * wide-int-print.cc: Likewise.
26459         * wide-int.cc: Likewise.
26460         * hash-map-tests.c: Remove include of signop.h.
26461         * hash-set-tests.c: Likewise.
26462         * rtl-tests.c: Likewise.
26463         * mkconfig.sh: Remove include of machmode.h.
26464         * genmodes.c (emit_insn_modes_h): Split emission of inline functions
26465         into...
26466         (emit_insn_modes_inline_h): ...this new function.  Emit the code
26467         into an insn-modes-inline.h header file, adding appropriate
26468         include guards and end comments.
26469         (emit_insn_modes_c_header): Remove include of machmode.h.
26470         (emit_min_insn_modes_c_header): Include coretypes.h rather than
26471         machmode.h.
26472         (main): Handle -i flag and call emit_insn_modes_inline_h when
26473         it is passed.
26475 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26477         * tree-ssa-strlen.c (strinfo): Rename the length field to
26478         nonzero_chars.  Add a full_string_p field.
26479         (compare_nonzero_chars, zero_length_string_p): New functions.
26480         (get_addr_stridx): Add an offset_out parameter.
26481         Use compare_nonzero_chars.
26482         (get_stridx): Update accordingly.  Use compare_nonzero_chars.
26483         (new_strinfo): Update after above changes to strinfo.
26484         (set_endptr_and_length): Set full_string_p.
26485         (get_string_length): Update after above changes to strinfo.
26486         (unshare_strinfo): Update call to new_strinfo.
26487         (maybe_invalidate): Likewise.
26488         (get_stridx_plus_constant): Change off to unsigned HOST_WIDE_INT.
26489         Use compare_nonzero_chars and zero_string_p.  Treat nonzero_chars
26490         as a uhwi instead of an shwi.  Update after above changes to
26491         strinfo and new_strinfo.
26492         (zero_length_string): Assert that chainsi contains full strings.
26493         Use zero_length_string_p.  Update call to new_strinfo.
26494         (adjust_related_strinfos): Update after above changes to strinfo.
26495         Copy full_string_p from origsi.
26496         (adjust_last_stmt): Use zero_length_string_p.
26497         (handle_builtin_strlen): Update after above changes to strinfo and
26498         new_strinfo.  Install the lhs as the string length if the previous
26499         entry didn't describe a full string.
26500         (handle_builtin_strchr): Update after above changes to strinfo
26501         and new_strinfo.
26502         (handle_builtin_strcpy): Likewise.
26503         (handle_builtin_strcat): Likewise.
26504         (handle_builtin_malloc): Likewise.
26505         (handle_pointer_plus): Likewise.
26506         (handle_builtin_memcpy): Likewise.  Track nonzero characters
26507         that aren't necessarily followed by a nul terminator.
26508         (handle_char_store): Likewise.
26510 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26512         PR tree-optimization/80769
26513         * tree-ssa-strlen.c (strinfo): Document that "stmt" is also used
26514         for malloc and calloc.  Document the new invariant that all related
26515         strinfos have delayed lengths or none do.
26516         (verify_related_strinfos): Move earlier in file.
26517         (set_endptr_and_length): New function, split out from...
26518         (get_string_length): ...here.  Also set the lengths of related
26519         strinfos.
26520         (zero_length_string): Assert that chainsi has known (rather than
26521         delayed) lengths.
26522         (adjust_related_strinfos): Likewise.
26524 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26526         PR tree-optimization/81136
26527         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Only
26528         assert that two references with the same misalignment have the same
26529         compile-time misalignment if those compile-time misalignments
26530         are known.
26532 2017-07-01  Andi Kleen  <ak@linux.intel.com>
26534         * print-tree.c (print_node): Print all attributes.
26536 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26538         * cfg.c (scale_bbs_frequencies): New function.
26539         * cfg.h (scale_bbs_frequencies): Declare it.
26540         * cfgloopanal.c (single_likely_exit): Cleanup.
26541         * cfgloopmanip.c (scale_loop_frequencies): Take profile_probability
26542         as parameter.
26543         (scale_loop_profile): Likewise.
26544         (loop_version): Likewise.
26545         (create_empty_loop_on_edge): Update.
26546         * cfgloopmanip.h (scale_loop_frequencies, scale_loop_profile,
26547         scale_loop_frequencies, scale_loop_profile, loopify,
26548         loop_version): Update prototypes.
26549         * modulo-sched.c (sms_schedule): Update.
26550         * predict.c (unlikely_executed_edge_p): Also check probability.
26551         (probably_never_executed_edge_p): Fix typo.
26552         * tree-if-conv.c (version_loop_for_if_conversion): Update.
26553         * tree-parloops.c (gen_parallel_loop): Update.
26554         * tree-ssa-loop-ivcanon.c (try_peel_loop): Update.
26555         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
26556         * tree-ssa-loop-split.c (split_loop): Update.
26557         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Update.
26558         * tree-vect-loop-manip.c (vect_do_peeling): Update.
26559         (vect_loop_versioning): Update.
26560         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
26562 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26564         * trans-mem.c (split_bb_make_tm_edge): Update profile.
26566 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26568         * tree-if-conv.c (combine_blocks): Use make_single_succ_edge
26569         to keep profile consistent.
26571 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26573         * cfgrtl.c (rtl_flow_call_edges_add): Update profile.
26574         * tree-cfg.c (gimple_flow_call_edges_add): Likewise.
26575         * profile-count.h (max_safe_multiplier): Make unsigned.
26576         (profile_count::guessed_zero): New.
26578 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26580         * bb-reorder.c (fix_up_crossing_landing_pad,
26581         fix_crossing_conditional_branches): Use make_single_succ_edge
26582         to keep profile consistent.
26584 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26586         * tree-vect-loop.c (optimize_mask_stores): Use make_single_succ_edge
26587         to update profile.
26589 2017-07-01  Jakub Jelinek  <jakub@redhat.com>
26591         PR sanitizer/81262
26592         * bb-reorder.c (fix_up_fall_thru_edges): Move variable declarations to
26593         the right scopes, make sure cond_jump isn't preserved between multiple
26594         iterations.  Search for fallthru edge whenever there are 3+ edges and
26595         use find_fallthru_edge for it.
26597 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26599         Patch by Alexander Monakov <amonakov@ispras.ru>
26600         * sel-sched-ir.c (compute_succs_info): Handle uninitialized
26601         probabilities consistently.
26603 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26605         * pa.c (pa_expand_compare_and_swap_loop): Update call of
26606         emit_cmp_and_jump_insns.
26608 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26610         PR ipa/81261
26611         * tree-inline.c (expand_call_inline): Combine profile statuses.
26613 2017-06-30  Andrew Pinski  <apinski@cavium.com>
26615         * tree-if-conv.c (predicate_scalar_phi): Update new_stmt if
26616         fold_stmt returned true.
26618 2017-06-30  Nathan Sidwell  <nathan@acm.org>
26620         * ggc.h (empty_string): Delete.
26621         * cfgexpand.c (expand_asm_stmt): Use plain "".
26622         * optabs.c (expand_asm_memory_barrier): Likewise.
26623         * stringpool.c (empty_string): Delete.
26624         (digit_vector, digit_string): Delete.
26625         (ggc_alloc_string): Use plain "", don't optimize single digit
26626         strings.  Use ggc_alloc_atomic.
26628 2017-06-30  Richard Earnshaw  <rearnsha@arm.com>
26630         * rtlanal.c (insn_rtx_cost): If a parallel contains exactly one
26631         comparison set and one other set, use the cost of the non-comparison
26632         set.
26634 2017-06-30  Nathan Sidwell  <nathan@acm.org>
26636         * ggc.h: Replace all 'static inline' with plain 'inline'.  Fix
26637         some formatting.
26639 2017-06-30  Peter Bergner  <bergner@vnet.ibm.com>
26641         * tree-cfg.c (group_case_labels_stmt): Merge scanning and compressing
26642         loops.  Remove now unneeded calls to gimple_switch_set_label() that
26643         just set removed labels to NULL_TREE.
26645 2017-06-30  Aldy Hernandez  <aldyh@redhat.com>
26647         * tree-ssanames.c (set_range_info_raw): Abstract from ...
26648         (set_range_info): ...here.  Only call set_range_info_raw if domain
26649         is useful.
26650         (set_nonzero_bits): Call set_range_info_raw.
26651         * tree-ssanames.h (set_range_info_raw): New.
26653 2017-06-30  Jakub Jelinek  <jakub@redhat.com>
26655         PR target/81225
26656         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): For V8FI,
26657         V16FI and VI8F_256 iterators, use <store_mask_predicate> instead
26658         of nonimmediate_operand and <store_mask_constraint> instead of m
26659         for the input operand.  For V8FI iterator, always split if input
26660         is a MEM.  For V16FI and V8SF_256 iterators, don't test if both
26661         operands are MEM if <mask_applied>.  For VI4F_256 iterator, use
26662         <store_mask_predicate> instead of register_operand and
26663         <store_mask_constraint> instead of v for the input operand.  Make
26664         sure both operands aren't MEMs for if not <mask_applied>.
26666 2017-06-30  Sylvestre Ledru  <sylvestre@debian.org>
26668         * lto-wrapper.c (copy_file) Close both file descriptors before
26669         exiting normally.
26671 2017-06-30  Martin Liska  <mliska@suse.cz>
26673         PR ipa/81214
26674         * multiple_target.c (create_dispatcher_calls): Make ifunc
26675         also for function that don't have calls or are not referenced.
26677 2017-06-30  Richard Biener  <rguenther@suse.de>
26679         * tree-vect-slp.c (vect_slp_analyze_node_operations): Only
26680         analyze the first scalar stmt.  Move vector type computation
26681         for the BB case here from ...
26682         * tree-vect-stmts.c (vect_analyze_stmt): ... here.  Guard
26683         live operation processing in the SLP case properly.
26685 2017-06-30  Richard Biener  <rguenther@suse.de>
26687         * graph.c (draw_cfg_node_succ_edges): Fix broken dot syntax.
26689 2017-06-30  Martin Liska  <mliska@suse.cz>
26691         PR sanitizer/81021
26692         * tree-eh.c (lower_resx): Call BUILT_IN_ASAN_HANDLE_NO_RETURN
26693         before BUILT_IN_UNWIND_RESUME when ASAN is used.
26695 2017-06-30  Yvan Roux  <yvan.roux@linaro.org>
26697         * doc/invoke.texi (AArch64): Add missing options and remove redundant
26698         ones.
26700 2017-06-30  Richard Biener  <rguenther@suse.de>
26702         PR tree-optimization/81249
26703         * tree-vect-loop.c (vect_create_epilog_for_reduction): Convert
26704         condition reduction result to original scalar type.
26706 2017-06-30  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26708         * profile-count.h (enum profile_quality): Fix typos and whitespace
26709         issues.
26711 2017-06-30  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26713         * config/s390/s390.c (s390_expand_setmem): Adjust to the new data
26714         type for branch probabilities.
26716 2017-06-29  Julian Brown  <julian@codesourcery.com>
26717             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
26719         * config/aarch64/aarch64-fusion-pairs.def: Add ALU_BRANCH entry.
26720         * config/aarch64/aarch64.c (AARCH64_FUSE_ALU_BRANCH): New fusion type.
26721         (thunderx2t99_tunings): Set AARCH64_FUSE_ALU_BRANCH flag.
26722         (aarch_macro_fusion_pair_p): Add support for AARCH64_FUSE_ALU_BRANCH.
26724 2017-06-29  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
26726         * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Push the
26727         check for CC usage into AARCH64_FUSE_CMP_BRANCH.
26728         * config/i386/i386.c (ix86_macro_fusion_pair_p): Push the check for
26729         CC usage from generic code to here.
26730         * sched-deps.c (sched_macro_fuse_insns): Move the condition for
26731         CC usage into the target macros.
26733 2017-06-29  Maya Rashish  <coypu@sdf.org>
26735         * config/netbsd.h (NETBSD_LIB_SPEC): Add -lc when creating shared
26736         objects.
26738 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26740         * arm/arm-builtins.c: Include profile-count.h
26741         * except.c (sjlj_emit_function_enter): Use
26742         profile_probability::unlikely.
26744 2017-06-29  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
26746         * config/rs6000/rs6000.c (toc_relative_expr_p): Make tocrel_base
26747         and tocrel_offset be pointer args rather than implicitly using
26748         static versions.
26749         (legitimate_constant_pool_address_p, rs6000_emit_move,
26750         const_load_sequence_p, adjust_vperm): Add local tocrel_base and
26751         tocrel_offset and use in toc_relative_expr_p call.
26752         (print_operand, print_operand_address): Use static tocrel_base_oac
26753         and tocrel_offset_oac.
26754         (rs6000_output_addr_const_extra): Use static tocrel_base_oac and
26755         tocrel_offset_oac.
26757 2017-06-29  Maya Rashish  <coypu@sdf.org>
26759         * config/vax/builtins.md (ffssi2_internal): Correct constraint.
26761 2017-06-29  Eric Botcazou  <ebotcazou@adacore.com>
26763         * expr.c (expand_expr) <normal_inner_ref>: When testing for unaligned
26764         objects, take into account only the alignment of 'op0' and 'mode1' if
26765         'op0' is a MEM.
26767 2017-06-29  Steve Ellcey  <sellcey@cavium.com>
26769         * ccmp.c (ccmp_tree_comparison_p): New function.
26770         (ccmp_candidate_p): Update to use above function.
26771         (get_compare_parts): New function.
26772         (expand_ccmp_next): Update to use new functions.
26773         (expand_ccmp_expr_1): Take tree arg instead of gimple, update to use
26774         new functions.
26775         (expand_ccmp_expr): Pass tree instead of gimple to expand_ccmp_expr_1,
26776         take mode as argument.
26777         * ccmp.h (expand_ccmp_expr): Add mode as argument.
26778         * expr.c (expand_expr_real_1): Pass mode as argument.
26780 2017-06-29  Segher Boessenkool  <segher@kernel.crashing.org>
26782         * combine.c (combine_instructions): Print insns to dump_file, together
26783         with their costs.
26785 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26787         * asan.c (asan_emit_stack_protection): Update.
26788         (create_cond_insert_point): Update.
26789         * auto-profile.c (afdo_propagate_circuit): Update.
26790         * basic-block.h (struct edge_def): Turn probability to
26791         profile_probability.
26792         (EDGE_FREQUENCY): Update.
26793         * bb-reorder.c (find_traces_1_round): Update.
26794         (better_edge_p): Update.
26795         (sanitize_hot_paths): Update.
26796         * cfg.c (unchecked_make_edge): Initialize probability to uninitialized.
26797         (make_single_succ_edge): Update.
26798         (check_bb_profile): Update.
26799         (dump_edge_info): Update.
26800         (update_bb_profile_for_threading): Update.
26801         * cfganal.c (connect_infinite_loops_to_exit): Initialize new edge
26802         probabilitycount to 0.
26803         * cfgbuild.c (compute_outgoing_frequencies): Update.
26804         * cfgcleanup.c (try_forward_edges): Update.
26805         (outgoing_edges_match): Update.
26806         (try_crossjump_to_edge): Update.
26807         * cfgexpand.c (expand_gimple_cond): Update make_single_succ_edge.
26808         (expand_gimple_tailcall): Update.
26809         (construct_init_block): Use make_single_succ_edge.
26810         (construct_exit_block): Use make_single_succ_edge.
26811         * cfghooks.c (verify_flow_info): Update.
26812         (redirect_edge_succ_nodup): Update.
26813         (split_edge): Update.
26814         (account_profile_record): Update.
26815         * cfgloopanal.c (single_likely_exit): Update.
26816         * cfgloopmanip.c (scale_loop_profile): Update.
26817         (set_zero_probability): Remove.
26818         (duplicate_loop_to_header_edge): Update.
26819         * cfgloopmanip.h (loop_version): Update prototype.
26820         * cfgrtl.c (try_redirect_by_replacing_jump): Update.
26821         (force_nonfallthru_and_redirect): Update.
26822         (update_br_prob_note): Update.
26823         (rtl_verify_edges): Update.
26824         (purge_dead_edges): Update.
26825         (rtl_lv_add_condition_to_bb): Update.
26826         * cgraph.c: (cgraph_edge::redirect_call_stmt_to_calle): Update.
26827         * cgraphunit.c (init_lowered_empty_function): Update.
26828         (cgraph_node::expand_thunk): Update.
26829         * cilk-common.c: Include profile-count.h
26830         * dojump.c (inv): Remove.
26831         (jumpifnot): Update.
26832         (jumpifnot_1): Update.
26833         (do_jump_1): Update.
26834         (do_jump): Update.
26835         (do_jump_by_parts_greater_rtx): Update.
26836         (do_compare_rtx_and_jump): Update.
26837         * dojump.h (jumpifnot, jumpifnot_1, jumpif_1, jumpif, do_jump,
26838         do_jump_1. do_compare_rtx_and_jump): Update prototype.
26839         * dwarf2cfi.c: Include profile-count.h
26840         * except.c (dw2_build_landing_pads): Use make_single_succ_edge.
26841         (sjlj_emit_dispatch_table): Likewise.
26842         * explow.c: Include profile-count.h
26843         * expmed.c (emit_store_flag_force): Update.
26844         (do_cmp_and_jump): Update.
26845         * expr.c (compare_by_pieces_d::generate): Update.
26846         (compare_by_pieces_d::finish_mode): Update.
26847         (emit_block_move_via_loop): Update.
26848         (store_expr_with_bounds): Update.
26849         (store_constructor): Update.
26850         (expand_expr_real_2): Update.
26851         (expand_expr_real_1): Update.
26852         * expr.h (try_casesi, try_tablejump): Update prototypes.
26853         * gimple-pretty-print.c (dump_probability): Update.
26854         (dump_profile): New.
26855         (dump_gimple_label): Update.
26856         (dump_gimple_bb_header): Update.
26857         * graph.c (draw_cfg_node_succ_edges): Update.
26858         * hsa-gen.c (convert_switch_statements): Update.
26859         * ifcvt.c (cheap_bb_rtx_cost_p): Update.
26860         (find_if_case_1): Update.
26861         (find_if_case_2): Update.
26862         * internal-fn.c (expand_arith_overflow_result_store): Update.
26863         (expand_addsub_overflow): Update.
26864         (expand_neg_overflow): Update.
26865         (expand_mul_overflow): Update.
26866         (expand_vector_ubsan_overflow): Update.
26867         * ipa-cp.c (good_cloning_opportunity_p): Update.
26868         * ipa-split.c (split_function): Use make_single_succ_edge.
26869         * ipa-utils.c (ipa_merge_profiles): Update.
26870         * loop-doloop.c (add_test): Update.
26871         (doloop_modify): Update.
26872         * loop-unroll.c (compare_and_jump_seq): Update.
26873         (unroll_loop_runtime_iterations): Update.
26874         * lra-constraints.c (lra_inheritance): Update.
26875         * lto-streamer-in.c (input_cfg): Update.
26876         * lto-streamer-out.c (output_cfg): Update.
26877         * mcf.c (adjust_cfg_counts): Update.
26878         * modulo-sched.c (sms_schedule): Update.
26879         * omp-expand.c (expand_omp_for_init_counts): Update.
26880         (extract_omp_for_update_vars): Update.
26881         (expand_omp_ordered_sink): Update.
26882         (expand_omp_for_ordered_loops): Update.
26883         (expand_omp_for_generic): Update.
26884         (expand_omp_for_static_nochunk): Update.
26885         (expand_omp_for_static_chunk): Update.
26886         (expand_cilk_for): Update.
26887         (expand_omp_simd): Update.
26888         (expand_omp_taskloop_for_outer): Update.
26889         (expand_omp_taskloop_for_inner): Update.
26890         * omp-simd-clone.c (simd_clone_adjust): Update.
26891         * optabs.c (expand_doubleword_shift): Update.
26892         (expand_abs): Update.
26893         (emit_cmp_and_jump_insn_1): Update.
26894         (expand_compare_and_swap_loop): Update.
26895         * optabs.h (emit_cmp_and_jump_insns): Update prototype.
26896         * predict.c (predictable_edge_p): Update.
26897         (edge_probability_reliable_p): Update.
26898         (set_even_probabilities): Update.
26899         (combine_predictions_for_insn): Update.
26900         (combine_predictions_for_bb): Update.
26901         (propagate_freq): Update.
26902         (estimate_bb_frequencies): Update.
26903         (force_edge_cold): Update.
26904         * profile-count.c (profile_count::dump): Add missing space into dump.
26905         (profile_count::debug): Add newline.
26906         (profile_count::differs_from_p): Explicitly convert to unsigned.
26907         (profile_count::stream_in): Update.
26908         (profile_probability::dump): New member function.
26909         (profile_probability::debug): New member function.
26910         (profile_probability::differs_from_p): New member function.
26911         (profile_probability::differs_lot_from_p): New member function.
26912         (profile_probability::stream_in): New member function.
26913         (profile_probability::stream_out): New member function.
26914         * profile-count.h (profile_count_quality): Rename to ...
26915         (profile_quality): ... this one.
26916         (profile_probability): New.
26917         (profile_count): Update.
26918         * profile.c (compute_branch_probabilities): Update.
26919         * recog.c (peep2_attempt): Update.
26920         * sched-ebb.c (schedule_ebbs): Update.
26921         * sched-rgn.c (find_single_block_region): Update.
26922         (compute_dom_prob_ps): Update.
26923         (schedule_region): Update.
26924         * sel-sched-ir.c (compute_succs_info): Update.
26925         * stmt.c (struct case_node): Update.
26926         (do_jump_if_equal): Update.
26927         (get_outgoing_edge_probs): Update.
26928         (conditional_probability): Update.
26929         (emit_case_dispatch_table): Update.
26930         (expand_case): Update.
26931         (expand_sjlj_dispatch_table): Update.
26932         (emit_case_nodes): Update.
26933         * targhooks.c: Update.
26934         * tracer.c (better_p): Update.
26935         (find_best_successor): Update.
26936         * trans-mem.c (expand_transaction): Update.
26937         * tree-call-cdce.c: Update.
26938         * tree-cfg.c (gimple_split_edge): Upate.
26939         (move_sese_region_to_fn): Upate.
26940         * tree-cfgcleanup.c (cleanup_control_expr_graph): Upate.
26941         * tree-eh.c (lower_resx): Upate.
26942         (cleanup_empty_eh_move_lp): Upate.
26943         * tree-if-conv.c (version_loop_for_if_conversion): Update.
26944         * tree-inline.c (copy_edges_for_bb): Update.
26945         (copy_cfg_body): Update.
26946         * tree-parloops.c (gen_parallel_loop): Update.
26947         * tree-profile.c (gimple_gen_ic_func_profiler): Update.
26948         (gimple_gen_time_profiler): Update.
26949         * tree-ssa-dce.c (remove_dead_stmt): Update.
26950         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
26951         * tree-ssa-loop-im.c (execute_sm_if_changed): Update.
26952         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update.
26953         (unloop_loops): Update.
26954         (try_peel_loop): Update.
26955         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
26956         * tree-ssa-loop-split.c (connect_loops): Update.
26957         (split_loop): Update.
26958         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Update.
26959         (hoist_guard): Update.
26960         * tree-ssa-phionlycprop.c (propagate_rhs_into_lhs): Update.
26961         * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Update.
26962         (value_replacement): Update.
26963         * tree-ssa-reassoc.c (branch_fixup): Update.
26964         * tree-ssa-tail-merge.c (replace_block_by): Update.
26965         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges): Update.
26966         (create_edge_and_update_destination_phis): Update.
26967         (compute_path_counts): Update.
26968         (recompute_probabilities): Update.
26969         (update_joiner_offpath_counts): Update.
26970         (freqs_to_counts_path): Update.
26971         (duplicate_thread_path): Update.
26972         * tree-switch-conversion.c (hoist_edge_and_branch_if_true): Update.
26973         (struct switch_conv_info): Update.
26974         (gen_inbound_check): Update.
26975         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
26976         (vect_do_peeling): Update.
26977         (vect_loop_versioning): Update.
26978         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
26979         (optimize_mask_stores): Update.
26980         * ubsan.c (ubsan_expand_null_ifn): Update.
26981         * value-prof.c (gimple_divmod_fixed_value): Update.
26982         (gimple_divmod_fixed_value_transform): Update.
26983         (gimple_mod_pow2): Update.
26984         (gimple_mod_pow2_value_transform): Update.
26985         (gimple_mod_subtract): Update.
26986         (gimple_mod_subtract_transform): Update.
26987         (gimple_ic): Update.
26988         (gimple_stringop_fixed_value): Update.
26989         (gimple_stringops_transform): Update.
26990         * value-prof.h: Update.
26992 2017-06-29  Carl Love  <cel@us.ibm.com>
26994         * config/rs6000/rs6000-c.c: Add support for built-in functions
26995         vector signed int vec_signed (vector float);
26996         vector signed long long vec_signed (vector double);
26997         vector signed int vec_signed2 (vector double, vector double);
26998         vector signed int vec_signede (vector double);
26999         vector signed int vec_signedo (vector double);
27000         * config/rs6000/rs6000.c (rs6000_generate_vsigned2_code): Add
27001         instruction generator.
27002         * config/rs6000/vsx.md (UNSPEC_VSX_XVCVSPSXWS, UNSPEC_VSX_XVCVSPSXDS,
27003         UNSPEC_VSX_VSIGNED2): Add UNSPECS.
27004         (vsx_xvcvspsxws, vsx_xvcvdpuxds_scale, vsx_xvcvspuxws, vsigned2_v2df):
27005         Add define_insn.
27006         (vsignedo_v2df, vsignede_v2df, vunsigned2_v2df, vunsignedo_v2df,
27007         vunsignede_v2df): Add define_expands.
27008         * config/rs6000/rs6000-builtin.def (VEC_SIGNED, VEC_UNSIGNED,
27009         VEC_SIGNED2, VEC_UNSIGNED2, VEC_SIGNEDE, VEC_UNSIGNEDE, VEC_SIGNEDO,
27010         VEC_UNSIGNEDO): Add definitions.
27011         * config/vsx.md (UNSPEC_VSX_XVCVSPSXWS, UNSPEC_VSX_XVCVSPSXDS,
27012         UNSPEC_VSX_VSIGNED2): Add UNSPECs.
27013         (vsx_xvcvspsxws, vsx_xvcvspuxws): Add define_insn.
27014         (vsigned2_v2df, vsigendo_v2df, vsignede_v2df,
27015         vunsigned2_v2df, vunsignedo_v2df, vunsignede_v2df): Add define_expands.
27016         * config/rs6000/altivec.h (vec_signed, vec_signed2,
27017         vec_signede and vec_signedo, vec_unsigned, vec_unsigned2,
27018         vec_unsignede, vec_unsignedo): Add builtin defines.
27019         * config/rs6000-protos.h (rs6000_generate_vsigned2_code): Add extern
27020         declaration.
27021         * doc/extend.texi: Update the built-in documentation file for the
27022         new built-in functions.
27024 2017-06-29  Richard Biener  <rguenther@suse.de>
27026         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Do not add
27027         reduction chains to LOOP_VINFO_REDUCTIONS.
27028         * tree-vect-slp.c (vect_analyze_slp): Continue looking for
27029         SLP reductions after processing reduction chains.
27031 2017-06-29  Nathan Sidwell  <nathan@acm.org>
27033         * builtins.c (fold_builtin_FUNCTION): Use
27034         lang_hooks.decl_printable_name.
27036 2017-06-29  Peter Bergner  <bergner@vnet.ibm.com>
27038         PR middle-end/81194
27039         * cfgexpand.c (expand_gimple_stmt_1): Handle switch statements
27040         with only one label.
27041         * stmt.c (expand_case): Assert NCASES is greater than one.
27043 2017-06-29  Richard Biener  <rguenther@suse.de>
27045         * tree-cfg.c (group_case_labels_stmt): Return whether we changed
27046         anything.
27047         (group_case_labels): Likewise.
27048         (find_taken_edge): Push sanity checking on val to workers...
27049         (find_taken_edge_cond_expr): ... here
27050         (find_taken_edge_switch_expr): ... and here, handle cases
27051         with just a default label.
27052         * tree-cfg.h (group_case_labels_stmt): Adjust prototype.
27053         (group_case_labels): Likewise.
27054         * tree-cfgcleanup.c (execute_cleanup_cfg_post_optimizing): When
27055         group_case_labels does anything cleanup the CFG again.
27057 2017-06-29  Bin Cheng  <bin.cheng@arm.com>
27059         PR tree-optimization/81196
27060         * tree-ssa-loop-niter.c (number_of_iterations_cond): Handle loop
27061         exit condition comparing two IVs.
27063 2017-06-29  Richard Earnshaw  <rearnsha@arm.com>
27065         * config/arm/parsecpu.awk (gen_comm_data): Add initializer for
27066         profile to the dummy entry at the end of the list of architectures.
27067         * config/arm/arm-cpu-cdata.h: Regenerated.
27069 2017-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27070             Michael Collison <michael.collison@arm.com>
27072         PR target/70119
27073         * config/aarch64/aarch64.md (*aarch64_<optab>_reg_<mode>3_mask1):
27074         New pattern.
27075         (*aarch64_reg_<mode>3_neg_mask2): New pattern.
27076         (*aarch64_reg_<mode>3_minus_mask): New pattern.
27077         (*aarch64_<optab>_reg_di3_mask2): New pattern.
27078         * config/aarch64/aarch64.c (aarch64_rtx_costs): Account for cost
27079         of shift when the shift amount is masked with constant equal to
27080         the size of the mode.
27081         * config/aarch64/predicates.md (subreg_lowpart_operator): New
27082         predicate.
27084 2017-06-29  Martin Liska  <mliska@suse.cz>
27086         * config/i386/i386.opt: Change range from [1,5] to [0,5].
27088 2017-06-29  Yury Gribov  <tetra2005@gmail.com>
27090         PR bootstrap/80565
27091         * ipa-cp.c (allocate_and_init_ipcp_value): Add initialization
27092         code.
27093         * ipa-inline.h
27094         (edge_growth_cache_entry::edge_growth_cache_entry): New
27095         function.
27096         (reset_edge_growth_cache): Update to use constructor.
27098 2017-06-28  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
27100         * config/aarch64/aarch64.h (AARCH64_EXPAND_ALIGNMENT): New.
27101         (DATA_ALIGNMENT): Update to use AARCH64_EXPAND_ALIGNMENT.
27102         (LOCAL_ALIGNMENT): Update to use AARCH64_EXPAND_ALIGNMENT.
27104 2017-06-28  Sebastian Peryt  <sebastian.peryt@intel.com>
27106         * config/i386/avx512vlintrin.h (_mm256_permutexvar_epi64)
27107         (_mm256_permutexvar_epi32, _mm256_permutex_epi64): New intrinsics.
27109 2017-06-28  Szabolcs Nagy  <szabolcs.nagy@arm.com>
27111         * config.gcc (*-linux-musl*): Add t-musl tmake_file.
27112         (*-linux-uclibc*): Add t-uclibc tmake_file.
27113         * config/t-musl: New.
27114         * config/t-uclibc: New.
27116 2017-06-28  Richard Earnshaw  <rearnsha@arm.com>
27118         * config/arm/parsecpu.awk (profile): Parse new keyword in an arch
27119         context.
27120         (gen_comm_data): Emit architectural setting of arch_prof.
27121         * config/arm/arm-cpus.in (armv6-m, armv6s-m, armv7-a, armv7ve): Set the
27122         profile.
27123         (armv7-r, armv7-m, armv7e-m, armv8-a, armv8.1-a, armv8.2-a): Likewise.
27124         (armv8-m.base, armv8-m.main): Likewise.
27125         * arm-protos.h (arm_build_target): Add profile field.
27126         (arch_option): Likewise.
27127         * config/arm/arm.c (arm_configure_build_target): Copy the profile to
27128         the active target.
27129         * config/arm/arm.h (TARGET_ARM_ARCH_PROFILE): Use
27130         arm_active_target.profile.
27132 2017-06-28  Richard Biener  <rguenther@suse.de>
27134         PR middle-end/81227
27135         * fold-const.c (negate_expr_p): Use TYPE_UNSIGNED, not
27136         TYPE_OVERFLOW_WRAPS.
27137         * match.pd (negate_expr_p): Likewise.
27138         * tree-ssa-reassoc.c (optimize_range_tests_diff): Use
27139         fold_build2, not fold_binary.
27141 2017-06-28  Wilco Dijkstra  <wdijkstr@arm.com>
27143         * config/aarch64/aarch64 (aarch64_expand_mov_immediate):
27144         Convert memory address to Pmode.
27145         (aarch64_print_operand): Assert MEM operands are always Pmode.
27147 2017-06-28  Wilco Dijkstra  <wdijkstr@arm.com>
27149         PR target/79665
27150         * config/arm/aarch-common.c (arm_no_early_alu_shift_dep):
27151         Remove redundant if.
27152         (aarch_forward_to_shift_is_not_shifted_reg): Remove.
27153         * config/arm/aarch-common-protos.h
27154         (aarch_forward_to_shift_is_not_shifted_re): Remove.
27155         * config/arm/cortex-a53.md: Use arm_no_early_alu_shift_dep in bypass.
27157 2017-06-28  Michael Meissner  <meissner@linux.vnet.ibm.com>
27159         PR ipa/81238
27160         * multiple_target.c (create_dispatcher_calls): Set the default
27161         clone to be static, not public.
27163 2017-06-28  Richard Biener  <rguenther@suse.de>
27165         * tree-vect-loop.c (vectorizable_reduction): Move special
27166         cond reduction IV var creation ...
27167         (vect_create_epilog_for_reduction): ... here.  Remove induction_index
27168         parameter.  Use STMT_VINFO_VECTYPE.
27169         * tree-vect-slp.c (vect_get_constant_vectors): Properly reset
27170         constant_p.
27172 2017-06-28  Martin Liska  <mliska@suse.cz>
27174         PR ipa/81128
27175         * ipa-visibility.c (non_local_p): Handle visibility.
27177 2017-06-28  Martin Liska  <mliska@suse.cz>
27179         PR driver/79659
27180         * common.opt: Add IntegerRange to various options.
27181         * opt-functions.awk (integer_range_info): New function.
27182         * optc-gen.awk: Add integer_range_info to cl_options struct.
27183         * opts-common.c (decode_cmdline_option): Handle
27184         CL_ERR_INT_RANGE_ARG.
27185         (cmdline_handle_error): Likewise.
27186         * opts.c (print_filtered_help): Show valid interval in
27187         when --help is provided.
27188         * opts.h (struct cl_option): Add range_min and range_max fields.
27189         * config/i386/i386.opt: Add IntegerRange for -mbranch-cost.
27191 2017-06-28  Marc Glisse  <marc.glisse@inria.fr>
27193         * match.pd ((X & ~Y) | (~X & Y)): Generalize to + and ^.
27194         (x * C EQ/NE y * C): New transformation.
27196 2017-06-28  Christophe Lyon  <christophe.lyon@linaro.org>
27198         * genmultilib (combination_space): Accept '+' in option names.
27200 2017-06-28  Martin Liska  <mliska@suse.cz>
27202         PR sanitizer/81224
27203         * asan.c (instrument_derefs): Bail out inner references
27204         that are hard register variables.
27206 2017-06-28  Jakub Jelinek  <jakub@redhat.com>
27208         PR target/81175
27209         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use def_builtin
27210         rather than def_builtin_pure for __builtin_ia32_gatherpf*.
27212 2017-06-28  Richard Biener  <rguenther@suse.de>
27214         * tree-vectorizer.h (vect_get_vec_defs): Remove.
27215         (vect_get_slp_defs): Adjust.
27216         * tree-vect-loop.c (get_initial_defs_for_reduction): Split
27217         out from ...
27218         * tree-vect-slp.c (vect_get_constant_vectors): ... here and
27219         simplify.
27220         * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
27221         get_initial_defs_for_reduction instead of vect_get_vec_defs.
27222         (vectorizable_reduction): Adjust.
27223         * tree-vect-slp.c (vect_get_constant_vectors): Remove reduction
27224         handling.
27225         (vect_get_slp_defs): Likewise.
27226         * tree-vect-stmts.c (vect_get_vec_defs): Make static and adjust.
27227         (vectorizable_bswap): Adjust.
27228         (vectorizable_call): Likewise.
27229         (vectorizable_conversion): Likewise.
27230         (vectorizable_assignment): Likewise.
27231         (vectorizable_shift): Likewise.
27232         (vectorizable_operation): Likewise.
27233         (vectorizable_store): Likewise.
27234         (vectorizable_condition): Likewise.
27235         (vectorizable_comparison): Likewise.
27237 2017-06-28  Michael Collison  <michael.collison@arm.com>
27239         PR target/68535
27240         * config/arm/arm.c (gen_ldm_seq): Remove last unnecessary
27241         set of base_reg
27242         (arm_gen_movmemqi): Removed unused variable 'i'.
27243         Convert 'for' loop into 'while' loop.
27244         (arm_expand_prologue): Remove last unnecessary set of insn.
27245         (thumb_pop): Remove unused variable 'pushed_words'.
27246         (thumb_exit): Remove last unnecessary set of regs_to_pop.
27248 2017-06-28  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
27250         * config/s390/predicates.md: Use s390_rel_address_ok_p.
27251         * config/s390/s390-protos.h: Add prototype of
27252         s390_rel_address_ok_p.
27253         * config/s390/s390.c (s390_got_symbol): New function.
27254         (s390_rel_address_ok_p): New function.
27255         (legitimize_pic_address): Use s390_rel_address_ok_p.
27256         (s390_load_got): Use s390_got_symbol.
27257         (s390_option_override): Issue error if
27258         -mno-pic-data-is-text-relative is used without -fpic/-fPIC.
27259         * config/s390/s390.h (TARGET_DEFAULT_PIC_DATA_IS_TEXT_RELATIVE):
27260         New macro.
27261         * config/s390/s390.opt: New option mpic-data-is-text-relative.
27263 2017-06-27  Andrew Pinski  <apinski@cavium.com>
27265         * match.pd (X >/>=/</<= 0 ? 1.0 : -1.0): New patterns.
27266         (X * copysign (1.0, X)): New pattern.
27267         (X * copysign (1.0, -X)): New pattern.
27268         (copysign (-1.0, CST)): New pattern.
27270 2017-06-27  Joseph Myers  <joseph@codesourcery.com>
27272         * genmultilib (combination_space): Remove variable.
27273         Validate reuse rules against regular expression for any sequence
27274         of multilib options in any order.
27276 2017-06-27  Michael Collison  <michael.collison@arm.com>
27278         * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Directly
27279         call aarch64_split_simd_combine.
27280         * (aarch64_combine_internal<mode>): Delete pattern.
27281         * config/aarch64/aarch64.c (aarch64_split_simd_combine):
27282         Allow register and subreg operands.
27284 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27286         * config/i386/vxworks.h (ASM_SPEC): Remove definition. No target
27287         specific need, just fallback on defaults.
27288         (ASM_OUTPUT_ALIGNED_BSS): Add #undef before #define.
27290 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27291             Olivier Hainque  <hainque@adacore.com>
27293         * config/i386/vxworks.h (DBX_REGISTER_NUMBER): Pick distinct
27294         map for 64bits.
27295         (TARGET_OS_CPP_BUILTINS): builtin_define CPU to X86_64 for 64bit
27296         targets. Pick a default if no particular attempt applied.
27297         (STACK_CHECK_PROTECT): Double for 64bit targets, which have
27298         larger contexts.
27300 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27302         * config.gcc (i*86-wrs-vxworks7): Handle new acceptable triplet.
27303         (x86_64-wrs-vxworks7): Likewise.
27305 2017-06-27  Marek Polacek  <polacek@redhat.com>
27307         PR sanitizer/81223
27308         * ubsan.c (instrument_null): Check get_base_address's result for null.
27310 2017-06-27  Marc Glisse  <marc.glisse@inria.fr>
27312         * match.pd ((A+-B)+(C-A), (A+B)-(A-C)): New transformations.
27314 2017-06-27  Marc Glisse  <marc.glisse@inria.fr>
27316         * builtin-types.def (BT_FENV_T_PTR, BT_CONST_FENV_T_PTR,
27317         BT_FEXCEPT_T_PTR, BT_CONST_FEXCEPT_T_PTR): New primitive types.
27318         (BT_FN_INT_FENV_T_PTR, BT_FN_INT_CONST_FENV_T_PTR,
27319         BT_FN_INT_FEXCEPT_T_PTR_INT, BT_FN_INT_CONST_FEXCEPT_T_PTR_INT):
27320         New function types.
27321         * builtins.def (BUILT_IN_FECLEAREXCEPT, BUILT_IN_FEGETENV,
27322         BUILT_IN_FEGETEXCEPTFLAG, BUILT_IN_FEGETROUND,
27323         BUILT_IN_FEHOLDEXCEPT, BUILT_IN_FERAISEEXCEPT,
27324         BUILT_IN_FESETENV, BUILT_IN_FESETEXCEPTFLAG,
27325         BUILT_IN_FESETROUND, BUILT_IN_FETESTEXCEPT,
27326         BUILT_IN_FEUPDATEENV): New builtins.
27327         * tree-core.h (TI_FENV_T_PTR_TYPE, TI_CONST_FENV_T_PTR_TYPE,
27328         TI_FEXCEPT_T_PTR_TYPE, TI_CONST_FEXCEPT_T_PTR_TYPE): New entries.
27329         * tree.h (fenv_t_ptr_type_node, const_fenv_t_ptr_type_node,
27330         fexcept_t_ptr_type_node, const_fexcept_t_ptr_type_node): New
27331         macros.
27332         (builtin_structptr_types): Adjust size.
27333         * tree.c (builtin_structptr_types): Add four entries.
27335 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27336             Olivier Hainque  <hainque@adacore.com>
27338         * config/vxworks.h (VXWORKS_LIB_SPEC): Incorporate ...
27339         (TLS_SYM): New local macro, forcing reference to __tls__ on
27340         link command lines for VxWorks 7 RTPs, triggering initialization
27341         of tlsLib.
27342         (VXWORKS_HAVE_TLS): New macro. State whether the target VxWorks
27343         OS features TLS support, true for RTPs on VxWorks 7.
27344         * config/vxworks.c (vxworks_override_options): Setup emutls
27345         accordingly.
27347 2017-06-27  Jakub Jelinek  <jakub@redhat.com>
27349         * predict.c (test_prediction_value_range): Use -1U instead of -1
27350         to avoid narrowing conversion warning.
27351         * dumpfile.c (dump_options): Wrap all value into dump_flags_t cast
27352         to avoid narrowing conversion warning.
27353         * opt-functions.awk (var_ref): Return (unsigned short) -1 instead of
27354         -1.
27355         * optc-gen.awk (END): Expect (unsigned short) -1 instead of -1.
27357 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27359         * config/vxworks.h (VXWORKS_LIBS_RTP): Alternative definition for
27360         64bit configurations.
27361         (PTR_DIFF_TYPE): Alternative definition for TARGET_LP64.
27362         (SIZE_TYPE): Likewise.
27363         * config/vxworks.c (vxworks_emutls_var_fields): Use
27364         long_unsigned_type_node instead of unsigned_type_node as the offset
27365         field type, which is "pointer" mode in emutls.c.
27367 2017-06-27  Jakub Jelinek  <jakub@redhat.com>
27369         PR sanitizer/81209
27370         * ubsan.c (ubsan_encode_value): Initialize DECL_CONTEXT on var.
27372         PR middle-end/81207
27373         * gimple-fold.c (replace_call_with_call_and_fold): Handle
27374         gimple_vuse copying separately from gimple_vdef copying.
27376 2017-06-27  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
27378         * value-prof.c (free_hist): Remove call to memset and the enclosing if
27379         condition.
27381 2017-06-26  Jerome Lambourg  <lambourg@adacore.com>
27382             Olivier Hainque  <hainque@adacore.com>
27384         * config.gcc (*-*-vxworks*): Add TARGET_VXWORKS7=1 to tm_defines
27385         for all vxworks7 targets.
27386         * config/vxworks.h (TARGET_VXWORKS7): If not defined, define to 0.
27387         (VXWORKS_ADDITIONAL_CPP_SPEC): Alternative definition for VXWORKS7.
27388         (VXWORKS_LIBS_RTP, VXWORKS_LIBS_RTP_DIR): New macros, allowing
27389         variations for VX6/VX7 and 32/64bits later on in ...
27390         (VXWORKS_LIB_SPEC): Leverage new macros.
27391         (VXWORKS_OS_CPP_BUILTINS): Define _VSB_CONFIG_FILE for VXWORKS7,
27392         as well as _ALLOW_KEYWORD_MACROS when "inline" is not a keyword.
27394 2017-06-26  Jerome Lambourg  <lambourg@adacore.com>
27396         * config/vxworks.h (VXWORKS_OS_CPP_BUILTINS): builtin_define
27397         _VX_TOOL_FAMILY and _VX_TOOL to gnu.
27399 2017-06-26  Carl Love  <cel@us.ibm.com>
27401         * config/rs6000/rs6000-c.c: Add support for built-in functions
27402         vector bool char vec_reve (vector bool char);
27403         vector signed char vec_reve (vector signed char);
27404         vector unsigned char vec_reve (vector unsigned char);
27405         vector bool int vec_reve (vector bool int);
27406         vector signed int vec_reve (vector signed int);
27407         vector unsigned int vec_reve (vector unsigned int);
27408         vector bool long long vec_reve (vector bool long long);
27409         vector signed long long vec_reve (vector signed long long);
27410         vector unsigned long long vec_reve (vector unsigned long long);
27411         vector bool short vec_reve (vector bool short);
27412         vector signed short vec_reve (vector signed short);
27413         vector double vec_reve (vector double);
27414         vector float vec_reve (vector float);
27415         * config/rs6000/rs6000-builtin.def (VREVE_V2DI, VREVE_V4SI,
27416         VREVE_V8HI, VREVE_V16QI, VREVE_V2DF, VREVE_V4SF, VREVE): New builtin.
27417         * config/rs6000/altivec.md (UNSPEC_VREVEV): New UNSPEC.
27418         (altivec_vreve): New pattern.
27419         * config/rs6000/altivec.h (vec_reve): New define.
27420         * doc/extend.texi (vec_rev): Update the built-in documentation file
27421         for the new built-in functions.
27423 2016-06-26  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27425         PR tree-optimization/71815
27426         * gimple-ssa-strength-reduction.c (uses_consumed_by_stmt): New
27427         function.
27428         (find_basis_for_candidate): Call uses_consumed_by_stmt rather than
27429         has_single_use.
27430         (slsr_process_phi): Likewise.
27431         (replace_uncond_cands_and_profitable_phis): Don't replace a
27432         multiply candidate with a stride of 1 (copy or cast).
27433         (phi_incr_cost): Call uses_consumed_by_stmt rather than
27434         has_single_use.
27435         (lowest_cost_path): Likewise.
27436         (total_savings): Likewise.
27438 2017-06-26  Richard Biener  <rguenther@suse.de>
27440         PR target/81175
27441         * config/i386/i386.c (ix86_init_mmx_sse_builtins):
27442         Use def_builtin_pure for all gather builtins.
27444 2017-06-26  Richard Biener  <rguenther@suse.de>
27446         PR tree-optimization/81203
27447         * tree-tailcall.c (find_tail_calls): Do not move stmts into
27448         non-dominating BBs.
27450 2017-06-26  Marek Polacek  <polacek@redhat.com>
27452         PR c/80116
27453         * doc/invoke.texi: Document -Wmultistatement-macros.
27455 2017-06-26  Christophe Lyon  <christophe.lyon@linaro.org>
27457         * doc/sourcebuild.texi (ARM-specific attributes): Document new
27458         arm_neon_ok_no_float_abi effective target.
27460 2017-06-26  Richard Biener  <rguenther@suse.de>
27462         PR tree-optimization/80928
27463         * cfghooks.c (duplicate_block): Do not copy BB_DUPLICATED flag.
27464         (copy_bbs): Set BB_DUPLICATED flag early.
27465         (execute_on_growing_pred): Do not execute for BB_DUPLICATED
27466         marked blocks.
27467         (execute_on_shrinking_pred): Likewise.
27468         * tree-ssa.c (ssa_redirect_edge): Do not look for PHI args in
27469         BB_DUPLICATED blocks.
27470         * tree-ssa-phionlycoprop.c (eliminate_degenerate_phis_1): Properly
27471         iterate over all PHIs considering removal of *gsi.
27473 2017-06-23  Jim Wilson  <jim.wilson@linaro.org>
27475         * doc/invoke.texi (AArch64 Options, -mtune): Re-add falkor and
27476         qdf24xx.
27478 2017-06-23  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
27480         * config/rs6000/rs6000-string.c: (expand_block_clear,
27481         do_load_for_compare, select_block_compare_mode,
27482         compute_current_alignment, expand_block_compare,
27483         expand_strncmp_align_check, expand_strn_compare,
27484         expand_block_move, rs6000_output_load_multiple)
27485         Move functions related to string/block move/compare
27486         to a separate file.
27487         * config/rs6000/rs6000.c: Move above functions to rs6000-string.c.
27488         * config/rs6000/rs6000-protos.h (rs6000_emit_dot_insn): Add prototype
27489         for this function which is now used in two files.
27490         * config/rs6000/t-rs6000: Add rule to compile rs6000-string.o.
27491         * config.gcc: Add rs6000-string.o to extra_objs for
27492         targets powerpc*-*-* and rs6000*-*-*.
27494 2017-06-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
27496         PR target/80510
27497         * config/rs6000/rs6000.md (ALTIVEC_DFORM): Do not allow DImode in
27498         32-bit, since indexed is not valid for DImode.
27499         (mov<mode>_hardfloat32): Reorder ISA 2.07 load/stores before ISA
27500         3.0 d-form load/stores to be the same as mov<mode>_hardfloat64.
27501         (define_peephole2 for Altivec d-form load): Add 32-bit support.
27502         (define_peephole2 for Altivec d-form store): Likewise.
27504         PR ipa/81185
27505         * multiple_target.c (create_dispatcher_calls): Only create the
27506         dispatcher call if the function is the default clone of a
27507         versioned function.
27509 2017-06-23  Segher Boessenkool  <segher@kernel.crashing.org>
27511         PR middle-end/80902
27512         * builtins.c (expand_builtin_atomic_fetch_op): If emitting code after
27513         a call, force the call to not be a tail call.
27515 2017-06-23  Jeff Law  <law@redhat.com>
27517         * doc/contrib.texi: Add entry for Steven Pemberton's work on
27518         enquire.
27520 2017-06-23  Will Schmidt  <will_schmidt@vnet.ibm.com>
27522         * config/rs6000/rs6000.c: Add include of ssa-propagate.h for
27523         update_call_from_tree().  (rs6000_gimple_fold_builtin): Add
27524         handling for early expansion of vector shifts (sl,sr,sra,rl).
27525         (builtin_function_type): Add vector shift right instructions
27526         to the unsigned argument list.
27528 2017-06-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
27530         rtl-optimizatoin/79286
27531         * ira.c (update_equiv_regs): Revert to using may_trap_or_fault_p again.
27532         * rtlanal.c (rtx_addr_can_trap_p_1): SYMBOL_REF_FUNCTION_P can never
27533         trap.  PIC register plus a const unspec without offset can never trap.
27535 2017-06-23  Marc Glisse  <marc.glisse@inria.fr>
27537         * tree.h (builtin_structptr_type): New type.
27538         (builtin_structptr_types): Declare new array.
27539         * tree.c (builtin_structptr_types): New array.
27540         (free_lang_data, build_common_tree_nodes): Use it.
27542 2017-06-23  Jonathan Wakely  <jwakely@redhat.com>
27544         PR c++/81187
27545         * doc/invoke.texi (-Wnoexcept-type): Fix name of option, from
27546         -Wnoexcept.
27548 2017-06-22  Matt Turner  <mattst88@gmail.com>
27550         * config/i386/driver-i386.c (host_detect_local_cpu): Add Kaby
27551         Lake models to skylake case.  Assume skylake for unknown
27552         models with clflushopt.
27554 2017-06-22  Jeff Law  <law@redhat.com>
27556         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Handle
27557         frame sizes that do not satisfy aarch64_uimm12_shift.
27559 2017-06-22  Jan Hubicka <hubicka@ucw.cz>
27561         * profile-count.h (apply_probability,
27562         apply_scale, probability_in): Fix checks for zero.
27564 2017-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27566         * incpath.c (add_sysroot_to_chain): Allow for $SYSROOT prefix.
27567         * doc/cppdiropts.texi (-I @var{dir}): Document it.
27569 2016-06-22  Richard Biener  <rguenther@suse.de>
27571         * tree-vect-loop.c (vect_model_reduction_cost): Handle
27572         COND_REDUCTION and INTEGER_INDUC_COND_REDUCTION without
27573         REDUC_MAX_EXPR support.
27574         (vectorizable_reduction): Likewise.
27575         (vect_create_epilog_for_reduction): Likewise.
27577 2017-06-22  James Greenhalgh  <james.greenhalgh@arm.com>
27579         * match.pd (A / (1 << B) -> A >> B): New.
27580         * generic-match-head.c: Include optabs-tree.h.
27581         * gimple-match-head.c: Likewise.
27582         * optabs-tree.h (target_supports_op_p): New.
27583         * optabs-tree.c (target_supports_op_p): New.
27585 2017-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27587         * configure.ac (gcc_cv_ld_static_dynamic): Also check stderr for
27588         $gcc_cv_ld --help output.
27589         (gcc_cv_ld_demangle): Likewise.
27590         (gcc_cv_ld_eh_frame_hdr): Likewise.
27591         (gcc_cv_ld_pie): Likewise.
27592         (gcc_cv_ld_as_needed): Likewise.  Prefer native forms unless $gnu_ld.
27593         (gcc_cv_ld_buildid): Likewise.
27594         (gcc_cv_ld_sysroot): Likewise.
27595         (ld_bndplt_support): Likewise.
27596         (ld_pushpopstate_support): Likewise.
27597         * configure: Regenerate.
27598         * config/sol2.h [!USE_GLD] (SYSROOT_SPEC): Define.
27600 2017-06-21  Jakub Jelinek  <jakub@redhat.com>
27602         PR target/81151
27603         * config/i386/sse.md (round<mode>2): Renumber match_dup and
27604         operands indexes to avoid gap between operands and match_dups.
27606 2017-06-21  Andrew Pinski  <apinski@cavium.com>
27608         * config/aarch64/aarch64-cost-tables.h (thunderx_extra_costs):
27609         Increment Arith_shift and Arith_shift_reg by 1.
27610         * config/aarch64/aarch64-tuning-flags.def (cheap_shift_extend):
27611         New tuning flag.
27612         * config/aarch64/aarch64.c (thunderx_tunings): Enable
27613         AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
27614         (aarch64_strip_extend): Add new argument and test for it.
27615         (aarch64_cheap_mult_shift_p): New function.
27616         (aarch64_rtx_mult_cost): Call aarch64_cheap_mult_shift_p and don't
27617         add a cost if it is true.
27618         Update calls to aarch64_strip_extend.
27619         (aarch64_rtx_costs): Update calls to aarch64_strip_extend.
27621 2017-06-21  Andrew Pinski  <apinski@cavium.com>
27623         * config/aarch64/aarch64-cores.def (thunderxt88p1): Use thunderxt88
27624         tunings.
27625         (thunderxt88): Likewise.
27626         * config/aarch64/aarch64.c (thunderxt88_prefetch_tune): New variable.
27627         (thunderx_prefetch_tune): New variable.
27628         (thunderx2t99_prefetch_tune): Update for the correct values.
27629         (thunderxt88_tunings): New variable.
27630         (thunderx_tunings): Use thunderx_prefetch_tune instead of
27631         generic_prefetch_tune.
27632         (thunderx2t99_tunings): Use AUTOPREFETCHER_WEAK.
27634 2017-06-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27636         * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>_lse,
27637         SHORT): Relax operand 3 to aarch64_reg_or_zero and constraint to Z.
27638         (aarch64_compare_and_swap<mode>_lse, GPI): Likewise.
27639         (aarch64_atomic_cas<mode>, SHORT): Likewise for operand 2.
27640         (aarch64_atomic_cas<mode>, GPI): Likewise.
27642 2017-06-21  Martin Liska  <mliska@suse.cz>
27644         * gimplify.c (gimplify_label_expr): Insert GIMPLE_PREDICT
27645         statements on cold and hot labels.
27646         * predict.c (tree_estimate_probability_bb): Remove the
27647         prediction from this place.
27649 2017-06-21  Martin Liska  <mliska@suse.cz>
27651         PR tree-optimization/79489
27652         * gimplify.c (maybe_add_early_return_predict_stmt): New
27653         function.
27654         (gimplify_return_expr): Call the function.
27655         * predict.c (tree_estimate_probability_bb): Remove handling
27656         of early return.
27657         * predict.def: Update comment about early return predictor.
27658         * gimple-predict.h (is_gimple_predict): New function.
27659         * predict.def: Change default value of early return to 66.
27660         * tree-tailcall.c (find_tail_calls): Skip GIMPLE_PREDICT
27661         statements.
27662         * passes.def: Put pass_strip_predict_hints to the beginning of
27663         IPA passes.
27665 2017-06-21  Pierre-Marie de Rodat  <derodat@adacore.com>
27667         * dwarf2out.c (gen_decl_die): Remove the guard to skip file-scope
27668         FUNCTION_DECL declarations.
27669         (dwarf2out_early_global_decl): Remove the guard to skip FUNCTION_DECL
27670         declarations.
27671         (dwaf2out_decl): Likewise.
27672         * godump.c (go_early_global_decl): Skip call to the real debug hook
27673         for FUNCTION_DECL declarations.
27674         * passes.c (rest_of_decl_compilation): Skip call to the
27675         early_global_decl debug hook for FUNCTION_DECL declarations, unless
27676         -fdump-go-spec is passed.
27678 2017-06-21  Marc Glisse  <marc.glisse@inria.fr>
27680         * config/i386/i386.c (struct builtin_isa): New field pure_p.
27681         Reorder for compactness.
27682         (def_builtin, def_builtin2, ix86_add_new_builtins): Handle pure_p.
27683         (def_builtin_pure, def_builtin_pure2): New functions.
27684         (ix86_init_mmx_sse_builtins) [__builtin_ia32_stmxcsr]: Mark as pure.
27686 2017-06-21  Marc Glisse  <marc.glisse@inria.fr>
27688         * match.pd (nop_convert): New predicate.
27689         ((A +- CST1) +- CST2): Allow some NOP conversions.
27691 2017-06-21  Jakub Jelinek  <jakub@redhat.com>
27693         PR c++/81130
27694         * gimplify.c (omp_add_variable): Don't force GOVD_SEEN for types
27695         with ctors/dtors if GOVD_SHARED is set.
27697 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27699         * config/aarch64/aarch64.md (movti_aarch64):
27700         Emit mov rather than orr.
27701         (movtf_aarch64): Likewise.
27702         * config/aarch64/aarch64-simd.md (aarch64_simd_mov):
27703         Emit mov rather than orr.
27705 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27707         * config/aarch64/aarch64-simd.md (aarch64_simd_dup):
27708         Swap alternatives, make integer dup more expensive.
27710 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27712         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
27713         Return true for non-tls symbols.
27715 2017-06-21  James Greenhalgh  <james.greenhalgh@arm.com>
27717         * config/aarch64/aarch64-cores.def (cortex-a55): New.
27718         (cortex-a75): Likewise.
27719         (cortex-a75.cortex-a55): Likewise.
27720         * config/aarch64/aarch64-tune.md: Regenerate.
27721         * doc/invoke.texi (-mtune): Document new values for -mtune.
27723 2017-06-21  Tom de Vries  <tom@codesourcery.com>
27725         * doc/sourcebuild.texi (Add Options, Features for dg-add-options): Add
27726         stack_size feature.
27727         (Effective-Target Keywords, Other attributes): Suggest using
27728         dg-add-options stack_size feature to get stack limit in stack_size
27729         effective target documentation.
27731 2017-06-21  Julian Brown  <julian@codesourcery.com>
27732             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
27734         * config/aarch64/aarch64-simd.md (aarch64_crypto_pmulldi)
27735         (aarch64_crypto_pmullv2di): Change type attribute to crypto_pmull.
27736         * config/aarch64/thunderx2t99.md (thunderx2t99_pmull): New
27737         reservation.
27738         * config/arm/cortex-a53.md (cortex_a53_advsimd_type): Add crypto_pmull to
27739         attribute type list for neon_multiply.
27740         * config/arm/cortex-a57.md (cortex_a57_neon_type): Add crypto_pmull to
27741         attribute type list for neon_multiply.
27742         * config/arm/crypto.md (crypto_vmullp64): Change type to crypto_pmull.
27743         * config/arm/exynos-m1.md (exynos_m1_neon_type): Add crypto_pmull to
27744         attribute type list for neon_multiply.
27745         * config/arm/types.md (crypto_pmull): Add.
27746         * config/arm/xgene1.md (xgene1_neon_pmull): Add crypto_pmull to
27747         attribute type list.
27749 2017-06-20  Andreas Tobler  <andreast@gcc.gnu.org>
27751         * config.gcc (armv6*-*-freebsd*): Change the target_cpu_cname to
27752         arm1176jzf-s.
27754 2017-06-20  Jakub Jelinek  <jakub@redhat.com>
27756         * ira-costs.c (find_costs_and_classes): Initialize cost_classes later
27757         to make sure not to dereference a NULL cost_classes_ptr pointer.
27759 2017-06-20  Carl Love  <cel@us.ibm.com>
27761         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
27762         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
27763         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
27764         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
27765         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
27766         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
27767         VMULOSW): New enum "unspec" values.
27768         (vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
27769         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si,
27770         altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
27771         altivec_vmulosw): New patterns.
27772         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
27773         VMULOSW): Add definitions.
27775 2017-06-20  Julia Koval  <julia.koval@intel.com>
27777         * config/i386/i386.c: Fix rounding expand for new pattern.
27778         * config/i386/subst.md: Fix pattern (parallel -> unspec).
27780 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27782         * config/aarch64/aarch64-option-extensions.def (rcpc): New.
27783         * config/aarch64/aarch64.h (AARCH64_FL_RCPC): New.
27785 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27787         * config/aarch64/aarch64-option-extensions.def (fp16): Fix expected
27788         feature string.
27790 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27792         * config/aarch64/aarch64-cores.def: Rearrange to sort by
27793         architecture, then by implementer ID.
27794         * config/aarch64/aarch64-tune.md: Regenerate.
27796 2017-06-20  Richard Biener  <rguenther@suse.de>
27798         PR middle-end/81097
27799         * fold-const.c (split_tree): Fold to type before negating.
27801 2017-06-20  David Malcolm  <dmalcolm@redhat.com>
27803         * diagnostic-show-locus.c
27804         (selftest::test_fixit_deletion_affecting_newline): New function.
27805         (selftest::diagnostic_show_locus_c_tests): Call it.
27807 2017-06-20  Andreas Schwab  <schwab@suse.de>
27809         PR target/80970
27810         * config/m68k/m68k.md (bsetdreg, bchgdreg, bclrdreg): Use "=d"
27811         instead of "+d".
27813 2017-06-20  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
27815         * config/arm/arm-c.c (arm_cpu_builtins): New block to define
27816         __ARM_FEATURE_COPROC according to support.
27818 2017-06-20  Jakub Jelinek  <jakub@redhat.com>
27820         * tree-chkp.c (chkp_get_hard_register_var_fake_base_address):
27821         Rewritten to avoid overflow for > 32-bit pointers.
27823         PR sanitizer/81125
27824         * ubsan.h (ubsan_encode_value): Workaround buggy clang++ parser
27825         by removing enum keyword.
27826         (ubsan_type_descriptor): Likewise.  Formatting fix.
27828         PR target/81121
27829         * config/i386/i386.md (TARGET_USE_VECTOR_CONVERTS float si->{sf,df}
27830         splitter): Require TARGET_SSE2 in the condition.
27832 2017-06-20  Michael Meissner  <meissner@linux.vnet.ibm.com>
27834         PR target/79799
27835         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Add support
27836         for doing vector set of SFmode on ISA 3.0.
27837         * config/rs6000/vsx.md (vsx_set_v4sf_p9): Likewise.
27838         (vsx_set_v4sf_p9_zero): Special case setting 0.0f to a V4SF
27839         element.
27840         (vsx_insert_extract_v4sf_p9): Add an optimization for inserting a
27841         SFmode value into a V4SF variable that was extracted from another
27842         V4SF variable without converting the element to double precision
27843         and back to single precision vector format.
27844         (vsx_insert_extract_v4sf_p9_2): Likewise.
27846 2017-06-19  Jakub Jelinek  <jakub@redhat.com>
27848         * tree-ssa-structalias.c (get_constraint_for_ptr_offset): Multiply
27849         in UWHI to avoid undefined overflow.
27851         PR sanitizer/81125
27852         * ubsan.h (enum ubsan_encode_value_phase): New.
27853         (ubsan_encode_value): Change second argument to
27854         enum ubsan_encode_value_phase with default value of
27855         UBSAN_ENCODE_VALUE_GENERIC.
27856         * ubsan.c (ubsan_encode_value): Change second argument to
27857         enum ubsan_encode_value_phase PHASE from bool IN_EXPAND_P,
27858         adjust uses, for UBSAN_ENCODE_VALUE_GENERIC use just
27859         create_tmp_var_raw instead of create_tmp_var and use a
27860         TARGET_EXPR.
27861         (ubsan_expand_bounds_ifn, ubsan_build_overflow_builtin,
27862         instrument_bool_enum_load, ubsan_instrument_float_cast): Adjust
27863         ubsan_encode_value callers.
27865         PR sanitizer/81111
27866         * ubsan.c (ubsan_encode_value): If current_function_decl is NULL,
27867         use create_tmp_var_raw instead of create_tmp_var, mark it addressable
27868         just by setting TREE_ADDRESSABLE on the result and use a TARGET_EXPR.
27870 2017-06-19  Richard Biener  <rguenther@suse.de>
27872         PR middle-end/81118
27873         * tree-cfgcleanup.c (cleanup_tree_cfg_noloop): Clear niter
27874         estimates if we changed anything.
27876 2017-06-19  Richard Biener  <rguenther@suse.de>
27878         PR tree-optimization/80887
27879         * tree-ssa-sccvn.c (mprts_hook_cnt): New global.
27880         (vn_lookup_simplify_result): Allow only mprts_hook_cnt succesful
27881         simplified lookups, then reset mprts_hook.
27882         (vn_nary_build_or_lookup_1): Set mprts_hook_cnt to 9 before
27883         simplifying.
27884         (try_to_simplify): Likewise.
27886 2017-06-19  Martin Liska  <mliska@suse.cz>
27888         PR sanitizer/80879
27889         * gimplify.c (gimplify_switch_expr):
27890         Initialize live_switch_vars for SWITCH_BODY == STATEMENT_LIST.
27892 2017-06-19  Martin Liska  <mliska@suse.cz>
27894         * doc/install.texi: Document that PGO runs in 4 stages.
27896 2017-06-19  Martin Liska  <mliska@suse.cz>
27898         PR ipa/80732
27899         * attribs.c (make_dispatcher_decl): Do not append '.ifunc'
27900         to dispatcher function name.
27901         * multiple_target.c (replace_function_decl): New function.
27902         (create_dispatcher_calls): Redirect both edges and references.
27904 2017-06-19  Jan Hubicka <hubicka@ucw.cz>
27906         * profile-count.c (profile_count::dump): Dump quality.
27907         (profile_count::differs_from_p): Update for unsigned val.
27908         * profile-count.h (profile_count_quality): New enum.
27909         (profile_count): Turn m_val to 62bit unsigned, add quality tracking.
27911 2017-06-19  Richard Biener  <rguenther@suse.de>
27913         * tree-ssa-loop-niter.h (estimate_numbers_of_iterations): Take
27914         struct function as arg.
27915         (estimate_numbers_of_iterations): Export overload with loop arg.
27916         (free_numbers_of_iterations_estimates_loop): Use an overload of
27917         free_numbers_of_iterations_estimates instead.
27918         * tree-cfg.c (remove_bb): Adjust.
27919         * tree-cfgcleanup.c (remove_forwarder_block_with_phi): Likewise.
27920         * tree-parloops.c (gen_parallel_loop): Likewise.
27921         * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables):
27922         Likewise.
27923         (tree_unroll_loops_completely): Likewise.
27924         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations_loop):
27925         Use an overload instead and export.
27926         (estimated_loop_iterations): Adjust.
27927         (max_loop_iterations): Likewise.
27928         (likely_max_loop_iterations): Likewise.
27929         (estimate_numbers_of_iterations): Take struct function as arg
27930         and adjust.
27931         (loop_exits_before_overflow): Adjust.
27932         (free_numbers_of_iterations_estimates_loop): Use an overload.
27933         * tree-vect-loop.c (vect_analyze_loop_form): Adjust.
27934         * tree-vectorizer.c (vect_free_loop_info_assumptions): Likewise.
27936 2017-06-19  Richard Biener  <rguenther@suse.de>
27938         PR ipa/81112
27939         * ipa-prop.c (find_constructor_constant_at_offset): Handle
27940         RANGE_EXPR conservatively.
27942 2017-06-16  Carl Love  <cel@us.ibm.com>
27944         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
27945         definitions for vec_float, vec_float2, vec_floato,
27946         vec_floate built-ins.
27947         * config/rs6000/vsx.md (define_c_enum "unspec"): Add RTL code
27948         for instructions vsx_xvcvsxws vsx_xvcvuxwsp, float2, floato and
27949         floate.
27950         * config/rs6000/rs6000-builtin.def (FLOAT2_V2DI, FLOATE_V2DF,
27951         FLOATE_2DI, FLOATO_V2DF, FLOATEE_V2DI, XVCVSXWSP_V4SF,
27952         UNS_FLOATO_V2DI, UNS_FLOATE_V2DI): Add definitions.
27953         * config/altivec.md (define_insn "p8_vmrgew_<mode>",
27954         define_mode_attr VF_sxddp): Add V4SF type to p8_vmrgew.
27955         * config/rs6000/altivec.h (vec_float, vec_float2, vec_floate,
27956         vec_floato): Add builtin defines.
27957         * doc/extend.texi (vec_float, vec_float2, vec_floate, vec_floato):
27958         Update the built-in documentation file for the new built-in
27959         functions.
27961 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27963         * config/arm/arm.opt (marm): Mark as the negative of of -mthumb.
27964         (mthumb): Mark as the negative of -marm.
27966 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27968         * doc/invoke.texi (ARM Options, -mcpu): Document supported
27969         extension options.
27970         (ARM Options, -mtune): Document that this accepts the same
27971         extension options as -mcpu.
27972         (ARM Options, -mfpu): Document addition of -mfpu=auto.
27974 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27976         * doc/invoke.texi (ARM Options, -march=): Document new syntax and
27977         permitted extensions.
27979 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27981         * config/arm/arm-cpus.in (armv7): Add extension +nofp.
27982         (armv7-r): Add aliases vfpv3xd and vfpv3-d16.
27983         (armv8-m.main): Add option +nodsp.
27984         * config/arm/arm-cpu-cdata.h: Regenerated.
27986 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27988         * config/arm/t-fuchsia: New file.
27989         * config.gcc (arm*-*-fuchsia*): Use it.
27991 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27993         * config/arm/t-symbian: Rewrite for new option infrastructure.
27995 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27997         * config/arm/t-phoenix (MULTILIB_REUSE): Clear variable.
27998         (MULTILIB_REQUIRED): Likewise.
28000 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28002         * config/arm/t-linux-eabi (MULTILIB_EXCEPTIONS): Set to empty.
28003         (MULTILIB_RESUE): Likewise.
28004         (MULTILIB_MATCHES): Likewise.
28005         (MULTLIB_REQUIRED): Likewise.
28007 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28009         * config/arm/t-rtems: Rewrite for new option framework.
28011 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28013         * config/arm/t-aprofile (v7_a_nosimd_variants, v7_a_simd_variants)
28014         (v7ve_nosimd_variatns, v7ve_vfpv3_simd_variants)
28015         (v7ve_vfpv4_simd_variants, v8_a_nosimd_variants, v8_a_simd_variants)
28016         (v8_1_a_simd_variants, v8_2_a_simd_variants): Move to ...
28017         * config/arm/t-multilib: ... here.
28018         (MULTILIB_OPTIONS): Add armv7 and armv7+fp architectures.
28019         (MULTILIB_MATCHES): Use armv7 libraries for armv7-r.  Also use for
28020         armv7-a and armv8*-a when A-profile libraries have not been built.
28021         * config/arm/t-rmprofile: Rewrite.
28023 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28025         * genmultilib (multilib_reuse): Allow an explicit period to be escaped
28026         with a backslash.  Remove the backslash after substituting unescaped
28027         periods.
28028         * doc/fragments.texi (MULTILIB_REUSE): Document it.
28030 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28032         * config.gcc: (arm*-*-*): When building a-profile libraries, force
28033         the driver to pass through the default setting of -mfloat-abi.
28034         * common/config/arm/arm-common.c (arm_target_thumb_only): Return -marm
28035         rather than NULL.
28036         * config/arm/t-multilib (MULTILIB_REUSE): Initialize to empty.
28037         (all_feat_combs): New rule.
28038         (MULTILIB_OPTIONS): Use explicit ARM and Thumb directories.  Rework
28039         default libraries.
28040         * config/arm/t-aprofile: Rewrite.
28042 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28044         * config/arm/arm.h (FPUTYPE_AUTO): Define.
28045         * config/arm/arm.c (arm_option_override): Use FPUTYPE_AUTO if the
28046         fpu is not specified by the user/command-line.
28047         * config/arm/bpabi.h (FPUTYPE_DEFAULT): Delete.
28048         * config/arm/netbsd-elf.h (FPUTYPE_DEFAULT): Delete.
28049         * config/arm/linux-elf.h (FPUTYPE_DEFAULT): Delete.
28050         * config/arm/vxworks.h (FPUTYPE_DEFAULT): Delete.
28051         * common/config/arm/arm-common.c (arm_canon_arch_option): Use
28052         FPUTYPE_AUTO insted of FPUTYPE_DEFAULT.
28054 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28056         * config/arm/elf.h (MULTILIB_DEFAULTS): Delete.
28057         * config/arm/t-arm-elf: Rewritten.
28059 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28061         * config/arm/arm.h (TARGET_HARD_FLOAT): Also check that we
28062         have some floating-point instructions.
28063         (TARGET_SOFT_FLOAT): Define as inverse of TARGET_HARD_FLOAT.
28064         (TARGET_MAYBE_HARD_FLOAT): New macro.
28065         * config/arm/arm-builtins.c (arm_init_builtins): Use
28066         TARGET_MAYBE_HARD_FLOAT.
28067         * config/arm/arm.c (arm_option_override): Use TARGET_HARD_FLOAT_ABI.
28069 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28071         * common/config/arm/arm-common.c: Define INCLUDE_LIST.
28072         (configargs.h): Include it.
28073         (arm_print_hint_for_fpu_option): New function.
28074         (arm_parse_fpu_option): New function.
28075         (candidate_extension): New class.
28076         (arm_canon_for_multilib): New function.
28077         * config/arm/arm.h (CANON_ARCH_SPEC_FUNCTION): New macro.
28078         (EXTRA_SPEC_FUNCTIONS): Add CANON_ARCH_SPEC_FUNCTION.
28079         (ARCH_CANONICAL_SPECS): New macro.
28080         (DRIVER_SELF_SPECS): Add ARCH_CANONICAL_SPECS.
28082 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28084         * config.gcc (arm*-*-*): Ensure both target_cpu_cname and with_cpu
28085         are set after handling multilib fragments.  Set target_cpu_default2
28086         from with_cpu.
28088 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28090         * config.gcc (arm*-*-fucshia*): Set target_cpu_cname to the real
28091         cpu name.
28092         (arm*-*-*): Set target_cpu_default2 to a quoted string.
28093         * config/arm/parsecpu.awk (check_cpu): Validate any extension
28094         options.
28095         (check_arch): Likewise.
28096         * config/arm/arm.c (arm_configure_build_target): Handle
28097         TARGET_CPU_DEFAULT being a string constant.  Scan any feature
28098         options in the default.
28100 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28102         * config/arm/arm-protos.h (cpu_arch_extension): Add field to record
28103         when an option is an alias of another.
28104         * config/arm/parsecpu.awk (optalias): New parser token.
28105         (gen_comm_data): Mark non-alias options as such.  Emit entries
28106         for extension aliases.
28107         * config/arm/arm-cpus.in (armv5e): Make vfpv2 an alias.
28108         (armv5te, armv5tej, armv6, armv6j, armv6k, armv6z): Likewise.
28109         (armv6kz, armv6zk, armv6t2): Likewise.
28110         (armv7): Make vfpv3-d16 an alias.
28111         (armv7-a): Make vfpv3-d16, neon and neon-vfpv3 aliases.  Sort in
28112         canonical order.
28113         (armv7ve): Make vfpv4-d16, neon-vfpv3 and neon-vfpv4 aliases.
28114         Sort in canonical order.
28115         (armv8-a): Sort in canonical order.
28116         (armv8.1-a, armv8.2-a):  Likewise.
28117         (generic-armv7-a): Make neon and neon-vfpv3 aliases.  Sort in
28118         canonical order.
28119         (cortex-a9): Sort in canonical order.
28120         * config/arm/arm.c (selftests.h): Include it.
28121         (arm_test_cpu_arch_data): New function.
28122         (arm_run_self_tests): New function.
28123         (TARGET_RUN_TARGET_SELFTESTS): Redefine.
28124         (targetm): Move declaration to the end of the file.
28125         * arm-cpu-cdata.h: Regenerated.
28127 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28129         * config/arm/arm.h (TARGET_MODE_SPECS): Add additional parameter to
28130         call to target_mode_check describing the type of option passed.
28131         * common/config/arm/arm-common.c (arm_arch_core_flag): Delete.
28132         (arm_target_thumb_only): Use arm_parse_arch_option_name or
28133         arm_parse_cpu_option_name to match parameters against list of
28134         available targets.
28135         * config/arm/parsecpu.awk (gen_comm_data): Don't generate
28136         arm_arch_core_flags data structure.
28137         * config/arm/arm-cpu_cdata.h: Regenerated.
28139 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28141         * common/config/arm/arm-common.c (arm_initialize_isa): Moved here from
28142         config/arm/arm.c.
28143         (arm_print_hint_for_cpu_option): Likewise.
28144         (arm_print_hint_for_arch_option): Likewise.
28145         (arm_parse_cpu_option_name): Likewise.
28146         (arm_parse_arch_option_name): Likewise.
28147         * config/arm/arm.c (arm_identify_fpu_from_isa): Use the computed number
28148         of entries in the all_fpus list.
28149         * config/arm/arm-protos.h (all_architectures, all_cores): Declare.
28150         (arm_parse_cpu_option_name): Declare.
28151         (arm_parse_arch_option_name): Declare.
28152         (arm_parse_option_features): Declare.
28153         (arm_intialize_isa): Declare.
28154         * config/arm/parsecpu.awk (gen_data): Move CPU and architecture
28155         data tables to ...
28156         (gen_comm_data): ... here.  Make definitions non-static.
28157         * config/arm/arm-cpu-data.h: Regenerated.
28158         * config/arm/arm-cpu-cdata.h: Regenerated.
28160 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28162         * config/arm/arm-protos.h (arm_build_target): Remove arch_core.
28163         (cpu_arch_extension): New structure.
28164         (cpu_arch_option, arch_option, cpu_option): New structures.
28165         * config/arm/parsecpu.awk (gen_headers): Build an enumeration of
28166         architecture types.
28167         (gen_data): Generate new format data tables.
28168         * config/arm/arm.c (cpu_tune): New structure.
28169         (cpu_option, processors): Delete.
28170         (arm_print_hint_for_core_or_arch): Delete.  Replace with ...
28171         (arm_print_hint_for_cpu_option): ... this and ...
28172         (arm_print_hint_for_arch_option): ... this.
28173         (arm_parse_arch_cpu_name): Delete.  Replace with ...
28174         (arm_parse_cpu_option_name): ... this and ...
28175         (arm_parse_arch_option_name): ... this.
28176         (arm_unrecognized_feature): Change type of target parameter to
28177         cpu_arch_option.
28178         (arm_parse_arch_cpu_features): Delete.  Replace with ...
28179         (arm_parse_option_features): ... this.
28180         (arm_configure_build_target): Rework to use new configuration data
28181         tables.
28182         (arm_print_tune_info): Rework for new configuration data tables.
28183         * config/arm/arm-cpu-data.h: Regenerated.
28184         * config/arm/arm-cpu.h: Regenerated.
28186 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28188         * Makefile.in (OBJS): Move sbitmap.o from here ...
28189         (OBJS-libcommon): ... to here.
28191 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28193         * config/arm/arm-isa.h (ISA_ALL_FPU_INTERNAL): Renamed from ISA_ALL_FPU.
28194         (ISA_ALL_CRYPTO): New macro.
28195         (ISA_ALL_SIMD): New macro
28196         (ISA_ALL_FP): New macro.
28197         * config/arm/arm.c (fpu_bitlist): Update initializer.
28198         * config/arm/arm-cpus.in: Use new ISA_ALL macros to disable crypto,
28199         simd or fp.
28200         (arm9e): Add fpu.  Add option for nofp
28201         (arm946e-s, arm966e-s, arm968e-s, arm10e, arm1020e, arm1022e): Likewise.
28202         (arm926ej-s, arm1026ej-s): Likewise.
28203         (generic-armv7-a): Add fpu.  Add options for simd, vfpv3, vfpv3-d16,
28204         vfpv3-fp16, vfpv3-d16-fp16, vfpv4, vfpv4-d16, neon, neon-vfp3,
28205         neon-fp16, neon-vfpv4, nofp and nosimd.
28206         (cortex-a5, cortex-a7): Add fpu.  Add options for nosimd and nofp.
28207         (cortex-a8): Add fpu.  Add option for nofp.
28208         (cortex-a9): Add fpu.  Add options for nosimd and nofp.
28209         (cortex-a12, cortex-a15, cortex-a17): Add fpu.  Add option for nofp.
28210         (cortex-r4f): Add fpu.
28211         (cortex-r5): Add fpu.  Add options for nofp.dp and nofp.
28212         (cortex-r7): Use idiv option from architecture.  Add fpu.  Add option
28213         for nofp.
28214         (cortex-r8): Likewise.
28215         (cortex-m4): Add fpu.  Add option for nofp.
28216         (cortex-a15.cortex-a7): Add fpu.  Add option for nofp.
28217         (cortex-a17.cortex-a7): Likewise.
28218         (cortex-a32): Add fpu.  Add options for crypto and nofp.
28219         (cortex-a35, cortex-a53): Likewise.
28220         (cortex-a57): Add fpu.  Add option for crypto.
28221         (cortex-a72, cortex-a73): Likewise.
28222         (exynos-m1): Likewise.
28223         (cortex-a57.cortex-a53, cortex-a72.cortex-a53): Likewise.
28224         (cortex-a73.cortex-a35, cortex-a73.cortex-a53): Likewise.
28225         (cortex-m33): Add fpu.  Add option for nofp.
28226         * config/arm/arm-cpu-cdata.h: Regenerated
28227         * config/arm/arm-cpu-data.h: Regenerated.
28229 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28231         * arm-cpus.in (armv5e): Add options fp, vfpv2 and nofp.
28232         (armv5te, armv5tej): Likewise.
28233         (armv6, armv6j, armv6k, armv6z, armv6kz, armv6zk, armv6t2): Likewise.
28234         (armv7): Add options fp and vfpv3-d16.
28235         (armv7-a): Add options fp, simd, vfpv3, vfpv3-d16, vfpv3-d16-fp16,
28236         vfpv3-fp16, vfpv4, vfpv4-d16, neon, neon-vfpv3, neon-fp16, neon-vfpv4,
28237         nofp and nosimd.
28238         (armv7ve): Likewise.
28239         (armv7-r): Add options fp, fp.sp, idiv, nofp and noidiv.
28240         (armv7e-m): Add options fp, fpv5, fp.dp and nofp.
28241         (armv8-a): Add nocrypto option.
28242         (armv8.1-a, armv8.2-a): Likewise.
28243         (armv8-m.main): add options fp, fp.dp and nofp.
28245 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28247         * config/arm/arm-cpus.in (armv8-a): Add options crc, simd crypto and
28248         nofp.
28249         (armv8-a+crc): Delete.
28250         (armv8.1-a): Add options simd, crypto and nofp.
28251         (armv8.2-a): Add options fp16, simd, crypto and nofp.
28252         (armv8.2-a+fp16): Delete.
28253         (armv8-m.main): Add option dsp.
28254         (armv8-m.main+dsp): Delete.
28255         (cortex-a8): Add fpu.  Add nofp option.
28256         (cortex-a9): Add fpu.  Add nofp and nosimd options.
28257         * config/arm/parsecpu.awk (gen_data): Generate option tables and
28258         link to main cpu and architecture data structures.
28259         (gen_comm_data): Only put isa attributes from the main architecture
28260         in common tables.
28261         (option): New statement for architecture and CPU entries.
28262         * arm.c (struct cpu_option): New structure.
28263         (struct processors): Add entry for options.
28264         (arm_unrecognized_feature): New function.
28265         (arm_parse_arch_cpu_name): Ignore any characters after the first
28266         '+' character.
28267         (arm_parse_arch_cpu_feature): New function.
28268         (arm_configure_build_target): Separate out any CPU and architecture
28269         features and parse separately.  Don't error out if -mfpu=auto is
28270         used with only an architecture string.
28271         (arm_print_asm_arch_directives): New function.
28272         (arm_file_start): Call it.
28273         * config/arm/arm-cpu-cdata.h: Regenerated.
28274         * config/arm/arm-cpu-data.h: Likewise.
28275         * config/arm/arm-tables.opt: Likewise.
28277 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28279         * config/arm/elf.h (ASM_SPEC): Only pass -mfpu through to the
28280         assembler when it is not -mfpu=auto.
28282 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28284         * config/arm/arm.h (BIG_LITTLE_SPEC): Delete macro.
28285         (ASM_REWRITE_SPEC_FUNCTIONS): New macro.
28286         (BIG_LITTLE_CPU_SPEC_FUNCTIONS): Delete macro.
28287         (ASM_CPU_SPEC): Rewrite.
28288         (MCPU_MTUNE_NATIVE_FUNCTIONS): New macro.
28289         (EXTRA_SPEC_FUNCTIONS): Move outside of ifdef.  Use
28290         MCPU_MTUNE_NATIVE_FUNCTIONS and ASM_REWRITE_SPEC_FUNCTIONS.  Remove
28291         reference to BIG_LITTLE_CPU_SPEC_FUNCTIONS.
28292         * common/config/arm/arm-common.c (arm_rewrite_selected_cpu): Ensure
28293         copied string is NUL-terminated.  Also strip any characters prefixed
28294         by '+'.
28295         (arm_rewrite_selected_arch): New function.
28296         (arm_rewrite_march): New function.
28298 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28300         * config/arm/arm.opt (x_arm_arch_string): New TargetSave option.
28301         (x_arm_cpu_string, x_arm_tune_string): Likewise.
28302         (march, mcpu, mtune): Convert to string-based options.
28303         * config/arm/arm.c (arm_print_hint_for_core_or_arch): New function.
28304         (arm_parse_arch_cpu_name): New function.
28305         (arm_configure_build_target): Use arm_parse_arch_cpu_name to
28306         identify selected architecture or CPU.
28307         (arm_option_save): New function.
28308         (TARGET_OPTION_SAVE): Redefine.
28309         (arm_option_restore): Restore string options.
28310         (arm_option_print): Print string options.
28312 2017-06-16  Martin Sebor  <msebor@redhat.com>
28314         PR tree-optimization/80933
28315         PR tree-optimization/80934
28316         * builtins.c (fold_builtin_3): Do not handle bcmp here.
28317         * gimple-fold.c (gimple_fold_builtin_bcmp): New function.
28318         (gimple_fold_builtin_bcopy, gimple_fold_builtin_bzero): Likewise.
28319         (gimple_fold_builtin): Call them.
28321 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28323         * gimple-ssa-isolate-paths.c (isolate_path): Set edge leading to path
28324         as unlikely; update profile.
28326 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28328         * predict.c (force_edge_cold): Handle declaring edges impossible
28329         more aggresively.
28331 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28333         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update
28334         profile.
28335         (try_unroll_loop_completely): Fix reporting.
28337 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28339         * tree-ssa-tail-merge.c (replace_block_by): Fix profile updating.
28341 2017-06-16  James Greenhalgh  <james.greenhalgh@arm.com>
28343         PR target/71778
28344         * config/arm/arm-builtins.c (arm_expand_builtin_args): Return TARGET
28345         if given a non-constant argument for an intrinsic which requires a
28346         constant.
28348 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28350         * profile.c (compare_freqs): New function.
28351         (branch_prob): Sort edge list.
28352         (find_spanning_tree): Assume that the list is priority sorted.
28354 2017-06-16  Richard Biener  <rguenther@suse.de>
28356         PR tree-optimization/81090
28357         * passes.def (pass_record_bounds): Remove.
28358         * tree-pass.h (make_pass_record_bounds): Likewise.
28359         * tree-ssa-loop.c (pass_data_record_bounds, pass_record_bounds,
28360         make_pass_record_bounds): Likewise.
28361         * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables): Do
28362         not free niter estimates at the beginning but at the end.
28363         * tree-scalar-evolution.c (scev_finalize): Free niter estimates.
28365 2017-06-16  Richard Biener  <rguenther@suse.de>
28367         * tree-switch-conversion.c (emit_case_bit_tests): Adjust
28368         initializer to workaround ICE in host GCC 4.8.
28370 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28372         * ipa-inline-transform.c (update_noncloned_frequencies): Update also
28373         counts.
28374         (clone_inlined_nodes): Update.
28376 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28378         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune): Update
28379         prefetch settings, and enable prefetching by default at -O3.
28381 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28383         * config/aarch64/aarch64.c (aarch64_override_options_internal):
28384         Set flag_prefetch_loop_arrays according to tuning data.
28386 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28388         * config/aarch64/aarch64-protos.h (struct cpu_prefetch_tune):
28389         New tune structure.
28390         (struct tune_params): Use cpu_prefetch_tune instead of cache_line_size.
28391         [Unrelated to main purpose of the patch] Place the pointer field last
28392         to enable type checking errors when tune structure are wrongly merged.
28393         * config/aarch64/aarch64.c (generic_prefetch_tune,)
28394         (exynosm1_prefetch_tune, qdf24xx_prefetch_tune,)
28395         (thunderx2t99_prefetch_tune): New tune constants.
28396         (tune_params *_tunings): Update all tunings (no functional change).
28397         (aarch64_override_options_internal): Set PARAM_SIMULTANEOUS_PREFETCHES,
28398         PARAM_L1_CACHE_SIZE, PARAM_L1_CACHE_LINE_SIZE, and PARAM_L2_CACHE_SIZE
28399         from tunings structures.
28401 2017-06-16  Jakub Jelinek  <jakub@redhat.com>
28403         PR sanitizer/81094
28404         * ubsan.c (instrument_null): Add T argument, use it instead
28405         of computing it based on IS_LHS.
28406         (instrument_object_size): Likewise.
28407         (pass_ubsan::execute): Adjust instrument_null and
28408         instrument_object_size callers to pass gimple_get_lhs or
28409         gimple_assign_rhs1 result to it.  Use instrument_null instead of
28410         calling get_base_address and instrument_mem_ref.  Handle
28411         aggregate call arguments for object-size sanitization.
28413 2017-06-16  Yury Gribov  <tetra2005@gmail.com>
28415         PR tree-optimization/81089
28416         * tree-vrp.c (is_masked_range_test): Validate operands of
28417         subexpression.
28419 2017-06-15  Martin Sebor  <msebor@redhat.com>
28421         PR c++/80560
28422         * dumpfile.c (dump_register): Avoid calling memset to initialize
28423         a class with a default ctor.
28424         * gcc.c (struct compiler): Remove const qualification.
28425         * genattrtab.c (gen_insn_reserv): Replace memset with initialization.
28426         * hash-table.h: Ditto.
28427         * ipa-cp.c (allocate_and_init_ipcp_value): Replace memset with
28428           assignment.
28429         * ipa-prop.c (ipa_free_edge_args_substructures): Ditto.
28430         * omp-low.c (lower_omp_ordered_clauses): Replace memset with
28431         default ctor.
28432         * params.h (struct param_info): Make struct members non-const.
28433         * tree-switch-conversion.c (emit_case_bit_tests): Replace memset
28434         with default initialization.
28435         * vec.h (vec_copy_construct, vec_default_construct): New helper
28436         functions.
28437         (vec<T>::copy, vec<T>::splice, vec<T>::reserve): Replace memcpy
28438         with vec_copy_construct.
28439         (vect<T>::quick_grow_cleared): Replace memset with default ctor.
28440         (vect<T>::vec_safe_grow_cleared, vec_safe_grow_cleared): Same.
28441         * doc/invoke.texi (-Wclass-memaccess): Document.
28443 2017-06-15  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
28445         * emit-rtl.h (is_leaf): Update comment about local
28446         register allocator.
28448 2017-06-15  Jozef Lawrynowicz   <jozef.l@somniumtech.com>
28450         PR target/78818
28451         * config/msp430/msp430.c (msp430_data_attr): Check that it's possible
28452         for a variable to have a section before checking if the section has a
28453         name.
28454         Set section to.persistent if persistent attribute is set.
28455         Warn if .persistent attribute is used on an automatic variable.
28457 2017-06-15  Eric Botcazou  <ebotcazou@adacore.com>
28459         PR rtl-optimization/80474
28460         * reorg.c (update_block): Do not ignore instructions in a delay slot.
28462 2017-06-15  Segher Boessenkool  <segher@kernel.crashing.org>
28464         * config/rs6000/rs6000.md (add<mode>3): Use reg_or_subregno instead
28465         of REGNO.
28467 2017-06-14  Maciej W. Rozycki  <macro@imgtec.com>
28469         * config/mips/mips.md (MIPS16_T_REGNUM): Remove constant.
28470         (casesi): Emit bounds checking as RTL.
28471         (casesi_internal_mips16_<mode>): Remove bounds checking.
28473 2017-06-14  Max Filippov  <jcmvbkbc@gmail.com>
28475         * config/xtensa/xtensa.c (xtensa_option_override): Append
28476         MASK_CONST16 to target_flags in the absence of TARGET_L32R.
28477         (hwloop_optimize, hwloop_fail, hwloop_pattern_reg,
28478          xtensa_doloop_hooks): Define unconditionally.
28479         (xtensa_reorg_loops): Only call reorg_loops in the presence of
28480         TARGET_LOOPS.
28481         * config/xtensa/xtensa.h (TARGET_L32R): New definition.
28482         (TARGET_DEFAULT): Remove XCHAL_HAVE_L32R condition and account
28483         for it in xtensa_option_override.
28484         (HARD_FRAME_POINTER_IS_FRAME_POINTER,
28485          HARD_FRAME_POINTER_IS_ARG_POINTER): New definitions.
28487 2017-06-14  Boris Kolpackov  <boris@codesynthesis.com>
28489         * doc/cppopts.texi: Document '-' special value to -MF.
28491 2017-06-14  Wilco Dijkstra  <wdijkstr@arm.com>
28493         * config/arm/cortex-a53.md (cortex_a53_fpalu) Adjust latency.
28494         (cortex_a53_fconst): Likewise.
28495         (cortex_a53_fpmul): Likewise.
28496         (cortex_a53_f_load_64): Likewise.
28497         (cortex_a53_f_load_many): Likewise.
28498         (cortex_a53_advsimd_alu): Likewise.
28499         (cortex_a53_advsimd_alu_q): Likewise.
28500         (cortex_a53_advsimd_mul): Likewise.
28501         (cortex_a53_advsimd_mul_q): Likewise.
28502         (fpmac bypass): Add new bypass for fpmac-fpmac case.
28503         Add missing fmul, r2f_cvt and fconst cases.
28505 2017-06-14  Richard Biener  <rguenther@suse.de>
28507         PR middle-end/81088
28508         * fold-const.c (split_tree): Drop TREE_OVERFLOW flag from
28509         literal constants.
28510         (fold_binary_loc): When associating do not treat pre-existing
28511         TREE_OVERFLOW on literal constants as a reason to allow
28512         TREE_OVERFLOW on associated literal constants.
28514 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28516         * config/sparc/sparc.h (MASK_ISA): Add MASK_LEON and MASK_LEON3.
28517         (MASK_FEATURES): New macro.
28518         * config/sparc/sparc.c (sparc_option_override): Remove the special
28519         handling of -mfpu and generalize it to all MASK_FEATURES switches.
28521 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28523         * simplify-rtx.c (simplify_binary_operation_1) <UDIV>: Do not simplify
28524         a division of 0 if non-call exceptions are enabled.
28526 2017-06-14  Andrew Pinski  <apinski@cavium.com>
28527             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
28529         PR target/71663
28530         * config/aarch64/aarch64.c (aarch64_expand_vector_init):
28531         Improve vector initialization code gen for only variable case.
28533 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28535         * config/sparc/driver-sparc.c (cpu_names): Add SPARC-T5 entry.
28537 2017-06-14  Richard Biener  <rguenther@suse.de>
28539         PR tree-optimization/81083
28540         * tree-ssa-sccvn.c (vn_reference_lookup_3): Do not use abnormals
28541         as values.
28543 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28545         * config/rs6000/rs6000.c: Update all comments that mentioned SPE.
28546         (rs6000_expand_builtin): Remove RS6000_BTC_EVSEL.
28547         * config/rs6000/rs6000.h (RS6000_BTC_EVSEL): Delete.
28548         * config/rs6000/vxworks.h (VXCPU_FOR_8548): Delete.  Adjust former use.
28549         * config/rs6000/vxworksae.h (VXCPU_FOR_8548): Delete.
28550         * config/rs6000/vxworksmils.h (VXCPU_FOR_8548): Delete.
28552 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28554         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete VECTOR_SPE.
28555         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete VECTOR_SPE.
28557 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28559         * config/rs6000/rs6000.h (FIXED_SCRATCH): Delete.
28561 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28563         * config/rs6000/t-rtems: Don't handle SPE.
28565 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28567         * config/rs6000/t-linux: Don't handle SPE.
28569 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28571         * config/rs6000/eabispe.h: Delete file.
28573 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28575         * config/rs6000/t-spe: Delete file.
28577 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28579         * config/rs6000/rs6000.c (SPE_CONST_OFFSET_OK): Delete.
28580         (rs6000_legitimate_offset_address_p): Return false for anything in
28581         V2SImode or V2SFmode.
28583 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28585         * config/rs6000/rs6000-modes.def: Remove all 8-byte vector modes
28586         except V2SF and V2SI.  Rearrange the vector modes, and add comments.
28587         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove V8QImode
28588         and V4HImode.
28589         (reg_offset_addressing_ok_p): Remove V4HImode and V1DImode.
28590         (rs6000_legitimate_offset_address_p): Ditto.
28591         (rs6000_emit_move): Ditto.
28592         (rs6000_init_builtins): Remove V4HI_type_node.
28594 2017-06-13  Martin Liska  <mliska@suse.cz>
28596         PR sanitize/78204
28597         * asan.c (asan_sanitize_stack_p): Use sanitize_flags_p.
28598         (gate_asan): Likewise.
28599         * asan.h (asan_no_sanitize_address_p): Remove the function.
28600         (sanitize_flags_p): New function.
28601         * builtins.def: Fix coding style.
28602         * common.opt: Use renamed enum value.
28603         * convert.c (convert_to_integer_1): Use sanitize_flags_p.
28604         * doc/extend.texi: Document no_sanitize attribute.
28605         * flag-types.h (enum sanitize_code): Rename SANITIZE_NONDEFAULT
28606         to SANITIZE_UNDEFINED_NONDEFAULT.
28607         * gcc.c (sanitize_spec_function): Use the renamed enum value.
28608         * gimple-fold.c (optimize_atomic_compare_exchange_p):
28609         Use sanitize_flags_p.
28610         * gimplify.c (gimplify_function_tree): Likewise.
28611         * ipa-inline.c (sanitize_attrs_match_for_inline_p): Likewise.
28612         * opts.c (parse_no_sanitize_attribute): New function.
28613         (common_handle_option): Use renamed enum value.
28614         * opts.h (parse_no_sanitize_attribute): Declare.
28615         * tree.c (sanitize_flags_p): New function.
28616         * tree.h: Declared here.
28617         * tsan.c: Use sanitize_flags_p.
28618         * ubsan.c (ubsan_expand_null_ifn): Likewise.
28619         (instrument_mem_ref): Likewise.
28620         (instrument_bool_enum_load): Likewise.
28621         (do_ubsan_in_current_function): Remove the function.
28622         (pass_ubsan::execute): Use sanitize_flags_p.
28623         * ubsan.h: Remove do_ubsan_in_current_function
28624         * tree-cfg.c (print_no_sanitize_attr_value): New function.
28625         (dump_function_to_file): Use it here.
28627 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28629         PR tree-optimization/80803
28630         PR tree-optimization/81063
28631         * tree-sra.c (subtree_mark_written_and_enqueue): Move up in the file.
28632         (propagate_subaccesses_across_link): Enqueue subtree whenever
28633         necessary instead of relying on the caller.
28635 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28637         * tree-sra.c (add_access_to_work_queue): Only enqueue accesses
28638         that have a first_link.
28639         (sort_and_splice_var_accesses): Do not check first_link before
28640         enquing.
28641         (subtree_mark_written_and_enqueue): Likewise.
28642         (propagate_all_subaccesses): Likewise and do not stop at first
28643         parent with a first_link.
28645 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28647         * tree-sra.c (dump_access_tree_1): Fix accidental dumping to stderr
28648         instead of f.
28650 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28652         * match.pd: New pattern.
28654 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28656         * tree-vrp.c (is_masked_range_test): New function.
28657         (register_edge_assert_for): Determine ranges for
28658         some bit tests.
28660 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28662         PR tree-optimization/67328
28663         * fold-const.c (maskable_range_p): New function.
28664         (build_range_check): Generate bittests if possible.
28666 2017-06-13  Martin Liska  <mliska@suse.cz>
28668         * gimple-pretty-print.c (dump_probability): Add new argument.
28669         (dump_edge_probability): Dump both probability and count.
28670         (dump_gimple_label): Likewise.
28671         (dump_gimple_bb_header): Likewise.
28673 2017-06-13  Georg-Johann Lay  <avr@gjlay.de>
28675         PR target/81072
28676         * config/avr/avr-devices.c: Fix indentation.
28677         * config/avr/gen-avr-mmcu-specs.c: Dito.
28679 2017-06-13  Richard Biener  <rguenther@suse.de>
28681         * tree-vect-loop.c (vect_model_reduction_cost): Do not fail,
28682         instead get vector type from stmt_info.
28683         (vectorizable_reduction): Adjust.  Remove dead code.
28685 2017-06-13  Richard Biener  <rguenther@suse.de>
28687         PR middle-end/81065
28688         * fold-const.c (extract_muldiv_1): Remove bogus distribution
28689         case of C * (x * C2 + C3).
28690         (fold_addr_of_array_ref_difference): Properly fold index difference.
28692 2017-06-12  David S. Miller  <davem@davemloft.net>
28694         PR target/80968
28695         * config/sparc/sparc.md (return expander): Emit frame blockage if
28696         function uses alloca.
28698 2017-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
28700         * combine.c (make_field_assignment): Check len rather than the mode
28701         precision when calling force_to_mode.
28703 2017-06-12  Georg-Johann Lay  <avr@gjlay.de>
28705         Support multilibs and devices that see flash in RAM address range.
28707         PR target/81072
28708         * config/avr/avr-arch.h (avr_arch_id) <ARCH_AVRXMEGA3>: New enum.
28709         (avr_mcu_t) <flash_pm_offset>: New field.
28710         (avr_device_specific_features) <AVR_ISA_RCALL>: New enum.
28711         * config/avr/avr.h (AVR_SHORT_CALLS): New define.
28712         (AVR_HAVE_JMP_CALL): Don't set if AVR_SHORT_CALLS.
28713         (AVR_TINY_PM_OFFSET): Remove macro.
28714         * config/avr/avr.opt (-mshort-calls): New option.
28715         * config/avr/gen-avr-mmcu-specs.c (print_mcu)
28716         [*self_spec]: Add / remove -mshort-calls depending on AVR_ISA_RCALL.
28717         * config/avr/avr-c.c (avr_cpu_cpp_builtins)
28718         <__AVR_SHORT_CALLS__>: Built-in define if AVR_SHORT_CALLS.
28719         <__AVR_HAVE_JMP_CALL__>: Use AVR_HAVE_JMP_CALL as condition
28720         instead of avr_arch->have_jmp_call.
28721         <__AVR_PM_BASE_ADDRESS__>: Built-in define if avr_arch->flash_pm_offset.
28722         [AVR_TINY] <__AVR_TINY_PM_BASE_ADDRESS__>: Use
28723         avr_arch->flash_pm_offset to define.
28724         * config/avr/avr-devices.c (avr_arch_types): Add initializers for
28725         new field flash_pm_offset.  Add entry for avrxmega3.
28726         (avr_texinfo): Add entry for avrxmega3.
28727         * config/avr/avr-mcus.def: Add entries for: avrxmega3,
28728         attiny212, attiny214,
28729         attiny412, attiny414, attiny416, attiny417,
28730         attiny814, attiny816, attiny817,
28731         attiny1614, attiny1616, attiny1617,
28732         attiny3214, attiny3216, attiny3217.
28733         * config/avr/avr.c (avr_assemble_integer)[AVR_TINY]: Use
28734         avr_arch->flash_pm_offset instead of AVR_TINY_PM_OFFSET.
28735         (avr_print_operand_address) [AVR_TINY]: Same.
28736         (avr_asm_init_sections) <readonly_data_section>: Only patch
28737         callback if avr_arch->flash_pm_offset = 0.
28738         (avr_asm_named_section) <avr_need_copy_data_p>: Skip setting it
28739         for rodata if avr_arch->flash_pm_offset != 0.
28740         (avr_encode_section_info) [AVR_TINY]: Adjust comment.
28741         * config/avr/genmultilib.awk (dir_rcall, opt_rcall): New vars.
28742         (opts) [AVR_ISA_RCALL]: Append opt_rcall.
28743         (m_options): Append opt_rcall.
28744         (m_dirnames): Append dir_rcall.
28745         * config/avr/t-multilib: Regenerate.
28747         * configure.ac [target=avr]: Check whether avrxmega3 default
28748         linker description file works as needed.
28749         * configure: Regenerate.
28750         * doc/avr-mmcu.texi: Regenerate.
28751         * doc/invoke.texi (AVR Options) <-mshort-calls>: Document it.
28752         <__AVR_ARCH__>: Document avrxmega3 and 103.
28753         <__AVR_HAVE_JMP_CALL__>: Adjust documentation.
28754         <__AVR_SHORT_CALLS__>: Document it.
28755         <__AVR_PM_BASE_ADDRESS__>: Document it.
28756         * doc/extend.texi (AVR Options) <-mshort-calls>: Document it.
28757         (AVR Variable Attributes) <progmem>: Document this is
28758         not needed for avrxmega3.
28759         (AVR Named Address Spaces) <__flash>: Dito.
28761 2017-06-12  Jan Hubicka  <hubicka@ucw.cz>
28763         * cgraph.c (cgraph_node::dump): Complain about profile insanities.
28765 2017-06-12  Doug Rupp  <rupp@adacore.com>
28767         * config.gcc (*-*-vxworks*): Set use_gcc_stdint to "provide".
28768         Append vxworks-stdint.h to the tm_file list.
28769         * config/vxworks-stdint.h: New file.
28771 2017-06-12  Martin Liska  <mliska@suse.cz>
28773         PR tree-optimization/81041
28774         * tree-profile.c (gimple_gen_ic_func_profiler):
28775         Create an extra BB in profile-generate
28776         (gimple_gen_time_profiler): Likewise.
28778 2017-06-12  Jakub Jelinek  <jakub@redhat.com>
28780         PR tree-optimization/81003
28781         * tree-ssa-reassoc.c (force_into_ssa_name): New function.
28782         (update_range_test): Use it instead of force_gimple_operand_gsi.
28784 2017-06-12  Richard Biener  <rguenther@suse.de>
28786         PR tree-optimization/81053
28787         * tree-vect-loop.c (vect_is_simple_reduction): Handle PHI
28788         with backedge value not defined in loop.  Simplify def stmt
28789         compute.
28791 2017-06-11  Tom de Vries  <tom@codesourcery.com>
28793         PR target/79939
28794         * config/nvptx/nvptx.c (nvptx_cannot_force_const_mem): New function.
28795         Return true.
28796         (TARGET_CANNOT_FORCE_CONST_MEM): Redefine to
28797         nvptx_cannot_force_const_mem.
28799 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28801         * opts.c (finish_options): Move test for flag_split_stack after
28802         it has been initialized.
28804 2017-06-11  Jason Merrill  <jason@redhat.com>
28806         * tree.h (id_equal): New.
28807         * dwarf2out.c, hsa-gen.c, ipa-devirt.c, omp-expand.c,
28808         omp-simd-clone.c, read-rtl-function.c, tree-chkp.c, tree.c: Use it
28809         instead of strcmp of IDENTIFIER_POINTER.
28811 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28813         * ipa-inline-transform.c: Include function.h, cfg.h and basic-block.h
28814         (mark_all_inlined_calls_cdtor): Fix formating.
28815         (inline_transform): Rescale profile before inlining.
28817 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28819         * cgraph.h (cgraph_edge::clone): Update prototype.
28820         * cgraphclones.c (cgraph_edge::clone): Update profile scaling.
28821         (cgraph_node::create_clone): Update.
28822         (cgraph_node::create_version_clone): Update.
28823         * tree-inline.c (copy_bb): Update.
28824         (expand_call_inline): Update.
28826 2017-06-10  Segher Boessenkool  <segher@kernel.crashing.org>
28828         * config/rs6000/rs6000.c (emit_vrsave_prologue): New function,
28829         factored out from ...
28830         (rs6000_emit_prologue): ... here.
28832 2017-06-10  Segher Boessenkool  <segher@kernel.crashing.org>
28834         * config/rs6000/rs6000.c (emit_split_stack_prologue): New function,
28835         factored out from ...
28836         (rs6000_emit_prologue): ... here.
28838 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28840         * predict.c (drop_profile): Also drop individual bb/edge and cgraph
28841         edge counts.
28842         (handle_missing_profiles): Fix computation of tp_first_run.
28843         (counts_to_freqs): Do not touch freqs when count is 0.
28845 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28847         * cgraphbuild.c (cgraph_edge::rebuild_references): Do not touch
28848         profile.
28850 2017-06-10  Tom de Vries  <tom@codesourcery.com>
28852         * doc/sourcebuild.texi (Effective-Target Keywords, Environment
28853         attributes): Document signal effective target.
28855 2017-06-10  Tom de Vries  <tom@codesourcery.com>
28857         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
28858         Document effective target stack_size.
28860 2017-06-09  David Malcolm  <dmalcolm@redhat.com>
28862         * diagnostic.c (diagnostic_report_diagnostic): Only add fixits
28863         to the edit_context if they can be auto-applied.
28865 2017-06-9  Ian Lance Taylor  <iant@golang.org>
28867         * opts.c (finish_options): If -fsplit-stack, disable implicit
28868         -forder-blocks-and-partition.
28869         * doc/invoke.texi (Optimize Options): Document that when using
28870         -fsplit-stack -forder-blocks-and-partition is not implicitly
28871         enabled.
28873 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28875         * builtin-attrs.def (ATTR_NORETURN_NOTHROW_LEAF_COLD_LIST,
28876         ATTR_CONST_NORETURN_NOTHROW_LEAF_COLD_LIST,
28877         ATTR_TMPURE_NORETURN_NOTHROW_LEAF_COLD_LIST): New.
28878         * builtins.def (abort, trap, unreachable): Declare cold.
28879         * calls.c (flags_from_decl_or_type): Lookup ECF_COLD.
28880         * tree-core.h (ECF_COLD): New.
28881         * tree.c (set_call_expr_flags): Handle ECF_COLD.
28882         (build_common_builtin_nodes): Mark unreachable and abort as cold.
28884 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28886         * predict.c (unlikely_executed_stmt_p): Cleanup.
28888 2017-06-09  Richard Biener  <rguenther@suse.de>
28890         * tree-ssa-loop-im.c (execute_sm): Do not force multi-threaded
28891         model if the ref is always written to.
28893 2017-06-09  Tamar Christina  <tamar.christina@arm.com>
28895         * config/aarch64/aarch64.md (lrint<GPF:mode><GPI:mode>2): New.
28897 2017-06-09  Tamar Christina  <tamar.christina@arm.com>
28899         * config/arm/arm.c (arm_rtx_costs_internal): Make sdiv more expensive
28900         than udiv.
28902 2017-06-09  Tom de Vries  <tom@codesourcery.com>
28904         PR target/80855
28905         * config/nvptx/nvptx.md (define_expand "mov<QHSDISDFM>"): Error out with
28906         "target cannot support label values" when encountering LABEL_REF.
28908 2017-06-09  Martin Liska  <mliska@suse.cz>
28910         * tree-profile.c (gimple_gen_ic_profiler): Update comment.
28911         (gimple_gen_ic_func_profiler): Emit direct comparison
28912         of __gcov_indirect_call_callee with NULL.
28913         (gimple_gen_time_profiler): Change probability from
28914         PROB_VERY_UNLIKELY to PROB_UNLIKELY.
28916 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28918         * profile.c (edge_gcov_counts): Turn to pointer.
28919         (compute_branch_probabilities, compute_branch_probabilities): Update.
28920         (branch_prob): Do not clear edge_gcov_count.
28921         * profile.h (edge_gcov_counts): Turn to pointer.
28922         (edge_gcov_count): Update.
28924 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28926         * gimple.h (gimple_check_failed): Mark cold.
28928 2017-06-09  Richard Biener  <rguenther@suse.de>
28930         PR tree-optimization/66623
28931         * tree-vect-loop.c (vect_is_simple_reduction): Cleanup,
28932         refactor check_reduction into two parts, properly computing
28933         whether we have to check reduction validity for outer loop
28934         vectorization.
28936 2017-06-09  Richard Biener  <rguenther@suse.de>
28938         PR tree-optimization/79483
28939         * graphite-scop-detection.c (order): New global.
28940         (get_order): Compute bb to order mapping that satisfies code
28941         generation constraints.
28942         (cmp_pbbs): New helper.
28943         (build_scops): Start domwalk at entry block, sort generated
28944         pbbs.
28946 2017-06-09  Richard Biener  <rguenther@suse.de>
28948         PR middle-end/81007
28949         * ipa-polymorphic-call.c
28950         (ipa_polymorphic_call_context::restrict_to_inner_class):
28951         Skip FIELD_DECLs with error_mark_node type.
28952         * passes.def (all_lowering_passes): Run pass_build_cgraph_edges
28953         last again.
28955 2017-06-09  Martin Liska  <mliska@suse.cz>
28957         * predict.c (struct branch_predictor): New struct.
28958         (test_prediction_value_range): New test.
28959         (predict_c_tests): New function.
28960         * selftest-run-tests.c (selftest::run_tests): Run the function.
28961         * selftest.h: Declare new tests.
28963 2017-06-09  Segher Boessenkool  <segher@kernel.crashing.org>
28965         PR target/80966
28966         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack): Assert that
28967         gen_add3_insn did not fail.
28968         * config/rs6000/rs6000.md (add<mode>3): If asked to add a constant to
28969         r0, construct that number in a temporary reg and add that reg to r0.
28970         If asked to put the result in r0 as well, fail.
28972 2017-06-08  Will Schmidt  <will_schmidt@vnet.ibm.com>
28974         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
28975         for early expansion of vec_eqv.
28977 2017-06-08  Jakub Jelinek  <jakub@redhat.com>
28979         PR middle-end/81005
28980         * ubsan.c (instrument_null): Avoid pointless code temporary.
28981         (pass_ubsan::execute): Instrument aggregate arguments of calls.
28983 2017-06-08  Uros Bizjak  <ubizjak@gmail.com>
28985         PR target/81015
28986         Revert:
28987         2016-12-14  Uros Bizjak  <ubizjak@gmail.com>
28989         PR target/59874
28990         * config/i386/i386.md (*ctzhi2): New insn_and_split pattern.
28991         (*clzhi2): Ditto.
28993 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28995         * predict.c (unlikely_executed_edge_p): Move ahead.
28996         (probably_never_executed_edge_p): Use it.
28998 2017-06-08  Alexander Ivchenko  <alexander.ivchenko@intel.com>
29000         PR middle-end/79988
29001         * tree-chkp.c (chkp_gimple_call_builtin_p): Remove
29002         gimple_call_builtin_p call.
29004 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
29006         * system.h (fancy_abort): Annotate by ATTRIBUTE_COLD.
29007         * rtl.h (rtl_check_failed_bounds, rtl_check_failed_type1,
29008         rtl_check_failed_type2, rtl_check_failed_code1,
29009         rtl_check_failed_code2, rtl_check_failed_code_mode,
29010         rtl_check_failed_block_symbol, cwi_check_failed_bounds,
29011         rtvec_check_failed_bounds, rtl_check_failed_flag,
29012         _fatal_insn_not_found, _fatal_insn): Likewise.
29013         * tree.h (tree_contains_struct_check_failed,
29014         tree_check_failed, tree_not_check_failed,
29015         tree_class_check_failed, tree_range_check_failed,
29016         tree_not_class_check_failed, tree_int_cst_elt_check_failed,
29017         tree_vec_elt_check_failed, phi_node_elt_check_failed,
29018         tree_operand_check_failed, omp_clause_check_failed,
29019         omp_clause_operand_check_failed, omp_clause_range_check_failed):
29020         Likewise.
29022 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
29024         * cgraph.c (cgraph_edge::maybe_hot_p): Do not check
29025         flag_branch_probabilities.
29026         * ipa-inline.c (edge_badness): Likewise.
29027         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
29028         * postreload-gcse.c (eliminate_partially_redundant_load): Likewise.
29029         * predict.c (maybe_hot_frequency_p): Likewise.
29030         (probably_never_executed): Likewise.
29031         * sched-ebb.c (schedule_ebbs): Likewise.
29032         * sched-rgn.c (find_single_block_region): Likewise.
29033         * tracer.c (tail_duplicate): Likewise.
29035 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
29037         * opts.c (finish_options): x_flag_reorder_blocks_and_partition no
29038         longer requires x_flag_profile_use.
29040 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
29042         * cfgrtl.c (cfg_layout_initialize): Check crtl->has_bb_partition
29043         instead of flag_reorder_blocks_and_partition.
29044         * dbxout.c (dbxout_function_end): Likewise.
29045         * dwarf2out.c (gen_subprogram_die): Likewise.
29046         * haifa-sched.c (sched_create_recovery_edges): Likewise.
29047         * hw-doloop.c (reorg_loops): Likewise.
29048         * varasm.c (assemble_start_function,
29049         assemble_end_function): Likewise.
29050         (decide_function_section): Do not check for
29051         flag_reorder_blocks_and_partition.
29053 2017-06-08  Alexander Ivchenko  <alexander.ivchenko@intel.com>
29055         * tree-chkp.c (chkp_get_hard_register_var_fake_base_address):
29056         New function.
29057         (chkp_get_hard_register_fake_addr_expr): Ditto.
29058         (chkp_build_addr_expr): Add check for hard reg case.
29059         (chkp_parse_array_and_component_ref): Ditto.
29060         (chkp_find_bounds_1): Ditto.
29061         (chkp_process_stmt): Don't generate bounds store for
29062         hard reg case.
29064 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
29066         * predict.c (maybe_hot_bb_p): Do not check profile status.
29067         (maybe_hot_edge_p): Likewise.
29068         (probably_never_executed): Check for zero counts even if profile
29069         is not read.
29070         (unlikely_executed_edge_p): New function.
29071         (unlikely_executed_stmt_p): New function.
29072         (unlikely_executed_bb_p): New function.
29073         (set_even_probabilities): Use unlikely predicates.
29074         (combine_predictions_for_bb): Likewise.
29075         (predict_paths_for_bb): Likewise.
29076         (predict_paths_leading_to_edge): Likewise.
29077         (determine_unlikely_bbs): New function.
29078         (estimate_bb_frequencies): Use it.
29079         (compute_function_frequency): Use zero counts even if profile is
29080         not read.
29081         * profile-count.h: Fix typo.
29083 2017-08-08  Julia Koval  <julia.koval@intel.com>
29085         * config/i386/avx512bwintrin.h (_mm512_mask_cvtepi16_storeu_epi8,
29086         _mm512_mask_cvtsepi16_storeu_epi8,
29087         _mm512_mask_cvtusepi16_storeu_epi8): New intrinsics.
29088         * config/i386/avx512vlbwintrin.h (_mm256_mask_cvtepi16_storeu_epi8,
29089         _mm_mask_cvtsepi16_storeu_epi8, _mm256_mask_cvtsepi16_storeu_epi8,
29090         _mm_mask_cvtusepi16_storeu_epi8, _mm256_mask_cvtusepi16_storeu_epi8,
29091         _mm_mask_cvtepi16_storeu_epi8): New intrinsics.
29092         * config/i386/i386-builtin-types.def (PV8Q, V8QI): New pointer type.
29093         (VOID_FTYPE_PV32QI_V32HI_USI, VOID_FTYPE_PV8QI_V8HI_UQI,
29094         VOID_FTYPE_PV16QI_V16HI_UHI): New function types.
29095         * config/i386/i386-builtin.def (__builtin_ia32_pmovwb128mem_mask,
29096         __builtin_ia32_pmovwb256mem_mask, __builtin_ia32_pmovswb128mem_mask,
29097         __builtin_ia32_pmovswb256mem_mask, __builtin_ia32_pmovuswb128mem_mask,
29098         __builtin_ia32_pmovuswb256mem_mask,
29099         __builtin_ia32_pmovuswb512mem_mask, __builtin_ia32_pmovswb512mem_mask)
29100         __builtin_ia32_pmovwb512mem_mask): New builtins.
29102 2017-08-08  Julia Koval  <julia.koval@intel.com>
29104         PR target/73350,80862
29105         * config/i386/subst.md (round): Fix round pattern.
29106         * config/i386/i386.c (ix86_erase_embedded_rounding):
29107         Fix erasing rounding for the fixed pattern.
29109 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
29111         * cfgbuild.c (find_many_sub_basic_blocks): Fix thinko.
29113 2017-06-08  Martin Liska  <mliska@suse.cz>
29115         PR gcov-profile/80911
29116         * gcov.c (block_info::block_info): New constructor.
29118 2017-06-07  Carl Love  <cel@us.ibm.com>
29120         * config/rs6000/rs6000-c: The return type of the following
29121         built-in functions was implemented as int not long long.  Fix sign
29122         of return value for the unsigned version of vec_mulo and vec_mule.
29123         vector unsigned long long vec_bperm (vector unsigned long long,
29124                                              vector unsigned char)
29125         vector signed long long vec_mule (vector signed int,
29126                                           vector signed int)
29127         vector unsigned long long vec_mule (vector unsigned int,
29128                                             vector unsigned int)
29129         vector signed long long vec_mulo (vector signed int,
29130                                           vector signed int)
29131         vector unsigned long long vec_mulo (vector unsigned int,
29132                                             vector unsigned int)
29133         * doc/extend.texi: Fix the documentation for the built-in
29134         functions.
29136 2017-06-07  Carl Love  <cel@us.ibm.com>
29138         PR target/80982
29139         * config/rs6000/altivec.md (double<mode>2): Fix the implementation of
29140         for BE.
29142 2017-06-07  Carl Love  <cel@us.ibm.com>
29144         * config/rs6000/altivec.md: Fix argument swizzle in vec_doublel
29145         support, Generate       doublehv for signed int/float for BE case only.
29147 2017-06-07  Alexander Monakov  <amonakov@ispras.ru>
29149         * doc/invoke.texi (mcx16): Rewrite.
29151 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29153         * config/rs6000/predicates.md (rs6000_nonimmediate_operand): Delete.
29154         * config/rs6000/rs6000.md (*movsi_internal1, movsi_from_sf,
29155         *mov<mode>_softfloat, and an anonymous splitter): Use
29156         nonimmediate_operand instead of rs6000_nonimmediate_operand.
29158 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29160         * config/rs6000/darwin.h (REGISTER_NAMES): Delete the SPE_ACC and
29161         SPEFSCR registers.
29162         * config/rs6000/rs6000.c (rs6000_reg_names, alt_reg_names): Ditto.
29163         (enum rs6000_reg_type): Delete SPE_ACC_TYPE and SPEFSCR_REG_TYPE.
29164         (rs6000_debug_reg_global): Adjust.
29165         (rs6000_init_hard_regno_mode_ok): Adjust.
29166         (rs6000_dbx_register_number): Adjust.
29167         * config/rs6000/rs6000.h (FIRST_PSEUDO_REGISTER): Change to 115.
29168         (FIXED_REGISTERS, CALL_USED_REGISTERS, CALL_REALLY_USED_REGISTERS):
29169         Remove SPE_ACC and SPEFSCR.
29170         (REG_ALLOC_ORDER): Ditto.
29171         (FRAME_POINTER_REGNUM): Change to 111.
29172         (enum reg_class): Remove the SPE_ACC and SPEFSCR registers.
29173         (REG_CLASS_NAMES): Ditto.
29174         (REG_CLASS_CONTENTS): Delete the SPE_ACC and SPEFSCR registers.
29175         (REGISTER_NAMES): Ditto.
29176         (ADDITIONAL_REG_NAMES): Ditto.
29177         (rs6000_reg_names): Ditto.
29178         * config/rs6000/rs6000.md: Renumber some register number
29179         define_constants.
29181 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29183         * config/rs6000/darwin.h (REGISTER_NAMES): Delete the SPE high
29184         registers.
29185         * config/rs6000/rs6000.c (rs6000_reg_names, alt_reg_names): Ditto.
29186         * config/rs6000/rs6000.h (FIRST_PSEUDO_REGISTER): Change from 149
29187         to 117.
29188         (DWARF_REG_TO_UNWIND_COLUMN): Do not define.
29189         (FIXED_REGISTERS, CALL_USED_REGISTERS, CALL_REALLY_USED_REGISTERS):
29190         Delete the SPE high registers.
29191         (REG_ALLOC_ORDER): Ditto.
29192         (enum reg_class): Remove SPE_HIGH_REGS.
29193         (REG_CLASS_NAMES): Ditto.
29194         (REG_CLASS_CONTENTS): Delete the SPE high registers.
29195         (REGISTER_NAMES): Ditto.
29196         (rs6000_reg_names): Ditto.
29197         * doc/tm.texi.in: Remove SPE as example.
29198         * doc/tm.texi: Regenerate.
29200 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29202         * config/rs6000/8540.md (ppc8540_brinc): Delete.
29203         * config/rs6000/e500mc.md (e500mc_brinc): Delete.
29204         * config/rs6000/e500mc64.md (e500mc64_brinc): Delete.
29205         * config/rs6000/rs6000.md (type): Remove "brinc".
29207 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29209         * config.gcc (powerpc*-*-*): Don't add spe.h to extra_headers.
29210         (powerpc*-linux*spe*): Use ${cpu_type} instead of rs6000.
29211         * config/rs6000/linuxspe.h: Delete file.
29212         * config/rs6000/rs6000.md: Don't include spe.md.
29213         * config/rs6000/spe.h: Delete file.
29214         * config/rs6000/spe.md: Delete file.
29215         * config/rs6000/t-rs6000: Remove spe.md.
29217 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29219         * config/rs6000/predicates.md (reg_or_mem_operand): Reformat.
29220         (reg_or_none500mem_operand): Delete.
29221         * config/rs6000/rs6000.md (extendsfdf2): Use reg_or_mem_operand
29222         instead of reg_or_none500mem_operand.
29224 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29226         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
29227         handling of SPE flags.
29228         * config/rs6000/rs6000.opt (-mspe, -mspe=no, -mspe=yes): Delete.
29230 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29232         * config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
29233         SPE ABI handling.
29234         * config/rs6000/paired.md (paired_negv2sf2): Rename to negv2sf2.
29235         (paired_absv2sf2, paired_addv2sf3, paired_subv2sf3, paired_mulv2sf3,
29236         paired_divv2sf3): Similar.
29237         * config/rs6000/predicates.md: Replace TARGET_SPE, TARGET_SPE_ABI,
29238         SPE_VECTOR_MODE and SPE_HIGH_REGNO_P by 0; simplify.
29239         * config/rs6000/rs6000-builtin.def: Delete RS6000_BUILTIN_E and
29240         RS6000_BUILTIN_S.
29241         Delete BU_SPE_1, BU_SPE_2, BU_SPE_3, BU_SPE_E, BU_SPE_P, and BU_SPE_X.
29242         Rename the paired_* instruction patterns.
29243         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Do not
29244         define __SPE__.
29245         * config/rs6000/rs6000-protos.h (invalid_e500_subreg): Delete.
29246         * config/rs6000/rs6000.c: Delete RS6000_BUILTIN_E and RS6000_BUILTIN_S.
29247         (struct rs6000_stack): Delete fields spe_gp_save_offset, spe_gp_size,
29248         spe_padding_size, and spe_64bit_regs_used.  Replace TARGET_SPE and
29249         TARGET_SPE_ABI with 0, simplify.  Replace SPE_VECTOR_MODE with
29250         PAIRED_VECTOR_MODE.
29251         (struct machine_function): Delete field spe_insn_chain_scanned_p.
29252         (spe_func_has_64bit_regs_p): Delete.
29253         (spe_expand_predicate_builtin): Delete.
29254         (spe_expand_evsel_builtin): Delete.
29255         (TARGET_DWARF_REGISTER_SPAN): Do not define.
29256         (TARGET_MEMBER_TYPE_FORCES_BLK): Do not define.
29257         (invalid_e500_subreg): Delete.
29258         (rs6000_legitimize_address): Always force_reg op2 as well, for
29259         paired single memory accesses.
29260         (rs6000_member_type_forces_blk): Delete.
29261         (rs6000_spe_function_arg): Delete.
29262         (rs6000_expand_unop_builtin): Delete SPE handling.
29263         (rs6000_expand_binop_builtin): Ditto.
29264         (spe_expand_stv_builtin): Delete.
29265         (bdesc_2arg_spe): Delete.
29266         (spe_expand_builtin): Delete.
29267         (spe_expand_predicate_builtin): Delete.
29268         (spe_expand_evsel_builtin): Delete.
29269         (rs6000_invalid_builtin): Remove RS6000_BTM_SPE handling.
29270         (spe_init_builtins): Delete.
29271         (spe_func_has_64bit_regs_p): Delete.
29272         (savres_routine_name): Delete "info" parameter.  Adjust callers.
29273         (rs6000_emit_stack_reset): Ditto.
29274         (rs6000_dwarf_register_span): Delete.
29275         * config/rs6000/rs6000.h (TARGET_SPE_ABI, TARGET_SPE,
29276         UNITS_PER_SPE_WORD, SPE_HIGH_REGNO_P, SPE_SIMD_REGNO_P,
29277         SPE_VECTOR_MODE, RS6000_BTM_SPE, RS6000_BUILTIN_E, RS6000_BUILTIN_S):
29278         Delete.
29279         * config/rs6000/rs6000.md (FIRST_SPE_HIGH_REGNO, LAST_SPE_HIGH_REGNO):
29280         Delete.
29281         * config/rs6000/rs6000.opt (-mabi=spe, -mabi=no-spe): Delete.
29282         * config/rs6000/spe.md: Delete every pattern that uses TARGET_SPE.
29283         * config/rs6000/vector.md (absv2sf2, negv2sf2, addv2sf3, subv2sf3,
29284         mulv2sf3, divv2sf3): Delete expanders.
29286 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29288         config/rs6000/rs6000.md (UNSPEC_MV_CR_GT): Delete.
29290 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29292         * config/rs6000/rs6000-protos.h (output_e500_flip_gt_bit): Delete.
29293         * config/rs6000/rs6000.c: Ditto.
29295 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29297         * config/rs6000/predicated.md (rs6000_cbranch_operator): Delete.
29298         * config/rs6000/rs6000.md: Replace rs6000_cbranch_operator by
29299         comparison_operator.
29301 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29303         * config/rs6000/rs6000.c: Remove everything related to -mfloat-gprs.
29304         * config/rs6000/rs6000.opt: Ditto.
29305         * config/rs6000/t-rtems: Ditto.
29307 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29309         * config/rs6000/predicates.md: Replace TARGET_E500_DOUBLE and
29310         TARGET_E500_SINGLE by 0, simplify.
29311         * config/rs6000/rs6000.c: Ditto.
29312         (rs6000_option_override_internal): Delete CHECK_E500_OPTIONS.
29313         (spe_build_register_parallel): Delete.
29314         * config/rs6000/rs6000.h: Delete TARGET_E500_SINGLE,
29315         TARGET_E500_DOUBLE, and CHECK_E500_OPTIONS.
29316         * config/rs6000/rs6000.md: Replace TARGET_E500_DOUBLE,
29317         TARGET_E500_SINGLE, and <E500_CONVERT> by 0, simplify.
29318         (E500_CONVERT): Delete.
29319         * config/rs6000/spe.md: Remove many patterns and all define_constants.
29321 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29323         * config/rs6000/darwin.md: Replace TARGET_FPRS by 1 and simplify.
29324         * config/rs6000/dfp.md: Ditto.
29325         (negdd2, *negdd2_fpr): Merge.
29326         (absdd2, *absdd2_fpr): Merge.
29327         (negtd2, *negtd2_fpr): Merge.
29328         (abstd2, *abstd2_fpr): Merge.
29329         * config/rs6000/e500.h: Delete file.
29330         * config/rs6000/predicates.md (rs6000_cbranch_operator): Replace
29331         TARGET_FPRS by 1 and simplify.
29332         * config/rs6000/rs6000-c.c: Ditto.
29333         * config/rs6000/rs6000.c: Ditto.  Also replace TARGET_SF_SPE and
29334         TARGET_DF_SPE by 0.
29335         * config/rs6000/rs6000.h: Ditto.  Delete TARGET_SF_SPE and
29336         TARGET_DF_SPE.
29337         * config/rs6000/rs6000.md: Ditto.
29338         (floatdidf2, *floatdidf2_fpr): Merge.
29339         (move_from_CR_gt_bit): Delete.
29340         * config/rs6000/spe.md: Replace TARGET_FPRS by 1 and simplify.
29341         (E500_CR_IOR_COMPARE): Delete.
29342         (All patterns that require !TARGET_FPRS): Delete.
29343         * config/rs6000/vsx.md: Replace TARGET_FPRS by 1 and simplify.
29345 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29347         * passes.def (pass_iv_canon): Move before pass_loop_distribution.
29349 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29351         * graphds.c (add_edge): Intitialize edge's attached data.
29352         (foll_in_subgraph, dfs_fst_edge, dfs_next_edge): New function
29353         pointer parameter.  Call pointed function on each edge during
29354         graph traversing.  Skip traversing the edge when the function
29355         returns true.
29356         (graphds_dfs, graphds_scc): Ditto.
29357         (for_each_edge): New parameter.  Pass the new parameter to callback
29358         function.
29359         * graphds.h (skip_edge_callback): New function pointer type.
29360         (graphds_dfs, graphds_scc): New function pointer parameter.
29361         (graphds_edge_callback, for_each_edge): New parameter.
29363 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29365         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test): Factor
29366         out code checking if runtime alias check is possible to below ...
29367         Call the new function.
29368         * tree-data-ref.c (runtime_alias_check_p): ... to new function.
29369         * tree-data-ref.h (runtime_alias_check_p): New decalaration.
29371 2017-06-07  Marek Polacek  <polacek@redhat.com>
29373         PR sanitizer/80932
29374         * fold-const.c (extract_muldiv_1) <case MINUS_EXPR>: Add
29375         TYPE_OVERFLOW_WRAPS check.
29377 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29379         * tree-vect-loop-manip.c (vect_do_peeling): Don't skip vector loop
29380         if versioning is required.
29381         * tree-vect-loop.c (vect_analyze_loop_2): Merge niter check for loop
29382         peeling with the check for versioning.
29384 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29386         * tree-vectorizer.h (vect_build_loop_niters): New parameter.
29387         * tree-vect-loop-manip.c (vect_build_loop_niters): New parameter.
29388         Set true to new parameter if new ssa variable is defined.
29389         (vect_gen_vector_loop_niters): Refactor.  Set range information
29390         for the new vector loop bound variable.
29391         (vect_do_peeling): Ditto.
29393 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29395         * tree-affine.c (ssa.h): Include header file.
29396         (tree_to_aff_combination): Handle (T1)(X - CST) when inner type
29397         has wrapping overflow behavior.
29399 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29401         * tree-affine.c (tree_to_aff_combination): Handle (T1)(X + X).
29403 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29405         (aff_combination_expand): Move (T1)(X *+- CST) simplification to ...
29406         (tree_to_aff_combination): ... here.
29408 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29410         * tree-ssa-loop-ivopts.c (ivopts_estimate_reg_pressure): New
29411         reg_pressure model function.
29412         (ivopts_global_cost_for_size): Delete.
29413         (determine_set_costs, iv_ca_recount_cost): Call new model function
29414         ivopts_estimate_reg_pressure.
29416 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29418         * config/aarch64/aarch64.c (aarch64_rtx_costs): Make sdiv more
29419         expensive than udiv.  Remove floating point cases from mod.
29421 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29423         * config/arm/aarch-cost-tables.h (cortexa53_extra_cost):
29424         Increase idiv cost.
29426 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29428         * config/aarch64/aarch64.md
29429         (copysignsf3): Fix mask generation.
29431 2017-06-07  Jakub Jelinek  <jakub@redhat.com>
29433         * dumpfile.h (enum tree_dump_index): Rename TDI_generic to
29434         TDI_gimple.
29435         (class dump_manager): Add register_dumps method.
29436         * dumpfile.c: Include langhooks.h.
29437         (dump_files): Use 0 instead of 3/4/5 for TDI_{original,gimple,nested}.
29438         (FIRST_AUTO_NUMBERED_DUMP): Decrease to 1.
29439         (FIRST_ME_AUTO_NUMBERED_DUMP): Define.
29440         (dump_manager::dump_register): Start with 512 entries instead of 32.
29441         (dump_manager::register_dumps): New method.
29442         * toplev.c (general_init): Instead of invoking register_dumps
29443         langhook, invoke register_dumps method on the dump manager.
29444         * gimplify.c (gimplify_function_tree): Use TDI_gimple instead of
29445         TDI_generic.
29447 2017-06-07  Richard Sandiford  <richard.sandiford@linaro.org>
29449         * doc/md.texi: Clarify the restrictions on a define_insn condition.
29450         Say that # requires an associated define_split to exist, and that
29451         the define_split must be suitable for use after register allocation.
29453 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29455         * cfgbuild.c (find_bb_boundaries): Initialize profile of split blocks.
29456         (compute_outgoing_frequencies): Also initialize zero counts.
29457         (find_many_sub_basic_blocks): Do not produce uninitialized profile
29458         around loops; preserve more of profile when nothing changes.
29460 2017-06-06  Jim Wilson  <jim.wilson@linaro.org>
29462         * config/aarch64/aarch64-cost-tables.h (qdf24xx_extra_costs): Move to
29463         here.
29464         * config/arm/aarch-cost-tables.h (qdf24xx_extra_costs): From here.
29465         * config/arm/arm-cpu-cdata.h: Regenerate.
29466         * config/arm/arm-cpu-data.h, config/arm/arm-cpu.h: Likewise.
29467         * config/arm/arm-tables.opt, config/arm/arm-tune.md: Likewise.
29468         * config/arm/arm-cpus.in: Delete falkor and qdf24xx entries.
29469         * config/arm/arm.c (arm_qdf24xx_tune): Delete.
29470         * config/arm/bpabi.h (BE8_LINK_SPEC): Delete falkor and qdf24xx
29471         support.
29472         * config/arm/t-aprofile (MULTILIB_MATCHES): Delete falkor and qdf24xx
29473         support.
29474         * config/arm/t-rmprofile: Likewise.
29475         * doc/invoke.texi (ARM Options): Drop falkor and qdf24xx support.
29477 2017-06-06  David S. Miller  <davem@davemloft.net>
29479         PR target/80968
29480         * config/sparc/sparc.c (sparc_expand_prologue): Emit frame
29481         blockage if function uses alloca.
29483 2017-06-06  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
29485         * tree-ssa-loop-prefetch.c (struct mem_ref_group, struct mem_ref):
29486         New "uid" fields to hold pretty-print IDs of group and ref.
29487         Memory references are now identified as <group_id>:<ref_id>
29488         instead of using [random] addresses.
29489         (dump_mem_details): Simplify, no functional change.
29490         (dump_mem_ref): Simplify and make output more concise.
29491         Replace couple of fprintf's throughout code with calls to dump_mem_ref.
29492         (find_or_create_group): Initialize group uid.
29493         (record_ref): Initialize ref uid.  Improve debug output.
29494         (prune_group_by_reuse, should_issue_prefetch_p,)
29495         (should_issue_prefetch_p, schedule_prefetches, issue_prefetch_ref,)
29496         (mark_nontemporal_store, determine_loop_nest_reuse):
29497         Improve debug output.
29499 2017-06-06  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
29501         * dbgcnt.def (prefetch): New debug counter.
29502         * tree-ssa-loop-prefetch.c (dbgcnt.h): New include.
29503         (schedule_prefetches): Stop issueing prefetches if debug counter
29504         tripped.
29506 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29508         * doc/sourcebuild.texi (Testsuites, C Language Testsuites,
29509         gcc.c-torture/compile): Remove mention of NO_LABEL_VALUES in fixme.
29511 2017-06-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29513         * config/aarch64/atomics.md (atomic_compare_and_swap<mode> expander):
29514         Use aarch64_reg_or_zero predicate for operand 4.
29515         (aarch64_compare_and_swap<mode> define_insn_and_split):
29516         Use aarch64_reg_or_zero predicate for operand 3.  Add 'Z' constraint.
29517         (aarch64_store_exclusive<mode>): Likewise for operand 2.
29519 2017-06-06  Thomas Preud'homme  <thomas.preudhomme@arm.com>
29521         * config/arm/arm.c (arm_compute_save_reg_mask): Rename into ...
29522         (arm_compute_save_core_reg_mask): This.
29523         (thumb1_compute_save_reg_mask): Rename into ...
29524         (thumb1_compute_save_core_reg_mask): This.
29525         (arm_compute_save_reg0_reg12_mask): Adapt comment.
29526         (arm_compute_frame_layout): Likewise.
29528 2017-06-06  Richard Biener  <rguenther@suse.de>
29530         PR tree-optimization/80974
29531         * tree-ssa-sccvn.c (set_ssa_val_to): Do not change but only
29532         keep or clear leaders SSA info.
29534 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29536         * config/nvptx/nvptx.c (split_mode_p): New function.
29537         (nvptx_declare_function_name, nvptx_print_operand): Use split_mode_p.
29539 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29541         * config/nvptx/nvptx.c (nvptx_print_operand): Use maybe_split_mode.
29543 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29545         PR bootstrap/80978
29546         * tree-cfg.c (execute_fixup_cfg): Fix condition on when to rescale
29547         profile.
29549 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29551         * shrink-wrap.c (handle_simple_exit): Update profile.
29552         (try_shrink_wrapping): Upate profile.
29554 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29556         * predict.c (tree_estimate_probability_bb): Add LOCAL_ONLY.
29557         (tree_guess_outgoing_edge_probabilities): New.
29558         * predict.h (tree_guess_outgoing_edge_probabilities): Declare.
29559         * tree-cfg.c (gimple_find_sub_bbs): Propagate profile.
29561 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29563         * ipa-split.c (split_function): Initialize return bb profile.
29565 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29567         * profile.c (compute_branch_probabilities): Also initialize
29568         EXIT_BLOCK profile.
29570 2017-06-06  Richard Biener  <rguenther@suse.de>
29572         PR tree-optimization/80928
29573         * tree-vect-loop.c (vect_update_vf_for_slp): Amend dumps.
29574         (vect_analyze_loop_operations): Properly guard analysis for
29575         pure SLP case.
29576         (vect_transform_loop): Likewise.
29577         (vect_analyze_loop_2): Also reset SLP type on PHIs.
29578         (vect_model_induction_cost): Do not cost for pure SLP.
29579         (vectorizable_induction): Pass in SLP node, implement SLP vectorization
29580         of induction in inner loop vectorization.
29581         * tree-vect-slp.c (vect_create_new_slp_node): Handle PHIs.
29582         (vect_get_and_check_slp_defs): Handle vect_induction_def.
29583         (vect_build_slp_tree): Likewise.  Handle PHIs as terminating the
29584         recursion.
29585         (vect_analyze_slp_cost_1): Cost induction.
29586         (vect_detect_hybrid_slp_stmts): Handle PHIs.
29587         (vect_get_slp_vect_defs): Likewise.
29588         * tree-vect-stmts.c (vect_analyze_stmt): Handle induction.
29589         (vect_transform_stmt): Handle SLP reductions.
29590         * tree-vectorizer.h (vectorizable_induction): Adjust.
29592 2017-06-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
29594         * config/rs6000/rs6000.c (make_resolver_func): Update
29595         init_lowered_empty_function call.
29597 2017-06-05  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29599         * doc/invoke.texi: Document the -fprofile-abs-path option.
29600         * common.opt (fprofile-abs-path): New option.
29601         * gcov-io.h (gcov_write_filename): Declare.
29602         * gcov-io.c (gcov_write_filename): New function.
29603         * coverage.c (coverage_begin_function): Use gcov_write_filename.
29604         * profile.c (output_location): Likewise.
29606 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29608         * shring-wrap.c: Revert accidental commit.
29610 2017-06-05  Volker Reichelt  <v.reichelt@netcologne.de>
29612         * doc/invoke.texi (-Wduplicated-branches): Add to warning list.
29614 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29616         * cfgexpand.c (expand_gimple_tailcall): Initialize profile of
29617         new edge.
29618         * ipa-inline.c (want_inline_self_recursive_call_p): Watch for missing
29619         profile in callgraph edge.
29620         * profile-count.h (apply_probability): If THIS is 0, then result is 0
29621         (apply_scale): Likewise.
29622         * tree-inline.c (copy_bb, copy_edges_for_bb, copy_cfg_body):
29623         Also scale profile when inlining function with zero profile.
29624         (initialize_cfun): Update exit block profile even when it is zero.
29625         * tree-ssa-threadupdate.c (clear_counts_path): Handle correctly case
29626         when profile is read.
29628 2017-06-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
29630         * config/rs6000/rs6000.c (toplevel): Include attribs.h.
29631         (CLONE_*): New constants to define the processors we can generate
29632         code for with the target_clone attribute.
29633         (rs6000_clone_map): New array to identify which clone processors
29634         the current program is running on.
29635         (TARGET_COMPARE_VERSION_PRIORITY): Define to enable the
29636         target_clone attribute.
29637         (TARGET_GENERATE_VERSION_DISPATCHER_BODY): Likewise.
29638         (TARGET_GET_FUNCTION_VERSIONS_DISPATCHER): Likewise.
29639         (TARGET_OPTION_FUNCTION_VERSIONS): Likewise.
29640         (cpu_expand_builtin): Add support for target_clone attribute.
29641         (rs6000_valid_attribute_p): Allow "default" attribute.
29642         (get_decl_name): New debug function to simplify printing the
29643         current function name in debugging statements.
29644         (rs6000_clone_priority): New functions to support the target_clone
29645         attribute, and be able to generate code to switch between ISA 2.05
29646         through ISA 3.0 (power6 through power9).
29647         (rs6000_compare_version_priority): Likewise.
29648         (rs6000_get_function_versions_dispatcher): Likewise.
29649         (make_resolver_func): Likewise.
29650         (add_condition_to_bb): Likewise.
29651         (dispatch_function_versions): Likewise.
29652         (rs6000_generate_version_dispatcher_body): Likewise.
29653         (rs6000_can_inline_p): Call get_decl_name for debugging usage.
29654         (fusion_gpr_load_p): Fix a spacing issue.
29655         * doc/extend.texi (Common Function Attributes): Document that the
29656         PowerPC supports the target_clone attribute.
29658 2017-06-05  Thomas Preud'homme  <thomas.preudhomme@arm.com>
29660         * config/arm/arm.h: explain F symbol found in description of ARM
29661         register allocation in its legend.
29663 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29665         * config/mips/frame-header-opt.c: Include profile-count.h.
29666         * config/riscv/riscv.c: Include profile-count.h
29668 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29670         * tree-ssa-loop-im.c (execute_sm_if_changed): Add FLAG_BBS parameter;
29671         update profile.
29672         (sm_set_flag_if_changed): Add bbs field.
29673         (execute_sm_if_changed_flag_set): Pass BBS.
29674         (execute_sm): Update.
29676 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29678         * config/aarch64/aarch64-simd.md (aarch64_store_lane0<mode>):
29679         New pattern.
29681 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29683         * config/aarch64/aarch64.md (sub<mode>3_compare1_imm): New define_insn.
29684         (peephole2): New peephole2 to emit the above.
29685         * config/aarch64/predicates.md (aarch64_sub_immediate): New predicate.
29687 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29689         * config/aarch64/aarch64.c (define_peephole2 above
29690         *sub_<shift>_<mode>): New peephole.
29692 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
29694         * config/i386/i386.c (make_resolver_func): Update.
29695         * Makefile.in: Add profile-count.h and profile-count.o
29696         * auto-profile.c (afdo_indirect_call): Update to new API.
29697         (afdo_set_bb_count): Update.
29698         (afdo_propagate_edge): Update.
29699         (afdo_propagate_circuit): Update.
29700         (afdo_calculate_branch_prob): Update.
29701         (afdo_annotate_cfg): Update.
29702         * basic-block.h: Include profile-count.h
29703         (struct edge_def): Turn count to profile_count.
29704         (struct basic_block_def): Likewie.
29705         (REG_BR_PROB_BASE): Move to profile-count.h
29706         (RDIV): Move to profile-count.h
29707         * bb-reorder.c (max_entry_count): Turn to profile_count.
29708         (find_traces): Update.
29709         (rotate_loop):Update.
29710         (connect_traces):Update.
29711         (sanitize_hot_paths):Update.
29712         * bt-load.c (migrate_btr_defs): Update.
29713         * cfg.c (RDIV): Remove.
29714         (init_flow): Use alloc_block.
29715         (alloc_block): Uninitialize count.
29716         (unchecked_make_edge): Uninitialize count.
29717         (check_bb_profile): Update.
29718         (dump_edge_info): Update.
29719         (dump_bb_info): Update.
29720         (update_bb_profile_for_threading): Update.
29721         (scale_bbs_frequencies_int): Update.
29722         (scale_bbs_frequencies_gcov_type): Update.
29723         (scale_bbs_frequencies_profile_count): New.
29724         * cfg.h (update_bb_profile_for_threading): Update.
29725         (scale_bbs_frequencies_profile_count): Declare.
29726         * cfgbuild.c (compute_outgoing_frequencies): Update.
29727         (find_many_sub_basic_blocks): Update.
29728         * cfgcleanup.c (try_forward_edges): Update.
29729         (try_crossjump_to_edge): Update.
29730         * cfgexpand.c (expand_gimple_tailcall): Update.
29731         (construct_exit_block): Update.
29732         * cfghooks.c (verify_flow_info): Update.
29733         (dump_bb_for_graph): Update.
29734         (split_edge): Update.
29735         (make_forwarder_block): Update.
29736         (duplicate_block): Update.
29737         (account_profile_record): Update.
29738         * cfgloop.c (find_subloop_latch_edge_by_profile): Update.
29739         (get_estimated_loop_iterations): Update.
29740         * cfgloopanal.c (expected_loop_iterations_unbounded): Update.
29741         (single_likely_exit): Update.
29742         * cfgloopmanip.c (scale_loop_profile): Update.
29743         (loopify): Update.
29744         (set_zero_probability): Update.
29745         (lv_adjust_loop_entry_edge): Update.
29746         * cfgrtl.c (force_nonfallthru_and_redirect): Update.
29747         (purge_dead_edges): Update.
29748         (rtl_account_profile_record): Update.
29749         * cgraph.c (cgraph_node::create): Uninitialize count.
29750         (symbol_table::create_edge): Uninitialize count.
29751         (cgraph_update_edges_for_call_stmt_node): Update.
29752         (cgraph_edge::dump_edge_flags): Update.
29753         (cgraph_node::dump): Update.
29754         (cgraph_edge::maybe_hot_p): Update.
29755         * cgraph.h: Include profile-count.h
29756         (create_clone), create_edge, create_indirect_edge): Update.
29757         (cgraph_node): Turn count to profile_count.
29758         (cgraph_edge0: Likewise.
29759         (make_speculative, clone): Update.
29760         (create_edge): Update.
29761         (init_lowered_empty_function): Update.
29762         * cgraphclones.c (cgraph_edge::clone): Update.
29763         (duplicate_thunk_for_node): Update.
29764         (cgraph_node::create_clone): Update.
29765         * cgraphunit.c (cgraph_node::analyze): Update.
29766         (cgraph_node::expand_thunk): Update.
29767         * final.c (dump_basic_block_info): Update.
29768         * gimple-streamer-in.c (input_bb): Update.
29769         * gimple-streamer-out.c (output_bb): Update.
29770         * graphite.c (print_global_statistics): Update.
29771         (print_graphite_scop_statistics): Update.
29772         * hsa-brig.c: Include basic-block.h.
29773         * hsa-dump.c: Include basic-block.h.
29774         * hsa-gen.c (T sum_slice): Update.
29775         (convert_switch_statements):Update.
29776         * hsa-regalloc.c: Include basic-block.h.
29777         * ipa-chkp.c (chkp_produce_thunks): Update.
29778         * ipa-cp.c (struct caller_statistics): Update.
29779         (init_caller_stats): Update.
29780         (gather_caller_stats): Update.
29781         (ipcp_cloning_candidate_p): Update.
29782         (good_cloning_opportunity_p): Update.
29783         (get_info_about_necessary_edges): Update.
29784         (dump_profile_updates): Update.
29785         (update_profiling_info): Update.
29786         (update_specialized_profile): Update.
29787         (perhaps_add_new_callers): Update.
29788         (decide_about_value): Update.
29789         (ipa_cp_c_finalize): Update.
29790         * ipa-devirt.c (struct odr_type_warn_count): Update.
29791         (struct decl_warn_count): Update.
29792         (struct final_warning_record): Update.
29793         (possible_polymorphic_call_targets): Update.
29794         (ipa_devirt): Update.
29795         * ipa-fnsummary.c (redirect_to_unreachable): Update.
29796         * ipa-icf.c (sem_function::merge): Update.
29797         * ipa-inline-analysis.c (do_estimate_edge_time): Update.
29798         * ipa-inline.c (compute_uninlined_call_time): Update.
29799         (compute_inlined_call_time): Update.
29800         (want_inline_small_function_p): Update.
29801         (want_inline_self_recursive_call_p): Update.
29802         (edge_badness): Update.
29803         (lookup_recursive_calls): Update.
29804         (recursive_inlining): Update.
29805         (inline_small_functions): Update.
29806         (dump_overall_stats): Update.
29807         (dump_inline_stats): Update.
29808         * ipa-profile.c (ipa_profile_generate_summary): Update.
29809         (ipa_propagate_frequency): Update.
29810         (ipa_profile): Update.
29811         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
29812         * ipa-utils.c (ipa_merge_profiles): Update.
29813         * loop-doloop.c (doloop_modify): Update.
29814         * loop-unroll.c (report_unroll): Update.
29815         (unroll_loop_runtime_iterations): Update.
29816         * lto-cgraph.c (lto_output_edge): Update.
29817         (lto_output_node): Update.
29818         (input_node): Update.
29819         (input_edge): Update.
29820         (merge_profile_summaries): Update.
29821         * lto-streamer-in.c (input_cfg): Update.
29822         * lto-streamer-out.c (output_cfg): Update.
29823         * mcf.c (create_fixup_graph): Update.
29824         (adjust_cfg_counts): Update.
29825         (sum_edge_counts): Update.
29826         * modulo-sched.c (sms_schedule): Update.
29827         * postreload-gcse.c (eliminate_partially_redundant_load): Update.
29828         * predict.c (maybe_hot_count_p): Update.
29829         (probably_never_executed): Update.
29830         (dump_prediction): Update.
29831         (combine_predictions_for_bb): Update.
29832         (propagate_freq): Update.
29833         (handle_missing_profiles): Update.
29834         (counts_to_freqs): Update.
29835         (rebuild_frequencies): Update.
29836         (force_edge_cold): Update.
29837         * predict.h: Include profile-count.h
29838         (maybe_hot_count_p, counts_to_freqs): UPdate.
29839         * print-rtl-function.c: Do not include cfg.h
29840         * print-rtl.c: Include basic-block.h
29841         * profile-count.c: New file.
29842         * profile-count.h: New file.
29843         * profile.c (is_edge_inconsistent): Update.
29844         (correct_negative_edge_counts): Update.
29845         (is_inconsistent): Update.
29846         (set_bb_counts): Update.
29847         (read_profile_edge_counts): Update.
29848         (compute_frequency_overlap): Update.
29849         (compute_branch_probabilities): Update; Initialize and deinitialize
29850         gcov_count tables.
29851         (branch_prob): Update.
29852         * profile.h (bb_gcov_counts, edge_gcov_counts): New.
29853         (edge_gcov_count): New.
29854         (bb_gcov_count): New.
29855         * shrink-wrap.c (try_shrink_wrapping): Update.
29856         * tracer.c (better_p): Update.
29857         * trans-mem.c (expand_transaction): Update.
29858         (ipa_tm_insert_irr_call): Update.
29859         (ipa_tm_insert_gettmclone_call): Update.
29860         * tree-call-cdce.c: Update.
29861         * tree-cfg.c (gimple_duplicate_sese_region): Update.
29862         (gimple_duplicate_sese_tail): Update.
29863         (gimple_account_profile_record): Update.
29864         (execute_fixup_cfg): Update.
29865         * tree-inline.c (copy_bb): Update.
29866         (copy_edges_for_bb): Update.
29867         (initialize_cfun): Update.
29868         (freqs_to_counts): Update.
29869         (copy_cfg_body): Update.
29870         (expand_call_inline): Update.
29871         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
29872         * tree-ssa-loop-ivcanon.c (unloop_loops): Update.
29873         (try_unroll_loop_completely): Update.
29874         (try_peel_loop): Update.
29875         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
29876         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations_loop): Update.
29877         * tree-ssa-loop-split.c (connect_loops): Update.
29878         * tree-ssa-loop-unswitch.c (hoist_guard): Update.
29879         * tree-ssa-reassoc.c (branch_fixup): Update.
29880         * tree-ssa-tail-merge.c (replace_block_by): Update.
29881         * tree-ssa-threadupdate.c (create_block_for_threading): Update.
29882         (compute_path_counts): Update.
29883         (update_profile): Update.
29884         (recompute_probabilities): Update.
29885         (update_joiner_offpath_counts): Update.
29886         (estimated_freqs_path): Update.
29887         (freqs_to_counts_path): Update.
29888         (clear_counts_path): Update.
29889         (ssa_fix_duplicate_block_edges): Update.
29890         (duplicate_thread_path): Update.
29891         * tree-switch-conversion.c (case_bit_test_cmp): Update.
29892         (struct switch_conv_info): Update.
29893         * tree-tailcall.c (decrease_profile): Update.
29894         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
29895         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
29896         * value-prof.c (check_counter): Update.
29897         (gimple_divmod_fixed_value): Update.
29898         (gimple_mod_pow2): Update.
29899         (gimple_mod_subtract): Update.
29900         (gimple_ic_transform): Update.
29901         (gimple_stringop_fixed_value): Update.
29902         * value-prof.h (gimple_ic): Update.
29904 2017-06-02  Carl Love  <cel@us.ibm.com>
29906         * config/rs6000/rs6000-c: Add support for built-in functions
29907         vector double vec_doublee (vector signed int);
29908         vector double vec_doublee (vector unsigned int);
29909         vector double vec_doublee (vector float);
29910         vector double vec_doubleh (vector signed int);
29911         vector double vec_doubleh (vector unsigned int);
29912         vector double vec_doubleh (vector float);
29913         vector double vec_doublel (vector signed int);
29914         vector double vec_doublel (vector unsigned int);
29915         vector double vec_doublel (vector float);
29916         vector double vec_doubleo (vector signed int);
29917         vector double vec_doubleo (vector unsigned int);
29918         vector double vec_doubleo (vector float);.
29919         * config/rs6000/rs6000-builtin.def: Add definitions for DOUBLEE,
29920         DOUBLEO, DOUBLEH, DOUBLEL, UNS_DOUBLEO, UNS_DOUBLEE, UNS_DOUBLEH,
29921         UNS_DOUBLEL.
29922         * config/rs6000/altivec.md: Add code generator for doublee<mode>2,
29923         unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2, doubleh<mode>2,
29924         unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2, add mode attribute
29925         VS_sxwsp.
29926         * config/rs6000/altivec.h: Add define for vec_doublee, vec_doubleo,
29927         vec_doublel, vec_doubleh.
29928         * doc/extend.texi: Update the built-in documentation file for the
29929         new built-in functions.
29931 2017-06-02  David Malcolm  <dmalcolm@redhat.com>
29933         PR jit/80954
29934         * ipa-inline-analysis.c (free_growth_caches): Set
29935         edge_removal_hook_holder to NULL after removing it.
29937 2017-06-02  Sudakshina Das  <sudi.das@arm.com>
29939         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Return CC_SWP for
29940         comparision with zero.
29942 2017-06-02  Will Schmidt  <will_schmidt@vnet.ibm.com>
29943         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
29944         for early expansion of vec_min and vec_max builtins.
29945         (builtin_function_type): Add min/max unsigned variants to those
29946         identified as having unsigned arguments.
29948 2017-06-02  Olivier Hainque  <hainque@adacore.com>
29950         * config/vx-common.h (DWARF_UNWIND_INFO): Switch #define to 1.
29952 2017-06-02  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29954         * config/aarch64/aarch64-simd.md (*aarch64_simd_vec_copy_lane<mode>):
29955         Use VALL_F16 iterator rather than VALL.
29957 2017-06-02  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29959         * config/aarch64/aarch64.c (aarch64_split_compare_and_swap):
29960         Emit CBNZ inside loop when doing a strong exchange and comparing
29961         against zero.  Generate the CC flags after the loop.
29963 2017-06-02  David Edelsohn  <dje.gcc@gmail.com>
29965         * dwarf2out.c (DWARF_INITIAL_LENGTH_SIZE_STR): New.
29966         (dl_section_ref): New.
29967         (dwarf2out_finish): Copy debug_line_section_label to dl_section_ref.
29968         On AIX, append an expression to subtract the size of the
29969         section length to dl_section_ref.
29971 2017-06-02  Will Schmidt  <will_schmidt@vnet.ibm.com>
29973         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
29974         for early expansion of vector absolute builtins.
29976 2017-06-02  Richard Biener  <rguenther@suse.de>
29978         * tree-vect-slp.c (vect_detect_hybrid_slp_2): Match up
29979         what we consider a relevant use stmt with vect_detect_hybrid_slp_stmts.
29981 2017-06-02  Richard Biener  <rguenther@suse.de>
29983         PR tree-optimization/80948
29984         * tree-tailcall.c (find_tail_calls): Track stmts to move in
29985         stmt order as well.
29987 2017-06-02  Richard Biener  <rguenther@suse.de>
29989         * tree-vect-loop.c (vect_analyze_loop_operations): Not relevant
29990         PHIs are ok.
29991         * tree-vect-stmts.c (process_use): Do not mark backedge defs
29992         for inductions as relevant.
29994 2017-06-02  Richard Biener  <rguenther@suse.de>
29996         * tree-vect-loop.c (get_initial_def_for_induction): Inline into ...
29997         (vectorizable_induction): ... this.  Remove dead code.
29999 2017-06-02  Eric Botcazou  <ebotcazou@adacore.com>
30001         * builtins. (expand_builtin_alloca): Remove second parameter and
30002         infer its value from the first parameter instead.
30003         (expand_builtin) <BUILT_IN_ALLOCA>: Adjust call to above.
30005 2017-06-02  Jakub Jelinek  <jakub@redhat.com>
30007         PR rtl-optimization/80903
30008         * loop-doloop.c (add_test): Unshare sequence.
30010 2017-06-02  Bernd Edlinger  <bernd.edlinger@hotmail.de>
30012         * doc/invoke.texi: Document the -Wsizeof-pointer-div warning.
30014 2017-06-01  Bernd Edlinger  <bernd.edlinger@hotmail.de>
30016         * config/i386/i386.c (x86_64_ms_sysv_extra_clobbered_registers): Make
30017         static.
30018         (xlogue_layout::get_stack_space_used, xlogue_layout::s_instances,
30019         xlogue_layout::get_instance, logue_layout::xlogue_layout,
30020         sp_valid_at, fp_valid_at, choose_basereg): Formatting.
30021         (xlogue_layout::get_stub_rtx): Make static.
30022         (xlogue_layout::get_stub_name): Avoid const-cast, make static.
30023         (xlogue_layout::compute_stub_managed_regs): Rename to...
30024         (xlogue_layout::count_stub_managed_regs): ...this.
30025         (xlogue_layout::is_stub_managed_reg): New function.
30026         (xlogue_layout::m_stub_names): Rename to...
30027         (xlogue_layout::s_stub_names): ...this, make static.
30028         (xlogue_layout::STUB_INDEX_OFFSET, xlogue_layout::MIN_REGS,
30029         xlogue_layout::MAX_REGS, xlogue_layout::MAX_EXTRA_REGS,
30030         xlogue_layout::VARIANT_COUNT, xlogue_layout::STUB_NAME_MAX_LEN,
30031         xlogue_layout::s_stub_names): Instantiate statics.
30032         (stub_managed_regs): Remove.
30033         (ix86_save_reg): Use xlogue_layout::compute_stub_managed_regs.
30034         (disable_call_ms2sysv_xlogues): Rename to...
30035         (warn_once_call_ms2sysv_xlogues): ...this, and warn only once.
30036         (ix86_initial_elimination_offset, ix86_expand_call): Fix call_ms2sysv
30037         warning logic.
30038         (ix86_static_chain): Make sure that ix86_static_chain_on_stack can't
30039         change after reload_completed.
30040         (ix86_can_use_return_insn_p): Use the ix86_frame data structure
30041         directly.
30042         (ix86_expand_prologue): Likewise.
30043         (ix86_expand_epilogue): Likewise.
30044         (ix86_expand_split_stack_prologue): Likewise.
30045         (ix86_compute_frame_layout): Remove frame parameter ...
30046         (TARGET_COMPUTE_FRAME_LAYOUT): ... and export it as a target hook.
30047         (ix86_finalize_stack_realign_flags): Call ix86_compute_frame_layout
30048         only if necessary.
30049         (ix86_init_machine_status): Don't set use_fast_prologue_epilogue_nregs.
30050         (ix86_frame): Move from here ...
30051         * config/i386/i386.h (ix86_frame): ... to here.
30052         (machine_function): Remove use_fast_prologue_epilogue_nregs, cache the
30053         complete ix86_frame data structure instead.  Remove some_ld_name.
30055 2017-06-01  Pierre-Marie de Rodat  <derodat@adacore.com>
30057         * dwarf2out.c (dwarf2out_late_global_decl): Add locations for
30058         symbols that hold a DECL_VALUE_EXPR.
30060 2017-06-01  Martin Jambor  <mjambor@suse.cz>
30062         PR tree-optimization/80898
30063         * tree-sra.c (process_subtree_disqualification): Removed.
30064         (disqualify_candidate): Do not acll
30065         process_subtree_disqualification.
30066         (subtree_mark_written_and_enqueue): New function.
30067         (propagate_all_subaccesses): Set grp_write of LHS subtree if the
30068         RHS has been disqualified and re-queue LHS if necessary.  Apart
30069         from that, ignore disqualified RHS.
30071 2017-06-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
30073         * config/s390/s390.c (s390_emit_epilogue): Disable early return
30074         address fetch for z10 or later.
30076 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30078         * config/arc/arc.md (tst_movb): Add guard when splitting.
30080 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30082         * config/arc/arc.c (arc_can_eliminate): Test against
30083         arc_frame_pointer_needed.
30085 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30087         * config/arc/arc.c (arc_expand_prologue): Emit a special barrier
30088         to prevent store reordering.
30089         * config/arc/arc.md (UNSPEC_ARC_STKTIE): Define.
30090         (type): Add block type.
30091         (stack_tie): Define special instruction to be used in
30092         expand_prologue.
30094 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30096         * config/arc/arc.md (commutative_binary_comparison): Remove 'I'
30097         constraint. It is not valid for the pattern.
30098         (noncommutative_binary_comparison): Likewise.
30100 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30102         * config/arc/simdext.md (movv2hi_insn): Change predicate to avoid
30103         scaled addresses.
30105 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30107         * config/arc/arc.c (arc_conditional_register_usage): Allow r30 to
30108         be used by the reg-alloc.
30110 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30112         * config/arc/arc.md (mulsi3): Avoid use of hard registers before
30113         reg-alloc when having mul64 or mul32x16 instructions.
30114         (mulsidi3): Likewise.
30115         (umulsidi3): Likewise.
30116         (mulsi32x16): New pattern.
30117         (mulsi64): Likewise.
30118         (mulsidi64): Likewise.
30119         (umulsidi64): Likewise.
30120         (MUL32x16_REG): Define.
30121         (mul64_600): Use MUL32x16_REG.
30122         (mac64_600): Likewise.
30123         (umul64_600): Likewise.
30124         (umac64_600): Likewise.
30126 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30128         * config/arc/arc.md (mulsi3_700): Make it commutative.
30130 2017-06-01  Jose E. Marchesi  <jose.marchesi@oracle.com>
30132         * config/sparc/sparc.md (*zero_extendsidi2_insn_sp64): Set insn
30133         type for movstouw.
30134         (*sign_extendsidi2_insn): Likewise for movstosw.
30136 2017-06-01  Pierre-Marie de Rodat  <derodat@adacore.com>
30138         * dwarf2out.c (get_discr_value): Call the get_debug_type hook on
30139         the type of the input discriminant value.  Convert the
30140         discriminant value of signedness vary.
30142 2017-06-01  Volker Reichelt  <v.reichelt@netcologne.de>
30144         * doc/invoke.texi (-Wcatch-value): Document new shortcut.
30145         Add to -Wall section.
30147 2017-06-01  Richard Biener  <rguenther@suse.de>
30149         PR middle-end/66313
30150         * fold-const.c (fold_plusminus_mult_expr): If the factored
30151         factor may be zero use a wrapping type for the inner operation.
30152         * tree-tailcall.c (independent_of_stmt_p): Pass in to_move bitmap
30153         and handle moved defs.
30154         (process_assignment): Properly guard the unary op case.  Return a
30155         tri-state indicating that moving the stmt before the call may allow
30156         to continue.  Pass through to_move.
30157         (find_tail_calls): Handle moving unrelated defs before
30158         the call.
30160 2017-05-31  Segher Boessenkool  <segher@kernel.crashing.org>
30162         PR target/80618
30163         * config/rs6000/vector.md (*vector_uneq<mode>): Write the nor in the
30164         splitter result in the canonical way.
30166 2017-05-31  Uros Bizjak  <ubizjak@gmail.com>
30168         * config/i386/i386.md (*zero_extendsidi2): Enable alternative (?r, *Yj)
30169         also for 32bit target.  Update insn attributes.
30170         (zero-extendsidi2 splitter): Allow all registers for operand 1.
30172 2017-05-31  Sebastian Peryt  <sebastian.peryt@intel.com>
30174         * config/i386/avx512fintrin.h (_mm_mask_max_sd)
30175         (_mm_maskz_max_sd, _mm_mask_max_ss, _mm_maskz_max_ss)
30176         (_mm_mask_min_sd, _mm_maskz_min_sd, _mm_mask_min_ss)
30177         (_mm_maskz_min_ss): New intrinsics.
30179 2017-05-31  Martin Liska  <mliska@suse.cz>
30181         * tree-vect-loop.c (vect_create_epilog_for_reduction):
30182         Change comment style to one we normally use.
30183         (vectorizable_reduction): Likewise.
30184         (vectorizable_induction): Likewise.
30185         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
30186         (vectorizable_call): Likewise.
30187         (vectorizable_simd_clone_call): Likewise.
30188         (vectorizable_conversion): Likewise.
30189         (vectorizable_assignment): Likewise.
30190         (vectorizable_shift): Likewise.
30191         (vectorizable_operation): Likewise.
30192         (vectorizable_store): Likewise.
30193         (vectorizable_load): Likewise.
30194         * tree-vectorizer.h: Likewise.
30196 2017-05-31  Alexander Monakov  <amonakov@ispras.ru>
30198         * passes.c (emergency_dump_function): New.
30199         * tree-pass.h (emergency_dump_function): Declare.
30200         * plugin.c (plugins_internal_error_function): Remove.
30201         * plugin.h (plugins_internal_error_function): Remove declaration.
30202         * toplev.c (internal_error_function): New static function.  Use it...
30203         (general_init): ...here.
30205 2017-05-31  Graham Markall  <graham.markall@embecosm.com>
30207         * config/arc/arc.c (arc_print_operand): Handle constant operands.
30208         (arc_rtx_costs): Add costs for new patterns.
30209         * config/arc/arc.md: Additional *add_n and *sub_n patterns.
30210         * config/arc/predicates.md: Add _1_2_3_operand predicate.
30212 2017-05-31  Richard Sandiford  <richard.sandiford@linaro.org>
30214         * tree-ssa-strlen.c (get_next_strinfo): New function.
30215         (get_stridx_plus_constant): Use it.
30216         (zero_length_string): Likewise.
30217         (adjust_related_strinfos): Likewise.
30218         (adjust_last_stmt): Likewise.
30220 2017-05-31  Richard Biener  <rguenther@suse.de>
30222         PR target/80880
30223         * config/i386/i386.c (ix86_expand_builtin): Remove assert
30224         for arg being an SSA name when expanding IX86_BUILTIN_BNDRET.
30226 2017-05-31  Richard Sandiford  <richard.sandiford@linaro.org>
30228         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Remove
30229         loop_vinfo argument and use of dependence distance vectors.
30230         Check instead whether the two references differ only in their
30231         initial value and assume that they have the same alignment if the
30232         difference is a multiple of the vector alignment.
30233         (vect_analyze_data_refs_alignment): Update call accordingly.
30235 2017-05-31  Martin Liska  <mliska@suse.cz>
30237         PR target/79155
30238         * config/i386/cpuid.h: Fix typo in a comment in cpuid.h.
30240 2017-05-31  Bin Cheng  <bin.cheng@arm.com>
30242         * tree-vect-loop-manip.c (create_intersect_range_checks_index)
30243         (create_intersect_range_checks): Move from ...
30244         * tree-data-ref.c (create_intersect_range_checks_index)
30245         (create_intersect_range_checks): ... to here.
30246         (create_runtime_alias_checks): New function factored from ...
30247         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks): ...
30248         here.  Call above function.
30249         * tree-data-ref.h (create_runtime_alias_checks): New function.
30251 2017-05-31  Bin Cheng  <bin.cheng@arm.com>
30253         * tree-data-ref.c (prune_runtime_alias_test_list): Relax minimal
30254         segment length for dr_b and compute it in wide_int.
30256 2017-05-31  Richard Biener  <rguenther@suse.de>
30258         PR tree-optimization/80906
30259         * graphite-isl-ast-to-gimple.c (copy_loop_close_phi_nodes): Get
30260         and pass through iv_map.
30261         (copy_bb_and_scalar_dependences): Adjust.
30262         (translate_pending_phi_nodes): Likewise.
30263         (copy_loop_close_phi_args): Handle code-generating IVs instead
30264         of ICEing.
30266 2017-05-30  David Malcolm  <dmalcolm@redhat.com>
30268         * diagnostic-color.c (color_dict): Add "type-diff".
30269         (parse_gcc_colors): Update comment.
30270         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
30271         -fdiagnostics-show-template-tree and -fno-elide-type.
30272         (GCC_COLORS): Add type-diff to example.
30273         (type-diff=): New.
30274         (-fdiagnostics-show-template-tree): New.
30275         (-fno-elide-type): New.
30276         * pretty-print.c (pp_format): Pass quote and formatters[argno] to
30277         the pp_format_decoder callback.  Call any m_format_postprocessor's
30278         "handle" method.
30279         (pretty_printer::pretty_printer): Initialize
30280         m_format_postprocessor.
30281         (pretty_printer::~pretty_printer): Delete any
30282         m_format_postprocessor.
30283         * pretty-print.h (printer_fn): Add bool and const char ** parameters.
30284         (class format_postprocessor): New class.
30285         (struct pretty_printer::format_decoder): Document the new parameters.
30286         (struct pretty_printer::m_format_postprocessor): New field.
30287         * tree-diagnostic.c (default_tree_printer): Update for new
30288         bool and const char ** params.
30289         * tree-diagnostic.h (default_tree_printer): Likewise.
30291 2017-05-30  Segher Boessenkool  <segher@kernel.crashing.org>
30293         * config/rs6000/predicates.md (cc_reg_not_micro_cr0_operand): Delete.
30294         (lwa_operand): Delete rs6000_gen_cell_microcode test.
30295         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
30296         rs6000_gen_cell_microcode code.
30297         (rs6000_final_prescan_insn): Delete.
30298         (rs6000_opt_vars): Delete the "gen-cell-microcode" and
30299         "warn-cell-microcode" entries.
30300         * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Delete.
30301         * config/rs6000/rs6000.md: Delete rs6000_gen_cell_microcode tests
30302         throughout.  Change cc_reg_not_micro_cr0_operand to
30303         cc_reg_not_cr0_operand throughout.
30304         (*extendhi<mode>2_noload): Delete.
30305         * config/rs6000/rs6000.opt (mgen-cell-microcode): Replace by stub.
30306         (mwarn-cell-microcode): Delete.
30307         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete
30308         -mgen-cell-microcode and -mwarn-cell-microcode.
30310 2017-05-30  Uros Bizjak  <ubizjak@gmail.com>
30312         PR target/80833
30313         * config/i386/constraints.md (Yd): New constraint.
30314         (Ye): Ditto.
30315         * config/i386/i386.md (*movti_internal): Add (?r, Ye)
30316         and (?Yd, r) alternatives.  Update insn attributes.
30317         * config/i386/i386.md (*movti_internal): Add (?r, *Ye)
30318         and (?*Yd, r) alternatives.  Update insn attributes.
30319         (double-mode inter-unit splitters): Add new GR<->XMM splitters.
30321 2017-05-30  Pierre-Marie de Rodat  <derodat@adacore.com>
30323         * gimplify.c (gimplify_modify_expr): Don't create a
30324         DECL_DEBUG_EXPR link if *FROM_P does not belong to the current
30325         function.
30327 2017-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
30329         * config/arm/arm-builtins.c (arm_expand_builtin): Remove const.
30331 2017-05-30  Richard Biener  <rguenther@suse.de>
30333         * tree-vectorizer.h (struct _stmt_vec_info): Add reduc_type
30334         and reduc_def fields.
30335         (STMT_VINFO_REDUC_TYPE): New define.
30336         (STMT_VINFO_REDUC_DEF): Likewise.
30337         (vect_force_simple_reduction): Adjust prototype.
30338         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Adjust.
30339         (vect_is_simple_reduction): Remove check_reduction argument.
30340         (vect_force_simple_reduction): Adjust and set
30341         STMT_VINFO_REDUC_TYPE and STMT_VINFO_REDUC_DEF.
30342         (vectorizable_reduction): Do not re-do reduction analysis
30343         but use STMT_VINFO_REDUC_TYPE and STMT_VINFO_REDUC_DEF.
30344         * tree-parloops.c (gather_scalar_reductions): Adjust.
30346 2017-05-30  Richard Biener  <rguenther@suse.de>
30348         PR middle-end/80901
30349         * cfgexpand.c (expand_gimple_cond): Match up loop fixup with
30350         split_edge code.
30352 2017-05-24  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30354         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs):
30355         Introduce unknown_misalignment parameter and remove vf.
30356         (vect_peeling_hash_get_lowest_cost):
30357         Pass unknown_misalignment parameter.
30358         (vect_enhance_data_refs_alignment):
30359         Fix unsupportable data ref treatment.
30361 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30363         * tree-vect-data-refs.c (vect_get_data_access_cost):
30364         Workaround for SLP handling.
30365         (vect_enhance_data_refs_alignment):
30366         Compute costs for doing no peeling at all, compare to the best
30367         peeling costs so far and avoid peeling if cheaper.
30369 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30371         * tree-vect-data-refs.c (vect_peeling_hash_choose_best_peeling):
30372         Return peeling info and set costs to zero for unlimited cost
30373         model.
30374         (vect_enhance_data_refs_alignment): Also inspect all datarefs
30375         with unknown misalignment. Compute and costs for unknown
30376         misalignment, compare them to the costs for known misalignment
30377         and choose the cheapest for peeling.
30379 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30381         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Rename.
30382         (vect_get_peeling_costs_all_drs): Create function.
30383         (vect_peeling_hash_get_lowest_cost):
30384         Use vect_get_peeling_costs_all_drs.
30385         (vect_peeling_supportable): Create function.
30386         (vect_enhance_data_refs_alignment): Use vect_peeling_supportable.
30388 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30390         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Create
30391         DR_HAS_NEGATIVE_STEP.
30392         (vect_update_misalignment_for_peel): Define DR_MISALIGNMENT.
30393         (vect_enhance_data_refs_alignment): Use.
30394         (vect_duplicate_ssa_name_ptr_info): Use.
30395         * tree-vectorizer.h (dr_misalignment): Use.
30396         (known_alignment_for_access_p): Use.
30398 2017-05-30  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
30400         PR target/78838
30401         * config/msp430/msp430.c (gen_prefix): Return NULL when section name is
30402         .lowtext.
30403         (has_section_name): New function.
30405 2017-05-30  Martin Liska  <mliska@suse.cz>
30407         PR other/80909
30408         * auto-profile.c (get_function_decl_from_block): Fix
30409         parenthesis.
30411 2017-05-30  Richard Biener  <rguenther@suse.de>
30413         PR middle-end/80876
30414         * cfgexpand.c (expand_gimple_cond): Fixup preserving loops again.
30416 2017-05-30  Martin Liska  <mliska@suse.cz>
30418         * dumpfile.c: Use newly added macro DUMP_FILE_INFO.
30419         * dumpfile.h (struct dump_file_info): Remove ctors.
30421 2017-05-30  Martin Liska  <mliska@suse.cz>
30423         * predict.def: Fix GNU coding style.
30425 2017-05-29  Max Filippov  <jcmvbkbc@gmail.com>
30427         * config/xtensa/xtensa.c (xtensa_initial_elimination_offset):
30428         Mark 'to' argument with ATTRIBUTE_UNUSED.
30430 2017-05-29  Max Filippov  <jcmvbkbc@gmail.com>
30432         * config/xtensa/xtensa.c (xtensa_emit_call): Use
30433         HOST_WIDE_INT_PRINT_HEX instead of 0x%lx format string.
30434         (print_operand): Use HOST_WIDE_INT_PRINT_DEC instead of %ld
30435         format string.
30437 2017-05-29  Eric Botcazou  <ebotcazou@adacore.com>
30439         * doc/install.texi (Options specification): Restore entry of
30440         --enable-sjlj-exceptions.
30442 2017-05-27  Michael Eager  <eager@eagercon.com>
30444         Revert:
30445         2016-01-21  Ajit Agarwal  <ajitkum@xilinx.com>
30447         See https://gcc.gnu.org/ml/gcc/2017-05/msg00221.html.
30449         * config/microblaze/microblaze.h
30450         (FIXED_REGISTERS): Update in macro.
30451         (CALL_USED_REGISTERS): Update in macro.
30453 2017-05-27  François-Xavier Coudett  <fxcoudert@gcc.gnu.org>
30455         * doc/install.texi: Add links to macOS binary distributions.
30457 2017-05-27  Jakub Jelinek  <jakub@redhat.com>
30459         PR bootstrap/80887
30460         Revert:
30461         2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30463         * match.pd ((A +- CST1) +- CST2): Allow some conversions.
30465 2017-05-26  Martin Liska  <mliska@suse.cz>
30467         * dumpfile.h (enum dump_kind): Renumber TDF_* flags to be contiguous.
30469 2017-05-26  Martin Liska  <mliska@suse.cz>
30471         * cfg.c (check_bb_profile): Do not use TDF_COMMENT and print
30472         always leading ';; '.
30473         (dump_bb_info): Likewise.
30474         (brief_dump_cfg): Likewise.
30475         * cfgrtl.c (print_rtl_with_bb): Do not use TDF_COMMENT.
30476         * dumpfile.c: Remove usage of TDF_VERBOSE.
30477         * dumpfile.h (enum dump_kind): Likewise.
30478         (dump_gimple_bb_header): Do not use TDF_COMMENT.
30479         * print-tree.c (debug_verbose): Remove.
30480         * tree-cfg.c (gimple_dump_cfg): Do not use TDF_COMMENT.
30481         (dump_function_to_file): Remove dumps guarded with TDF_VERBOSE.
30482         * tree-diagnostic.c (default_tree_printer): Replace
30483         TDF_DIAGNOSTIC with TDF_SLIM.
30485 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30487         * tree-vect-loop-manip.c (create_intersect_range_checks_index): Pass
30488         in parameter loop, rather than loop_vinfo.
30489         (create_intersect_range_checks): Ditto.
30490         (vect_create_cond_for_alias_checks): Update call to above functions.
30492 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30494         PR tree-optimization/80815
30495         * tree-data-ref.c (prune_runtime_alias_test_list): Simplify condition
30496         for merging runtime alias checks.  Handle negative DR_STEPs.
30498 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30500         * tree-vect-data-refs.c (Operator==, comp_dr_with_seg_len_pair):
30501         Move from ...
30502         * tree-data-ref.c (Operator==, comp_dr_with_seg_len_pair): To here.
30503         * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list): Factor
30504         out code pruning runtime alias checks.
30505         * tree-data-ref.c (prune_runtime_alias_test_list): New function
30506         factored out from above.
30507         * tree-vectorizer.h (struct dr_with_seg_len, dr_with_seg_len_pair_t):
30508         Move from ...
30509         * tree-data-ref.h (struct dr_with_seg_len, dr_with_seg_len_pair_t):
30510         ... to here.
30511         (prune_runtime_alias_test_list): New decalaration.
30513 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30515         * tree-vect-data-refs.c (compare_tree): Rename and move ...
30516         * tree-data-ref.c (data_ref_compare_tree): ... to here.
30517         * tree-data-ref.h (data_ref_compare_tree): New decalaration.
30518         * tree-vect-data-refs.c (dr_group_sort_cmp): Update uses.
30519         (operator==, comp_dr_with_seg_len_pair): Ditto.
30520         (vect_prune_runtime_alias_test_list): Ditto.
30522 2017-05-26  Martin Liska  <mliska@suse.cz>
30524         PR ipa/80663
30525         * params.def: Bound partial-inlining-entry-probability param.
30527 2017-05-26  Marek Polacek  <polacek@redhat.com>
30529         PR sanitizer/80875
30530         * fold-const.c (fold_binary_loc) <case MULT_EXPR>: Check if OP1
30531         can be negated.
30533 2017-05-26  Richard Biener  <rguenther@suse.de>
30535         PR tree-optimization/80842
30536         * tree-ssa-ccp.c (set_lattice_value): Always meet with the old
30537         value.
30539 2017-05-26  Richard Biener  <rguenther@suse.de>
30541         PR tree-optimization/80844
30542         * tree-vectorizer.c (adjust_simduid_builtins): Propagate results.
30544 2017-05-25  Sebastian Peryt  <sebastian.peryt@intel.com>
30546         * doc/md.texi (Machine Constraints): Update x86 family
30547         machine constraints section to match 'config/i386/constraints.md'.
30549 2017-05-25  Volker Reichelt  <v.reichelt@netcologne.de>
30551         * doc/invoke.texi (-Wcatch-value=): Document new warning option.
30553 2017-05-25  Nathan Sidwell  <nathan@acm.org>
30555         * doc/invoke.texi (--enable-languages): Update documentation.
30557 2017-05-25  Martin Liska  <mliska@suse.cz>
30559         * dumpfile.c: Add TDF_FOLDING.
30560         * dumpfile.h (enum dump_kind): Likewise.
30561         * genmatch.c (dt_simplify::gen_1): Use it.
30563 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30565         * match.pd (view_convert (convert@0 @1)): Handle zero-extension.
30567 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30569         * match.pd ((A +- CST1) +- CST2): Allow some conversions.
30570         * tree.c (drop_tree_overflow): Handle COMPLEX_CST and VECTOR_CST.
30572 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30574         * fold-const.c (fold_binary_loc) [(A & C) == D]: Remove transformation.
30575         * match.pd (X == C): Rewrite it here.
30576         (with_possible_nonzero_bits, with_possible_nonzero_bits2,
30577         with_certain_nonzero_bits2): New predicates.
30578         * tree-ssanames.c (get_nonzero_bits): Handle INTEGER_CST.
30580 2017-05-24  Nathan Sidwell  <nathan@acm.org>
30582         * lto-streamer-in.c (lto_input_data_block): Adjust T const cast to
30583         avoid warning.
30585         * auto-profile.c (afdo_propagate): Adjust T const cast to avoid
30586         warning.
30588 2017-05-24  Segher Boessenkool  <segher@kernel.crashing.org>
30590         * config/powerpcspe: New port.  Files are copied from the rs6000
30591         port, with "rs6000" in filenames replaced by "powerpcspe".
30593 2017-05-24  Wilco Dijkstra  <wdijkstr@arm.com>
30595         PR rtl-optimization/80754
30596         * lra-remat.c (do_remat): Add overlap checks for dst_regno.
30598 2017-05-24  Sheldon Lobo  <smlobo@sheldon.us.oracle.com>
30600         * config/sparc/sparc.md (length): Return the correct value for -mflat
30601         sibcalls to match output_sibcall.
30603 2017-05-24  Segher Boessenkool  <segher@kernel.crashing.org>
30605         PR bootstrap/80860
30606         PR bootstrap/80843
30607         * config/rs6000/rs6000.c (struct machine_function): Add new field
30608         n_components.
30609         (rs6000_get_separate_components): Init that field, use it.
30610         (rs6000_components_for_bb): Use the field.
30612 2017-05-24  Thomas Preud'homme  <thomas.preudhomme@arm.com>
30614         * config/arm/arm.c (arm_expand_prologue): Fix typo in comment.
30616 2017-05-24  Peter Bergner  <bergner@vnet.ibm.com>
30618         PR middle-end/80823
30619         * tree-cfg.c (group_case_labels_stmt): Delete increment of "i";
30621 2017-05-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
30623         PR target/80725
30624         * config/s390/s390.c (s390_check_qrst_address): Check incoming
30625         address against address_operand predicate.
30626         * config/s390/s390.md ("*indirect_jump"): Swap alternatives.
30628 2017-05-24  Eric Botcazou  <ebotcazou@adacore.com>
30630         * var-tracking.c (track_expr_p): Do not return 0 for tracked record
30631         parameters passed indirectly.
30633 2017-05-23  Uros Bizjak  <ubizjak@gmail.com>
30635         * config/i386/i386.md (*movdi_internal): Remove SSE4
30636         alternative 18 (?r, *v).  Update insn attributes.
30637         (*movsi_internal): Remove SSE4 alternative 13 (?r, *v).
30638         Update insn attributes.
30639         (*zero_extendsidi2): Remove SSE4 alternative (?r, *x).
30640         Update insn attributes.
30641         * config/i386/sse.md (vec_extract<ssevecmodelower>_0): Remove SSE4
30642         alternative 1 (r, v). Remove isa attribute.
30643         * config/i386/i386.c (dimode_scalar_chain::make_vector_copies):
30644         Always move value through stack for !TARGET_INTER_UNIT_MOVES_TO_VEC
30645         and !TARGET_INTER_UNIT_MOVES_TO_VEC targets.
30647 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30649         * doc/sourcebuild.texi (Directives, Verify compiler message): Document
30650         dg-line directive.
30652 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30654         * cgraphunit.c (symbol_table::process_new_functions): Update.
30655         * ipa-fnsummary.c (pass_data_inline_parameters): Remove.
30656         (inline_generate_summary): Rename to ...
30657         (ipa_fn_summary_generate): ... this one.
30658         (inline_read_summary): Rename to ...
30659         (ipa_fn_summary_read): ... this one.
30660         (inline_write_summary): Rename to ...
30661         (ipa_fn_summary_write): ... this one.
30662         (inline_free_summary): Rename to ...
30663         (ipa_free_fn_summary): ... this one.
30664         (pass_data_local_fn_summary, pass_local_fn_summary,
30665         make_pass_local_fn_summary, pass_data_ipa_free_fn_summary,
30666         pass_ipa_free_fn_summary, make_pass_ipa_free_fn_summary,
30667         pass_data_ipa_fn_summary, pass_ipa_fn_summary,
30668         make_pass_ipa_fn_summary): New.
30669         * ipa-fnsummary.h (inline_generate_summary, inline_read_summary,
30670         inline_write_summary, inline_free_summary): Remove.
30671         (ipa_free_fn_summary) : New.
30672         * ipa-inline.c (ipa_inline): Update.
30673         (pass_ipa_inline): Do not generate summaries.
30674         * ipa.c (pass_data_ipa_free_fn_summary, pass_ipa_free_fn_summary):
30675         Remove.
30676         * passes.def: Replace pass_inline_parameters by pass_local_fn_summary
30677         and add pass_ipa_fn_summary.
30678         * tree-pass.h (make_pass_ipa_fn_summary, make_pass_local_fn_summary):
30679         New.
30680         (make_pass_inline_parameters): Remove.
30682 2017-05-23  Thomas Schwinge  <thomas@codesourcery.com>
30684         * omp-low.c (struct omp_context): Remove "default_kind" member.
30685         Adjust all users.
30687         * omp-offload.c (execute_oacc_device_lower): Remove the
30688         parallelism dimensions function attributes for unparallelized
30689         OpenACC kernels constructs.
30691 2017-05-23  Martin Liska  <mliska@suse.cz>
30693         * cgraph.c (cgraph_node::get_create): Use symtab_node::dump_{asm_,}name
30694         functions.
30695         (cgraph_edge::make_speculative): Likewise.
30696         (cgraph_edge::resolve_speculation): Likewise.
30697         (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
30698         (cgraph_node::dump): Likewise.
30699         * cgraph.h: Likewise.
30700         * cgraphunit.c (analyze_functions): Likewise.
30701         (symbol_table::compile): Likewise.
30702         * ipa-cp.c (print_all_lattices): Likewise.
30703         (determine_versionability): Likewise.
30704         (initialize_node_lattices): Likewise.
30705         (ipcp_verify_propagated_values): Likewise.
30706         (estimate_local_effects): Likewise.
30707         (update_profiling_info): Likewise.
30708         (create_specialized_node): Likewise.
30709         (perhaps_add_new_callers): Likewise.
30710         (decide_about_value): Likewise.
30711         (decide_whether_version_node): Likewise.
30712         (identify_dead_nodes): Likewise.
30713         (ipcp_store_bits_results): Likewise.
30714         * ipa-devirt.c (dump_targets): Likewise.
30715         (ipa_devirt): Likewise.
30716         * ipa-icf.c (sem_item::dump): Likewise.
30717         (sem_function::equals): Likewise.
30718         (sem_variable::equals): Likewise.
30719         (sem_item_optimizer::read_section): Likewise.
30720         (sem_item_optimizer::execute): Likewise.
30721         (congruence_class::dump): Likewise.
30722         * ipa-inline-analysis.c (dump_ipa_call_summary): Likewise.
30723         (dump_inline_summary): Likewise.
30724         (estimate_node_size_and_time): Likewise.
30725         (inline_analyze_function): Likewise.
30726         * ipa-inline-transform.c (inline_call): Likewise.
30727         * ipa-inline.c (report_inline_failed_reason): Likewise.
30728         (want_early_inline_function_p): Likewise.
30729         (edge_badness): Likewise.
30730         (update_edge_key): Likewise.
30731         (inline_small_functions): Likewise.
30732         * ipa-profile.c (ipa_profile): Likewise.
30733         * ipa-prop.c (ipa_print_node_jump_functions): Likewise.
30734         (ipa_make_edge_direct_to_target): Likewise.
30735         (remove_described_reference): Likewise.
30736         (ipa_impossible_devirt_target): Likewise.
30737         (propagate_controlled_uses): Likewise.
30738         (ipa_print_node_params): Likewise.
30739         (ipcp_transform_function): Likewise.
30740         * ipa-pure-const.c (pure_const_read_summary): Likewise.
30741         (propagate_pure_const): Likewise.
30742         * ipa-reference.c (generate_summary): Likewise.
30743         (read_write_all_from_decl): Likewise.
30744         (propagate): Likewise.
30745         (ipa_reference_read_optimization_summary): Likewise.
30746         * ipa-utils.c (ipa_merge_profiles): Likewise.
30747         * ipa.c (walk_polymorphic_call_targets): Likewise.
30748         (symbol_table::remove_unreachable_nodes): Likewise.
30749         (ipa_single_use): Likewise.
30750         * passes.c (execute_todo): Likewise.
30751         * predict.c (drop_profile): Likewise.
30752         * symtab.c (symtab_node::get_dump_name): New function.
30753         (symtab_node::dump_name): Likewise.
30754         (symtab_node::dump_asm_name): Likewise.
30755         (symtab_node::dump_references): Likewise.
30756         (symtab_node::dump_referring): Likewise.
30757         (symtab_node::dump_base): Likewise.
30758         (symtab_node::debug_symtab): Likewise.
30759         * tree-sra.c (convert_callers_for_node): Likewise.
30760         * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
30761         * value-prof.c (init_node_map): Likewise.
30763 2017-05-23  Martin Liska  <mliska@suse.cz>
30765         * cgraph.h: Move symtab_node::dump_table to symbol_table::dump
30766         and symtab_node::debug_symtab to symbol_table::debug.
30767         * cgraphunit.c (analyze_functions): Use the renamed function.
30768         (symbol_table::compile): Likewise.
30769         * ipa-cp.c (ipcp_verify_propagated_values): Likewise.
30770         * ipa-icf.c (sem_item_optimizer::execute): Likewise.
30771         * passes.c (execute_todo): Likewise.
30772         * symtab.c (symbol_table::dump): New function.
30773         * tree-ssa-structalias.c (ipa_pta_execute): Use the renamed function.
30775 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30777         * ipa-fnsummary.c (estimate_node_size_and_time): Do not sanity check
30778         that nonconst implies exec.
30780 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30782         * ipa-inline-analysis.c (cgraph_2edge_hook_list, cgraph_edge_hook_list,
30783         inline_edge_removal_hook, inline_edge_duplication_hook): Remove.
30784         (inline_edge_summary_vec): Turn into ...
30785         (ipa_call_summaries): ... this one.
30786         (redirect_to_unreachable, edge_set_predicate,
30787         evaluate_properties_for_edge, inline_summary_alloc,
30788         reset_ipa_call_summary, reset_inline_summary,
30789         inline_summary_t::duplicate): Update.
30790         (inline_edge_duplication_hook): Turn to ...
30791         (ipa_call_summary_t::duplicate): ... this one.
30792         (inline_edge_removal_hook): Turn to ...
30793         (ipa_call_summary_t::remove): ... this one.
30794         (dump_inline_edge_summary): Turn to ...
30795         (dump_ipa_call_summary): ... this one.
30796         (estimate_function_body_sizes): Update.
30797         (inline_update_callee_summaries): Update.
30798         (remap_edge_change_prob): Update.
30799         (remap_edge_summaries): Update.
30800         (inline_merge_summary): Update.
30801         (do_estimate_edge_time): Update.
30802         (inline_generate_summary): Update.
30803         (inline_read_section): Update.
30804         (inline_read_summary): Update.
30805         (inline_free_summary): Update.
30806         * ipa-inline.c (can_inline_edge_p): Update.
30807         (compute_inlined_call_time): Update.
30808         (want_inline_small_function_p): Update.
30809         (edge_badness): Update.
30810         (early_inliner): Update.
30811         * ipa-inline.h (inline_edge_summary): Turn to ...
30812         (ipa_call_summary): ... this one.
30813         (ipa_call_summary_t): New class.
30814         (inline_edge_summary_t, inline_edge_summary_vec): Remove.
30815         (ipa_call_summaries): New.
30816         (inline_edge_summary): Remove.
30817         (estimate_edge_growth): Update.
30818         * ipa-profile.c (ipa_propagate_frequency_1): Update.
30819         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
30820         * ipa-split.c (execute_split_functions): Update.
30821         * ipa.c (symbol_table::remove_unreachable_nodes): Update.
30823 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30825         * doc/sourcebuild.texi (Effective-Target Keywords, Other hardware
30826         attributes): Document rdrand effective target.
30828 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30830         * doc/sourcebuild.texi (Effective-Target Keywords, Other hardware
30831         attributes): Sort alphabetically.
30833 2017-05-23  Georg-Johann Lay  <avr@gjlay.de>
30835         * config/avr/genmultilib.awk: Use gsub instead of gensub.
30837 2017-05-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
30839         PR target/80718
30840         * config/rs6000/vsx.md (vsx_splat_<mode>, VSX_D iterator): Split
30841         V2DF/V2DI splat into two separate patterns, one that handles
30842         registers, and the other that only handles memory.  Drop support
30843         for splatting from a GPR on ISA 2.07 and then splitting the
30844         splat into direct move and splat.
30845         (vsx_splat_<mode>_reg): Likewise.
30846         (vsx_splat_<mode>_mem): Likewise.
30848 2017-05-22  Segher Boessenkool  <segher@kernel.crashing.org>
30850         * cfgcleanup.c (bb_is_just_return): Allow CLOBBERs.
30852 2017-05-22  Jakub Jelinek  <jakub@redhat.com>
30854         PR middle-end/80809
30855         * omp-low.c (finish_taskreg_remap): New function.
30856         (finish_taskreg_scan): If unit size of ctx->record_type
30857         is non-constant, unshare the size expression and replace
30858         decls in it with possible outer var refs.
30860         PR middle-end/80809
30861         * gimplify.c (omp_add_variable): For GOVD_DEBUG_PRIVATE use
30862         GOVD_SHARED rather than GOVD_PRIVATE with it.
30863         (gimplify_adjust_omp_clauses_1, gimplify_adjust_omp_clauses): Expect
30864         GOVD_SHARED rather than GOVD_PRIVATE with GOVD_DEBUG_PRIVATE.
30866         PR middle-end/80853
30867         * omp-low.c (lower_reduction_clauses): Pass OMP_CLAUSE_PRIVATE
30868         as last argument to build_outer_var_ref for pointer bases of array
30869         section reductions.
30871 2017-05-19  Martin Sebor  <msebor@redhat.com>
30873         * print-tree.c (print_node): Print DECL_READ_P flag.
30875 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30877         * Makefile.in: Add ipa-fnsummary.o and ipa-fnsummary.h
30878         * auto-profile.c: Replace ipa-inline.h by ipa-fnsummary.h
30879         * cgraph.c: Likewise.
30880         * cgraphunit.c: Likewise.
30881         * gengtype.c: Likewise.
30882         * ipa-cp.c: Likewise.
30883         * ipa-devirt.c: Likewise.
30884         * ipa-icf.c: Likewise.
30885         * ipa-predicate.c: Likewise.
30886         * ipa-profile.c: Likewise.
30887         * ipa-prop.c: Likewise.
30888         * ipa-split.c: Likewise.
30889         * ipa.c: Likewise.
30890         * ipa-inline-analysis.c (inline_summaries, ipa_call_summaries,
30891         edge_predicate_pool, dump_inline_hints,
30892         inline_summary::account_size_time, redirect_to_unreachable,
30893         edge_set_predicate, set_hint_predicate,
30894         evaluate_conditions_for_known_args, evaluate_properties_for_edge,
30895         inline_summary_alloc, ipa_call_summary::reset, inline_summary::reset,
30896         inline_summary_t::remove, remap_hint_predicate_after_duplication,
30897         inline_summary_t::duplicate, ipa_call_summary_t::duplicate,
30898         ipa_call_summary_t::remove, initialize_growth_caches,
30899         free_growth_caches, dump_ipa_call_summary, dump_inline_summary,
30900         debug_inline_summary, dump_inline_summaries, initialize_inline_failed,
30901         mark_modified, unmodified_parm_1, unmodified_parm,
30902         unmodified_parm_or_parm_agg_item, eliminated_by_inlining_prob,
30903         set_cond_stmt_execution_predicate, set_switch_stmt_execution_predicate,
30904         compute_bb_predicates, will_be_nonconstant_expr_predicate,
30905         will_be_nonconstant_predicate, record_modified_bb_info,
30906         get_minimal_bb, record_modified, param_change_prob,
30907         phi_result_unknown_predicate, predicate_for_phi_result,
30908         array_index_predicate, clobber_only_eh_bb_p, fp_expression_p,
30909         estimate_function_body_sizes, compute_inline_parameters,
30910         compute_inline_parameters_for_curren, pass_data_inline_parameters,
30911         estimate_node_size_and_time, estimate_ipcp_clone_size_and_time,
30912         inline_update_callee_summaries, remap_edge_change_prob,
30913         remap_edge_summaries, remap_hint_predicate, inline_merge_summary,
30914         inline_update_overall_summary, inline_indirect_intraprocedural_analysis,
30915         inline_analyze_function, inline_summary_t::insert,
30916         inline_generate_summary, read_ipa_call_summary, inline_read_section,
30917         inline_read_summary, write_ipa_call_summary, inline_write_summary,
30918         inline_free_summary): Move to ipa-fnsummary.h
30919         (predicate_t): Remove.
30920         * ipa-fnsummary.c: New file.
30921         * ipa-inline.h:  Do not include sreal.h and ipa-predicate.h
30922         (enum inline_hints_vals, inline_hints, agg_position_info,
30923         INLINE_SIZE_SCALE, size_time_entry, inline_summary, inline_summary_t,
30924         inline_summaries, ipa_call_summary, ipa_call_summary_t,
30925         ipa_call_summaries, debug_inline_summary, dump_inline_summaries,
30926         dump_inline_summary, dump_inline_hints, inline_generate_summary,
30927         inline_read_summary, inline_write_summary, inline_free_summary,
30928         inline_analyze_function, initialize_inline_failed,
30929         inline_merge_summary, inline_update_overall_summary,
30930         compute_inline_parameters): Move to ipa-fnsummary.h
30931         * ipa-fnsummary.h: New file.
30932         * ipa-inline-transform.h: Include ipa-inline.h.
30933         * ipa-inline.c: LIkewise.
30935 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30937         * ipa-inline.c (edge_badness): Use inlined_time instead of
30938         inline_summaries->get.
30940 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30942         * ipa-inline.c (edge_badness): Use estimate_size_after_inlining.
30944 2017-05-22  Nathan Sidwell  <nathan@acm.org>
30946         * doc/invoke.texi (fdump-translation-unit): Delete documentation.
30947         (fdump-lang): Document 'raw' option.
30948         * dumpfile.h (TDI_tu): Delete.
30949         * dumpfile.c (dump_files): Remove translation-unit.
30950         (FIRST_AUTO_NUMBERED_DUMP): Decrement.
30952 2017-05-22  Georg-Johann Lay  <avr@gjlay.de>
30954         * config/avr/t-avr (AWK) [t-multilib]: Remove "-v FORMAT=Makefile"
30955         command option from $(AWK) call.
30956         * config/avr/genmultilib.awk: Simplify and rewrite so that it
30957         generates MULTILIB_REQUIRED instead of MULTILIB_EXCEPTIONS.
30958         [FORMAT]: Remove handling of variable.
30959         * config/avr/t-multilib: Regenerate.
30961 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30963         * ipa-inline-analysis.c (inline_summary::reset): Do not reset
30964         self_time.
30965         (dump_inline_summary): Do not print self_time.
30966         (estimate_function_body_sizes): Do not set self_time.
30967         (compute_inline_parameters): Likewise.
30968         (inline_read_section, inline_write_summary): Do not stream self_time.
30969         * ipa-inline.h (inline_summary): Drop self_time.
30971 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30973         * ipa-inline-analysis.c (account_size_time): Rename to ...
30974         (inline_summary::account_size_time): ... this one.
30975         (reset_ipa_call_summary): Turn to ...
30976         (ipa_call_summary::reset): ... this one.
30977         (reset_inline_summary): Turn to ...
30978         (inline_summary::reset): ... this one.
30979         (inline_summary_t::remove): Update.
30980         (inline_summary_t::duplicate): Update.
30981         (ipa_call_summary_t::remove): Update.
30982         (dump_inline_summary): Update.
30983         (estimate_function_body_sizes): Update.
30984         (compute_inline_parameters): Update.
30985         (estimate_node_size_and_time): Update.
30986         (inline_merge_summary): Update.
30987         (inline_update_overall_summary): Update.
30988         (inline_read_section): Update.
30989         (inline_write_summary): Update.
30990         * ipa-inline.h (inline_summary): Rename entry to size_time_table;
30991         add account_size_time and reset member functions.
30992         (ipa_call_summary): Add reset function.
30993         * ipa-predicate.h (predicate::operator &): Constify.
30995 2017-05-22  Richard Biener  <rguenther@suse.de>
30997         * df-scan.c (df_insn_refs_verify): Speedup when not verifying.
30999 2017-05-19  Jason Merrill  <jason@redhat.com>
31001         * tree.c (make_tree_vec_stat, grow_tree_vec_stat): Use size_t.
31003 2017-05-19  Marek Polacek  <polacek@redhat.com>
31005         PR sanitizer/80800
31006         * fold-const.c (extract_muldiv_1) <case TRUNC_DIV_EXPR>: Add
31007         TYPE_OVERFLOW_WRAPS checks.
31009 2017-05-19  Thomas Schwinge  <thomas@codesourcery.com>
31011         * tree-core.h (enum omp_clause_default_kind): Add
31012         "OMP_CLAUSE_DEFAULT_PRESENT".
31013         * tree-pretty-print.c (dump_omp_clause): Handle it.
31014         * gimplify.c (enum gimplify_omp_var_data): Add
31015         "GOVD_MAP_FORCE_PRESENT".
31016         (gimplify_adjust_omp_clauses_1): Map it to
31017         "GOMP_MAP_FORCE_PRESENT".
31018         (oacc_default_clause): Handle "OMP_CLAUSE_DEFAULT_PRESENT".
31020         * gimplify.c (oacc_default_clause): Clarify.
31022 2017-05-19  Nathan Sidwell  <nathan@acm.org>
31024         LANG_HOOK_REGISTER_DUMPS
31025         * toplev.c (general_init): Call register dump lang hook.
31026         * doc/invoke.texi: Document -fdump-lang option family.
31027         * dumpfile.c (dump_files): Remove class dump here.
31028         (FIRST_AUTO_NUMBERED_DUMP): Adjust.
31029         * dumpfile.h (tree_dump_index): Remove TDI_class.
31030         * langhooks-def.h (lhd_register_dumps): Declare.
31031         (LANG_HOOKS_REGISTER_DUMPS): Define.
31032         (LANG_HOOKS_INITIALIZER): Add it.
31033         * langhooks.c (lhd_register_dumps): Define.
31034         * langhooks.h (struct lang_hooks): Add register_dumps.
31036 2017-05-19  Nathan Sidwell  <nathan@acm.org>
31038         * context.h (context::set_passes): New.
31039         * context.c (context::context): Do not create pass manager.
31040         * toplev.c (general_init): Create pass manager here.
31042 2017-05-19  Segher Boessenkool  <segher@kernel.crashing.org>
31044         * config/rs6000/rs6000.md (splitter to load of -1 and mask): Don't
31045         use this splitter if two add or or instructions would also work for
31046         the constant we want to generate.
31048 2017-05-19  Richard Biener  <rguenther@suse.de>
31050         PR build/80821
31051         * genmatch.c (dt_node::gen_kids_1): Add missing scope around
31052         predicate evaluation.
31054 2017-05-19  Jan Hubicka  <hubicka@ucw.cz>
31056         * ipa-inline.h (ipa_call_summary): Turn sizes into signed;
31057         add ctor.
31058         * ipa-inline.c (want_inline_small_function_p): Do not cast to
31059         unsigned.
31061 2017-05-19  Jan Hubicka  <hubicka@ucw.cz>
31063         * ipa-inline-analysis.c (cgraph_2edge_hook_list, cgraph_edge_hook_list,
31064         inline_edge_removal_hook, inline_edge_duplication_hook): Remove.
31065         (inline_edge_summary_vec): Turn into ...
31066         (ipa_call_summaries): ... this one.
31067         (redirect_to_unreachable, edge_set_predicate,
31068         evaluate_properties_for_edge, inline_summary_alloc,
31069         reset_ipa_call_summary, reset_inline_summary,
31070         inline_summary_t::duplicate): Update.
31071         (inline_edge_duplication_hook): Turn to ...
31072         (ipa_call_summary_t::duplicate): ... this one.
31073         (inline_edge_removal_hook): Turn to ...
31074         (ipa_call_summary_t::remove): ... this one.
31075         (dump_inline_edge_summary): Turn to ...
31076         (dump_ipa_call_summary): ... this one.
31077         (estimate_function_body_sizes): Update.
31078         (inline_update_callee_summaries): Update.
31079         (remap_edge_change_prob): Update.
31080         (remap_edge_summaries): Update.
31081         (inline_merge_summary): Update.
31082         (do_estimate_edge_time): Update.
31083         (inline_generate_summary): Update.
31084         (inline_read_section): Update.
31085         (inline_read_summary): Update.
31086         (inline_free_summary): Update.
31087         * ipa-inline.c (can_inline_edge_p): Update.
31088         (compute_inlined_call_time): Update.
31089         (want_inline_small_function_p): Update.
31090         (edge_badness): Update.
31091         (early_inliner): Update.
31092         * ipa-inline.h (inline_edge_summary): Turn to ...
31093         (ipa_call_summary): ... this one.
31094         (ipa_call_summary_t): New class.
31095         (inline_edge_summary_t, inline_edge_summary_vec): Remove.
31096         (ipa_call_summaries): New.
31097         (inline_edge_summary): Remove.
31098         (estimate_edge_growth): Update.
31099         * ipa-profile.c (ipa_propagate_frequency_1): Update.
31100         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
31101         * ipa-split.c (execute_split_functions): Update.
31102         * ipa.c (symbol_table::remove_unreachable_nodes): Update.
31104 2017-05-19  Richard Biener  <rguenther@suse.de>
31106         PR middle-end/80764
31107         * cfgexpand.c (expand_gimple_cond): Fix loop fixup.
31109 2017-05-18  Segher Boessenkool  <segher@kernel.crashing.org>
31111         * config/rs6000/rs6000.c (struct machine_function): Add field
31112         fpr_is_wrapped_separately.
31113         (rs6000_get_separate_components): Use 64 components.  Handle the
31114         new FPR components.
31115         (rs6000_components_for_bb): Handle the FPR components.
31116         (rs6000_emit_prologue_components): Handle the FPR components.
31117         (rs6000_emit_epilogue_components): Handle the FPR components.
31118         (rs6000_set_handled_components): Handle the FPR components.
31119         (rs6000_emit_prologue): Don't output prologue code for those FPRs
31120         that are already separately shrink-wrapped.
31121         (rs6000_emit_epilogue): Don't output epilogue code for those FPRs
31122         that are already separately shrink-wrapped.
31124 2017-05-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
31126         PR target/80510
31127         * config/rs6000/predicates.md (simple_offsettable_mem_operand):
31128         New predicate.
31130         * config/rs6000/rs6000.md (ALTIVEC_DFORM): New iterator.
31131         (define_peephole2 for Altivec d-form load): Add peepholes to catch
31132         cases where the register allocator uses a move and an offsettable
31133         memory operation to/from a FPR register on ISA 2.06/2.07.
31134         (define_peephole2 for Altivec d-form store): Likewise.
31136 2017-05-18  Uros Bizjak  <ubizjak@gmail.com>
31138         PR target/80799
31139         * config/i386/mmx.md (*mov<mode>_internal): Enable
31140         alternatives 11, 12, 13 and 14 also for 32bit targets.
31141         Remove alternatives 15, 16, 17 and 18.
31142         * config/i386/sse.md (vec_concatv2di): Change
31143         alternative (!x, *y) to (x, ?!*Yn).
31145 2017-05-18  Paolo Carlini  <paolo.carlini@oracle.com>
31147         * dumpfile.h (enum dump_kind): Remove stray comma.
31149 2017-05-18  Jan Hubicka  <hubicka@ucw.cz>
31151         * Makefile.in: Add ipa-predicate.o and ipa-predicate.h
31152         * ipa-inline-analysis.c (NUM_CONDITIONS): turn into
31153         predicate::num_conditions
31154         (IS_NOT_CONSTANT): turn into predicate::is_not_constant.
31155         (CHANGED): turn into predicate::changed.
31156         (agg_position_info): Move to ipa-predicate.h
31157         (add_condition, predicate::add_clause, predicate::operator &=,
31158         predicate::or_with, predicate::evaluate, predicate::probability,
31159         dump_condition, dump_clause, predicate::dump,
31160         predicate::remap_after_duplication, predicate::remap_after_inlining,
31161         predicate::stream_in, predicate::stream_out): Move to ipa-predicate.c
31162         (evaluate_conditions_for_known_args): Update.
31163         (set_cond_stmt_execution_predicate): Update.
31164         * ipa-inline.h: Include ipa-predicate.h
31165         (condition, inline_param_summary, conditions, agg_position_info,
31166         predicate): Move to ipa-predicate.h
31167         * ipa-predicate.c: New file.
31168         * ipa-predicate.h: New file.
31170 2017-05-18  Wilco Dijkstra  <wdijkstr@arm.com>
31172         * final.c (leaf_function_p): Check we are not in a sequence.
31174 2017-05-18  Martin Liska  <mliska@suse.cz>
31176         * cfgrtl.c (rtl_verify_edges): Remove usage of TDF_RTL.
31177         * dumpfile.c (dump_register): Use new enum dump_kind.
31178         (get_dump_file_name): Likewise.
31179         (dump_enable_all): Likewise.
31180         (dump_switch_p_1): Likewise.
31181         (enable_rtl_dump_file): Remove usage of TDF_RTL.
31182         * dumpfile.h (enum dump_kind): New enum type.
31183         (struct dump_file_info): Create constructor and
31184         format fields and comments.
31185         * passes.c (pass_manager::register_one_dump_file):
31186         Use num dump_kind.
31187         * statistics.c (statistics_early_init): Likewise.
31188         * tree-ssa-loop-prefetch.c (dump_mem_details): Replace
31189         TDF_TREE with TDF_SLIM.
31190         (gather_memory_references_ref): Likewise.
31192 2017-05-18  Martin Liska  <mliska@suse.cz>
31194         * vec.h (struct vnull): Use it.
31196 2017-05-18  Jan Hubicka  <hubicka@ucw.cz>
31198         * ipa-inline-analysis.c (predicate_conditions): Move to ipa-inline.h
31199         (true_predicate, false_predicate, true_predicate_p,
31200         false_predicate_p): Remove.
31201         (single_cond_predicate, not_inlined_predicate): Turn to member function
31202         in ipa-inline.h
31203         (add_condition): Update.
31204         (add_clause): Turn to...
31205         (predicate::add_clause): ... this one; update; allow passing NULL
31206         as parameter.
31207         (and_predicates): Turn to ...
31208         (predicate::operator &=): ... this one.
31209         (predicates_equal_p): Move to predicate::operator == in ipa-inline.h
31210         (or_predicates): Turn to ...
31211         (predicate::or_with): ... this one.
31212         (evaluate_predicate): Turn to ...
31213         (predicate::evaluate): ... this one.
31214         (predicate_probability): Turn to ...
31215         (predicate::probability): ... this one.
31216         (dump_condition): Update.
31217         (dump_predicate): Turn to ...
31218         (predicate::dump): ... this one.
31219         (account_size_time): Update.
31220         (edge_set_predicate): Update.
31221         (set_hint_predicate): UPdate.
31222         (evaluate_conditions_for_known_args): Update.
31223         (evaluate_properties_for_edge): Update.
31224         (remap_predicate_after_duplication): Turn to...
31225         (predicate::remap_after_duplication): ... this one.
31226         (remap_hint_predicate_after_duplication): Update.
31227         (inline_summary_t::duplicate): UPdate.
31228         (dump_inline_edge_summary): Update.
31229         (dump_inline_summary): Update.
31230         (set_cond_stmt_execution_predicate): Update.
31231         (set_switch_stmt_execution_predicate): Update.
31232         (compute_bb_predicates): Update.
31233         (will_be_nonconstant_expr_predicate): Update.
31234         (will_be_nonconstant_predicate): Update.
31235         (phi_result_unknown_predicate): Update.
31236         (predicate_for_phi_result): Update.
31237         (array_index_predicate): Update.
31238         (estimate_function_body_sizes): Update.
31239         (estimate_node_size_and_time): Update.
31240         (estimate_ipcp_clone_size_and_time): Update.
31241         (remap_predicate): Rename to ...
31242         (predicate::remap_after_inlining): ... this one.
31243         (remap_hint_predicate): Update.
31244         (inline_merge_summary): Update.
31245         (inline_update_overall_summary): Update.
31246         (estimate_size_after_inlining): Update.
31247         (read_predicate): Rename to ...
31248         (predicate::stream_in): ... this one.
31249         (read_inline_edge_summary): Update.
31250         (write_predicate): Rename to ...
31251         (predicate::stream_out): ... this one.
31252         (write_inline_edge_summary): Update.
31253         * ipa-inline.h (MAX_CLAUSES): Turn to predicate::max_clauses.
31254         (clause_t): Turn to uint32_t
31255         (predicate): Turn to class; implement constructor and operators
31256         ==, !=, &
31257         (size_time_entry): Update.
31258         (inline_summary): Update.
31259         (inline_edge_summary): Update.
31261 2017-05-18  Marc Glisse  <marc.glisse@inria.fr>
31263         * fold-const.c (fold_binary_loc): Move transformation...
31264         * match.pd (C - X CMP X): ... here.
31266 2017-05-18  Sheldon Lobo  <sheldon.lobo@oracle.com>
31268         * config/sparc/sparc.c (sparc_option_override): Set function
31269         alignment for -mcpu=niagara7 to 64 to match the I$ line.
31270         * config/sparc/sparc.h (BRANCH_COST): Set the SPARC M7 branch
31271         latency to 1.
31272         * config/sparc/sparc.h (BRANCH_COST): Set the SPARC T4 branch
31273         latency to 2.
31274         * config/sparc/sol2.h: Fix a ASM_CPU32_DEFAULT_SPEC typo.
31276 2017-05-18  Marek Polacek  <polacek@redhat.com>
31278         PR sanitizer/80797
31279         * ubsan.c (instrument_null): Unwrap ADDR_EXPRs.
31280         (pass_ubsan::execute): Call gimple_assign_single_p instead of
31281         gimple_assign_load_p.
31283 2017-05-17  Segher Boessenkool  <segher@kernel.crashing.org>
31285         PR middle-end/80692
31286         * real.c (do_compare): Give decimal_do_compare preference over
31287         comparing just the signs.
31289 2017-05-17  Uros Bizjak  <ubizjak@gmail.com>
31291         * doc/md.texi (Canonicalization of Instructions): Describe the
31292         canonical form of instructions that inherently set a condition
31293         code register.
31295 2017-05-17  Peter Bergner  <bergner@vnet.ibm.com>
31297         PR middle-end/80775
31298         * tree-cfg.c: Move deletion of unreachable case statements to after
31299         the merging of consecutive case labels.
31301 2017-05-17  Thomas Preud'homme  <thomas.preudhomme@arm.com>
31303         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Refer
31304         readers to __gnu_cmse_nonsecure_call libcall for saving, clearing and
31305         restoring of callee-saved registers.
31307 2017-05-17  Eric Botcazou  <ebotcazou@adacore.com>
31309         * compare-elim.c (equivalent_reg_at_start): Adjust outdated comment.
31310         * config/visium/visium.c (single_set_and_flags): Likewise.
31311         * config/visium/visium.md (Substitutions): Likewise.
31313 2017-05-17  Martin Liska  <mliska@suse.cz>
31315         * cfg.c: Introduce dump_flags_t type and
31316         use it instead of int type.
31317         * cfg.h: Likewise.
31318         * cfghooks.c: Likewise.
31319         * cfghooks.h (struct cfg_hooks): Likewise.
31320         * cfgrtl.c: Likewise.
31321         * cfgrtl.h: Likewise.
31322         * cgraph.c (cgraph_node::get_body): Likewise.
31323         * coretypes.h: Likewise.
31324         * domwalk.c: Likewise.
31325         * domwalk.h: Likewise.
31326         * dumpfile.c (struct dump_option_value_info): Likewise.
31327         (dump_enable_all): Likewise.
31328         (dump_switch_p_1): Likewise.
31329         (opt_info_switch_p): Likewise.
31330         * dumpfile.h (enum tree_dump_index): Likewise.
31331         (struct dump_file_info): Likewise.
31332         * genemit.c: Likewise.
31333         * generic-match-head.c: Likewise.
31334         * gengtype.c (open_base_files): Likewise.
31335         * gimple-pretty-print.c: Likewise.
31336         * gimple-pretty-print.h: Likewise.
31337         * graph.c (print_graph_cfg): Likewise.
31338         * graphite-scop-detection.c (dot_all_sese): Likewise.
31339         * ipa-devirt.c (build_type_inheritance_graph): Likewise.
31340         * loop-unroll.c (report_unroll): Likewise.
31341         * passes.c (pass_manager::register_one_dump_file): Likewise.
31342         * print-tree.c: Likewise.
31343         * statistics.c: Likewise.
31344         * tree-cfg.c: Likewise.
31345         * tree-cfg.h: Likewise.
31346         * tree-dfa.c: Likewise.
31347         * tree-dfa.h: Likewise.
31348         * tree-dump.c (dump_function): Likewise.
31349         * tree-dump.h (struct dump_info): Likewise.
31350         * tree-pretty-print.c: Likewise.
31351         * tree-pretty-print.h: Likewise.
31352         * tree-ssa-live.c: Likewise.
31353         * tree-ssa-live.h: Likewise.
31354         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Likewise.
31355         * tree-vect-loop.c: Likewise.
31356         * tree-vect-slp.c: Likewise.
31358 2017-05-16  James Greenhalgh  <james.greenhalgh@arm.com>
31359             Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
31361         PR tree-optimization/80457
31362         * tree-vect-stmts.c (vect_model_simple_cost): Model the cost
31363         of all arguments to a statement as scalar_to_vec operations.
31364         (vectorizable_call): Adjust call to vect_model_simple_cost for
31365         new parameter.
31366         (vectorizable_conversion): Likewise.
31367         (vectorizable_assignment): Likewise.
31368         (vectorizable_shift): Likewise.
31369         (vectorizable_operation): Likewise.
31370         (vectorizable_comparison): Likewise.
31371         (vect_is_simple_cond): Record the def types for operands.
31372         (vectorizable_condition): Likewise, call vect_model_simple_cost.
31373         * tree-vectorizer.h (vect_model_simple_cost): Add new parameter
31374         for statement argument count.
31376 2017-05-16  Carl Love  <cel@us.ibm.com>
31378         * config/rs6000/rs6000-c: Add support for built-in functions
31379         vector unsigned long long vec_bperm (vector unsigned long long,
31380                                              vector unsigned char)
31381         vector signed long long vec_mule (vector signed int,
31382                                           vector signed int)
31383         vector unsigned long long vec_mule (vector unsigned int,
31384                                             vector unsigned int)
31385         vector signed long long vec_mulo (vector signed int,
31386                                           vector signed int)
31387         vector unsigned long long vec_mulo (vector unsigned int,
31388                                             vector unsigned int)
31389         vector signed char vec_sldw (vector signed char,
31390                                      vector signed char,
31391                                      const int)
31392         vector unsigned char vec_sldw (vector unsigned char,
31393                                        vector unsigned char,
31394                                        const int)
31395         vector signed short vec_sldw (vector signed short,
31396                                       vector signed short,
31397                                       const int)
31398         vector unsigned short vec_sldw (vector unsigned short,
31399                                         vector unsigned short,
31400                                         const int)
31401         vector signed int vec_sldw (vector signed int,
31402                                     vector signed int,
31403                                     const int)
31404         vector unsigned int vec_sldw (vector unsigned int,
31405                                       vector unsigned int,
31406                                       const int)
31407         vector signed long long vec_sldw (vector signed long long,
31408                                           vector signed long long,
31409                                           const int)
31410         vector unsigned long long vec_sldw (vector unsigned long long,
31411                                             vector unsigned long long,
31412                                             const int)
31413         * config/rs6000/rs6000-c: Add support for built-in functions
31414         * config/rs6000/rs6000-builtin.def: Add definition for SLDW.
31415         * config/rs6000/altivec.h: Add defintion for vec_sldw.
31416         * doc/extend.texi: Update the built-in documentation for the
31417         new built-in functions.
31419 2017-05-16  Marek Polacek  <polacek@redhat.com>
31421         PR sanitizer/80536
31422         PR sanitizer/80386
31423         * tree.c (save_expr): Don't fold the expression.
31425 2017-05-16  Uros Bizjak  <ubizjak@gmail.com>
31427         * config/i386/i386.md (*movsi_internal): Split (?rm,*y) alternative
31428         to (?r,*Yn) and (?m,*y) alternatives, and (?*y,rm) to (?*Ym,r)
31429         and (?*y,m).  Update insn attributes.
31431 2017-05-16  Martin Liska  <mliska@suse.cz>
31433         * cgraph.c (cgraph_edge::resolve_speculation): Add default value for
31434         flags argument of print_gimple_stmt, print_gimple_expr,
31435         print_generic_stmt and print_generic_expr.
31436         * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
31437         * coretypes.h: Likewise.
31438         * except.c (dump_eh_tree): Likewise.
31439         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
31440         * gimple-pretty-print.h: Likewise.
31441         * gimple-ssa-backprop.c (dump_usage_prefix): Likewise.
31442         (backprop::push_to_worklist): Likewise.
31443         (backprop::pop_from_worklist): Likewise.
31444         (backprop::process_use): Likewise.
31445         (backprop::intersect_uses): Likewise.
31446         (note_replacement): Likewise.
31447         * gimple-ssa-store-merging.c
31448         (pass_store_merging::terminate_all_aliasing_chains): Likewise.
31449         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
31450         (pass_store_merging::execute): Likewise.
31451         * gimple-ssa-strength-reduction.c (dump_candidate): Likewise.
31452         (ssa_base_cand_dump_callback): Likewise.
31453         (dump_incr_vec): Likewise.
31454         (replace_refs): Likewise.
31455         (replace_mult_candidate): Likewise.
31456         (create_add_on_incoming_edge): Likewise.
31457         (create_phi_basis): Likewise.
31458         (insert_initializers): Likewise.
31459         (all_phi_incrs_profitable): Likewise.
31460         (introduce_cast_before_cand): Likewise.
31461         (replace_one_candidate): Likewise.
31462         * gimplify.c (gimplify_expr): Likewise.
31463         * graphite-isl-ast-to-gimple.c (is_valid_rename): Likewise.
31464         (set_rename): Likewise.
31465         (rename_uses): Likewise.
31466         (copy_loop_phi_nodes): Likewise.
31467         (add_close_phis_to_merge_points): Likewise.
31468         (copy_loop_close_phi_args): Likewise.
31469         (copy_cond_phi_args): Likewise.
31470         (graphite_copy_stmts_from_block): Likewise.
31471         (translate_pending_phi_nodes): Likewise.
31472         * graphite-poly.c (print_pdr): Likewise.
31473         (dump_gbb_cases): Likewise.
31474         (dump_gbb_conditions): Likewise.
31475         (print_scop_params): Likewise.
31476         * graphite-scop-detection.c (build_cross_bb_scalars_def): Likewise.
31477         (build_cross_bb_scalars_use): Likewise.
31478         (gather_bbs::before_dom_children): Likewise.
31479         * hsa-dump.c (dump_hsa_immed): Likewise.
31480         * ipa-cp.c (print_ipcp_constant_value): Likewise.
31481         (get_replacement_map): Likewise.
31482         * ipa-inline-analysis.c (dump_condition): Likewise.
31483         (estimate_function_body_sizes): Likewise.
31484         * ipa-polymorphic-call.c (check_stmt_for_type_change): Likewise.
31485         (ipa_polymorphic_call_context::get_dynamic_type): Likewise.
31486         * ipa-prop.c (ipa_dump_param): Likewise.
31487         (ipa_print_node_jump_functions_for_edge): Likewise.
31488         (ipa_modify_call_arguments): Likewise.
31489         (ipa_modify_expr): Likewise.
31490         (ipa_dump_param_adjustments): Likewise.
31491         (ipa_dump_agg_replacement_values): Likewise.
31492         (ipcp_modif_dom_walker::before_dom_children): Likewise.
31493         * ipa-pure-const.c (check_stmt): Likewise.
31494         (pass_nothrow::execute): Likewise.
31495         * ipa-split.c (execute_split_functions): Likewise.
31496         * omp-offload.c (dump_oacc_loop_part): Likewise.
31497         (dump_oacc_loop): Likewise.
31498         * trans-mem.c (tm_log_emit): Likewise.
31499         (tm_memopt_accumulate_memops): Likewise.
31500         (dump_tm_memopt_set): Likewise.
31501         (dump_tm_memopt_transform): Likewise.
31502         * tree-cfg.c (gimple_verify_flow_info): Likewise.
31503         (print_loop): Likewise.
31504         * tree-chkp-opt.c (chkp_print_addr): Likewise.
31505         (chkp_gather_checks_info): Likewise.
31506         (chkp_get_check_result): Likewise.
31507         (chkp_remove_check_if_pass): Likewise.
31508         (chkp_use_outer_bounds_if_possible): Likewise.
31509         (chkp_reduce_bounds_lifetime): Likewise.
31510         * tree-chkp.c (chkp_register_addr_bounds): Likewise.
31511         (chkp_mark_completed_bounds): Likewise.
31512         (chkp_register_incomplete_bounds): Likewise.
31513         (chkp_mark_invalid_bounds): Likewise.
31514         (chkp_maybe_copy_and_register_bounds): Likewise.
31515         (chkp_build_returned_bound): Likewise.
31516         (chkp_get_bound_for_parm): Likewise.
31517         (chkp_build_bndldx): Likewise.
31518         (chkp_get_bounds_by_definition): Likewise.
31519         (chkp_generate_extern_var_bounds): Likewise.
31520         (chkp_get_bounds_for_decl_addr): Likewise.
31521         * tree-chrec.c (chrec_apply): Likewise.
31522         * tree-data-ref.c (dump_data_reference): Likewise.
31523         (dump_subscript): Likewise.
31524         (dump_data_dependence_relation): Likewise.
31525         (analyze_overlapping_iterations): Likewise.
31526         * tree-inline.c (expand_call_inline): Likewise.
31527         (tree_function_versioning): Likewise.
31528         * tree-into-ssa.c (dump_defs_stack): Likewise.
31529         (dump_currdefs): Likewise.
31530         (dump_names_replaced_by): Likewise.
31531         (dump_update_ssa): Likewise.
31532         (update_ssa): Likewise.
31533         * tree-object-size.c (pass_object_sizes::execute): Likewise.
31534         * tree-parloops.c (build_new_reduction): Likewise.
31535         (try_create_reduction_list): Likewise.
31536         (ref_conflicts_with_region): Likewise.
31537         (oacc_entry_exit_ok_1): Likewise.
31538         (oacc_entry_exit_single_gang): Likewise.
31539         * tree-pretty-print.h: Likewise.
31540         * tree-scalar-evolution.c (set_scalar_evolution): Likewise.
31541         (get_scalar_evolution): Likewise.
31542         (add_to_evolution): Likewise.
31543         (get_loop_exit_condition): Likewise.
31544         (analyze_evolution_in_loop): Likewise.
31545         (analyze_initial_condition): Likewise.
31546         (analyze_scalar_evolution): Likewise.
31547         (instantiate_scev): Likewise.
31548         (number_of_latch_executions): Likewise.
31549         (gather_chrec_stats): Likewise.
31550         (final_value_replacement_loop): Likewise.
31551         (scev_const_prop): Likewise.
31552         * tree-sra.c (dump_access): Likewise.
31553         (disqualify_candidate): Likewise.
31554         (create_access): Likewise.
31555         (reject): Likewise.
31556         (maybe_add_sra_candidate): Likewise.
31557         (create_access_replacement): Likewise.
31558         (analyze_access_subtree): Likewise.
31559         (analyze_all_variable_accesses): Likewise.
31560         (sra_modify_assign): Likewise.
31561         (initialize_constant_pool_replacements): Likewise.
31562         (find_param_candidates): Likewise.
31563         (decide_one_param_reduction): Likewise.
31564         (replace_removed_params_ssa_names): Likewise.
31565         * tree-ssa-ccp.c (ccp_fold_stmt): Likewise.
31566         * tree-ssa-copy.c (dump_copy_of): Likewise.
31567         (copy_prop_visit_cond_stmt): Likewise.
31568         * tree-ssa-dce.c (mark_operand_necessary): Likewise.
31569         * tree-ssa-dom.c (pass_dominator::execute): Likewise.
31570         (record_equivalences_from_stmt): Likewise.
31571         * tree-ssa-dse.c (compute_trims): Likewise.
31572         (delete_dead_call): Likewise.
31573         (delete_dead_assignment): Likewise.
31574         * tree-ssa-forwprop.c (forward_propagate_into_gimple_cond): Likewise.
31575         (forward_propagate_into_cond): Likewise.
31576         (pass_forwprop::execute): Likewise.
31577         * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
31578         * tree-ssa-loop-im.c (invariantness_dom_walker::before_dom_children):
31579         Likewise.
31580         (move_computations_worker): Likewise.
31581         (execute_sm): Likewise.
31582         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Likewise.
31583         (remove_exits_and_undefined_stmts): Likewise.
31584         (remove_redundant_iv_tests): Likewise.
31585         * tree-ssa-loop-ivopts.c (dump_use): Likewise.
31586         (adjust_iv_update_pos): Likewise.
31587         * tree-ssa-math-opts.c (bswap_replace): Likewise.
31588         * tree-ssa-phiopt.c (factor_out_conditional_conversion): Likewise.
31589         (value_replacement): Likewise.
31590         * tree-ssa-phiprop.c (phiprop_insert_phi): Likewise.
31591         * tree-ssa-pre.c (print_pre_expr): Likewise.
31592         (get_representative_for): Likewise.
31593         (create_expression_by_pieces): Likewise.
31594         (insert_into_preds_of_block): Likewise.
31595         (eliminate_insert): Likewise.
31596         (eliminate_dom_walker::before_dom_children): Likewise.
31597         (eliminate): Likewise.
31598         (remove_dead_inserted_code): Likewise.
31599         * tree-ssa-propagate.c (substitute_and_fold): Likewise.
31600         * tree-ssa-reassoc.c (get_rank): Likewise.
31601         (eliminate_duplicate_pair): Likewise.
31602         (eliminate_plus_minus_pair): Likewise.
31603         (eliminate_not_pairs): Likewise.
31604         (undistribute_ops_list): Likewise.
31605         (eliminate_redundant_comparison): Likewise.
31606         (update_range_test): Likewise.
31607         (optimize_range_tests_var_bound): Likewise.
31608         (optimize_vec_cond_expr): Likewise.
31609         (rewrite_expr_tree): Likewise.
31610         (rewrite_expr_tree_parallel): Likewise.
31611         (linearize_expr): Likewise.
31612         (break_up_subtract): Likewise.
31613         (linearize_expr_tree): Likewise.
31614         (attempt_builtin_powi): Likewise.
31615         (attempt_builtin_copysign): Likewise.
31616         (transform_stmt_to_copy): Likewise.
31617         (transform_stmt_to_multiply): Likewise.
31618         (dump_ops_vector): Likewise.
31619         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Likewise.
31620         (print_scc): Likewise.
31621         (set_ssa_val_to): Likewise.
31622         (visit_reference_op_store): Likewise.
31623         (visit_use): Likewise.
31624         (sccvn_dom_walker::before_dom_children): Likewise.
31625         (run_scc_vn): Likewise.
31626         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
31627         Likewise.
31628         (expr_hash_elt::print): Likewise.
31629         (const_and_copies::pop_to_marker): Likewise.
31630         (const_and_copies::record_const_or_copy_raw): Likewise.
31631         * tree-ssa-structalias.c (compute_dependence_clique): Likewise.
31632         * tree-ssa-uninit.c (collect_phi_def_edges): Likewise.
31633         (dump_predicates): Likewise.
31634         (find_uninit_use): Likewise.
31635         (warn_uninitialized_phi): Likewise.
31636         (pass_late_warn_uninitialized::execute): Likewise.
31637         * tree-ssa.c (verify_vssa): Likewise.
31638         (verify_ssa): Likewise.
31639         (maybe_optimize_var): Likewise.
31640         * tree-vrp.c (dump_value_range): Likewise.
31641         (dump_all_value_ranges): Likewise.
31642         (dump_asserts_for): Likewise.
31643         (register_edge_assert_for_2): Likewise.
31644         (vrp_visit_cond_stmt): Likewise.
31645         (vrp_visit_switch_stmt): Likewise.
31646         (vrp_visit_stmt): Likewise.
31647         (vrp_visit_phi_node): Likewise.
31648         (simplify_cond_using_ranges_1): Likewise.
31649         (fold_predicate_in): Likewise.
31650         (evrp_dom_walker::before_dom_children): Likewise.
31651         (evrp_dom_walker::push_value_range): Likewise.
31652         (evrp_dom_walker::pop_value_range): Likewise.
31653         (execute_early_vrp): Likewise.
31655 2017-05-16  Richard Biener  <rguenther@suse.de>
31657         * dwarf2out.c (loc_list_from_tree_1): Do not create
31658         DW_OP_GNU_variable_value for DECL_IGNORED_P decls.
31660 2017-05-16  Richard Biener  <rguenther@suse.de>
31662         * dwarf2out.c (resolve_variable_value_in_expr): Lookup DIE
31663         just generated.
31664         (note_variable_value_in_expr): If we resolved the decl ref
31665         do not push to the stack.
31667 2017-05-16  Matthew Wahab  <matthew.wahab@arm.com>
31669         * config/arm/arm_neon.h (vadd_f16): Use standard arithmetic
31670         operations in fast-math mode.
31671         (vaddq_f16): Likewise.
31672         (vmul_f16): Likewise.
31673         (vmulq_f16): Likewise.
31674         (vsub_f16): Likewise.
31675         (vsubq_f16): Likewise.
31676         * config/arm/neon.md (add<mode>3): New.
31677         (sub<mode>3): New.
31678         (fma:<VH:mode>3): New.  Also remove outdated comment.
31679         (mul<mode>3): New.
31681 2017-05-16  Martin Liska  <mliska@suse.cz>
31683         PR ipa/79849.
31684         PR ipa/79850.
31685         * ipa-devirt.c (warn_types_mismatch): Fix typo.
31686         (odr_types_equivalent_p): Likewise.
31688 2017-05-15  Sylvestre Ledru  <sylvestre@debian.org>
31690         * plugin.c (try_init_one_plugin): Fix ressource leaks (CID 726637).
31692 2017-05-15  Uros Bizjak  <ubizjak@gmail.com>
31694         PR target/80425
31695         * config/i386.i386.md (*zero_extendsidi2): Do not penalize
31696         non-interunit SSE move alternatives with '?'.
31697         (zero-extendsidi peephole2): New peephole to skip intermediate
31698         general register in SSE zero-extend sequence.
31700 2017-05-15  Jeff Law  <law@redhat.com>
31702         * reorg.c (relax_delay_slots): Create a new variable to hold
31703         the temporary target rather than clobbering TARGET_LABEL.
31705         * config/tilegx/tilegx.c (tilegx_expand_unaligned_load): Add
31706         missing argument to extract_bit_field call.
31707         * config/tilepro/tilepro.c (tilepro_expand_unaligned_load): Likewise.
31709 2017-05-15  Martin Liska  <mliska@suse.cz>
31711         PR driver/31468
31712         * gcc.c (process_command): Do not allow empty argument of -o option.
31714 2017-05-15  Renlin Li  <renlin.li@arm.com>
31716         * config/aarch64/aarch64-protos.h (aarch64_expand_call): Declare.
31717         * config/aarch64/aarch64.c (aarch64_expand_call): Define.
31718         * config/aarch64/constraints.md (Usf): Add long call check.
31719         * config/aarch64/aarch64.md (call): Use aarch64_expand_call.
31720         (call_value): Likewise.
31721         (sibcall): Likewise.
31722         (sibcall_value): Likewise.
31723         (call_insn): New.
31724         (call_value_insn): New.
31725         (sibcall_insn): Update rtx pattern.
31726         (sibcall_value_insn): Likewise.
31727         (call_internal): Remove.
31728         (call_value_internal): Likewise.
31729         (sibcall_internal): Likewise.
31730         (sibcall_value_internal): Likewise.
31731         (call_reg): Likewise.
31732         (call_symbol): Likewise.
31733         (call_value_reg): Likewise.
31734         (call_value_symbol): Likewise.
31736 2017-05-14  Krister Walfridsson  <krister.walfridsson@gmail.com>
31738         PR target/80600
31739         * config/netbsd.h (NETBSD_LIBGCC_SPEC): Always add -lgcc.
31741 2017-05-14  Uros Bizjak  <ubizjak@gmail.com>
31743         * config/i386.i386.c (ix86_cc_modes_compatible): CCNOmode is
31744         compatible with CCGOCmode and with CCZmode.
31746 2017-05-14  Martin Sebor  <msebor@redhat.com>
31748         PR middle-end/77671
31749         * gimple-fold.c (gimple_fold_builtin_sprintf): Make extern.
31750         (gimple_fold_builtin_snprintf): Same.
31751         * gimple-fold.h (gimple_fold_builtin_sprintf): Declare.
31752         (gimple_fold_builtin_snprintf): Same.
31753         * gimple-ssa-sprintf.c (get_format_string): Correct the detection
31754         of character types.
31755         (is_call_safe): New function.
31756         (try_substitute_return_value): Call it.
31757         (try_simplify_call): New function.
31758         (pass_sprintf_length::handle_gimple_call): Call it.
31760 2017-05-14  Martin Sebor  <msebor@redhat.com>
31762         PR middle-end/80669
31763         * builtins.c (expand_builtin_stpncpy): Simplify.
31765 2017-05-14  Daniel Santos  <daniel.santos@pobox.com>
31767         * config/i386/i386.opt: Add option -mcall-ms2sysv-xlogues.
31768         * config/i386/i386.h
31769         (x86_64_ms_sysv_extra_clobbered_registers): Change type to unsigned.
31770         (NUM_X86_64_MS_CLOBBERED_REGS): New macro.
31771         (struct machine_function): Add new members call_ms2sysv,
31772         call_ms2sysv_pad_in, call_ms2sysv_pad_out and call_ms2sysv_extra_regs.
31773         (struct machine_frame_state): New fields sp_realigned and
31774         sp_realigned_offset.
31775         * config/i386/i386.c
31776         (enum xlogue_stub): New enum.
31777         (enum xlogue_stub_sets): New enum.
31778         (class xlogue_layout): New class.
31779         (struct ix86_frame): New fields stack_realign_allocate_offset,
31780         stack_realign_offset and outlined_save_offset.  Modify comments to
31781         detail stack layout when using out-of-line stubs.
31782         (ix86_target_string): Add -mcall-ms2sysv-xlogues option.
31783         (ix86_option_override_internal): Add sorry() for TARGET_SEH and
31784         -mcall-ms2sysv-xlogues.
31785         (stub_managed_regs): New static variable.
31786         (ix86_save_reg): Add new parameter ignore_outlined to optionally omit
31787         registers managed by out-of-line stub.
31788         (disable_call_ms2sysv_xlogues): New function.
31789         (ix86_compute_frame_layout): Modify re-alignment calculations, disable
31790         m->call_ms2sysv when appropriate and compute frame layout for
31791         out-of-line stubs.
31792         (sp_valid_at, fp_valid_at): New inline functions.
31793         (choose_basereg): New function.
31794         (choose_baseaddr): Add align parameter, use choose_basereg and modify
31795         all callers.
31796         (ix86_emit_save_reg_using_mov, ix86_emit_restore_sse_regs_using_mov):
31797         Use align parameter of choose_baseaddr to generated aligned SSE movs
31798         when possible.
31799         (pro_epilogue_adjust_stack): Modify to track
31800         machine_frame_state::sp_realigned.
31801         (ix86_nsaved_regs): Modify to accommodate changes to ix86_save_reg.
31802         (ix86_nsaved_sseregs): Likewise.
31803         (ix86_emit_save_regs): Likewise.
31804         (ix86_emit_save_regs_using_mov): Likewise.
31805         (ix86_emit_save_sse_regs_using_mov): Likewise.
31806         (get_scratch_register_on_entry): Likewise.
31807         (gen_frame_set): New function.
31808         (gen_frame_load): Likewise.
31809         (gen_frame_store): Likewise.
31810         (emit_outlined_ms2sysv_save): Likewise.
31811         (emit_outlined_ms2sysv_restore): Likewise.
31812         (ix86_expand_prologue): Modify stack re-alignment code and call
31813         emit_outlined_ms2sysv_save when appropriate.
31814         (ix86_emit_leave): Clear machine_frame_state::sp_realigned.  Add
31815         parameter rtx_insn *insn, which allows the function to be used to only
31816         generate the notes.
31817         (ix86_expand_epilogue): Modify validity checks of frame and stack
31818         pointers, and call emit_outlined_ms2sysv_restore when appropriate.
31819         (ix86_expand_call): Modify to enable m->call_ms2sysv when appropriate.
31820         * config/i386/predicates.md
31821         (save_multiple): New predicate.
31822         (restore_multiple): Likewise.
31823         * config/i386/sse.md
31824         (save_multiple<mode>): New pattern.
31825         (save_multiple_realign<mode>): Likewise.
31826         (restore_multiple<mode>): Likewise.
31827         (restore_multiple_and_return<mode>): Likewise.
31828         (restore_multiple_leave_return<mode>): Likewise.
31829         * Makefile.in: Export HOSTCXX and HOSTCXXFLAGS to site.exp
31831 2017-05-14  Julia Koval  <julia.koval@intel.com>
31833         * config/i386/i386-builtin-types.def (VOID_FTYPE_INT_INT64): New type.
31834         * config/i386/i386-builtin.def (__builtin_ia32_xgetbv)
31835         (__builtin_ia32_xsetbv): New builtins.
31836         * config/i386/i386.c (ix86_expand_special_args_builtin):
31837         Process new types.
31838         (ix86_expand_builtin): Special expand for new intrinsics.
31839         * config/i386/i386.md (UNSPECV_XGETBV, UNSPECV_XSETBV): New.
31840         (xsetbv, xsetbv_rex64, xgetbv, xgetbv_rex64): New insn patterns.
31841         * config/i386/xsaveintrin.h (_xsetbv, _getbv): New intrinsics.
31843 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31845         * cfganal.c (inverted_post_order_compute): Change argument type
31846         to vec *.
31847         * cfganal.h (inverted_post_order_compute): Adjust prototype.
31848         * df-core.c (rest_of_handle_df_initialize): Adjust.
31849         (rest_of_handle_df_finish): Likewise.
31850         (df_analyze_1): Likewise.
31851         (df_analyze): Likewise.
31852         (loop_inverted_post_order_compute): Change argument to be a vec *.
31853         (df_analyze_loop): Adjust.
31854         (df_get_n_blocks): Likewise.
31855         (df_get_postorder): Likewise.
31856         * df.h (struct df_d): Change field to be a vec.
31857         * lcm.c (compute_laterin): Adjust.
31858         (compute_available): Likewise.
31859         * lra-lives.c (lra_create_live_ranges_1): Likewise.
31860         * tree-ssa-dce.c (remove_dead_stmt): Likewise.
31861         * tree-ssa-pre.c (compute_antic): Likewise.
31863 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31865         * cfganal.c (connect_infinite_loops_to_exit): Adjust.
31866         (depth_first_search::depth_first_search): Change structure init
31867         function to this constructor.
31868         (depth_first_search::add_bb): Rename function to this member.
31869         (depth_first_search::execute): Likewise.
31870         (flow_dfs_compute_reverse_finish): Adjust.
31872 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31874         * ddg.c (find_nodes_on_paths): Use auto_sbitmap.
31875         (longest_simple_path): Likewise.
31876         * shrink-wrap.c (spread_components): Likewise.
31877         (disqualify_problematic_components): Likewise.
31878         (emit_common_heads_for_components): Likewise.
31879         (emit_common_tails_for_components): Likewise.
31880         (insert_prologue_epilogue_for_components): Likewise.
31882 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31884         * tree-ssa-dse.c (dse_dom_walker): Make m_live_byes a
31885         auto_sbitmap.
31887 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31889         * df-core.c (df_set_blocks): Start using auto_bitmap.
31890         (df_compact_blocks): Likewise.
31891         * df-problems.c (df_rd_confluence_n): Likewise.
31892         * df-scan.c (df_insn_rescan_all): Likewise.
31893         (df_process_deferred_rescans): Likewise.
31894         (df_update_entry_block_defs): Likewise.
31895         (df_update_exit_block_uses): Likewise.
31896         (df_entry_block_bitmap_verify): Likewise.
31897         (df_exit_block_bitmap_verify): Likewise.
31898         (df_scan_verify): Likewise.
31899         * lra-constraints.c (lra_constraints): Likewise.
31900         (undo_optional_reloads): Likewise.
31901         (lra_undo_inheritance): Likewise.
31902         * lra-remat.c (calculate_gen_cands): Likewise.
31903         (do_remat): Likewise.
31904         * lra-spills.c (assign_spill_hard_regs): Likewise.
31905         (spill_pseudos): Likewise.
31906         * tree-ssa-pre.c (bitmap_set_and): Likewise.
31907         (bitmap_set_subtract_values): Likewise.
31909 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31911         * haifa-sched.c (estimate_shadow_tick): Replace manual bitmap
31912         management with auto_bitmap.
31913         (fix_inter_tick): Likewise.
31914         (fix_recovery_deps): Likewise.
31915         * ira.c (add_store_equivs): Likewise.
31916         (find_moveable_pseudos): Likewise.
31917         (split_live_ranges_for_shrink_wrap): Likewise.
31918         * print-rtl.c (rtx_reuse_manager::rtx_reuse_manager): Likewise.
31919         (rtx_reuse_manager::seen_def_p): Likewise.
31920         (rtx_reuse_manager::set_seen_def): Likewise.
31921         * print-rtl.h (class rtx_reuse_manager): Likewise.
31923 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31925         * bt-load.c (combine_btr_defs): Use auto_bitmap to manage bitmap
31926         lifetime.
31927         (migrate_btr_def): Likewise.
31928         * cfgloop.c (get_loop_body_in_bfs_order): Likewise.
31929         * df-core.c (loop_post_order_compute): Likewise.
31930         (loop_inverted_post_order_compute): Likewise.
31931         * hsa-common.h: Likewise.
31932         * hsa-gen.c (hsa_bb::~hsa_bb): Likewise.
31933         * init-regs.c (initialize_uninitialized_regs): Likewise.
31934         * ipa-inline.c (resolve_noninline_speculation): Likewise.
31935         (inline_small_functions): Likewise.
31936         * ipa-reference.c (ipa_reference_write_optimization_summary): Likewise.
31937         * ira.c (combine_and_move_insns): Likewise.
31938         (build_insn_chain): Likewise.
31939         * loop-invariant.c (find_invariants): Likewise.
31940         * lower-subreg.c (propagate_pseudo_copies): Likewise.
31941         * predict.c (tree_predict_by_opcode): Likewise.
31942         (predict_paths_leading_to): Likewise.
31943         (predict_paths_leading_to_edge): Likewise.
31944         (estimate_loops_at_level): Likewise.
31945         (estimate_loops): Likewise.
31946         * shrink-wrap.c (try_shrink_wrapping): Likewise.
31947         (spread_components): Likewise.
31948         * tree-cfg.c (remove_edge_and_dominated_blocks): Likewise.
31949         * tree-loop-distribution.c (rdg_build_partitions): Likewise.
31950         * tree-predcom.c (tree_predictive_commoning_loop): Likewise.
31951         * tree-ssa-coalesce.c (coalesce_ssa_name): Likewise.
31952         * tree-ssa-phionlycprop.c (pass_phi_only_cprop::execute): Likewise.
31953         * tree-ssa-pre.c (remove_dead_inserted_code): Likewise.
31954         * tree-ssa-sink.c (nearest_common_dominator_of_uses): Likewise.
31955         * tree-ssa-threadupdate.c (compute_path_counts): Likewise.
31956         (mark_threaded_blocks): Likewise.
31957         (thread_through_all_blocks): Likewise.
31958         * tree-ssa.c (verify_ssa): Likewise.
31959         (execute_update_addresses_taken): Likewise.
31960         * tree-ssanames.c (verify_ssaname_freelists): Likewise.
31962 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31964         * cfganal.c (mark_dfs_back_edges): Replace manual stack with
31965         auto_vec.
31966         (post_order_compute): Likewise.
31967         (inverted_post_order_compute): Likewise.
31968         (pre_and_rev_post_order_compute_fn): Likewise.
31970 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31972         * genrecog.c (int_set::int_set): Explicitly construct our
31973         auto_vec base class.
31974         * vec.h (auto_vec::auto_vec): New constructor.
31976 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31978         * bitmap.h (class auto_bitmap): New constructor taking
31979         bitmap_obstack * argument.
31981 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31983         * bitmap.h (class auto_bitmap): Change type of m_bits to
31984         bitmap_head, and adjust ctor / dtor and member operators.
31986 2017-05-13  Uros Bizjak  <ubizjak@gmail.com>
31988         * compare-elim.c (equivalent_reg_at_start): Return NULL_RTX
31989         when returned register mode doesn't match original mode.
31991 2017-05-12  Jeff Law  <law@redhat.com>
31992             Jakub Jelinek  <jakub@redhat.com>
31994         * config/mn10300/mn10300.c (mn10300_match_ccmode): Fix where
31995         we look for cc setter after the compare-elim changes.
31996         * config/mn10300/mn10300.md (addsi3_flags): Fix order of patterns
31997         within the vector to match what compare-elim now expects.
31998         (subsi3_flags, andsi3_flags, iorsi3_flags): Likewise.
31999         (xorsi3_flags, one_cmplsi2_flags): Likewise.
32001         * config/rx/rx.c (rx_match_ccmode): Fix where we look cc setter
32002         after the compare-elim changes.
32003         * config/rx/rx.md (abssi2_flags): Fix order of patterns within
32004         the vector to match what compare-elim now expects.
32005         (addsi3_flags, adc_flags, addsi3_flags peepholes): Likewise.
32006         (andsi3_flags, negsi2_flags, one_cmplsi2_flags): Likewise.
32007         (iorsi3_flags, rotlsi3_flags, rotrsi3_flags): Likewise.
32008         (ashrsi3_flags, lshrsi3_flags, ashlsi3_flags): Likewise.
32009         (ssaddsi3, subsi3_flags, sbb_flags, xorsi3_flags): Likewise.
32011         * config/visium/visium.c (single_set_and_flags): Fix where
32012         we look for cc setter after the compare-elim changes.
32013         * config/visium/visium.md (flags_subst_logic): Fix order of patterns
32014         with the vector to match what compare-elim now expects.
32015         (flags_subst_arith, add<mode>3_insn_set_carry): Likewise.
32016         (add<mode>3_insn_set_overflow, addsi3_insn_set_carry): Likewise.
32017         (addsi3_insn_set_overflow, sub<mode>3_insn_set_carry): Likewise.
32018         (sub<mode>3_insn_set_overflow, subsi3_insn_set_carry): Likewise.
32019         (subsi3_insn_set_overflow, negsi2_insn_set_carry): Likewise.
32020         (neg<mode>2_insn_set_overflow): Likewise.
32022 2017-05-12  Jim Wilson  <jim.wilson@linaro.org>
32024         PR middle-end/79794
32025         * expmed.c (extract_bit_field_1): Add alt_rtl argument.  Before
32026         maybe_expand_insn call, set ops[0].target.  If still set after call,
32027         set alt_rtl.  Add extra arg to recursive calls.
32028         (extract_bit_field): Add alt_rtl argument.  Pass to
32029         extract_bit_field.
32030         * expmed.h (extract_bit_field): Fix prototype.
32031         * expr.c (emit_group_load_1, copy_blkmode_from_reg)
32032         (copy_blkmode_to_reg, read_complex_part, store_field): Pass extra NULL
32033         to extract_bit_field_calls.
32034         (expand_expr_real_1): Pass alt_rtl to expand_expr_real instead of 0.
32035         Pass alt_rtl to extract_bit_field calls.
32036         * calls.c (store_unaligned_arguments_into_psuedos)
32037         load_register_parameters): Pass extra NULL to extract_bit_field calls.
32038         * optabs.c (maybe_legitimize_operand): Clear op->target when call
32039         gen_reg_rtx.
32040         * optabs.h (struct expand_operand): Add target bitfield.
32042 2017-05-12  Uros Bizjak  <ubizjak@gmail.com>
32044         * compare-elim.c (try_eliminate_compare): Canonicalize
32045         operation with embedded compare to
32046         [(set (reg:CCM) (compare:CCM (operation) (immediate)))
32047          (set (reg) (operation)].
32049         * config/i386/i386.c (TARGET_FLAGS_REGNUM): New define.
32051 2017-05-12  Uros Bizjak  <ubizjak@gmail.com>
32053         PR target/80723
32054         * config/i386/i386.c (ix86_rtx_cost) [case PLUS]: Ignore the
32055         cost of adding a carry flag for ADC instruction.
32056         [case MINUS]: Ignore the cost of subtracting a carry flag
32057         for SBB instruction.
32059 2017-05-12  Steven Munroe  <munroesj@gcc.gnu.org>
32061         * config.gcc (powerpc*-*-*): Add bmi2intrin.h, bmiintrin.h,
32062         and x86intrin.h
32063         * config/rs6000/bmiintrin.h: New file.
32064         * config/rs6000/bmi2intrin.h: New file.
32065         * config/rs6000/x86intrin.h: New file.
32067 2017-05-12  Jeff Law  <law@redhat.com>
32069         * tree-vrp.c (vrp_dom_walker::before_dom_children): Push unwinding
32070         markers.
32072 2017-05-12  Peter Bergner  <bergner@vnet.ibm.com>
32074         PR middle-end/80707
32075         * tree-cfg.c: Remove cfg edges of unreachable case statements.
32077 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
32079         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling for
32080         early expansion of vector divide builtins.
32081         (builtin_function_type): Add VSX_BUILTIN_UDIV_V2DI to the list of
32082         builtins identified as having unsigned arguments.
32084 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
32086         * config/rs6000/rs6000.c (gimple-fold.h): New #include.
32087         (rs6000_gimple_fold_builtin): Add handling for early GIMPLE
32088         expansion of vector logical operations (and, andc, or, xor,
32089         nor, orc, nand).
32091 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
32093         * gimple-fold.c (create_tmp_reg_or_ssa_name): Remove static declaration.
32094         * gimple-fold.h (create_tmp_reg_or_ssa_name): New prototype.
32096 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
32098         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling for
32099         early GIMPLE expansion of vector multiplies.
32101 2017-05-12  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
32103         * config/arm/arm.md (movsi): Add TARGET_32BIT in addition to the
32104         TARGET_HAVE_MOVT conditional.
32105         (movt splitter): Likewise.
32107 2017-05-12  Richard Biener  <rguenther@suse.de>
32109         * tree-ssa-sccvn.h (has_VN_INFO): Declare.
32110         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
32111         Fold all stmts not inplace.
32113 2017-05-12  Richard Biener  <rguenther@suse.de>
32115         PR tree-optimization/80713
32116         * tree-ssa-pre.c (remove_dead_inserted_code): Clear
32117         inserted_exprs bit for not removed stmts.
32119 2017-05-12  Thomas Schwinge  <thomas@codesourcery.com>
32121         PR middle-end/69921
32122         * tree-parloops.c (create_parallel_loop): Set "oacc kernels
32123         parallelized" attribute for parallelized OpenACC kernels.
32124         * omp-offload.c (execute_oacc_device_lower): Use it.
32126         * omp-expand.c (expand_omp_target) <GF_OMP_TARGET_KIND_OACC_KERNELS>:
32127         Set "oacc kernels" attribute.
32128         * omp-general.c (oacc_set_fn_attrib): Remove is_kernel formal
32129         parameter.  Adjust all users.
32130         (oacc_fn_attrib_kernels_p): Remove function.
32131         * omp-offload.c (execute_oacc_device_lower): Look for "oacc
32132         kernels" attribute instead of calling oacc_fn_attrib_kernels_p.
32133         * tree-ssa-loop.c (gate_oacc_kernels): Likewise.
32134         * tree-parloops.c (create_parallel_loop): If oacc_kernels_p,
32135         assert "oacc kernels" attribute is set.
32137 2017-05-11  Carl Love  <cel@us.ibm.com>
32139         * config/rs6000/rs6000-c: Add support for built-in functions
32140         vector unsigned char vec_popcnt (vector signed char)
32141         vector unsigned char vec_popcnt (vector unsigned char)
32142         vector unsigned short vec_popcnt (vector signed short)
32143         vector unsigned short vec_popcnt (vector unsigned short)
32144         vector unsigned int vec_popcnt (vector signed int)
32145         vector unsigned int vec_popcnt (vector unsigned int)
32146         vector unsigned long long vec_popcnt (vector signed long long)
32147         vector unsigned long long vec_popcnt (vector unsigned long long)
32148         vector signed long long vec_slo (vector signed long long,
32149                                          vector signed char)
32150         vector signed long long vec_slo (vector signed long long,
32151                                          vector unsigned char)
32152         vector unsigned long long vec_slo (vector unsigned long long,
32153                                            vector signed char)
32154         vector unsigned long long vec_slo (vector unsigned long long,
32155                                            vector unsigned char)
32156         * config/rs6000/rs6000-builtin.def: Add definitions for VPOPCNTUB,
32157         VPOPCNTUH, VPOPCNTUW, and VPOPCNTUD overloads.
32158         * config/rs6000/altivec.h: Add define for vec_popcnt, vec_popcntb,
32159         vec_popcnth, vec_popcntw and vec_popcntd built-in functions.
32160         * doc/extend.texi: Update the built-in documentation file for the
32161         new built-in functions.
32163 2017-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
32165         * attribs.h (sorted_attr_string): Move machine independent
32166         functions for target clone support from the i386 port to common
32167         code.  Rename ix86_function_versions to common_function_versions.
32168         Rename make_name to make_unique_name.
32169         (common_function_versions): Likewise.
32170         (make_unique_name): Likewise.
32171         (make_dispatcher_decl): Likewise.
32172         (is_function_default_version): Likewise.
32173         * attribs.c (attr_strcmp): Likewise.
32174         (sorted_attr_string): Likewise.
32175         (common_function_versions): Likewise.
32176         (make_unique_name): Likewise.
32177         (make_dispatcher_decl): Likewise.
32178         (is_function_default_version): Likewise.
32179         * config/i386/i386.c (attr_strcmp): Likewise.
32180         (sorted_attr_string): Likewise.
32181         (ix86_function_versions): Likewise.
32182         (make_name): Likewise.
32183         (make_dispatcher_decl): Likewise.
32184         (is_function_default_version): Likewise.
32185         (TARGET_OPTION_FUNCTION_VERSIONS): Update target function hook.
32187 2017-05-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32189         PR target/80695
32190         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost):
32191         Account for direct move costs for vec_construct of integer
32192         vectors.
32194 2017-05-11  Uros Bizjak  <ubizjak@gmail.com>
32196         PR target/80706
32197         * config/i386/sync.md (UNSPEC_LDX_ATOMIC): New unspec.
32198         (UNSPEC_STX_ATOMIC): Ditto.
32199         (loaddi_via_sse): New insn.
32200         (storedi_via_sse): Ditto.
32201         (atomic_loaddi_fpu): Emit loaddi_via_sse and storedi_via_sse.
32202         Update corresponding peephole2 patterns.
32203         (atomic_storedi_fpu): Ditto.
32205 2017-05-11  Julia Koval  <julia.koval@intel.com>
32207         * config/i386/avx512fintrin.h (_mm_mask_rsqrt14_sd)
32208         (_mm_maskz_rsqrt14_sd, _mm_mask_rsqrt14_ss, _mm_maskz_rsqrt14_ss):
32209         New intrinsics.
32210         * config/i386/i386-builtin.def (__builtin_ia32_rsqrt14sd_mask)
32211         (__builtin_ia32_rsqrt14ss_mask): New builtins.
32212         * config/i386/sse.md (rsqrt14_<mode>_mask): New pattern.
32214 2017-05-11  Nathan Sidwell  <nathan@acm.org>
32216         * graphite-poly.c: Include dumpfile.h.
32218         * dumpfle.h (dump_function): Declare here ...
32219         * tree-dump.h (dump_function): ... not here.
32220         * dumpfile.c: #include tree-cfg.h.
32221         (dump_function): Move here from ...
32222         * tree-dump.c (dump_function): ... here.
32223         * gimplify.c: #include splay-tree.h, not tree-dump.h.
32224         * graphite-poly.c: Don't include tree-dump.h.
32225         * cgraphclones.c: Include dumpfile.h not tree-dump.h.
32226         * print-tree.c: Likewise.
32227         * stor-layout.c: Likewise.
32228         * tree-nested.c: Likewise.
32230         * dumpfile.c (dump_start): Use TDF_FLAGS.
32231         (dump_enable_all): Fix TDF_KIND check thinko.
32233 2017-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
32235         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
32236         array entries to represent two legal parameterizations of the
32237         overloaded __builtin_cmpb function, as represented by the
32238         P6_OV_BUILTIN_CMPB constant.
32239         (altivec_resolve_overloaded_builtin): Add special case handling
32240         for the __builtin_cmpb function, as represented by the
32241         P6_OV_BUILTIN_CMPB constant.
32242         * config/rs6000/rs6000-builtin.def (BU_P6_2): New macro.
32243         (BU_P6_64BIT_2): New macro.
32244         (BU_P6_OVERLOAD_2): New macro
32245         (CMPB_32): Add 32-bit compare-bytes support for 32-bit only targets.
32246         (CMPB): Add 64-bit compare-bytes support for 32-bit and 64-bit targets.
32247         (CMPB): Add overload support to represent both 32-bit and 64-bit
32248         compare-bytes function.
32249         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add
32250         support for TARGET_CMPB.
32251         * config/rs6000/rs6000.h: Add support for RS6000_BTM_CMPB.
32252         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
32253         documentation of the __builtin_cmpb overloaded built-in function.
32255 2017-05-11  Richard Biener  <rguenther@suse.de>
32257         PR tree-optimization/80705
32258         * tree-vect-data-refs.c (vect_analyze_data_refs): DECL_NONALIASED
32259         bases are not vectorizable.
32261 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32263         * tree-ssa-loop-ivopts.c (determine_set_costs): Skip non-interger
32264         when counting register pressure.
32266 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32268         * tree-ssa-loop-ivopts.c (compare_cost_pair): New.
32269         (iv_ca_more_deps): Renamed to ...
32270         (iv_ca_compare_deps): ... this.
32271         (iv_ca_extend): Extend iv_ca if NEW_CP is cheaper than OLD_CP.
32273 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32275         * tree-ssa-loop-ivopts.c (find_interesting_uses): Move inv vars dump
32276         to ...
32277         (determine_group_iv_costs): ... here.
32278         (find_inv_vars_cb): Record inv var if it's not recorded before.
32280 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32282         * tree-ssa-loop-ivopts.c (allow_ip_end_pos_p): Refine comments.
32283         (get_shiftadd_cost): Ditto.
32285 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32287         * tree-ssa-address.c: Include header file.
32288         (move_hint_to_base): Return TRUE if BASE_HINT is moved to memory
32289         address.
32290         (add_to_parts): Refactor.
32291         (addr_to_parts): New parameter.  Update use of move_hint_to_base.
32292         (create_mem_ref): Update use of addr_to_parts.  Re-associate addr
32293         in new order.
32295 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32297         PR tree-optimization/53090
32298         * tree-ssa-loop-ivopts.c (enum comp_iv_rewrite): New enum value
32299         COMP_IV_EXPR_2.
32300         (extract_cond_operands): Detect condition with IV on both sides
32301         and return COMP_IV_EXPR_2.
32302         (find_interesting_uses_cond): Add iv_use for both IVs in condition.
32303         (rewrite_use_compare): Simplify by removing call to function
32304         extract_cond_operands.
32306 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32308         * tree-ssa-loop-ivopts.c (enum comp_iv_rewrite): New.
32309         (extract_cond_operands): Detect condition comparing against non-
32310         invariant bound and return appropriate enum value.
32311         (find_interesting_uses_cond): Update use of extract_cond_operands.
32312         Handle its return value accordingly.
32313         (determine_group_iv_cost_cond, rewrite_use_compare): Ditto.
32315 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32317         * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Re-associate
32318         nonlinear iv_use computation in loop invariant sensitive way.
32320 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32322         * tree-ssa-loop-ivopts.c (relate_compare_use_with_all_cands): New.
32323         (find_iv_candidates): Call relate_compare_use_with_all_cands.
32325 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32327         * tree-ssa-loop-ivopts.c (struct iv_cand): New field inv_exprs.
32328         (dump_cand): Support iv_cand.inv_exprs.
32329         (add_candidate_1): Record invariant exprs in iv_cand.inv_exprs
32330         for candidates.
32331         (iv_ca_set_no_cp, iv_ca_set_cp, free_loop_data): Support
32332         iv_cand.inv_exprs.
32334 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32336         * tree-ssa-loop-ivopts.c (multiplier_allowed_in_address_p): Move
32337         from ...
32338         * tree-ssa-address.c (multiplier_allowed_in_address_p): ... to here
32339         as local function.  Include necessary header files.
32340         * tree-ssa-loop-ivopts.h (multiplier_allowed_in_address_p): Delete.
32342 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32344         * tree-ssa-loop-ivopts.c (autoinc_possible_for_pair): Simplify.
32346 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32348         * tree-ssa-loop-ivopts.c (force_expr_to_var_cost): Handle more
32349         operators: TRUNC_DIV_EXPR, BIT_AND_EXPR, BIT_IOR_EXPR, LSHIFT_EXPR,
32350         RSHIFT_EXPR and BIT_NOT_EXPR.
32352 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32354         * tree-ssa-loop-ivopts.c (get_loop_invariant_expr): Simplify.
32355         (adjust_setup_cost): New parameter supporting round up adjustment.
32356         (struct address_cost_data): Delete.
32357         (force_expr_to_var_cost): Don't bound cost with spill_cost.
32358         (split_address_cost, ptr_difference_cost): Delete.
32359         (difference_cost, compare_aff_trees, record_inv_expr): Delete.
32360         (struct ainc_cost_data): New struct.
32361         (get_address_cost_ainc): New function.
32362         (get_address_cost, get_computation_cost): Reimplement.
32363         (determine_group_iv_cost_address): Record inv_expr for all uses of
32364         a group.
32365         (determine_group_iv_cost_cond): Call get_loop_invariant_expr.
32366         (iv_ca_has_deps): Reimplemented to ...
32367         (iv_ca_more_deps): ... this.  Check if NEW_CP introduces more deps
32368         than OLD_CP.
32369         (iv_ca_extend): Call iv_ca_more_deps.
32371 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32373         * tree-ssa-address.c (struct mem_address): Move to header file.
32374         (valid_mem_ref_p, move_fixed_address_to_symbol): Make it global.
32375         * tree-ssa-address.h (struct mem_address): Move from C file.
32376         (valid_mem_ref_p, move_fixed_address_to_symbol): Declare.
32378 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32380         * tree-affine.h (aff_combination_type): New interface.
32381         (aff_combination_zero_p): Remove static.
32382         (aff_combination_const_p): New interface.
32383         (aff_combination_singleton_var_p): New interfaces.
32385 2017-05-11  Richard Biener  <rguenther@suse.de>
32387         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
32388         Skip unreachable blocks and destinations.
32389         (eliminate): Move stmt removal and fixup ...
32390         (fini_eliminate): ... here.  Skip inserted exprs.
32391         (pass_pre::execute): Move fini_pre after fini_eliminate.
32392         * tree-ssa-tailmerge.c: Include tree-cfgcleanup.h.
32393         (tail_merge_optimize): Run cleanup_tree_cfg if requested by
32394         PRE to get rid of dead code that has invalid SSA form and
32395         split critical edges again.
32397 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32399         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
32401 2017-05-11  Richard Biener  <rguenther@suse.de>
32403         * passes.c (execute_function_todo): Verify loops if they are
32404         said to be up-to-date.
32405         * cfgexpand.c (pass_expand::execute): Discard loops for -dx.
32406         * trans-mem.c (pass_tm_edges::execute): Mark loops for fixup.
32408 2017-05-10  John David Anglin  <danglin@gcc.gnu.org>
32410         PR target/80090
32411         * config/pa/pa.c (pa_assemble_integer): When outputting a SYMBOL_REF,
32412         handle calling assemble_external ourself.
32414         PR target/79027
32415         * config/pa/pa.c (pa_cannot_change_mode_class): Reject changes to/from
32416         modes with zero size.  Enhance comment.
32418 2017-05-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32420         * config/rs6000/rs6000.c (altivec_init_builtins): Define POWER8
32421         built-ins for vec_xl and vec_xst with short and char pointer
32422         arguments.
32424 2017-05-10  Sebastian Peryt  <sebastian.peryt@intel.com>
32426         * config/i386/avx512fintrin.h (_mm_mask_max_round_sd)
32427         (_mm_maskz_max_round_sd, _mm_mask_max_round_ss)
32428         (_mm_maskz_max_round_ss, _mm_mask_min_round_sd)
32429         (_mm_maskz_min_round_sd, _mm_mask_min_round_ss)
32430         (_mm_maskz_min_round_ss): New intrinsics.
32431         * config/i386/i386-builtin-types.def (V2DF, V2DF, V2DF, V2DF, UQI, INT)
32432         (V4SF, V4SF, V4SF, V4SF, UQI, INT): New function type aliases.
32433         * config/i386/i386-builtin.def (__builtin_ia32_maxsd_mask_round)
32434         (__builtin_ia32_maxss_mask_round, __builtin_ia32_minsd_mask_round)
32435         (__builtin_ia32_minss_mask_round): New builtins.
32436         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32437         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
32438         * config/i386/sse.md (<sse>_vm<code><mode>3<round_saeonly_name>):
32439         Rename to ...
32440         (<sse>_vm<code><mode>3<mask_name><round_saeonly_name>): ... this.
32441         (v<maxmin_float><ssescalarmodesuffix>\t{<round_saeonly_op3>%2, %1, %0|%0, %1, %<iptr>2<round_saeonly_op3>}):
32442         Change to ...
32443         (v<maxmin_float><ssescalarmodesuffix>\t{<round_saeonly_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_saeonly_mask_op3>}):
32444         ... this.
32446 2017-05-10  Sebastian Peryt  <sebastian.peryt@intel.com>
32448         * config/i386/avx512fintrin.h (_mm_mask_mul_round_sd)
32449         (_mm_maskz_mul_round_sd, _mm_mask_mul_round_ss)
32450         (_mm_maskz_mul_round_ss, _mm_mask_div_round_sd)
32451         (_mm_maskz_div_round_sd, _mm_mask_div_round_ss)
32452         (_mm_maskz_div_round_ss, _mm_mask_mul_sd, _mm_maskz_mul_sd)
32453         (_mm_mask_mul_ss, _mm_maskz_mul_ss, _mm_mask_div_sd)
32454         (_mm_maskz_div_sd, _mm_mask_div_ss, _mm_maskz_div_ss): New intrinsics.
32455         * config/i386/i386-builtin-types.def (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32456         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): New function type aliases.
32457         * config/i386/i386-builtin.def (__builtin_ia32_divsd_mask_round)
32458         (__builtin_ia32_divss_mask_round, __builtin_ia32_mulsd_mask_round)
32459         (__builtin_ia32_mulss_mask_round): New builtins.
32460         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32461         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
32462         * config/i386/sse.md (<sse>_vm<multdiv_mnemonic><mode>3<round_name>):
32463         Rename to ...
32464         (<sse>_vm<multdiv_mnemonic><mode>3<mask_name><round_name>): ... this.
32465         (v<multdiv_mnemonic><ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|%0, %1, %<iptr>2<round_op3>}):
32466         Change to ...
32467         (v<multdiv_mnemonic><ssescalarmodesuffix>\t{<round_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}):
32468         ... this.
32470 2017-05-10  Julia Koval  <julia.koval@intel.com>
32472         * config/i386/avxintrin.h (_mm256_set_m128, _mm256_set_m128d)
32473         (_mm256_set_m128i, _mm256_setr_m128, _mm256_setr_m128d)
32474         (_mm256_setr_m128i): New intrinsics.
32476 2017-05-10  Julia Koval  <julia.koval@intel.com>
32478         * config/i386/avx512fintrin.h (_mm_mask_rcp14_sd)
32479         (_mm_maskz_rcp14_sd, _mm_mask_rcp14_ss)
32480         (_mm_maskz_rcp14_ss): New intrinsics.
32481         * config/i386/i386-builtin.def (__builtin_ia32_rcp14sd_mask)
32482         (__builtin_ia32_rcp14ss_mask): New builtins.
32483         * config/i386/sse.md (srcp14<mode>_mask): New pattern.
32485 2017-05-10  Peter Bergner  <bergner@vnet.ibm.com>
32487         PR tree-optimization/51513
32488         * tree-cfg.c (gimple_seq_unreachable_p): New function.
32489         (assert_unreachable_fallthru_edge_p): Use it.
32490         (group_case_labels_stmt): Likewise.
32491         * tree-cfg.h: Prototype it.
32492         * stmt.c: Include cfghooks.h and tree-cfg.h.
32493         (emit_case_dispatch_table) <gap_label>: New local variable.
32494         Use it to fill dispatch table gaps.
32495         Test for default_label before updating probabilities.
32496         (expand_case) <default_label>: Remove unneeded initialization.
32497         Test for unreachable default case statement and remove its edge.
32498         Set default_label accordingly.
32499         * tree-ssa-ccp.c (optimize_unreachable): Update comment.
32501 2017-05-10  Carl Love  <cel@us.ibm.com>
32503         * config/rs6000/rs6000-c: Add support for built-in functions
32504         vector signed char      vec_neg (vector signed char)
32505         vector signed short int vec_neg (vector short int)
32506         vector signed int       vec_neg (vector signed int)
32507         vector signed long long vec_neg (vector signed long long)
32508         vector float            vec_neg (vector float)
32509         vector double           vec_neg (vector double)
32510         * config/rs6000/rs6000-builtin.def: Add definitions for NEG function
32511         overload.
32512         * config/rs6000/altivec.h: Add define for vec_neg
32513         * doc/extend.texi: Update the built-in documentation for the
32514         new built-in functions.
32516 2017-05-10  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
32518         PR tree-optimization/77644
32519         * match.pd (sqrt(x) cmp sqrt(y) -> x cmp y): New pattern.
32521 2017-05-10  Nathan Sidwell  <nathan@acm.org>
32523         * dumpfile.h (TDI_lang_all): New.
32524         (TDF_KIND): New. Renumber others
32525         (TDF_LANG, TDF_TREE, TDF_RTL, TDF_IPA): Enumerate value, rather
32526         than bits.
32527         * dumpfile.c (dump_files): Mark language dumps as TDF_LANG.  add
32528         lang-all.
32529         (get_dump_file_name): Adjust suffix generation.
32530         (dump_enable_all): Use TDF_KIND.
32531         * doc/invoke.texi (-fdump-lang-all): Document.
32533         * dumpfile.h: Tabify.
32535 2017-05-10  Wilco Dijkstra  <wdijkstr@arm.com>
32537         PR target/80671
32538         * config/aarch64/cortex-a57-fma-steering.c (merge_forest):
32539         Move member access before delete.
32541 2017-05-10  Alexandre Oliva  <aoliva@redhat.com>
32543         * tree-inline.c (expand_call_inline): Split block at stmt
32544         before the call.
32546 2017-05-09  Michael Meissner  <meissner@linux.vnet.ibm.com>
32548         PR target/68163
32549         * config/rs6000/rs6000.md (f32_lr): Delete mode attributes that
32550         are now unused after splitting mov{sf,sd}_hardfloat.
32551         (f32_lr2): Likewise.
32552         (f32_lm): Likewise.
32553         (f32_lm2): Likewise.
32554         (f32_li): Likewise.
32555         (f32_li2): Likewise.
32556         (f32_lv): Likewise.
32557         (f32_sr): Likewise.
32558         (f32_sr2): Likewise.
32559         (f32_sm): Likewise.
32560         (f32_sm2): Likewise.
32561         (f32_si): Likewise.
32562         (f32_si2): Likewise.
32563         (f32_sv): Likewise.
32564         (f32_dm): Likewise.
32565         (f32_vsx): Likewise.
32566         (f32_av): Likewise.
32567         (mov<mode>_hardfloat): Split into separate movsf and movsd pieces.
32568         For movsf, order stores so the VSX stores occur before the GPR
32569         store which encourages the register allocator to use a traditional
32570         FPR instead of a GPR.  For movsd, order the stores so that the GPR
32571         store comes before the VSX stores to allow the power6 to work.
32572         This is due to the power6 not having a 32-bit integer store
32573         instruction from a FPR.
32574         (movsf_hardfloat): Likewise.
32575         (movsd_hardfloat): Likewise.
32577 2017-05-09  Martin Sebor  <msebor@redhat.com>
32579         PR translation/80280
32580         * config/sol2-c.c (cmn_err_flag_specs): Initialize new data member
32581         added in r247778.
32583         PR translation/80280
32584         * config/i386/msformat-c.c (ms_printf_flag_specs): Initialize new
32585         data member added in r247778.
32586         (ms_scanf_flag_specs, ms_strftime_flag_specs): Same.
32588 2017-05-09  Nathan Sidwell  <nathan@acm.org>
32590         * tree.h (tree_fits_shwi_p, tree_fits_uhwi_p): Unconditionally pure.
32592         * ipa-devirt.c (default_hash_traits<type_pair>): Skip struct-scope
32593         typedefs.
32595 2017-05-09  Marek Polacek  <polacek@redhat.com>
32597         * doc/invoke.texi: Fix typo.
32599 2017-05-09  Richard Biener  <rguenther@suse.de>
32601         * tree-vrp.c (vrp_val_is_max): Adjust comment.
32602         (vrp_val_is_min): Likewise.
32603         (set_value_range_to_value): Likewise.
32604         (set_value_range_to_nonnegative): Likewise.
32605         (gimple_assign_nonzero_p): Likewise.
32606         (gimple_stmt_nonzero_p): Likewise.
32607         (vrp_int_const_binop): Likewise.  Remove unreachable case.
32608         (adjust_range_with_scev): Adjust comments.
32609         (compare_range_with_value): Likewise.
32610         (extract_range_from_phi_node): Likewise.
32611         (test_for_singularity): Likewise.
32613 2017-05-09  Richard Biener  <rguenther@suse.de>
32615         * tree-vrp.c (get_single_symbol): Add assert that we don't
32616         get overflowed constants as invariant part.
32617         (compare_values_warnv): Add comment before the TREE_NO_WARNING
32618         checks.  Use wi::cmp instead of recursing for integer constants.
32619         (compare_values): Just ignore whether we assumed undefined
32620         overflow instead of failing the compare.
32621         (extract_range_for_var_from_comparison_expr): Add comment before the
32622         TREE_NO_WARNING sets.
32623         (test_for_singularity): Likewise.
32624         (extract_range_from_comparison): Do not disable optimization
32625         when we assumed undefined overflow.
32626         (extract_range_basic): Remove init of unused var.
32628 2017-05-09  Richard Biener  <rguenther@suse.de>
32630         * tree-vrp.c (vrp_int_const_binop): Use wide-ints and simplify.
32631         (extract_range_from_multiplicative_op_1): Adjust.
32632         (extract_range_from_binary_expr_1): Use int_const_binop.
32634 2017-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
32636         PR target/80101
32637         * config/rs6000/power6.md: Replace store_data_bypass_p calls with
32638         rs6000_store_data_bypass_p in seven define_bypass directives and
32639         in several comments.
32640         * config/rs6000/rs6000-protos.h: Add prototype for
32641         rs6000_store_data_bypass_p function.
32642         * config/rs6000/rs6000.c (rs6000_store_data_bypass_p): New
32643         function implements slightly different (rs6000-specific) semantics
32644         than store_data_bypass_p, returning false rather than aborting
32645         with assertion error when arguments do not satisfy the
32646         requirements of store data bypass.
32647         (rs6000_adjust_cost): Replace six calls of store_data_bypass_p with
32648         rs6000_store_data_bypass_p.
32650 2017-05-08  Max Filippov  <jcmvbkbc@gmail.com>
32652         * config/xtensa/xtensa-protos.h
32653         (xtensa_initial_elimination_offset): New declaration.
32654         * config/xtensa/xtensa.c (xtensa_initial_elimination_offset):
32655         New function. Move its body from the INITIAL_ELIMINATION_OFFSET
32656         macro definition, add case for FRAME_POINTER_REGNUM when
32657         FRAME_GROWS_DOWNWARD.
32658         * config/xtensa/xtensa.h (FRAME_GROWS_DOWNWARD): New macro definition.
32659         (INITIAL_ELIMINATION_OFFSET): Replace body with call to
32660         xtensa_initial_elimination_offset.
32662 2017-05-08  Nathan Sidwell  <nathan@acm.org>
32664         * doc/invoke.texi: Alphabetize -fdump options.
32666 2017-05-08  Martin Sebor  <msebor@redhat.com>
32668         PR translation/80280
32669         * config/sol2-c.c (solaris_pragma_align): Correct quoting.
32671 2017-05-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
32673         * target.def (compute_frame_layout): New optional target hook.
32674         * doc/tm.texi.in (TARGET_COMPUTE_FRAME_LAYOUT): Add hook.
32675         * doc/tm.texi (TARGET_COMPUTE_FRAME_LAYOUT): Add documentation.
32676         * lra-eliminations.c (update_reg_eliminate): Call compute_frame_layout
32677         target hook.
32678         * reload1.c (verify_initial_elim_offsets): Likewise.
32679         * config/arm/arm.c (TARGET_COMPUTE_FRAME_LAYOUT): Define.
32680         (use_simple_return_p): Call arm_compute_frame_layout if needed.
32681         (arm_get_frame_offsets): Split up into this ...
32682         (arm_compute_frame_layout): ... and this function.
32684 2017-05-08  Richard Sandiford  <richard.sandiford@arm.com>
32686         * config/aarch64/constraints.md (Usa): New constraint.
32687         * config/aarch64/aarch64.md (*movsi_aarch64, *movdi_aarch64): Use it.
32689 2017-05-08  Thomas Preud'homme  <thomas.preudhomme@arm.com>
32691         * config.gcc (arm*-*-*): Set TM_MULTILIB_CONFIG from
32692         with_multilib_list after it has been checked.
32694 2017-05-08  Richard Biener  <rguenther@suse.de>
32696         * tree-ssa-pre.c (bitmap_set_and): Avoid bitmap copy.
32697         (bitmap_set_subtract_values): Likewise.
32699 2017-05-08  Richard Biener  <rguenther@suse.de>
32701         * tree-vrp.c (gimple_assign_nonzero_warnv_p): Rename to ...
32702         (gimple_assign_nonzero): ... this and remove strict_overflow_p
32703         argument.
32704         (gimple_stmt_nonzero_warnv_p): Rename to ...
32705         (gimple_stmt_nonzero_p): ... this and remove strict_overflow_p
32706         argument.
32707         (vrp_stmt_computes_nonzero): Remove strict_overflow_p argument.
32708         (extract_range_basic): Adjust, do not disable propagation on
32709         strict overflow sensitive simplification.
32710         (vrp_visit_cond_stmt): Likewise.
32712 2017-05-05  Jan Hubicka  <hubicka@ucw.cz>
32714         * ipa-inline-analysis.c (estimate_function_body_sizes): Recompute
32715         body size unconditionally.
32717 2017-05-07  Jeff Law  <law@redhat.com>
32719         Revert:
32720         2017-05-06  Jeff Law  <law@redhat.com>
32721         PR tree-optimization/78496
32722         * tree-vrp.c (simplify_assert_expr_using_ranges): Remove debugging
32723         code.
32725         PR tree-optimization/78496
32726         * tree-vrp.c (simplify_assert_expr_using_ranges): New function.
32727         (simplify_stmt_using_ranges): Call it.
32728         (vrp_dom_walker::before_dom_children): Extract equivalences
32729         from an ASSERT_EXPR with an equality comparison against a
32730         constant.
32732 2017-05-06  Jeff Law  <law@redhat.com>
32734         PR tree-optimization/78496
32735         * tree-vrp.c (simplify_assert_expr_using_ranges): Remove debugging
32736         code.
32738         PR tree-optimization/78496
32739         * tree-vrp.c (simplify_assert_expr_using_ranges): New function.
32740         (simplify_stmt_using_ranges): Call it.
32741         (vrp_dom_walker::before_dom_children): Extract equivalences
32742         from an ASSERT_EXPR with an equality comparison against a
32743         constant.
32745 2017-05-06  Richard Sandiford  <richard.sandiford@linaro.org>
32747         * lra-constraints.c (lra_copy_reg_equiv): New function.
32748         (split_reg): Use it to copy equivalence information from the
32749         original register to the spill register.
32751 2017-05-06  Richard Sandiford  <richard.sandiford@linaro.org>
32753         PR rtl-optimization/75964
32754         * simplify-rtx.c (simplify_const_relational_operation): Remove
32755         invalid handling of comparisons of integer ABS.
32757 2017-05-06  Uros Bizjak  <ubizjak@gmail.com>
32759         * config/i386/i386.c (ext_80387_constant_init): Do not explicitly
32760         initialize to zero.
32761         (init_regs): Remove declaration.
32762         (function_arg_advance_32): Initialize error_p as boolean variable.
32764 2017-05-05  Nathan Sidwell  <nathan@acm.org>
32766         * store-motion.c (remove_reachable_equiv_notes): Reformat long
32767         lines.  Use for (;;).
32769 2017-05-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32771         * config/rs6000/rs6000.c (rs6000_vect_nonmem): New static var.
32772         (rs6000_init_cost): Initialize rs6000_vect_nonmem.
32773         (rs6000_add_stmt_cost): Update rs6000_vect_nonmem.
32774         (rs6000_finish_cost): Avoid vectorizing simple copy loops with
32775         VF=2 that require versioning.
32777 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32779         * diagnostic.h (CARET_LINE_MARGIN): Convert from macro to const
32780         int.
32782 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32784         * diagnostic.h (diagnostic_override_option_index): Convert from
32785         macro to inline function.
32787 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32789         * diagnostic.c (last_module_changed_p): New function.
32790         (set_last_module): New function.
32791         (diagnostic_report_current_module): Convert macro usage to
32792         the above functions.
32793         * diagnostic.h (diagnostic_context::last_module): Strengthen
32794         from const line_map * to const line_map_ordinary *.
32795         (diagnostic_last_module_changed): Delete macro.
32796         (diagnostic_set_last_module): Delete macro.
32798 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32800         * diagnostic.c (diagnostic_impl): Replace report_diagnostic
32801         with diagnostic_report_diagnostic.
32802         (diagnostic_n_impl_richloc): Likewise.
32803         * diagnostic.h (report_diagnostic): Delete macro.
32804         * rtl-error.c (diagnostic_for_asm): Replace report_diagnostic
32805         with diagnostic_report_diagnostic.
32806         * substring-locations.c (format_warning_va): Likewise.
32808 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32810         * diagnostic.c (diagnostic_report_diagnostic): Eliminate
32811         save/restor of format_spec.  Move option-printing code to...
32812         (print_option_information): ...this new function, and
32813         reimplement by simply printing to the pretty_printer,
32814         rather than appending to the format string.
32816 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32818         * diagnostic.c (diagnostic_report_diagnostic): Split out pragma
32819         handling logic into...
32820         (update_effective_level_from_pragmas): ...this new function.
32822 2017-05-04  Andrew Waterman  <andrew@sifive.com>
32824         * config/riscv/riscv.opt (mstrict-align): New option.
32825         * config/riscv/riscv.h (STRICT_ALIGNMENT): Use it.  Update comment.
32826         (SLOW_UNALIGNED_ACCESS): Define.
32827         (riscv_slow_unaligned_access): Declare.
32828         * config/riscv/riscv.c (riscv_tune_info): Add slow_unaligned_access
32829         field.
32830         (riscv_slow_unaligned_access): New variable.
32831         (rocket_tune_info): Set slow_unaligned_access to true.
32832         (optimize_size_tune_info): Set slow_unaligned_access to false.
32833         (riscv_cpu_info_table): Add entry for optimize_size_tune_info.
32834         (riscv_valid_lo_sum_p): Use TARGET_STRICT_ALIGN.
32835         (riscv_option_override): Set riscv_slow_unaligned_access.
32836         * doc/invoke.texi: Add -mstrict-align to RISC-V.
32838 2017-05-04  Kito Cheng  <kito.cheng@gmail.com>
32840         * config/riscv/riscv.md: Unify indentation.
32842 2017-05-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
32844         PR target/79038
32845         PR target/79202
32846         PR target/79203
32847         * config/rs6000/rs6000.md (u code attribute): Add FIX and
32848         UNSIGNED_FIX.
32849         (extendsi<mode>2): Add support for doing sign extension via
32850         VUPKHSW and XXPERMDI if the value is in Altivec registers and we
32851         don't have ISA 3.0 instructions.
32852         (extendsi<mode>2 splitter): Likewise.
32853         (fix_trunc<mode>si2): If we are at ISA 2.07 (VSX small integer),
32854         generate the normal insns since SImode can now go in vector
32855         registers.  Disallow the special UNSPECs needed for previous
32856         machines to hide SImode being used.  Add new insns
32857         fctiw{,w}_<mode>_smallint if SImode can go in vector registers.
32858         (fix_trunc<mode>si2_stfiwx): Likewise.
32859         (fix_trunc<mode>si2_internal): Likewise.
32860         (fixuns_trunc<mode>si2): Likewise.
32861         (fixuns_trunc<mode>si2_stfiwx): Likewise.
32862         (fctiw<u>z_<mode>_smallint): Likewise.
32863         (fctiw<u>z_<mode>_mem): New combiner pattern to prevent conversion
32864         of floating point to 32-bit integer from doing a direct move to
32865         the GPR registers to do a store.
32866         (fctiwz_<mode>): Break long line.
32868 2017-05-05  Bin Cheng  <bin.cheng@arm.com>
32870         * Makefile.in (GTFILES): Add tree-ssa-loop-ivopts.c.
32871         * tree-ssa-loop-ivopts.c (compute_max_addr_offset): Delete.
32872         (addr_list, addr_offset_valid_p): New.
32873         (split_address_groups): Check offset validity with above function.
32874         (gt-tree-ssa-loop-ivopts.h): Include header file.
32876 2017-05-05  Nathan Sidwell  <nathan@acm.org>
32878         * config.gcc (arm*-*-*): Add missing 'fi'.
32880 2017-05-05  Steve Ellcey  <sellcey@cavium.com>
32882         * doc/invoke.texi (-fopt-info): Explicitly say order of options
32883         included in -fopt-info does not matter.
32884         * doc/optinfo.texi (-fopt-info): Fix description of default
32885         behavour. Explicitly say order of options included in -fopt-info
32886         does not matter.
32888 2017-05-05  Thomas Preud'homme  <thomas.preudhomme@arm.com>
32890         * config.gcc: Allow combinations of aprofile and rmprofile values for
32891         --with-multilib-list.
32892         * config/arm/t-multilib: New file.
32893         * config/arm/t-aprofile: Remove initialization of MULTILIB_*
32894         variables.  Remove setting of ISA and floating-point ABI in
32895         MULTILIB_OPTIONS and MULTILIB_DIRNAMES.  Set architecture and FPU in
32896         MULTI_ARCH_OPTS_A and MULTI_ARCH_DIRS_A rather than MULTILIB_OPTIONS
32897         and MULTILIB_DIRNAMES respectively.  Add comment to introduce all
32898         matches.  Add architecture matches for marvel-pj4 and generic-armv7-a
32899         CPU options.
32900         * config/arm/t-rmprofile: Likewise except for the matches changes.
32901         * doc/install.texi (--with-multilib-list): Document the combination of
32902         aprofile and rmprofile values and warn about pitfalls in doing that.
32904 2017-05-05  Wilco Dijkstra  <wdijkstr@arm.com>
32906         * config/aarch64/aarch64.md (movsi_aarch64): Remove '*' from r=w.
32907         (movdi_aarch64): Likewise.
32909 2017-05-05  Jakub Jelinek  <jakub@redhat.com>
32911         PR tree-optimization/80632
32912         * tree-switch-conversion.c (struct switch_conv_info): Add target_vop
32913         field.
32914         (build_arrays): Initialize it for virtual phis.
32915         (fix_phi_nodes): Use it for virtual phis.
32917         PR tree-optimization/80558
32918         * tree-vrp.c (extract_range_from_binary_expr_1): Optimize
32919         [x, y] op z into [x op, y op z] for op & or | if conditions
32920         are met.
32922 2017-05-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
32923             Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
32925         PR target/71607
32926         * config/arm/arm.md (use_literal_pool): Remove.
32927         (64-bit immediate split): No longer takes cost into consideration
32928         if arm_disable_literal_pool is enabled.
32929         * config/arm/arm.c (arm_tls_referenced_p): Add diagnostic if TLS is
32930         used when arm_disable_literal_pool is enabled.
32931         (arm_max_const_double_inline_cost): Remove use of
32932         arm_disable_literal_pool.
32933         (push_minipool_fix): Add assert.
32934         (arm_reorg): Add return if arm_disable_literal_pool is enabled.
32935         * config/arm/vfp.md (no_literal_pool_df_immediate): New.
32936         (no_literal_pool_sf_immediate): New.
32938 2017-05-05  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
32940         PR tree-optimization/80613
32941         * tree-ssa-dce.c (propagate_necessity): Remove cases for
32942         BUILT_IN_STRDUP and BUILT_IN_STRNDUP.
32944 2017-05-05  Richard Biener  <rguenther@suse.de>
32946         * tree-ssa-pre.c (get_or_alloc_expr_for): Simplify.
32948 2017-05-05  Georg-Johann Lay  <avr@gjlay.de>
32950         * config/avr/avr.md [flag_strict_overflow]: Remove any occurence
32951         of this flag from insn conditions due to removal from r247495.
32953 2017-05-05  Wilco Dijkstra  <wdijkstr@arm.com>
32955         * config/arm/aarch-common.c (arm_early_load_addr_dep_ptr):
32956         New function.
32957         (arm_early_store_addr_dep_ptr): Likewise.
32958         * config/arm/aarch-common-protos.h
32959         (arm_early_load_addr_dep_ptr): Add prototype.
32960         (arm_early_store_addr_dep_ptr): Likewise.
32961         * config/arm/cortex-a53.md: Add new bypasses.
32963 2017-05-05  Jakub Jelinek  <jakub@redhat.com>
32965         * tree.c (next_type_uid): Change type to unsigned.
32966         (type_hash_canon): Decrement back next_type_uid if
32967         freeing a type node with the highest TYPE_UID.  For INTEGER_TYPEs
32968         also ggc_free TYPE_MIN_VALUE, TYPE_MAX_VALUE and TYPE_CACHED_VALUES
32969         if possible.
32971 2017-05-04  Martin Sebor  <msebor@redhat.com>
32973         * builtins.c: Fix a trivial typo in a comment.
32975         PR middle-end/79234
32976         * builtins.c (check_sizes): Adjust to handle reading past the end.
32977         Avoid printing excessive upper bound of ranges.  Use %E to print
32978         tree nodes instead of converting them to %wu.
32979         (expand_builtin_memchr): New function.
32980         (compute_dest_size): Rename...
32981         (compute_objsize): ...to this.
32982         (expand_builtin_memcpy): Adjust.
32983         (expand_builtin_mempcpy): Adjust.
32984         (expand_builtin_strcat): Adjust.
32985         (expand_builtin_strcpy): Adjust.
32986         (check_strncat_sizes): Adjust.
32987         (expand_builtin_strncat): Adjust.
32988         (expand_builtin_strncpy): Adjust and simplify.
32989         (expand_builtin_memset): Adjust.
32990         (expand_builtin_bzero): Adjust.
32991         (expand_builtin_memcmp): Adjust.
32992         (expand_builtin): Handle memcmp.
32993         (maybe_emit_chk_warning): Check strncat just once.
32995 2017-05-04  Martin Sebor  <msebor@redhat.com>
32997         PR preprocessor/79214
32998         PR middle-end/79222
32999         PR middle-end/79223
33000         * builtins.c (check_sizes): Add inlining context and issue
33001         warnings even when -Wno-system-headers is set.
33002         (check_strncat_sizes): Same.
33003         (expand_builtin_strncat): Same.
33004         (expand_builtin_memmove): New function.
33005         (expand_builtin_stpncpy): Same.
33006         (expand_builtin): Handle memmove and stpncpy.
33008 2017-05-04  Bin Cheng  <bin.cheng@arm.com>
33010         * tree-ssa-loop-ivopts.c (struct cost_pair): Remove field inv_expr
33011         which is not used any more.
33013 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
33015         * config/aarch64/aarch64.c (generic_tunings): Update prefetch model.
33017 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
33019         * config/aarch64/aarch64.c (cortexa35_tunings): Set jump alignment to 4.
33020         (cortexa53_tunings): Likewise.
33021         (cortexa57_tunings): Likewise.
33022         (cortexa72_tunings): Likewise.
33023         (cortexa73_tunings): Likewise.
33025 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
33027         * config/aarch64/aarch64.c (generic_tunings): Set jump alignment to 4.
33028         Set loop alignment to 8.
33030 2017-05-04  Martin Sebor  <msebor@redhat.com>
33032         PR translation/80280
33033         * builtins.c (expand_builtin_object_size): Add missing quoting to
33034         %D and like directives.
33035         * hsa-gen.c (hsa_type_for_scalar_tree_type): Same.
33036         (hsa_type_for_tree_type): Same.
33037         (verify_function_arguments): Same.
33038         * symtab.c (symbol_table::change_decl_assembler_name): Same.
33039         * varasm.c (get_section): Same.
33040         (mark_weak): Same.
33042 2017-05-04  Martin Sebor  <msebor@redhat.com>
33044         PR translation/80280
33045         * config/i386/i386.c (ix86_function_versions): Quote a %D directive.
33047 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
33049         * config/aarch64/aarch64.c (generic_addrcost_table):
33050         Change HI/TI mode setting.
33052 2017-05-04  Martin Jambor  <mjambor@suse.cz>
33054         PR tree-optimization/80622
33055         * tree-sra.c (comes_initialized_p): New function.
33056         (build_accesses_from_assign): Only set write lazily when
33057         comes_initialized_p is false.
33058         (analyze_access_subtree): Use comes_initialized_p.
33059         (propagate_subaccesses_across_link): Assert !comes_initialized_p
33060         instead of testing for PARM_DECL.
33062 2017-05-04  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
33064         * config/aarch64/aarch64.md (prefetch); Adjust predicate and
33065         constraint on operand 0 to allow more general addressing modes.
33066         Adjust output template.
33067         * config/aarch64/aarch64.c (aarch64_address_valid_for_prefetch_p):
33068         New function.
33069         * config/aarch64/aarch64-protos.h
33070         (aarch64_address_valid_for_prefetch_p): Declare prototype.
33071         * config/aarch64/constraints.md (Dp): New address constraint.
33072         * config/aarch64/predicates.md (aarch64_prefetch_operand): New
33073         predicate.
33075 2017-05-04  Jan Hubicka  <hubicka@ucw.cz>
33077         * ipa-cp.c (perform_estimation_of_a_value): Drop base_time parameter;
33078         update use of estimate_ipcp_clone_size_and_time.
33079         (estimate_local_effects): Update use of
33080         estimate_ipcp_clone_size_and_time and perform_estimation_of_a_value.
33081         * ipa-inline.h (estimate_ipcp_clone_size_and_time): Update prototype.
33082         * ipa-inline-analysis.c (estimate_ipcp_clone_size_and_time):
33083         Return nonspecialized time.
33085 2017-05-04  Richard Biener  <rguenther@suse.de>
33087         * tree-ssa-alias.c (get_continuation_for_phi): Improve looking
33088         for the last VUSE which def dominates the PHI.  Directly call
33089         maybe_skip_until.
33090         (get_continuation_for_phi_1): Remove.
33092 2017-05-04  Richard Sandiford  <richard.sandiford@linaro.org>
33094         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): Add commentary
33095         to explain the use of truncating division.  Cap the number of
33096         iterations to the maximum given by nb_iterations_upper_bound,
33097         if defined.
33099 2017-05-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
33101         * configure.ac (--enable-mingw-wildcard): Add new configurable feature.
33102         * configure: Regenerate.
33103         * config.in: Regenerate.
33104         * config/i386/driver-mingw32.c: new file.
33105         * config/i386/x-mingw32: Add rule to build driver-mingw32.o.
33106         * config.host: Link driver-mingw32.o on MinGW host.
33107         * doc/install.texi: Document new --enable-mingw-wildcard configure
33108         option.
33110 2017-05-04  Marek Polacek  <polacek@redhat.com>
33112         PR tree-optimization/80612
33113         * calls.c (get_size_range): Check for INTEGRAL_TYPE_P.
33115 2017-05-04  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
33116             Andre Simoes Dias Vieira  <andre.simoesdiasvieira@arm.com>
33118         * config/arm/arm.md (movsi): Change TARGET_32BIT to TARGET_HAVE_MOVT.
33119         (movt splitter): Likewise.
33120         * config/arm/arm.c (arm_option_check_internal): Change arm_arch_thumb2
33121         to TARGET_HAVE_MOVT, and merge with -mslow-flash-data check.
33122         (const_ok_for_arm): Change else to else if (TARGET_THUMB2) and add else
33123         block for Thumb-1 with MOVT.
33124         (thumb2_legitimate_address_p): Move code block ...
33125         (can_avoid_literal_pool_for_label_p): ... into this new function.
33126         (thumb1_legitimate_address_p): Add check for TARGET_HAVE_MOVT and
33127         literal pool.
33128         (thumb_legitimate_constant_p): Add conditional on TARGET_HAVE_MOVT
33129         * doc/invoke.texi (-mpure-code): Change "ARMv7-M targets" for
33130         "M-profile targets with the MOVT instruction".
33132 2017-05-04  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
33134         * config/arm/arm-builtins.c (arm_init_builtins): Rename
33135         __builtin_arm_ldfscr to __builtin_arm_get_fpscr, and rename
33136         __builtin_arm_stfscr to __builtin_arm_set_fpscr.
33138 2017-05-04  Martin Liska  <mliska@suse.cz>
33140         * tree-vrp.c (simplify_cond_using_ranges_2): Remove unused
33141         variable cond_code.
33143 2017-05-04  Richard Biener  <rguenther@suse.de>
33145         * tree.c (array_at_struct_end_p): Handle arrays at struct
33146         end with flexarrays more conservatively.  Refactor and treat
33147         arrays of arrays or aggregates more strict.  Fix
33148         VIEW_CONVERT_EXPR handling.  Remove allow_compref argument.
33149         * tree.c (array_at_struct_end_p): Adjust prototype.
33150         * emit-rtl.c (set_mem_attributes_minus_bitpos): Adjust.
33151         * gimple-fold.c (get_range_strlen): Likewise.
33152         * tree-chkp.c (chkp_may_narrow_to_field): Likewise.
33154 2017-05-04  Richard Biener  <rguenther@suse.de>
33156         PR tree-optimization/31130
33157         * tree-vrp.c (needs_overflow_infinity): Remove as always returning
33158         false.
33159         (supports_overflow_infinity): Likewise.
33160         (is_negative_overflow_infinity): Likewise.
33161         (is_positive_overflow_infinity): Likewise.
33162         (is_overflow_infinity): Likewise.
33163         (stmt_overflow_infinity): Likewise.
33164         (overflow_infinity_range_p): Likewise.
33165         (usable_range_p): Remove as always returning true.
33166         (make_overflow_infinity): Remove.
33167         (negative_overflow_infinity): Likewise.
33168         (positive_overflow_infinity): Likewise.
33169         (avoid_overflow_infinity): Likewise.
33170         (set_value_range): Adjust accordingly.
33171         (set_value_range_to_nonnegative): Likewise, remove now unused
33172         overflow_infinity arg.
33173         (vrp_operand_equal_p): Adjust.
33174         (update_value_range): Likewise.
33175         (range_int_cst_singleton_p): Likewise.
33176         (operand_less_p): Likewise.
33177         (compare_values_warnv): Likewise.
33178         (extract_range_for_var_from_comparison_expr): Likewise.
33179         (vrp_int_const_binop): Likewise.
33180         (zero_nonzero_bits_from_vr): Likewise.
33181         (extract_range_from_multiplicative_op_1): Likewise.
33182         (extract_range_from_binary_expr_1): Likewise.
33183         (extract_range_from_unary_expr): Likewise.
33184         (extract_range_from_comparison): Likewise.
33185         (extract_range_basic): Likewise.
33186         (adjust_range_with_scev): Likewise.
33187         (compare_ranges): Likewise.
33188         (compare_range_with_value): Likewise.
33189         (dump_value_range): Likewise.
33190         (test_for_singularity): Likewise, remove strict_overflow_p parameter
33191         never used.
33192         (simplify_cond_using_ranges): Adjust.
33194 2017-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
33196         * brig-builtins.def: Added a builtin for class_f64.
33197         * builtin-types.def: Added a builtin type needed by class_f64.
33199 2017-05-03  Jason Merrill  <jason@redhat.com>
33201         * timevar.def: Add TV_CONSTEXPR.
33203 2017-05-03  David Malcolm  <dmalcolm@redhat.com>
33205         * common.opt (fdiagnostics-parseable-fixits): Fix typo.
33207 2017-05-03  Martin Jambor  <mjambor@suse.cz>
33209         * ipa-prop.c (ipa_update_after_lto_read): Removed.
33210         * ipa-prop.h (ipa_update_after_lto_read): Remove declaration.
33211         * ipa-cp.c (ipcp_propagate_stage): Do not call
33212         ipa_update_after_lto_read.
33213         * ipa-inline.c (ipa_inline): Likewise.
33215 2017-05-03  Martin Jambor  <mjambor@suse.cz>
33217         * ipa-prop.h (ipa_edge_args): Make a class.  Mark with for_user GTY
33218         tag.  Added a default constructor and a destructor.
33219         (ipa_edge_args_sum_t): New class;
33220         (ipa_edge_args_sum): Declare.
33221         (ipa_edge_args_vector): Remove declaration.
33222         (IPA_EDGE_REF): Use ipa_edge_args_sum.
33223         (ipa_free_edge_args_substructures): Remove declaration.
33224         (ipa_check_create_edge_args): Use ipa_edge_args_sum.
33225         (ipa_edge_args_info_available_for_edge_p): Likewise.
33226         * ipa-prop.c (ipa_edge_args_vector): Removed.
33227         (edge_removal_hook_holder): Likewise.
33228         (edge_duplication_hook_holder): Likewise.
33229         (ipa_edge_args_sum): New variable.
33230         (ipa_propagate_indirect_call_infos): Test ipa_edge_args_sum instead of
33231         ipa_edge_args_vector.
33232         (ipa_free_edge_args_substructures): Likewise.
33233         (ipa_free_all_edge_args): Free ipa_edge_args_sum instead of
33234         ipa_edge_args_vector.
33235         (ipa_edge_removal_hook): Turned into method
33236         ipa_edge_args_sum_t::remove.
33237         (ipa_edge_duplication_hook): Turned into method
33238         ipa_edge_args_sum_t::duplicate.
33239         (ipa_register_cgraph_hooks): Create ipa_edge_args_sum instead of
33240         registering edge hooks.
33241         (ipa_unregister_cgraph_hooks): Do not unregister edge hooks.
33242         * ipa-inline-analysis.c (estimate_function_body_sizes): Test
33243         ipa_edge_args_sum instead of ipa_edge_args_vector.
33244         * ipa-profile.c (ipa_profile): Likewise.
33246 2017-05-03  Martin Jambor  <mjambor@suse.cz>
33248         * symbol-summary.h (function_summary): New method exists.
33249         (function_summary::symtab_removal): Deallocate through release.
33250         (call_summary): New class.
33251         (gt_ggc_mx): New overload.
33252         (gt_pch_nx): Likewise.
33253         (gt_pch_nx): Likewise.
33255 2017-05-03  Jeff Law  <law@redhat.com>
33257         PR tree-optimization/78496
33258         * tree-vrp.c (simplify_cond_using_ranges_1): Renamed
33259         from simplify_cond_using_ranges.  Split off code to walk
33260         backwards through casts into ...
33261         (simplify_cond_using_ranges_2): New function.
33262         (simplify_stmt_using_ranges): Call simplify_cond_using_ranges_1.
33263         (execute_vrp): After identifying jump threads, call
33264         simplify_cond_using_ranges_2.
33266 2017-05-03  Jan Hubicka  <hubicka@ucw.cz>
33268         PR bootstrap/80609
33269         * ipa-inline.h (inline_summary): Add ctor.
33270         (create_ggc): Do not use ggc_cleared_alloc.
33272 2017-05-03  Jeff Downs  <heydowns@somuchpressure.net>
33273             Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
33275         * gcc.c (handle_braces): Support escaping in switch matching
33276         text.
33277         * doc/invoke.texi (Spec Files): Document it.
33278         Remove superfluous @code markup in items.
33280 2017-05-03  David Malcolm  <dmalcolm@redhat.com>
33282         * diagnostic-show-locus.c (struct column_range): New struct.
33283         (get_affected_columns): New function.
33284         (get_printed_columns): New function.
33285         (struct correction): New struct.
33286         (correction::ensure_capacity): New function.
33287         (correction::ensure_terminated): New function.
33288         (struct line_corrections): New struct.
33289         (line_corrections::~line_corrections): New dtor.
33290         (line_corrections::add_hint): New function.
33291         (layout::print_trailing_fixits): Reimplement in terms of the new
33292         classes.
33293         (selftest::test_overlapped_fixit_printing): New function.
33294         (selftest::diagnostic_show_locus_c_tests): Call it.
33296 2017-05-03  Nathan Sidwell  <nathan@acm.org>
33298         Canonicalize canonical type hashing
33299         * tree.h (type_hash_canon_hash): Declare.
33300         * tree.c (type_hash_list, attribute_hash_list): Move into
33301         type_hash_canon_hash.
33302         (build_type_attribute_qual_variant): Break out hash code calc into
33303         type_hash_canon_hash.
33304         (type_hash_canon_hash): New.  Generic type hash computation.
33305         (build_range_type_1, build_array_type_1, build_function_type,
33306         build_method_type_directly, build_offset_type, build_complex_type,
33307         make_vector_type): Call it.
33309 2017-05-03  Richard Biener  <rguenther@suse.de>
33311         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
33312         When all DRs have unknown misaligned do not always peel
33313         when there is a store but apply the same costing model as if
33314         there were only loads.
33316 2017-05-03  Richard Biener  <rguenther@suse.de>
33318         Revert
33319         PR tree-optimization/80492
33320         * tree-ssa-alias.c (decl_refs_may_alias_p): Handle
33321         compare_base_decls returning dont-know properly.
33323 2017-05-03  Thomas Preud'homme  <thomas.preudhomme@arm.com>
33325         * config/arm/iterators.md (CCSI): New mode iterator.
33326         (arch): New mode attribute.
33327         * config/arm/sync.md (atomic_compare_and_swap<mode>_1): Rename into ...
33328         (atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1): This and ...
33329         (atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): This.  Use CCSI
33330         code iterator for success result mode.
33331         * config/arm/arm.c (arm_expand_compare_and_swap): Adapt code to use
33332         the corresponding new insn generators.
33334 2017-05-03  Bin Cheng  <bin.cheng@arm.com>
33336         Revert r247509
33337         2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33338         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
33340 2017-05-03  Richard Sandiford  <richard.sandiford@linaro.org>
33342         * tree-data-ref.h (SUB_CONFLICTS_IN_A): Wrap SUB argument in brackets.
33343         (SUB_CONFLICTS_IN_B, SUB_LAST_CONFLICT, SUB_DISTANCE): Likewise.
33344         (DDR_A): Wrap DDR argument in brackets.
33345         (DDR_B, DDR_AFFINE_P, DDR_ARE_DEPENDENT, DDR_SUBSCRIPTS): Likewise.
33346         (DDR_LOOP_NEST, DDR_INNER_LOOP, DDR_SELF_REFERENCE): Likewise.
33347         (DDR_REVERSED_P): Likewise.
33349 2017-05-03  Jakub Jelinek  <jakub@redhat.com>
33351         PR tree-optimization/79472
33352         * tree-switch-conversion.c (struct switch_conv_info): Add
33353         contiguous_range and default_case_nonstandard fields.
33354         (collect_switch_conv_info): Compute contiguous_range and
33355         default_case_nonstandard fields, don't clear final_bb if
33356         contiguous_range and only the default case doesn't have the required
33357         structure.
33358         (check_all_empty_except_final): Set default_case_nonstandard instead
33359         of failing if contiguous_range and the default case doesn't have empty
33360         block.
33361         (check_final_bb): Add SWTCH argument, don't fail if contiguous_range
33362         and only the default case doesn't have the required constants.  Skip
33363         virtual phis.
33364         (gather_default_values): Skip virtual phis.  Allow non-NULL CASE_LOW
33365         if default_case_nonstandard.
33366         (build_constructors): Build constant 1 just once.  Assert that default
33367         values aren't inserted in between cases if contiguous_range.  Skip
33368         virtual phis.
33369         (build_arrays): Skip virtual phis.
33370         (prune_bbs): Add DEFAULT_BB argument, don't remove that bb.
33371         (fix_phi_nodes): Don't add e2f phi arg if default_case_nonstandard.
33372         Handle virtual phis.
33373         (gen_inbound_check): Handle default_case_nonstandard case.
33374         (process_switch): Adjust check_final_bb caller.  Call
33375         gather_default_values with the first non-default case instead of
33376         default case if default_case_nonstandard.
33378 2017-05-02  Nathan Sidwell  <nathan@acm.org>
33380         * ggc-page.c (move_ptes_to_front): Replace unsigned >0 with i--
33381         check.  Fix formatting.
33383 2017-05-02  Jan Hubicka  <hubicka@ucw.cz>
33385         * ipa-inline-analysis.c (estimate_node_size_and_time): Allow roundoff
33386         errors when comparing specialized and unspecialized times.
33388 2017-05-02  David Malcolm  <dmalcolm@redhat.com>
33390         * diagnostic-show-locus.c
33391         (layout::should_print_annotation_line_p): Make private.
33392         (layout::print_annotation_line): Make private.
33393         (layout::annotation_line_showed_range_p): Make private.
33394         (layout::show_ruler): Make private.
33395         (layout::print_source_line): Make private.  Pass in line and
33396         line_width, rather than calling location_get_source_line.  Drop
33397         returned value.
33398         (layout::print_leading_fixits): New method.
33399         (layout::print_any_fixits): Rename to...
33400         (layout::print_trailing_fixits): ...this, and make private.
33401         Don't print newline fixits.
33402         (diagnostic_show_locus): Move logic for printing one row into...
33403         (layout::print_line): ...this new function.  Move the
33404         location_get_source_line call and error-handling from
33405         print_source_line to here.  Call print_leading_fixits, and rename
33406         print_any_fixits to print_trailing_fixits.
33407         (selftest::test_fixit_insert_containing_newline): Update now that
33408         newlines are partially supported.
33409         (selftest::test_fixit_insert_containing_newline_2): New test.
33410         (selftest::test_fixit_replace_containing_newline): Update comments.
33411         (selftest::diagnostic_show_locus_c_tests): Call the new test.
33412         * edit-context.c (class added_line): New class.
33413         (class edited_line): Describe newline handling in comment.
33414         (edited_line::actually_edited_p): New method.
33415         (edited_line::print_content): Delete redundant decl.
33416         (edited_line::m_predecessors): New field.
33417         (edited_file::print_content): Call edited_line::print_content.
33418         (edited_file::print_diff): Update to support newlines.
33419         (edited_file::print_diff_hunk): Likewise.
33420         (edited_file::print_run_of_changed_lines): New function.
33421         (edited_file::print_diff_line): Convert to...
33422         (print_diff_line): ...this.
33423         (edited_file::get_effective_line_count): New function.
33424         (edited_line::edited_line): Initialize new field m_predecessors.
33425         (edited_line::~edited_line): Clean up m_predecessors.
33426         (edited_line::apply_fixit): Handle newlines.
33427         (edited_line::get_effective_line_count): New function.
33428         (edited_line::print_content): New function.
33429         (edited_line::print_diff_lines): New function.
33430         (selftest::test_applying_fixits_insert_containing_newline): New
33431         test.
33432         (selftest::test_applying_fixits_replace_containing_newline): New
33433         test.
33434         (selftest::insert_line): New function.
33435         (selftest::test_applying_fixits_multiple_lines): Add example of
33436         inserting a line.
33437         (selftest::edit_context_c_tests): Call the new tests.
33439 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33441         * tree-ssa-loop-ivopts.c (get_scaled_computation_cost_at): Delete
33442         parameter cand.  Update dump information.
33443         (get_computation_cost): Update uses.
33445 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33447         * tree-ssa-loop-ivopts.c (get_computation_aff_1): New.
33448         (get_computation_aff): Reorder parameters.  Use get_computation_aff_1.
33449         (get_computation_at, rewrite_use_address): Update use of
33450         get_computation_aff.
33452 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33454         * tree-ssa-loop-ivopts.c (get_computation_at): Reorder parameters.
33455         (get_computation): Delete.
33456         (get_computation_cost): Implement like get_computation_cost_at.
33457         Use get_computation_at.
33458         (get_computation_cost_at): Delete.
33459         (rewrite_use_nonlinear_expr): Use get_computation_at.
33460         (rewrite_use_compare, remove_unused_ivs): Ditto.
33462 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33464         * tree-ssa-loop-ivopts.c (rewrite_use_address): Simple refactor.
33466 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33468         * tree-ssa-loop-ivopts.c (struct iv_ca): Rename n_regs to n_invs.
33469         (ivopts_global_cost_for_size): Rename parameter and update uses.
33470         (iv_ca_recount_cost): Update uses.
33471         (iv_ca_set_remove_invs, iv_ca_set_no_cp): Record invariants and
33472         candidates seperately in n_invs and n_cands.
33473         (iv_ca_set_add_invs, iv_ca_set_cp, iv_ca_new): Ditto.
33475 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33477         * tree-ssa-loop-ivopts.c (struct walk_tree_data): New.
33478         (find_inv_vars_cb): New.
33479         (find_depends): Renamed to ...
33480         (find_inv_vars): ... this.
33481         (add_candidate_1, force_var_cost): Call find_inv_vars.
33482         (split_address_cost, determine_group_iv_cost_cond): Ditto.
33484 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33486         * tree-ssa-loop-ivopts.c (struct cost_pair): Rename depends_on to
33487         inv_vars.  Add inv_exprs.
33488         (struct iv_cand): Rename depends_on to inv_vars.
33489         (struct ivopts_data): Rename max_inv_id/n_invariant_uses to
33490         max_inv_var_id/n_inv_var_uses.  Move max_inv_expr_id around.
33491         Refactor field used_inv_exprs from has_map to array n_inv_expr_uses.
33492         (dump_cand): Dump inv_vars.
33493         (tree_ssa_iv_optimize_init): Support inv_vars and inv_exprs.
33494         (record_invariant, find_depends, add_candidate_1): Ditto.
33495         (set_group_iv_cost, force_var_cost): Ditto.
33496         (split_address_cost, ptr_difference_cost, difference_cost): Ditto.
33497         (get_computation_cost_at, get_computation_cost): Ditto.
33498         (determine_group_iv_cost_generic): Ditto.
33499         (determine_group_iv_cost_address): Ditto.
33500         (determine_group_iv_cost_cond, autoinc_possible_for_pair): Ditto.
33501         (determine_group_iv_costs): Ditto.
33502         (iv_ca_recount_cost): Update call to ivopts_global_cost_for_size.
33503         (iv_ca_set_remove_invariants): Renamed to ...
33504         (iv_ca_set_remove_invs): ... this.  Support inv_vars and inv_exprs.
33505         (iv_ca_set_no_cp): Use iv_ca_set_remove_invs.
33506         (iv_ca_set_add_invariants):  Renamed to ...
33507         (iv_ca_set_add_invs): ... this.  Support inv_vars and inv_exprs.
33508         (iv_ca_set_cp): Use iv_ca_set_add_invs.
33509         (iv_ca_has_deps): Support inv_vars and inv_exprs.
33510         (iv_ca_new, iv_ca_free, iv_ca_dump, free_loop_data): Ditto.
33511         (create_new_ivs): Remove useless dump.
33513 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33515         * tree-ssa-loop-ivopts.c (get_computation_cost_at): Remove pseudo
33516         iv_cand code.
33517         (determine_group_iv_cost_cond, determine_iv_cost): Ditto.
33518         (iv_ca_set_no_cp, create_new_iv): Ditto.
33520 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33522         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
33524 2017-05-02  Xi Ruoyao  <ryxi@stu.xidian.edu.cn>
33526         * tree.h (EXPR_CILK_SPAWN): Use macro TREE_CHECK2 instead of
33527         function tree_check2.
33529 2017-05-02  Martin Liska  <mliska@suse.cz>
33531         * doc/gcov.texi: Add missing preposition.
33532         * gcov.c (function_info::function_info): Properly fill up
33533         all member variables.
33535 2017-05-02  Tamar Christina  <tamar.christina@arm.com>
33537         * expr.c (expand_expr_real_2): Re-cost if previous costs are the same.
33539 2017-05-02  Tamar Christina  <tamar.christina@arm.com>
33541         * simplify-rtx.c (simplify_binary_operation_1): Add LSHIFTRT case.
33543 2017-05-02  Martin Liska  <mliska@suse.cz>
33545         PR lto/77954.
33546         * lto-streamer-in.c (lto_read_tree_1): Remove
33547         LTO_STREAMER_DEBUG.
33548         * lto-streamer.c (struct tree_hash_entry): Likewise.
33549         (struct tree_entry_hasher): Likewise.
33550         (tree_entry_hasher::hash): Likewise.
33551         (tree_entry_hasher::equal): Likewise.
33552         (lto_streamer_init): Likewise.
33553         (lto_orig_address_map): Likewise.
33554         (lto_orig_address_get): Likewise.
33555         (lto_orig_address_remove): Likewise.
33556         * lto-streamer.h: Likewise.
33557         * tree-streamer-in.c (streamer_alloc_tree): Likewise.
33558         * tree-streamer-out.c (streamer_write_tree_header): Likewise.
33560 2017-05-02  Sebastian Peryt  <sebastian.peryt@intel.com>
33562         * config/i386/avx512fintrin.h (_mm_mask_add_round_sd)
33563         (_mm_maskz_add_round_sd, _mm_mask_add_round_ss)
33564         (mm_maskz_add_round_ss, _mm_mask_sub_round_sd)
33565         (mm_maskz_sub_round_sd, _mm_mask_sub_round_ss)
33566         (mm_maskz_sub_round_ss, _mm_mask_add_sd)
33567         (mm_maskz_add_sd, _mm_mask_add_ss, _mm_maskz_add_ss)
33568         (mm_mask_sub_sd, _mm_maskz_sub_sd, _mm_mask_sub_ss)
33569         (mm_maskz_sub_ss): New intrinsics.
33570         * config/i386/i386-builtin-types.def (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
33571         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): New function type aliases.
33572         * config/i386/i386-builtin.def (__builtin_ia32_addsd_mask_round)
33573         (__builtin_ia32_addss_mask_round, __builtin_ia32_subsd_mask_round)
33574         (__builtin_ia32_subss_mask_round): New builtins.
33575         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
33576         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
33577         * config/i386/sse.md (<sse>_vm<plusminus_insn><mode>3<round_name>):
33578         Renamed to ...
33579         (<sse>_vm<plusminus_insn><mode>3<mask_name><round_name>): ... this.
33580         (v<plusminus_mnemonic><ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|%0, %1, %<iptr>2<round_op3>}):
33581         Changed to ...
33582         (v<plusminus_mnemonic><ssescalarmodesuffix>\t{<round_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}):
33583         ... this.
33585 2017-05-02  Martin Jambor  <mjambor@suse.cz>
33587         PR tree-optimization/78687
33588         * tree-sra.c (access): New field parent.
33589         (process_subtree_disqualification): New function.
33590         (disqualify_candidate): Call it.
33591         (build_accesses_from_assign): Reset write flag if creating an
33592         assighnment link.
33593         (build_access_subtree): Fill in parent field and also prpagate
33594         down grp_write flag.
33595         (create_artificial_child_access): New parameter set_grp_write, set
33596         grp_write to its value.
33597         (propagate_subaccesses_across_link): Also propagate grp_write flag
33598         values.
33599         (propagate_all_subaccesses): Push the closest parent back to work
33600         queue if add_access_to_work_queue returned true.
33602 2017-05-02  Richard Biener  <rguenther@suse.de>
33604         * common.opt (fstrict-overflow): Alias negative to fwrapv.
33605         * doc/invoke.texi (fstrict-overflow): Remove all traces of
33606         -fstrict-overflow documentation.
33607         * tree.h (TYPE_OVERFLOW_UNDEFINED): Do not test flag_strict_overflow.
33608         (POINTER_TYPE_OVERFLOW_UNDEFINED): Test !flag_wrapv instead of
33609         flag_strict_overflow.
33610         * ipa-inline.c (can_inline_edge_p): Do not test flag_strict_overflow.
33611         * lto-opts.c (lto_write_options): Do not stream it.
33612         * lto-wrapper.c (merge_and_complain): Do not handle it.
33613         * opts.c (default_options_table): Do not set -fstrict-overflow.
33614         (finish_options): Likewise do not clear it when sanitizing.
33615         * simplify-rtx.c (simplify_const_relational_operation): Do not
33616         test flag_strict_overflow.
33618 2017-05-02  Uros Bizjak  <ubizjak@gmail.com>
33620         * config/alpha/alpha.md (*add<mode>3_ieee): Merge to add<mode>3
33621         using enabled attribute.
33622         (*sub<mode>3_ieee): Merge to sub<mode>3 using enabled attribute.
33623         (*mul<mode>3_ieee): Merge to mul<mode>3 using enabled attribute.
33624         (*div<mode>3_ieee): Merge to div<mode>3 using enabled attribute.
33625         (*sqrt<mode>2_ieee): Merge to sqrt<mode>2 using enabled attribute.
33626         (*fix_truncdfdi_ieee): Merge to *fix_truncdfdi2 using enabled attribute.
33627         (*fix_truncsfdi_ieee): Merge to *fix_truncsfdi2 using enabled attribute.
33628         (*floatdisf_ieee): Merge to floatdisf2 using enabled attribute.
33629         (*floatdidf_ieee): Merge to floatdidf2 using enabled attribute.
33630         (*truncdfsf2_ieee): Merge to truncdfsf2 using enabled attribute.
33631         (*cmpdf_ieee): Merge to *cmpdf_internal using enabled attribute.
33633 2017-05-02  Uros Bizjak  <ubizjak@gmail.com>
33635         * config/i386/i386.c (ix86_code_end): Use {FIRST,LAST}_INT_REG.
33637 2017-05-02  Richard Biener  <rguenther@suse.de>
33639         PR tree-optimization/80591
33640         Revert
33641         2017-04-10  Richard Biener  <rguenther@suse.de>
33643         * tree-ssa-structalias.c (find_func_aliases): Properly handle
33644         asm inputs.
33646 2017-05-02  Richard Biener  <rguenther@suse.de>
33648         PR tree-optimization/80549
33649         * tree-cfgcleanup.c (mfb_keep_latches): New helper.
33650         (cleanup_tree_cfg_noloop): Create forwarders to known loop
33651         headers if they do not have a preheader.
33653 2017-05-02  Martin Liska  <mliska@suse.cz>
33655         PR other/80589
33656         * common.opt: Fix typo.
33657         * doc/invoke.texi: Likewise.
33659 2017-05-01  Jan Beulich  <jbeulich@suse.com>
33661         * config/i386/sse.md (xop_vpermil2<mode>3): Do not allow operand
33662         swapping, add (x,x,m,x,n) alternative.
33664 2017-05-01  Nathan Sidwell  <nathan@acm.org>
33666         * calls.c (combine_pending_stack_adjustment_and_call): Remove
33667         unnecessary unadjusted_alignment check.
33669 2017-05-01  Xi Ruoyao  <ryxi@stu.xidian.edu.cn>
33671         PR c++/80038
33672         * cilk_common.c (expand_builtin_cilk_detach): Move pedigree
33673         operations here.
33674         * gimplify.c (gimplify_cilk_detach): New function.
33675         (gimplify_call_expr, gimplify_modify_expr): Call it as needed.
33676         * tree-core.h: Document EXPR_CILK_SPAWN.
33677         * tree.h (EXPR_CILK_SPAWN): Define.
33679 2017-05-01  David Malcolm  <dmalcolm@redhat.com>
33681         * diagnostic-show-locus.c (layout::get_expanded_location): Rewrite
33682         to use new fixit_hint representation, using the "replace" logic.
33683         (get_line_span_for_fixit_hint): Likewise.
33684         (layout::print_any_fixits): Likewise.
33685         (selftest::test_one_liner_many_fixits): Rename to...
33686         (selftest::test_one_liner_many_fixits_1): ...this, and update
33687         comment and expected output to reflect that the multiple fix-it
33688         hints are now consolidated into one insertion.
33689         (selftest::test_one_liner_many_fixits_2): New test.
33690         (selftest::test_diagnostic_show_locus_one_liner): Update for
33691         above.
33692         (selftest::test_fixit_consolidation): Update for fix-it API
33693         change.
33694         * diagnostic.c (print_parseable_fixits): Likewise.
33695         * edit-context.c (edited_line::m_line_events): Convert from
33696         auto_vec <line_event *> to auto_vec <line_event>.
33697         (class line_event): Convert from abstract base class to a concrete
33698         class, taking over the role of replace_event.
33699         (class insert_event): Delete.
33700         (class replace_event): Rename to class line_event.  Convert to
33701         half-open range.
33702         (edit_context::add_fixits): Reimplement.
33703         (edit_context::apply_insert): Delete.
33704         (edit_context::apply_replace): Rename to...
33705         (edit_context::apply_fixit): ...this.  Convert to half-open range.
33706         (edited_file::apply_insert): Delete.
33707         (edited_file::apply_replace): Rename to...
33708         (edited_file::apply_fixit): ...this.
33709         (edited_line::~edited_line): Drop deletion of events.
33710         (edited_line::apply_insert): Delete.
33711         (edited_line::apply_replace): Rename to...
33712         (edited_line::apply_fixit): ...this.  Convert to half-open range.
33713         Update for change to type of m_line_events.
33714         * edit-context.h (edit_context::apply_insert): Delete.
33715         (edit_context::apply_replace): Rename to...
33716         (edit_context::apply_fixit): ...this.
33718 2017-05-01  Martin Sebor  <msebor@redhat.com>
33720         * gimple-ssa-sprintf.c (format_integer): Set knownrange when it's
33721         known.
33723 2017-05-01  Uros Bizjak  <ubizjak@gmail.com>
33725         PR target/68491
33726         * config/i386/cpuid.h (__get_cpuid): Always return 0 when
33727         __get_cpuid_max returns 0.
33728         (__get_cpuid_count): Ditto.
33730 2017-05-01  Eric Botcazou  <ebotcazou@adacore.com>
33732         * tree.c (substitute_in_expr) <tcc_vl_exp>: Also inline a call if the
33733         replacement expression is another instance of one of its arguments.
33735 2017-05-01  Jakub Jelinek  <jakub@redhat.com>
33737         PR target/79430
33738         * rtlanal.c (reg_set_p): If reg is a stack_pointer_rtx, also
33739         check for stack push/pop autoinc.
33740         * config/i386/i386.c (ix86_agi_dependent): Return false
33741         if the only reason why modified_in_p returned true is that
33742         addr is SP based and set_insn is a push or pop.
33744 2017-04-29  Jan Hubicka  <hubicka@ucw.cz>
33746         * ipa-inline.c (compute_inlined_call_time): Remove now unnecesary
33747         overflow check.
33749 2017-04-29  Jan Hubicka  <hubicka@ucw.cz>
33751         PR ipa/79224
33752         * ipa-inline-analysis.c (dump_predicate): Add optional parameter NL.
33753         (account_size_time): Use two predicates - exec_pred and
33754         nonconst_pred_ptr.
33755         (evaluate_conditions_for_known_args): Compute both clause and
33756         nonspec_clause.
33757         (evaluate_properties_for_edge): Evaulate both clause and nonspec_clause.
33758         (inline_summary_t::duplicate): Update.
33759         (estimate_function_body_sizes): Caluculate exec and nonconst predicates
33760         separately.
33761         (compute_inline_parameters): Likewise.
33762         (estimate_edge_size_and_time): Update caluclation of time.
33763         (estimate_node_size_and_time): Compute both time and nonspecialized
33764         time.
33765         (estimate_ipcp_clone_size_and_time): Update.
33766         (inline_merge_summary): Update.
33767         (do_estimate_edge_time): Update.
33768         (do_estimate_edge_size): Update.
33769         (do_estimate_edge_hints): Update.
33770         (inline_read_section, inline_write_summary): Stream both new predicates.
33771         * ipa-inline.c (compute_uninlined_call_time): Take uninlined_call_time
33772         as argument.
33773         (compute_inlined_call_time): Cleanup.
33774         (big_speedup_p): Update.
33775         (edge_badness): Update.
33776         * ipa-inline.h (INLINE_TIME_SCALE): Remove.
33777         (size_time_entry): Replace predicate by exec_predicate and
33778         nonconst_predicate.
33779         (edge_growth_cache_entry): Cache both time nad nonspecialized time.
33780         (estimate_edge_time): Return also nonspec_time.
33781         (reset_edge_growth_cache): Update.
33783 2017-04-29  Jakub Jelinek  <jakub@redhat.com>
33785         PR rtl-optimization/80491
33786         * ifcvt.c (noce_process_if_block): When looking for x setter
33787         with missing else_bb, don't check only the insn right before
33788         cond_earliest, but look for the last insn that x is modified in
33789         within the same bb.
33791         PR rtl-optimization/80491
33792         * alias.c (memory_modified_in_insn_p): Return true for CALL_INSNs.
33794 2017-04-29  Marc Glisse  <marc.glisse@inria.fr>
33796         PR tree-optimization/80487
33797         * tree-ssa-alias.c (stmt_kills_ref_p): Handle stpncpy and strncpy.
33799 2017-04-29  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
33801         PR tree-optimization/79697
33802         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Check if callee
33803         is BUILT_IN_STRDUP, BUILT_IN_STRNDUP, BUILT_IN_REALLOC.
33804         (propagate_necessity): Check if def_callee is BUILT_IN_STRDUP or
33805         BUILT_IN_STRNDUP.
33806         * gimple-fold.c (gimple_fold_builtin_realloc): New function.
33807         (gimple_fold_builtin): Call gimple_fold_builtin_realloc.
33809 2017-04-28  Martin Sebor  <msebor@redhat.com>
33811         PR tree-optimization/80523
33812         * gimple-ssa-sprintf.c (target_to_host_charmap): New global variable.
33813         (init_target_to_host_charmap, target_to_host, target_strtol10): New
33814         functions.
33815         (maybe_warn, format_directive, parse_directive): Use new functions.
33816         (pass_sprintf_length::execute): Call init_target_to_host_charmap.
33818 2017-04-28  Marc Glisse  <marc.glisse@inria.fr>
33820         * match.pd (X+Z OP Y+Z, X-Z OP Y-Z, Z-X OP Z-Y): New transformations.
33822 2017-04-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
33824         * configure.ac (SYSTEM_HEADER_DIR, BUILD_SYSTEM_HEADER_DIR,
33825         target_header_dir): Set correctly.
33826         * configure: Regenerated.
33827         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
33828         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
33829         instead of SYSTEM_HEADER_DIR.
33831 2017-04-28  Jan Hubicka  <hubicka@ucw.cz>
33833         * ipa-cp.c (perform_estimation_of_a_value): Turn time to sreal.
33834         (estimate_local_effects): Likewise.
33835         * ipa-inline.c (compute_inlined_call_time, want_inline_small_function_p,
33836         edge_badness, inline_small_functions, dump_overall_stats): LIkewise.
33837         * ipa-inline.h (edge_growth_cache_entry, estimate_time_after_inlining,
33838         estimate_ipcp_clone_size_and_time, do_estimate_edge_time,
33839         do_estimate_edge_time, estimate_edge_time): Likewise.
33840         * ipa-inline-analysis.c (estimate_node_size_and_time,
33841         estimate_ipcp_clone_size_and_time, do_estimate_edge_time): Likewise.
33842         (estimate_time_after_inlining): Remove.
33844 2017-04-28  Martin Liska  <mliska@suse.cz>
33846         * doc/gcov.texi: Enhance documentation of gcov.
33848 2017-04-28  Martin Liska  <mliska@suse.cz>
33850         * doc/gcov.texi: Sort options in alphabetic order.
33851         * doc/gcov-dump.texi: Likewise.
33852         * doc/gcov-tool.texi: Likewise.
33853         * gcov.c (print_usage): Likewise.
33854         * gcov-dump.c (print_usage): Likewise.
33855         * gcov-tool.c (print_merge_usage_message): Likewise.
33856         (print_rewrite_usage_message): Likewise.
33857         (print_overlap_usage_message): Likewise.
33859 2017-04-28  Martin Liska  <mliska@suse.cz>
33861         PR gcov-profile/53915
33862         * gcov.c (format_gcov): Print 'NAN %' when top > bottom.
33864 2017-04-28  Martin Liska  <mliska@suse.cz>
33866         PR gcov-profile/79891
33867         * gcov.c (add_line_counts): Assign BBs to lines just if the BB
33868         is marked by compiler as living on a line.
33869         (get_cycles_count): Remove usage of the union.
33870         (output_intermediate_file): Likewise.
33871         (find_source): Fix GNU coding style.
33872         (accumulate_line_counts): Remove old non-all block mode.
33873         (output_lines): Remove usage of the union.
33874         * profile.c (output_location): Include all BBs, even if
33875         belonging to a same line (and file) as a previous BB.
33877 2017-04-28  Martin Liska  <mliska@suse.cz>
33879         * gcov.c (process_args): Handle new argument 'w'.
33880         (read_graph_file): Assign ID to BBs.
33881         (output_branch_count): Display BB # if verbose flag is set.
33882         (output_lines): Likewise for arcs.
33883         (print_usage): Add '--verbose' option help.
33884         * doc/gcov.texi: Document --verbose (-w) option.
33886 2017-04-28  Martin Liska  <mliska@suse.cz>
33888         * gcov.c (struct block_location_info): New struct.
33889         (process_file): Fill up the new structure.
33890         (read_graph_file): Replace usage of encoding by the newly added
33891         struct.
33892         (add_line_counts): Likewise.
33893         (accumulate_line_counts): Remove usage of the union.
33894         (function_info::function_info): New function.
33895         (function_info::~function_info): Likewise.
33896         (process_file): Call delete instead of release_function.
33897         (release_function): Release the function.
33898         (release_structures): Call delete instead of release_function.
33899         (solve_flow_graph): Replace usage of num_blocks.
33900         (find_exception_blocks): Likewise.
33901         (output_lines): Fix GNU coding style.
33903 2017-04-28  Martin Liska  <mliska@suse.cz>
33905         PR driver/56469
33906         * coverage.c (coverage_remove_note_file): New function.
33907         * coverage.h: Declare the function.
33908         * toplev.c (finalize): Clean if an error has been seen.
33910 2017-04-28  Martin Liska  <mliska@suse.cz>
33912         PR gcov-profile/80031
33913         * gcov-dump.c (tag_blocks): Just print number of basic blocks.
33914         * gcov-io.h (GCOV_TAG_BLOCKS_NUM): Remove unused macro.
33915         * gcov.c (read_graph_file): Read just number of blocks.
33916         * profile.c (branch_prob): Do not stream 0 flags per a basic
33917         block.
33919 2017-04-28  Martin Liska  <mliska@suse.cz>
33921         * gcov-dump.c (tag_*): Add new argument to declarations.
33922         (dump_gcov_file): Likewise.
33923         (tag_blocks): Add and use new argument depth.
33924         (tag_arcs): Likewise.
33925         (tag_lines): Likewise.
33926         (tag_counters): Likewise.
33927         (tag_summary): Likewise.
33928         (dump_working_sets): Use depth to do a proper indentation.
33930 2017-04-28  Jakub Jelinek  <jakub@redhat.com>
33932         PR bootstrap/80531
33933         * cgraph.h (symtab_node::debug_symtab): No longer inline.
33934         * symtab.c (symtab_node::debug_symtab): Move definition here.
33936 2017-04-28  Richard Biener  <rguenther@suse.de>
33938         * lto-streamer.h (LTO_major_version): Bump to 7.
33940 2017-04-28  Richard Biener  <rguenther@suse.de>
33942         * tree-vrp.c (assert_info): New struct.
33943         (add_assert_info): New helper.
33944         (register_edge_assert_for_2): Refactor to add asserts to a vector
33945         of assert_info.
33946         (register_edge_assert_for_1): Likewise.
33947         (register_edge_assert_for): Likewise.
33948         (finish_register_edge_assert_for): New helper actually registering
33949         asserts where live on edge.
33950         (find_conditional_asserts): Adjust.
33951         (find_switch_asserts): Likewise.
33952         (evrp_dom_walker::try_find_new_range): Generalize.
33953         (evrp_dom_walker::before_dom_children): Use register_edge_assert_for.
33955 2017-04-27  Marek Polacek  <polacek@redhat.com>
33957         PR sanitizer/80349
33958         * fold-const.c (fold_binary_loc) <case EQ_EXPR, NE_EXPR>: Convert
33959         arg10 and arg11 to itype.
33961 2017-04-27  Jonathan Wakely  <jwakely@redhat.com>
33963         * doc/extend.texi (Object Size Checking): Improve grammar.
33965 2017-04-27  Richard Earnshaw  <rearnsha@arm.com>
33967         PR target/80530
33968         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Ensure
33969         that the logic for permitting reciprocal estimates matches that
33970         in use_rsqrt_p.
33972 2017-04-27  Jakub Jelinek  <jakub@redhat.com>
33974         PR c++/80534
33975         * tree.c (type_cache_hasher::equal): Only compare
33976         TYPE_TYPELESS_STORAGE flag on non-aggregate element types.
33977         (build_array_type_1): Only hash TYPE_TYPELESS_STORAGE flag on
33978         non-aggregate element types.
33979         * tree.h (TYPE_TYPELESS_STORAGE): Fix comment typo, add more details
33980         about the flag on ARRAY_TYPEs in the comment, formatting fix.
33982 2017-04-27  Richard Biener  <rguenther@suse.de>
33984         PR middle-end/80533
33985         * emit-rtl.c (set_mem_attributes_minus_bitpos): When
33986         stripping ARRAY_REFs from MEM_EXPR make sure we're not
33987         keeping a reference to a trailing array.
33989 2017-04-27  Richard Biener  <rguenther@suse.de>
33991         PR middle-end/80539
33992         * tree-chrec.c (chrec_fold_plus_poly_poly): Deal with not
33993         being in loop-closed SSA form conservatively.
33994         (chrec_fold_multiply_poly_poly): Likewise.
33996 2017-04-27  Tamar Christina  <tamar.christina@arm.com>
33998         PR middle-end/79665
33999         * expr.c (expand_expr_real_2): Move TRUNC_MOD_EXPR, FLOOR_MOD_EXPR,
34000         CEIL_MOD_EXPR, ROUND_MOD_EXPR cases.
34002 2017-04-27  Jakub Jelinek  <jakub@redhat.com>
34004         PR target/77728
34005         * config/aarch64/aarch64.c (struct aarch64_fn_arg_alignment): Remove.
34006         (aarch64_function_arg_alignment): Return unsigned int again, but still
34007         ignore TYPE_FIELDS chain decls other than FIELD_DECLs.
34008         (aarch64_layout_arg): Adjust aarch64_function_arg_alignment caller.
34009         Don't emit -Wpsabi note.
34010         (aarch64_function_arg_boundary): Likewise.
34011         (aarch64_gimplify_va_arg_expr): Adjust aarch64_function_arg_alignment
34012         caller.
34014 2017-04-26  Nathan Sidwell  <nathan@acm.org>
34016         * tree.h (crc32_unsigned_n): Declare.
34017         (crc32_unsigned, crc32_unsigned): Make inline.
34018         * tree.c (crc32_unsigned_bits): Replace with ...
34019         (crc32_unsigned_n): ... this.
34020         (crc32_unsigned, crc32_byte): Remove.
34021         (crc32_string): Remove unnecessary braces.
34023 2017-04-25  Jan Hubicka  <hubicka@ucw.cz>
34025         * ipa-cp.c (estimate_local_effects): Convert sreal to int.
34026         * ipa-inline-analysis.c (MAX_TIME): Remove.
34027         (account_size_time): Use sreal for time.
34028         (dump_inline_summary): Update.
34029         (estimate_function_body_sizes): Update.
34030         (estimate_edge_size_and_time): Update.
34031         (estimate_calls_size_and_time): Update.
34032         (estimate_node_size_and_time): Update.
34033         (inline_merge_summary): Update.
34034         (inline_update_overall_summary): Update.
34035         (estimate_time_after_inlining): Update.
34036         (inline_read_section): Update.
34037         (inline_write_summary): Update.
34038         * ipa-inline.c (compute_uninlined_call_time): Update.
34039         (compute_inlined_call_time): Update.
34040         (recursive_inlining): Update.
34041         (inline_small_functions): Update.
34042         (dump_overall_stats): Update.
34043         * ipa-inline.h: Include sreal.h.
34044         (size_time_entry): Turn time to sreal.
34045         (inline_summary): Turn self_time nad time to sreal.
34047 2017-04-25  Jan Hubicka  <hubicka@ucw.cz>
34049         * sreal.c: Include backend.h, tree.h, gimple.h, cgraph.h and
34050         data-streamer.h
34051         (sreal::stream_out, sreal::stream_in): New.
34052         * sreal.h (sreal::stream_out, sreal::stream_in): Declare.
34054 2017-04-25  Jakub Jelinek  <jakub@redhat.com>
34056         * Makefile.in (s-options): Invoke opt-gather.awk with LC_ALL=C in the
34057         environment.
34059 2017-04-25  Uros Bizjak  <ubizjak@gmail.com>
34061         PR target/70799
34062         * config/i386/i386.c (dimode_scalar_to_vector_candidate_p):
34063         Handle ASHIFTRT.
34064         (dimode_scalar_chain::compute_convert_gain): Ditto.
34065         (dimode_scalar_chain::make_vector_copies): Ditto.
34066         (dimode_scalar_chain::convert_reg): Ditto.
34067         (dimode_scalar_chain::convert_insn): Ditto.
34068         * config/i386/sse.md (VI24_AVX512BW_1): Remove mode iterator.
34069         (VI248_AVX512BW_1): New mode iterator.
34070         (<mask_codefor>ashr<mode>3<mask_name>): Merge insn pattern with
34071         <mask_codefor>ashrv2di3<mask_name> insn using VI248_AVX512BW_1
34072         mode iterator.
34074 2017-04-25  Martin Sebor  <msebor@redhat.com>
34076         PR tree-optimization/80497
34077         * gimple-ssa-sprintf.c (get_int_range): Avoid assuming all integer
34078         constants are representable in HOST_WIDE_INT.
34079         (parse_directive): Ditto.
34081 2017-04-25  Martin Sebor  <msebor@redhat.com>
34083         PR bootstrap/80486
34084         * dominance.c (dom_info::m_n_basic_blocks): Change type to unsigned.
34085         (new_zero_array): Adjust signature.
34086         (dom_info::dom_init): Used unsigned rather that size_t.
34087         (dom_info::dom_info): Same.
34089 2017-04-25  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
34090             Jakub Jelinek  <jakub@redhat.com>
34092         PR target/77728
34093         * config/arm/arm.c: Include gimple.h.
34094         (aapcs_layout_arg): Emit -Wpsabi note if arm_needs_doubleword_align
34095         returns negative, increment ncrn only if it returned positive.
34096         (arm_needs_doubleword_align): Return int instead of bool,
34097         ignore DECL_ALIGN of non-FIELD_DECL TYPE_FIELDS chain
34098         members, but if there is any such non-FIELD_DECL
34099         > PARM_BOUNDARY aligned decl, return -1 instead of false.
34100         (arm_function_arg): Emit -Wpsabi note if arm_needs_doubleword_align
34101         returns negative, increment nregs only if it returned positive.
34102         (arm_setup_incoming_varargs): Likewise.
34103         (arm_function_arg_boundary): Emit -Wpsabi note if
34104         arm_needs_doubleword_align returns negative, return
34105         DOUBLEWORD_ALIGNMENT only if it returned positive.
34107 2017-04-25  Marek Polacek  <polacek@redhat.com>
34109         PR sanitizer/80349
34110         * fold-const.c (fold_binary_loc) <case BIT_IOR_EXPR>: Convert arg0's
34111         first argument to type.
34113 2017-04-25  Bill Seurer  <seurer@linux.vnet.ibm.com>
34115         PR target/80482
34116         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Change
34117         type checks to test for compatibility instead of equality.
34119 2017-04-25  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
34120             Jakub Jelinek  <jakub@redhat.com>
34122         PR target/77728
34123         * config/aarch64/aarch64.c (struct aarch64_fn_arg_alignment): New
34124         type.
34125         (aarch64_function_arg_alignment): Return aarch64_fn_arg_alignment
34126         struct.  Ignore DECL_ALIGN of decls other than FIELD_DECL for
34127         the alignment computation, but return their maximum in warn_alignment.
34128         (aarch64_layout_arg): Adjust aarch64_function_arg_alignment caller.
34129         Emit a -Wpsabi note if warn_alignment is 16 bytes, but alignment
34130         is smaller.
34131         (aarch64_function_arg_boundary): Likewise.  Simplify using MIN/MAX.
34132         (aarch64_gimplify_va_arg_expr): Adjust aarch64_function_arg_alignment
34133         caller.
34135 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34137         * config/arc/simdext.md (dmpyh): Fix typo.
34139 2017-04-25  Richard Biener  <rguenther@suse.de>
34141         PR tree-optimization/80492
34142         * alias.c (compare_base_decls): Handle registers with asm
34143         specification conservatively.
34144         * tree-ssa-alias.c (decl_refs_may_alias_p): Handle
34145         compare_base_decls returning dont-know properly.
34147 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34149         * config/arc/arc.c (LEGITIMATE_OFFSET_ADDRESS_P): Delete macro.
34150         (legitimate_offset_address_p): New function.
34151         (arc_legitimate_address_p): Use above function.
34153 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34155         * config/arc/arc.c (arc_output_mi_thunk): Emit PIC calls.
34157 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34159         * config/arc/arc.c (arc_conditional_register_usage): Use ACCL,
34160         ACCH registers whenever they are available.
34162 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34164         * config/arc/arc.c (arc_conditional_register_usage): Make D0, D1
34165         double regs fix when not used.
34167 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34169         * config/arc/arc.h (REGNO_OK_FOR_BASE_P): Consider also extension
34170         core registers.
34171         (REG_OK_FOR_INDEX_P_NONSTRICT): Likewise.
34172         (REG_OK_FOR_BASE_P_NONSTRICT): Likewise.
34174 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34176         * config/arc/arc.c (arc_output_addsi): Check for h-register class
34177         when emitting short ADD instructions.
34179 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34181         * config/arc/arc.md (cmpsi_cc_insn_mixed): Use 'h' register
34182         constraint.
34183         (cmpsi_cc_c_insn): Likewise.
34184         (cbranchsi4_scratch): Compute proper instruction length using
34185         compact_hreg_operand.
34186         * config/arc/predicates.md (compact_hreg_operand): New predicate.
34188 2017-04-25  Richard Biener  <rguenther@suse.de>
34190         PR middle-end/80509
34191         * passes.c (pass_manager::pass_manager): Initialize
34192         m_name_to_pass_map.
34194 2017-04-25  Richard Biener  <rguenther@suse.de>
34196         PR tree-optimization/79201
34197         * tree-ssa-sink.c (statement_sink_location): Handle calls.
34199 2017-04-25  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
34201         PR target/80464
34202         * config/s390/vector.md: Split MEM->GPR vector moves for
34203         non-s_operand addresses.
34205 2017-04-25  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
34207         PR target/79895
34208         * config/s390/predicates.md (reload_const_wide_int_operand): New
34209         predicate.
34210         * config/s390/s390.md ("movti"): Remove d/P alternative.
34211         ("movti_bigconst"): New pattern definition.
34213 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
34215         PR target/80080
34216         * s390-protos.h (s390_expand_cs_hqi): Removed.
34217         (s390_expand_cs, s390_expand_atomic_exchange_tdsi): New prototypes.
34218         * config/s390/s390.c (s390_emit_compare_and_swap): Handle all integer
34219         modes as well as CCZ1mode and CCZmode.
34220         (s390_expand_atomic_exchange_tdsi, s390_expand_atomic): Adapt to new
34221         signature of s390_emit_compare_and_swap.
34222         (s390_expand_cs_hqi): Likewise, make static.
34223         (s390_expand_cs_tdsi): Generate an explicit compare before trying
34224         compare-and-swap, in some cases.
34225         (s390_expand_cs): Wrapper function.
34226         (s390_expand_atomic_exchange_tdsi): New backend specific expander for
34227         atomic_exchange.
34228         (s390_match_ccmode_set): Allow CCZmode <-> CCZ1 mode.
34229         * config/s390/s390.md ("atomic_compare_and_swap<mode>"): Merge the
34230         patterns for small and large integers.  Forbid symref memory operands.
34231         Move expander to s390.c.  Require cc register.
34232         ("atomic_compare_and_swap<DGPR:mode><CCZZ1:mode>_internal")
34233         ("*atomic_compare_and_swap<TDI:mode><CCZZ1:mode>_1")
34234         ("*atomic_compare_and_swapdi<CCZZ1:mode>_2")
34235         ("*atomic_compare_and_swapsi<CCZZ1:mode>_3"): Use s_operand to forbid
34236         symref memory operands.  Remove CC mode and call s390_match_ccmode
34237         instead.
34238         ("atomic_exchange<mode>"): Allow and implement all integer modes.
34240 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
34242         * config/s390/s390.md (define_peephole2): New peephole to help
34243         combining the load-and-test pattern with volatile memory.
34245 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
34247         * config/s390/s390.md ("cstorecc4"): Use load-on-condition and deal
34248         with CCZmode for TARGET_Z196.
34250 2017-04-25  Jakub Jelinek  <jakub@redhat.com>
34252         PR rtl-optimization/80501
34253         * combine.c (make_compound_operation_int): Set subreg_code to SET
34254         even for AND with mask of the sign bit of mode.
34256         PR rtl-optimization/80500
34257         * loop-unroll.c (combine_var_copies_in_loop_exit): Call copy_rtx on
34258         sum's initial value.
34260 2017-04-25  Julian Brown  <julian@codesourcery.com>
34261             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
34263         * config/aarch64/thunderx2t99.md (thunderx2t99_crc): New Reservation.
34265 2017-04-25  Marc Glisse  <marc.glisse@inria.fr>
34267         * fold-const.c (tree_single_nonzero_warnv_p): Handle SSA_NAME.
34269 2017-04-25  Julian Brown  <julian@codesourcery.com>
34270             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
34272         * config/aarch64/thunderx2t99.md (thunderx2t99_aes): New Reservation.
34273         (thunderx2t99_sha): New Reservation.
34275 2017-04-25  Julian Brown  <julian@codesourcery.com>
34276             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
34278         * config/aarch64/aarch64-simd.md (aarch64_simd_vec_set<mode>): Fix
34279         type for 1-element load.
34281 2017-04-24  Marc Glisse  <marc.glisse@inria.fr>
34283         * match.pd (X/[ex]C CMP Y/[ex]C): New transformation.
34285 2017-04-24  Martin Jambor  <mjambor@suse.cz>
34287         PR tree-optimization/80293
34288         * tree-sra.c (scalarizable_type_p): New parameter const_decl, make
34289         char arrays not totally scalarizable if it is false.
34290         (analyze_all_variable_accesses): Pass correct value in the new
34291         parameter.  Add a statistics counter.
34293 2017-04-24  Jan Hubicka  <hubicka@ucw.cz>
34295         PR middle-end/79931
34296         * ipa-devirt.c (dump_possible_polymorphic_call_targets): Fix ICE.
34298 2017-04-24  Richard Biener  <rguenther@suse.de>
34300         PR tree-optimization/80494
34301         * tree-scalar-evolution.c (analyze_scalar_evolution_1): Bail
34302         out for complex types.
34304 2017-04-24  Richard Biener  <rguenther@suse.de>
34306         * tree-ssa-sccvn.h (run_scc_vn): Adjust prototype.
34307         * tree-ssa-sccvn.c (print_scc): Print SCC size.
34308         (extract_and_process_scc_for_name): Never fail but drop SCC to varying.
34309         (DFS): Adjust and never fail.
34310         (sccvn_dom_walker::fail): Remove.
34311         (sccvn_dom_walker::before_dom_children): Adjust.
34312         (run_scc_vn): Likewise and never fail.
34313         * tree-ssa-pre.c (pass_pre::execute): Adjust.
34314         (pass_fre::execute): Likewise.
34316 2017-04-24  Richard Biener  <rguenther@suse.de>
34318         PR tree-optimization/79725
34319         * tree-ssa-sink.c (statement_sink_location): Return whether
34320         failure reason was zero uses.  Move that check later.
34321         (sink_code_in_bb): Deal with zero uses by removing the stmt
34322         if possible.
34324 2017-04-24  Richard Biener  <rguenther@suse.de>
34326         PR c++/2972
34327         * tree-ssa-uninit.c (warn_uninitialized_vars): Handle some
34328         pointer-based references.
34330 2017-04-24  Richard Biener  <rguenther@suse.de>
34332         PR bootstrap/79814
34333         * pass_manager.h (pass_manager::operator new): Remove.
34334         (pass_manager::operator delete): Likewise.
34335         * passes.c (pass_manager::operator new): Remove.
34336         (pass_manager::operator delete): Likewise.
34337         (pass_manager::pass_manager): Zero individual pass members.
34339 2017-04-23  Uros Bizjak  <ubizjak@gmail.com>
34341         PR target/70799
34342         * config/i386/i386.c (dimode_scalar_to_vector_candidate_p)
34343         <case ASHIFT, case LSHIFTRT>: Also consider variable shifts.
34344         Check "XEXP (src, 1)" operand here.
34345         <case PLUS, case MINUS, case IOR, case XOR, case AND>:
34346         Check "XEXP (src, 1)" operand here.
34347         (dimode_scalar_chain::make_vector_copies): Detect count register
34348         of a shift instruction.  Zero extend count register from QImode
34349         to DImode to satisfy vector shift pattern count operand predicate.
34350         Substitute vector shift count operand with a DImode copy.
34351         (dimode_scalar_chain::convert_reg): Ditto, zero-extend from
34352         vector register.
34354 2017-04-21  Uros Bizjak  <ubizjak@gmail.com>
34356         * config/i386/i386.md (*extzvqi_mem_rex64): Move above *extzv<mode>.
34357         Remove UNSPEC_NOREX_MEM tag.  Update corresponding peephole2 pattern.
34358         (*insvqi_1_mem_rex64): Move above insv<mode>_1.  Remove
34359         UNSPEC_NOREX_MEM tag.  Update corresponding peephole2 pattern.
34360         (UNSPEC_NOREX_MEM): Remove definition.
34362 2017-04-21  Richard Biener  <rguenther@suse.de>
34364         PR tree-optimization/79547
34365         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
34366         Handle strlen, strcmp, strncmp, strcasecmp, strncasecmp, memcmp,
34367         bcmp, strspn, strcspn, __builtin_object_size and __builtin_constant_p
34368         without any constraints.
34370 2017-04-21  Richard Biener  <rguenther@suse.de>
34372         PR tree-optimization/78847
34373         * fold-const.c (split_tree): Handle POINTER_PLUS_EXPR.
34375 2017-04-21  Richard Biener  <rguenther@suse.de>
34377         * tree.h (build_qualified_type): Annotate with CXX_MEM_STAT_INFO.
34378         (build_distinct_type_copy): Likewise.
34379         (build_variant_type_copy): Likewise.
34380         * tree.c (build_qualified_type): Pass down mem-stat info.
34381         (build_distinct_type_copy): Likewise.
34382         (build_variant_type_copy): Likewise.
34384 2017-04-21  Richard Biener  <rguenther@suse.de>
34386         PR tree-optimization/80237
34387         * tree-ssa-pre.c (find_leader_in_sets): Add third set argument,
34388         defaulted to NULL.
34389         (phi_translate_1): Also allow a leader in AVAIL_OUT of pred
34390         for a simplified result.
34392 2016-04-21  Richard Biener  <rguenther@suse.de>
34394         * tree-ssa-loop-ivcanon.c (constant_after_peeling): Do not require
34395         sth as strict as a simple_iv but a chrec without symbols and an
34396         operand defined in the loop we are peeling (and not some subloop).
34397         (propagate_constants_for_unrolling): Propagate all constants.
34399 2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
34401         PR target/79804
34402         * config/i386/i386.c (print_reg): Remove assert for disalowed
34403         regno values, call output_operand_lossage instead.
34405 2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
34407         PR target/78090
34408         * config/i386/constraints.md (Yc): New register constraint.
34409         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
34410         Use Yc constraint for alternative 2 of operand 0.  Remove
34411         preferred_for_speed attribute.
34413 2017-04-20  Alexander Monakov  <amonakov@ispras.ru>
34415         * omp-low.c (lower_lastprivate_clauses): Correct handling of linear and
34416         lastprivate clauses in SIMT case.
34418 2017-04-20  Volker Reichelt  <v.reichelt@netcologne.de>
34420         * doc/invoke.texi (-Wextra-semi): Document new warning option.
34422 2017-04-20  Richard Biener  <rguenther@suse.de>
34424         PR tree-optimization/57796
34425         * tree-vect-stmts.c (vect_model_store_cost): Cost scatters
34426         as N scalar stores.
34427         (vect_model_load_cost): Cost gathers as N scalar loads.
34429 2017-04-20  Richard Biener  <rguenther@suse.de>
34431         * ggc-page.c (ggc_allocated_p): Rename to ...
34432         (safe_lookup_page_table_entry): ... this and return the lookup
34433         result.
34434         (gt_ggc_m_S): Use safe_lookup_page_table_entry.
34436 2017-04-20  Richard Biener  <rguenther@suse.de>
34438         PR tree-optimization/80453
34439         * tree-ssa-sccvn.h (struct vn_phi_s): Add cclhs and ccrhs members.
34440         * tree-ssa-sccvn.c (cond_stmts_equal_p): Use recorded lhs and rhs
34441         from the conditions.
34442         (vn_phi_eq): Pass them down.
34443         (vn_phi_lookup): Record them.
34444         (vn_phi_insert): Likewise.
34446 2017-04-20  Matthew Fortune  <matthew.fortune@imgtec.com>
34448         * config/mips/mips.c (mips_expand_vec_perm_const): Re-fix
34449         uninitialized variable warning to avoid buffer overrun.
34451 2017-04-20  Alexander Monakov  <amonakov@ispras.ru>
34453         PR other/71250
34454         * doc/invoke.texi (-Wmissing-field-initializers): Mention that warning
34455         is suppressed for '{ 0 }' in C.
34457 2017-04-20  Jakub Jelinek  <jakub@redhat.com>
34459         * BASE-VER: Set to 8.0.0.
34461 2017-04-20  Thomas Preud'homme  <thomas.preudhomme@arm.com>
34463         * config/arm/arm.c (arm_elf_asm_cdtor): Create non-default
34464         priority .init_array and .fini_array section with SECTION_NOTYPE
34465         flag.
34467 2017-04-20  Jakub Jelinek  <jakub@redhat.com>
34469         PR middle-end/80423
34470         * tree.h (build_array_type): Add typeless_storage default argument.
34471         * tree.c (type_cache_hasher::equal): Also compare
34472         TYPE_TYPELESS_STORAGE flag for ARRAY_TYPEs.
34473         (build_array_type): Add typeless_storage argument, set
34474         TYPE_TYPELESS_STORAGE to it, if shared also hash it, and pass to
34475         recursive call.
34476         (build_nonshared_array_type): Adjust build_array_type_1 caller.
34477         (build_array_type): Likewise.  Add typeless_storage argument.
34479 2017-04-19  Eric Botcazou  <ebotcazou@adacore.com>
34480             Jakub Jelinek  <jakub@redhat.com>
34482         PR tree-optimization/80426
34483         * tree-vrp.c (extract_range_from_binary_expr_1): For an additive
34484         operation on symbolic operands, also compute the overflow for the
34485         invariant part when the operation degenerates into a negation.
34487 2017-04-19  Jakub Jelinek  <jakub@redhat.com>
34489         PR debug/80461
34490         * dwarf2out.c (modified_type_die, gen_type_die_with_usage):
34491         Check for t with zero TYPE_QUALS_NO_ADDR_SPACE.
34493         PR debug/80436
34494         * tree-ssa-loop-manip.c (find_uses_to_rename_def): Ignore debug uses.
34496 2017-04-19  Georg-Johann Lay  <avr@gjlay.de>
34498         PR target/80462
34499         * config/avr/avr.c (tree.h): Include it.
34500         (cgraph.h): Include it.
34501         (avr_encode_section_info): Don't warn for uninitialized progmem
34502         variable if it's just an alias.
34504 2017-04-19  Richard Biener  <rguenther@suse.de>
34506         PR ipa/65972
34507         * auto-profile.c (afdo_vpt_for_early_inline): Update SSA
34508         when needed by AutoPGO.
34510 2017-04-19  Paulo J. Matos  <paulo@matos-sorge.com>
34512         PR lto/50345
34513         * doc/lto.texi: Remove an extra 'that'.
34515 2017-04-19  Segher Boessenkool  <segher@kernel.crashing.org>
34517         PR rtl-optimization/80429
34518         * ira.c (split_live_ranges_for_shrink_wrap): Don't split regs that
34519         are only used in debug insns.
34521 2017-04-19  Eric Botcazou  <ebotcazou@adacore.com>
34522             Vladimir Makarov  <vmakarov@redhat.com>
34524         * config/sparc/predicates.md (input_operand): Add comment.  Return
34525         true for any memory operand when LRA is in progress.
34526         * config/sparc/sparc.c (sparc_expand_move): Minor formatting fix.
34528 2017-04-18  Jeff Law  <law@redhat.com>
34530         PR target/74563
34531         * mips.md ({return,simple_return}_internal): Do not overwrite
34532         operands[0].
34534 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34536         PR tree-optimization/80443
34537         * tree-vrp.c (intersect_ranges): For signed 1-bit precision type,
34538         instead of adding 1, subtract -1 and similarly instead of subtracting
34539         1 add -1.
34541 2017-04-18  Richard Sandiford  <richard.sandiford@arm.com>
34543         PR rtl-optimization/80357
34544         * haifa-sched.c (tmp_bitmap): New variable.
34545         (model_recompute): Handle duplicate use records.
34546         (alloc_global_sched_pressure_data): Initialize tmp_bitmap.
34547         (free_global_sched_pressure_data): Free it.
34549 2017-04-18  Bernd Edlinger  <bernd.edlinger@hotmail.de>
34551         Revert:
34552         2017-02-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
34553         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
34554         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
34555         instead of SYSTEM_HEADER_DIR.
34557 2017-04-18  Jeff Law  <law@redhat.com>
34559         PR middle-end/80422
34560         * cfgcleanup.c (try_crossjump_to_edge): Verify SRC1 and SRC2 have
34561         predecessors after walking up the insn chain.
34563 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34565         PR debug/80263
34566         * dwarf2out.c (modified_type_die): Try harder not to emit internal
34567         sizetype type into debug info.
34569 2017-04-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
34571         PR target/80099
34572         * config/rs6000/rs6000.c (rs6000_expand_vector_extract): Eliminate
34573         unneeded test for TARGET_UPPER_REGS_SF.
34574         * config/rs6000/vsx.md (vsx_extract_v4sf_var): Likewise.
34576 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34578         PR sanitizer/80444
34579         * sancov.c (sancov_pass): Use gsi_start_nondebug_after_labels_bb
34580         instead of gsi_after_labels.
34582 2017-04-18  Jeff Law  <law@redhat.com>
34584         * regcprop.c (maybe_mode_change): Avoid creating copies of the
34585         stack pointer.
34587         Revert:
34588         2017-04-13  Jeff Law  <law@redhat.com>
34589         * config/mips.mips.md (zero_extendsidi2): Do not allow SP to appear
34590         in operands[1] if it is a MEM and TARGET_MIPS16 is active.
34592 2017-04-18  Georg-Johann Lay  <avr@gjlay.de>
34594         PR target/79453
34595         * config/avr/avr.c (intl.h): Include it.
34596         (avr_pgm_check_var_decl) [reason]: Wrap diagnostic snippets into _().
34598 2017-04-18  Martin Liska  <mliska@suse.cz>
34600         PR gcov-profile/78783
34601         * gcov-tool.c (gcov_output_files): Validate that destination
34602         file is either removed by the tool or by a user.
34604 2017-04-14  Andrew Burgess  <andrew.burgess@embecosm.com>
34605             Guy Benyei  <guybe@mellanox.com>
34607         * config/arc/arc.c (arc_reorg): Move loop_end_id into a more local
34608         block, and do not negate it, the stored id is already negative.
34610 2017-04-14  Andrew Burgess  <andrew.burgess@embecosm.com>
34612         * config/arc/arc.md (doloop_begin_i): Use @pcl assembler syntax.
34614 2017-04-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
34616         PR target/80098
34617         * config/rs6000/rs6000-cpus.def (OTHER_P9_VECTOR_MASKS): Define
34618         masks of options that should be turned off if the VSX vector
34619         options are turned off.
34620         (OTHER_P8_VECTOR_MASKS): Likewise.
34621         (OTHER_VSX_VECTOR_MASKS): Likewise.
34622         * config/rs6000/rs6000.c (rs6000_option_override_internal): Call
34623         rs6000_disable_incompatible_switches to validate no type switches
34624         like -mvsx.
34625         (rs6000_incompatible_switch): New function to disallow turning on
34626         other vector options if -mno-vsx, -mno-power8-vector, or
34627         -mno-power9-vector are specified.
34629 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34631         * config/arc/arc.h (CRT_CALL_STATIC_FUNCTION): Use long calls.
34633 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34635         * config/arc/arc-protos.h (arc_decl_pretend_args): Remove.
34636         * config/arc/arc.c (arc_decl_pretend_args): Likewise.
34637         * config/arc/arc.h (CFA_FRAME_BASE_OFFSET): Likewise.
34638         (ARG_POINTER_CFA_OFFSET): Likewise.
34640 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34642         * config/arc/arc.c (arc_mode_dependent_address_p): Relax
34643         conditions to take advantage of various optimizations.
34645 2017-04-13  Jeff Law  <law@redhat.com>
34647         * config/mips.mips.md (zero_extendsidi2): Do not allow SP to appear
34648         in operands[1] if it is a MEM and TARGET_MIPS16 is active.
34649         (zero_extendsidi2_dext): Likewise.
34651 2017-04-13  Jakub Jelinek  <jakub@redhat.com>
34653         PR sanitizer/80403
34654         * fold-const.c (fold_ternary_loc): Revert
34655         use op0 instead of fold_convert_loc (loc, type, arg0) part of
34656         2017-04-12 change.
34658 2017-04-13  Vladimir Makarov  <vmakarov@redhat.com>
34660         PR rtl-optimization/80343
34661         * lra-remat.c (update_scratch_ops): Assign original hard reg to
34662         new scratch pseudo.
34664 2017-04-13  Denis Khalikov <d.khalikov@partner.samsung.com>
34666         PR sanitizer/80414
34667         * ubsan.c (ubsan_expand_bounds_ifn): Pass original index
34668         to ubsan_encode_value.
34670 2017-04-13  Jeff Law  <law@redhat.com>
34672         * reload1.c (eliminate_regs_1): Call gen_rtx_raw_SUBREG for SUBREGs
34673         appearing in DEBUG_INSNs.
34675 2017-04-13  Martin Liska  <mliska@suse.cz>
34677         PR gcov-profile/80413
34678         * gcov-io.c (gcov_write_string): Copy to buffer just when
34679         allocated size is greater than zero.
34681 2017-04-13  Jakub Jelinek  <jakub@redhat.com>
34683         PR debug/80321
34684         * dwarf2out.c (decls_for_scope): Ignore declarations of
34685         current_function_decl in BLOCK_NONLOCALIZED_VARS.
34687 2017-04-12  Jan Hubicka  <hubicka@ucw.cz>
34689         PR lto/69953
34690         * ipa-visibility.c (non_local_p): Fix typos.
34691         (localize_node): When localizing symbol in same comdat group,
34692         dissolve the group only when we know external symbols are going
34693         to be privatized.
34694         (function_and_variable_visibility): Do not localize DECL_EXTERNAL.
34696 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34698         PR tree-optimization/79390
34699         * optabs.c (emit_conditional_move): If the preferred op2/op3 operand
34700         order does not result in usable sequence, retry with reversed operand
34701         order.
34703         PR sanitizer/80403
34704         PR sanitizer/80404
34705         PR sanitizer/80405
34706         * fold-const.c (fold_ternary_loc): Use op1 instead of arg1 as argument
34707         to fold_build2_loc.  Convert TREE_OPERAND (tem, 0) to type.  Use
34708         op0 instead of fold_convert_loc (loc, type, arg0).
34710 2017-04-12  Jeff Law  <law@redhat.com>
34712         * genattrtab.c (write_eligible_delay): Verify DELAY_INSN still
34713         has a delay slot in the generated code.
34715         * config/cris/cris.md (cris_preferred_reload_class): Return
34716         GENNONACR_REGS rather than GENERAL_REGS.
34718 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34720         PR c/80163
34721         * expr.c <CASE_CONVERT>: For EXPAND_INITIALIZER determine SIGN_EXTEND
34722         vs. ZERO_EXTEND based on signedness of treeop0's type rather than
34723         signedness of the result type.
34725 2017-04-12  Richard Biener  <rguenther@suse.de>
34726             Jeff Law  <law@redhat.com>
34728         PR tree-optimization/80359
34729         * tree-ssa-dse.c (maybe_trim_partially_dead_store): Do not
34730         trim stores to TARGET_MEM_REFs.
34732 2017-04-12  Richard Biener  <rguenther@suse.de>
34734         PR tree-optimization/79390
34735         * gimple-ssa-split-paths.c (is_feasible_trace): Restrict
34736         threading case even more.
34738 2017-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
34740         PR target/80382
34741         * config/rs6000/sync.md (atomic_load<mode>, atomic_store<mode): Test
34742         for quad_address_p for TImode, instead of just not indexed_address.
34744 2017-04-12  Richard Biener  <rguenther@suse.de>
34745             Bernd Edlinger  <bernd.edlinger@hotmail.de>
34747         PR middle-end/79671
34748         * alias.c (component_uses_parent_alias_set_from): Handle
34749         TYPE_TYPELESS_STORAGE.
34750         (get_alias_set): Likewise.
34751         * tree-core.h (tree_type_common): Add typeless_storage flag.
34752         * tree.h (TYPE_TYPELESS_STORAGE): New macro.
34753         * stor-layout.c (place_union_field): Set TYPE_TYPELESS_STORAGE
34754         for types containing members with TYPE_TYPELESS_STORAGE.
34755         (place_field): Likewise.
34756         (layout_type): Likewise for ARRAY_TYPE.
34757         * lto-streamer-out.c (hash_tree): Hash TYPE_TYPELESS_STORAGE.
34758         * tree-streamer-in.c (unpack_ts_type_common_value_fields): Stream
34759         TYPE_TYPELESS_STORAGE.
34760         * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
34762 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34764         PR sanitizer/80349
34765         * fold-const.c (fold_binary_loc) <case BIT_IOR_EXPR>: Convert arg0's
34766         first argument to type.
34768 2017-04-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
34770         PR target/80376
34771         PR target/80315
34772         * config/rs6000/rs6000.c (rs6000_expand_unop_builtin): Return
34773         CONST0_RTX (mode) rather than const0_rtx where appropriate.
34774         (rs6000_expand_binop_builtin): Likewise.
34775         (rs6000_expand_ternop_builtin): Likewise; also add missing
34776         vsx_xxpermdi_* variants; also fix typo (arg1 => arg2) for
34777         vshasigma built-ins.
34778         * doc/extend.texi: Document that vec_xxpermdi's third argument
34779         must be a constant.
34781 2017-04-11  Uros Bizjak  <ubizjak@gmail.com>
34783         * config/i386/i386.c (dimode_scalar_chain::compute_convert_gain):
34784         Use shift_const cost parameter when calculating gain of STV shifts.
34786 2017-04-11  Vladimir Makarov  <vmakarov@redhat.com>
34788         PR rtl-optimization/70478
34789         * lra-constraints.c (process_alt_operands): Check memory for
34790         disfavoring memory insn operand.
34792 2017-04-11  Jakub Jelinek  <jakub@redhat.com>
34794         PR middle-end/80100
34795         * simplify-rtx.c (simplify_binary_operation_1) <case IOR>: Perform
34796         left shift in unsigned HOST_WIDE_INT type.
34798         PR rtl-optimization/80385
34799         * simplify-rtx.c (simplify_unary_operation_1): Don't transform
34800         (not (neg X)) into (plus X -1) for complex or non-integral modes.
34802         PR libgomp/80394
34803         * omp-low.c (scan_omp_task): Don't optimize away empty tasks
34804         if they have any depend clauses.
34806 2017-04-11  Martin Liska  <mliska@suse.cz>
34808         PR ipa/80212
34809         * cgraph.c (cgraph_node::dump): Dump calls_comdat_local.
34810         * ipa-split.c (split_function): Create a local comdat symbol
34811         if caller is in a comdat group.
34813 2017-04-11  Martin Liska  <mliska@suse.cz>
34815         PR ipa/80212
34816         * ipa-cp.c (determine_versionability): Handle calls_comdat_local
34817         flags.
34819 2017-04-11  Martin Sebor  <msebor@redhat.com>
34821         PR middle-end/80364
34822         * gimple-ssa-sprintf.c (get_int_range): Remove second argument and
34823         always use the int type.  Use INTEGRAL_TYPE_P() rather than testing
34824         for INTEGER_TYPE.
34825         (directive::set_width, directive::set_precision, format_character):
34826         Adjust.
34827         (parse_directive): Use INTEGRAL_TYPE_P() rather than testing for
34828         INTEGER_TYPE.
34830 2017-04-11  Richard Earnshaw  <rearnsha@arm.com>
34832         PR target/80389
34833         * config/arm/arm.c (arm_configure_build_target): When -mcpu and -arch
34834         conflict, set target->arch_name instead of target->cpu_name.
34836 2017-04-11  Richard Biener  <rguenther@suse.de>
34838         PR tree-optimization/80374
34839         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Use
34840         build_zero_cst, remove fold_convertible_p check again.
34842 2017-04-11  Martin Liska  <mliska@suse.cz>
34844         PR sanitizer/70878
34845         * ubsan.c (instrument_object_size): Do not instrument register
34846         variables.
34848 2017-04-11  Jakub Jelinek  <jakub@redhat.com>
34850         PR target/80381
34851         * config/i386/i386-builtin-types.def
34852         (V16HI_FTYPE_V16HI_INT_V16HI_UHI_COUNT,
34853         V16HI_FTYPE_V16HI_V8HI_V16HI_UHI_COUNT,
34854         V16SI_FTYPE_V16SI_INT_V16SI_UHI_COUNT,
34855         V16SI_FTYPE_V16SI_V4SI_V16SI_UHI_COUNT,
34856         V2DI_FTYPE_V2DI_INT_V2DI_UQI_COUNT,
34857         V2DI_FTYPE_V2DI_V2DI_V2DI_UQI_COUNT,
34858         V32HI_FTYPE_V32HI_INT_V32HI_USI_COUNT,
34859         V32HI_FTYPE_V32HI_V8HI_V32HI_USI_COUNT,
34860         V4DI_FTYPE_V4DI_INT_V4DI_UQI_COUNT,
34861         V4DI_FTYPE_V4DI_V2DI_V4DI_UQI_COUNT,
34862         V4SI_FTYPE_V4SI_INT_V4SI_UQI_COUNT,
34863         V4SI_FTYPE_V4SI_V4SI_V4SI_UQI_COUNT,
34864         V8DI_FTYPE_V8DI_INT_V8DI_UQI_COUNT,
34865         V8DI_FTYPE_V8DI_V2DI_V8DI_UQI_COUNT,
34866         V8HI_FTYPE_V8HI_INT_V8HI_UQI_COUNT,
34867         V8HI_FTYPE_V8HI_V8HI_V8HI_UQI_COUNT,
34868         V8SI_FTYPE_V8SI_INT_V8SI_UQI_COUNT,
34869         V8SI_FTYPE_V8SI_V4SI_V8SI_UQI_COUNT): New function type aliases.
34870         * config/i386/i386-builtin.def (__builtin_ia32_pslld512_mask,
34871         __builtin_ia32_pslldi512_mask, __builtin_ia32_psllq512_mask,
34872         __builtin_ia32_psllqi512_mask, __builtin_ia32_psrad512_mask,
34873         __builtin_ia32_psradi512_mask, __builtin_ia32_psraq512_mask,
34874         __builtin_ia32_psraqi512_mask, __builtin_ia32_psrld512_mask,
34875         __builtin_ia32_psrldi512_mask, __builtin_ia32_psrlq512_mask,
34876         __builtin_ia32_psrlqi512_mask, __builtin_ia32_psllwi128_mask,
34877         __builtin_ia32_pslldi128_mask, __builtin_ia32_psllqi128_mask,
34878         __builtin_ia32_psllw128_mask, __builtin_ia32_pslld128_mask,
34879         __builtin_ia32_psllq128_mask, __builtin_ia32_psllwi256_mask,
34880         __builtin_ia32_psllw256_mask, __builtin_ia32_pslldi256_mask,
34881         __builtin_ia32_pslld256_mask, __builtin_ia32_psllqi256_mask,
34882         __builtin_ia32_psllq256_mask, __builtin_ia32_psradi128_mask,
34883         __builtin_ia32_psrad128_mask, __builtin_ia32_psradi256_mask,
34884         __builtin_ia32_psrad256_mask, __builtin_ia32_psraqi128_mask,
34885         __builtin_ia32_psraq128_mask, __builtin_ia32_psraqi256_mask,
34886         __builtin_ia32_psraq256_mask, __builtin_ia32_psrldi128_mask,
34887         __builtin_ia32_psrld128_mask, __builtin_ia32_psrldi256_mask,
34888         __builtin_ia32_psrld256_mask, __builtin_ia32_psrlqi128_mask,
34889         __builtin_ia32_psrlq128_mask, __builtin_ia32_psrlqi256_mask,
34890         __builtin_ia32_psrlq256_mask, __builtin_ia32_psrawi256_mask,
34891         __builtin_ia32_psraw256_mask, __builtin_ia32_psrawi128_mask,
34892         __builtin_ia32_psraw128_mask, __builtin_ia32_psrlwi256_mask,
34893         __builtin_ia32_psrlw256_mask, __builtin_ia32_psrlwi128_mask,
34894         __builtin_ia32_psrlw128_mask, __builtin_ia32_psllwi512_mask,
34895         __builtin_ia32_psllw512_mask, __builtin_ia32_psrawi512_mask,
34896         __builtin_ia32_psraw512_mask, __builtin_ia32_psrlwi512_mask,
34897         __builtin_ia32_psrlw512_mask): Use _COUNT suffixed function type
34898         aliases.
34899         * config/i386/i386.c (ix86_expand_args_builtin): Rename last_arg_count
34900         flag to second_arg_count, handle 4 argument function type _COUNT
34901         aliases, handle second_arg_count on second argument rather than last.
34903 2017-04-10  Jeff Law  <law@redhat.com>
34905         PR tree-optimization/80374
34906         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Do not try to
34907         record anything if we can not convert integer_zero_node to the
34908         desired type.
34910 2017-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
34912         PR target/80108
34913         * config/rs6000/rs6000.c (rs6000_option_override_internal):
34914         Enhance special handling given to the TARGET_P9_MINMAX option in
34915         relation to certain other options.
34917 2017-04-10  Bin Cheng  <bin.cheng@arm.com>
34919         PR tree-optimization/80153
34920         * tree-ssa-loop-ivopts.c (add_iv_candidate_for_use): Check and
34921         remove POINTER_PLUS_EXPR's base part directly, rather than through
34922         aff_tree.
34924 2017-04-10  Richard Biener  <rguenther@suse.de>
34925             Bin Cheng  <bin.cheng@arm.com>
34927         PR tree-optimization/80153
34928         * tree-affine.c (aff_combination_to_tree): Get base pointer from
34929         the first element of pointer type aff_tree.  Build result expr in
34930         aff_tree's type.
34931         (add_elt_to_tree): Convert to type unconditionally.  Remove other
34932         fold_convert calls.
34933         * tree-ssa-loop-ivopts.c (alloc_iv): Pass in consistent types.
34934         (rewrite_use_nonlinear_expr): Check invariant using iv information.
34936 2017-04-10  Richard Biener  <rguenther@suse.de>
34938         * tree-ssa-structalias.c (find_func_aliases): Properly handle
34939         asm inputs.
34941 2017-04-10  Vladimir Makarov  <vmakarov@redhat.com>
34943         PR rtl-optimization/70478
34944         * lra-constraints.c (curr_small_class_check): New.
34945         (update_and_check_small_class_inputs): New.
34946         (process_alt_operands): Update curr_small_class_check.  Disfavor
34947         alternative insn memory operands.  Check available regs for small
34948         class operands.
34950 2017-03-31  Matthew Fortune  <matthew.fortune@imgtec.com>
34952         PR target/80057
34953         * config/mips/mips.opt (-mvirt): Update description.
34954         * doc/invoke.texi (-mvirt): Likewise.
34956 2017-04-10  Richard Biener  <rguenther@suse.de>
34958         PR middle-end/80362
34959         * fold-const.c (fold_binary_loc): Look at unstripped ops when
34960         looking for NEGATE_EXPR in -A / -B to A / B folding.
34962 2017-04-10  Martin Liska  <mliska@suse.cz>
34964         PR gcov-profile/80224
34965         * gcov.c (print_usage): Fix usage string.
34966         (get_gcov_intermediate_filename): Remove.
34967         (output_gcov_file): Use both for normal and intermediate format.
34968         (generate_results): Do not initialize special file for
34969         intermediate format.
34971 2017-04-10  Richard Biener  <rguenther@suse.de>
34973         PR tree-optimization/80304
34974         * tree-ssa-loop-im.c (ref_indep_loop_p_1): Also recurse
34975         for safelen.
34977 2017-04-10  Nathan Sidwell  <nathan@acm.org>
34979         PR target/79905
34980         * config/rs6000/rs6000.c (rs6000_vector_type): New.
34981         (rs6000_init_builtins): Use it.
34983 2016-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
34985         * config/arm/arm.md (<mrc>): Add mode to SET source.
34986         (<mrrc>): Likewise.
34988 2017-04-10  Richard Biener  <rguenther@suse.de>
34990         PR middle-end/80344
34991         * gimplify.c (is_gimple_mem_rhs_or_call): Allow CLOBBERs.
34993 2017-04-10  Jakub Jelinek  <jakub@redhat.com>
34995         PR target/80324
34996         * config/i386/avx512fintrin.h (_mm512_reduce_add_epi32,
34997         _mm512_reduce_mul_epi32, _mm512_reduce_and_epi32,
34998         _mm512_reduce_or_epi32, _mm512_mask_reduce_add_epi32,
34999         _mm512_mask_reduce_mul_epi32, _mm512_mask_reduce_and_epi32,
35000         _mm512_mask_reduce_or_epi32, _mm512_reduce_min_epi32,
35001         _mm512_reduce_max_epi32, _mm512_reduce_min_epu32,
35002         _mm512_reduce_max_epu32, _mm512_mask_reduce_min_epi32,
35003         _mm512_mask_reduce_max_epi32, _mm512_mask_reduce_min_epu32,
35004         _mm512_mask_reduce_max_epu32, _mm512_reduce_add_ps,
35005         _mm512_reduce_mul_ps, _mm512_mask_reduce_add_ps,
35006         _mm512_mask_reduce_mul_ps, _mm512_reduce_min_ps, _mm512_reduce_max_ps,
35007         _mm512_mask_reduce_min_ps, _mm512_mask_reduce_max_ps,
35008         _mm512_reduce_add_epi64, _mm512_reduce_mul_epi64,
35009         _mm512_reduce_and_epi64, _mm512_reduce_or_epi64,
35010         _mm512_mask_reduce_add_epi64, _mm512_mask_reduce_mul_epi64,
35011         _mm512_mask_reduce_and_epi64, _mm512_mask_reduce_or_epi64,
35012         _mm512_reduce_min_epi64, _mm512_reduce_max_epi64,
35013         _mm512_mask_reduce_min_epi64, _mm512_mask_reduce_max_epi64,
35014         _mm512_reduce_min_epu64, _mm512_reduce_max_epu64,
35015         _mm512_mask_reduce_min_epu64, _mm512_mask_reduce_max_epu64,
35016         _mm512_reduce_add_pd, _mm512_reduce_mul_pd, _mm512_mask_reduce_add_pd,
35017         _mm512_mask_reduce_mul_pd, _mm512_reduce_min_pd, _mm512_reduce_max_pd,
35018         _mm512_mask_reduce_min_pd, _mm512_mask_reduce_max_pd): New intrinsics.
35020 2017-04-08  Vladimir Makarov  <vmakarov@redhat.com>
35022         PR rtl-optimization/70478
35023         * lra-constraints.c: Reverse the last patch.
35025 2017-04-08  Andreas Tobler  <andreast@gcc.gnu.org>
35027         * config/aarch64/aarch64-freebsd.h: Define MCOUNT_NAME.
35028         Add comment for WCHAR_T.
35030 2017-04-08  Martin Liska  <mliska@suse.cz>
35032         Revert:
35033         2017-04-07  Martin Liska  <mliska@suse.cz>
35035         PR ipa/80212
35036         * ipa-split.c (split_function): Add function part to a same comdat
35037         group.
35039 2017-04-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
35041         PR target/80358
35042         * config/rs6000/rs6000.c (expand_block_compare): Fix boundary check.
35044 2017-04-07  Pat Haugen  <pthaugen@us.ibm.com>
35046         * rs6000/rs6000.c (vec_load_pendulum): Rename...
35047         (vec_pairing): ...to this.
35048         (power9_sched_reorder2): Rewrite code for pairing vector/vecload insns.
35049         (rs6000_sched_init): Adjust for name change.
35050         (struct rs6000_sched_context): Likewise.
35051         (rs6000_init_sched_context): Likewise.
35052         (rs6000_set_sched_context): Likewise.
35054 2017-04-07  Jakub Jelinek  <jakub@redhat.com>
35056         PR target/80322
35057         PR target/80323
35058         PR target/80325
35059         PR target/80326
35060         * config/i386/avxintrin.h (_mm256_cvtsd_f64, _mm256_cvtss_f32): New
35061         intrinsics.
35062         * config/i386/avx512fintrin.h (_mm512_int2mask, _mm512_mask2int,
35063         _mm512_abs_ps, _mm512_mask_abs_ps, _mm512_abs_pd, _mm512_mask_abs_pd,
35064         _mm512_cvtsd_f64, _mm512_cvtss_f32): Likewise.
35066 2017-04-07  Andreas Tobler  <andreast@gcc.gnu.org>
35068         * config/aarch64/aarch64-freebsd.h: Define WCHAR_TYPE.
35070 2017-04-07  Vladimir Makarov  <vmakarov@redhat.com>
35072         PR rtl-optimization/70703
35073         * ira-color.c (update_conflict_hard_regno_costs): Use
35074         int64_t instead of HOST_WIDE_INT.
35076 2017-04-07  Vladimir Makarov  <vmakarov@redhat.com>
35078         PR rtl-optimization/70478
35079         * lra-constraints.c (process_alt_operands): Disfavor alternative
35080         insn memory operands.
35082 2017-04-07  Jeff Law  <law@redhat.com>
35084         * config/iq2000/iq2000.c (final_prescan_insn): Do not separate a
35085         CALL and NOTE_INSN_CALL_ARG_LOCATION.
35087 2017-04-07  Martin Liska  <mliska@suse.cz>
35089         PR target/79889
35090         * config/aarch64/aarch64.c (aarch64_process_target_attr):
35091         Show error message instead of an ICE.
35093 2017-04-07  Martin Liska  <mliska@suse.cz>
35095         PR ipa/80212
35096         * ipa-split.c (split_function): Add function part to a same comdat
35097         group.
35099 2017-04-07  Richard Biener  <rguenther@suse.de>
35101         PR middle-end/80341
35102         * tree.c (get_unwidened): Also handle ! for_type case for
35103         INTEGER_CSTs.
35104         * convert.c (do_narrow): Split out from ...
35105         (convert_to_integer_1): ... here.  Do not pass final truncation
35106         type to get_unwidened for TRUNC_DIV_EXPR.
35108 2017-04-07  Richard Biener  <rguenther@suse.de>
35110         * tree-affine.c (wide_int_ext_for_comb): Take type rather
35111         than aff_tree.
35112         (aff_combination_const): Adjust.
35113         (aff_combination_scale): Likewise.
35114         (aff_combination_add_elt): Likewise.
35115         (aff_combination_add_cst): Likewise.
35116         (aff_combination_convert): Likewise.
35117         (add_elt_to_tree): Likewise.  Remove unused argument.
35118         (aff_combination_to_tree): Adjust calls to add_elt_to_tree.
35120 2017-04-07  Sebastian Huber  <sebastian.huber@embedded-brains.de>
35122         * config/arm/arm.h (ARM_DEFAULT_SHORT_ENUMS): Provide default
35123         definition.
35124         * config/arm/arm.c (arm_default_short_enums): Use
35125         ARM_DEFAULT_SHORT_ENUMS.
35126         * config/arm/rtems.h (ARM_DEFAULT_SHORT_ENUMS): Define.
35128 2017-04-06  Jakub Jelinek  <jakub@redhat.com>
35130         PR debug/80234
35131         * dwarf2out.c (gen_member_die): Handle C++17 inline static data
35132         members with redundant out-of-class redeclaration.
35134 2017-04-06  Uros Bizjak  <ubizjak@gmail.com>
35136         PR target/80286
35137         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): New pattern.
35138         * config/i386/i386.md (*zero_extendsidi2):
35139         Add (?*x,*x) and (?*v,*v) alternatives.
35141 2017-04-06  Uros Bizjak  <ubizjak@gmail.com>
35143         PR target/79733
35144         * config/i386/i386.c (ix86_expand_builtin)
35145         <case IX86_BUILTIN_K{,OR}TEST{C,Z}{8,16,32,64}>: Determine insn operand
35146         mode from insn data. Convert operands to insn operand mode.
35147         Copy operands that don't satisfy insn predicate to a register.
35149 2017-04-06  Sam Thursfield  <sam.thursfield@codethink.co.uk>
35151         * config/rs6000/x-aix: Increase memory limit for genautomata on AIX.
35152         Update comments.
35154 2017-04-06  Richard Biener  <rguenther@suse.de>
35156         PR tree-optimization/80334
35157         * tree-ssa-loop-ivopts.c (rewrite_use_address): Properly
35158         preserve alignment of accesses.
35160 2017-04-06  Richard Biener  <rguenther@suse.de>
35162         PR tree-optimization/80262
35163         * tree-sra.c (build_ref_for_offset): Preserve address-space
35164         information.
35165         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address):
35166         Drop useless address-space information on MEM_REF offsets.
35168 2017-04-05  Andreas Schwab  <schwab@linux-m68k.org>
35170         * builtins.def (BUILT_IN_UPDATE_SETJMP_BUF): Fix type.
35172 2017-04-05  Vladimir Makarov  <vmakarov@redhat.com>
35174         PR rtl-optimization/70703
35175         * ira-color.c (update_conflict_hard_regno_costs): Use
35176         HOST_WIDE_INT instead of long.
35178 2017-04-05  Uros Bizjak  <ubizjak@gmail.com>
35180         PR target/80298
35181         * config/i386/mmintrin.h: Add -msse target option when __SSE__ is
35182         not defined for x86_64 target.  Add -mmmx target option when __SSE2__
35183         is not defined.
35184         * config/i386/mm3dnow.h: Add -msse target when __SSE__ is not defined
35185         for x86_64 target.  Handle -m3dnowa option.
35187 2017-04-05  Vladimir Makarov  <vmakarov@redhat.com>
35189         PR rtl-optimization/70703
35190         * ira-color.c (update_costs_from_allocno): Use the smallest mode.
35191         (update_conflict_hard_regno_costs): Use long instead of unsigned
35192         arithmetic for cost calculation.
35194 2017-04-05  Jakub Jelinek  <jakub@redhat.com>
35195             Bernd Edlinger  <bernd.edlinger@hotmail.de>
35197         PR sanitizer/80308
35198         * asan.c (asan_store_shadow_bytes): Fix location of last_chunk_value
35199         for big endian.
35201 2017-04-05  Eric Botcazou  <ebotcazou@adacore.com>
35203         PR target/78002
35204         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Replace
35205         ptr_mode with Pmode throughout.
35206         * config/aarch64/aarch64.md (probe_stack_range_<PTR:mode): Rename
35207         into probe_stack_range and use DImode.
35209 2017-04-05  Dominik Vogt  <vogt@linux.vnet.ibm.com>
35211         PR target/79890
35212         * config/s390/s390.c (s390_register_info_gprtofpr): Return if
35213         call_eh_return is true.
35215 2017-04-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35217         * config/s390/s390-c.c (s390_resolve_overloaded_builtin):
35218         Initialize last_match_fntype_index.
35220 2017-04-05  Jakub Jelinek  <jakub@redhat.com>
35222         PR target/80310
35223         * tree-nvr.c: Include internal-fn.h.
35224         (pass_return_slot::execute): Ignore internal calls without
35225         direct optab.
35227 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
35228             Richard Biener  <rguenther@suse.de>
35230         PR c++/80297
35231         * genmatch.c (capture::gen_transform): For GENERIC unshare_expr
35232         captures used multiple times, except for the last use.
35233         * generic-match-head.c: Include gimplify.h.
35235 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
35237         PR tree-optimization/79390
35238         * target.h (struct noce_if_info): Declare.
35239         * targhooks.h (default_noce_conversion_profitable_p): Declare.
35240         * target.def (noce_conversion_profitable_p): New target hook.
35241         * ifcvt.h (struct noce_if_info): New type, moved from ...
35242         * ifcvt.c (struct noce_if_info): ... here.
35243         (noce_conversion_profitable_p): Renamed to ...
35244         (default_noce_conversion_profitable_p): ... this.  No longer
35245         static nor inline.
35246         (noce_try_store_flag_constants, noce_try_addcc,
35247         noce_try_store_flag_mask, noce_try_cmove, noce_try_cmove_arith,
35248         noce_convert_multiple_sets): Use targetm.noce_conversion_profitable_p
35249         instead of noce_conversion_profitable_p.
35250         * config/i386/i386.c: Include ifcvt.h.
35251         (ix86_option_override_internal): Don't override
35252         PARAM_MAX_RTL_IF_CONVERSION_INSNS default.
35253         (ix86_noce_conversion_profitable_p): New function.
35254         (TARGET_NOCE_CONVERSION_PROFITABLE_P): Redefine.
35255         * config/i386/x86-tune.def (X86_TUNE_ONE_IF_CONV_INSN): Adjust comment.
35256         * doc/tm.texi.in (TARGET_NOCE_CONVERSION_PROFITABLE_P): Add.
35257         * doc/tm.texi: Regenerated.
35259 2017-04-04  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35261         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Grammar
35262         correction.
35264 2017-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
35266         PR target/80307
35267         * config/arm/arm.c (thumb1_rtx_costs): Give a cost of 32
35268         instructions for small multiply cores.
35270 2017-04-04  Jeff Law  <law@redhat.com>
35272         * config/mips/mips.c (mips_multi_add): Zero initialize the newly
35273         added member.
35274         (mips_expand_vec_perm_const): Initialize elements in orig_perm
35275         that are not set by the loop over the elements.
35277 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
35279         PR target/80286
35280         * config/i386/i386.c (ix86_expand_args_builtin): If op has scalar
35281         int mode, convert_modes it to mode as unsigned, otherwise use
35282         lowpart_subreg to mode rather than SImode.
35283         * config/i386/sse.md (<mask_codefor>ashr<mode>3<mask_name>,
35284         ashr<mode>3, ashr<mode>3<mask_name>, <shift_insn><mode>3<mask_name>):
35285         Use DImode instead of SImode for the shift count operand.
35286         * config/i386/mmx.md (mmx_ashr<mode>3, mmx_<shift_insn><mode>3):
35287         Likewise.
35289 2017-04-04  Richard Biener  <rguenther@suse.de>
35291         PR middle-end/80281
35292         * match.pd (A + (-B) -> A - B): Make sure to preserve unsigned
35293         arithmetic done for the negate or the plus.  Simplify.
35294         (A - (-B) -> A + B): Likewise.
35295         * fold-const.c (split_tree): Make sure to not negate pointers.
35297 2017-04-04  Segher Boessenkool  <segher@kernel.crashing.org>
35299         PR rtl-optimization/60818
35300         * simplify-rtx.c (simplify_binary_operation_1): Do not replace
35301         a compare of comparisons with the thing compared if this results
35302         in a different machine mode.
35304 2017-04-03  Jonathan Wakely  <jwakely@redhat.com>
35306         * alias.c (base_alias_check): Fix typo in comment.
35307         * cgraph.h (class ipa_polymorphic_call_context): Likewise.
35308         * cgraphunit.c (symbol_table::compile): Likewise.
35309         * collect2.c (maybe_run_lto_and_relink): Likewise.
35310         * config/arm/arm.c (arm_thumb1_mi_thunk): Likewise.
35311         * config/avr/avr-arch.h (avr_arch_info_t): Likewise.
35312         * config/avr/avr.c (avr_map_op_t): Likewise.
35313         * config/cr16/cr16.h (DATA_ALIGNMENT): Likewise.
35314         * config/epiphany/epiphany.c (TARGET_ARG_PARTIAL_BYTES): Likewise.
35315         * config/epiphany/epiphany.md (movcc): Likewise.
35316         * config/i386/i386.c (legitimize_pe_coff_extern_decl): Likewise.
35317         * config/m68k/m68k.c (struct _sched_ib, m68k_sched_variable_issue):
35318         Likewise.
35319         * config/mips/mips.c (mips_save_restore_reg): Likewise.
35320         * config/rx/rx.c (rx_is_restricted_memory_address): Likewise.
35321         * config/s390/s390.c (Z10_EARLYLOAD_DISTANCE): Likewise.
35322         * config/sh/sh.c (sh_rtx_costs): Likewise.
35323         * fold-const.c (fold_truth_andor): Likewise.
35324         * genautomata.c (collapse_flag): Likewise.
35325         * gengtype.h (struct type::u::s): Likewise.
35326         * gensupport.c (has_subst_attribute, add_mnemonic_string): Likewise.
35327         * input.c (FORMAT_AMOUNT): Likewise.
35328         * ipa-cp.c (class ipcp_lattice, agg_replacements_to_vector)
35329         (known_aggs_to_agg_replacement_list): Likewise.
35330         * ipa-inline-analysis.c: Likewise.
35331         * ipa-inline.h (estimate_edge_time, estimate_edge_hints): Likewise.
35332         * ipa-polymorphic-call.c
35333         (ipa_polymorphic_call_context::restrict_to_inner_class): Likewise.
35334         * loop-unroll.c (analyze_insn_to_expand_var): Likewise.
35335         * lra.c (lra_optional_reload_pseudos, lra_subreg_reload_pseudos):
35336         Likewise.
35337         * modulo-sched.c (apply_reg_moves): Likewise.
35338         * omp-expand.c (build_omp_regions_1): Likewise.
35339         * trans-mem.c (struct tm_wrapper_hasher): Likewise.
35340         * tree-ssa-loop-ivopts.c (may_eliminate_iv): Likewise.
35341         * tree-ssa-loop-niter.c (maybe_lower_iteration_bound): Likewise.
35342         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
35343         * value-prof.c: Likewise.
35344         * var-tracking.c (val_reset): Likewise.
35346 2017-04-03  Richard Biener  <rguenther@suse.de>
35348         PR tree-optimization/80275
35349         * fold-const.c (split_address_to_core_and_offset): Handle
35350         POINTER_PLUS_EXPR.
35352 2017-04-03  Eric Botcazou  <ebotcazou@adacore.com>
35354         * tree-nested.c (get_descriptor_type): Make sure that the alignment of
35355         descriptors is at least equal to that of functions.
35357 2017-04-02  Uros Bizjak  <ubizjak@gmail.com>
35359         * config/i386/sse.md (movdi_to_sse): Add missing DONE.
35361 2017-04-02  Uros Bizjak  <ubizjak@gmail.com>
35363         PR target/80250
35364         * config/i386/sse.md (mov<IMOD4:mode>): Remove insn pattern.
35365         (mov<IMOD4:mode>): New expander.
35366         (*mov<IMOD4:mode>_internal): New insn and split pattern.
35368 2017-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
35370         PR rtl-optimization/79405
35371         * fwprop.c (propagations_left): New variable.
35372         (forward_propagate_into): Decrement it.
35373         (fwprop_init): Initialize it.
35374         (fw_prop): If the variable has reached zero, stop propagating.
35375         (fwprop_addr): Ditto.
35377 2017-03-31  Jakub Jelinek  <jakub@redhat.com>
35379         PR debug/79255
35380         * dwarf2out.c (decls_for_scope): If BLOCK_NONLOCALIZED_VAR is
35381         a FUNCTION_DECL, pass it as decl instead of origin to
35382         process_scope_var.
35384 2017-03-31  Alexander Monakov  <amonakov@ispras.ru>
35386         * config/nvptx/nvptx.c (nvptx_output_softstack_switch): Correct format
35387         string.
35389 2017-03-31  Pat Haugen  <pthaugen@us.ibm.com>
35391         PR target/80107
35392         * config/rs6000/rs6000.md (extendhi<mode>2): Add test for
35393         TARGET_VSX_SMALL_INTEGER.
35395 2017-03-31  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35397         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
35398         reference to the OpenPOWER 64-Bit ELF V2 ABI Specification.
35400 2017-03-31  Matthew Fortune  <matthew.fortune@imgtec.com>
35402         * config/mips/mips-msa.md (msa_vec_extract_<msafmt_f>): Update
35403         extraction from odd-numbered MSA register.
35405 2017-03-31  Jakub Jelinek  <jakub@redhat.com>
35407         PR middle-end/80173
35408         * expmed.c (store_bit_field_1): Don't attempt to create
35409         a word subreg out of hard registers wider than word if they
35410         have HARD_REGNO_NREGS of 1 for their mode.
35412         PR middle-end/80163
35413         * varasm.c (initializer_constant_valid_p_1): Disallow sign-extending
35414         conversions to integer types wider than word and pointer.
35416         PR debug/80025
35417         * cselib.h (rtx_equal_for_cselib_1): Add depth argument.
35418         (rtx_equal_for_cselib_p): Pass 0 to it.
35419         * cselib.c (cselib_hasher::equal): Likewise.
35420         (rtx_equal_for_cselib_1): Add depth argument.  If depth
35421         is 128, don't look up VALUE locs and punt.  Increment
35422         depth in recursive calls when walking VALUE locs.
35424 2017-03-31  Bernd Edlinger  <bernd.edlinger@hotmail.de>
35426         * gcov.c (md5sum_to_hex): Fix output of MD5 hex bytes.
35427         (make_gcov_file_name): Use the canonical path name for generating
35428         the MD5 value.
35429         (read_line): Fix handling of files with ascii null bytes.
35431 2017-03-30  Matthew Fortune  <matthew.fortune@imgtec.com>
35433         * config/mips/mips.c (mips_expand_vector_init): Create a const_vector
35434         to initialise a vector register instead
35435         of using a const_int.
35437 2017-03-30  Jakub Jelinek  <jakub@redhat.com>
35439         PR translation/80189
35440         * gimplify.c (omp_default_clause): Use %qs instead of %s in
35441         diagnostic messages.
35443 2017-03-30  Peter Bergner  <bergner@vnet.ibm.com>
35445         PR target/80246
35446         * config/rs6000/dfp.md (dfp_dxex_<mode>): Update mode of operand 0.
35447         (dfp_diex_<mode>): Update mode of operand 1.
35448         * doc/extend.texi (dxex, dxexq): Document change to return type.
35449         (diex, diexq): Document change to argument type.
35451 2017-03-30  Martin Jambor  <mjambor@suse.cz>
35453         PR ipa/77333
35454         * cgraph.h (cgraph_build_function_type_skip_args): Declare.
35455         * cgraph.c (redirect_call_stmt_to_callee): Set gimple fntype so that
35456         it reflects the signature changes performed at the callee side.
35457         * cgraphclones.c (build_function_type_skip_args): Make public, renamed
35458         to cgraph_build_function_type_skip_args.
35459         (build_function_decl_skip_args): Adjust call to the above function.
35461 2017-03-30  Jakub Jelinek  <jakub@redhat.com>
35463         PR target/80206
35464         * config/i386/sse.md
35465         (<extract_type>_vextract<shuffletype><extract_suf>_mask): Use
35466         register as dest whenever it is a MEM not rtx_equal_p to the
35467         corresponding dup operand, and when forcing into reg move the
35468         reg into the memory afterwards.
35469         (<extract_type_2>_vextract<shuffletype><extract_suf_2>_mask):
35470         Likewise.  Use <ssehalfvecmode> instead of <ssequartermode>
35471         for the force_reg mode.
35472         (avx512vl_vextractf128<mode>): Use register as dest either
35473         always when a MEM, or when it is a MEM not rtx_equal_p to the
35474         corresponding dup operand, or even not when it is a CONST_VECTOR
35475         depending on the mode and lo vs. hi.
35476         (avx512dq_vextract<shuffletype>64x2_1_maskm): Remove extraneous
35477         parens.
35478         (avx512f_vextract<shuffletype>32x4_1_maskm): Likewise.
35479         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name>):
35480         Likewise.  Require that operands[2] is even.
35481         (<mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name>):
35482         Remove extraneous parens.  Require that operands[2] is a multiple
35483         of 4.
35484         (vec_extract_lo_<mode><mask_name>): Don't bother testing if
35485         operands[0] is a MEM if <mask_applied>, the predicates/constraints
35486         disallow memory then.
35488 2017-03-30  Richard Biener  <rguenther@suse.de>
35490         PR tree-optimization/77498
35491         * tree-ssa-pre.c (phi_translate_1): Do not allow simplifications
35492         to non-constants over backedges.
35494 2017-03-29  Segher Boessenkool  <segher@kernel.crashing.org>
35496         PR rtl-optimization/80233
35497         * combine.c (combine_instructions): Only take NONDEBUG_INSN_P insns
35498         as last_combined_insn.  Do not test for BARRIER_P separately.
35500 2017-03-29  Andreas Schwab  <schwab@suse.de>
35502         PR ada/80146
35503         * calls.c (prepare_call_address): Convert funexp to Pmode before
35504         copying to temp reg.
35506 2017-03-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35508         PR tree-optimization/80158
35509         * gimple-ssa-strength-reduction.c (replace_mult_candidate):
35510         Handle possible future case of more than one alternate
35511         interpretation.
35512         (replace_rhs_if_not_dup): Likewise.
35513         (replace_one_candidate): Likewise.
35515 2017-03-28  Vladimir Makarov  <vmakarov@redhat.com>
35517         PR rtl-optimization/80193
35518         * ira.c (ira): Do not check allocation for LRA.
35520 2017-03-28  Alexander Monakov  <amonakov@ispras.ru>
35522         * config/nvptx/nvptx-protos.h (nvptx_output_simt_enter): Declare.
35523         (nvptx_output_simt_exit): Declare.
35524         * config/nvptx/nvptx.c (nvptx_init_unisimt_predicate): Use
35525         cfun->machine->unisimt_location.  Handle NULL unisimt_predicate.
35526         (init_softstack_frame): Move initialization of crtl->is_leaf to...
35527         (nvptx_declare_function_name): ...here.  Emit declaration of local
35528         memory space buffer for omp_simt_enter insn.
35529         (nvptx_output_unisimt_switch): New.
35530         (nvptx_output_softstack_switch): New.
35531         (nvptx_output_simt_enter): New.
35532         (nvptx_output_simt_exit): New.
35533         * config/nvptx/nvptx.h (struct machine_function): New fields
35534         has_simtreg, unisimt_location, simt_stack_size, simt_stack_align.
35535         * config/nvptx/nvptx.md (UNSPECV_SIMT_ENTER): New unspec.
35536         (UNSPECV_SIMT_EXIT): Ditto.
35537         (omp_simt_enter_insn): New insn.
35538         (omp_simt_enter): New expansion.
35539         (omp_simt_exit): New insn.
35540         * config/nvptx/nvptx.opt (msoft-stack-reserve-local): New option.
35542         * internal-fn.c (expand_GOMP_SIMT_ENTER): New.
35543         (expand_GOMP_SIMT_ENTER_ALLOC): New.
35544         (expand_GOMP_SIMT_EXIT): New.
35545         * internal-fn.def (GOMP_SIMT_ENTER): New internal function.
35546         (GOMP_SIMT_ENTER_ALLOC): Ditto.
35547         (GOMP_SIMT_EXIT): Ditto.
35548         * target-insns.def (omp_simt_enter): New insn.
35549         (omp_simt_exit): Ditto.
35550         * omp-low.c (struct omplow_simd_context): New fields simt_eargs,
35551         simt_dlist.
35552         (lower_rec_simd_input_clauses): Implement SIMT privatization.
35553         (lower_rec_input_clauses): Likewise.
35554         (lower_lastprivate_clauses): Handle SIMT privatization.
35556         * omp-offload.c: Include langhooks.h, tree-nested.h, stor-layout.h.
35557         (ompdevlow_adjust_simt_enter): New.
35558         (find_simtpriv_var_op): New.
35559         (execute_omp_device_lower): Handle IFN_GOMP_SIMT_ENTER,
35560         IFN_GOMP_SIMT_ENTER_ALLOC, IFN_GOMP_SIMT_EXIT.
35562         * tree-inline.h (struct copy_body_data): New field dst_simt_vars.
35563         * tree-inline.c (expand_call_inline): Handle SIMT privatization.
35564         (copy_decl_for_dup_finish): Ditto.
35566         * tree-ssa.c (execute_update_addresses_taken): Handle GOMP_SIMT_ENTER.
35568 2017-03-28  Uros Bizjak  <ubizjak@gmail.com>
35570         PR target/53383
35571         * config/i386/i386.c (ix86_option_override_internal): Always
35572         allow -mpreferred-stack-boundary=3 for 64-bit targets.
35574 2017-03-28  Bin Cheng  <bin.cheng@arm.com>
35576         * tree-vect-loop.c (optimize_mask_stores): Add bb to the right loop.
35578 2017-03-28  Bin Cheng  <bin.cheng@arm.com>
35580         * tree-vect-loop-manip.c (slpeel_add_loop_guard): New param and
35581         mark new edge's irreducible flag accordign to it.
35582         (vect_do_peeling): Check loop preheader edge's irreducible flag
35583         and pass it to function slpeel_add_loop_guard.
35585 2017-03-28  Richard Sandiford  <richard.sandiford@arm.com>
35587         PR tree-optimization/80218
35588         * tree-call-cdce.c (shrink_wrap_one_built_in_call_with_conds):
35589         Update block frequencies and counts.
35591 2017-03-28  Richard Biener  <rguenther@suse.de>
35593         PR tree-optimization/78644
35594         * tree-ssa-ccp.c (evaluate_stmt): When we may not use the value
35595         of a simplification result we may not use it at all.
35597 2017-03-28  Richard Biener  <rguenther@suse.de>
35599         PR ipa/80205
35600         * tree-inline.c (copy_phis_for_bb): Do not create PHI node
35601         without arguments, generate default definition of a SSA name.
35603 2017-03-28  Richard Biener  <rguenther@suse.de>
35605         PR middle-end/80222
35606         * gimple-fold.c (gimple_fold_indirect_ref): Do not touch
35607         TYPE_REF_CAN_ALIAS_ALL references.
35608         * fold-const.c (fold_indirect_ref_1): Likewise.
35610 2017-03-28  Martin Liska  <mliska@suse.cz>
35612         PR ipa/80104
35613         * cgraphunit.c (cgraph_node::expand_thunk): Mark argument of a
35614         thunk call as DECL_GIMPLE_REG_P when vector or complex type.
35616 2017-03-28  Claudiu Zissulescu  <claziss@synopsys.com>
35617             Thomas Petazzoni  <thomas.petazzoni@free-electrons.com>
35619         * config/arc/arc.h (CPP_SPEC): Add subtarget_cpp_spec.
35620         (EXTRA_SPECS): Define.
35621         (SUBTARGET_EXTRA_SPECS): Likewise.
35622         (SUBTARGET_CPP_SPEC): Likewise.
35623         * config/arc/elf.h (EXTRA_SPECS): Renamed to
35624         SUBTARGET_EXTRA_SPECS.
35625         * config/arc/linux.h (SUBTARGET_CPP_SPEC): Define.
35627 2017-03-28  Claudiu Zissulescu  <claziss@synopsys.com>
35629         * config/arc/simdext.md (vst64_insn): Update pattern.
35630         (vld32wh_insn): Likewise.
35631         (vld32wl_insn): Likewise.
35632         (vld64_insn): Likewise.
35633         (vld32_insn): Likewise.
35635 2017-03-28  Marek Polacek  <polacek@redhat.com>
35637         PR sanitizer/80067
35638         * fold-const.c (fold_comparison): Use protected_set_expr_location
35639         instead of SET_EXPR_LOCATION.
35641 2017-03-28  Markus Trippelsdorf  <markus@trippelsdorf.de>
35643         * tree.c (add_expr): Avoid name lookup warning.
35645 2017-03-27  Jeff Law  <law@redhat.com>
35647         PR tree-optimization/80216
35648         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Fix typo in
35649         function name.  Limit recursion depth.
35650         (record_temporary_equivalences): Corresponding changes.
35652 2017-03-27  Jonathan Wakely  <jwakely@redhat.com>
35654         * doc/invoke.texi (-Wno-narrowing): Reorder so default behavior is
35655         covered first.
35657 2017-03-27  Jakub Jelinek  <jakub@redhat.com>
35659         PR target/80102
35660         * reg-notes.def (REG_CFA_NOTE): Define.  Use it for CFA related
35661         notes.
35662         * cfgcleanup.c (reg_note_cfa_p): New array.
35663         (insns_have_identical_cfa_notes): New function.
35664         (old_insns_match_p): Don't cross-jump in between /f
35665         and non-/f instructions.  If both i1 and i2 are frame related,
35666         verify all CFA notes, their order and content.
35668 2017-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
35670         PR target/78543
35671         * config/rs6000/rs6000.md (bswaphi2_extenddi): Combine bswap
35672         HImode and SImode with zero extend to DImode to one insn.
35673         (bswap<mode>2_extenddi): Likewise.
35674         (bswapsi2_extenddi): Likewise.
35675         (bswaphi2_extendsi): Likewise.
35676         (bswaphi2): Combine bswap HImode and SImode into one insn.
35677         Separate memory insns from swapping register.
35678         (bswapsi2): Likewise.
35679         (bswap<mode>2): Likewise.
35680         (bswaphi2_internal): Delete, no longer used.
35681         (bswapsi2_internal): Likewise.
35682         (bswap<mode>2_load): Split bswap HImode/SImode into separate load,
35683         store, and gpr<-gpr swap insns.
35684         (bswap<mode>2_store): Likewise.
35685         (bswaphi2_reg): Register only splitter, combine with the splitter.
35686         (bswaphi2 splitter): Likewise.
35687         (bswapsi2_reg): Likewise.
35688         (bswapsi2 splitter): Likewise.
35689         (bswapdi2): If we have the LDBRX and STDBRX instructions, split
35690         the insns into load, store, and register/register insns.
35691         (bswapdi2_ldbrx): Likewise.
35692         (bswapdi2_load): Likewise.
35693         (bswapdi2_store): Likewise.
35694         (bswapdi2_reg): Likewise.
35696 2017-03-27  Gunther Nikl  <gnikl@users.sourceforge.net>
35698         * system.h (HAVE_DESIGNATED_INITIALIZERS): Fix non C++ case.
35699         (HAVE_DESIGNATED_UNION_INITIALIZERS): Likewise.
35701 2017-03-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
35703         PR target/80103
35704         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Edit and
35705         add comments.
35706         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
35707         special handling for target option conflicts between dform
35708         options (-mpower9-dform, -mpower9-dform-vector,
35709         -mpower9-dform-scalar) and -mno-direct-move.
35711 2017-03-27  Richard Biener  <rguenther@suse.de>
35713         PR tree-optimization/80181
35714         * tree-ssa-ccp.c (likely_value): UNDEFINED ^ X is UNDEFINED.
35716 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35718         * config/arc/predicates.md (move_double_src_operand): Replace the
35719         call to move_double_src_operand with a call to address_operand.
35721 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35723         * config/arc/elf.h (ARGET_ARC_TP_REGNO_DEFAULT): Define.
35724         * config/arc/linux.h (ARGET_ARC_TP_REGNO_DEFAULT): Likewise.
35725         * config/arc/arc.opt (mtp-regno): Use ARGET_ARC_TP_REGNO_DEFAULT.
35727 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35729         * config/arc/predicates.md (long_immediate_loadstore_operand):
35730         Consider scaled addresses cases.
35732 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35734         * config/arc/arc.c (arc_epilogue_uses): BLINK should be also
35735         restored when in interrupt.
35736         * config/arc/arc.md (simple_return): ARCv2 rtie instruction
35737         doesn't have delay slot.
35739 2017-03-27  Richard Biener  <rguenther@suse.de>
35741         PR ipa/79776
35742         * tree-ssa-structalias.c (associate_varinfo_to_alias): Skip
35743         inlined thunk clones.
35745 2017-03-27  Jakub Jelinek  <jakub@redhat.com>
35747         PR sanitizer/80168
35748         * asan.c (instrument_derefs): Copy over last operand from
35749         original COMPONENT_REF to the new COMPONENT_REF with
35750         DECL_BIT_FIELD_REPRESENTATIVE.
35751         * ubsan.c (instrument_object_size): Likewise.
35753 2017-03-27  Richard Biener  <rguenther@suse.de>
35755         PR tree-optimization/80170
35756         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Make
35757         sure DR/SCEV didnt fold in constants we do not see when looking
35758         at the reference base alignment.
35760 2017-03-27  Richard Biener  <rguenther@suse.de>
35762         PR middle-end/80171
35763         * gimple-fold.c (fold_ctor_reference): Properly guard against
35764         NULL return value from canonicalize_constructor_val.
35766 2017-03-25  Uros Bizjak  <ubizjak@gmail.com>
35768         PR target/80180
35769         * config/i386/i386.c (ix86_expand_builtin)
35770         <IX86_BUILTIN_RDSEED{16,32,64}_STEP>: Do not expand arg0 between
35771         flags reg setting and flags reg using instructions.
35772         <IX86_BUILTIN_RDRAND{16,32,64}_STEP>: Ditto.  Use non-flags reg
35773         clobbering instructions to zero extend op2.
35775 2017-03-25  Gerald Pfeifer  <gerald@pfeifer.com>
35777         * doc/install.texi (Configuration) <--with-aix-soname>:
35778         Update link to AIX ld.
35780 2017-03-25  Bernd Schmidt  <bschmidt@redhat.com>
35782         PR rtl-optimization/80160
35783         PR rtl-optimization/80159
35784         * lra-assigns.c (must_not_spill_p): Tighten new test to also take
35785         reg_alternate_class into account.
35787 2017-03-24  Vladimir Makarov  <vmakarov@redhat.com>
35789         PR target/80148
35790         * lra-assigns.c (assign_by_spills): Add spilled non-reload pseudos
35791         to consider in curr_insn_transform.
35793 2017-03-24  Jakub Jelinek  <jakub@redhat.com>
35795         * genrecog.c (validate_pattern): Add VEC_SELECT validation.
35796         * genmodes.c (emit_min_insn_modes_c): Call emit_mode_nunits
35797         and emit_mode_inner.
35799 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35801         * config/s390/s390-builtins.def: Add VXE builtins.  Add a flags
35802         argument to the overloaded builtin variants.  Use the new flag to
35803         deprecate certain builtin variants.
35804         * config/s390/s390-builtin-types.def: Add new builtin types.
35805         * config/s390/s390-builtins.h: Support new flags field for
35806         overloaded builtins.
35807         * config/s390/s390-c.c (OB_DEF_VAR): New flags field.
35808         (s390_macro_to_expand): Enable vector float data type.
35809         (s390_cpu_cpp_builtins_internal): Indicate support of the new
35810         builtins by incrementing the __VEC__ version number.
35811         (s390_expand_overloaded_builtin): Support expansion of vec_xl and
35812         vec_xst.
35813         (s390_resolve_overloaded_builtin): Emit error messages depending
35814         on the builtin flags.
35815         * config/s390/s390.c (s390_expand_builtin): Support additional
35816         flags argument.  Change error message to match the messages
35817         emitted in s390-c.c.
35818         * config/s390/s390.md: New UNSPEC_* constants.
35819         (op_type): Add new instruction types.
35820         * config/s390/vecintrin.h: Add new builtins and test data class
35821         constants.
35822         * config/s390/vx-builtins.md (V_HW_32_64): Add V4SF.
35823         (V_HW_4, VEC_HW, VECF_HW): New mode iterators.
35824         (VEC_INEXACT, VEC_NOINEXACT): New constants.
35825         ("vec_splats<mode>", "vec_insert<mode>", "vec_promote<mode>")
35826         ("vec_insert_and_zero<mode>", "vec_mergeh<mode>")
35827         ("vec_mergel<mode>"): V_HW -> VEC_HW.
35829         ("vlrlrv16qi", "vstrlrv16qi", "vbpermv16qi", "vec_msumv2di")
35830         ("vmslg", "*vftci<mode>_cconly", "vftci<mode>_intcconly")
35831         ("*vftci<mode>", "vftci<mode>_intcc", "vec_double_s64")
35832         ("vec_double_u64", "vfmin<mode>", "vfmax<mode>"): New definition.
35834         ("and_av2df3", "and_cv2df3", "vec_andc_av2df3")
35835         ("vec_andc_cv2df3", "xor_av2df3", "xor_cv2df3", "vec_nor_av2df3")
35836         ("vec_nor_cv2df3", "ior_av2df3", "ior_cv2df3", "vec_nabs")
35837         ("*vftcidb", "*vftcidb_cconly", "vftcidb"): Remove definition.
35839         ("vec_all_<fpcmpcc:code>v2df", "vec_any_<fpcmpcc:code>v2df")
35840         ("vec_scatter_elementv4si_DI", "vec_cmp<fpcmp:code>v2df")
35841         ("vec_di_to_df_s64", "vec_di_to_df_u64", "vec_df_to_di_u64")
35842         ("vfidb", "*vldeb", "*vledb", "*vec_cmp<insn_cmp>v2df_cconly")
35843         ("vec_cmpeqv2df_cc", "vec_cmpeqv2df_cc", "vec_cmphv2df_cc")
35844         ("vec_cmphev2df_cc", "*vec_cmpeqv2df_cc")
35845         ("*vec_cmphv2df_cc", "*vec_cmphev2df_cc"): Enable new modes as ...
35847         ("vec_all_<fpcmpcc:code><mode>", "vec_any_<fpcmpcc:code><mode>")
35848         ("vec_scatter_element<V_HW_4:mode>_DI")
35849         ("vec_cmp<fpcmp:code><mode>", "vcdgb", "vcdlgb", "vclgdb")
35850         ("vec_fpint<mode>", "vflls")
35851         ("vflrd", "*vec_cmp<insn_cmp><mode>_cconly", "vec_cmpeq<mode>_cc")
35852         ("vec_cmpeq<mode>_cc", "vec_cmph<mode>_cc", "vec_cmphe<mode>_cc")
35853         ("*vec_cmpeq<mode>_cc", "*vec_cmph<mode>_cc")
35854         ("*vec_cmphe<mode>_cc"): ... these.
35856         ("vec_ctd_s64", "vec_ctsl", "vec_ctul", "vec_st2f"): Use rounding
35857         mode constant instead of magic value.
35859 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35861         * config/s390/s390.c (s390_expand_vec_compare): Support other
35862         vector floating point modes than just V2DF.
35863         (s390_expand_vcond): Likewise.
35864         (s390_hard_regno_mode_ok): Allow SFmode values in VRs.
35865         (s390_cannot_change_mode_class): Prevent mode changes between TF
35866         and V1TF in vector registers.
35867         * config/s390/s390.md (DF, SF): New mode attributes.
35868         ("*cmp<mode>_ccs", "add<mode>3", "sub<mode>3", "mul<mode>3")
35869         ("fma<mode>4", "fms<mode>4", "div<mode>3", "*neg<mode>2"): Add
35870         SFmode support for VRs.
35871         * config/s390/vector.md (V_HW, V_HW2, VT_HW, ti*, nonvec): Add new
35872         vector fp modes.
35873         (VFT, VF_HW): New mode iterators.
35874         (vw, sdx): New mode attributes.
35875         ("addv2df3", "subv2df3", "mulv2df3", "divv2df3", "sqrtv2df2")
35876         ("fmav2df4","fmsv2df4", "negv2df2", "absv2df2", "*negabsv2df2")
35877         ("smaxv2df3", "sminv2df3", "*vec_cmp<VFCMP_HW_OP:code>v2df_nocc")
35878         ("vec_cmpuneqv2df", "vec_cmpltgtv2df", "vec_orderedv2df")
35879         ("vec_unorderedv2df"): Adjust the v2df only patterns to support
35880         also the new vector floating point modes.  Renaming to ...
35882         ("add<mode>3", "sub<mode>3", "mul<mode>3", "div<mode>3")
35883         ("sqrt<mode>2", "fma<mode>4", "fms<mode>4", "neg<mode>2")
35884         ("abs<mode>2", "negabs<mode>2", "smax<mode>3")
35885         ("smin<mode>3", "*vec_cmp<VFCMP_HW_OP:code><mode>_nocc")
35886         ("vec_cmpuneq<mode>", "vec_cmpltgt<mode>", "vec_ordered<mode>")
35887         ("vec_unordered<mode>"): ... these.
35889         ("neg_fma<mode>4", "neg_fms<mode>4", "*smax<mode>3_vxe")
35890         ("*smin<mode>3_vxe", "*sminv2df3_vx", "*vec_extendv4sf")
35891         ("*vec_extendv2df"): New insn definitions.
35893 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35895         * config/s390/s390.md ("*adddi3_sign", "*subdi3_sign", "mulditi3")
35896         ("mulditi3_2", "*muldi3_sign"): New patterns.
35897         ("muldi3", "*muldi3", "mulsi3", "*mulsi3"): Add an expander and
35898         rename the pattern definition.
35900 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35902         * config/s390/s390.md ("indirect_jump"): Turn insn definition into
35903         expander.
35904         ("*indirect_jump", "*indirect2_jump"): New pattern definitions.
35906 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35908         * config/s390/s390.c (s390_expand_vec_init): Use vllezl
35909         instruction if possible.
35910         * config/s390/vector.md (vec_halfnumelts): New mode
35911         attribute.
35912         ("*vec_vllezlf<mode>"): New pattern.
35914 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35916         * config/s390/vector.md ("popcountv16qi2", "popcountv8hi2")
35917         ("popcountv4si2", "popcountv2di2"): Rename to ...
35918         ("popcount<mode>2", "popcountv8hi2_vx", "popcountv4si2_vx")
35919         ("popcountv2di2_vx"): ... these and add !TARGET_VXE to the
35920         condition.
35921         ("popcount<mode>2_vxe"): New pattern.
35923 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35925         * common/config/s390/s390-common.c (processor_flags_table): Add
35926         arch12.
35927         * config.gcc: Add arch12.
35928         * config/s390/driver-native.c (s390_host_detect_local_cpu):
35929         Default to arch12 for unknown CPU model numbers.
35930         * config/s390/s390-builtins.def: Add B_VXE builtin flag.
35931         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Adjust
35932         PROCESSOR_max sanity check.
35933         * config/s390/s390-opts.h (enum processor_type): Add
35934         PROCESSOR_ARCH12.
35935         * config/s390/s390.c (processor_table): Add arch12.
35936         (s390_expand_builtin): Add check for B_VXE flag.
35937         (s390_issue_rate): Add PROCESSOR_ARCH12.
35938         (s390_get_sched_attrmask): Likewise.
35939         (s390_get_unit_mask): Likewise.
35940         (s390_sched_score): Enable z13 scheduling for arch12.
35941         (s390_sched_reorder): Likewise.
35942         (s390_sched_variable_issue): Likewise.
35943         * config/s390/s390.h (enum processor_flags): Add PF_ARCH12 and
35944         PF_VXE.
35945         (s390_tune_attr): Use z13 scheduling also for arch12.
35946         (TARGET_CPU_ARCH12, TARGET_CPU_ARCH12_P, TARGET_CPU_VXE)
35947         (TARGET_CPU_VXE_P, TARGET_ARCH12, TARGET_ARCH12_P, TARGET_VXE)
35948         (TARGET_VXE_P): New macros.
35949         * config/s390/s390.md: Add arch12 to cpu attribute.  Add arch12
35950         and vxe to cpu_facility.  Add arch12 and vxe to enabled attribute.
35951         * config/s390/s390.opt: Add arch12 as processor_type.
35953 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35955         * config/s390/s390.md
35956         ("fixuns_truncdddi2", "fixuns_trunctddi2")
35957         ("fixuns_trunc<BFP:mode><GPR:mode>2"): Merge into ...
35958         ("fixuns_trunc<FP:mode><GPR:mode>2"): New expander.
35960         ("fixuns_trunc<BFP:mode><GPR:mode>2", "fixuns_trunc<mode>si2"):
35961         Rename expanders to ...
35963         ("fixuns_trunc<BFP:mode><GPR:mode>2_emu")
35964         ("fixuns_truncdddi2_emu"): ... these.
35966         ("fixuns_trunc<mode>si2_emu"): New expander.
35968         ("*fixuns_truncdfdi2_z13"): Rename to ...
35969         ("*fixuns_truncdfdi2_vx"): ... this.
35971 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35973         * config/s390/2964.md: Remove the single element vector compare
35974         instructions which are no longer used.
35975         * config/s390/s390.c (s390_select_ccmode): Remove handling of
35976         vector CCmodes.
35977         (s390_canonicalize_comparison): Remove handling of DFmode
35978         compares.
35979         (s390_expand_vec_compare_scalar): Remove function.
35980         (s390_emit_compare): Don't call s390_expand_vec_compare_scalar.
35981         * config/s390/s390.md ("*vec_cmp<insn_cmp>df_cconly"): Remove
35982         pattern.
35983         ("*cmp<mode>_ccs"): Add wfcdb instruction.
35985 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35987         * config/s390/s390.md ("mov<mode>_64dfp" DD_DF): Use vleig for loading a
35988         FP zero.
35989         ("*mov<mode>_64" DD_DF): Remove the vector instructions. These
35990         will anyway by matched by mov<mode>_64dfp.
35992 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35994         * config/s390/s390.md ("mov<mode>" SD_SF): Change vleg/vsteg to
35995         vlef/vstef.  Add missing operand to vleif.
35997 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35999         * config/s390/s390.c (s390_expand_vec_init): Enable vector load
36000         pair for all vector types with 64 bit elements.
36001         * config/s390/vx-builtins.md (V_HW_64): Move mode iterator to ...
36002         * config/s390/vector.md (V_HW_64): ... here.
36003         (V_128_NOSINGLE): New mode iterator.
36004         ("vec_init<V_HW:mode>"): Use V_128 as mode iterator.
36005         ("*vec_splat<mode>"): Use V_128_NOSINGLE mode iterator.
36006         ("*vec_tf_to_v1tf", "*vec_ti_to_v1ti"): New pattern definitions.
36007         ("*vec_load_pairv2di"): Change to ...
36008         ("*vec_load_pair<mode>"): ... this one.
36010 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
36012         * config/s390/constraints.md: Add comments.
36013         (jKK): Reject element sizes > 8 bytes.
36014         * config/s390/s390.c (s390_split_ok_p): Enable splitting also for
36015         s_operands.
36016         * config/s390/s390.md: Add the s_operand checks formerly in
36017         s390_split_ok_p to various splitters where they are still
36018         required.
36019         * config/s390/vector.md ("mov<mode>" V_128): Add GPR alternatives
36020         for 128 bit vectors.  Plus two splitters.
36022 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
36024         * config/s390/s390.md: Rename the cpu facilty vec to vx throughout
36025         the file.
36027 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
36029         PR target/79893
36030         * config/s390/s390-c.c (s390_adjust_builtin_arglist): Issue an
36031         error if the boundary argument is not constant.
36033 2017-03-24  Jakub Jelinek  <jakub@redhat.com>
36035         PR rtl-optimization/80112
36036         * loop-doloop.c (doloop_condition_get): Don't check condition
36037         if cmp isn't SET with IF_THEN_ELSE src.
36039 2017-03-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36041         PR tree-optimization/80158
36042         * gimple-ssa-strength-reduction.c (replace_mult_candidate): When
36043         replacing a candidate statement, also replace it for the
36044         candidate's alternate interpretation.
36045         (replace_rhs_if_not_dup): Likewise.
36046         (replace_one_candidate): Likewise.
36048 2017-03-24  Richard Biener  <rguenther@suse.de>
36050         PR tree-optimization/80167
36051         * graphite-isl-ast-to-gimple.c
36052         (translate_isl_ast_to_gimple::is_valid_rename): Handle default-defs
36053         properly.
36054         (translate_isl_ast_to_gimple::get_rename): Likewise.
36056 2017-03-23  Kelvin Nilsen  <kelvin@gcc.gnu.org>
36058         * config/rs6000/rs6000.c (rs6000_option_override_internal): Change
36059         handling of certain combinations of target options, including the
36060         combinations -mpower8-vector vs. -mno-vsx, -mpower9-vector vs.
36061         -mno-power8-vector, and -mpower9_dform vs. -mno-power9-vector.
36063 2017-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36065         PR target/71436
36066         * config/arm/arm.md (*load_multiple): Add reload_completed to
36067         matching condition.
36069 2017-03-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36070             Richard Biener  <rguenth@suse.de>
36072         PR tree-optimization/79908
36073         PR tree-optimization/80136
36074         * tree-stdarg.c (expand_ifn_va_arg_1): For a VA_ARG whose LHS has
36075         been cast away, gimplify_and_add suffices.
36077 2017-03-23  Markus Trippelsdorf  <markus@trippelsdorf.de>
36079         * tree-vrp.c (identify_jump_threads): Delete avail_exprs.
36081 2017-03-23  Richard Biener  <rguenther@suse.de>
36083         PR tree-optimization/80032
36084         * gimplify.c (gimple_push_cleanup): Forced unconditional
36085         cleanups still have to go to the conditional_cleanups
36086         sequence.
36088 2017-03-22  Jakub Jelinek  <jakub@redhat.com>
36090         PR tree-optimization/80072
36091         * tree-ssa-reassoc.c (struct operand_entry): Change id field type
36092         to unsigned int.
36093         (next_operand_entry_id): Change type to unsigned int.
36094         (sort_by_operand_rank): Make sure to return the right return value
36095         even if unsigned fields are bigger than INT_MAX.
36096         (struct oecount): Change cnt and id type to unsigned int.
36097         (oecount_hasher::equal): Formatting fix.
36098         (oecount_cmp): Make sure to return the right return value
36099         even if unsigned fields are bigger than INT_MAX.
36100         (undistribute_ops_list): Change next_oecount_id type to unsigned int.
36102         PR c++/80129
36103         * gimplify.c (gimplify_modify_expr_rhs) <case COND_EXPR>: Clear
36104         TREE_READONLY on result if writing it more than once.
36106         PR sanitizer/80110
36107         * doc/invoke.texi (-fsanitize=thread): Document that with
36108         -fnon-call-exceptions atomics are not able to throw
36109         exceptions.
36111         PR sanitizer/80110
36112         * tsan.c: Include tree-eh.h.
36113         (instrument_builtin_call): Call maybe_clean_eh_stmt or
36114         maybe_clean_or_replace_eh_stmt where needed.
36115         (instrument_memory_accesses): Add cfg_changed argument.
36116         Call gimple_purge_dead_eh_edges on each block and set *cfg_changed
36117         if it returned true.
36118         (tsan_pass): Adjust caller.  Return TODO_cleanup_cfg if cfg_changed.
36120         PR rtl-optimization/63191
36121         * config/i386/i386.c (ix86_delegitimize_address): Turn into small
36122         wrapper function, moved the whole old content into ...
36123         (ix86_delegitimize_address_1): ... this.  New inline function.
36124         (ix86_find_base_term): Use ix86_delegitimize_address_1 with
36125         true as last argument instead of ix86_delegitimize_address.
36127 2017-03-22  Wilco Dijkstra  <wdijkstr@arm.com>
36129         * config/aarch64/aarch64.c (generic_branch_cost): Copy
36130         cortexa57_branch_cost.
36132 2017-03-22  Wilco Dijkstra  <wdijkstr@arm.com>
36134         * config/aarch64/aarch64.c (generic_tunings): Add AES fusion.
36136 2017-03-21  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
36138         PR target/80123
36139         * doc/md.texi (Constraints): Document wA constraint.
36140         * config/rs6000/constraints.md (wA): New.
36141         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Add wA reg_class.
36142         (rs6000_init_hard_regno_mode_ok): Init wA constraint.
36143         * config/rs6000/rs6000.h (RS6000_CONSTRAINT_wA): New.
36144         * config/rs6000/vsx.md (vsx_splat_<mode>): Use wA constraint.
36146 2017-03-22  Cesar Philippidis  <cesar@codesourcery.com>
36148         PR c++/80029
36149         * gimplify.c (is_oacc_declared): New function.
36150         (oacc_default_clause): Use it to set default flags for acc declared
36151         variables inside parallel regions.
36152         (gimplify_scan_omp_clauses): Strip firstprivate pointers for acc
36153         declared variables.
36154         (gimplify_oacc_declare): Gimplify the declare clauses.  Add the
36155         declare attribute to any decl as necessary.
36157 2017-03-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
36159         PR target/80082
36160         * config/arm/arm-isa.h (isa_bit_lpae): New feature bit.
36161         (ISA_ARMv7ve): Add isa_bit_lpae to the definition.
36162         * config/arm/arm-protos.h (arm_arch7ve): Rename into ...
36163         (arm_arch_lpae): This.
36164         * config/arm/arm.c (arm_arch7ve): Rename into ...
36165         (arm_arch_lpae): This.  Define it in term of isa_bit_lpae.
36166         * config/arm/arm.h (TARGET_HAVE_LPAE): Redefine in term of
36167         arm_arch_lpae.
36169 2017-03-22  Martin Liska  <mliska@suse.cz>
36171         PR target/79906
36172         * config/rs6000/rs6000.c (rs6000_inner_target_options): Show
36173         error message instead of an ICE.
36175 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36177         * doc/extend.texi (6.11 Additional Floating Types): Revise.
36179 2017-03-21  Kelvin Nilsen  <kelvin@gcc.gnu.org>
36181         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Add
36182         comments.
36183         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
36184         comments.
36186 2017-03-21  Martin Sebor  <msebor@redhat.com>
36188         * doc/extend.texi: Use "cannot" instead of "can't."
36189         * doc/hostconfig.texi: Same.
36190         * doc/install.texi: Same.
36191         * doc/invoke.texi: Same.
36192         * doc/loop.texi: Same.
36193         * doc/md.texi: Same.
36194         * doc/objc.texi: Same.
36195         * doc/rtl.texi: Same.
36196         * doc/tm.texi: Same.
36197         * doc/tm.texi.in: Same.
36198         * doc/trouble.texi: Same.
36200 2017-03-21  Alexandre Oliva  <aoliva@redhat.com>
36202         PR debug/63238
36203         * dwarf2out.c (struct checksum_attributes): Add at_alignment.
36204         (collect_checksum_attributes): Set it.
36205         (die_checksum_ordered): Use it.
36207 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36209         PR tree-optimization/79908
36210         * tree-stdarg.c (expand_ifn_va_arg_1): Revert the following
36211         change: For a VA_ARG whose LHS has been cast away, use
36212         force_gimple_operand to construct the side effects.
36214 2017-03-21  David Malcolm  <dmalcolm@redhat.com>
36216         PR translation/80001
36217         * omp-offload.c (oacc_loop_fixed_partitions): Make diagnostics
36218         more amenable to translation.
36219         (oacc_loop_auto_partitions): Likewise.
36221 2017-03-21  Marek Polacek  <polacek@redhat.com>
36222             Martin Sebor  <msebor@redhat.com>
36224         PR tree-optimization/80109
36225         * gimple-ssa-warn-alloca.c (alloca_call_type): Only call get_range_info
36226         on INTEGRAL_TYPE_P.
36228 2017-03-21  Jakub Jelinek  <jakub@redhat.com>
36229             Segher Boessenkool  <segher@kernel.crashing.org>
36231         PR target/80125
36232         * combine.c (can_combine_p): Revert the 2017-03-20 change, only
36233         check reg_used_between_p between insn and one of succ or succ2
36234         depending on if succ is artificial insn not inserted into insn
36235         stream.
36237 2017-03-21  Martin Liska  <mliska@suse.cz>
36239         PR gcov-profile/80081
36240         * Makefile.in: Add gcov-dump and fix installation of gcov-tool.
36241         * doc/gcc.texi: Include gcov-dump stuff.
36242         * doc/gcov-dump.texi: New file.
36244 2017-03-21  Toma Tabacu  <toma.tabacu@imgtec.com>
36246         PR rtl-optimization/79150
36247         * config/mips/mips.c (mips_block_move_loop): Emit a NOP after the
36248         conditional jump, if the jump is the last insn of the loop.
36250 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36251             Richard Biener  <rguenth@suse.de>
36253         PR tree-optimization/79908
36254         * tree-stdarg.c (expand_ifn_va_arg_1): For a VA_ARG whose LHS has
36255         been cast away, use force_gimple_operand to construct the side
36256         effects.
36258 2017-03-21  Martin Liska  <mliska@suse.cz>
36260         PR libfortran/79956
36261         * simplify-rtx.c (simplify_immed_subreg): Initialize a variable
36262         to NULL.
36264 2017-03-21  Brad Spengler <spender@grsecurity.net>
36266         PR plugins/80094
36267         * plugin.c (htab_hash_plugin): New function.
36268         (add_new_plugin): Use it and adjust.
36269         (parse_plugin_arg_opt): Adjust.
36270         (init_one_plugin): Likewise.
36272 2017-03-21  Richard Biener  <rguenther@suse.de>
36274         PR tree-optimization/80032
36275         * gimplify.c (gimple_push_cleanup): Add force_uncond parameter,
36276         if set force the cleanup to happen unconditionally.
36277         (gimplify_target_expr): Push inserted clobbers with force_uncond
36278         to avoid them being removed by control-dependent DCE.
36280 2017-03-21  Richard Biener  <rguenther@suse.de>
36282         PR tree-optimization/80122
36283         * tree-inline.c (copy_bb): Do not expans va-arg packs or
36284         va_arg_pack_len when the inlined call stmt requires pack
36285         expansion itself.
36286         * tree-inline.h (struct copy_body_data): Make call_stmt a gcall *.
36288 2017-03-21  Jakub Jelinek  <jakub@redhat.com>
36290         PR sanitizer/78158
36291         * tsan.c (instrument_builtin_call): If the memory model argument
36292         is not a constant, assume it is valid.
36294         PR c/67338
36295         * fold-const.c (round_up_loc): Negate divisor in unsigned type to
36296         avoid UB.
36298 2017-03-20  Segher Boessenkool  <segher@kernel.crashing.org>
36300         PR rtl-optimization/79910
36301         * combine.c (can_combine_p): Do not allow combining an I0 or I1
36302         if its dest is used by an insn before I2 (other than the combined
36303         insns themselves, which are properly handled already).
36305 2017-03-20  Segher Boessenkool  <segher@kernel.crashing.org>
36307         Revert:
36308         2017-03-17  Bernd Schmidt  <bschmidt@redhat.com>
36310         * combine.c (record_used_regs): New static function.
36311         (try_combine): Handle situations where there is an additional
36312         instruction between I2 and I3 which needs to have a LOG_LINK
36313         updated.
36315         Revert:
36316         2017-03-17  Jim Wilson  <jim.wilson@linaro.org>
36318         * combine.c (try_combine): Delete redundant i1 test.  Call
36319         prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
36321 2017-03-20  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
36323         PR target/80083
36324         * config/rs6000/rs6000.md (*movsi_internal1): Fix constraints for
36325         alternatives 13/14.
36327 2017-03-20  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36329         PR tree-optimization/80054
36330         * gimple-ssa-strength-reduction.c (all_phi_incrs_profitable): Fail
36331         the optimization if a PHI or any of its arguments is not dominated
36332         by the candidate's basis.  Use gphi* rather than gimple* as
36333         appropriate.
36334         (replace_profitable_candidates): Clean up a gimple* variable that
36335         should be a gphi* variable.
36337 2017-03-20  Martin Sebor  <msebor@redhat.com>
36339         PR c++/52477
36340         * doc/extend.texi (attribute constructor): Document present limitation.
36342 2017-03-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
36344         PR target/79963
36345         * config/rs6000/altivec.h (vec_all_ne): Under __cplusplus__ and
36346         __POWER9_VECTOR__ #ifdef control, change template definition to
36347         use Power9-specific built-in function.
36348         (vec_any_eq): Likewise.
36349         * config/rs6000/vector.md (vector_ae_v2di_p): Change the flag used
36350         to control outcomes from this test.
36351         (vector_ae_<mode>p): For VEC_F modes, likewise.
36353 2017-03-20  Ian Lance Taylor  <iant@google.com>
36355         * config/i386/i386.c (ix86_function_regparm): Save an extra
36356         register for -fsplit-stack with DECL_STATIC_CHAIN.
36358 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com>
36360         PR target/79912
36361         * config/riscv/riscv.c (riscv_preferred_reload_class): Remove.
36362         (TARGET_PREFERRED_RELOAD_CLASS): Likewise.
36364 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com>
36366         * config/riscv/riscv.c (riscv_print_operand): Use "fence
36367         iorw,ow".
36368         * config/riscv/sync.mc (mem_thread_fence_1): Use "fence
36369         iorw,iorw".
36371 2017-03-20  Marek Polacek  <polacek@redhat.com>
36373         PR sanitizer/80063
36374         * asan.c (DEF_SANITIZER_BUILTIN): Use do { } while (0).
36376 2017-03-20  Richard Biener  <rguenther@suse.de>
36378         PR tree-optimization/80113
36379         * graphite-isl-ast-to-gimple.c (copy_loop_phi_nodes): Do not
36380         allocate extra SSA name for PHI def.
36381         (add_close_phis_to_outer_loops): Likewise.
36382         (add_close_phis_to_merge_points): Likewise.
36383         (copy_loop_close_phi_args): Likewise.
36384         (copy_cond_phi_nodes): Likewise.
36386 2017-03-20  Martin Liska  <mliska@suse.cz>
36388         PR middle-end/79753
36389         * tree-chkp.c (chkp_build_returned_bound): Do not build
36390         returned bounds for a LHS that's not a BOUNDED_P type.
36392 2017-03-20  Martin Liska  <mliska@suse.cz>
36394         PR target/79769
36395         PR target/79770
36396         * tree-chkp.c (chkp_find_bounds_1): Handle REAL_CST,
36397         COMPLEX_CST and VECTOR_CST.
36399 2017-03-20  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
36401         PR target/78857
36402         * config/s390/s390.md ("cmp<mode>_ccs_0"): Add a clobber of the
36403         target operand.  A new splitter adds the clobber statement in case
36404         the target operand is dead anyway.
36406 2017-03-19  Gerald Pfeifer  <gerald@pfeifer.com>
36408         * doc/install.texi (Specific) <sparc-*-linux*>: No longer refer
36409         to age-old versions of binutils and glibc.
36411 2017-03-18  Segher Boessenkool  <segher@kernel.crashing.org>
36413         * doc/contrib.texi (Contributors): Remove duplicate entry for myself.
36415 2017-03-18  Gerald Pfeifer  <gerald@pfeifer.com>
36417         * doc/contrib.texi (Contributors): Add Segher Boessenkool.
36419 2017-03-18  Gerald Pfeifer  <gerald@pfeifer.com>
36421         * doc/install.texi (Specific) <arm-*-eabi>: Remove old
36422         requirement for binutils 2.13.
36424 2017-03-17  Jim Wilson  <jim.wilson@linaro.org>
36426         * combine.c (try_combine): Delete redundant i1 test.  Call
36427         prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
36429 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com
36431         * doc/install.texi (Specific) <riscv32-*-elf>: Add riscv32-*-elf,
36432         riscv32-*-linux, riscv64-*-elf, riscv64-*-linux to the table of
36433         contents.
36434         <riscv64-*-elf>: Re-arrange section
36435         <riscv32-*-elf>: Add a note about requiring binutils 2.28.
36436         <riscv32-*-linux>: Likewise.
36437         <riscv64-*-elf>: Likewise
36438         <riscv64-*-linux>: Likewise.
36440 2017-03-17  Richard Earnshaw  <rearnsha@arm.com>
36442         PR target/80052
36443         * aarch64.opt(verbose-cost-dump): Fix typo.
36445 2017-03-17  Pat Haugen  <pthaugen@us.ibm.com>
36447         PR target/79951
36448         * config/rs6000/rs6000.md (copysign<mode>3_fcpsgn): Test
36449         for VECTOR_UNIT_VSX_P (<MODE>mode) too.
36451 2017-03-17  Bernd Schmidt  <bschmidt@redhat.com>
36453         * reload.c (find_reloads): When reloading a nonoffsettable address,
36454         use RELOAD_OTHER for it and its address reloads.
36456         PR rtl-optimization/79910
36457         * combine.c (record_used_regs): New static function.
36458         (try_combine): Handle situations where there is an additional
36459         instruction between I2 and I3 which needs to have a LOG_LINK
36460         updated.
36462 2017-03-17  Jeff Law  <law@redhat.com>
36464         PR tree-optimization/71437
36465         * tree-vrp.c (simplify_stmt_for_jump_threading): Lookup the
36466         conditional in the hash table first.
36467         (vrp_dom_walker::before_dom_children): Extract condition from
36468         ASSERT_EXPR.  Record condition, its inverion and any implied
36469         conditions as well.
36471 2017-03-17  Marek Polacek  <polacek@redhat.com>
36472             Markus Trippelsdorf  <markus@trippelsdorf.de>
36474         PR tree-optimization/80079
36475         * gimple-ssa-store-merging.c (class pass_store_merging): Initialize
36476         m_stores_head.
36478 2017-03-17  Richard Biener  <rguenther@suse.de>
36480         PR middle-end/80075
36481         * tree-eh.c (stmt_could_throw_1_p): Only handle gimple assigns.
36482         Properly verify the LHS before the RHS possibly claims to be
36483         handled.
36484         (stmt_could_throw_p): Hande gimple conds fully here.  Clobbers
36485         do not throw.
36487 2017-03-17  Martin Jambor  <mjambor@suse.cz>
36489         * doc/invoke.texi (Option Options): Include -fipa-vrp in the list.
36490         (List of -O2 options): Likewise.
36491         (-fipa-bit-cp): Replace "ipa" with "interprocedural."
36492         (-fipa-vrp) New.
36494 2017-03-17  Tom de Vries  <tom@codesourcery.com>
36496         * gcov-dump.c (print_usage): Print bug_report_url.
36498 2017-03-17  Richard Biener  <rguenther@suse.de>
36500         PR middle-end/80050
36501         * genmatch.c (parser::next): Remove pointless check for CPP_EOF.
36502         (parser::peek): Likewise.
36504 2017-03-17  Richard Biener  <rguenther@suse.de>
36506         PR tree-optimization/80048
36507         * sese.c (free_sese_info): Properly release rename_map and
36508         copied_bb_map elements.
36510 2017-03-16  Alexandre Oliva  <aoliva@redhat.com>
36512         * gimple-ssa-store-merging.c (struct imm_store_chain_info):
36513         Add linked-list forward and backlinks.  Insert on
36514         construction, remove on destruction.
36515         (class pass_store_merging): Add m_stores_head field.
36516         (pass_store_merging::terminate_and_process_all_chains):
36517         Iterate over m_stores_head list.
36518         (pass_store_merging::terminate_all_aliasing_chains):
36519         Likewise.
36520         (pass_store_merging::execute): Check for debug stmts first.
36521         Push new chains onto the m_stores_head stack.
36523 2017-03-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
36525         PR target/71294
36526         * config/rs6000/vsx.md (vsx_splat_<mode>, VSX_D iterator): Allow a
36527         SPLAT operation on ISA 2.07 64-bit systems that have direct move,
36528         but no MTVSRDD support, by doing MTVSRD and XXPERMDI.
36530 2017-03-16  Jeff Law  <law@redhat.com>
36532         PR tree-optimization/71437
36533         * tree-ssa-dom.c (dom_opt_dom_walker): Remove thread_across_edge
36534         member function.  Implementation moved into after_dom_children
36535         member function and into the threader's thread_outgoing_edges
36536         function.
36537         (dom_opt_dom_walker::after_dom_children): Simplify by moving
36538         some code into new thread_outgoing_edges.
36539         * tree-ssa-threadedge.c (thread_across_edge): Make static and simplify
36540         definition.  Simplify marker handling (do it here).   Assume we always
36541         have the available expression and the const/copies tables.
36542         (thread_outgoing_edges): New function extracted from tree-ssa-dom.c
36543         and tree-vrp.c
36544         * tree-ssa-threadedge.h (thread_outgoing_edges): Declare.
36545         * tree-vrp.c (equiv_stack): No longer file scoped.
36546         (vrp_dom_walker): New class.
36547         (vrp_dom_walker::before_dom_children): New member function.
36548         (vrp_dom_walker::after_dom_children): Likewise.
36549         (identify_jump_threads):  Setup domwalker.  Use it rather than
36550         walking edges in a random order by hand.  Simplify setup/finalization.
36551         (finalize_jump_threads): Remove.
36552         (vrp_finalize): Do not call identify_jump_threads here.
36553         (execute_vrp): Do it here instead and call thread_through_all_blocks
36554         here too.
36556         PR tree-optimization/71437
36557         * tree-ssa-dom.c (pfn_simplify): Add basic_block argument.  All
36558         callers changed.
36559         (simplify_stmt_for_jump_threading): Add basic_block argument.  All
36560         callers changed.
36561         (lhs_of_dominating_assert): Moved from here into tree-vrp.c.
36562         (dom_opt_dom_walker::thread_across_edge): Remove
36563         handle_dominating_asserts argument.  All callers changed.
36564         (record_temporary_equivalences_from_stmts_at_dest): Corresponding
36565         changes.  Remove calls to lhs_of_dominating_assert.  Other
36566         uses of handle_dominating_asserts turn into unconditional code
36567         (simplify_control_stmt_condition_1): Likewise.
36568         (simplify_control_stmt_condition): Likewise.
36569         (thread_through_normal_block, thread_across_edge): Likewise.
36570         * tree-ssa-threadedge.h (thread_across_edge): Corresponding changes.
36571         * tree-vrp.c (lhs_of_dominating_assert): Move here.  Return original
36572         object if it is not an SSA_NAME.
36573         (simplify_stmt_for_jump_threading): Call lhs_of_dominating_assert
36574         before calling into the VRP specific simplifiers.
36575         (identify_jump_threads): Remove handle_dominating_asserts
36576         argument.
36578 2017-03-16  Jakub Jelinek  <jakub@redhat.com>
36580         PR fortran/79886
36581         * tree-diagnostic.c (default_tree_printer): No longer static.
36582         * tree-diagnostic.h (default_tree_printer): New prototype.
36584 2017-03-16  Tamar Christina  <tamar.christina@arm.com>
36586         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>)
36587         Change ins into fmov.
36589 2017-03-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36591         * config/aarch64/iterators.md (h_con): Return "x" for V4HF and V8HF.
36592         * config/aarch64/aarch64-simd.md (*aarch64_fma4_elt_from_dup<mode>):
36593         Use h_con constraint for operand 1.
36594         (*aarch64_fnma4_elt_from_dup<mode>): Likewise.
36595         (*aarch64_mulx_elt_from_dup<mode>): Likewise for operand 2.
36597 2017-03-15  Jeff Law  <law@redhat.com>
36599         PR tree-optimization/71437
36600         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): New function.
36601         (record_temporary_equivalences): Use it.
36603         PR tree-optimization/71437
36604         * tree-ssa-dom.c (struct cond_equivalence): Moved from here into
36605         tree-ssa-scopedtables.
36606         (lookup_avail_expr, build_and_record_new_cond): Likewise.
36607         (record_conditions, record_cond, vuse_eq): Likewise.
36608         (record_edge_info): Adjust to API tweak of record_conditions.
36609         (simplify_stmt_for_jump_threading): Similarly for lookup_avail_expr.
36610         (record_temporary_equivalences, optimize_stmt): Likewise.
36611         (eliminate_redundant_computations): Likewise.
36612         (record_equivalences_from_stmt): Likewise.
36613         * tree-ssa-scopedtables.c: Include options.h and params.h.
36614         (vuse_eq): New function, moved from tree-ssa-dom.c
36615         (build_and_record_new_cond): Likewise.
36616         (record_conditions): Likewise.  Accept vector of conditions rather
36617         than edge_equivalence structure for first argument.
36618         for the first argument.
36619         (avail_exprs_stack::lookup_avail_expr): New member function, moved
36620         from tree-ssa-dom.c.
36621         (avail_exprs_stack::record_cond): Likewise.
36622         * tree-ssa-scopedtables.h (struct cond_equivalence): Moved here
36623         from tree-ssa-dom.c.
36624         (avail_exprs_stack): Add new member functions lookup_avail_expr
36625         and record_cond.
36626         (record_conditions): Declare.
36628 2017-03-15  Vladimir Makarov  <vmakarov@redhat.com>
36630         PR target/80017
36631         * lra-constraints.c (process_alt_operands): Increase reject for
36632         reloading an input/output operand.
36634 2017-03-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
36636         PR target/79038
36637         * config/rs6000/rs6000.md (float<QHI:mode><IEEE128:mode>2): Define
36638         insns to convert from signed/unsigned char/short to IEEE 128-bit
36639         floating point.
36640         (floatuns<QHI:mode><IEEE128:mode>2): Likewise.
36642 2017-03-15  Uros Bizjak  <ubizjak@gmail.com>
36644         PR target/80019
36645         * config/i386/i386.c (ix86_vector_duplicate_value): Create
36646         subreg of inner mode for values already in registers.
36648 2017-03-15  Bernd Schmidt  <bschmidt@redhat.com>
36650         * config/c6x/c6x.c (hwloop_optimize): Handle case where the old
36651         iteration reg is used after the loop.
36653 2017-03-14  Martin Sebor  <msebor@redhat.com>
36655         PR tree-optimization/79800
36656         * gimple-ssa-sprintf.c (format_floating: Add argument.  Handle
36657         precision in negative-positive range.
36658         (format_floating): Call non-const overload with adjusted precision.
36660 2017-03-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
36662         PR target/79947
36663         * config/rs6000/rs6000.h (TARGET_FRSQRTES): Add check for
36664         -mpowerpc-gfxopt.
36666 2017-03-14  Martin Sebor  <msebor@redhat.com>
36668         PR middle-end/80020
36669         * builtin-attrs.def (ATTR_ALLOC_SIZE_2_NOTHROW_LIST): New macro.
36670         * builtins.def (aligned_alloc): Use it.
36672         PR c/79936
36673         * Makefile.in (GTFILES): Add calls.c.
36674         * calls.c: Include "gt-calls.h".
36676 2017-03-14  Bernd Schmidt  <bschmidt@redhat.com>
36678         PR rtl-optimization/79728
36679         * regs.h (struct target_regs): New field
36680         x_contains_allocatable_regs_of_mode.
36681         (contains_allocatable_regs_of_mode): New macro.
36682         * reginfo.c (init_reg_sets_1): Initialize it, and change
36683         contains_reg_of_mode so it includes global regs as well.
36684         * reload.c (push_reload): Use contains_allocatable_regs_of_mode
36685         rather than contains_regs_of_mode.
36687 2017-03-14  Martin Liska  <mliska@suse.cz>
36689         * doc/invoke.texi: Document options that can't be combined with
36690         -fcheck-pointer-bounds.
36692 2017-03-14  Martin Liska  <mliska@suse.cz>
36694         PR middle-end/79831
36695         * doc/invoke.texi (-Wchkp): Document the option.
36697 2017-03-14  Martin Liska  <mliska@suse.cz>
36699         * Makefile.in: Install gcov-dump.
36701 2017-03-14  Martin Liska  <mliska@suse.cz>
36703         * multiple_target.c (expand_target_clones): Bail out for
36704         an invalid attribute.
36706 2017-03-14  Richard Biener  <rguenther@suse.de>
36708         * alias.c (struct alias_set_entry): Pack properly.
36709         * cfgloop.h (struct loop): Likewise.
36710         * cse.c (struct set): Likewise.
36711         * ipa-utils.c (struct searchc_env): Likewise.
36712         * loop-invariant.c (struct invariant): Likewise.
36713         * lra-remat.c (struct cand): Likewise.
36714         * recog.c (struct change_t): Likewise.
36715         * rtl.h (struct address_info): Likewise.
36716         * symbol-summary.h (function_summary): Likewise.
36717         * tree-loop-distribution.c (struct partition): Likewise.
36718         * tree-object-size.c (struct object_size_info): Likewise.
36719         * tree-ssa-loop-ivopts.c (struct cost_pair): Likewise.
36720         * tree-ssa-threadupdate.c (struct ssa_local_info_t): Likewise.
36721         * tree-vect-data-refs.c (struct _vect_peel_info): Likewise.
36722         * tree-vect-slp.c (struct _slp_oprnd_info): Likewise.
36723         * tree-vect-stmts.c (struct simd_call_arg_info): Likewise.
36724         * tree-vectorizer.h (struct _loop_vec_info): Likewise.
36725         (struct _stmt_vec_info): Likewise.
36727 2017-03-14  Martin Liska  <mliska@suse.cz>
36729         PR target/79892
36730         * multiple_target.c (create_dispatcher_calls): Check that
36731         a target can create a function dispatcher.
36733 2017-03-14  Martin Liska  <mliska@suse.cz>
36735         PR lto/66295
36736         * multiple_target.c (expand_target_clones): Drop local.local
36737         flag for default implementation.
36739 2017-03-14  Richard Biener  <rguenther@suse.de>
36741         PR tree-optimization/80030
36742         * tree-vect-stmts.c (vectorizable_store): Plug memleak.
36744 2017-03-13  Kito Cheng  <kito.cheng@gmail.com>
36746         * config/riscv/riscv.c (riscv_emit_float_compare>: Use
36747         gcc_fallthrough() instead of __attribute__((fallthrough));
36749 2017-03-13  Gerald Pfeifer  <gerald@pfeifer.com>
36751         * doc/gcc.texi: Remove "up" link to (DIR).
36752         * doc/gccint.texi: Ditto.
36754 2017-03-13  Gerald Pfeifer  <gerald@pfeifer.com>
36756         * doc/install.texi (Specific) <avr>: Remove reference to
36757         binutils 2.13.
36759 2017-03-13  Jeff Law  <law@redhat.com>
36761         * config/riscv/riscv.c (riscv_emit_float_compare): Use fallthru
36762         attribute rather than comments.
36764         * config/pdp11/pdp11.md (movmemhi): Adjust operand numbers to
36765         match_scratch operand is highest.
36767 2017-03-13  Martin Liska  <mliska@suse.cz>
36769         PR middle-end/78339
36770         * ipa-pure-const.c (warn_function_noreturn): If the declarations
36771         is a CHKP clone, use original declaration.
36773 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36775         * config/arc/arc.c (arc_init): Use multiplier whenever we have it.
36776         (arc_conditional_register_usage): Use a different allocation order
36777         when optimizing for size.
36778         * common/config/arc/arc-common.c (arc_option_optimization_table):
36779         Section anchors default on when optimizing for size.
36781 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36783         * config/arc/arc.md (*tst_bitfield_tst): Fix pattern.
36785 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36787         * config/arc/arc.c (arc_output_addsi): Emit code density adds.
36788         * config/arc/arc.md (cpu_facility): Add cd variant.
36789         (*movqi_insn): Add code density variant.
36790         (*movhi_insn): Likewise.
36791         (*movqi_insn): Likewise.
36792         (*addsi3_mixed): Likewise.
36793         (subsi3_insn): Likewise.
36795 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36797         * config/arc/arc.md (movsi_cond_exec): Update constraint.
36799 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36801         * config/arc/arc.c (arc_legitimize_pic_address): Handle PIC
36802         expressions with MINUS and UNARY ops.
36804 2017-03-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36806         PR target/79911
36807         * config/arm/neon.md (vec_sel_widen_ssum_lo<VQI:mode><VW:mode>3):
36808         Rename to...
36809         (vec_sel_widen_ssum_lo<mode><V_half>3): ... This. Avoid mismatch
36810         between vec_select and vector argument.
36811         (vec_sel_widen_ssum_hi<VQI:mode><VW:mode>3): Rename to...
36812         (vec_sel_widen_ssum_hi<mode><V_half>3): ... This. Likewise.
36813         (vec_sel_widen_usum_lo<VQI:mode><VW:mode>3): Rename to...
36814         (vec_sel_widen_usum_lo<mode><V_half>3): ... This.
36815         (vec_sel_widen_usum_hi<VQI:mode><VW:mode>3): Rename to...
36816         (vec_sel_widen_usum_hi<mode><V_half>3): ... This.
36818 2017-03-13  Richard Biener  <rguenther@suse.de>
36820         PR other/79991
36821         * params.def (vect-max-peeling-for-alignment): Fix typo.
36823 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36825         * doc/install.texi (Specific) <mips-*-*>: Remove description of
36826         issue that only occurred with binutils below 2.18.
36828 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36830         * doc/install.texi (Specific) <cris-axis-elf>: No longer
36831         refer to binutils 2.11/2.12 minimum.
36833 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36835         * doc/install.texi (Specific) <powerpc-*-*>: Remove link to
36836         ftp.kernel.org and simplify binutils requirement.
36838 2017-03-11  Gerald Pfeifer  <gerald@pfeifer.com>
36840         * doc/invoke.texi (Warning Options): Fix spelling of link-time
36841         optimization.
36842         (Optimize Options): Ditto.  Also remove redundancy.
36844 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36846         PR translation/79848
36847         * ipa-devirt.c (warn_types_mismatch): Simplify uses of "%<%s%>" to
36848         "%qs".
36849         * ipa-pure-const.c (suggest_attribute): Likewise.  Convert _
36850         to G_ to avoid double translation.
36852 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36854         PR translation/79923
36855         * auto-profile.c (get_combined_location): Convert leading
36856         character of diagnostics to lower case and remove trailing period.
36857         (read_profile): Likewise for various diagnostics.
36858         * config/arm/arm.c (arm_option_override): Remove trailing period
36859         from various diagnostics.
36860         * config/msp430/msp430.c (msp430_expand_delay_cycles): Likewise.
36861         (msp430_expand_delay_cycles): Likewise.
36863 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36865         PR target/79925
36866         * config/aarch64/aarch64.c (aarch64_validate_mcpu): Quote the
36867         full command-line argument, rather than just "str".
36868         (aarch64_validate_march): Likewise.
36869         (aarch64_validate_mtune): Likewise.
36871 2017-03-10  Bernd Schmidt  <bschmidt@redhat.com>
36873         PR rtl-optimization/78911
36874         * lra-assigns.c (must_not_spill_p): New function.
36875         (spill_for): Use it.
36877 2017-03-10  Jakub Jelinek  <jakub@redhat.com>
36879         PR tree-optimization/79981
36880         * tree-vrp.c (extract_range_basic): Handle IMAGPART_EXPR of
36881         ATOMIC_COMPARE_EXCHANGE ifn result.
36882         (stmt_interesting_for_vrp, vrp_visit_stmt): Handle
36883         IFN_ATOMIC_COMPARE_EXCHANGE.
36885 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36887         PR driver/79875
36888         * opts.c (parse_sanitizer_options): Add missing question mark to
36889         "did you mean" message.
36891 2017-03-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36893         * config/rs6000/rs6000-builtin.def (VMULEUB_UNS): Remove orphaned
36894         built-in.
36895         (VMULEUH_UNS): Likewise.
36896         (VMULOUB_UNS): Likewise.
36897         (VMULOUH_UNS): Likewise.
36898         * config/rs6000/rs6000.c (builtin_function_type): Remove
36899         references to ALTIVEC_BUILTIN_VMUL[EO]U[BH]_UNS.
36901 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36903         PR bootstrap/79952
36904         * read-rtl-function.c (function_reader::read_rtx_operand): Update
36905         x with result of extra_parsing_for_operand_code_0.
36906         (function_reader::extra_parsing_for_operand_code_0): Convert
36907         return type from void to rtx, returning x.  When reading
36908         SYMBOL_REF with SYMBOL_FLAG_HAS_BLOCK_INFO, reallocate x to the
36909         larger size containing struct block_symbol.
36911 2017-03-10  Segher Boessenkool  <segher@kernel.crashing.org>
36913         * config/rs6000/rs6000.c (rs6000_option_override_internal): Disallow
36914         -mfloat128-hardware without -m64.
36916 2017-03-10  Will Schmidt <will_schmidt@vnet.ibm.com>
36918         PR target/79941
36919         * config/rs6000/rs6000.c (builtin_function_type): Add VMUL*U[HB]
36920         entries to the case statement that marks unsigned arguments to
36921         overloaded functions.
36923 2017-03-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
36925         * config/rs6000/rs6000.c (rs6000_option_override_internal): Fix
36926         two typographic errors in the handling of TARGET_UPPER_REGS_DI.
36928 2017-03-10  Pat Haugen  <pthaugen@us.ibm.com>
36930         PR target/79907
36931         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Test
36932         TARGET_UPPER_REGS_DI when setting 'wi' constraint regclass.
36934 2017-03-10  Martin Liska  <mliska@suse.cz>
36936         PR target/65705
36937         PR target/69804
36938         * toplev.c (process_options): Enable MPX with LSAN and UBSAN.
36939         * tree-chkp.c (chkp_walk_pointer_assignments): Verify that
36940         FIELD != NULL.
36942 2017-03-10  Olivier Hainque  <hainque@adacore.com>
36944         * tree-switch-conversion (array_value_type): Start by resetting
36945         candidate type to it's main variant.
36947 2017-03-10  Jakub Jelinek  <jakub@redhat.com>
36949         PR rtl-optimization/79909
36950         * combine.c (try_combine): Use simplify_replace_rtx on individual
36951         CALL_INSN_FUNCTION_USAGE elements instead of replace_rtx on copy_rtx
36952         of the whole CALL_INSN_FUNCTION_USAGE.
36954         PR tree-optimization/79972
36955         * gimple-ssa-warn-alloca.c (alloca_call_type): Only call
36956         get_range_info on SSA_NAMEs.  Formatting fixes.
36958 2017-03-10  Richard Biener  <rguenther@suse.de>
36959             Jakub Jelinek  <jakub@redhat.com>
36961         PR tree-optimization/77975
36962         * tree-ssa-loop-niter.c (get_base_for): Allow phi argument from latch
36963         edge to be constant.
36964         (get_val_for): For constant x return it.  Formatting fix.
36965         (loop_niter_by_eval): Avoid pointless looping if the next iteration
36966         would use the same bases as the current one.
36968 2017-03-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36970         * config/rs6000/rs6000.c (rs6000_gen_le_vsx_permute): Use rotate
36971         instead of vec_select for V1TImode.
36972         * conifg/rs6000/vsx.md (VSX_LE): Remove mode iterator that is no
36973         longer needed.
36974         (VSX_LE_128): Add V1TI to this mode iterator.
36975         (*vsx_le_perm_load_<mode>): Change to use VSX_D mode iterator.
36976         (*vsx_le_perm_store_<mode>): Likewise.
36977         (pre-reload splitter for VSX stores): Likewise.
36978         (post-reload splitter for VSX stores): Likewise.
36979         (*vsx_xxpermdi2_le_<mode>): Likewise.
36980         (*vsx_lxvd2x2_le_<mode>): Likewise.
36981         (*vsx_stxvd2x2_le_<mode>): Likewise.
36983 2017-03-09  Michael Eager  <eager@eagercon.com>
36985         Correct failures with --enable-checking=yes,rtl.
36987         * config/microblaze/microblaze.c (microblaze_expand_shift):
36988         Replace GET_CODE test with CONST_INT_P and INTVAL test with
36989         test for const0_rtx.
36990         * config/microblaze/microblaze.md (ashlsi3_byone, ashrsi3_byone,
36991         lshrsi3_byone): Replace INTVAL with test for const1_rtx.
36993 2017-03-09  Richard Biener  <rguenther@suse.de>
36995         PR tree-optimization/79977
36996         * graphite-scop-detection.c (scop_detection::merge_sese):
36997         Handle the case of extra exits to blocks dominating the entry.
36999 2017-03-09  Toma Tabacu  <toma.tabacu@imgtec.com>
37001         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
37002         Document rdynamic.
37004 2017-03-09  Vladimir Makarov  <vmakarov@redhat.com>
37006         PR rtl-optimization/79949
37007         * lra-constraints.c (process_alt_operands): Check memory when
37008         trying to predict a cycle.  Print about the overall increase.
37010 2017-03-09  Richard Biener  <rguenther@suse.de>
37012         PR middle-end/79971
37013         * gimple-expr.c (useless_type_conversion_p): Preserve
37014         TYPE_SATURATING for fixed-point types.
37016 2017-03-09  Richard Biener  <rguenther@suse.de>
37018         PR ipa/79970
37019         * ipa-prop.c (ipa_modify_formal_parameters): Avoid changing
37020         alignment of BLKmode params.
37022 2017-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
37024         PR target/79913
37025         * config/aarch64/iterators.md (VALL_F16_NO_V2Q): New mode iterator.
37026         (VALL_NO_V2Q): Likewise.
37027         (VDQF_DF): Delete.
37028         * config/aarch64/aarch64-simd.md
37029         (aarch64_dup_lane_<vswap_width_name><mode>): Use VALL_F16_NO_V2Q
37030         iterator.
37031         (*aarch64_simd_vec_copy_lane_<vswap_width_name><mode>): Use
37032         VALL_NO_V2Q mode iterator.
37033         (*aarch64_vgetfmulx<mode>): Use VDQF iterator.
37035 2017-03-09  Martin Liska  <mliska@suse.cz>
37037         PR tree-optimization/79631
37038         * tree-chkp-opt.c (chkp_is_constant_addr): Call
37039         tree_int_cst_sign_bit just for INTEGER constants.
37041 2017-03-09  Martin Liska  <mliska@suse.cz>
37043         PR target/65705
37044         PR target/69804
37045         * toplev.c (process_options): Disable -fcheck-pointer-bounds with
37046         sanitizers.
37048 2017-03-09  Marek Polacek  <polacek@redhat.com>
37050         PR c++/79672
37051         * tree.c (inchash::add_expr): Handle TREE_VEC.
37053 2017-03-09  Martin Liska  <mliska@suse.cz>
37055         PR ipa/79764
37056         (chkp_narrow_size_and_offset): New function.
37057         (chkp_parse_array_and_component_ref): Support BIT_FIELD_REF.
37058         (void chkp_parse_bit_field_ref): New function.
37059         (chkp_make_addressed_object_bounds): Add case for BIT_FIELD_REF.
37060         (chkp_process_stmt): Use chkp_parse_bit_field_ref.
37062 2017-03-09  Martin Liska  <mliska@suse.cz>
37064         PR ipa/79761
37065         * tree-chkp.c (chkp_get_bound_for_parm): Get bounds for a param.
37066         (chkp_find_bounds_1): Remove gcc_unreachable.
37068 2017-03-09  Jakub Jelinek  <jakub@redhat.com>
37070         PR sanitizer/79944
37071         * asan.c (get_mem_refs_of_builtin_call): For BUILT_IN_ATOMIC* and
37072         BUILT_IN_SYNC*, determine the access type from the size suffix and
37073         always build a MEM_REF with that type.  Handle forgotten
37074         BUILT_IN_SYNC_FETCH_AND_NAND_16 and BUILT_IN_SYNC_NAND_AND_FETCH_16.
37076         PR target/79932
37077         * config/i386/avx512vlintrin.h (_mm256_cmpge_epi32_mask,
37078         _mm256_cmpge_epi64_mask, _mm256_cmpge_epu32_mask,
37079         _mm256_cmpge_epu64_mask, _mm256_cmple_epi32_mask,
37080         _mm256_cmple_epi64_mask, _mm256_cmple_epu32_mask,
37081         _mm256_cmple_epu64_mask, _mm256_cmplt_epi32_mask,
37082         _mm256_cmplt_epi64_mask, _mm256_cmplt_epu32_mask,
37083         _mm256_cmplt_epu64_mask, _mm256_cmpneq_epi32_mask,
37084         _mm256_cmpneq_epi64_mask, _mm256_cmpneq_epu32_mask,
37085         _mm256_cmpneq_epu64_mask, _mm256_mask_cmpge_epi32_mask,
37086         _mm256_mask_cmpge_epi64_mask, _mm256_mask_cmpge_epu32_mask,
37087         _mm256_mask_cmpge_epu64_mask, _mm256_mask_cmple_epi32_mask,
37088         _mm256_mask_cmple_epi64_mask, _mm256_mask_cmple_epu32_mask,
37089         _mm256_mask_cmple_epu64_mask, _mm256_mask_cmplt_epi32_mask,
37090         _mm256_mask_cmplt_epi64_mask, _mm256_mask_cmplt_epu32_mask,
37091         _mm256_mask_cmplt_epu64_mask, _mm256_mask_cmpneq_epi32_mask,
37092         _mm256_mask_cmpneq_epi64_mask, _mm256_mask_cmpneq_epu32_mask,
37093         _mm256_mask_cmpneq_epu64_mask, _mm_cmpge_epi32_mask,
37094         _mm_cmpge_epi64_mask, _mm_cmpge_epu32_mask, _mm_cmpge_epu64_mask,
37095         _mm_cmple_epi32_mask, _mm_cmple_epi64_mask, _mm_cmple_epu32_mask,
37096         _mm_cmple_epu64_mask, _mm_cmplt_epi32_mask, _mm_cmplt_epi64_mask,
37097         _mm_cmplt_epu32_mask, _mm_cmplt_epu64_mask, _mm_cmpneq_epi32_mask,
37098         _mm_cmpneq_epi64_mask, _mm_cmpneq_epu32_mask, _mm_cmpneq_epu64_mask,
37099         _mm_mask_cmpge_epi32_mask, _mm_mask_cmpge_epi64_mask,
37100         _mm_mask_cmpge_epu32_mask, _mm_mask_cmpge_epu64_mask,
37101         _mm_mask_cmple_epi32_mask, _mm_mask_cmple_epi64_mask,
37102         _mm_mask_cmple_epu32_mask, _mm_mask_cmple_epu64_mask,
37103         _mm_mask_cmplt_epi32_mask, _mm_mask_cmplt_epi64_mask,
37104         _mm_mask_cmplt_epu32_mask, _mm_mask_cmplt_epu64_mask,
37105         _mm_mask_cmpneq_epi32_mask, _mm_mask_cmpneq_epi64_mask,
37106         _mm_mask_cmpneq_epu32_mask, _mm_mask_cmpneq_epu64_mask): Move
37107         definitions outside of __OPTIMIZE__ guarded section.
37109         PR target/79932
37110         * config/i386/avx512bwintrin.h (_mm512_packs_epi32,
37111         _mm512_maskz_packs_epi32, _mm512_mask_packs_epi32,
37112         _mm512_packus_epi32, _mm512_maskz_packus_epi32,
37113         _mm512_mask_packus_epi32): Move definitions outside of __OPTIMIZE__
37114         guarded section.
37116 2017-03-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
37118         * config/s390/vx-builtins.md ("vfee<mode>", "vfeez<mode>")
37119         ("vfenez<mode>"): Add missing constraints.
37121 2017-03-08  Martin Sebor  <msebor@redhat.com>
37123         PR target/79928
37124         * config/nds32/nds32.c (nds32_option_override):
37125         Fix misspelled diagnostic.
37127 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
37129         PR c/79940
37130         * gimplify.c (gimplify_omp_for): Replace index var in outer
37131         taskloop statement with an artificial variable and add
37132         OMP_CLAUSE_PRIVATE clause for it.
37134 2017-03-08  Richard Biener  <rguenther@suse.de>
37136         PR tree-optimization/79955
37137         * tree-ssa-uninit.c (warn_uninitialized_vars): Do not warn
37138         for accesses that are completely outside of the variable.
37140 2017-03-08  Andrew Haley  <aph@redhat.com>
37142         PR tree-optimization/79943
37143         * tree-ssa-loop-split.c (compute_new_first_bound): When
37144         calculating the new upper bound, (END-BEG) should be added, not
37145         subtracted.
37147 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
37149         * config/avr/avr.md (setmemhi): Make sure match_dup
37150         operand number comes before match_scratch.
37152 2017-03-08  Richard Biener  <rguenther@suse.de>
37154         PR tree-optimization/79920
37155         * tree-vect-slp.c (vect_create_mask_and_perm): Remove and inline
37156         with ncopies == 1 to ...
37157         (vect_transform_slp_perm_load): ... here.  Properly compute
37158         all element loads by iterating VF times over the group.  Do
37159         not handle ncopies (computed in a broken way) in
37160         vect_create_mask_and_perm.
37162 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
37164         PR sanitizer/79904
37165         * internal-fn.c (expand_vector_ubsan_overflow): If arg0 or arg1
37166         is a uniform vector, use uniform_vector_p return value instead of
37167         building ARRAY_REF on folded VIEW_CONVERT_EXPR to array type.
37169 2017-03-07  Marek Polacek  <polacek@redhat.com>
37171         PR middle-end/79809
37172         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Use HOST_WIDE_INT.
37173         (alloca_call_type): Likewise.
37175 2017-03-07  Martin Liska  <mliska@suse.cz>
37177         * gcov.c (process_args): Put comment to correct location.
37179 2017-03-07  Martin Liska  <mliska@suse.cz>
37181         PR middle-end/68270
37182         * tree-chkp.c (chkp_may_narrow_to_field): Add new argument ref.
37183         Use array_at_struct_end_p instead of DECL_CHAIN (field).
37184         (chkp_narrow_bounds_for_field): Likewise.
37185         (chkp_parse_array_and_component_ref): Pass one more argument to
37186         call.
37188 2017-03-07  Richard Biener  <rguenther@suse.de>
37190         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Preserve
37191         preheaders.
37193 2017-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
37195         * config/i386/i386.c (ix86_local_alignment): Align most aggregates
37196         of 16 bytes and more to 16 bytes, not those of 16 bits and more.
37198 2017-03-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
37200         PR c/79855
37201         * params.def (PARAM_STORE_MERGING_ALLOW_UNALIGNED): Add full stop
37202         to end of description.
37203         (PARAM_MAX_STORES_TO_MERGE): Likewise.
37205 2017-03-07  Jakub Jelinek  <jakub@redhat.com>
37207         PR rtl-optimization/79901
37208         * config/i386/sse.md (*avx512bw_<code><mode>3<mask_name>): Renamed to
37209         ...
37210         (*avx512f_<code><mode>3<mask_name>): ... this.
37211         (<code><mode>3 with maxmin code iterator): Use VI8_AVX2_AVX512F
37212         iterator instead of VI8_AVX2_AVX512BW.
37214         PR rtl-optimization/79901
37215         * expr.c (expand_expr_real_2): For vector MIN/MAX, if there is no
37216         min/max expander, expand it using expand_vec_cond_expr.
37218         PR sanitizer/79897
37219         * ubsan.c (ubsan_encode_value): Call mark_addressable on the
37220         temporary.
37222 2017-03-06  Jakub Jelinek  <jakub@redhat.com>
37224         PR c++/79821
37225         * dwarf2out.h (dw_vec_const): Change array type from unsigned char *
37226         to void * for PCH reasons.
37227         * dwarf2out.c (output_loc_operands, output_die): Cast
37228         v.val_vec.array to unsigned char *.
37230 2017-03-06  John David Anglin  <danglin@gcc.gnu.org>
37232         PR target/77850
37233         * config/pa/pa-64.h (PAD_VARARGS_DOWN): Don't pad down complex and
37234         vector types.
37236 2017-03-06  Vladimir Makarov  <vmakarov@redhat.com>
37238         PR rtl-optimization/79571
37239         * lra-constraints.c (process_alt_operands): Calculate static
37240         reject and subtract it from overall when only addresses will be
37241         reloaded.
37243 2017-03-06  Julia Koval  <julia.koval@intel.com>
37245         PR target/79793
37246         * config/i386/i386.c (ix86_minimum_incoming_stack_boundary): Set
37247         incoming stack boundary to 128 for 64-bit targets.
37249 2017-03-06  Richard Biener  <rguenther@suse.de>
37251         PR tree-optimization/79894
37252         * tree-vectorizer.c (vectorize_loops): Set loop_vectorized_call
37253         to NULL after folding it.
37255 2017-03-06  Richard Biener  <rguenther@suse.de>
37257         PR tree-optimization/79824
37258         * tree-vect-stmts.c (get_group_load_store_type): Fix alignment
37259         check disabling peeling for gaps.
37261 2017-03-06  Toma Tabacu  <toma.tabacu@imgtec.com>
37263         * doc/sourcebuild.texi (Effective-Target Keywords, Environment
37264         attributes): Document gettimeofday.
37266 2017-03-06  Robin Dapp  <rdapp@linux.vnet.ibm.com>
37268         * config/s390/s390.c (s390_option_override_internal): Set
37269         PARAM_MIN_VECT_LOOP_BOUND
37271 2017-03-06  Robin Dapp  <rdapp@linux.vnet.ibm.com>
37273         * config/s390/s390.c (s390_asm_output_function_label): Use nopr %r0.
37274         * config/s390/s390.md: Likewise.
37276 2017-03-06  Jakub Jelinek  <jakub@redhat.com>
37278         PR target/79812
37279         * config/i386/sse.md (VI8F_256_512): Remove mode iterator.
37280         (<avx2_avx512>_perm<mode>): Rename to ...
37281         (avx2_perm<mode>): ... this.  Use VI8F_256 iterator instead
37282         of VI8F_256_512.
37283         (<avx512>_perm<mode>_mask): Rename to ...
37284         (avx512vl_perm<mode>_mask): ... this.  Use VI8F_256 iterator instead
37285         of VI8F_256_512.
37286         (<avx2_avx512>_perm<mode>_1<mask_name>): Rename to ...
37287         (avx2_perm<mode>_1<mask_name): ... this.  Use VI8F_256 iterator
37288         instead of VI8F_256_512.
37289         (avx512f_perm<mode>): New define_expand.
37290         (avx512f_perm<mode>_mask): Likewise.
37291         (avx512f_perm<mode>_1<mask_name>): New define_insn.
37292         (<avx512>_vec_dup<mode>_1): Fix up vec_select mode.
37294 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37296         * config/mips/mips-msa.md (msa_fmax_a_<msafmt>, msa_fmin_a_<msafmt>,
37297         msa_max_a_<msafmt>, msa_min_a_<msafmt>): Introduce mode interator for
37298         if_then_else.
37299         (smin<mode>3, smax<mode>3): Change operand print code from 'B' to 'E'.
37301 2017-03-06  Martin Liska  <mliska@suse.cz>
37303         PR sanitize/79783
37304         * asan.c (asan_expand_poison_ifn): Do not expand ASAN_POISON
37305         when having a SSA NAME w/o VAR_DECL assigned to it.
37307 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37309         * config/mips/mips-msa.md (msa_dotp_<su>_d, msa_dpadd_<su>_d,
37310         msa_dpsub_<su>_d): Fix MODE for vec_select.
37312 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37314         * config/mips/mips.c (mips_gen_const_int_vector): Change type of last
37315         argument.
37316         * config/mips/mips-protos.h (mips_gen_const_int_vector): Likewise.
37318 2017-03-06  Richard Biener  <rguenther@suse.de>
37320         * lto-streamer.c (lto_check_version): Use %qs in diagnostics.
37321         * plugin.c (register_plugin_info): Likewise.
37322         * tree-chkp.c (chkp_make_static_const_bounds): Likewise.
37324 2017-03-05  Jakub Jelinek  <jakub@redhat.com>
37326         * config/i386/sse.md (sse_storehps, sse_storelps,
37327         avx_<castmode><avxsizesuffix>_<castmode>,
37328         avx512f_<castmode><avxsizesuffix>_<castmode>,
37329         avx512f_<castmode><avxsizesuffix>_256<castmode>): Require
37330         in condition that at least one operand is not a MEM.
37332 2017-03-03  Jakub Jelinek  <jakub@redhat.com>
37334         PR middle-end/79805
37335         * internal-fn.def (ATOMIC_BIT_TEST_AND_SET, ATOMIC_BIT_TEST_AND_RESET,
37336         ATOMIC_BIT_TEST_AND_COMPLEMENT, ATOMIC_COMPARE_EXCHANGE): Remove
37337         ECF_NOTHROW.
37338         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Set
37339         gimple_call_nothrow_p flag based on whether original builtin can throw.
37340         If it can, emit following stmts on the fallthrough edge.
37341         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Similarly, except
37342         don't create new bb if inserting just debug stmts on the edge, try to
37343         insert them on the fallthru bb or just reset debug stmts.
37345 2017-03-03  Segher Boesssenkool  <segher@kernel.crashing.org>
37347         PR target/43763
37348         * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Save and
37349         restore recog_data (including the operand rtxes inside it) around
37350         the call to get_insn_template.
37352 2017-03-03  Martin Sebor  <msebor@redhat.com>
37354         PR tree-optimization/79699
37355         * context.c (context::~context): Free MPFR caches to avoid
37356         a memory leak on program exit.
37358 2017-03-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
37360         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
37361         Use wide_int::ulow () instead of .elt (0).
37363 2017-03-03  Uros Bizjak  <ubizjak@gmail.com>
37365         * config/i386/i386.md (*pushtf): Change *roF constraint to *roC.
37366         (*pushxf): Limit oF constraint to 32bit targets and add oC
37367         constraint for 64bit targets.
37368         (pushxf splitter): Use PUSH_ROUNDING to calculate stack adjustment.
37369         (*pushdf): Change rmF constraint to rmC.
37371 2017-03-03  Martin Liska  <mliska@suse.cz>
37373         * tree-ssa-loop-prefetch.c (pass_loop_prefetch::execute):
37374         Remove unused variable.
37376 2017-03-03  Jakub Jelinek  <jakub@redhat.com>
37378         PR target/79807
37379         * config/i386/i386.c (ix86_expand_multi_arg_builtin): If target
37380         is a memory operand, increase num_memory.
37381         (ix86_expand_args_builtin): Likewise.
37383 2017-03-03  Jan Hubicka  <jh@suse.cz>
37385         PR lto/79760
37386         * ipa-devirt.c (maybe_record_node): Properly handle
37387         __cxa_pure_virtual visibility.
37389 2017-03-03  Martin Liska  <mliska@suse.cz>
37391         PR tree-optimization/79803
37392         * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Remove
37393         assert.
37394         (pass_loop_prefetch::execute): Disabled optimization if an
37395         assumption about L1 cache size is not met.
37397 2017-03-03  Martin Liska  <mliska@suse.cz>
37399         PR rtl-optimization/79574
37400         * gcse.c (struct gcse_expr): Use HOST_WIDE_INT instead of int.
37401         (hash_scan_set): Likewise.
37402         (dump_hash_table): Likewise.
37403         (hoist_code): Likewise.
37405 2017-03-03  Richard Biener  <rguenther@suse.de>
37407         * fixed-value.c (fixed_from_string): Restore use of elt (1)
37408         in place of uhigh ().
37409         (fixed_convert_from_real): Likewise.
37411 2017-03-03  Uros Bizjak  <ubizjak@gmail.com>
37413         PR target/79514
37414         * config/i386/i386.md (*pushxf_rounded): Use Pmode instead of DImode.
37416 2017-03-03  Richard Biener  <rguenther@suse.de>
37418         PR middle-end/79818
37419         * match.pd ( X +- C1 CMP C2 -> X CMP C2 -+ C1): Add missing
37420         TYPE_OVERFLOW_UNDEFINED check.
37422 2017-03-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
37424         * config/rs6000/vector.md (vector_ne_<mode>_p): Correct operand
37425         numbers.
37426         (vector_ae_<mode>_p): Likewise.
37427         (vector_nez_<mode>_p): Likewise.
37428         (vector_ne_v2di_p): Likewise.
37429         (vector_ae_v2di_p): Likewise.
37430         (vector_ne_<mode>_p): Likewise.
37431         * config/rs6000/vsx.md (vsx_tsqrt<mode>2_fg): Correct operand
37432         numbers.
37433         (vsx_tsqrt<mode>2_fe): Likewise.
37435 2017-03-02  Uros Bizjak  <ubizjak@gmail.com>
37437         PR target/79514
37438         * config/i386/i386.md (*pushxf_rounded): New insn_and_split pattern.
37440 2017-03-02  Jakub Jelinek  <jakub@redhat.com>
37442         PR rtl-optimization/79780
37443         * cprop.c (one_cprop_pass): When second and further conditional trap
37444         in a single basic block is turned into an unconditional trap, turn it
37445         into a deleted note to avoid RTL verification failures.
37447 2017-03-02  Richard Biener  <rguenther@suse.de>
37449         * fold-const.c (const_binop): Use ulow () instead of elt (0).
37451 2017-03-02  Richard Biener  <rguenther@suse.de>
37453         PR tree-optimization/79345
37454         PR c++/42000
37455         * tree-ssa-alias.c (walk_aliased_vdefs_1): Take a limit
37456         param and abort the walk, returning -1 if it is hit.
37457         (walk_aliased_vdefs): Take a limit param and pass it on.
37458         * tree-ssa-alias.h (walk_aliased_vdefs): Add a limit param,
37459         defaulting to 0 and return a signed int.
37460         * tree-ssa-uninit.c (struct check_defs_data): New struct.
37461         (check_defs): New helper.
37462         (warn_uninitialized_vars): Use walk_aliased_vdefs to warn
37463         about uninitialized memory.
37464         * fixed-value.c (fixed_from_string): Use ulow/uhigh to avoid
37465         bogus uninitialized warning.
37466         (fixed_convert_from_real): Likewise.
37468 2017-03-02  Bin Cheng  <bin.cheng@arm.com>
37470         PR tree-optimization/66768
37471         * tree-ssa-loop-ivopts.c (find_interesting_uses_address): Skip addr
37472         iv_use if base object can't be determined.
37474 2017-03-02  Jakub Jelinek  <jakub@redhat.com>
37476         PR tree-optimization/79345
37477         * gensupport.h (struct pattern_stats): Add min_scratch_opno field.
37478         * gensupport.c (get_pattern_stats_1) <case MATCH_SCRATCH>: Update it.
37479         (get_pattern_stats): Initialize it.
37480         * genemit.c (gen_expand): Verify match_scratch numbers come after
37481         match_operand/match_dup numbers.
37482         * config/i386/i386.md (<s>mul<mode>3_highpart): Swap match_dup and
37483         match_scratch numbers.
37484         * config/i386/sse.md (avx2_gathersi<mode>, avx2_gatherdi<mode>):
37485         Likewise.
37486         * config/s390/s390.md (trunctdsd2): Likewise.
37488 2017-03-02  Richard Biener  <rguenther@suse.de>
37490         * wide-int.h (wide_int_storage::operator=): Implement in terms
37491         of wi::copy.
37493 2017-03-02  Richard Biener  <rguenther@suse.de>
37495         PR tree-optimization/79777
37496         * tree-ssa-pre.c (eliminate_insert): Give up if we simplify
37497         the to insert expression to sth existing.
37499 2017-03-01  Martin Sebor  <msebor@redhat.com>
37501         PR middle-end/79692
37502         * gimple-ssa-sprintf.c
37503         (directive::known_width_and_precision): New function.
37504         (format_integer): Use it.
37505         (get_mpfr_format_length): Consider the full range of precision
37506         when computing %g output with the # flag.  Set the likely byte
37507         count to 3 rather than 1 when precision is indeterminate.
37508         (format_floating): Correct the lower bound of precision.
37510 2017-03-01  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
37512         * doc/invoke.texi: Document default code model for 64-bit Linux.
37514 2017-03-01  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
37516         PR target/79752
37517         * config/rs6000/rs6000.md (peephole2 for udiv/umod): Should emit
37518         udiv rather than div since input pattern is unsigned.
37520 2017-03-01  Uros Bizjak  <ubizjak@gmail.com>
37522         * config/i386/i386.c (print_reg): Warn for values of
37523         unsupported size in integer register.
37525 2017-03-01  Michael Meissner  <meissner@linux.vnet.ibm.com>
37527         PR target/79439
37528         * config/rs6000/predicates.md (current_file_function_operand): Do
37529         not allow self calls to be local if the function is replaceable.
37531 2017-03-01  Kelvin Nilsen  <kelvin@gcc.gnu.org>
37533         PR target/79395
37534         * config/rs6000/altivec.h (vec_ctz and others): Change the
37535         preprocessor macro that controls conditional compilation from
37536         _ARCH_PWR9 to __POWER9_VECTOR__.
37537         (vec_all_ne): Change parameterization of __altivec_scalar_pred
37538         macro expansion under preprocessor #ifdef __POWER9_VECTOR__
37539         control (instead of _ARCH_PWR9 control) so that template
37540         definition uses power9-specific function.
37541         (vec_any_eq): Likewise.
37542         (vec_all_ne): Change macro definition to use a power9-specific
37543         expansion under #ifdef __POWER9_VECTOR__ control (instead of
37544         _ARCH_PWR9 control).
37545         (vec_any_eq) Likewise.
37546         * config/rs6000/rs6000-builtin.def (CMPNEF): Remove BU_P9V_AV_2
37547         expansion for CMPNEF to remove support for xvcmpnesp instruction.
37548         (CMPNED): Remove BU_P9V_AV2 expansion for CMPNED to remove
37549         support for xvcmpnedp instruction.
37550         (VCMPNEB_P): Replace BU_P9V_AV_P macro expansion with BU_P9V_AV_2
37551         macro expansion so that Power9 implementation of vec_all_ne does
37552         not use the AltiVec predicate framework.
37553         (VCMPNEH_P): Likewise.
37554         (VCMPNEW_P): Likewise.
37555         (VCMPNED_P): Likewise.
37556         (VCMPNEFP_P): Likewise.
37557         (VCMPNEDP_P): Likewise.
37558         (VCMPAEB_P): Add BU_P9V_AV_2 macro expansion to change
37559         implementation of vec_any_eq to not use AltiVec predicate
37560         framework.
37561         (VCMPAEH_P): Likewise.
37562         (VCMPAEW_P): Likewise.
37563         (VCMPAED_P): Likewise.
37564         (VCMPAEFP_P): Likewise.
37565         (VCMPAEDP_P): Likewise.
37566         (VCMPNE_P): Replace BU_P9V_OVERLOAD_P macro expansion with
37567         BU_P9V_OVERLOAD_2 so that Power9 implementation of vec_all_ne does
37568         not use the AltiVec predicate framework.
37569         (VCMPAE_P): Add BU_P9V_OVERLOAD_2 macro to change implementation
37570         of vec_any_eq to not use AltiVec predicate framework.
37571         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Add
37572         support for predefined __POWER9_VECTOR__ macro to indicate that
37573         Power9 instruction selection is enabled.
37574         (altivec_overloaded_builtins): Remove extraneous
37575         ALTIVEC_BUILTIN_VEC_CMPNE entry for overloaded
37576         function argument types RS6000_BTI_bool_V16QI and
37577         RS6000_BTI_bool_V16QI.  Remove erroneous ALTIVEC_BUILTIN_VEC_CMPNE
37578         entry for overloaded function argument types RS6000_BTI_bool_V4SI
37579         andRS6000_BTI_bool_V4SI, mapping to P9V_BUILTIN_CMPNEB.  Remove
37580         two entries mapping to P9V_BUITIN_CMPNED and one entry mapping to
37581         P9V_BUILTIN_CMPNEF to force use of instructions not specific to
37582         Power9 for implementations of vec_cmpne.  Change the signature for
37583         all definitions of the overloaded P9V_BUILTIN_VEC_CMPNE_P function
37584         (representing vec_all_ne) to remove the previously described first
37585         argument of type RS6000_BTI_INTSI, as this was an artifact of
37586         reliance on the AltiVec predicate framework, which is no longer
37587         used in the implementation of these functions.  Add
37588         P9V_BUILTIN_VEC_VCMPAE_P entries (representing the vec_anyeq
37589         function) to match all of the P9V_BUILTIN_VEC_VCMNE_P entries
37590         since, unlike the AltiVec predicate framework implementation, we
37591         do not share function descriptors between vec_alle and vec_anyeq.
37592         (altivec_resolve_overloaded_builtin): Add SFmode and DFmode to the
37593         set of modes that receive special treatment even when
37594         TARGET_P9_VECTOR is true.  The special treatment emits code that
37595         does not depend on Power9 instructions.
37596         * config/rs6000/vector.md (vector_ne_<mode>_p): Change this
37597         define_expand to not rely on AltiVec predicate framework.
37598         (vector_ae_<mode>p): New define_expand to represent vec_any_eq
37599         function.
37600         (vector_ne_v2di_p): Change this define_expand to not rely on
37601         AltiVec predicate framework.
37602         (vector_ae_v2di_p): New define_expand to represent vec_any_eq
37603         function.
37604         (vector_ne_<mode>_p): Change this define_expand to not rely on
37605         AltiVec predicate framework.
37606         (vector_ae_<mode>p): New define_expand to represent vec_any_eq
37607         function.
37608         * config/rs6000/vsx.md (*vsx_ne_<mode>_p): For modes VSX_EXTRACT_I
37609         (V16QI, V8HI, V4SI), correct a typo in the code emitted for this
37610         define_insn pattern.
37611         (*vsx_ne_<mode>_p): For modes VSX_F (V4SF and V2DF), remove this
37612         define_insn pattern because the xvcmpne<VSs>. instruction is not
37613         supported.
37614         (vcmpne<VSs>): Remove this define_insn because xvcmpne<VSs>
37615         instruction is not supported.
37617 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37619         * config/nvptx/nvptx.c: Include intl.h.
37621 2017-03-01  Martin Jambor  <mjambor@suse.cz>
37623         PR lto/78140
37624         * ipa-prop.h (ipa_bits): Removed field known.
37625         (ipa_jump_func): Removed field vr_known.  Changed fields bits and m_vr
37626         to pointers.  Adjusted their comments to warn about their sharing.
37627         (ipcp_transformation_summary): Change bits to a vector of pointers.
37628         (ipa_check_create_edge_args): Moved to ipa-prop.c, declare.
37629         (ipa_get_ipa_bits_for_value): Declare.
37630         * tree-vrp.h (value_range): Mark as GTY((for_user)).
37631         * ipa-prop.c (ipa_bit_ggc_hash_traits): New.
37632         (ipa_bits_hash_table): Likewise.
37633         (ipa_vr_ggc_hash_traits): Likewise.
37634         (ipa_vr_hash_table): Likewise.
37635         (ipa_print_node_jump_functions_for_edge): Adjust for bits and m_vr
37636         being pointers and vr_known being removed.
37637         (ipa_set_jf_unknown): Likewise.
37638         (ipa_get_ipa_bits_for_value): New function.
37639         (ipa_set_jfunc_bits): Likewise.
37640         (ipa_get_value_range): New overloaded functions.
37641         (ipa_set_jfunc_vr): Likewise.
37642         (ipa_compute_jump_functions_for_edge): Use the above functions to
37643         construct bits and vr parts of jump functions.
37644         (ipa_check_create_edge_args): Move here from ipa-prop.h, also allocate
37645         ipa_bits_hash_table and ipa_vr_hash_table if they do not already
37646         exist.
37647         (ipcp_grow_transformations_if_necessary): Also allocate
37648         ipa_bits_hash_table and ipa_vr_hash_table if they do not already
37649         exist.
37650         (ipa_node_params_t::duplicate): Do not copy bits, just pointers to
37651         them.  Fix too long lines.
37652         (ipa_write_jump_function): Adjust for bits and m_vr being pointers and
37653         vr_known being removed.
37654         (ipa_read_jump_function): Use new setter functions to construct bits
37655         and vr parts of jump functions or set them to NULL.
37656         (write_ipcp_transformation_info): Adjust for bits being pointers.
37657         (read_ipcp_transformation_info): Likewise.
37658         (ipcp_update_bits): Likewise.  Fix excessively long lines a trailing
37659         space.
37660         Include gt-ipa-prop.h.
37661         * ipa-cp.c (propagate_bits_across_jump_function): Adjust for bits
37662         being pointers.
37663         (ipcp_store_bits_results): Likewise.
37664         (propagate_vr_across_jump_function): Adjust for m_vr being a pointer.
37665         Do not write to existing jump functions but use a temporary instead.
37667 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37669         PR c++/79681
37670         * fold-const.c (make_bit_field_ref): If orig_inner is COMPONENT_REF,
37671         attempt to use its first operand as BIT_FIELD_REF base.
37673 2017-03-01  Richard Biener  <rguenther@suse.de>
37675         PR middle-end/79721
37676         * tree-chrec.c (chrec_evaluate): Perform computation of Newtons
37677         interpolating formula in wrapping arithmetic.
37678         (chrec_apply): Convert chrec_evaluate return value to wanted type.
37680 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37682         PR tree-optimization/79734
37683         * tree-vect-generic.c (expand_vector_condition): Optimize
37684         AVX512 vector boolean VEC_COND_EXPRs into bitwise operations.
37685         Handle VEC_COND_EXPR where comparison has different inner width from
37686         type's inner width.
37688 2017-02-28  Sandra Loosemore  <sandra@codesourcery.com>
37690         * doc/invoke.texi (ARC Options): Copy-edit to fix punctuation,
37691         markup, and similar issues.  Remove @opindex entries for things
37692         that aren't options.  Add missing -mmpy-option entries.
37694 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37696         PR tree-optimization/79737
37697         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): If bitlen is
37698         a multiple of BITS_PER_UNIT and !BYTES_BIG_ENDIAN, clear
37699         tmpbuf[byte_size - 1].  Call natice_encode_expr with byte_size - 1
37700         instead of byte_size.  Formatting fix.
37701         (shift_bytes_in_array_right): Formatting fix.
37703 2017-02-28  Eric Botcazou  <ebotcazou@adacore.com>
37705         PR target/79749
37706         * config/sparc/sparc.c (sparc_frame_pointer_required): Add missing
37707         condition on optimize for the leaf function test.
37709 2017-02-28  Martin Liska  <mliska@suse.cz>
37711         PR lto/79625
37712         * read-rtl-function.c (function_reader::handle_unknown_directive):
37713         Bail out when one uses -flto.
37715 2017-02-28  Martin Liska  <mliska@suse.cz>
37717         * common.opt: Replace space with tabular for options of <number>
37718         type.
37719         * config/i386/i386.opt: Show <number> value for
37720         -mlarge-data-threshold.
37721         * opts.c (print_filtered_help): Do not display number in hexadecimal
37722         format.
37724 2017-02-28  Martin Liska  <mliska@suse.cz>
37726         * common.opt: Fix --help=option -Q for options which are of
37727         an enum type.
37729 2017-02-28  Uros Bizjak  <ubizjak@gmail.com>
37731         * config/i386/i386.c (print_reg): Error out for values
37732         of 8-bit size in invalid integer register.
37734 2017-02-28  Martin Sebor  <msebor@redhat.com>
37736         PR tree-optimization/79691
37737         * passes.def (pass_all_optimizations_g): Enable pass_sprintf_length.
37739 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37741         PR target/79729
37742         * config/i386/i386.c (ix86_print_operand) <case 'R'>: Replace
37743         gcc_unreachable with output_operand_lossage.
37745 2017-02-28  Richard Biener  <rguenther@suse.de>
37747         PR tree-optimization/79740
37748         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Allow redundant
37749         inserts.
37750         (visit_nary_op): Insert the nary into the hashtable if we
37751         pattern-matched sth.
37752         * tree-ssa-pre.c (eliminate_insert): Robustify.
37754 2017-02-28  Richard Biener  <rguenther@suse.de>
37756         PR middle-end/79731
37757         * fold-const.c (decode_field_reference): Reject out-of-bound
37758         accesses.
37760 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37762         * config/i386/i386.c: Include intl.h.
37763         (ix86_option_override_internal): Use cond ? G_("...") : G_("...")
37764         instead of just cond ? "..." : "...".
37765         * config/nvptx/nvptx.c (nvptx_goacc_validate_dims): Likewise.
37766         * coverage.c (read_counts_file): Likewise.
37767         * omp-offload.c: Include intl.h.
37768         (oacc_loop_fixed_partitions): Use cond ? G_("...") : G_("...") instead
37769         of just cond ? "..." : "...".
37770         * gcov.c (read_count_file): Use cond ? N_("...") : N_("...") instead
37771         of just cond ? "..." : "...".
37773 2017-02-28  Richard Earnshaw  <rearnsha@arm.com>
37775         PR target/79742
37776         * config/arm/parsecpu.awk (gen_data): Set tuning target to 'tune for'
37777         entry, if present.
37778         * config/arm/arm-cpus.in (cortex-m0plus.small-multiply): Correct
37779         'tune for' CPU name.
37780         * config/arm/arm-cpu-data.h: Regenerated.
37782 2017-02-28  Richard Biener  <rguenther@suse.de>
37784         PR tree-optimization/79732
37785         * tree-inline.c (expand_call_inline): Do not shadow var.
37787 2017-02-28  Richard Biener  <rguenther@suse.de>
37789         PR tree-optimization/79723
37790         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Preserve
37791         address-space properly.
37793 2017-02-28  Thomas Schwinge  <thomas@codesourcery.com>
37795         * doc/optinfo.texi (Optimization groups): Fix option used for
37796         OPTGROUP_ALL.
37797         * doc/invoke.texi (-fopt-info): Document "omp".
37798         * dumpfile.h: Sort OPTGROUP_OMP before OPTGROUP_VEC.
37799         (OPTGROUP_ALL): Add OPTGROUP_OMP.
37800         * hsa-gen.c (pass_data_gen_hsail): Use OPTGROUP_OMP.
37801         * ipa-hsa.c (pass_data_ipa_hsa): Likewise.
37802         * omp-simd-clone.c (pass_data_omp_simd_clone): Likewise.
37804         * dumpfile.h (OPTGROUP_OPENMP): Rename to OPTGROUP_OMP.  Adjust
37805         all users.
37806         * dumpfile.c (optgroup_options): Instead of "openmp", associate
37807         OPTGROUP_OMP with "omp".
37809 2017-02-27  Pat Haugen  <pthaugen@us.ibm.com>
37811         PR target/79544
37812         * config/rs6000/rs6000-c.c (struct altivec_builtin_types): Use VSRAD
37813         for arithmetic shift of unsigned V2DI.
37815 2017-02-27  Claudiu Zissulescu  <claziss@synopsys.com>
37817         * config.gcc (arc*-): Clean up, use arc/big.h, arc/elf.h, and
37818         arc/linux.h headers.
37819         * config/arc/arc.h (TARGET_OS_CPP_BUILTINS): Remove.
37820         (LINK_SPEC): Likewise.
37821         (ARC_TLS_EXTRA_START_SPEC): Likewise.
37822         (EXTRA_SPECS): Likewise.
37823         (STARTFILE_SPEC): Likewise.
37824         (ENDFILE_SPEC): Likewise.
37825         (LIB_SPEC): Likewise.
37826         (TARGET_SDATA_DEFAULT): Likewise.
37827         (TARGET_MMEDIUM_CALLS_DEFAULT): Likewise.
37828         (MULTILIB_DEFAULTS): Likewise.
37829         (DWARF2_UNWIND_INFO): Likewise.
37830         * config/arc/big.h: New file.
37831         * config/arc/elf.h: Likewise.
37832         * config/arc/linux.h: Likewise.
37833         * config/arc/t-uClibc: Remove.
37835 2017-02-27  Bin Cheng  <bin.cheng@arm.com>
37837         PR tree-optimization/77536
37838         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): New function.
37839         (tree_transform_and_unroll_loop): Use above function to compute the
37840         estimated niter of unrolled loop and use it when scaling profile.
37841         Also use count info rather than frequency if it's non-zero.
37842         * tree-ssa-loop-manip.h niter_for_unrolled_loop(): New declaration.
37843         * tree-vect-loop.c (scale_profile_for_vect_loop): New function.
37844         (vect_transform_loop): Call above function.
37846 2017-02-27  Richard Biener  <rguenther@suse.de>
37848         PR tree-optimization/45397
37849         * tree-ssa-pre.c (eliminate_insert): Handle BIT_AND_EXPR.
37850         * tree-ssa-sccvn.c (valueized_wider_op): New helper.
37851         (visit_nary_op): Add pattern matching for CSEing sign-changed
37852         or truncated operations with wider ones.
37854 2017-02-27  Richard Biener  <rguenther@suse.de>
37856         PR tree-optimization/79690
37857         * tree-vect-stmts.c (vectorizable_store): Use vector type
37858         built from the DR with address-space.
37860 2017-02-26  Gerald Pfeifer  <gerald@pfeifer.com>
37862         * doc/invoke.texi (Optimize Options): Refine the description
37863         of asan-use-after-return.
37865 2017-02-25  Alan Modra  <amodra@gmail.com>
37867         PR rtl-optimization/79584
37868         * lra-constraints.c (base_to_reg): Reload ad->base, the entire
37869         base, not ad->base_term, the reg within base.  Remove assertion
37870         that ad->base == ad->base_term.  Replace gen_int_mode using
37871         bogus mode with const0_rtx.
37873 2017-02-25  Jakub Jelinek  <jakub@redhat.com>
37875         PR middle-end/79396
37876         * tree-eh.c (operation_could_trap_p, stmt_could_throw_1_p): Handle
37877         FMA_EXPR like tcc_binary or tcc_unary.
37879         * tree-ssa-loop-niter.c (number_of_iterations_exit): Simplify warning.
37881         PR debug/77589
37882         * dwarf2out.c (struct dw_loc_list_struct): Add noted_variable_value
37883         bitfield.
37884         (size_of_loc_descr): Handle DW_OP_GNU_variable_value.
37885         (output_loc_operands): Handle DW_OP_call_ref and
37886         DW_OP_GNU_variable_value.
37887         (struct variable_value_struct): New type.
37888         (struct variable_value_hasher): Likewise.
37889         (variable_value_hash): New variable.
37890         (string_types): Remove.
37891         (copy_loc_descr): New function.
37892         (add_loc_descr_to_each): Clarify comment.  Use copy_loc_descr.
37893         (prepend_loc_descr_to_each): New function.
37894         (add_loc_list): Fix comment typo.  Use prepend_loc_descr_to_each
37895         instead of add_loc_descr_to_each if the first argument is single
37896         location list and the second has multiple.
37897         (resolve_args_picking_1): Handle DW_OP_GNU_variable_value.
37898         (loc_list_from_tree_1): For early_dwarf, emit DW_OP_GNU_variable_value
37899         when looking for variable value which doesn't have other location info.
37900         (loc_list_from_tree): Formatting fix.
37901         (gen_array_type_die): Simplify DW_AT_string_length handling.
37902         (adjust_string_types): Remove.
37903         (gen_subprogram_die): Don't call adjust_string_types nor test/set
37904         string_types.  Call resolve_variable_values.
37905         (prune_unused_types_walk_loc_descr): Handle DW_OP_GNU_variable_value.
37906         (resolve_addr_in_expr): Likewise.  Add A argument.
37907         (copy_deref_exprloc): Remove deref argument.  Adjust for the
37908         original expression being DW_OP_GNU_variable_value with optionally
37909         DW_OP_stack_value after it instead of DW_OP_call4 with DW_OP_deref
37910         optionally after it.
37911         (optimize_string_length): Rework for DW_OP_GNU_variable_value.
37912         (resolve_addr): Adjust optimize_string_length and resolve_addr_in_expr
37913         callers.  Set remove_AT_byte_size if removing DW_AT_string_length.
37914         (variable_value_hasher::hash, variable_value_hasher::equal): New
37915         methods.
37916         (resolve_variable_value_in_expr, resolve_variable_value,
37917         resolve_variable_values, note_variable_value_in_expr,
37918         note_variable_value): New functions.
37919         (dwarf2out_early_finish): Call note_variable_value on all toplevel
37920         DIEs.
37922 2017-02-24  Jakub Jelinek  <jakub@redhat.com>
37924         PR c/79677
37925         * opts.h (handle_generated_option): Add GENERATED_P argument.
37926         * opts-common.c (handle_option): Adjust function comment.
37927         (handle_generated_option): Add GENERATED_P argument, pass it to
37928         handle_option.
37929         (control_warning_option): Pass false to handle_generated_option
37930         GENERATED_P.
37931         * opts.c (maybe_default_option): Pass true to handle_generated_option
37932         GENERATED_P.
37933         * optc-gen.awk: Likewise.
37935 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37937         * config/sh/sh.md (tstsi_t): If operands[0] is a SUBREG instead of
37938         a REG, look at the REG it is a SUBREG of.
37939         (splitter for cmpeqsi_t): Ditto.
37941 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37943         * config/pa/pa.c (pa_combine_instructions): Do not share RTL.  Make
37944         the special USEs with the pattern of the insn, not the insn itself.
37946 2017-02-24  Matthew Fortune  <matthew.fortune@imgtec.com>
37948         PR target/79473
37949         * doc/invoke.texi: Document -mload-store-pairs.
37951 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37952             Sandra Loosemore  <sandra@codesourcery.com>
37954         * config/nios2/nios2.c (nios2_simple_const_p): Returns false if the
37955         argument isn't a CONST_INT.
37956         (nios2_alternate_compare_const): Assert op is a CONST_INT.
37957         (nios2_valid_compare_const_p): Assert op is a CONST_INT.
37958         (nios2_validate_compare): Bypass alternate compare logic if *op2
37959         is not a CONST_INT.
37960         (ldstwm_operation_p): Return false if first_base is not a REG or
37961         if first_offset is not a CONST_INT.
37963 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37965         * config/cris/cris.md: Use correct operand in a define_peephole2.
37967 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37969         * config/c6x/c6x.c (predicate_insn): Do not incorrectly share RTL.
37971 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37973         * config/arc/arc.c (arc_ccfsm_advance): Only take the PATTERN of
37974         this_insn if it is an INSN or JUMP_INSN.
37975         (force_offsettable): Look at base, not at addr.
37976         * config/arc/predicates.md (brcc_nolimm_operator): Don't call INTVAL
37977         on things that aren't necessarily CONST_INTs.
37979 2017-02-24  Uros Bizjak  <ubizjak@gmail.com>
37981         * doc/invoke.texi (x86 Options, -mfpmath=sse): Mention that
37982         -mfpmath=sse is the default also for x86-32 targets with SSE2
37983         instruction set when @option{-ffast-math} is enabled
37985 2017-02-24  Jeff Law  <law@redhat.com>
37987         PR rtl-optimizatoin/79286
37988         * ira.c (update_equiv_regs): Drop may_trap_p exception to
37989         dominance test.
37991 2017-02-24  Richard Biener  <rguenther@suse.de>
37993         PR tree-optimization/79389
37994         * gimple-ssa-split-paths.c (is_feasible_trace): Properly skip
37995         debug insns.
37997 2017-02-24  Aldy Hernandez  <aldyh@redhat.com>
37999         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update
38000         function comment to reflect reality.
38001         (loop_exits_before_overflow): Fix typo in function description.
38003 2017-02-24  Richard Biener  <rguenther@suse.de>
38005         PR tree-optimization/79389
38006         * gimple-ssa-split-paths.c (is_feasible_trace): Verify more
38007         properly that a threading opportunity exists.  Detect conditional
38008         copy/constant propagation opportunities.
38010 2017-02-23  Eric Botcazou  <ebotcazou@adacore.com>
38012         * config/visium/visium.md (type): Add trap.
38013         (b): New mode attribute.
38014         (*btst): Rename into...
38015         (*btst<mode>): ...this and adjust.
38016         (*cbranchsi4_btst_insn): Rename into...
38017         (*cbranch<mode>4_btst_insn): ...this and adjust.
38018         (trap): New define_insn.
38020 2017-02-23  Jakub Jelinek  <jakub@redhat.com>
38022         PR tree-optimization/79389
38023         * ifcvt.c (struct noce_if_info): Add rev_cond field.
38024         (noce_reversed_cond_code): New function.
38025         (noce_emit_store_flag): Use rev_cond if non-NULL instead of
38026         reversed_comparison_code.  Formatting fix.
38027         (noce_try_store_flag): Test rev_cond != NULL in addition to
38028         reversed_comparison_code.
38029         (noce_try_store_flag_constants): Likewise.
38030         (noce_try_store_flag_mask): Likewise.
38031         (noce_try_addcc): Use rev_cond if non-NULL instead of
38032         reversed_comparison_code.
38033         (noce_try_cmove_arith): Likewise.  Formatting fixes.
38034         (noce_try_minmax, noce_try_abs): Clear rev_cond.
38035         (noce_find_if_block): Initialize rev_cond.
38036         (find_cond_trap): Call noce_get_condition with then_bb == trap_bb
38037         instead of false as last argument never attempt to reverse it
38038         afterwards.
38040 2017-02-23  Bin Cheng  <bin.cheng@arm.com>
38042         PR tree-optimization/79663
38043         * tree-predcom.c (combine_chains): Process refs in reverse order
38044         only for ZERO length chains, and add explaining comment.
38046 2017-02-23  Jeff Law  <law@redhat.com>
38048         PR tree-optimization/79578
38049         * tree-ssa-dse.c (clear_bytes_written_by): Use OEP_ADDRESS_OF
38050         in call to operand_equal_p.
38052 2017-01-23  Dominique d'Humieres  <dominiq@lps.ens.fr>
38054         PR target/71017
38055         * config/i386/cpuid.h: Fix another undefined behavior.
38057 2017-02-23  Richard Biener  <rguenther@suse.de>
38059         PR tree-optimization/79683
38060         * tree-vect-stmts.c (vect_analyze_stmt): Do not overwrite
38061         vector types for data-refs.
38063 2017-02-23  Martin Liska  <mliska@suse.cz>
38065         * params.def (PARAM_MIN_NONDEBUG_INSN_UID): Change default to 0.
38067 2017-02-23  Jakub Jelinek  <jakub@redhat.com>
38069         PR middle-end/79665
38070         * internal-fn.c (get_range_pos_neg): Moved to ...
38071         * tree.c (get_range_pos_neg): ... here.  No longer static.
38072         * tree.h (get_range_pos_neg): New prototype.
38073         * expr.c (expand_expr_real_2) <case TRUNC_DIV_EXPR>: If both arguments
38074         are known to be in between 0 and signed maximum inclusive, try to
38075         expand both unsigned and signed divmod and use the cheaper one from
38076         those.
38078 2017-02-22  Jeff Law  <law@redhat.com>
38080         PR tree-optimization/79578
38081         * tree-ssa-dse.c (clear_bytes_written_by): Use operand_equal_p
38082         to compare base operands.
38084 2017-02-22  Segher Boessenkool  <segher@kernel.crashing.org>
38086         PR target/79211
38087         * config/rs6000/rs6000.md (*fsel<SFDF:mode><SFDF2:mode>4): Use
38088         gpc_reg_operand instead of fpr_reg_operand.
38090 2017-02-22  Sameera Deshpande  <sameera.deshpande@imgtec.com>
38092         * config/mips/mips.c (mips_return_in_memory): Force FP
38093         vector types to be returned in memory for o32 ABI.
38095 2017-02-22  Jakub Jelinek  <jakub@redhat.com>
38097         * dwarf2out.c (gen_variable_die): For -gdwarf-5, use DW_TAG_variable
38098         instead of DW_TAG_member for static data member declarations and don't
38099         set no_linkage_name for static inline data members.
38100         (gen_member_die): For -gdwarf-5 don't change DW_TAG_variable
38101         to DW_TAG_member.
38103 2017-02-22  Martin Liska  <mliska@suse.cz>
38105         * doc/invoke.texi: Replace inequality signs with square brackets
38106         for -Wnormalized.
38108 2017-02-22  Matthew Fortune  <matthew.fortune@imgtec.com>
38110         PR target/78660
38111         * lra-constraints.c (simplify_operand_subreg): Handle
38112         WORD_REGISTER_OPERATIONS targets.
38114 2017-02-22  Jakub Jelinek  <jakub@redhat.com>
38116         PR target/70465
38117         * reg-stack.c (emit_swap_insn): Treat (float_extend:?F (mem:?F))
38118         and (const_double:?F) like (mem:?F) for the purpose of fxch %st(1)
38119         elimination by swapping fld*.
38121 2017-02-22  Richard Biener  <rguenther@suse.de>
38123         PR tree-optimization/79673
38124         * tree-ssa-pre.c (compute_avail): Use wide_int_to_tree to
38125         convert the [TARGET_]MEM_REF offset INTEGER_CST, scrapping off
38126         irrelevant address-space qualifiers and avoiding a
38127         ADDR_SPACE_CONVERT_EXPR from fold_convert.
38129 2017-02-22  Richard Biener  <rguenther@suse.de>
38131         PR tree-optimization/79666
38132         * tree-vrp.c (extract_range_from_binary_expr_1): Make sure
38133         to not symbolically negate if that may introduce undefined
38134         overflow.
38136 2017-02-22  Martin Liska  <mliska@suse.cz>
38138         PR lto/79587
38139         * data-streamer-in.c (streamer_read_gcov_count): Remove assert.
38140         * data-streamer-out.c (streamer_write_gcov_count_stream):
38141         Likewise.
38142         * value-prof.c (stream_out_histogram_value): Make assert more
38143         precise based on type of counter.
38145 2017-02-21  Uros Bizjak  <ubizjak@gmail.com>
38147         PR target/79593
38148         * config/i386/i386.md (standard_x87sse_constant_load splitter):
38149         Use nonimmediate_operand instead of memory_operand for operand 1.
38150         (float-extend standard_x87sse_constant_load splitter): Ditto.
38152 2017-02-21  Jeff Law  <law@redhat.com>
38154         PR tree-optimization/79621
38155         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior): Ignore
38156         blocks with edges to themselves.
38158 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
38160         PR target/79633
38161         * tree-chkp-opt.c (chkp_optimize_string_function_calls): Use
38162         is_gimple_call instead of comparing gimple_code with GIMPLE_CALL.
38163         Use gimple_call_builtin_p.
38165         PR target/79570
38166         * sel-sched.c (moveup_expr_cached): Don't call sel_bb_head
38167         on temporarily removed DEBUG_INSNs.
38169         PR tree-optimization/79649
38170         * tree-loop-distribution.c (classify_partition): Give up on
38171         non-generic address space loads/stores.
38173 2017-02-21  Aldy Hernandez  <aldyh@redhat.com>
38175         * doc/loop.texi (Loop manipulation): Remove nonexistent
38176         tree_ssa_loop_version from the documentation.
38177         * cfgloopmanip.c (loop_version): Document CONDITION_BB argument.
38179 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
38181         PR target/79494
38182         * config/i386/i386.c (ix86_expand_split_stack_prologue): Call
38183         make_reg_eh_region_note_nothrow_nononlocal on call_insn.
38184         * config/rs6000/rs6000.c: Include except.h.
38185         (rs6000_expand_split_stack_prologue): Call
38186         make_reg_eh_region_note_nothrow_nononlocal on the call insn.
38188 2017-02-21  Martin Jambor  <mjambor@suse.cz>
38190         PR lto/79579
38191         * ipa-prop.c (ipa_prop_write_jump_functions): Bail out if no edges
38192         have been analyzed.
38194 2017-02-21  Martin Jambor  <mjambor@suse.cz>
38196         * common.opt (-fipa-cp-alignment): Mark as ignored and preserved
38197         for backward compatibility only.
38198         * doc/invoke.texi (Option Summary): Remove all references to
38199         -fipa-cp-alignment.
38201 2017-02-21  Matthew Fortune  <matthew.fortune@imgtec.com>
38203         PR target/78660
38204         Revert:
38205         2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38207         * lra-constraints.c (curr_insn_transform): Handle
38208         WORD_REGISTER_OPERATIONS requirements when reloading SUBREGs.
38210 2017-02-21  Martin Liska  <mliska@suse.cz>
38212         * config/i386/i386.opt: Replace -masm-dialect with -masm.
38214 2017-02-21  Thomas Schwinge  <thomas@codesourcery.com>
38216         PR translation/79638
38217         * config/nvptx/nvptx.c (ENTRY_TEMPLATE): Single out "%ntid.y".
38219 2017-02-21  Eric Botcazou  <ebotcazou@adacore.com>
38221         PR ada/67205
38222         * config/arm/arm.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): Define.
38223         (arm_function_ok_for_sibcall): Return false for an indirect call by
38224         descriptor if all the argument registers are used.
38225         (arm_relayout_function): Use FUNCTION_ALIGNMENT macro to adjust the
38226         alignment of the function.
38228 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
38230         PR tree-optimization/61441
38231         * simplify-rtx.c (simplify_const_unary_operation): For
38232         -fsignaling-nans and sNaN operand, return NULL_RTX rather than
38233         the sNaN unmodified.
38235 2017-02-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
38237         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
38238         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
38239         instead of SYSTEM_HEADER_DIR.
38241 2017-02-20  Gerald Pfeifer  <gerald@pfeifer.com>
38242             Martin LiÅ¡ka  <mliska@suse.cz>
38244         * doc/invoke.texi (use-after-scope-direct-emission-threshold):
38245         Fix typos and grammar, use active voice, and clarify.
38247 2017-02-20  Marek Polacek  <polacek@redhat.com>
38249         PR middle-end/79537
38250         * gimplify.c (gimplify_expr): Handle unused *&&L;.
38252         PR sanitizer/79558
38253         * ubsan.c (ubsan_type_descriptor): Check if TYPE_MAX_VALUE is null.
38255 2017-02-20  Jakub Jelinek  <jakub@redhat.com>
38257         PR target/79568
38258         * config/i386/i386.c (ix86_expand_builtin): Handle
38259         OPTION_MASK_ISA_AVX512VL and OPTION_MASK_ISA_64BIT in
38260         ix86_builtins_isa[fcode].isa as a requirement of those
38261         flags and any other flag in the bitmask.
38262         (ix86_init_mmx_sse_builtins): Use 0 instead of
38263         ~OPTION_MASK_ISA_64BIT as mask.
38264         * config/i386/i386-builtin.def (__builtin_ia32_rdtsc,
38265         __builtin_ia32_rdtscp, __builtin_ia32_pause, __builtin_ia32_bsrsi,
38266         __builtin_ia32_rdpmc, __builtin_ia32_rolqi, __builtin_ia32_rolhi,
38267         __builtin_ia32_rorqi, __builtin_ia32_rorhi): Likewise.
38269 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38271         PR target/78012
38272         * lra-constraints.c (split_reg): Check requested split mode
38273         is supported by the register.
38275 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38277         * lra-constraints.c (simplify_operand_subreg): Remove early
38278         return false.
38280 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38282         PR target/78660
38283         * lra-constraints.c (curr_insn_transform): Tighten condition
38284         for converting SUBREG reloads from OP_OUT to OP_INOUT.
38286 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38288         PR target/78660
38289         * lra-constraints.c (curr_insn_transform): Handle
38290         WORD_REGISTER_OPERATIONS requirements when reloading SUBREGs.
38292 2017-02-19  Uros Bizjak  <ubizjak@gmail.com>
38294         Revert:
38295         2016-05-30  Uros Bizjak  <ubizjak@gmail.com>
38297         * config/i386/sync.md (mfence_nosse): Use "lock orl $0, -4(%esp)".
38299 2017-02-19  Jonathan Wakely  <jwakely@redhat.com>
38301         PR c++/69523
38302         * doc/invoke.texi (C++ Dialect Options) [-Wliteral-suffix]: Update
38303         description.
38305 2017-02-19  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
38307         * gimple-pretty-print.c (dump_ternary_rhs): Adjust gimple dump format
38308         for FMA_EXPR.
38310 2017-02-18  Jakub Jelinek  <jakub@redhat.com>
38312         * final.c (last_columnnum, override_columnnum): New variables.
38313         (final_start_function): Set last_columnnum, pass it to begin_prologue
38314         hook and pass 0 to dwarf2out_begin_prologue.
38315         (final_scan_insn): Update override_columnnum.  Pass last_columnnum
38316         to source_line debug hook.
38317         (notice_source_line): Compute last_columnnum and for debug_column_info
38318         return true on column changes.
38319         * debug.h (struct gcc_debug_hooks): Add column argument to
38320         source_line and begin_prologue hooks.
38321         (debug_nothing_int_charstar_int_bool): Remove prototype.
38322         (debug_nothing_int_int_charstar,
38323         debug_nothing_int_int_charstar_int_bool): New prototypes.
38324         (dwarf2out_begin_prologue): Add column argument.
38325         * debug.c (do_nothing_debug_hooks): Adjust source_line and
38326         begin_prologue hooks.
38327         (debug_nothing_int_charstar_int_bool): Remove.
38328         (debug_nothing_int_int_charstar,
38329         debug_nothing_int_int_charstar_int_bool): New functions.
38330         * dwarf2out.c (dwarf2out_begin_prologue): Add column argument, pass it
38331         through to dwarf2out_source_line.
38332         (dwarf2_lineno_debug_hooks): Adjust begin_prologue hook.
38333         (dwarf2out_source_line): Add column argument, emit it if requested.
38334         * sdbout.c (sdbout_source_line, sdbout_begin_prologue): Add column
38335         arguments.
38336         * xcoffout.h (xcoffout_begin_prologue, xcoffout_source_line): Likewise.
38337         * xcoffout.c (xcoffout_begin_prologue, xcoffout_source_line): Likewise.
38338         * vmsdbgout.c (vmsdbgout_begin_prologue): Add column argument, pass it
38339         through to dwarf2out_begin_prologue.
38340         (vmsdbgout_source_line): Add column argument, pass it through to
38341         dwarf2out_source_line.
38342         * dbxout.c (dbxout_begin_prologue): Add column argument, adjust
38343         dbxout_source_line caller.
38344         (dbxout_source_line): Add column argument.
38346         * common.opt (gno-column-info, gcolumn-info): New options.
38347         * dwarf2out.c (dwarf2_lineno_debug_hooks): Formatting fix.
38348         (check_die): Also test for multiple DW_AT_decl_column attributes.
38349         (add_src_coords_attributes, dwarf2out_imported_module_or_decl_1): Add
38350         DW_AT_decl_column if requested.
38351         (gen_subprogram_die): Compare and/or add also DW_AT_decl_column
38352         if requested.
38353         (gen_variable_die): Likewise.
38354         (add_call_src_coords_attributes): Add DW_AT_call_column if requested.
38355         * doc/invoke.texi (-gcolumn-info, -gno-column-info): Document.
38357         PR target/79569
38358         * config/i386/i386.opt (m3dnowa): Replace Undocumented with Report.
38359         * common/config/i386/i386-common.c (OPTION_MASK_ISA_3DNOW_A_SET): Define.
38360         (ix86_handle_option): Handle OPT_m3dnowa.
38361         * doc/invoke.texi (-m3dnowa): Document.
38362         * doc/extend.texi (__builtin_ia32_pmulhuw, __builtin_ia32_pf2iw): Use
38363         -m3dnowa instead of -m3dnow -march=athlon.
38365         PR target/79559
38366         * config/i386/i386.c (ix86_print_operand): Use output_operand_lossage
38367         instead of gcc_assert for K, r and R code checks.  Formatting fixes.
38369 2017-02-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
38371         PR target/79261
38372         * config/rs6000/rs6000.c (rs6000_expand_ternop_builtin): Add
38373         support for CODE_FOR_vsx_xxpermdi_v2d[fi]_be.
38374         * config/rs6000/rs6000.md (reload_gpr_from_vsx<mode>): Call
38375         generator for vsx_xxpermdi_<mode>_be.
38376         * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Remove logic to
38377         force big-endian semantics.
38378         (vsx_xxpermdi_<mode>_be): New define_expand with same
38379         implementation as previous version of vsx_xxpermdi_<mode>.
38381 2017-02-17  Jakub Jelinek  <jakub@redhat.com>
38383         PR tree-optimization/79327
38384         * gimple-ssa-sprintf.c (format_integer): Remove likely_adjust
38385         variable, its initialization and use.
38387 2017-02-17  Julia Koval  <julia.koval@intel.com>
38389         * common/config/i386/i386-common.c (OPTION_MASK_ISA_RDPID_SET): New.
38390         (OPTION_MASK_ISA_PKU_UNSET): New.
38391         (ix86_handle_option): Handle -mrdpid.
38392         * config/i386/cpuid.h (bit_RDPID): New.
38393         * config/i386/driver-i386.c (host_detect_local_cpu):
38394         Detect RDPID feature.
38395         * config/i386/i386-builtin.def (__builtin_ia32_rdpid): New.
38396         * config/i386/i386-c.c (ix86_target_macros_internal):
38397         Handle RDPID flag.
38398         * config/i386/i386.c (ix86_target_string): Add -mrdpid to isa2_opts.
38399         (ix86_valid_target_attribute_inner_p): Add "rdpid".
38400         (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
38401         * config/i386/i386.h (TARGET_RDPID, TARGET_RDPID_P): New.
38402         * config/i386/i386.md (define_insn "rdpid"): New.
38403         * config/i386/i386.opt Add -mrdpid.
38404         * config/i386/immintrin.h (_rdpid_u32): New.
38406 2017-02-17  Vladimir Makarov  <vmakarov@redhat.com>
38408         PR rtl-optimization/79541
38409         * lra-constraints.c (curr_insn_transform): Remove wrong asm insn
38410         instead of transforming it into USE.
38412 2017-02-17  Segher Boessenkool  <segher@kernel.crashing.org>
38414         * config/rs6000/rs6000.md (extendsfdf2): Remove default arguments.
38415         If HONOR_SNANS (SFmode) force the input to a register.
38416         (*extendsfdf2_fpr): Add !HONOR_SNANS (SFmode) condition.
38417         (*extendsfdf2_snan): New pattern, used when using SNaNs; it generates
38418         an frsp or similar insn.
38420 2017-02-17  Martin Liska  <mliska@suse.cz>
38422         PR rtl-optimization/79577
38423         * params.def (selsched-max-sched-times): Increase minimum to 1.
38425 2017-02-17  Martin Liska  <mliska@suse.cz>
38427         PR rtl-optimization/79574
38428         * gcse.c (want_to_gcse_p): Prevent integer overflow.
38430 2017-02-17  Martin Liska  <mliska@suse.cz>
38432         PR tree-optimization/79529
38433         * tree-ssa-loop-unswitch.c (is_maybe_undefined): Use
38434         ssa_defined_default_def_p to handle cases which are implicitly
38435         defined.
38436         * tree-ssa.c (ssa_defined_default_def_p): New function.
38437         (ssa_undefined_value_p): Use ssa_defined_default_def_p to handle cases
38438         which are implicitly defined.
38439         * tree-ssa.h (ssa_defined_default_def_p): Declare.
38441 2017-02-17  Richard Biener  <rguenther@suse.de>
38443         PR middle-end/79576
38444         * params.def (max-ssa-name-query-depth): Limit to 10.
38446 2017-02-17  Richard Biener  <rguenther@suse.de>
38448         PR tree-optimization/79552
38449         * tree-ssa-structalias.c (visit_loadstore): Properly verify
38450         default defs.
38452 2017-02-17  Richard Biener  <rguenther@suse.de>
38454         PR bootstrap/79567
38455         * genmatch.c (output_line_directive): Handle DIR_SEPARATOR_2.
38457 2017-02-17  Marek Polacek  <polacek@redhat.com>
38459         PR middle-end/79536
38460         * fold-const.c (fold_negate_expr_1): Renamed from fold_negate_expr.
38461         (fold_negate_expr): New wrapper.
38463 2017-02-16  Sandra Loosemore  <sandra@codesourcery.com>
38465         * doc/invoke.texi (C++ Dialect Options) [-Wno-non-template-friend]: 
38466         Correct terminology and de-emphasize pre-standard behavior.
38468 2017-02-16  Alan Modra  <amodra@gmail.com>
38470         PR rtl-optimization/79286
38471         * ira.c (def_dominates_uses): New function.
38472         (update_equiv_regs): Don't create an equivalence for insns that
38473         may trap where the register def does not dominate the use.
38475 2017-02-16  Vladimir Makarov  <vmakarov@redhat.com>
38477         PR rtl-optimization/78127
38478         * lra.c (lra): Call lra_eliminate before finish the loop after
38479         lra_constraint.
38481 2017-02-16  Richard Biener  <rguenther@suse.de>
38483         * graphite.h: Do not include isl/isl_val_gmp.h, instead include
38484         isl/isl_val.h.
38485         * graphite-isl-ast-to-gimple.c (gmp_cst_to_tree): Remove.
38486         (gcc_expression_from_isl_expr_int): Use generic isl_val interface.
38487         * graphite-sese-to-poly.c: Do not include isl/isl_val_gmp.h.
38488         (isl_val_int_from_wi): New function.
38489         (extract_affine_gmp): Rename to ...
38490         (extract_affine_wi): ... this, take a widest_int.
38491         (extract_affine_int): Just wrap extract_affine_wi.
38492         (add_param_constraints): Use isl_val_int_from_wi.
38493         (add_loop_constraints): Likewise, and extract_affine_wi.
38495 2017-02-15  Jeff Law  <law@redhat.com>
38497         PR middle-end/79521
38498         * ira-costs.c (scan_one_insn): Check have_regs_of_mode before calling
38499         ira_init_register_move_cost_if_necessary.
38501 2017-02-15  Martin Sebor  <msebor@redhat.com>
38503         PR middle-end/32003
38504         * doc/invoke.texi (-fdump-final-insns): Replace option accidentally
38505         removed in a prior commit.
38507 2017-02-15  Bin Cheng  <bin.cheng@arm.com>
38509         PR tree-optimization/79347
38510         * tree-vect-loop-manip.c (vect_do_peeling): Maintain profile
38511         counters during peeling.
38513 2017-02-15  Thomas Schwinge  <thomas@codesourcery.com>
38515         * Makefile.in (site.exp): Remove "set ISLVER".
38517 2017-02-15  Jakub Jelinek  <jakub@redhat.com>
38519         PR target/79487
38520         * real.c (real_from_integer): Call real_convert even for decimal.
38522 2017-02-15  Dominik Vogt  <vogt@linux.vnet.ibm.com>
38524         PR target/79421
38525         * config/s390/s390.c: define TARGET_CUSTOM_FUNCTION_DESCRIPTORS.
38527 2017-02-14  Andrew Pinski  <apinski@cavium.com>
38529         * config/aarch64/aarch64-cores.def (thunderx2t99): Move to under 'C"
38530         cores and change the partno/implementer to be correct.
38531         (thunderx2t99p1): New core which replaces thunderx2t99 and still has
38532         the 'B" as the implementer.
38533         * config/aarch64/aarch64-tune.md: Regenerate.
38535 2017-02-14  Carl Love  <cel@us.ibm.com>
38537         * config/rs6000/rs6000.c: Add case statement entry to make the
38538         xvcvuxdsp built-in argument unsigned.
38539         * config/rs6000/vsx.md: Fix the source and return operand types so they
38540         match the instruction definitions from the ISA document.  Fix typo
38541         in the instruction generation for the (define_insn "vsx_xvcvuxdsp"
38542         statement.
38544 2017-02-14  Vladimir Makarov  <vmakarov@redhat.com>
38546         PR target/79282
38547         * lra-int.h (struct lra_operand_data, struct lra_insn_reg): Add
38548         member early_clobber_alts.
38549         * lra-lives.c (reg_early_clobber_p): New.
38550         (process_bb_lives): Use it.
38551         * lra.c (new_insn_reg): New arg early_clobber_alts.  Use it.
38552         (debug_operand_data): Initialize early_clobber_alts.
38553         (setup_operand_alternative): Set up early_clobber_alts.
38554         (collect_non_operand_hard_regs): Ditto.  Pass early clobber
38555         alternatives to new_insn_reg.
38556         (add_regs_to_insn_regno_info): Add arg early_clobber_alts.  Use
38557         it.
38558         (lra_update_insn_regno_info): Pass the new arg.
38560 2017-02-14  Jakub Jelinek  <jakub@redhat.com>
38562         PR middle-end/79505
38563         * omp-offload.c (free_oacc_loop): Release loop->ifns vector.
38564         (new_oacc_loop_raw): Don't clear already cleared fields.
38566         PR target/79481
38567         * config/i386/avx512pfintrin.h (_mm512_prefetch_i32gather_pd,
38568         _mm512_prefetch_i32gather_ps, _mm512_prefetch_i64gather_pd,
38569         _mm512_prefetch_i64gather_ps): New inline functions and macros.
38571 2017-02-14  Uros Bizjak  <ubizjak@gmail.com>
38573         PR target/79495
38574         * config/i386/i386.md (*movxf_internal): Add (o,rC) alternative.
38576 2017-02-14  H.J. Lu  <hongjiu.lu@intel.com>
38578         PR target/79498
38579         * config/i386/i386.c (timode_scalar_chain::convert_insn): Insert
38580         the extra instruction to the right place to store 128-bit constant
38581         when needed.
38583 2017-02-14  Martin Sebor  <msebor@redhat.com>
38585         PR middle-end/79448
38586         * gimple-ssa-sprintf.c (format_directive): Avoid issuing INT_MAX
38587           warning for strings of unknown length.
38589 2017-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
38591         * config.gcc (supported_defaults) [powerpc*-*-*]: Update.
38593 2017-02-14  Jeff Law  <law@redhat.com>
38595         PR target/79404
38596         * ira-costs.c (scan_one_insn): Initialize register move costs
38597         for pseudos seen in USE/CLOBBER insns.
38599         PR tree-optimization/79095
38600         * tree-vrp.c (extract_range_from_binary_expr_1): For EXACT_DIV_EXPR,
38601         if the numerator has the range ~[0,0] make the resultant range ~[0,0].
38602         (extract_range_from_binary_expr): For MINUS_EXPR with no derived range,
38603         if the operands are known to be not equal, then the resulting range
38604         is ~[0,0].
38605         (intersect_ranges): If the new range is ~[0,0] and the old range is
38606         wide, then prefer ~[0,0].
38607         * tree-vrp.c (overflow_comparison_p_1): New function.
38608         (overflow_comparison_p): New function.
38609         * tree-vrp.c (register_edge_assert_for_2): Register additional asserts
38610         if NAME is used in an overflow test.
38611         (vrp_evaluate_conditional_warnv_with_ops): If the ops represent an
38612         overflow check that can be expressed as an equality test, then adjust
38613         ops to be that equality test.
38615 2017-02-14  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
38617         * config/s390/s390-builtin-types.def: Remove flags argument.
38618         * config/s390/s390.c (s390_init_builtins): Likewise.
38620 2017-02-14  Martin Liska  <mliska@suse.cz>
38622         * tree-ssa-loop-unswitch.c (hoist_guard): Release get_loop_body
38623         vector.  Fix trailing white spaces.
38625 2017-02-14  James Greenhalgh  <james.greenhalgh@arm.com>
38627         * config/aarch64/aarch64.c (aarch64_simd_container_mode): Handle
38628         HFmode.
38630 2017-02-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38632         PR rtl-optimization/68664
38633         * config/arm/arm.c (arm_sched_can_speculate_insn):
38634         New function.  Declare prototype.
38635         (TARGET_SCHED_CAN_SPECULATE_INSN): Define.
38637 2017-02-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38639         PR rtl-optimization/68664
38640         * config/aarch64/aarch64.c (aarch64_sched_can_speculate_insn):
38641         New function.
38642         (TARGET_SCHED_CAN_SPECULATE_INSN): Define.
38644 2017-02-14  Amit Pawar  <amit.pawar@amd.com>
38646         * config/i386/i386.c (znver1_cost): Fix the alignment for function and
38647         max skip bytes for function, loop and jump.
38649 2017-02-14  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
38651         * gimple-pretty-print.c (dump_unary_rhs): Adjust dump format for
38652         ABS_EXPR for gimple dump.
38654 2017-02-14  Jakub Jelinek  <jakub@redhat.com>
38656         PR target/79462
38657         * config/sh/sh.c (expand_cbranchdi4): Don't clear operands[4].
38659         PR tree-optimization/79408
38660         * tree-vrp.c (simplify_div_or_mod_using_ranges): Handle also the
38661         case when on TRUNC_MOD_EXPR op0 is INTEGER_CST.
38662         (simplify_stmt_using_ranges): Call simplify_div_or_mod_using_ranges
38663         also if rhs1 is INTEGER_CST.
38665 2017-02-14  Richard Biener  <rguenther@suse.de>
38667         PR middle-end/79432
38668         * tree-into-ssa.c (insert_phi_nodes): When the function can
38669         have abnormal edges rewrite SSA names with broken use-def
38670         dominance out of SSA and register them for PHI insertion.
38672 2017-02-13  Martin Sebor  <msebor@redhat.com>
38674         PR middle-end/79496
38675         * gimple-ssa-sprintf.c (pass_sprintf_length::handle_gimple_call): Avoid
38676         clearing info.nowrite flag when snprintf size argument is a range.
38678 2017-02-13  Jakub Jelinek  <jakub@redhat.com>
38680         * cprop.c (cprop_jump): Add missing space in string literal.
38681         * tree-ssa-structalias.c (rewrite_constraints): Likewise.
38682         (get_constraint_for_component_ref): Likewise.
38683         * df-core.c (df_worklist_dataflow_doublequeue): Likewise.
38684         * tree-outof-ssa.c (insert_partition_copy_on_edge): Likewise.
38685         * lra-constraints.c (process_alt_operands): Likewise.
38686         * ipa-inline.c (inline_small_functions): Likewise.
38687         * tree-ssa-sccvn.c (visit_reference_op_store): Likewise.
38688         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
38689         * trans-mem.c (diagnose_tm_1_op): Likewise.
38690         * omp-grid.c (grid_find_single_omp_among_assignments): Likewise.
38691         (grid_parallel_clauses_gridifiable): Likewise.
38693         * config/nvptx/mkoffload.c (process): Add space in between
38694         , and %d.
38696         * config/i386/i386.h (REG_CLASS_NAMES): Add , in between
38697         "MOD4_SSE_REGS" and "ALL_REGS".
38699         * spellcheck.c (test_data): Add , in between "foo" and "food".
38701 2017-02-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
38703         PR target/79449
38704         * config/rs6000/rs6000.c (expand_block_compare): Make sure runtime
38705         boundary crossing check and subsequent code generation agree.
38707 2017-02-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38709         * config/aarch64/aarch64.c (has_memory_op): Delete.
38710         (aarch64_madd_needs_nop): Use contains_mem_rtx_p instead of
38711         has_memory_op.
38713 2017-02-13  Jakub Jelinek  <jakub@redhat.com>
38715         PR rtl-optimization/79388
38716         PR rtl-optimization/79450
38717         * combine.c (distribute_notes): When removing TEM_INSN for which
38718         corresponding dest has last value recorded, invalidate that last
38719         value.
38721 2017-02-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38723         * config/arm/arm.c (arm_print_tune_info): Use ASM_COMMENT_START instead
38724         of explicit '@'.  Add missing assembly comment marker on branch costs
38725         printout.
38727 2017-02-13  Nathan Sidwell  <nathan@acm.org>
38729         * gengtype-lex.l (<in_struct>): Add '/'.
38731 2017-02-13  Martin Liska  <mliska@suse.cz>
38733         PR c/79471
38734         * calls.c (expand_call): Replace XALLOCAVEC with XCNEWVEC.
38736 2017-02-13  Richard Biener  <rguenther@suse.de>
38738         * configure.ac (HAVE_ISL_OPTIONS_SET_SCHEDULE_SERIALIZE_SCCS):
38739         Remove.
38740         * configure: Re-generate.
38741         * config.in: Likewise.
38742         * graphite-dependences.c: Simplify as if
38743         HAVE_ISL_OPTIONS_SET_SCHEDULE_SERIALIZE_SCCS was defined.
38744         * graphite-isl-ast-to-gimple.c: Likewise.
38745         * graphite-optimize-isl.c: Likewise.
38746         * graphite-poly.c: Likewise.
38747         * graphite-sese-to-poly.c: Likewise.
38748         * graphite.h: Likewise.
38749         * toplev.c: Include isl/version.h and use isl_version () for
38750         printing the ISL version.
38751         * doc/install.texi: Update ISL requirement.
38753 2017-02-12  Gerald Pfeifer  <gerald@pfeifer.com>
38755         * doc/standards.texi (Standards): Update reference to
38756         Objective-C 2.0.
38758 2017-02-12  Gerald Pfeifer  <gerald@pfeifer.com>
38760         * doc/extend.texi (Named Address Spaces): sourceware.org now
38761         defaults to https.
38762         * doc/install.texi (Binaries): Ditto.
38763         (Specific): Ditto.
38765 2017-02-11  Sandra Loosemore  <sandra@codesourcery.com>
38767         * doc/cpp.texi: Replace "stringify"/"stringification" with C
38768         standard terminology "stringize"/"stringizing" throughout.
38769         * doc/cppinternals.texi: Likewise.
38771 2017-02-11  Sandra Loosemore  <sandra@codesourcery.com>
38773         * doc/extend.texi: Fix some spelling mistakes and typos.
38774         * doc/invoke.texi: Likewise.
38776 2017-02-11  Jan Hubicka  <hubicka@ucw.cz>
38778         PR ipa/79224
38779         * params.def (inline-min-speedup) Change from 10 to 8.
38781 2017-02-11  Jakub Jelinek  <jakub@redhat.com>
38783         * doc/invoke.texi (fopenmp): Bump OpenMP version from 4.0 to
38784         4.5.
38786 2017-02-11  Jan Hubicka  <hubicka@ucw.cz>
38788         PR ipa/79224
38789         * ipa-inline-analysis.c (get_minimal_bb): New function.
38790         (record_modified): Use it.
38791         (remap_edge_change_prob): Handle also ancestor functions.
38793 2017-02-11  Gerald Pfeifer  <gerald@pfeifer.com>
38795         * doc/contrib.texi (Contributors): Remove broken link into
38796         the Mauve CVS repository.
38798 2017-02-11  Jakub Jelinek  <jakub@redhat.com>
38800         PR middle-end/79454
38801         * internal-fn.c (expand_vector_ubsan_overflow): Use piece-wise
38802         result computation whenever lhs doesn't have vector mode, not
38803         just when it has BLKmode.
38805 2017-02-10  Gerald Pfeifer  <gerald@pfeifer.com>
38807         * doc/makefile.texi (profiledbootstrap): Refer to the
38808         installation instructions only in textual form.
38810 2017-02-10  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
38812         PR target/79295
38813         * config/rs6000/altivec.md (bcd<bcd_add_sub>): Fix constraints.
38815 2017-02-10  Gerald Pfeifer  <gerald@pfeifer.com>
38817         * doc/install.texi (Specific): Use https for blackfin.uclinux.org.
38818         (Specific): Update mingw-w64 reference.
38819         (Binaries): Ditto.
38820         (Specific): Remove broken link to Renesas RX processor.
38822 2017-02-10  Richard Biener  <rguenther@suse.de>
38824         * toplev.c (process_options): Do not mention obsolete graphite
38825         options when printing sorry message about missing graphite support.
38826         Mention -floop-nest-optimize.
38828 2017-02-10  Christophe Lyon  <christophe.lyon@linaro.org>
38830         * config/aarch64/arm_neon.h (vtst_p8): Rewrite without asm.
38831         (vtst_p16): Likewise.
38832         (vtstq_p8): Likewise.
38833         (vtstq_p16): Likewise.
38834         (vtst_p64): New.
38835         (vtstq_p64): Likewise.
38836         * config/arm/arm_neon.h (vgetq_lane_p64): New.
38837         (vset_lane_p64): New.
38838         (vsetq_lane_p64): New.
38840 2017-02-10  Jakub Jelinek  <jakub@redhat.com>
38842         PR tree-optimization/79411
38843         * tree-ssa-reassoc.c (is_reassociable_op): Return false if
38844         stmt operands are SSA_NAMEs used in abnormal phis.
38845         (can_reassociate_p): Return false if op is SSA_NAME used in abnormal
38846         phis.
38848 2017-02-09  Jan Hubicka  <hubicka@ucw.cz>
38850         PR ipa/70795
38851         * cgraphunit.c (cgraph_node::add_new_function): Set externally_visible
38852         flag if needed.
38854 2017-02-09  Jan Hubicka  <hubicka@ucw.cz>
38856         * tree-ssa-loop-unswitch.c (hoist_guard): Update profile.
38858 2017-02-09  Jakub Jelinek  <jakub@redhat.com>
38860         * omp-offload.c (oacc_loop_auto_partitions): Use || instead of |
38861         to avoid warning.
38863         PR c/79413
38864         * gimplify.h (is_gimple_sizepos): Only test for INTEGER_CST constants,
38865         not arbitrary TREE_CONSTANT.
38867         PR c/79431
38868         * gimplify.c (gimplify_adjust_omp_clauses): Ignore
38869         "omp declare target link" attribute unless is_global_var.
38870         * omp-offload.c (find_link_var_op): Likewise.
38872 2017-02-09  Nathan Sidwell  <nathan@codesourcery.com>
38873             Chung-Lin Tang  <cltang@codesourcery.com>
38875         * gimplify.c (gimplify_scan_omp_clauses): No special handling for
38876         OMP_CLAUSE_TILE.
38877         (gimplify_adjust_omp_clauses): Don't delete TILE.
38878         (gimplify_omp_for): Deal with TILE.
38879         * internal-fn.c (expand_GOACC_TILE): New function.
38880         * internal-fn.def (GOACC_DIM_POS): Comment may be overly conservative.
38881         (GOACC_TILE): New.
38882         * omp-expand.c (struct oacc_collapse): Add tile and outer fields.
38883         (expand_oacc_collapse_init): Add LOC paramter.  Initialize tile
38884         element fields.
38885         (expand_oacc_collapse_vars): Add INNER parm, adjust for tiling,
38886         avoid DIV for outermost collapse var.
38887         (expand_oacc_for): Insert tile element loop as needed.  Adjust.
38888         Remove out of date comments, fix whitespace.
38889         * omp-general.c (omp_extract_for_data): Deal with tiling.
38890         * omp-general.h (enum oacc_loop_flags): Add OLF_TILE flag,
38891         adjust OLF_DIM_BASE value.
38892         (struct omp_for_data): Add tiling field.
38893         * omp-low.c (scan_sharing_clauses): Allow OMP_CLAUSE_TILE.
38894         (lower_oacc_head_mark): Add OLF_TILE as appropriate.  Ensure 2 levels
38895         for auto loops.  Remove default auto determining, moved to
38896         oacc_loop_fixed_partitions.
38897         * omp-offload.c (struct oacc_loop): Change 'ifns' to vector of call
38898         stmts, add e_mask field.
38899         (oacc_dim_call): New function, abstracted out from oacc_thread_numbers.
38900         (oacc_thread_numbers): Use oacc_dim_call.
38901         (oacc_xform_tile): New.
38902         (new_oacc_loop_raw): Initialize e_mask, adjust for ifns vector.
38903         (finish_oacc_loop): Adjust for ifns vector.
38904         (oacc_loop_discover_walk): Append loop abstraction sites to list,
38905         add case for GOACC_TILE fns.
38906         (oacc_loop_xform_loop): Delete.
38907         (oacc_loop_process): Iterate over call list directly, and add
38908         handling for GOACC_TILE fns.
38909         (oacc_loop_fixed_partitions): Determine default auto, deal with TILE,
38910         dump partitioning.
38911         (oacc_loop_auto_partitions): Add outer_assign parm. Assign all but
38912         vector partitioning to outer loops.  Assign 2 partitions to loops
38913         when available. Add TILE handling.
38914         (oacc_loop_partition): Adjust oacc_loop_auto_partitions call.
38915         (execite_oacc_device_lower): Process GOACC_TILE fns,
38916         ignore unknown specs.
38917         * tree-nested.c (convert_nonlocal_omp_clauses): Allow OMP_CLAUSE_TILE.
38918         * tree.c (omp_clause_num_ops): Adjust TILE ops.
38919         * tree.h (OMP_CLAUSE_TILE_ITERVAR, OMP_CLAUSE_TILE_COUNT): New.
38921 2017-02-09  Gerald Pfeifer  <gerald@pfeifer.com>
38923         * configure.ac (ACX_BUGURL): Update.
38924         * configure: Regenerate.
38926 2017-02-09  Richard Biener  <rguenther@suse.de>
38928         PR tree-optimization/69823
38929         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
38930         Properly enumerate all BBs in the region.  Use auto_vec/auto_bitmap.
38932 2017-02-09  Andrew Burgess  <andrew.burgess@embecosm.com>
38934         * config/arc/arc-c.def: Add __NPS400__ definition.
38935         * config/arc/arc.h (CPP_SPEC): Don't define __NPS400__ here.
38936         (TARGET_NPS400): Define.
38938 2017-02-09  Andrew Burgess  <andrew.burgess@embecosm.com>
38940         * config/arc/arc-arch.h (arc_arch_t): Move unchanged to earlier in
38941         file.
38942         (arc_cpu_t): Change base_architecture field, arch, to a arc_arc_t
38943         pointer, arch_info.
38944         (arc_cpu_types): Fill the arch_info field with a pointer into the
38945         arc_arch_types table.
38946         (arc_selected_cpu): Declare.
38947         * config/arc/arc.c (arc_selected_cpu): Make global.
38948         (arc_selected_arch): Delete.
38949         (arc_base_cpu): Delete.
38950         (arc_override_options): Remove references to deleted variables,
38951         update access to arch information.
38952         (ARC_OPT): Update access to arch information.
38953         (ARC_OPTX): Likewise.
38954         * config/arc/arc.h (arc_base_cpu): Remove declaration.
38955         (TARGET_ARC600): Update access to arch information.
38956         (TARGET_ARC601): Likewise.
38957         (TARGET_ARC700): Likewise.
38958         (TARGET_EM): Likewise.
38959         (TARGET_HS): Likewise.
38960         * config/arc/driver-arc.c (arc_cpu_to_as): Update access to arch
38961         information.
38963 2017-02-08  Pat Haugen  <pthaugen@us.ibm.com>
38965         PR target/78604
38966         * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Invert
38967         condition/operands for integer GE/LE/GEU/LEU operations.
38969 2017-02-08  Segher Boessenkool  <segher@kernel.crashing.org>
38971         PR translation/79397
38972         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Fix spelling
38973         of AltiVec.
38975 2017-02-08  Martin Jambor  <mjambor@suse.cz>
38977         PR ipa/79375
38978         * ipa-prop.c (ipa_alloc_node_params): Make static, return bool
38979         whether allocation happened.
38980         (ipa_initialize_node_params): Do not call ipa_alloc_node_params if
38981         nothing was allocated.
38983 2017-02-08  Jakub Jelinek  <jakub@redhat.com>
38985         PR tree-optimization/79408
38986         * tree-vrp.c (simplify_div_or_mod_using_ranges): If op1 is not
38987         constant, but SSA_NAME with a known integer range, use the minimum
38988         of that range instead of op1 to determine if modulo can be replaced
38989         with its first operand.
38991 2016-02-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38993         * config/riscv/riscv.c (riscv_build_integer_1): Avoid use of INT16_MAX.
38995 2017-02-08  Richard Biener  <rguenther@suse.de>
38997         PR tree-optimization/71824
38998         * graphite-scop-detection.c (scop_detection::build_scop_breadth):
38999         Check all loops contained in the merged region.
39001 2017-02-07  Andrew Pinski  <apinski@cavium.com>
39003         * config/aarch64/aarch64.md (popcount<mode>2): New pattern.
39005 2017-02-07  Andrew Pinski  <apinski@cavium.com>
39007         * config/aarch64/aarch64-cores.def (thunderx): Disable LSE.
39008         (thunderxt88): Likewise.
39009         (thunderxt81): Disable LSE and change v8.1 to v8.
39010         (thunderxt83): Likewise.
39012 2017-02-07  Jakub Jelinek  <jakub@redhat.com>
39013             Richard Biener  <rguenther@suse.de>
39015         PR middle-end/79399
39016         * ira-int.h (struct target_ira_int): Change x_max_struct_costs_size
39017         type from int to size_t.
39018         * ira-costs.c (struct_costs_size): Change type from int to size_t.
39020 2017-02-07  Jakub Jelinek  <jakub@redhat.com>
39022         PR rtl-optimization/79386
39023         * cprop.c (bypass_conditional_jumps): Initialize
39024         bypass_last_basic_block already before splitting bbs after
39025         unconditional traps...
39026         (bypass_conditional_jumps): ... rather than here.
39028         PR target/79299
39029         * config/i386/sse.md (xtg_mode, gatherq_mode): New mode attrs.
39030         (*avx512f_gathersi<mode>, *avx512f_gathersi<mode>_2,
39031         *avx512f_gatherdi<mode>, *avx512f_gatherdi<mode>_2): Use them,
39032         fix -masm=intel patterns.
39034 2017-02-07  Richard Biener  <rguenther@suse.de>
39036         PR tree-optimization/79256
39037         PR middle-end/79278
39038         * builtins.c (get_object_alignment_2): Use min_align_of_type
39039         to extract alignment for MEM_REFs to honor BIGGEST_FIELD_ALIGNMENT
39040         and ADJUST_FIELD_ALIGN.
39042         * doc/tm.texi.in (ADJUST_FIELD_ALIGN): Adjust to take additional
39043         type parameter.
39044         * doc/tm.texi: Regenerate.
39045         * stor-layout.c (layout_decl): Adjust.
39046         (update_alignment_for_field): Likewise.
39047         (place_field): Likewise.
39048         (min_align_of_type): Likewise.
39049         * config/arc/arc.h (ADJUST_FIELD_ALIGN): Adjust.
39050         * config/epiphany/epiphany.h (ADJUST_FIELD_ALIGN): Likewise.
39051         * config/epiphany/epiphany.c (epiphany_adjust_field_align): Likewise.
39052         * config/frv/frv.h (ADJUST_FIELD_ALIGN): Likewise.
39053         * config/frv/frv.c (frv_adjust_field_align): Likewise.
39054         * config/i386/i386.h (ADJUST_FIELD_ALIGN): Likewise.
39055         * config/i386/i386.c (x86_field_alignment): Likewise.
39056         * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Likewise.
39057         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Likewise.
39058         * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Likewise.
39059         * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
39060         * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Likewise.
39061         * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
39062          Likewise.
39064         Revert
39065         2017-01-30  Richard Biener  <rguenther@suse.de>
39067         PR tree-optimization/79256
39068         * targhooks.c (default_builtin_vector_alignment_reachable): Honor
39069         BIGGEST_FIELD_ALIGNMENT and ADJUST_FIELD_ALIGN to fix up bogus
39070         alignment on TYPE.
39072 2017-02-07  Toma Tabacu  <toma.tabacu@imgtec.com>
39074         * config/mips/mips.c (mips_expand_builtin_insn): Convert the QImode
39075         argument of the pshufh, psllh, psllw, psrah, psraw, psrlh, psrlw
39076         builtins to SImode and emit a zero-extend, if necessary.
39078 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
39080         * docs/invoke.texi (RISC-V Options): Alphabetize.
39082 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
39084         * doc/invoke.texi (RISC-V Options): Use two spaces to separate
39085         options.
39087 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
39089         * config/riscv/riscv.c: New file.
39090         * common/config/riscv/riscv-common.c: Likewise.
39091         * config.gcc: Likewise.
39092         * config/riscv/constraints.md: Likewise.
39093         * config/riscv/elf.h: Likewise.
39094         * config/riscv/generic.md: Likewise.
39095         * config/riscv/linux.h: Likewise.
39096         * config/riscv/multilib-generator: Likewise.
39097         * config/riscv/peephole.md: Likewise.
39098         * config/riscv/pic.md: Likewise.
39099         * config/riscv/predicates.md: Likewise.
39100         * config/riscv/riscv-builtins.c: Likewise.
39101         * config/riscv/riscv-c.c: Likewise.
39102         * config/riscv/riscv-ftypes.def: Likewise.
39103         * config/riscv/riscv-modes.def: Likewise.
39104         * config/riscv/riscv-opts.h: Likewise.
39105         * config/riscv/riscv-protos.h: Likewise.
39106         * config/riscv/riscv.h: Likewise.
39107         * config/riscv/riscv.md: Likewise.
39108         * config/riscv/riscv.opt: Likewise.
39109         * config/riscv/sync.md: Likewise.
39110         * config/riscv/t-elf-multilib: Likewise.
39111         * config/riscv/t-linux: Likewise.
39112         * config/riscv/t-linux-multilib: Likewise.
39113         * config/riscv/t-riscv: Likewise.
39114         * configure.ac: Likewise.
39115         * doc/contrib.texi: Add Kito Cheng, Palmer Dabbelt, and Andrew
39116         Waterman as RISC-V maintainers.
39117         * doc/install.texi: Add RISC-V entries.
39118         * doc/invoke.texi: Add RISC-V options section.
39119         * doc/md.texi: Add RISC-V constraints section.
39120         * configure: Regenerated.
39122 2017-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
39124         PR target/66144
39125         * config/rs6000/vector.md (vcond<mode><mode>): Allow the true and
39126         false values to be constant vectors with all 0 or all 1 bits set.
39127         (vcondu<mode><mode>): Likewise.
39128         * config/rs6000/predicates.md (vector_int_reg_or_same_bit): New
39129         predicate.
39130         (fpmask_comparison_operator): Update comment.
39131         (vecint_comparison_operator): New predicate.
39132         * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Optimize
39133         vector conditionals when the true and false values are constant
39134         vectors with all 0 bits or all 1 bits set.
39136 2017-02-06  Martin Sebor  <msebor@redhat.com>
39138         PR  tree-optimization/79376
39139         * gimple-fold.c (get_range_strlen): Set the minimum length to zero.
39141 2017-02-06  Uros Bizjak  <ubizjak@gmail.com>
39143         * config/i386/sse.md (vector modes -> vec_extract* splitter): Use
39144         explicit subreg RTX with operand 1.  Use VECTOR_MODE_P predicate
39145         to simplify split condition.
39147 2017-02-06  Jakub Jelinek  <jakub@redhat.com>
39149         * omp-expand.c (oxpand_omp_atomic_fetch_op,
39150         expand_omp_atomic_pipeline): Return false if can_atomic_load_p is
39151         false.
39153 2017-02-06  Segher Boessenkool  <segher@kernel.crashing.org>
39155         PR rtl-optimization/68664
39156         * target.def (can_speculate_insn): New hook.
39157         * doc/tm.texi.in (TARGET_SCHED_CAN_SPECULATE_INSN): New hook.
39158         * doc/tm.texi: Regenerate.
39159         * sched-rgn.c (can_schedule_ready_p): Use the new hook.
39160         * config/rs6000/rs6000.c (TARGET_SCHED_CAN_SPECULATE_INSN): New macro.
39161         (rs6000_sched_can_speculate_insn): New function.
39163 2017-02-06  Jakub Jelinek  <jakub@redhat.com>
39165         PR tree-optimization/79284
39166         * tree-vectorizer.h (VECT_SCALAR_BOOLEAN_TYPE_P): Define.
39167         * tree-vect-stmts.c (vect_get_vec_def_for_operand,
39168         vectorizable_mask_load_store, vectorizable_operation,
39169         vect_is_simple_cond, get_same_sized_vectype): Use it instead
39170         of comparing TREE_CODE of a type against BOOLEAN_TYPE.
39171         * tree-vect-patterns.c (check_bool_pattern, search_type_for_mask_1,
39172         vect_recog_bool_pattern, vect_recog_mask_conversion_pattern): Likewise.
39173         * tree-vect-slp.c (vect_get_constant_vectors): Likewise.
39174         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
39175         Remove redundant gimple_code (stmt) == GIMPLE_ASSIGN test after
39176         is_gimple_assign (stmt).  Replace another such test with
39177         is_gimple_assign (stmt).
39179 2017-02-06  Georg-Johann Lay  <avr@gjlay.de>
39181         PR target/78883
39182         * config/avr/avr.c (rtl-iter.h): Include it.
39183         (TARGET_LEGITIMATE_COMBINED_INSN): New hook define...
39184         (avr_legitimate_combined_insn): ...and implementation.
39186 2017-02-06  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39188         * config/s390/predicates.md ("larl_operand"): Use macros from hwint.h.
39189         * config/s390/s390.c (s390_const_operand_ok)
39190         (s390_canonicalize_comparison, s390_extract_part)
39191         (s390_single_part, s390_contiguous_bitmask_nowrap_p)
39192         (s390_contiguous_bitmask_p, s390_rtx_costs)
39193         (legitimize_pic_address): Likewise.
39194         * config/s390/s390.md ("clzdi2", "clztidi2"): Likewise.
39195         * config/s390/vx-builtins.md ("vec_genbytemaskv16qi")
39196         ("vec_permi<mode>", "vfae<mode>", "*vfaes<mode>", "vstrc<mode>")
39197         ("*vstrcs<mode>"): Use UINTVAL() to set unsigned HOST_WIDE_INT.
39198         * config/s390/vector.md ("vec_vfenes<mode>"): Likewise.
39200 2017-02-06  Georg-Johann Lay  <avr@gjlay.de>
39202         * config/avr/avr.md (*addhi3_zero_extend): Add alternative where
39203         REGNO($0) == REGNO($1).
39205 2017-02-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39207         * config/s390/linux.h(SIZE_TYPE): Add comment.
39209 2017-02-06  Julian Brown  <julian@codesourcery.com>
39210             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
39211             Virendra Pathak  <virendra.pathak@broadcom.com>
39213         * config/aarch64/aarch64-cores.def: Change the scheduler
39214         to Thunderx2t99.
39215         * config/aarch64/aarch64.md: Include thunderx2t99.md.
39216         * config/aarch64/thunderx2t99.md: New file.
39218 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
39220         * doc/standards.texi (Go Language): Update link to language
39221         standard.
39223 2017-02-05  Jan Hubicka  <hubicka@ucw.cz>
39225         * tree-eh.c (lower_resx): Sanitize profile.
39226         (cleanup_empty_eh_move_lp): Likewise.
39228 2017-02-05  Jan Hubicka  <hubicka@ucw.cz>
39230         PR tree-ssa/79347
39231         * cfgloopmanip.c (lv_adjust_loop_entry_edge, loop_version): Add
39232         ELSE_PROB.
39233         * cfgloopmanip.h (loop_version): Update prototype.
39234         * modulo-sched.c (sms_schedule): Update call of loop_version.
39235         * tree-if-conv.c(version_loop_for_if_conversion): Likewise.
39236         * tree-parloops.c (gen_parallel_loop): Likewise.
39237         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Likewise.
39238         * tree-ssa-loop-split.c (split_loop): Likewise.
39239         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Likewise.
39240         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
39242 2017-02-05  Martin Liska  <mliska@suse.cz>
39244         PR bootstrap/78985
39245         * config/s390/s390.c (s390_gimplify_va_arg): Initialize local
39246         variable to NULL.
39247         (print_operand_address): Initialize a struct to zero.
39249 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
39251         * doc/contrib.texi (Contributors): Refer to Hans Boehm's
39252         garbage collector only in textual form.
39254 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
39256         * doc/extend.texi (x86 specific memory model extensions for
39257         transactional memory): Simplify a phrase.
39259 2017-02-05  Eric Botcazou  <ebotcazou@adacore.com>
39261         PR target/79353
39262         * config/sparc/sync.md (atomic_loaddi_1): Replace 'U' constraint with
39263         'r', 'm' constraint with 'T' and !TARGET_ARCH64 with TARGET_ARCH32.
39264         (atomic_storedi_1): Likewise.
39266 2017-02-04  Jakub Jelinek  <jakub@redhat.com>
39268         PR tree-optimization/79338
39269         * tree-parloops.c (gather_scalar_reductions): Don't call
39270         vect_analyze_loop_form for loop->inner before destroying loop's
39271         loop_vinfo.
39273 2017-02-03  Martin Sebor  <msebor@redhat.com>
39275         PR tree-optimization/79327
39276         * gimple-ssa-sprintf.c (tree_digits): Avoid adding the base prefix
39277         when precision has resulted in leading zeros.
39278         (format_integer): Adjust the likely counter to assume an unknown
39279         argument that may be zero is non-zero.
39281 2017-02-03  Jason Merrill  <jason@redhat.com>
39283         PR c++/78689
39284         * tree-inline.c (copy_tree_body_r) [COND_EXPR]: Revert change to
39285         avoid copying non-taken branch.
39287 2017-02-03  Jakub Jelinek  <jakub@redhat.com>
39289         PR tree-optimization/79340
39290         * tree-vect-loop.c (vectorizable_reduction): Release
39291         vec_defs elements after safe_splicing them into other vectors.
39292         Formatting fixes.
39294         PR tree-optimization/79327
39295         * gimple-ssa-sprintf.c (adjust_range_for_overflow): If returning
39296         true, always set *argmin and *argmax to TYPE_{MIN,MAX}_VALUE of
39297         dirtype.
39298         (format_integer): Use wide_int_to_tree instead of build_int_cst
39299         + to_?hwi.  If argmin is NULL, just set argmin and argmax to
39300         TYPE_{MIN,MAX}_VALUE of argtype.  Simplify and fix computation
39301         of shortest and longest sequence.
39303 2017-02-03  Uros Bizjak  <ubizjak@gmail.com>
39305         * config/i386/i386.c (dimode_scalar_chain::convert_reg):
39306         Use pextrd for TARGET_SSE4_1 when creating scalar copy.
39308 2017-02-03  Walter Lee  <walt@tilera.com>
39310         PR target/78862
39311         * config/tilegx/tilegx.md (tilegx_expand_prologue): Add blockage
39312         after initial stackframe link reg save.
39313         * config/tilepro/tilepro.md (tilepro_expand_prologue): Likewise.
39315 2017-02-03  Jakub Jelinek  <jakub@redhat.com>
39317         PR target/79354
39318         * config/rs6000/rs6000.md (movsi_from_sf): Use wb constraint instead of
39319         wu for stxssp alternative.
39321 2017-02-03  Martin Sebor  <msebor@redhat.com>
39323         PR tree-optimization/79352
39324         * gimple-fold.c (get_range_strlen): Add argument.
39325         (get_range_strlen): Change return type to bool.
39326         (get_maxval_strlen): Pass in a dummy argument.
39327         * gimple-fold.h (get_range_strlen): Change return type to bool.
39328         * gimple-ssa-sprintf.c (get_string_length): Set unlikely counter.
39329         * tree.h (array_at_struct_end_p): Add argument.
39330         * tree.c (array_at_struct_end_p): Handle it.
39332 2017-02-03  Martin Liska  <mliska@suse.cz>
39334         PR lto/66295
39335         * multiple_target.c (create_dispatcher_calls): Redirect edge
39336         from a caller of a dispatcher.
39337         (expand_target_clones): Make the clones local.
39338         (ipa_target_clone): Do both target clones and resolvers.
39339         (ipa_dispatcher_calls): Remove the pass.
39340         (pass_dispatcher_calls::gate): Likewise.
39341         (make_pass_dispatcher_calls): Likewise.
39342         * passes.def (pass_target_clone): Put as very first IPA early
39343         pass.
39345 2017-02-03  Martin Liska  <mliska@suse.cz>
39347         * symtab.c (symtab_node::binds_to_current_def_p): Bail out
39348         in case of a function with ifunc attribute.
39350 2017-02-03  Martin Liska  <mliska@suse.cz>
39352         * cgraph.c (cgraph_node::dump): Dump function version info.
39353         * symtab.c (symtab_node::dump_base): Add missing new line.
39355 2017-02-02  Jan Hubicka  <hubicka@ucw.cz>
39357         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): New function.
39358         (ifcombine_ifandif): Use it.
39360 2017-02-03  Martin Liska  <mliska@suse.cz>
39362         * doc/invoke.texi: Document default value for
39363         use-after-scope-direct-emission-threshold.
39365 2017-02-03  Martin Liska  <mliska@suse.cz>
39367         PR tree-optimization/79339
39368         * gimple-ssa-sprintf.c (format_floating_max): Call mpfr_clear.
39369         (format_floating): Likewise.
39371 2017-02-03  Martin Liska  <mliska@suse.cz>
39373         PR ipa/79337
39374         * ipa-prop.c (ipa_node_params_t::insert): Remove current
39375         implementation.
39376         (ipa_node_params_t::remove): Likewise.
39377         * ipa-prop.h (ipa_node_params::ipa_node_params): Make default
39378         initialization from removed ipa_node_params_t::insert.
39379         (ipa_node_params::~ipa_node_params): Move from removed
39380         ipa_node_params_t::release.
39381         * symbol-summary.h (symbol_summary::m_released): New member.
39382         Do not release a summary twice.  Do not allow to call finalizer
39383         for types of a summary that live in GGC memory.
39385 2017-02-02  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
39387         * config/aarch64/aarch64.c (thunderx2t99_tunings): Enable AES and
39388         cmp_branch fusion.
39390 2017-02-02  Martin Sebor  <msebor@redhat.com>
39392         PR middle-end/79275
39393         * gimple-ssa-sprintf.c (get_string_length): Set lower bound to zero.
39394         (format_string): Tighten up the range of output for non-constant
39395         strings and correct the expected range for wide non-constant strings.
39397 2017-02-02  Martin Sebor  <msebor@redhat.com>
39399         * doc/invoke.texi (-maccumulate-args): Fix bad grammar.
39401         PR middle-end/32003
39402         * doc/invoke.texi (-fdump-tree-): Remove pass-specific options from
39403         index.
39404         (-fdump-tree-@var): Add to index and document how to come up
39405         with pass-specific option and dump file names.
39406         (-fdump-passes): Clarify where to look for output.
39408 2017-02-02  Jan Hubicka  <hubicka@ucw.cz>
39410         PR middle-end/77445
39411         * tree-ssa-threadbackward.c (profitable_jump_thread_path): Dump
39412         statistics of the analyzed path; allow threading for speed when
39413         any of BBs along the path are optimized for speed.
39415 2017-02-02  Eric Botcazou  <ebotcazou@adacore.com>
39417         PR middle-end/78468
39418         * emit-rtl.c (init_emit): Add ??? comment for problematic alignment
39419         settings of the virtual registers.
39421         Revert again
39422         2016-08-23  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39424         * explow.c (get_dynamic_stack_size): Take known alignment of stack
39425         pointer + STACK_DYNAMIC_OFFSET into account when calculating the size
39426         needed.
39428 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39430         * config/s390/vx-builtins.md ("vec_ceil", "vec_floor")
39431         ("vec_trunc", "vec_roundc", "vec_round"): Remove expanders.
39433 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39435         * config/s390/s390.md: Add missing comments with the expanded
39436         mnemonics.
39437         * config/s390/vector.md: Likewise.
39438         * config/s390/vx-builtins.md: Likewise.
39440 2017-02-02  Jakub Jelinek  <jakub@redhat.com>
39442         PR target/79197
39443         * config/rs6000/rs6000.md (*fixuns_trunc<mode>di2_fctiduz): Rename to ...
39444         (fixuns_trunc<mode>di2): ... this, remove previous expander.  Put all
39445         conditions on a single line.
39447 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39449         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Rename
39450         __S390_VX__ to __VX__.
39452 2017-02-01  Andrew Pinski  <apinski@cavium.com>
39454         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost): Pass
39455         stmt_info to record_stmt_cost.
39456         (vect_get_known_peeling_cost): Pass stmt_info if known to
39457         record_stmt_cost.
39458         * config/aarch64/aarch64-protos.h (cpu_vector_cost): Split
39459         cpu_vector_cost field into
39460         scalar_int_stmt_cost and scalar_fp_stmt_cost.  Split vec_stmt_cost
39461         field into vec_int_stmt_cost and vec_fp_stmt_cost.
39462         * config/aarch64/aarch64.c (generic_vector_cost): Update for the
39463         splitting of scalar_stmt_cost and vec_stmt_cost.
39464         (thunderx_vector_cost): Likewise.
39465         (cortexa57_vector_cost): LIkewise.
39466         (exynosm1_vector_cost): Likewise.
39467         (xgene1_vector_cost): Likewise.
39468         (thunderx2t99_vector_cost): Improve after the splitting of the two
39469         fields.
39470         (aarch64_builtin_vectorization_cost): Update for the splitting of
39471         scalar_stmt_cost and vec_stmt_cost.
39473 2017-02-01  Torvald Riegel  <triegel@redhat.com>
39474             Richard Henderson  <rth@redhat.com>
39476         * builtins.c (fold_builtin_atomic_always_lock_free): Make "lock-free"
39477         conditional on existance of a fast atomic load.
39478         * optabs-query.c (can_atomic_load_p): New function.
39479         * optabs-query.h (can_atomic_load_p): Declare it.
39480         * optabs.c (expand_atomic_exchange): Always delegate to libatomic if
39481         no fast atomic load is available for the particular size of access.
39482         (expand_atomic_compare_and_swap): Likewise.
39483         (expand_atomic_load): Likewise.
39484         (expand_atomic_store): Likewise.
39485         (expand_atomic_fetch_op): Likewise.
39486         * testsuite/lib/target-supports.exp
39487         (check_effective_target_sync_int_128): Remove x86 because it provides
39488         no fast atomic load.
39489         (check_effective_target_sync_int_128_runtime): Likewise.
39491 2017-02-01  Richard Biener  <rguenther@suse.de>
39493         * graphite.c: Include tree-vectorizer.h for find_loop_location.
39494         (graphite_transform_loops): Provide opt-info for optimized nests.
39495         * tree-parloop.c (parallelize_loops): Provide opt-info for
39496         parallelized loops.
39498 2017-02-01  Richard Biener  <rguenther@suse.de>
39500         PR middle-end/79315
39501         * tree-cfg.c (move_stmt_op): Never set TREE_BLOCK when it
39502         was not set before.
39504 2017-02-01  Richard Biener  <rguenther@suse.de>
39506         PR tree-optimization/71824
39507         * graphite-scop-detection.c (scop_detection::build_scop_breadth):
39508         Verify the loops are valid in the merged SESE region.
39509         (scop_detection::can_represent_loop_1): Check analyzing the
39510         evolution of the number of iterations in the region succeeds.
39512 2017-01-31  Ian Lance Taylor  <iant@golang.org>
39514         * config/i386/i386.c (ix86_expand_split_stack_prologue): Add
39515         REG_ARGS_SIZE note to 32-bit push insns and call insn.
39517 2017-01-31  David Malcolm  <dmalcolm@redhat.com>
39519         PR preprocessor/79210
39520         * input.c (get_substring_ranges_for_loc): Replace line_width
39521         assertion with error-handling.
39523 2017-01-31  Richard Biener  <rguenther@suse.de>
39525         PR tree-optimization/77318
39526         * graphite-sese-to-poly.c (extract_affine): Fix assert.
39527         (create_pw_aff_from_tree): Take loop parameter.
39528         (add_condition_to_pbb): Pass loop of the condition to
39529         create_pw_aff_from_tree.
39531 2017-01-31  Jakub Jelinek  <jakub@redhat.com>
39533         * config/s390/s390.c (s390_asan_shadow_offset): New function.
39534         (TARGET_ASAN_SHADOW_OFFSET): Redefine.
39536 2017-01-31  Michael Meissner  <meissner@linux.vnet.ibm.com>
39538         PR target/78597
39539         PR target/79038
39540         * config/rs6000/rs6000-protos.h (convert_float128_to_int): Delete,
39541         no longer used.
39542         (convert_int_to_float128): Likewise.
39543         * config/rs6000/rs6000.c (convert_float128_to_int): Likewise.
39544         (convert_int_to_float128): Likewise.
39545         * config/rs6000/rs6000.md (UNSPEC_IEEE128_MOVE): Likewise.
39546         (UNSPEC_IEEE128_CONVERT): Likewise.
39547         (floatsi<mode>2, FLOAT128 iterator): Bypass calling
39548         rs6000_expand_float128_convert if we have IEEE 128-bit hardware.
39549         Use local variables for IBM extended format.
39550         (fix_trunc<mode>si2, FLOAT128 iterator): Likewise.
39551         (fix_trunc<mode>si2_fprs): Likewise.
39552         (fixuns_trunc<IEEE128:mode><SDI:mode>2): Likewise.
39553         (floatuns<IEEE128:mode>2, IEEE128 iterator): Likewise.
39554         (fix<uns>_<mode>si2_hw): Rework the IEEE 128-bt hardware support
39555         to know that we can now have integers of all sizes in vector
39556         registers.
39557         (fix<uns>_<mode>di2_hw): Likewise.
39558         (float<uns>_<mode>si2_hw): Likewise.
39559         (fix_<mode>si2_hw): Likewise.
39560         (fixuns_<mode>si2_hw): Likewise.
39561         (float<uns>_<mode>di2_hw): Likewise.
39562         (float_<mode>di2_hw): Likewise.
39563         (float_<mode>si2_hw): Likewise.
39564         (floatuns_<mode>di2_hw): Likewise.
39565         (floatuns_<mode>si2_hw): Likewise.
39566         (xscvqp<su>wz_<mode>): Delete, no longer used.
39567         (xscvqp<su>dz_<mode>): Likewise.
39568         (xscv<su>dqp_<mode>): Likewise.
39569         (ieee128_mfvsrd_64bit): Likewise.
39570         (ieee128_mfvsrd_32bit): Likewise.
39571         (ieee128_mfvsrwz): Likewise.
39572         (ieee128_mtvsrw): Likewise.
39573         (ieee128_mtvsrd_64bit): Likewise.
39574         (ieee128_mtvsrd_32bit): Likewise.
39576 2017-01-31  Martin Liska  <mliska@suse.cz>
39578         PR ipa/79285
39579         * ipa-prop.c (ipa_free_all_node_params): Call release method
39580         instead of ~sumbol_summary to not to trigger double times
39581         dtor of hash_map.
39583 2017-01-31  Aldy Hernandez  <aldyh@redhat.com>
39585         PR tree-optimization/71691
39586         * bitmap.h (class auto_bitmap): New.
39587         * tree-ssa-loop-unswitch.c (tree_may_unswitch_on): Call
39588         is_maybe_undefined instead of ssa_undefined_value_p.
39590 2017-01-31  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39592         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Rename
39593         __S390_ARCH_LEVEL__ to __ARCH__.
39595 2017-01-31  Jakub Jelinek  <jakub@redhat.com>
39597         PR tree-optimization/79267
39598         * value-prof.c (gimple_ic): Only drop lhs for noreturn calls
39599         if should_remove_lhs_p is true.
39601 2017-01-30  Alexandre Oliva  <aoliva@redhat.com>
39603         PR debug/63238
39604         * dwarf2out.c (clone_as_declaration): Drop DW_AT_alignment.
39605         (add_alignment_attribute): New.
39606         (base_type_die): Add alignment attribute.
39607         (subrange_type_die): Likewise.
39608         (modified_type_die): Likewise.
39609         (gen_array_type_die): Likewise.
39610         (gen_descr_array_type_die: Likewise.
39611         (gen_enumeration_type_die): Likewise.
39612         (gen_subprogram_die): Likewise.
39613         (gen_variable_die): Likewise.
39614         (gen_field_die): Likewise.
39615         (gen_ptr_to_mbr_type_die): Likewise.
39616         (gen_struct_or_union_type_die): Likewise.
39617         (gen_subroutine_type_die): Likewise.
39618         (gen_typedef_die): Likewise.
39619         (base_type_cmp): Compare alignment attribute.
39621 2017-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
39623         PR target/79170
39624         * config/rs6000/altivec.md (*setb_internal): Rename to setb_signed.
39625         (setb_unsigned) New pattern for setb with CCUNS.
39626         * config/rs6000/rs6000.c (expand_block_compare): Use a different
39627         subfc./subfe sequence to avoid overflow problems.  Generate a
39628         shorter sequence with cmpld/setb for power9.
39629         * config/rs6000/rs6000.md (subf<mode>3_carry_dot2): Add a new pattern
39630         for generating subfc. instruction.
39631         (cmpstrsi): Add TARGET_POPCNTD predicate as the generate sequence
39632         now uses this instruction.
39634 2017-01-30  Ian Lance Taylor  <iant@google.com>
39636         PR debug/79289
39637         * dwarf2out.c (gen_type_die_with_usage): When picking a variant
39638         for FUNCTION_TYPE/METHOD_TYPE, use the first matching one.
39640 2017-01-30  Martin Sebor  <msebor@redhat.com>
39642         * gimple-ssa-sprintf.c (fmtresult::adjust_for_width_or_precision):
39643         Move constant to the right of a relational operator.
39644         (get_mpfr_format_length, format_character, format_string): Ditto.
39645         (should_warn_p, maybe_warn): Same.
39647         * doc/invoke.texi (-Wformat-truncation=1): Fix typo.
39649 2017-01-30  Maxim Ostapenko  <m.ostapenko@samsung.com>
39651         PR lto/79061
39652         * asan.c (get_translation_unit_decl): Remove function.
39653         (asan_add_global): Force has_dynamic_init to zero in LTO mode.
39655 2017-01-30  Martin Liska  <mliska@suse.cz>
39657         PR gcov-profile/79259
39658         * opts.c (common_handle_option): Enable flag_ipa_bit_cp w/
39659         -fprofile-generate.
39661 2017-01-30  Martin Liska  <mliska@suse.cz>
39663         PR bootstrap/78985
39664         * config/aarch64/cortex-a57-fma-steering.c (func_fma_steering::analyze):
39665         Initialize variables with NULL value.
39667 2017-01-30  Richard Earnshaw  <rearnsh@arm.com>
39669         PR target/79260
39670         * config.gcc (arm*-*-*): Add arm/arm-flags.h and arm/arm-isa.h to
39671         tm_p_file.
39672         * arm/arm-protos.h: Don't directly include arm-flags.h and arm-isa.h.
39674 2017-01-30  Richard Biener  <rguenther@suse.de>
39676         PR tree-optimization/79276
39677         * tree-vrp.c (process_assert_insertions): Properly adjust common
39678         when removing a duplicate.
39680 2017-01-30  Richard Biener  <rguenther@suse.de>
39682         PR tree-optimization/79256
39683         * targhooks.c (default_builtin_vector_alignment_reachable): Honor
39684         BIGGEST_FIELD_ALIGNMENT and ADJUST_FIELD_ALIGN to fix up bogus
39685         alignment on TYPE.
39686         * tree.c (build_aligned_type): Set TYPE_USER_ALIGN.
39688 2017-01-30  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39690         PR target/79240
39691         * config/s390/s390.md ("*r<noxa>sbg_<mode>_srl_bitmask")
39692         ("*r<noxa>sbg_<mode>_sll_bitmask")
39693         ("*extzv_<mode>_srl<clobbercc_or_nocc>")
39694         ("*extzv_<mode>_sll<clobbercc_or_nocc>"):
39695         Use contiguous_bitmask_nowrap_operand.
39697 2017-01-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
39699         PR target/79268
39700         * config/rs6000/altivec.h (vec_xl): Revise #define.
39701         (vec_xst): Likewise.
39703 2017-01-27  Uros Bizjak  <ubizjak@gmail.com>
39705         * config/i386/i386.c (print_reg): Use REGNO instead of true_regnum.
39707 2017-01-27  Bernd Schmidt  <bschmidt@redhat.com>
39709         PR rtl-optimization/79194
39710         * cprop.c (one_cprop_pass): Move deletion of code after unconditional
39711         traps before call to bypass_conditional_jumps.
39713 2017-01-27  Vladimir Makarov  <vmakarov@redhat.com>
39715         PR tree-optimization/71374
39716         * lra-constraints.c (check_conflict_input_operands): New.
39717         (match_reload): Use it.
39719 2017-01-27  Vladimir Makarov  <vmakarov@redhat.com>
39721         PR target/79131
39722         * lra-assigns.c (find_hard_regno_for_1): Take endianess for into
39723         account to calculate conflict_set.
39725 2017-01-27  Bin Cheng  <bin.cheng@arm.com>
39727         PR rtl-optimization/78559
39728         * combine.c (try_combine): Discard REG_EQUAL and REG_EQUIV for
39729         other_insn in combine.
39731 2017-01-27  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
39733         * builtin-types.def: Use unsigned_char_type_node for BT_UINT8.  Use
39734         uint16_type_node for BT_UINT16.
39736 2017-01-27  David Malcolm  <dmalcolm@redhat.com>
39738         * doc/sourcebuild.texi (Testsuites): Add "GIMPLE Tests" and
39739         "RTL Tests" to menu.
39740         (GIMPLE Tests): New node.
39741         (RTL Tests): New node.
39743 2017-01-27  Richard Biener  <rguenther@suse.de>
39745         PR tree-optimization/79245
39746         * tree-loop-distribution.c (distribute_loop): Apply cost
39747         modeling also to detected patterns.
39749 2017-01-27  Richard Biener  <rguenther@suse.de>
39751         PR tree-optimization/71433
39752         * tree-vrp.c (register_new_assert_for): Revert earlier changes.
39753         (compare_assert_loc): New function.
39754         (process_assert_insertions): Sort and optimize assert locations
39755         to remove duplicates and push down identical assertions on
39756         edges to their destination block.
39758 2017-01-27  Richard Biener  <rguenther@suse.de>
39760         PR tree-optimization/79244
39761         * tree-vrp.c (remove_range_assertions): Forcefully propagate
39762         out SSA names even if abnormal.
39764 2017-01-27  Jakub Jelinek  <jakub@redhat.com>
39766         * realmpfr.h: Poison MPFR_RND{N,Z,U,D}.
39767         * gimple-ssa-sprintf.c (format_floating_max): Use GMP_RNDN
39768         instead of MPFR_RNDN.
39770 2017-01-27  Richard Earnshaw  <rearnsha@arm.com>
39772         PR target/79239
39773         * arm.c (arm_option_override): Don't call build_target_option_node
39774         until after doing all option overrides.
39775         (arm_valid_target_attribute_tree): Likewise.
39777 2017-01-27  Martin Liska  <mliska@suse.cz>
39779         * doc/invoke.texi (-fprofile-arcs): Document profiling support
39780         for {cd}tors and C++ {cd}tors.
39782 2017-01-27  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39784         * config/s390/s390.md ("*setmem_long_and")
39785         ("*setmem_long_and_31z"): Use zero_extend instead of and.
39787 2017-01-26  Martin Sebor  <msebor@redhat.com>
39789         * gimple-ssa-sprintf.c (format_floating): Simplify the computation
39790         of precision.
39792 2017-01-26  Martin Sebor  <msebor@redhat.com>
39794         * gimple-ssa-sprintf.c (format_floating): Test HAVE_XFmode and
39795         HAVE_DFmode before using XFmode or DFmode.
39796         (parse_directive): Avoid using the z length modifier to avoid
39797         the ISO C++98 does not support the â€˜z’ gnu_printf length modifier.
39799         PR middle-end/78703
39800         * gimple-ssa-sprintf.c (adjust_for_width_or_precision): Change
39801         to accept adjustment as an array.
39802         (get_int_range): New function.
39803         (struct directive): Make width and prec arrays.
39804         (directive::set_width, directive::set_precision): Call get_int_range.
39805         (format_integer, format_floating): Handle width and precision ranges.
39806         (format_string, parse_directive): Same.
39808 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39810         PR debug/79129
39811         * dwarf2out.c (generate_skeleton_bottom_up): For children with
39812         comdat_type_p set, just clone them, but keep the children in the
39813         original DIE.
39815         PR debug/78835
39816         * dwarf2out.c (prune_unused_types): Mark all functions with DIEs
39817         which have direct callers with -fvar-tracking-assignments enabled
39818         in the current TU.
39819         (resolve_addr): Avoid adding skeleton DIEs for DW_AT_call_origin
39820         inside of type units.
39822 2017-01-26  Martin Sebor  <msebor@redhat.com>
39824         PR middle-end/78703
39825         * gimple-ssa-sprintf.c (struct result_range): Add likely and
39826         unlikely counters.
39827         (struct format_result): Replace number_chars, number_chars_min,
39828         and number_chars_max with a single member of struct result_range.
39829         Remove bounded.
39830         (format_result::operator+=): Adjust.
39831         (struct fmtresult): Remove bounded.  Handle likely and unlikely
39832         counters.
39833         (fmtresult::adjust_for_width_or_precision): New function.
39834         (fmtresult:type_max_digits): New function.
39835         (bytes_remaining): Handle likely and unlikely counters.
39836         (min_bytes_remaining): Remove.
39837         (format_percent): Simplify.
39838         (format_integer, format_floating): Set likely and unlikely counters.
39839         (get_string_length, format_character, format_string): Same.
39840         (format_plain, should_warn_p): New function.
39841         (maybe_warn): Call should_warn_p.  Update diagnostic messages
39842         and handle those for all directives, including plain strings.
39843         (format_directive): Handle likely and unlikely counters.
39844         Remove unnecessary quoting from diagnostics.  Add an informational
39845         note.
39846         (add_bytes): Remove.
39847         (pass_sprintf_length::compute_format_length): Simplify.
39848         (try_substitute_return_value): Handle likely and unlikely counters.
39850 2017-01-26  Carl Love  <cel@us.ibm.com>
39852         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Remove
39853         bogus entries for the P8V_BUILTIN_VEC_VGBBD built-ins
39855 2017-01-26  Vladimir Makarov  <vmakarov@redhat.com>
39857         PR target/79131
39858         * lra-assigns.c (setup_live_pseudos_and_spill_after_risky): Take
39859         endianess for subregs into account.
39860         * lra-constraints.c (lra_constraints): Do risky transformations
39861         always on the first iteration.
39862         * lra-lives.c (check_pseudos_live_through_calls): Add arg
39863         last_call_used_reg_set.
39864         (process_bb_lives): Define and use last_call_used_reg_set.
39865         * lra.c (lra): Always continue after lra_constraints on the first
39866         iteration.
39868 2017-01-26  Kirill Yukhin  <kirill.yukhin@gmail.com>
39870         * gcc.target/i386/avx512bw-kshiftlq-2.c: Use unsigned long long
39871         constant.
39872         * gcc.target/i386/avx512bw-kshiftrq-2.c: Ditto.
39874 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39876         * config/i386/avx512fintrin.h (_ktest_mask16_u8,
39877         _ktestz_mask16_u8, _ktestc_mask16_u8, _kadd_mask16): Move to ...
39878         * config/i386/avx512dqintrin.h (_ktest_mask16_u8,
39879         _ktestz_mask16_u8, _ktestc_mask16_u8, _kadd_mask16): ... here.
39880         * config/i386/i386-builtin.def (__builtin_ia32_ktestchi,
39881         __builtin_ia32_ktestzhi, __builtin_ia32_kaddhi): Use
39882         OPTION_MASK_ISA_AVX512DQ instead of OPTION_MASK_ISA_AVX512F.
39883         * config/i386/sse.md (SWI1248_AVX512BWDQ2): New mode iterator.
39884         (kadd<mode>, ktest<mode>): Use it instead of SWI1248_AVX512BWDQ.
39886 2017-01-26  Marek Polacek  <polacek@redhat.com>
39888         PR c/79199
39889         * fold-const.c (operand_equal_p) [COND_EXPR]: Use OP_SAME_WITH_NULL
39890         for the third operand.
39892 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39894         PR middle-end/79236
39895         * omp-low.c (struct omp_context): Add simt_stmt field.
39896         (scan_omp_for): Return omp_context *.
39897         (scan_omp_simd): Set simt_stmt on the non-_simt_ SIMD
39898         context to the _simt_ SIMD stmt.
39899         (lower_omp_for): For combined SIMD with sibling _simt_
39900         SIMD, make sure to use the same decls in _looptemp_
39901         clauses as in the sibling.
39903 2017-01-26  David Sherwood  <david.sherwood@arm.com>
39905         PR middle-end/79212
39906         * gimplify.c (omp_notice_variable): Add GOVD_SEEN flag to variables in
39907         all contexts.
39909 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39911         PR target/70465
39912         * reg-stack.c (emit_swap_insn): Instead of fld a; fld b; fxchg %st(1);
39913         emit fld b; fld a; if possible.
39915         * brig-builtins.def: Update copyright years.
39916         * config/arm/arm_acle_builtins.def: Update copyright years.
39918 2017-01-25  Michael Meissner  <meissner@linux.vnet.ibm.com>
39920         PR target/79179
39921         * config/rs6000/vsx.md (vsx_extract_<mode>_store): Use wY
39922         constraint instead of o for the stxsd instruction.
39924 2017-01-25  Carl Love  <cel@us.ibm.com>
39926         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Fix order
39927         of entries for ALTIVEC_BUILTIN_VEC_PACKS and P8V_BUILTIN_VEC_VGBBD.
39929 2017-01-25  Jonathan Wakely  <jwakely@redhat.com>
39931         * doc/invoke.texi (C++ Dialect Options): Fix typo.
39933 2017-01-25  Richard Biener  <rguenther@suse.de>
39935         PR tree-optimization/69264
39936         * target.def (vector_alignment_reachable): Improve documentation.
39937         * doc/tm.texi: Regenerate.
39938         * targhooks.c (default_builtin_vector_alignment_reachable): Simplify
39939         and add a comment.
39940         * tree-vect-data-refs.c (vect_supportable_dr_alignment): Revert
39941         earlier changes with respect to TYPE_USER_ALIGN.
39942         (vector_alignment_reachable_p): Likewise.  Improve dumping.
39944 2016-01-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
39946         PR target/79145
39947         * config/arm/arm.md (xordi3): Force constant operand into a register
39948         for TARGET_IWMMXT.
39950 2016-01-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
39952         * doc/invoke.texi (-fstore-merging): Correct default optimization
39953         levels at which it is enabled.
39954         (-O): Move -fstore-merging from list to...
39955         (-O2): ... Here.
39957 2017-01-25  Richard Biener  <rguenther@suse.de>
39959         PR debug/78363
39960         * omp-expand.c: Include debug.h.
39961         (expand_omp_taskreg): Make sure to generate early debug before
39962         outlining anything from a function.
39963         (expand_omp_target): Likewise.
39964         (grid_expand_target_grid_body): Likewise.
39966 2017-01-25  Maxim Ostapenko  <m.ostapenko@samsung.com>
39968         PR lto/79061
39969         * asan.c (get_translation_unit_decl): New function.
39970         (asan_add_global): Extract modules file name from globals
39971         TRANSLATION_UNIT_DECL name.
39973 2017-01-24  Eric Botcazou  <ebotcazou@adacore.com>
39975         PR target/77439
39976         * config/arm/arm.c (arm_function_ok_for_sibcall): Add back restriction
39977         for long calls with APCS frame and VFP.
39979 2017-01-24  David Malcolm  <dmalcolm@redhat.com>
39981         * cfg.c (original_copy_tables_initialized_p): New function.
39982         * cfg.h (original_copy_tables_initialized_p): New decl.
39983         * cfgrtl.c (relink_block_chain): Guard the call to
39984         free_original_copy_tables with a call to
39985         original_copy_tables_initialized_p.
39986         * cgraph.h (symtab_node::native_rtl_p): New decl.
39987         * cgraphunit.c (symtab_node::native_rtl_p): New function.
39988         (symtab_node::needed_p): Don't assert for early assembly output
39989         for __RTL functions.
39990         (cgraph_node::finalize_function): Set "force_output" for __RTL
39991         functions.
39992         (cgraph_node::analyze): Bail out early for __RTL functions.
39993         (analyze_functions): Update assertion to support __RTL functions.
39994         (cgraph_node::expand): Bail out early for __RTL functions.
39995         * final.c (rest_of_clean_state): Don't call delete_tree_ssa for
39996         __RTL functions.
39997         * function.h (struct function): Update comment for field
39998         "pass_startwith".
39999         * gimple-expr.c: Include "tree-pass.h".
40000         (gimple_has_body_p): Return false for __RTL functions.
40001         * Makefile.in (OBJS): Add run-rtl-passes.o.
40002         * pass_manager.h (gcc::pass_manager::get_rest_of_compilation): New
40003         accessor.
40004         (gcc::pass_manager::get_clean_slate): New accessor.
40005         * passes.c: Include "insn-addr.h".
40006         (should_skip_pass_p): Add logging.  Update logic for running
40007         "expand" to be compatible with both __GIMPLE and __RTL.  Guard
40008         property-provider override so it is only done for gimple passes.
40009         Don't skip dfinit.
40010         (skip_pass): New function.
40011         (execute_one_pass): Call skip_pass when skipping passes.
40012         * read-md.c (md_reader::read_char): Support filtering
40013         the input to a subset of line numbers.
40014         (md_reader::md_reader): Initialize fields
40015         m_first_line and m_last_line.
40016         (md_reader::read_file_fragment): New function.
40017         * read-md.h (md_reader::read_file_fragment): New decl.
40018         (md_reader::m_first_line): New field.
40019         (md_reader::m_last_line): New field.
40020         * read-rtl-function.c (function_reader::create_function): Only
40021         create cfun if it doesn't already exist.  Set PROP_rtl on cfun's
40022         curr_properties.  Set DECL_INITIAL to a dummy block.
40023         (read_rtl_function_body_from_file_range): New function.
40024         * read-rtl-function.h (read_rtl_function_body_from_file_range):
40025         New decl.
40026         * run-rtl-passes.c: New file.
40027         * run-rtl-passes.h: New file.
40029 2017-01-24  Jeff Law  <law@redhat.com>
40031         * config/microblaze/microblaze.h (ASM_FORMAT_PRIVATE_NAME): Increase
40032         buffer size.
40034 2017-01-24  Bin Cheng  <bin.cheng@arm.com>
40036         PR tree-optimization/79159
40037         * tree-ssa-loop-niter.c (get_cst_init_from_scev): New function.
40038         (record_nonwrapping_iv): Improve boundary using above function if no
40039         value range information.
40041 2017-01-24  Pekka Jääskeläinen  <pekka@parmance.com>
40042             Martin Jambor  <mjambor@suse.cz>
40044         * brig-builtins.def: New file.
40045         * builtins.def (DEF_HSAIL_BUILTIN): New macro.
40046         (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
40047         (DEF_HSAIL_SAT_BUILTIN): Likewise.
40048         (DEF_HSAIL_INTR_BUILTIN): Likewise.
40049         (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
40050         * builtin-types.def (BT_INT8): New.
40051         (BT_INT16): Likewise.
40052         (BT_UINT8): Likewise.
40053         (BT_UINT16): Likewise.
40054         (BT_FN_ULONG): Likewise.
40055         (BT_FN_UINT_INT): Likewise.
40056         (BT_FN_UINT_ULONG): Likewise.
40057         (BT_FN_UINT_LONG): Likewise.
40058         (BT_FN_UINT_PTR): Likewise.
40059         (BT_FN_ULONG_PTR): Likewise.
40060         (BT_FN_INT8_FLOAT): Likewise.
40061         (BT_FN_INT16_FLOAT): Likewise.
40062         (BT_FN_UINT32_FLOAT): Likewise.
40063         (BT_FN_UINT16_FLOAT): Likewise.
40064         (BT_FN_UINT8_FLOAT): Likewise.
40065         (BT_FN_UINT64_FLOAT): Likewise.
40066         (BT_FN_UINT16_UINT32): Likewise.
40067         (BT_FN_UINT32_UINT16): Likewise.
40068         (BT_FN_UINT16_UINT16_UINT16): Likewise.
40069         (BT_FN_INT_PTR_INT): Likewise.
40070         (BT_FN_UINT_PTR_UINT): Likewise.
40071         (BT_FN_LONG_PTR_LONG): Likewise.
40072         (BT_FN_ULONG_PTR_ULONG): Likewise.
40073         (BT_FN_VOID_UINT64_UINT64): Likewise.
40074         (BT_FN_UINT8_UINT8_UINT8): Likewise.
40075         (BT_FN_INT8_INT8_INT8): Likewise.
40076         (BT_FN_INT16_INT16_INT16): Likewise.
40077         (BT_FN_INT_INT_INT): Likewise.
40078         (BT_FN_UINT_FLOAT_UINT): Likewise.
40079         (BT_FN_FLOAT_UINT_UINT): Likewise.
40080         (BT_FN_ULONG_UINT_UINT): Likewise.
40081         (BT_FN_ULONG_UINT_PTR): Likewise.
40082         (BT_FN_ULONG_ULONG_ULONG): Likewise.
40083         (BT_FN_UINT_UINT_UINT): Likewise.
40084         (BT_FN_VOID_UINT_PTR): Likewise.
40085         (BT_FN_UINT_UINT_PTR: Likewise.
40086         (BT_FN_UINT32_UINT64_PTR): Likewise.
40087         (BT_FN_INT_INT_UINT_UINT): Likewise.
40088         (BT_FN_UINT_UINT_UINT_UINT): Likewise.
40089         (BT_FN_UINT_UINT_UINT_PTR): Likewise.
40090         (BT_FN_UINT_ULONG_ULONG_UINT): Likewise.
40091         (BT_FN_ULONG_ULONG_ULONG_ULONG): Likewise.
40092         (BT_FN_LONG_LONG_UINT_UINT): Likewise.
40093         (BT_FN_ULONG_ULONG_UINT_UINT): Likewise.
40094         (BT_FN_VOID_UINT32_UINT64_PTR): Likewise.
40095         (BT_FN_VOID_UINT32_UINT32_PTR): Likewise.
40096         (BT_FN_UINT_UINT_UINT_UINT_UINT): Likewise.
40097         (BT_FN_UINT_FLOAT_FLOAT_FLOAT_FLOAT): Likewise.
40098         (BT_FN_ULONG_ULONG_ULONG_UINT_UINT): Likewise.
40099         * doc/frontends.texi: List BRIG FE.
40100         * doc/install.texi (Testing): Add BRIG tesring requirements.
40101         * doc/invoke.texi (Overall Options): Mention BRIG.
40102         * doc/standards.texi (Standards): Doucment BRIG HSA version.
40104 2017-01-24  Richard Biener  <rguenther@suse.de>
40106         PR translation/79208
40107         * ipa-devirt.c (odr_types_equivalent_p): Fix typo in diagnostic.
40109 2017-01-24  Martin Jambor  <mjambor@suse.cz>
40111         PR bootstrap/79198
40112         * ipa-prop.c (ipa_free_all_node_params): Call summary destructor.
40113         * ipa-prop.c (ipa_node_params_t::insert): Initialize fields known_csts
40114         and known_contexts.
40116 2017-01-24  Aldy Hernandez  <aldyh@redhat.com>
40118         PR middle-end/79123
40119         * gimple-ssa-warn-alloca.c (alloca_call_type): Make sure
40120         casts from signed to unsigned really don't have a range.
40122 2017-01-24  Markus Trippelsdorf  <markus@trippelsdorf.de>
40124         * gimple-ssa-sprintf.c (format_floating): Change MPFR_RNDx to
40125         GMP_RNDx for compatiblity.
40127 2017-01-24  Martin Liska  <mliska@suse.cz>
40129         PR bootstrap/79132
40130         * tree-ssa-reassoc.c (rewrite_expr_tree_parallel): Insert assert
40131         that would prevent us to call alloca with -1 as argument.
40133 2017-01-24  Jakub Jelinek  <jakub@redhat.com>
40135         * dwarf2out.c (output_compilation_unit_header, output_file_names):
40136         Avoid -Wformat-security warning.
40138 2017-01-23  Andrew Pinski  <apinski@cavium.com>
40140         * config/aarch64/aarch64.c (thunderx2t99_addrcost_table): Improve
40141         cost table.
40143 2017-01-23  Martin Sebor  <msebor@redhat.com>
40145         PR middle-end/78703
40146         * gimple-ssa-sprintf.c (warn_level): New global.
40147         (format_integer): Use it here and throughout the rest of the file.
40148         Use the same switch to compute sign as base.
40149         (maybe_warn): New function.
40150         (format_directive): Factor out warnings into maybe_warn.
40151         Add debugging output.  Use warn_level.
40152         (add_bytes): Use warn_level.
40153         (pass_sprintf_length::compute_format_length): Add debugging output.
40154         (try_substitute_return_value): Same.
40155         (pass_sprintf_length::handle_gimple_call): Set and use warn_level.
40157         PR middle-end/78703
40158         * gimple-ssa-sprintf.c (struct format_result): Remove constant member.
40159         (struct fmtresult, format_integer, format_floating): Adjust.
40160         (fmtresult::fmtresult): Set max correctly in two argument ctor.
40161         (get_string_length, format_string,format_directive): Same.
40162         (pass_sprintf_length::compute_format_length): Same.
40163         (try_substitute_return_value): Simplify slightly.
40165         PR middle-end/78703
40166         * gimple-ssa-sprintf.c (pass_sprintf_length::gate): Adjust formatting.
40167         (fmtresult::operator+=): Outlined.
40168         (struct fmtresult): Add ctors.
40169         (struct conversion_spec): Rename...
40170         (struct directive): ...to this.  Add and remove data members.
40171         (directive::set_width, directive::set_precision): New functions.
40172         (format_percent): Use fmtresult ctor.
40173         (get_width_and_precision): Remove.
40174         (format_integer): Make naming changes.  Avoid computing width and
40175         precision.
40176         (format_floating): Same.  Adjust indentation.
40177         (format_character, format_none): New functions.
40178         (format_string): Moved character handling to format_character.
40179         (format_directive): Remove arguments, change return type.
40180         (parse_directive): New function.
40181         (pass_sprintf_length::compute_format_length): Move directive
40182         parsing to parse_directive.
40184 2017-01-23  Jakub Jelinek  <jakub@redhat.com>
40186         * tree.h (assign_assembler_name_if_neeeded): Rename to ...
40187         (assign_assembler_name_if_needed): ... this.
40188         * tree.c (assign_assembler_name_if_neeeded): Rename to ...
40189         (assign_assembler_name_if_needed): ... this.
40190         (free_lang_data_in_cgraph): Adjust callers.
40191         * cgraphunit.c (cgraph_node::analyze): Likewise.
40192         * omp-expand.c (expand_omp_taskreg, expand_omp_target):
40193         Likewise.
40195 2017-01-23  Richard Biener  <rguenther@suse.de>
40197         PR tree-optimization/79088
40198         PR tree-optimization/79188
40199         * tree-ssa-threadupdate.c (mark_threaded_blocks): Move code
40200         resetting loop bounds after last path deletion.  Reset loop
40201         bounds of the target loop, make code match the comments.
40202         * tree-ssa-threadbackwards.c (pass_early_thread_jumps::execute):
40203         Make sure loops need no fixups.
40205 2017-01-23  Kelvin Nilsen  <kelvin@gcc.gnu.org>
40207         * config/rs6000/rs6000-builtin.def (VSIEDPF): Add scalar insert
40208         exponent support with double type for first argument.
40209         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Changed
40210         type returned by __builtin_vec_extract_sig,
40211         __builtin_vec_extract_sig_sp, and __builtin_vec_extract_sig_dp
40212         functions from "vector int" to "vector unsigned int" or from
40213         "vector long long int" to "vector unsigned long long int".
40214         Changed type returned by __builtin_vec_extract_exp,
40215         __builtin_vec_extract_exp_sp, and __builtin_vec_extract_exp_dp
40216         functions from "vector int" to "vector unsigned int" or from
40217         "vector long long int" to "vector unsigned long long int".
40218         Changed return type of __builtin_vec_test_data_class,
40219         __builtin_vec_test_data_class_sp, and
40220         __builtin_vec_test_data_class_dp from "vector int" to
40221         "vector bool int" or from "vector long long int" to "vector bool
40222         long long int" and changed second argument type from "unsigned
40223         int" to "int".  Added new overloaded function forms "vector float
40224         __builtin_vec_insert_exp (vector float, vector unsigned int)" and
40225         "vector float __builtin_vec_insert_exp_sp (vector float, vector
40226         unsigned int)" and "vector double __builtin_vec_insert_exp (vector
40227         double, vector unsigned long long int)" and "vector double
40228         __builtin_vec_insert_exp_dp (vector double, vector unsigned long
40229         long int)".  Changed return type of
40230         __builtin_scalar_test_data_class and
40231         __builtin_scalar_test_data_class_sp and
40232         __builtin_scalar_test_data_class_dp from "unsigned int" to "bool
40233         int" and changed second argument from "unsigned int" to "int".
40234         Changed type returned by __builtin_scalar_test_neg,
40235         __builtin_scalar_test_neg_sp, and __builtin_scalar_test_neg_dp
40236         from "int" to "bool int".  Added new overloaded function form
40237         "double __builtin_scalar_insert_exp (double, unsigned long long int)".
40238         * config/rs6000/vsx.md (xsiexpdpf): New insn for scalar insert
40239         exponent double-precision with floating point first argument.
40240         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
40241         documentation of scalar_test_data_class, scalar_test_neg,
40242         scalar_extract_sig, scalar_extract_exp, scalar_insert_exp,
40243         vector_extract_exp, vec_extract_sig, vec_insert_exp, and
40244         vec_test_data_class built-in functions to reflect refinements in
40245         their type signatures.
40247 2017-01-23  Andreas Tobler  <andreast@gcc.gnu.org>
40249         * config/aarch64/aarch64.c (aarch64_elf_asm_constructor): Increase
40250         size of buf.
40251         (aarch64_elf_asm_destructor): Likewise.
40253 2017-01-23  Bernd Schmidt  <bschmidt@redhat.com>
40255         PR rtl-optimization/78634
40256         * config/i386/i386.c (ix86_max_noce_ifcvt_seq_cost): New function.
40257         (TARGET_MAX_NOCE_IFCVT_SEQ_COST): Define.
40258         * ifcvt.c (noce_try_cmove): Add missing cost check.
40260         PR rtl-optimization/71724
40261         * combine.c (if_then_else_cond): Look for situations where it is
40262         beneficial to undo the work of one of the recursive calls.
40264 2017-01-23  Bin Cheng  <bin.cheng@arm.com>
40266         PR tree-optimization/70754
40267         * tree-predcom.c (stmt_combining_refs): New parameter INSERT_BEFORE.
40268         (reassociate_to_the_same_stmt): New parameter INSERT_BEFORE.  Insert
40269         combined stmt before it if not NULL.
40270         (combine_chains): Process refs reversely and compute dominance point
40271         for root ref.
40273 2017-01-23  Martin Liska  <mliska@suse.cz>
40275         PR tree-optimization/79196
40276         * tree-ssa-strlen.c (fold_strstr_to_memcmp): Rename to ...
40277         (fold_strstr_to_strncmp): ... this.  Fold the pattern to strncmp
40278         instead of memcmp.
40279         (strlen_optimize_stmt): Call the renamed function.
40281 2017-01-23  Michael Matz  <matz@suse.de>
40283         PR tree-optimization/78384
40284         * tree-ssa-loop-split.c (patch_loop_exit): Use correct edge.
40286 2017-01-23  Richard Biener  <rguenther@suse.de>
40288         PR tree-optimization/79186
40289         * tree-vrp.c (register_new_assert_for): Make sure we've seen
40290         both incoming edges before moving an assert.
40292 2017-01-23  Martin Jambor  <mjambor@suse.cz>
40294         * ipa-prop.c (load_from_param_1): Removed.
40295         (load_from_unmodified_param): Bits from load_from_param_1 put back
40296         here.
40297         (load_from_param): Removed.
40298         (compute_complex_assign_jump_func): Removed stmt2 and just replaced it
40299         with stmt.  Reverted back to use of load_from_unmodified_param.
40301 2017-01-23  Martin Jambor  <mjambor@suse.cz>
40303         PR ipa/79108
40304         * ipa-prop.h (ipa_param_descriptor): Anotate with with GTY(()).
40305         (ipa_node_params): Annotate with GTY((for_user)).  Make descriptors
40306         field a pointer to garbage collected vector, mark lattices and
40307         ipcp_orig_node with GTY((skip)).
40308         (ipa_get_param_count): Adjust to descriptors being a pointer.
40309         (ipa_get_param): Likewise.
40310         (ipa_get_type): Likewise.
40311         (ipa_get_param_move_cost): Likewise.
40312         (ipa_set_param_used): Likewise.
40313         (ipa_get_controlled_uses): Likewise.
40314         (ipa_set_controlled_uses): Likewise.
40315         (ipa_is_param_used): Likewise.
40316         (ipa_node_params_t): Move into garbage collector.  New methods insert
40317         and remove.
40318         (ipa_node_params_sum): Annotate wth GTY(()).
40319         (ipa_check_create_node_params): Adjust to ipa_node_params_sum being
40320         garbage collected.
40321         (ipa_load_from_parm_agg): Adjust declaration.
40322         * ipa-icf.c (param_used_p): Adjust to descriptors being a pointer.
40323         * ipa-profile.c (ipa_profile): Likewise.
40324         * ipa-prop.c (ipa_get_param_decl_index_1): Likewise.
40325         (ipa_populate_param_decls): Make descriptors parameter garbage
40326         collected.
40327         (ipa_dump_param): Adjust to descriptors being a pointer.
40328         (ipa_alloc_node_params): Likewise.
40329         (ipa_initialize_node_params): Likewise.
40330         (load_from_param_1): Make descriptors parameter garbage collected.
40331         (load_from_unmodified_param): Likewise.
40332         (load_from_param): Likewise.
40333         (ipa_load_from_parm_agg): Likewise.
40334         (ipa_node_params::~ipa_node_params): Removed.
40335         (ipa_free_all_node_params): Remove call to delete operator.
40336         (ipa_node_params_t::insert): New.
40337         (ipa_node_params_t::remove): Likewise.
40338         (ipa_node_params_t::duplicate): Adjust to descriptors being a pointer,
40339         copy known_csts and known_contexts vectors.
40340         (ipa_read_node_info): Adjust to descriptors being a pointer.
40341         (ipcp_modif_dom_walker): Make m_descriptors field garbage
40342         collected.
40343         (ipcp_transform_function): Make descriptors variable garbage
40344         collected.
40346 2017-01-23  Andrew Senkevich  <andrew.senkevich@intel.com>
40348         * config/i386/avx512bwintrin.h: Add k-mask test, kortest intrinsics.
40349         * config/i386/avx512dqintrin.h: Ditto.
40350         * config/i386/avx512fintrin.h: Ditto.
40351         * config/i386/i386.c: Handle new builtins.
40352         * config/i386/i386-builtin.def: Add new builtins.
40353         * config/i386/sse.md (ktest<mode>, kortest<mode>): New.
40354         (UNSPEC_KORTEST, UNSPEC_KTEST): New.
40356 2017-01-23  Jakub Jelinek  <jakub@redhat.com>
40357             Martin Liska  <mliska@suse.cz>
40359         * asan.h: Define ASAN_USE_AFTER_SCOPE_ATTRIBUTE.
40360         * asan.c (asan_expand_poison_ifn): Support stores and use
40361         appropriate ASAN report function.
40362         * internal-fn.c (expand_ASAN_POISON_USE): New function.
40363         * internal-fn.def (ASAN_POISON_USE): Declare.
40364         * tree-into-ssa.c (maybe_add_asan_poison_write): New function.
40365         (maybe_register_def): Create ASAN_POISON_USE when sanitizing.
40366         * tree-ssa-dce.c (eliminate_unnecessary_stmts): Remove
40367         ASAN_POISON calls w/o LHS.
40368         * tree-ssa.c (execute_update_addresses_taken): Create clobber
40369         for ASAN_MARK (UNPOISON, &x, ...) in order to prevent usage of a LHS
40370         from ASAN_MARK (POISON, &x, ...) coming to a PHI node.
40371         * gimplify.c (asan_poison_variables): Add attribute
40372         use_after_scope_memory to variables that really needs to live
40373         in memory.
40374         * tree-ssa.c (is_asan_mark_p): Do not rewrite into SSA when
40375         having the attribute.
40377 2017-01-23  Martin Liska  <mliska@suse.cz>
40379         * asan.c (create_asan_shadow_var): New function.
40380         (asan_expand_poison_ifn): Likewise.
40381         * asan.h (asan_expand_poison_ifn): New declaration.
40382         * internal-fn.c (expand_ASAN_POISON): Likewise.
40383         * internal-fn.def (ASAN_POISON): New builtin.
40384         * sanopt.c (pass_sanopt::execute): Expand
40385         asan_expand_poison_ifn.
40386         * tree-inline.c (copy_decl_for_dup_finish): Make function
40387         external.
40388         * tree-inline.h (copy_decl_for_dup_finish): Likewise.
40389         * tree-ssa.c (is_asan_mark_p): New function.
40390         (execute_update_addresses_taken): Rewrite local variables
40391         (identified just by use-after-scope as addressable) into SSA.
40393 2017-01-22  Gerald Pfeifer  <gerald@pfeifer.com>
40395         * doc/install.texi (Specific): opensource.apple.com uses https
40396         now. Remove trailing slash.
40398 2017-01-22  Gerald Pfeifer  <gerald@pfeifer.com>
40400         * README.Portability: Remove note on an Irix compatibility issue.
40402 2017-01-22  Dimitry Andric <dim@FreeBSD.org>
40404         * gcov.c (INCLUDE_ALGORITHM): Define.
40405         (INCLUDE_VECTOR): Define.
40406         No longer include <vector> and <algorithm> directly.
40408 2017-01-21  Gerald Pfeifer  <gerald@pfeifer.com>
40410         * doc/extend.texi (Thread-Local): Change www.akkadia.org reference
40411         to https.
40412         * doc/invoke.texi (Code Gen Options): Ditto.
40414 2017-01-21  Jan Hubicka  <hubicka@ucw.cz>
40416         PR lto/78407
40417         * cfg.c (update_bb_profile_for_threading): Fix updating of probablity.
40419 2017-01-21  Bernd Schmidt  <bschmidt@redhat.com>
40421         rtl-optimization/79125
40422         * cprop.c (local_cprop_pass): Handle cases where we make an
40423         unconditional trap.
40425 2017-01-20  Segher Boessenkool  <segher@kernel.crashing.org>
40427         PR target/61729
40428         PR target/77850
40429         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Adjust address to
40430         read from, for big endian.
40432 2017-01-20  Jiong Wang  <jiong.wang@arm.com>
40434         * config/aarch64/aarch64-builtins.c (aarch64_init_builtins): Register
40435         register pauth builtins for LP64 only.
40437 2017-01-20  Marek Polacek  <polacek@redhat.com>
40439         PR c/79152
40440         * gimplify.c (should_warn_for_implicit_fallthrough): Handle consecutive
40441         non-case labels.
40443 2017-01-20  Alexander Monakov  <amonakov@ispras.ru>
40445         * omp-expand.c (expand_omp_simd): Clear PROP_gimple_lomp_dev regardless
40446         of safelen status.
40447         * omp-offload.c (pass_omp_device_lower::gate): Use PROP_gimple_lomp_dev.
40448         * passes.c (dump_properties): Handle PROP_gimple_lomp_dev.
40449         * tree-inline.c (expand_call_inline): Propagate PROP_gimple_lomp_dev.
40451 2017-01-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
40453         PR target/71270
40454         * config/arm/arm.c (neon_valid_immediate): Reject vector constants
40455         in big-endian mode when they are not a single duplicated value.
40457 2017-01-20  Richard Biener  <rguenther@suse.de>
40459         * BASE-VER: Bump to 7.0.1.
40461 2017-01-20  Alexander Monakov  <amonakov@ispras.ru>
40463         * omp-low.c (omplow_simd_context): New struct.  Use it...
40464         (lower_rec_simd_input_clauses): ...here and...
40465         (lower_rec_input_clauses): ...here to hold common data.  Adjust all
40466         references to idx, lane, max_vf, is_simt.
40468 2017-01-20  Graham Markall  <graham.markall@embecosm.com>
40470         * config/arc/arc.h (LINK_SPEC): Use arclinux_nps emulation when
40471         mcpu=nps400.
40473 2017-01-20  Martin Jambor  <mjambor@suse.cz>
40475         * hsa.h: Renaed to hsa-common.h.  Adjusted a comment.
40476         * hsa.c: Renaed to hsa-common.c.  Change include of gt-hsa.h to
40477         gt-hsa-common.h.
40478         * Makefile.in (OBJS): Rename hsa.o to hsa-common.o.
40479         (GTFILES): Rename hsa.c to hsa-common.c.
40480         * hsa-brig.c: Change include of hsa.h to hsa-common.h.
40481         * hsa-dump.c: Likewise.
40482         * hsa-gen.c: Likewise.
40483         * hsa-regalloc.c: Likewise.
40484         * ipa-hsa.c: Likewise.
40485         * omp-expand.c: Likewise.
40486         * omp-low.c: Likewise.
40487         * toplev.c: Likewise.
40489 2017-01-20  Marek Polacek  <polacek@redhat.com>
40491         PR c/64279
40492         * doc/invoke.texi: Document -Wduplicated-branches.
40493         * fold-const.c (operand_equal_p): Handle MODIFY_EXPR, INIT_EXPR,
40494         COMPOUND_EXPR, PREDECREMENT_EXPR, PREINCREMENT_EXPR,
40495         POSTDECREMENT_EXPR, POSTINCREMENT_EXPR, CLEANUP_POINT_EXPR, EXPR_STMT,
40496         STATEMENT_LIST, and RETURN_EXPR.  For non-pure non-const functions
40497         return 0 only when not OEP_LEXICOGRAPHIC.
40498         (fold_build_cleanup_point_expr): Use the expression
40499         location when building CLEANUP_POINT_EXPR.
40500         * tree-core.h (enum operand_equal_flag): Add OEP_LEXICOGRAPHIC.
40501         * tree.c (add_expr): Handle error_mark_node.
40503 2017-01-20  Martin Liska  <mliska@suse.cz>
40505         PR lto/69188
40506         * tree-profile.c (init_ic_make_global_vars): Do not call
40507         finalize_decl.
40508         (gimple_init_gcov_profiler): Likewise.
40510 2017-01-20  Martin Liska  <mliska@suse.cz>
40512         PR ipa/71190
40513         * cgraph.h (maybe_create_reference): Remove argument and
40514         update comment.
40515         * cgraphclones.c (cgraph_node::create_virtual_clone): Remove one
40516         argument.
40517         * ipa-cp.c (create_specialized_node): Likewise.
40518         * symtab.c (symtab_node::maybe_create_reference): Handle
40519         VAR_DECLs and ADDR_EXPRs and select ipa_ref_use type.
40521 2017-01-20  Martin Liska  <mliska@suse.cz>
40523         * read-rtl-function.c (function_reader::create_function): Use
40524         build_decl instread of build_decl_stat.
40526 2017-01-20  Andrew Senkevich  <andrew.senkevich@intel.com>
40528         * config/i386/avx512bwintrin.h: Add k-mask registers shift intrinsics.
40529         * config/i386/avx512dqintrin.h: Ditto.
40530         * config/i386/avx512fintrin.h: Ditto.
40531         * config/i386/i386-builtin-types.def: Add new types.
40532         * config/i386/i386.c: Handle new types.
40533         * config/i386/i386-builtin.def (__builtin_ia32_kshiftliqi)
40534         (__builtin_ia32_kshiftlihi, __builtin_ia32_kshiftlisi)
40535         (__builtin_ia32_kshiftlidi, __builtin_ia32_kshiftriqi)
40536         (__builtin_ia32_kshiftrihi, __builtin_ia32_kshiftrisi)
40537         (__builtin_ia32_kshiftridi): New.
40538         * config/i386/sse.md (k<code><mode>): Rename *k<code><mode>.
40540 2017-01-19  Segher Boessenkool  <segher@kernel.crashing.org>
40542         PR target/78875
40543         PR target/79140
40544         * config/rs6000/rs6000.c (TARGET_STACK_PROTECT_GUARD): Unconditionally
40545         define to rs6000_init_stack_protect_guard.
40546         (rs6000_init_stack_protect_guard): New function.
40548 2017-01-19  Matthew Fortune  <matthew.fortune@imgtec.com>
40549             Yunqiang Su  <yunqiang.su@imgtec.com>
40551         * config.gcc (supported_defaults): Add madd4.
40552         (with_madd4): Add validation.
40553         (all_defaults): Add madd4.
40554         * config/mips/mips.opt (mmadd4): New option.
40555         * config/mips/mips.h (OPTION_DEFAULT_SPECS): Add a default for
40556         mmadd4.
40557         (TARGET_CPU_CPP_BUILTINS): Add builtin_define for
40558         __mips_no_madd4.
40559         (ISA_HAS_UNFUSED_MADD4): Gate with mips_madd4.
40560         (ISA_HAS_FUSED_MADD4): Likewise.
40561         * doc/invoke.texi (-mmadd4): Document the new option.
40562         * doc/install.texi (--with-madd4): Document the new option.
40564 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40566         * config/aarch64/aarch64-builtins.c (enum aarch64_builtins): New
40567         entries for AARCH64_PAUTH_BUILTIN_XPACLRI,
40568         AARCH64_PAUTH_BUILTIN_PACIA1716, AARCH64_PAUTH_BUILTIN_AUTIA1716.
40569         (aarch64_init_pauth_hint_builtins): New.
40570         (aarch64_init_builtins): Call aarch64_init_pauth_hint_builtins.
40571         (aarch64_expand_builtin): Expand new builtins.
40573 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40575         * reg-notes.def (CFA_TOGGLE_RA_MANGLE): New reg-note.
40576         * combine-stack-adj.c (no_unhandled_cfa): Handle
40577         REG_CFA_TOGGLE_RA_MANGLE.
40578         * dwarf2cfi.c (dwarf2out_frame_debug): Handle REG_CFA_TOGGLE_RA_MANGLE.
40579         * config/aarch64/aarch64.c (aarch64_expand_prologue): Generates DWARF
40580         info for return address signing.
40581         (aarch64_expand_epilogue): Likewise.
40583 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40585         * config/aarch64/aarch64-opts.h (aarch64_function_type): New enum.
40586         * config/aarch64/aarch64-protos.h
40587         (aarch64_return_address_signing_enabled): New declaration.
40588         * config/aarch64/aarch64.c (aarch64_return_address_signing_enabled):
40589         New function.
40590         (aarch64_expand_prologue): Sign return address before it's pushed onto
40591         stack.
40592         (aarch64_expand_epilogue): Authenticate return address fetched from
40593         stack.
40594         (aarch64_override_options): Sanity check for ILP32 and ISA level.
40595         (aarch64_attributes): New function attributes for "sign-return-address".
40596         * config/aarch64/aarch64.md (UNSPEC_AUTI1716, UNSPEC_AUTISP,
40597         UNSPEC_PACI1716, UNSPEC_PACISP, UNSPEC_XPACLRI): New unspecs.
40598         ("*do_return"): Generate combined instructions according to key index.
40599         ("<pauth_mnem_prefix>sp", "<pauth_mnem_prefix1716", "xpaclri"): New.
40600         * config/aarch64/iterators.md (PAUTH_LR_SP, PAUTH_17_16): New integer
40601         iterators.
40602         (pauth_mnem_prefix, pauth_hint_num_a): New integer attributes.
40603         * config/aarch64/aarch64.opt (msign-return-address=): New.
40604         * doc/extend.texi (AArch64 Function Attributes): Documents
40605         "sign-return-address=".
40606         * doc/invoke.texi (AArch64 Options): Documents "-msign-return-address=".
40608 2017-01-19  Matthew Fortune  <matthew.fortune@imgtec.com>
40610         * doc/invoke.texi: Add missing -mlxc1-sxc1 options to
40611         overall option summary.
40613 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40615         * config/aarch64/aarch64-arches.def: New entry for "armv8.3-a".
40616         * config/aarch64/aarch64.h (AARCH64_FL_V8_3, AARCH64_FL_FOR_ARCH8_3,
40617         AARCH64_ISA_V8_3, TARGET_ARMV8_3): New.
40618         * doc/invoke.texi (AArch64 Options): Document "armv8.3-a".
40620 2017-01-19  Michael Meissner  <meissner@linux.vnet.ibm.com>
40622         * config/rs6000/rs6000-cpus.def (ISA_3_0_MASKS_SERVER): Enable
40623         -mpower9-minmax by default for -mcpu=power9.
40624         (ISA_3_MASKS_IEEE): Require -mvsx-small-integer to enable IEEE
40625         128-bit floating point.
40627 2017-01-20  Alan Modra  <amodra@gmail.com>
40629         * config/rs6000/rs6000.md (cmpstrnsi, cmpstrsi): Fail if
40630         optimizing for size.
40632 2017-01-20  Alan Modra  <amodra@gmail.com>
40634         PR target/79144
40635         * config/rs6000/rs6000.c (expand_strn_compare): Get the asm name
40636         for strcmp and strncmp from corresponding builtin decl.
40638 2017-01-19  Uros Bizjak  <ubizjak@gmail.com>
40640         * config.gcc (x86_64-*-rtems*): Use i386/rtemself.h
40641         instead of i386/rtems-64.h.
40642         * config/i386/rtems-64.h: Remove.
40644 2017-01-19  Uros Bizjak  <ubizjak@gmail.com>
40646         PR target/78478
40647         Revert:
40648         2013-11-05  Uros Bizjak  <ubizjak@gmail.com>
40650         * config/i386/rtemself.h (LONG_DOUBLE_TYPE_SIZE): New define.
40652 2017-01-19  Tamar Christina  <tamar.christina@arm.com>
40654         * config/aarch64/aarch64.c (aarch64_simd_gen_const_vector_dup):
40655         Change int to HOST_WIDE_INT.
40656         * config/aarch64/aarch64-protos.h
40657         (aarch64_simd_gen_const_vector_dup): Likewise.
40658         * config/aarch64/aarch64-simd.md: Add copysign<mode>3.
40660 2017-01-19  David Malcolm  <dmalcolm@redhat.com>
40662         * langhooks-def.h (lhd_type_for_size): New decl.
40663         (LANG_HOOKS_TYPE_FOR_SIZE): Define as lhd_type_for_size.
40664         * langhooks.c (lhd_type_for_size): New function, taken from
40665         lto_type_for_size.
40667 2017-01-19  Pat Haugen  <pthaugen@us.ibm.com>
40669         * config/rs6000/power9.md (power9-alu): Remove 'cmp' type and add
40670         define_bypass for CR latency.
40671         (power9-cracked-alu): Update bypass latency and remove power9-branch.
40672         (power9-alu2): Add define_bypass for CR latency.
40673         (power9-cmp): New.
40674         (power9-mul): Update insn latency.
40675         (power9-mul-compare): Update insn latency, bypass latency and remove
40676         power9-branch.
40678 2016-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
40680         * config/aarch64/aarch64-protos.h (aarch64_nopcrelative_literal_loads):
40681         Delete.
40682         * config/aarch64/aarch64.md
40683         (aarch64_reload_movcp<GPF_TF:mode><P:mode>): Delete reference to
40684         aarch64_nopcrelative_literal_loads.
40685         (aarch64_reload_movcp<VALL:mode><P:mode>): Likewise.
40687 2017-01-19  Chenghua Xu  <paul.hua.gm@gmail.com>
40689         * config/mips/mips.h (ISA_HAS_FUSED_MADD4): Enable for
40690         TARGET_LOONGSON_3A.
40691         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_LOONGSON_3A.
40693 2017-01-19  Doug Gilmore  <doug.gilmore@imgtec.com>
40695         PR target/78176
40696         * config.gcc (supported_defaults): Add lxc1-sxc1.
40697         (with_lxc1_sxc1): Add validation.
40698         (all_defaults): Add lxc1-sxc1.
40699         * config/mips/mips.opt (mlxc1-sxc1): New option.
40700         * config/mips/mips.h (OPTION_DEFAULT_SPECS): Add a default for
40701         mlxc1-sxc1.
40702         (TARGET_CPU_CPP_BUILTINS): Add builtin_define for
40703         __mips_no_lxc1_sxc1.
40704         (ISA_HAS_LXC1_SXC1): Gate with mips_lxc1_sxc1.
40705         * doc/invoke.texi (-mlxc1-sxc1): Document the new option.
40706         * doc/install.texi (--with-lxc1-sxc1): Document the new option.
40708 2017-01-19  Richard Biener  <rguenther@suse.de>
40710         PR tree-optimization/72488
40711         * tree-ssa-sccvn.c (run_scc_vn): When we abort the VN make
40712         sure to restore SSA info.
40713         * tree-ssa.c (verify_ssa): Verify SSA info is not shared.
40715 2017-01-19  Richard Earnshaw  <rearnsha@arm.com>
40717         PR rtl-optimization/79121
40718         * expr.c (expand_expr_real_2, case LSHIFT_EXPR): Look at the signedness
40719         of the inner type when shifting an extended value.
40721 2017-01-17  Jan Hubicka  <hubicka@ucw.cz>
40723         PR lto/78407
40724         * symtab.c (symtab_node::equal_address_to): Fix comparing of
40725         interposable aliases.
40727 2017-01-18  Peter Bergner  <bergner@vnet.ibm.com>
40729         PR target/78516
40730         * config/rs6000/spe.md (mov_si<mode>_e500_subreg0): Fix constraints.
40731         Use the evmergelohi instruction.
40732         (mov_si<mode>_e500_subreg4_2_le): Likewise.
40733         (mov_sitf_e500_subreg8_2_be): Likewise.
40734         (mov_sitf_e500_subreg12_2_le): Likewise.
40735         (mov_si<mode>_e500_subreg0_2_le): Fix constraints.
40736         (mov_si<mode>_e500_subreg4_2_be): Likewise.
40737         (mov_sitf_e500_subreg8_2_le): Likewise.
40738         (mov_sitf_e500_subreg12_2_be): Likewise.
40740 2017-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40742         * config/rs6000/altivec.md (altivec_vbpermq): Change "type"
40743         attribute from vecsimple to vecperm.
40744         (altivec_vbpermq2): Likewise.
40746 2017-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40748         PR target/79040
40749         * config/rs6000/altivec.h: Fix typo of vec_cntlz to vec_cnttz.
40751 2017-01-18  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
40752         * config/rs6000/rs6000-protos.h (expand_strn_compare): Add arg.
40753         * config/rs6000/rs6000.c (expand_strn_compare): Add ability to expand
40754         strcmp. Fix bug where comparison didn't stop with zero byte. Fix
40755         case where N arg is SIZE_MAX.
40756         * config/rs6000/rs6000.md (cmpstrnsi): Args to expand_strn_compare.
40757         (cmpstrsi): Add pattern.
40759 2017-01-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
40761         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
40762         __builtin_vec_revb builtins.
40763         * config/rs6000/rs6000-builtins.def (P9V_BUILTIN_XXBRQ_V16QI): Add
40764         built-in functions to support generation of the ISA 3.0 XXBR<x>
40765         vector byte reverse instructions.
40766         (P9V_BUILTIN_XXBRQ_V1TI): Likewise.
40767         (P9V_BUILTIN_XXBRD_V2DI): Likewise.
40768         (P9V_BUILTIN_XXBRD_V2DF): Likewise.
40769         (P9V_BUILTIN_XXBGW_V4SI): Likewise.
40770         (P9V_BUILTIN_XXBGW_V4SF): Likewise.
40771         (P9V_BUILTIN_XXBGH_V8HI): Likewise.
40772         (P9V_BUILTIN_VEC_REVB): Likewise.
40773         * config/rs6000/vsx.md (p9_xxbrq_v1ti): New insns/expanders to
40774         generate the ISA 3.0 XXBR<x> vector byte reverse instructions.
40775         (p9_xxbrq_v16qi): Likewise.
40776         (p9_xxbrd_<mode>, VSX_D iterator): Likewise.
40777         (p9_xxbrw_<mode>, VSX_W iterator): Likewise.
40778         (p9_xxbrh_v8hi): Likewise.
40779         * config/rs6000/altivec.h (vec_revb): Define if ISA 3.0.
40780         * doc/extend.texi (RS/6000 Altivec Built-ins): Document the
40781         vec_revb built-in functions.
40783 2017-01-18  Uros Bizjak  <ubizjak@gmail.com>
40785         PR rtl-optimization/78952
40786         * config/i386/i386.md (any_extract): New code iterator.
40787         (*insvqi_2): Use any_extract for source operand.
40788         (*insvqi_3): Use any_shiftrt for source operand.
40790 2017-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
40792         * config/aarch64/aarch64.c (aarch64_sched_adjust_priority)
40793         New function.
40794         (TARGET_SCHED_ADJUST_PRIORITY): Define target hook.
40796 2017-01-18  Matthias Klose  <doko@ubuntu.com>
40798         * doc/install.texi: Allow default for --with-target-bdw-gc-include.
40800 2016-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40802         * config/rs6000/altivec.h (vec_bperm): Change #define.
40803         * config/rs6000/altivec.md (UNSPEC_VBPERMD): New enum constant.
40804         (altivec_vbpermq2): New define_insn.
40805         (altivec_vbpermd): Likewise.
40806         * config/rs6000/rs6000-builtin.def (VBPERMQ2): New monomorphic
40807         function interface.
40808         (VBPERMD): Likewise.
40809         (VBPERM): New polymorphic function interface.
40810         * config/rs6000/r6000-c.c (altivec_overloaded_builtins_table):
40811         Add entries for P9V_BUILTIN_VEC_VBPERM.
40812         * doc/extend.texi: Add interfaces for vec_bperm.
40814 2017-01-18  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
40816         * config/s390/s390-c.c (s390_expand_overloaded_builtin): Downcase
40817         first letter of error messages.
40818         (s390_resolve_overloaded_builtin): Likewise.
40819         * config/s390/s390.c (s390_expand_builtin): Likewise.
40820         (s390_invalid_arg_for_unprototyped_fn): Likewise.
40821         (s390_valid_target_attribute_inner_p): Likewise.
40822         * config/s390/s390.md ("tabort"): Likewise.
40824 2017-01-18  Toma Tabacu  <toma.tabacu@imgtec.com>
40826         * config/mips/mips.h (ISA_HAS_DIV3): Remove unused macro.
40827         (ISA_AVOID_DIV_HILO): New macro.
40828         (ISA_HAS_DIV): Use new ISA_AVOID_DIV_HILO macro.
40829         (ISA_HAS_DDIV): Likewise.
40831 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40833         * doc/invoke.texi (fabi-version): Correct number of occurrences.
40835 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40837         * doc/invoke.texi (fabi-version): Spelling fix.
40839 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40841         PR c++/70182
40842         * doc/invoke.texi (fabi-version): Mention mangling fix for
40843         operator names.
40845 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40847         PR c++/77489
40848         * doc/invoke.texi (fabi-version): Document discriminator mangling.
40850 2017-01-17  Segher Boessenkool  <segher@kernel.crashing.org>
40852         PR target/78875
40853         * config/rs6000/rs6000-opts.h (stack_protector_guard): New enum.
40854         * config/rs6000/rs6000.c (rs6000_option_override_internal): Handle
40855         the new options.
40856         * config/rs6000/rs6000.md (stack_protect_set): Handle the new more
40857         flexible settings.
40858         (stack_protect_test): Ditto.
40859         * config/rs6000/rs6000.opt (mstack-protector-guard=,
40860         mstack-protector-guard-reg=, mstack-protector-guard-offset=): New
40861         options.
40862         * doc/invoke.texi (Option Summary) [RS/6000 and PowerPC Options]:
40863         Add -mstack-protector-guard=, -mstack-protector-guard-reg=, and
40864         -mstack-protector-guard-offset=.
40865         (RS/6000 and PowerPC Options): Ditto.
40867 2017-01-17  Uros Bizjak  <ubizjak@gmail.com>
40869         * config/i386/i386.h (MASK_CLASS_P): New define.
40870         * config/i386/i386.c (inline_secondary_memory_needed): Ensure that
40871         there are no registers from different register sets also when
40872         mask registers are used.  Update function comment.
40873         * config/i386/i386.md (*movsi_internal): Split (*k/*krm) alternative
40874         to (*k/*r) and (*k/*km) alternatives.
40876 2017-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
40878         * config/aarch64/aarch64.md (eh_return): Remove pattern and splitter.
40879         * config/aarch64/aarch64.h (AARCH64_EH_STACKADJ_REGNUM): Remove.
40880         (EH_RETURN_HANDLER_RTX): New define.
40881         * config/aarch64/aarch64.c (aarch64_frame_pointer_required):
40882         Force frame pointer in EH return functions.
40883         (aarch64_expand_epilogue): Add barrier for eh_return.
40884         (aarch64_final_eh_return_addr): Remove.
40885         (aarch64_eh_return_handler_rtx): New function.
40886         * config/aarch64/aarch64-protos.h (aarch64_final_eh_return_addr):
40887         Remove.
40888         (aarch64_eh_return_handler_rtx): New prototype.
40890 2017-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40892         * config/rs6000/altivec.h (vec_rlmi): New #define.
40893         (vec_vrlnm): Likewise.
40894         (vec_rlnm): Likewise.
40895         * config/rs6000/altivec.md (UNSPEC_VRLMI): New UNSPEC enum value.
40896         (UNSPEC_VRLNM): Likewise.
40897         (VIlong): New mode iterator.
40898         (altivec_vrl<VI_char>mi): New define_insn.
40899         (altivec_vrl<VI_char>nm): Likewise.
40900         * config/rs6000/rs6000-builtin.def (VRLWNM): New monomorphic
40901         function entry.
40902         (VRLDNM): Likewise.
40903         (RLNM): New polymorphic function entry.
40904         (VRLWMI): New monomorphic function entry.
40905         (VRLDMI): Likewise.
40906         (RLMI): New polymorphic function entry.
40907         * config/rs6000/r6000-c.c (altivec_overloaded_builtin_table): Add
40908         new entries for P9V_BUILTIN_VEC_RLMI and P9V_BUILTIN_VEC_RLNM.
40909         * doc/extend.texi: Add description of vec_rlmi, vec_rlnm, and
40910         vec_vrlnm.
40912 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
40914         PR debug/78839
40915         * dwarf2out.c (field_byte_offset): Restore the
40916         PCC_BITFIELD_TYPE_MATTERS behavior for INTEGER_CST DECL_FIELD_OFFSET
40917         and DECL_FIELD_BIT_OFFSET.  Use fold_build2 instead of build2 + fold.
40918         (analyze_variants_discr, gen_variant_part): Use fold_build2 instead
40919         of build2 + fold.
40921 2017-01-17  Eric Botcazou  <ebotcazou@adacore.com>
40923         PR ada/67205
40924         * config/aarch64/aarch64.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): Define
40926 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
40928         PR debug/71669
40929         * dwarf2out.c (add_data_member_location_attribute): For constant
40930         offset bitfield emit for -gdwarf-5 DW_AT_data_bit_offset attribute
40931         instead of DW_AT_data_member_location, DW_AT_bit_offset and
40932         DW_AT_byte_size attributes.
40934 2017-01-17  Eric Botcazou  <ebotcazou@adacore.com>
40936         * config/rs6000/rs6000.c (rs6000_emit_move): Also use a TOC reference
40937         after forcing to constant memory when the code model is medium.
40939 2017-01-17  Julia Koval  <julia.koval@intel.com>
40941         PR target/76731
40942         * config/i386/avx512fintrin.h
40943         (_mm512_i32gather_ps): Change __addr type to void const*.
40944         (_mm512_mask_i32gather_ps): Ditto.
40945         (_mm512_i32gather_pd): Ditto.
40946         (_mm512_mask_i32gather_pd): Ditto.
40947         (_mm512_i64gather_ps): Ditto.
40948         (_mm512_mask_i64gather_ps): Ditto.
40949         (_mm512_i64gather_pd): Ditto.
40950         (_mm512_mask_i64gather_pd): Ditto.
40951         (_mm512_i32gather_epi32): Ditto.
40952         (_mm512_mask_i32gather_epi32): Ditto.
40953         (_mm512_i32gather_epi64): Ditto.
40954         (_mm512_mask_i32gather_epi64): Ditto.
40955         (_mm512_i64gather_epi32): Ditto.
40956         (_mm512_mask_i64gather_epi32): Ditto.
40957         (_mm512_i64gather_epi64): Ditto.
40958         (_mm512_mask_i64gather_epi64): Ditto.
40959         (_mm512_i32scatter_ps): Change __addr type to void*.
40960         (_mm512_mask_i32scatter_ps): Ditto.
40961         (_mm512_i32scatter_pd): Ditto.
40962         (_mm512_mask_i32scatter_pd): Ditto.
40963         (_mm512_i64scatter_ps): Ditto.
40964         (_mm512_mask_i64scatter_ps): Ditto.
40965         (_mm512_i64scatter_pd): Ditto.
40966         (_mm512_mask_i64scatter_pd): Ditto.
40967         (_mm512_i32scatter_epi32): Ditto.
40968         (_mm512_mask_i32scatter_epi32): Ditto.
40969         (_mm512_i32scatter_epi64): Ditto.
40970         (_mm512_mask_i32scatter_epi64): Ditto.
40971         (_mm512_i64scatter_epi32): Ditto.
40972         (_mm512_mask_i64scatter_epi32): Ditto.
40973         (_mm512_i64scatter_epi64): Ditto.
40974         (_mm512_mask_i64scatter_epi64): Ditto.
40975         * config/i386/avx512pfintrin.h
40976         (_mm512_mask_prefetch_i32gather_pd): Change __addr type to void const*.
40977         (_mm512_mask_prefetch_i32gather_ps): Ditto.
40978         (_mm512_mask_prefetch_i64gather_pd): Ditto.
40979         (_mm512_mask_prefetch_i64gather_ps): Ditto.
40980         (_mm512_prefetch_i32scatter_pd): Change __addr type to void*.
40981         (_mm512_prefetch_i32scatter_ps): Ditto.
40982         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
40983         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
40984         (_mm512_prefetch_i64scatter_pd): Ditto.
40985         (_mm512_prefetch_i64scatter_ps): Ditto.
40986         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
40987         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
40988         * config/i386/avx512vlintrin.h
40989         (_mm256_mmask_i32gather_ps): Change __addr type to void const*.
40990         (_mm_mmask_i32gather_ps): Ditto.
40991         (_mm256_mmask_i32gather_pd): Ditto.
40992         (_mm_mmask_i32gather_pd): Ditto.
40993         (_mm256_mmask_i64gather_ps): Ditto.
40994         (_mm_mmask_i64gather_ps): Ditto.
40995         (_mm256_mmask_i64gather_pd): Ditto.
40996         (_mm_mmask_i64gather_pd): Ditto.
40997         (_mm256_mmask_i32gather_epi32): Ditto.
40998         (_mm_mmask_i32gather_epi32): Ditto.
40999         (_mm256_mmask_i32gather_epi64): Ditto.
41000         (_mm_mmask_i32gather_epi64): Ditto.
41001         (_mm256_mmask_i64gather_epi32): Ditto.
41002         (_mm_mmask_i64gather_epi32): Ditto.
41003         (_mm256_mmask_i64gather_epi64): Ditto.
41004         (_mm_mmask_i64gather_epi64): Ditto.
41005         (_mm256_i32scatter_ps): Change __addr type to void*.
41006         (_mm256_mask_i32scatter_ps): Ditto.
41007         (_mm_i32scatter_ps): Ditto.
41008         (_mm_mask_i32scatter_ps): Ditto.
41009         (_mm256_i32scatter_pd): Ditto.
41010         (_mm256_mask_i32scatter_pd): Ditto.
41011         (_mm_i32scatter_pd): Ditto.
41012         (_mm_mask_i32scatter_pd): Ditto.
41013         (_mm256_i64scatter_ps): Ditto.
41014         (_mm256_mask_i64scatter_ps): Ditto.
41015         (_mm_i64scatter_ps): Ditto.
41016         (_mm_mask_i64scatter_ps): Ditto.
41017         (_mm256_i64scatter_pd): Ditto.
41018         (_mm256_mask_i64scatter_pd): Ditto.
41019         (_mm_i64scatter_pd): Ditto.
41020         (_mm_mask_i64scatter_pd): Ditto.
41021         (_mm256_i32scatter_epi32): Ditto.
41022         (_mm256_mask_i32scatter_epi32): Ditto.
41023         (_mm_i32scatter_epi32): Ditto.
41024         (_mm_mask_i32scatter_epi32): Ditto.
41025         (_mm256_i32scatter_epi64): Ditto.
41026         (_mm256_mask_i32scatter_epi64): Ditto.
41027         (_mm_i32scatter_epi64): Ditto.
41028         (_mm_mask_i32scatter_epi64): Ditto.
41029         (_mm256_i64scatter_epi32): Ditto.
41030         (_mm256_mask_i64scatter_epi32): Ditto.
41031         (_mm_i64scatter_epi32): Ditto.
41032         (_mm_mask_i64scatter_epi32): Ditto.
41033         (_mm256_i64scatter_epi64): Ditto.
41034         (_mm256_mask_i64scatter_epi64): Ditto.
41035         (_mm_i64scatter_epi64): Ditto.
41036         (_mm_mask_i64scatter_epi64): Ditto.
41037         * config/i386/i386-builtin-types.def (V16SF_V16SF_PCFLOAT_V16SI_HI_INT)
41038         (V8DF_V8DF_PCDOUBLE_V8SI_QI_INT, V8SF_V8SF_PCFLOAT_V8DI_QI_INT)
41039         (V8DF_V8DF_PCDOUBLE_V8DI_QI_INT, V16SI_V16SI_PCINT_V16SI_HI_INT)
41040         (V8DI_V8DI_PCINT64_V8SI_QI_INT, V8SI_V8SI_PCINT_V8DI_QI_INT)
41041         (V8DI_V8DI_PCINT64_V8DI_QI_INT, V2DF_V2DF_PCDOUBLE_V4SI_QI_INT)
41042         (V4DF_V4DF_PCDOUBLE_V4SI_QI_INT, V2DF_V2DF_PCDOUBLE_V2DI_QI_INT)
41043         (V4DF_V4DF_PCDOUBLE_V4DI_QI_INT, V4SF_V4SF_PCFLOAT_V4SI_QI_INT)
41044         (V8SF_V8SF_PCFLOAT_V8SI_QI_INT, V4SF_V4SF_PCFLOAT_V2DI_QI_INT)
41045         (V4SF_V4SF_PCFLOAT_V4DI_QI_INT, V2DI_V2DI_PCINT64_V4SI_QI_INT)
41046         (V4DI_V4DI_PCINT64_V4SI_QI_INT, V2DI_V2DI_PCINT64_V2DI_QI_INT)
41047         (V4DI_V4DI_PCINT64_V4DI_QI_INT, V4SI_V4SI_PCINT_V4SI_QI_INT)
41048         (V8SI_V8SI_PCINT_V8SI_QI_INT, V4SI_V4SI_PCINT_V2DI_QI_INT)
41049         (V4SI_V4SI_PCINT_V4DI_QI_INT, VOID_PFLOAT_HI_V16SI_V16SF_INT)
41050         (VOID_PFLOAT_QI_V8SI_V8SF_INT, VOID_PFLOAT_QI_V4SI_V4SF_INT)
41051         (VOID_PDOUBLE_QI_V8SI_V8DF_INT, VOID_PDOUBLE_QI_V4SI_V4DF_INT)
41052         (VOID_PDOUBLE_QI_V4SI_V2DF_INT, VOID_PFLOAT_QI_V8DI_V8SF_INT)
41053         (VOID_PFLOAT_QI_V4DI_V4SF_INT, VOID_PFLOAT_QI_V2DI_V4SF_INT)
41054         (VOID_PDOUBLE_QI_V8DI_V8DF_INT, VOID_PDOUBLE_QI_V4DI_V4DF_INT)
41055         (VOID_PDOUBLE_QI_V2DI_V2DF_INT, VOID_PINT_HI_V16SI_V16SI_INT)
41056         (VOID_PINT_QI_V8SI_V8SI_INT, VOID_PINT_QI_V4SI_V4SI_INT)
41057         (VOID_PLONGLONG_QI_V8SI_V8DI_INT, VOID_PLONGLONG_QI_V4SI_V4DI_INT)
41058         (VOID_PLONGLONG_QI_V4SI_V2DI_INT, VOID_PINT_QI_V8DI_V8SI_INT)
41059         (VOID_PINT_QI_V4DI_V4SI_INT, VOID_PINT_QI_V2DI_V4SI_INT)
41060         (VOID_PLONGLONG_QI_V8DI_V8DI_INT, VOID_QI_V8SI_PCINT64_INT_INT)
41061         (VOID_PLONGLONG_QI_V4DI_V4DI_INT, VOID_PLONGLONG_QI_V2DI_V2DI_INT)
41062         (VOID_HI_V16SI_PCINT_INT_INT, VOID_QI_V8DI_PCINT64_INT_INT)
41063         (VOID_QI_V8DI_PCINT_INT_INT): Remove.
41064         (V16SF_V16SF_PCVOID_V16SI_HI_INT,  V8DF_V8DF_PCVOID_V8SI_QI_INT)
41065         (V8SF_V8SF_PCVOID_V8DI_QI_INT, V8DF_V8DF_PCVOID_V8DI_QI_INT)
41066         (V16SI_V16SI_PCVOID_V16SI_HI_INT, V8DI_V8DI_PCVOID_V8SI_QI_INT)
41067         (V8SI_V8SI_PCVOID_V8DI_QI_INT, V8DI_V8DI_PCVOID_V8DI_QI_INT)
41068         (VOID_PVOID_HI_V16SI_V16SF_INT, VOID_PVOID_QI_V8SI_V8DF_INT)
41069         (VOID_PVOID_QI_V8DI_V8SF_INT, VOID_PVOID_QI_V8DI_V8DF_INT)
41070         (VOID_PVOID_HI_V16SI_V16SI_INT, VOID_PVOID_QI_V8SI_V8DI_INT)
41071         (VOID_PVOID_QI_V8DI_V8SI_INT, VOID_PVOID_QI_V8DI_V8DI_INT)
41072         (V2DF_V2DF_PCVOID_V4SI_QI_INT, V4DF_V4DF_PCVOID_V4SI_QI_INT)
41073         (V2DF_V2DF_PCVOID_V2DI_QI_INT, V4DF_V4DF_PCVOID_V4DI_QI_INT
41074         (V4SF_V4SF_PCVOID_V4SI_QI_INT, V8SF_V8SF_PCVOID_V8SI_QI_INT)
41075         (V4SF_V4SF_PCVOID_V2DI_QI_INT, V4SF_V4SF_PCVOID_V4DI_QI_INT)
41076         (V2DI_V2DI_PCVOID_V4SI_QI_INT, V4DI_V4DI_PCVOID_V4SI_QI_INT)
41077         (V2DI_V2DI_PCVOID_V2DI_QI_INT, V4DI_V4DI_PCVOID_V4DI_QI_INT)
41078         (V4SI_V4SI_PCVOID_V4SI_QI_INT, V8SI_V8SI_PCVOID_V8SI_QI_INT)
41079         (V4SI_V4SI_PCVOID_V2DI_QI_INT, V4SI_V4SI_PCVOID_V4DI_QI_INT)
41080         (VOID_PVOID_QI_V8SI_V8SF_INT, VOID_PVOID_QI_V4SI_V4SF_INT)
41081         (VOID_PVOID_QI_V4SI_V4DF_INT, VOID_PVOID_QI_V4SI_V2DF_INT)
41082         (VOID_PVOID_QI_V4DI_V4SF_INT, VOID_PVOID_QI_V2DI_V4SF_INT)
41083         (VOID_PVOID_QI_V4DI_V4DF_INT, VOID_PVOID_QI_V2DI_V2DF_INT)
41084         (VOID_PVOID_QI_V8SI_V8SI_INT, VOID_PVOID_QI_V4SI_V4SI_INT)
41085         (VOID_PVOID_QI_V4SI_V4DI_INT, VOID_PVOID_QI_V4SI_V2DI_INT)
41086         (VOID_PVOID_QI_V4DI_V4SI_INT, VOID_PVOID_QI_V2DI_V4SI_INT)
41087         (VOID_PVOID_QI_V4DI_V4DI_INT, VOID_PVOID_QI_V2DI_V2DI_INT)
41088         (VOID_QI_V8SI_PCVOID_INT_INT, VOID_HI_V16SI_PCVOID_INT_INT)
41089         (VOID_QI_V8DI_PCVOID_INT_INT): Add.
41090         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Adjust builtin
41091         definitions accordingly.
41093 2017-01-17  Kito Cheng  <kito.cheng@gmail.com>
41094             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
41096         PR target/79079
41097         * internal-fn.c (expand_mul_overflow): Use convert_modes instead of
41098         gen_lowpart.
41100 2017-01-17  Vladimir Makarov  <vmakarov@redhat.com>
41102         PR target/79058
41103         * ira-conflicts.c (ira_build_conflicts): Update total conflict
41104         hard regs for inner regno.
41106 2017-01-17  Martin Liska  <mliska@suse.cz>
41108         PR ipa/71207
41109         * ipa-polymorphic-call.c (contains_type_p): Fix wrong
41110         assumption and add comment.
41112 2017-01-17  Nathan Sidwell  <nathan@acm.org>
41114         * ipa-visibility.c (localize_node): New function, broken out of ...
41115         (function_and_variable_visibility): ... here. Call it.
41117 2017-01-17  Jan Hubicka  <hubicka@ucw.cz>
41119         PR middle-end/77445
41120         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges):
41121         correctly set frequency of oudgoing edge.
41122         (duplicate_thread_path): Fix profile updating.
41124 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
41126         PR other/79046
41127         * configure.ac: Add GCC_BASE_VER.
41128         * Makefile.in (version): Use @get_gcc_base_ver@ instead of cat to get
41129         version from BASE-VER file.
41130         (CFLAGS-gcc.o): Add -DBASEVER=$(BASEVER_s).
41131         (gcc.o): Depend on $(BASEVER).
41132         * common.opt (dumpfullversion): New option.
41133         * gcc.c (driver_handle_option): Handle OPT_dumpfullversion.
41134         * doc/invoke.texi: Document -dumpfullversion.
41135         * doc/install.texi: Document --with-gcc-major-version-only.
41136         * configure: Regenerated.
41138 2017-01-17  Richard Biener  <rguenther@suse.de>
41140         PR tree-optimization/71433
41141         * tree-vrp.c (register_new_assert_for): Merge same asserts
41142         on all incoming edges.
41143         (process_assert_insertions_for): Handle insertions at the
41144         beginning of BBs.
41146 2017-01-17  Gerald Pfeifer  <gerald@pfeifer.com>
41148         * config/i386/cygwin.h (LIBGCJ_SONAME): Remove.
41149         * config/i386/mingw32.h (LIBGCJ_SONAME): Remove.
41151 2017-01-17  Kaz Kojima  <kkojima@gcc.gnu.org>
41153         PR target/78633
41154         * config/sh/sh.md (cmpeqsi_t+1): Call copy_rtx to avoid invalid
41155         RTL sharing.
41157 2017-01-17  Alan Modra  <amodra@gmail.com>
41159         PR target/79066
41160         * config/rs6000/rs6000.md (elf_high, elf_low): Disable when pic.
41161         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack): Don't allow
41162         symbolic stack limit when pic.
41164 2017-01-16  Martin Sebor  <msebor@redhat.com>
41166         PR tree-optimization/78608
41167         * gimple-ssa-sprintf.c (tree_digits): Avoid negating TYPE_MIN.
41169 2017-01-16  Jeff Law  <law@redhat.com>
41171         Revert:
41172         2016-12-02  Tadek Kijkowski  <tkijkowski@gmail.com>
41173         * Makefile.in (PREPROCESSOR_DEFINES): Add a level of indirection
41174         for several include directories that may be relative to sysroot.
41175         * config/i386/x-mingw32 (gplus_includedir): Define.
41176         (gplus_tool_includedir, gplus_backward_include_dir): Likewise.
41177         (native_system_includedir): Likewise.
41178         * config/i386/mingw32.h (STANDARD_STARTFILE_PREFIX_1): Do not
41179         override if TARGET_SYSTEM_ROOT is defined.
41180         (NATIVE_SYSTEM_HEADER_DIR): Likewise.
41182         PR tree-optimization/79090
41183         PR tree-optimization/33562
41184         PR tree-optimization/61912
41185         PR tree-optimization/77485
41186         * tree-ssa-dse.c (compute_trims): Accept STMT argument.  Dump STMT
41187         and computed trims into the dump file.
41189 2017-01-17  Uros Bizjak  <ubizjak@gmail.com>
41191         * config/i386/i386.h (LIMIT_RELOAD_CLASS): Remove.
41193 2017-01-16  Jakub Jelinek  <jakub@redhat.com>
41195         PR c/79089
41196         * gimplify.c (gimplify_init_constructor): If want_value and
41197         object == lhs, unshare lhs to avoid invalid tree sharing.  Formatting
41198         fix.
41200         PR target/79080
41201         * loop-doloop.c (doloop_modify): Call unshare_all_rtl_in_chain on
41202         sequence.  Formatting fixes.
41203         (doloop_optimize): Formatting fixes.
41205         PR driver/49726
41206         * gcc.c (debug_level_greater_than_spec_func): New function.
41207         (static_spec_functions): Add debug-level-gt spec function.
41208         (ASM_DEBUG_SPEC, cpp_options): Use %:debug-level-gt(0) instead of
41209         !g0.
41210         * config/darwin.h (DSYMUTIL_SPEC, ASM_DEBUG_SPEC): Likewise.
41211         * config/darwin9.h (DSYMUTIL_SPEC, ASM_DEBUG_SPEC): Likewise.
41212         * common.opt (g, gcoff, gdwarf, gdwarf-, ggdb, gno-pubnames,
41213         gpubnames, ggnu-pubnames, gno-record-gcc-switches,
41214         grecord-gcc-switches, gno-strict-dwarf, gstrict-dwarf, gstabs,
41215         gstabs+, gtoggle, gvms, gxcoff, gxcoff+): Add Driver flag.
41217 2017-01-16  Uros Bizjak  <ubizjak@gmail.com>
41219         * config/i386/i386.h (HARD_REGNO_CALLER_SAVE_MODE): Apply HImode and
41220         QImode fixups to general and mask registers only.
41222 2017-01-16  Carl Love  <cel@us.ibm.com>
41224         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Add support
41225         for built-in functions
41226         vector signed char vec_nabs (vector signed char)
41227         vector signed short vec_nabs (vector signed short)
41228         vector signed int vec_nabs (vector signed int)
41229         vector signed long long vec_nabs (vector signed long long)
41230         vector float vec_nabs (vector float)
41231         vector double vec_nabs (vector double)
41232         * config/rs6000/rs6000-builtin.def: Add definitions for NABS functions
41233         and NABS overload.
41234         * config/rs6000/altivec.md: New define_expand nabs<mode>2 types
41235         * config/rs6000/altivec.h: New define for vec_nabs built-in function.
41236         * doc/extend.texi: Update the documentation file for the new built-in
41237         functions.
41239 2017-01-16  Martin Sebor  <msebor@redhat.com>
41241         * gimple-ssa-sprintf.c (format_directive): Correct a typo in a warning
41242         message.
41244 2017-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
41246         * config/rs6000/rs6000.c (rtx_is_swappable_p): Change
41247         UNSPEC_VSX__XXSPLTD to require special splat handling.
41249 2017-01-16  David Malcolm  <dmalcolm@redhat.com>
41251         PR bootstrap/78616
41252         * system.h: Poison strndup.
41254 2017-01-16  Alan Modra  <amodra@gmail.com>
41256         PR target/79098
41257         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Don't
41258         use a switch.
41260 2017-01-16  Georg-Johann Lay  <avr@gjlay.de>
41262         * config/avr/avr.h (BRANCH_COST) [reload_completed]: Increase by 4.
41264 2017-01-15  Uros Bizjak  <ubizjak@gmail.com>
41266         * config/i386/i386.c (ix86_legitimate_combined_insn): Do not
41267         call recog here.  Assert that INSN_CODE (insn) is non-negative.
41269 2017-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
41271         PR target/72749
41272         * cfgrtl.c (rtl_split_edge): Also patch jump insns that jump to the
41273         fallthrough.
41274         * haifa-sched.c (dump_insn_stream): Don't crash if there is a label
41275         in the currently scheduled RTL fragment.
41277 2017-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
41279         PR rtl-optimization/78751
41280         * ifcvt.c (find_cond_trap): If we generated a non-existing insn,
41281         give up.
41283 2017-01-14  Jeff Law  <law@redhat.com>
41285         PR tree-optimization/79090
41286         * tree-ssa-dse.c (valid_ao_ref_for_dse): Reject zero length and
41287         variable length stores.
41288         (compute_trims): Delete dead assignment to *trim_tail.
41289         (dse_dom_walker::dse_optimize_stmt): Optimize mem* calls with
41290         zero length.
41292 2017-01-14  Bernd Schmidt  <bschmidt@redhat.com>
41294         PR rtl-optimization/78626
41295         PR rtl-optimization/78727
41296         * cprop.c (one_cprop_pass): Collect unconditional traps in the middle
41297         of a block, and split such blocks after everything else is finished.
41299 2017-01-14  Alan Modra  <amodra@gmail.com>
41301         PR target/72749
41302         * combine.c (recog_for_combine_1): Set INSN_CODE before calling
41303         target legitimate_combined_insn.
41304         * config/rs6000/rs6000.c (TARGET_LEGITIMATE_COMBINED_INSN): Define.
41305         (rs6000_legitimate_combined_insn): New function.
41306         * config/rs6000/rs6000.md (UNSPEC_DOLOOP): Delete, and remove
41307         all uses.
41308         (ctr<mode>_internal3): Rename from *ctr<mode>_internal5.
41309         (ctr<mode>_internal4): Rename from *ctr<mode>_internal6.
41310         (ctr<mode>_internal1, ctr<mode>_internal2): Remove '*' from name.
41312 2017-01-14  Gerald Pfeifer  <gerald@pfeifer.com>
41314         * doc/frontends.texi (G++ and GCC): Remove references to Java.
41316 2017-01-13  Jeff Law  <law@redhat.com>
41318         PR tree-optimization/33562
41319         PR tree-optimization/61912
41320         PR tree-optimization/77485
41321         * tree-ssa-dse.c (delete_dead_call): Accept gsi rather than
41322         a statement.
41323         (delete_dead_assignment): Likewise.
41324         (dse_dom_walker::dse_optimize_stmt): Pass in the gsi rather than
41325         statement to delete_dead_call and delete_dead_assignment.
41327 2017-01-13  David Malcolm  <dmalcolm@redhat.com>
41329         PR c/78304
41330         * substring-locations.c (format_warning_va): Strengthen case 1 so
41331         that both endpoints of the substring must be within the format
41332         range for just the substring to be printed.
41334 2017-01-13  Uros Bizjak  <ubizjak@gmail.com>
41336         * config/i386/i386.opt (msgx): Use ix86_isa_flags2 variable.
41337         * config/i386/i386.c (ix86_target_string): Add missing options
41338         to isa_opts and reorder options by implied ISAs.  Rename isa_opts2 to
41339         isa2_opts, ix86_flag_opts to flag2_opts, ix86_target_other to
41340         flags_other and ix86_target_other to flags2_other.  Display unknown
41341         isa2 options.
41342         (ix86_valid_target_attribute_inner_p): Add missing options and
41343         reorder options by implied ISAs, as in ix86_target_string.
41345 2017-01-13  Richard Sandiford  <richard.sandiford@arm.com>
41347         * hash-table.h (hash_table::too_empty_p): New function.
41348         (hash_table::expand): Use it.
41349         (hash_table::traverse): Likewise.
41350         (hash_table::empty_slot): Use sizeof (value_type) instead of
41351         sizeof (PTR) to convert bytes to elements.  Shrink the table
41352         if the current size is excessive for the current number of
41353         elements.
41355 2017-01-13  Richard Sandiford  <richard.sandiford@arm.com>
41357         * ira-costs.c (record_reg_classes): Break from the inner loop
41358         early once alt_fail is known to be true.  Update outer loop
41359         handling accordingly.
41361 2017-01-13  Jeff Law  <law@redhat.com>
41363         * tree-ssa-dse.c (decrement_count): New function.
41364         (increment_start_addr, maybe_trim_memstar_call): Likewise.
41365         (dse_dom_walker::optimize_stmt): Call maybe_trim_memstar_call directly
41366         when we know the partially dead statement is a mem* function.
41368         PR tree-optimization/61912
41369         PR tree-optimization/77485
41370         * tree-ssa-dse.c: Include expr.h.
41371         (maybe_trim_constructor_store): New function.
41372         (maybe_trim_partially_dead_store): Call maybe_trim_constructor_store.
41374         PR tree-optimization/33562
41375         PR tree-optimization/61912
41376         PR tree-optimization/77485
41377         * doc/invoke.texi: Document new dse-max-object-size param.
41378         * params.def (PARM_DSE_MAX_OBJECT_SIZE): New PARAM.
41379         * tree-ssa-dse.c: Include params.h.
41380         (dse_store_status): New enum.
41381         (initialize_ao_ref_for_dse): New, partially extracted from
41382         dse_optimize_stmt.
41383         (valid_ao_ref_for_dse, normalize_ref): New.
41384         (setup_live_bytes_from_ref, compute_trims): Likewise.
41385         (clear_bytes_written_by, maybe_trim_complex_store): Likewise.
41386         (maybe_trim_partially_dead_store): Likewise.
41387         (maybe_trim_complex_store): Likewise.
41388         (dse_classify_store): Renamed from dse_possibly_dead_store_p.
41389         Track what bytes live from the original store.  Return tri-state
41390         for dead, partially dead or live.
41391         (dse_dom_walker): Add constructor, destructor and new private members.
41392         (delete_dead_call, delete_dead_assignment): New extracted from
41393         dse_optimize_stmt.
41394         (dse_optimize_stmt): Make a member of dse_dom_walker.
41395         Use initialize_ao_ref_for_dse.
41397         PR tree-optimization/33562
41398         PR tree-optimization/61912
41399         PR tree-optimization/77485
41400         * sbitmap.h (bitmap_count_bits): Prototype.
41401         (bitmap_clear_range, bitmap_set_range): Likewise.
41402         * sbitmap.c (bitmap_clear_range): New function.
41403         (bitmap_set_range, sbitmap_popcount, bitmap_count_bits): Likewise.
41405 2017-01-13  Martin Liska  <mliska@suse.cz>
41407         PR ipa/79043
41408         * function.c (set_cfun): Add new argument force.
41409         * function.h (set_cfun): Likewise.
41410         * ipa-inline-transform.c (inline_call): Use the function when
41411         strict alising from is dropped for function we inline to.
41413 2017-01-13  Richard Biener  <rguenther@suse.de>
41415         * tree-pretty-print.c (dump_generic_node): Fix inverted condition
41416         for dumping GIMPLE INTEGER_CSTs.
41418 2017-01-13  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
41420         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define __STDC_VERSION__
41421         to 201112L since C++17.
41423 2017-01-13  Maxim Ostapenko  <m.ostapenko@samsung.com>
41425         PR sanitizer/78887
41426         * asan.c (asan_needs_odr_indicator_p): Don't emit ODR indicators
41427         if -fsanitize=kernel-address is present.
41429 2017-01-13  Richard Biener  <rguenther@suse.de>
41431         * tree-pretty-print.c (dump_generic_node): Dump INTEGER_CSTs
41432         as _Literal ( type ) number in case usual suffixes do not
41433         preserve all information.
41435 2017-01-13  Richard Biener  <rguenther@suse.de>
41437         PR tree-optimization/77283
41438         * gimple-ssa-split-paths.c: Include gimple-ssa.h, tree-phinodes.h
41439         and ssa-iterators.h.
41440         (is_feasible_trace): Implement a cost model based on joiner
41441         PHI node uses.
41443 2017-01-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
41445         PR target/79004
41446         * config/rs6000/rs6000.md (FP_ISA3): Do not optimize converting
41447         char or short to __float128/_Float128 directly.
41449 2017-01-12  Martin Sebor  <msebor@redhat.com>
41451         to -Wformat-overflow.
41452         * gimple-ssa-sprintf.c (pass_sprintf_length::gate): Adjust.
41453         (min_bytes_remaining): Same.
41454         (get_string_length): Same.
41455         (format_string): Same.
41456         (format_directive): Same.
41457         (add_bytes): Same.
41458         (pass_sprintf_length::handle_gimple_call): Same.
41460 2017-01-12  Jakub Jelinek  <jakub@redhat.com>
41462         * gimple-ssa-sprintf.c (try_substitute_return_value): Remove
41463         info.nowrite calls with no lhs that can't throw.  Return bool
41464         whether gsi_remove has been called or not.
41465         (pass_sprintf_length::handle_gimple_call): Return bool whether
41466         try_substitute_return_value called gsi_remove.  Formatting fix.
41467         (pass_sprintf_length::execute): Don't use gsi_remove if
41468         handle_gimple_call returned true.
41470         PR bootstrap/79069
41471         * cfgrtl.c (rtl_tidy_fallthru_edge): For any_uncondjump_p that can't
41472         be removed due to side-effects, don't remove following barrier nor
41473         turn the successor edge into fallthru edge.
41475 2017-01-12  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
41477         PR target/79044
41478         * config/rs6000/rs6000.c (insn_is_swappable_p): Mark
41479         element-reversing loads and stores as not swappable.
41481 2017-01-12  Nathan Sidwell  <nathan@acm.org>
41482             Nicolai Stange  <nicstange@gmail.com>
41484         * combine.c (try_combine): Don't ignore result of overlap checking
41485         loop.  Combine overlap & asm check into single loop.
41487 2017-01-12  Richard Biener  <rguenther@suse.de>
41489         * tree-pretty-print.c (dump_generic_node): Provide -gimple
41490         variant for MEM_REF.  Sanitize INTEGER_CST for -gimple.
41492 2017-01-12  Richard Biener  <rguenther@suse.de>
41494         * tree.c (initialize_tree_contains_struct): Make TS_OPTIMIZATION
41495         and TS_TARGET_OPTION directly derive from TS_BASE.
41496         * tree-core.h (tree_optimization_option): Derive from tree_base.
41497         (tree_target_option): Likewise.
41499 2017-01-11  Uros Bizjak  <ubizjak@gmail.com>
41501         * config/i386/i386.c (memory_address_length): Increase len
41502         only when rip_relative_addr_p returns false.
41504 2017-01-11  Julia Koval  <julia.koval@intel.com>
41506         * common/config/i386/i386-common.c (OPTION_MASK_ISA_SGX_UNSET): New.
41507         (OPTION_MASK_ISA_SGX_SET): New.
41508         (ix86_handle_option): Handle OPT_msgx.
41509         * config.gcc: Added sgxintrin.h.
41510         * config/i386/driver-i386.c (host_detect_local_cpu): Detect sgx.
41511         * config/i386/i386-c.c (ix86_target_macros_internal): Define __SGX__.
41512         * config/i386/i386.c (ix86_target_string): Add -msgx.
41513         (PTA_SGX): New.
41514         (ix86_option_override_internal): Handle new options.
41515         (ix86_valid_target_attribute_inner_p): Add sgx.
41516         * config/i386/i386.h (TARGET_SGX, TARGET_SGX_P): New.
41517         * config/i386/i386.opt: Add msgx.
41518         * config/i386/sgxintrin.h: New file.
41519         * config/i386/x86intrin.h: Add sgxintrin.h.
41521 2017-01-11  Jakub Jelinek  <jakub@redhat.com>
41523         PR c++/71537
41524         * fold-const.c (maybe_nonzero_address): Return 1 for function
41525         local objects.
41526         (tree_single_nonzero_warnv_p): Don't handle function local objects
41527         here.
41529         PR c++/72813
41530         * gcc.c (default_compilers): Don't add -o %g.s for -S -save-temps
41531         of c-header.
41533 2017-01-11  David Malcolm  <dmalcolm@redhat.com>
41535         PR driver/78877
41536         * opts.c: Include "spellcheck.h"
41537         (struct string_fragment): New struct.
41538         (struct edit_distance_traits<const string_fragment &>): New
41539         struct.
41540         (get_closest_sanitizer_option): New function.
41541         (parse_sanitizer_options): Offer suggestions for unrecognized arguments.
41543 2017-01-11  Jakub Jelinek  <jakub@redhat.com>
41545         * dwarf2out.c (DWARF_COMPILE_UNIT_HEADER_SIZE): For DWARF5 decrease
41546         by 12.
41547         (DWARF_COMDAT_TYPE_UNIT_HEADER_SIZE): Always
41548         DWARF_COMPILE_UNIT_HEADER_SIZE plus 12.
41549         (DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE): Define.
41550         (calc_base_type_die_sizes): Use DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE
41551         for initial die_offset if dwarf_split_debug_info.
41552         (output_comp_unit): Use DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE for
41553         initial next_die_offset if dwo_id is non-NULL.  Don't emit padding
41554         fields.
41555         (output_skeleton_debug_sections): Formatting fix.  Use
41556         DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE instead of
41557         DWARF_COMPILE_UNIT_HEADER_SIZE.  Don't emit padding.
41559 2017-01-11  Wilco Dijkstra  <wdijkstr@arm.com>
41561         * config/arm/cortex-a53.md: Add bypasses for
41562         cortex_a53_r2f_cvt.
41563         (cortex_a53_r2f): Only use for transfers.
41564         (cortex_a53_f2r): Likewise.
41565         (cortex_a53_r2f_cvt): Add reservation for conversions.
41566         (cortex_a53_f2r_cvt): Likewise.
41568 2017-01-11  Tamar Christina  <tamar.christina@arm.com>
41570         * config/arm/arm_neon.h: Add __artificial__ and gnu_inline
41571         to all inlined functions, change static to extern.
41573 2017-01-11  Christophe Lyon  <christophe.lyon@linaro.org>
41575         PR target/78253
41576         * config/arm/arm.c (legitimize_pic_address): Handle reference to
41577         weak symbol.
41578         (arm_assemble_integer): Likewise.
41580 2017-01-11  Richard Earnshaw  <rearnsha@arm.com>
41582         * config.gcc: Use new awk script to check CPU, FPU and architecture
41583         parameters for --with-... options.
41584         * config/arm/parsecpu.awk: New file
41585         * config/arm/arm-cpus.in: New file.
41586         * config/arm/arm-opts.h: Include arm-cpu.h instead of processing .def
41587         files.
41588         * config/arm/arm.c: Include arm-cpu-data.h instead of processing .def
41589         files.
41590         * config/arm/t-arm: Update dependency rules.
41591         * common/config/arm/arm-common.c: Include arm-cpu-cdata.h instead
41592         of processing .def files.
41593         * config/arm/genopt.sh: Deleted.
41594         * config/arm/gentune.sh: Deleted.
41595         * config/arm/arm-cores.def: Deleted.
41596         * config/arm/arm-arches.def: Deleted.
41597         * config/arm/arm-fpus.def: Deleted.
41598         * config/arm/arm-tune.md: Regenerated.
41599         * config/arm/arm-tables.opt: Regenerated.
41600         * config/arm/arm-cpu.h: New generated file.
41601         * config/arm/arm-cpu-data.h: New generated file.
41602         * config/arm/arm-cpu-cdata.h: New generated file.
41604 2017-01-11  Maxim Ostapenko  <m.ostapenko@samsung.com>
41606         PR lto/79042
41607         * lto-cgraph.c (lto_output_varpool_node): Pack dynamically_initialized
41608         bit.
41609         (input_varpool_node): Unpack dynamically_initialized bit.
41611 2017-01-11  Eric Botcazou  <ebotcazou@adacore.com>
41613         PR rtl-optimization/79032
41614         * lra-constraints.c (simplify_operand_subreg): In the MEM case, test
41615         the alignment of the adjusted memory reference against that of MODE,
41616         instead of the alignment of the original memory reference.
41618 2017-01-11  Martin Jambor  <mjambor@suse.cz>
41620         * hsa.c (hsa_callable_function_p): Revert addition of DECL_ARTIFICIAL
41621         test.
41622         * ipa-hsa.c (process_hsa_functions): Only duplicate non-artificial
41623         decorated functions.
41625 2017-01-11  Richard Biener  <rguenther@suse.de>
41627         * tree-vrp.c (evrp_dom_walker::before_dom_children): Also
41628         set range/nonnull info for PHI results.  Do not set it on
41629         stmts marked for removal.
41631 2017-01-10  Eric Botcazou  <ebotcazou@adacore.com>
41633         * expr.c (store_field): In the bitfield case, fetch the return value
41634         from the registers before applying a single big-endian adjustment.
41635         Always do a final load for a BLKmode value not larger than a word.
41637 2017-01-10  David Malcolm  <dmalcolm@redhat.com>
41639         PR c++/77949
41640         * input.c (selftest::test_accessing_ordinary_linemaps): Verify
41641         that we correctly handle column numbers greater than
41642         LINE_MAP_MAX_COLUMN_NUMBER.
41644 2017-01-10  Martin Sebor  <msebor@redhat.com>
41646         PR middle-end/78245
41647         * gimple-ssa-sprintf.c (get_destination_size): Call
41648         {init,fini}object_sizes.
41649         * tree-object-size.c (addr_object_size): Adjust.
41650         (pass_through_call): Adjust.
41651         (pass_object_sizes::execute): Adjust.
41652         * tree-object-size.h (fini_object_sizes): Declare.
41654 2017-01-10  Martin Sebor  <msebor@redhat.com>
41656         PR tree-optimization/78775
41657         * builtins.c (get_size_range): Move...
41658         * calls.c: ...to here.
41659         (alloc_max_size): Accept zero argument.
41660         (operand_signed_p): Remove.
41661         (maybe_warn_alloc_args_overflow): Call get_size_range.
41662         * calls.h (get_size_range): Declare.
41664 2017-01-10  Joe Seymour  <joe.s@somniumtech.com>
41666         * config/msp430/driver-msp430.c (msp430_mcu_data): Sync with data
41667         from TI's devices.csv file as of September 2016.
41668         * config/msp430/msp430.c (msp430_mcu_data): Likewise.
41670 2017-01-10  Sandra Loosemore  <sandra@codesourcery.com>
41672         * doc/extend.texi: Tweak formatting to fix overfull hbox warnings.
41673         * doc/invoke.texi: Likewise.
41674         * doc/md.texi: Likewise.
41675         * doc/objc.texi: Likewise.
41677 2017-01-10  Joshua Conner  <joshconner@google.com>
41679         * config/arm/fuchsia-elf.h: New file.
41680         * config/fuchsia.h: New file.
41681         * config.gcc (*-*-fuchsia*): Set native_system_header_dir.
41682         (aarch64*-*-fuchsia*, arm*-*-fuchsia*, x86_64-*-fuchsia*): Add to
41683         targets.
41684         * config.host: (aarch64*-*-fuchsia*, arm*-*-fuchsia*): Add to hosts.
41686 2016-01-10  Richard Biener  <rguenther@suse.de>
41688         PR tree-optimization/79034
41689         * tree-call-cdce.c (shrink_wrap_one_built_in_call_with_conds):
41690         Propagate out degenerate PHIs in the joiner.
41692 2017-01-10  Martin Liska  <mliska@suse.cz>
41694         * ipa-icf.c (sort_sem_items_by_decl_uid): New function.
41695         (sort_congruence_classes_by_decl_uid): Likewise.
41696         (sort_congruence_class_groups_by_decl_uid): Likewise.
41697         (sem_item_optimizer::merge_classes): Sort class, groups in these
41698         classes and members in the groups by DECL_UID of declarations.
41699         This would make merge operations stable.
41701 2017-01-10  Martin Liska  <mliska@suse.cz>
41703         * ipa-icf.c (sem_item_optimizer::sem_item_optimizer): Remove
41704         usage of m_classes_vec.
41705         (sem_item_optimizer::~sem_item_optimizer):  Likewise.
41706         (sem_item_optimizer::get_group_by_hash): Likewise.
41707         (sem_item_optimizer::subdivide_classes_by_equality): Likewise.
41708         (sem_item_optimizer::subdivide_classes_by_sensitive_refs): Likewise.
41709         (sem_item_optimizer::verify_classes): Likewise.
41710         (sem_item_optimizer::process_cong_reduction): Likewise.
41711         (sem_item_optimizer::dump_cong_classes): Likewise.
41712         (sem_item_optimizer::merge_classes): Likewise.
41713         * ipa-icf.h (congruence_class_hash): Rename from
41714         congruence_class_group_hash.  Remove declaration of m_classes_vec.
41716 2017-01-10  Andrew Senkevich  <andrew.senkevich@intel.com>
41718         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VPOPCNTDQ_SET,
41719         OPTION_MASK_ISA_AVX512VPOPCNTDQ_UNSET): New.
41720         * config.gcc: Add avx512vpopcntdqintrin.h.
41721         * config/i386/avx512vpopcntdqintrin.h: New.
41722         * config/i386/cpuid.h (bit_AVX512VPOPCNTDQ): New.
41723         * config/i386/i386-builtin-types.def: Add new types.
41724         * config/i386/i386-builtin.def (__builtin_ia32_vpopcountd_v16si,
41725         __builtin_ia32_vpopcountd_v16si_mask, __builtin_ia32_vpopcountq_v8di,
41726         __builtin_ia32_vpopcountq_v8di_mask): New.
41727         * config/i386/i386-c.c (ix86_target_macros_internal): Define
41728         __AVX512VPOPCNTDQ__.
41729         * config/i386/i386.c (ix86_target_string): Add -mavx512vpopcntdq.
41730         (PTA_AVX512VPOPCNTDQ): Define.
41731         * config/i386/i386.h (TARGET_AVX512VPOPCNTDQ,
41732         TARGET_AVX512VPOPCNTDQ_P): Define.
41733         * config/i386/i386.opt: Add mavx512vpopcntdq.
41734         * config/i386/immintrin.h: Include avx512vpopcntdqintrin.h.
41735         * config/i386/sse.md (define_insn "vpopcount<mode><mask_name>"): New.
41737 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
41739         PR middle-end/77484
41740         * predict.def (PRED_CALL): Set to 67.
41742 2017-01-09  Eric Botcazou  <ebotcazou@adacore.com>
41744         * expr.c (store_field): In the bitfield case, if the value comes from
41745         a function call and is of an aggregate type returned in registers, do
41746         not modify the field mode; extract the value in all cases if the mode
41747         is BLKmode and the size is not larger than a word.
41749 2017-01-09  Dominique d'Humieres  <dominiq@lps.ens.fr>
41751         PR target/71017
41752         * config/i386/cpuid.h: Fix undefined behavior.
41754 2017-01-04  Jeff Law  <law@redhat.com>
41756         PR tree-optimization/79007
41757         PR tree-optimization/67955
41758         * tree-ssa-alias.c (same_addr_size_stores_p): Only need to be
41759         conservative for pt.null when flag_non_call_exceptions is on.
41761 2017-01-09  Jakub Jelinek  <jakub@redhat.com>
41763         PR translation/79019
41764         PR translation/79020
41765         * params.def (PARAM_INLINE_MIN_SPEEDUP,
41766         PARAM_IPA_CP_SINGLE_CALL_PENALTY,
41767         PARAM_USE_AFTER_SCOPE_DIRECT_EMISSION_THRESHOLD): Fix typos
41768         in descriptions.
41769         * config/avr/avr.opt (maccumulate-args): Likewise.
41770         * config/msp430/msp430.opt (mwarn-mcu): Likewise.
41771         * common.opt (freport-bug): Likewise.
41772         * cif-code.def (CIF_FINAL_ERROR): Likewise.
41773         * doc/invoke.texi (ipa-cp-single-call-penalty): Likewise.
41774         * config/s390/s390.c (s390_invalid_binary_op): Fix spelling in
41775         translatable string.
41776         * config/i386/i386.c (function_value_32): Likewise.
41777         * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Likewise.
41778         * config/msp430/msp430.c (msp430_option_override, msp430_attr):
41779         Likewise.
41780         * config/msp430/driver-msp430.c (msp430_select_hwmult_lib): Likewise.
41781         * common/config/msp430/msp430-common.c (msp430_handle_option):
41782         Likewise.
41783         * symtab.c (symtab_node::verify_base): Likewise.
41784         * opts.c (set_debug_level): Likewise.
41785         * tree.c (verify_type_variant): Likewise.  Fix typo in comment.
41786         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Add
41787         missing whitespace to translatable strings.
41788         * config/avr/avr.md (bswapsi2): Fix typo in comment.
41789         * config/sh/superh.h: Likewise.
41790         * config/i386/xopintrin.h: Likewise.
41791         * config/i386/znver1.md: Likewise.
41792         * config/rs6000/rs6000.c (struct rs6000_opt_mask): Likewise.
41793         * ipa-inline-analysis.c (compute_inline_parameters): Likewise.
41794         * double-int.h (struct double_int): Likewise.
41795         * double-int.c (div_and_round_double): Likewise.
41796         * wide-int.cc: Likewise.
41797         * tree-ssa.c (non_rewritable_mem_ref_base): Likewise.
41798         * tree-ssa-sccvn.c (vn_reference_lookup_3): Likewise.
41799         * cfgcleanup.c (crossjumps_occured): Renamed to ...
41800         (crossjumps_occurred): ... this.
41801         (try_crossjump_bb, try_head_merge_bb, try_optimize_cfg, cleanup_cfg):
41802         Adjust all uses.
41804         PR tree-optimization/78899
41805         * tree-if-conv.c (version_loop_for_if_conversion): Instead of
41806         returning bool return struct loop *, NULL for failure and the new
41807         loop on success.
41808         (versionable_outer_loop_p): Don't version outer loop if it has
41809         dont_vectorized bit set.
41810         (tree_if_conversion): When versioning outer loop, ensure
41811         tree_if_conversion is performed also on the inner loop of the
41812         non-vectorizable outer loop copy.
41813         * tree-vectorizer.c (set_uid_loop_bbs): Formatting fix.  Fold
41814         LOOP_VECTORIZED in inner loop of the scalar outer loop and
41815         prevent vectorization of it.
41816         (vectorize_loops): For outer + inner LOOP_VECTORIZED, ensure
41817         the outer loop vectorization of the non-scalar version is attempted
41818         before vectorization of the inner loop in scalar version.  If
41819         outer LOOP_VECTORIZED guarded loop is not vectorized, prevent
41820         vectorization of its inner loop.
41821         * tree-vect-loop-manip.c (rename_variables_in_bb): If outer_loop
41822         has 2 inner loops, rename also on edges from bb whose single pred
41823         is outer_loop->header.  Fix typo in function comment.
41825 2017-01-09  Martin Sebor  <msebor@redhat.com>
41827         PR bootstrap/79033
41828         * asan.c (asan_emit_stack_protection): Increase local buffer size
41829         to avoid snprintf truncation warning.
41831 2017-01-09  Andrew Pinski  <apinski@cavium.com>
41833         * config/aarch64/aarch64-cores.def: Add thunderx2t99.  Change vulcan
41834         to reference thunderx2t99 for the tuning structure
41835         * config/aarch64/aarch64-cost-tables.h (vulcan_extra_costs):
41836         Rename to ...
41837         (thunderx2t99_extra_costs): This.
41838         * config/aarch64/aarch64-tune.md: Regenerate.
41839         * config/aarch64/aarch64.c (vulcan_addrcost_table): Rename to ...
41840         (vulcan_addrcost_table): This.
41841         (vulcan_regmove_cost): Rename to ...
41842         (thunderx2t99_regmove_cost): This.
41843         (vulcan_vector_cost): Rename to ...
41844         (thunderx2t99_vector_cost): this.
41845         (vulcan_branch_cost): Rename to ...
41846         (thunderx2t99_branch_cost): This.
41847         (vulcan_tunings): Rename to ...
41848         (thunderx2t99_tunings): This and s/vulcan/thunderx2t99 .
41849         * doc/invoke.texi (AARCH64/mtune): Add thunderx2t99.
41851 2017-01-09  Martin Jambor  <mjambor@suse.cz>
41853         PR ipa/78365
41854         PR ipa/78599
41855         * ipa-prop.h (ipa_jump_func): Swap positions of vr_known and m_vr.
41856         * ipa-cp.c (ipa_vr_operation_and_type_effects): New function.
41857         (propagate_vr_accross_jump_function): Use the above function for all
41858         value range computations for pass-through jump functions and type
41859         converasion from explicit value range values.
41860         (ipcp_propagate_stage): Do not attempt to deduce types of formal
41861         parameters from TYPE_ARG_TYPES.
41862         * ipa-prop.c (ipa_write_jump_function): Remove trailing whitespace.
41863         (ipa_write_node_info): Stream type of the actual argument.
41864         (ipa_read_node_info): Likewise. Also remove trailing whitespace.
41866 2017-01-09  Martin Liska  <mliska@suse.cz>
41868         PR pch/78970
41869         * gcc.c (driver_handle_option): Handle OPT_E and set have_E.
41870         (lookup_compiler): Do not show error message with have_E.
41872 2017-01-09  Jakub Jelinek  <jakub@redhat.com>
41874         PR tree-optimization/78938
41875         * tree-vect-stmts.c (vectorizable_condition): For non-masked COND_EXPR
41876         where comp_vectype is VECTOR_BOOLEAN_TYPE_P, use
41877         BIT_{NOT,XOR,AND,IOR}_EXPR on the comparison operands instead of
41878         {EQ,NE,GE,GT,LE,LT}_EXPR directly inside of VEC_COND_EXPR.  Formatting
41879         fixes.
41881 2017-01-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
41883         * tree-ssa-address.c (gen_addr_rtx): Don't handle index if it
41884         is const0_rtx.
41886 2017-01-09  Richard Biener  <rguenther@suse.de>
41888         PR tree-optimization/78997
41889         * tree-vect-slp.c (vect_mask_constant_operand_p): Handle SSA
41890         name condition properly.
41892 2017-01-09  Richard Biener  <rguenther@suse.de>
41894         PR debug/79000
41895         * dwarf2out.c (is_cxx): New overload with context.
41896         (is_naming_typedef_decl): Use it.
41898 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41900         * invoke.texi (Option Summary): Correct spacing in option lists
41901         and add line breaks to fix over-long lines.
41903 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41905         PR middle-end/17660
41907         * extend.texi (Common Variable Attributes): Add xref to GCC
41908         Internals manual to explain mode attribute keywords.
41910 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41912         PR other/16519
41913         * doc/invoke.texi (Option Summary): Move -pthread to Linker Options
41914         and Preprocessor Options.
41915         (Options for Linking): Document -pthread here....
41916         (RS/6000 and PowerPC Options): ...not here.
41917         (Solaris 2 Options): ...or here.
41918         * doc/cppopts.texi: Document -pthread.
41920 2017-01-08  Martin Sebor  <msebor@redhat.com>
41922         PR middle-end/77708
41923         * doc/invoke.texi (Warning Options): Document -Wformat-truncation.
41924         * gimple-ssa-sprintf.c (call_info::reval_used, call_info::warnopt):
41925         New member functions.
41926         (format_directive): Used them.
41927         (add_bytes): Same.
41928         (pass_sprintf_length::handle_gimple_call): Same.
41929         * graphite-sese-to-poly.c (tree_int_to_gmp): Increase buffer size
41930         to avoid truncation for any argument.
41931         (extract_affine_mul): Same.
41932         * tree.c (get_file_function_name): Same.
41934 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
41936         PR middle-end/77484
41937         * predict.def (PRED_INDIR_CALL): Set to 86.
41939 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41941         PR preprocessor/54124
41942         * doc/cppopts.texi: Reformat -d subtable to list the full name
41943         of the options.  Add cross-reference to the docs for the general
41944         compiler -d options.
41945         * doc/invoke.texi (Developer Options): Add cross-reference to the
41946         preprocessor-specific -d option documentation.
41948 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41950         PR preprocessor/13498
41951         * doc/cpp.texi (Search Path): Rewrite to remove obsolete and
41952         redudant material, and reflect new command-line options.
41953         (System Headers): Likewise.
41955 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41957         * doc/cppdiropts.texi: Merge documentation of -I, -iquote,
41958         -isystem, and -idirafter.  Copy-edit.
41959         * doc/cppopts.texi: Copy-edit.  Remove contradiction about
41960         default for -ftrack-macro-expansion.  Delete obsolete and
41961         badly-formatted implementation details about -fdebug-cpp output.
41962         * doc/cppwarnopts.texi: Copy-edit.
41964 2017-01-07  David Malcolm  <dmalcolm@redhat.com>
41966         PR c++/72803
41967         * input.c (selftest::test_accessing_ordinary_linemaps): Verify
41968         that the transition from a max line width >= 1<<10 to narrower
41969         lines works correctly.
41971 2017-01-07  Alexandre Oliva  <aoliva@redhat.com>
41973         * doc/options.texi (PerFunction): New.
41974         * opt-functions.awk (switch_flags): Map both Optimization and
41975         PerFunction to CL_OPTIMIZATION.
41976         * opth-gen.awk: Test for PerFunction flag along with
41977         Optimization.
41978         * optc-save-gen.awk: Likewise.  Introduce var_opt_hash and set
41979         it only when the latter is present.  Skip those that don't in
41980         the hash function generator.
41981         * common.opt (fvar-tracking): Mark as PerFunction instead of
41982         Optimization.
41983         (fvar-tracking-assignments): Likewise.
41984         (fvar-tracking-assignments-toggle): Likewise.
41985         (fvar-tracking-uninit): Likewise.
41987 2017-01-07  Jakub Jelinek  <jakub@redhat.com>
41989         PR translation/79018
41990         * params.def (PARAM_MAX_STORES_TO_MERGE): Add missing space between
41991         the and store.
41993 2017-01-06  Mikael Pettersson  <mikpelinux@gmail.com>
41995         PR target/57583
41996         * config/m68k/m68k.opt (LONG_JUMP_TABLE_OFFSETS): New option.
41997         * config/m68k/linux.h (ASM_RETURN_CASE_JUMP): Handle
41998         TARGET_LONG_JUMP_TABLE_OFFSETS.
41999         * config/m68k/m68kelf.h (ASM_RETURN_CASE_JUMP): Likewise.
42000         * config/m68k/netbsd-elf.h (ASM_RETURN_CASE_JUMP): Likewise.
42001         * config/m68k/m68k.h (CASE_VECTOR_MODE): Likewise.
42002         (ASM_OUTPUT_ADDR_DIFF_ELF): Likewise.
42003         * config/m68k/m68k.md (tablejump expander): Likewise.
42004         (*tablejump_pcrel_hi): Renamed from unnamed insn, reject
42005         TARGET_LONG_JUMP_TABLE_OFFSETS.
42006         (*tablejump_pcrel_si): New insn, handle TARGET_LONG_JUMP_TABLE_OFFSETS.
42007         * doc/invoke.texi (M68K options): Add -mlong-jump-table-offsets.
42009 2017-01-06  Edgar E. Iglesias <edgar.iglesias@xilinx.com>
42010             David Holsgrove <david.holsgrove@xilinx.com>
42012         * common/config/microblaze/microblaze-common.c
42013         (TARGET_EXCEPT_UNWIND_INFO): Remove.
42014         * config/microblaze/microblaze-protos.h (microblaze_eh_return):
42015         New prototype.
42016         * config/microblaze/microblaze.c (microblaze_must_save_register)
42017         (microblaze_expand_epilogue, microblaze_return_addr): Handle
42018         calls_eh_return.
42019         (microblaze_eh_return): New function.
42020         * config/microblaze/microblaze.h (RETURN_ADDR_OFFSET)
42021         (EH_RETURN_DATA_REGNO, MB_EH_STACKADJ_REGNUM)
42022         (EH_RETURN_STACKADJ_RTX, ASM_PREFERRED_EH_DATA_FORMAT): New macros.
42023         * config/microblaze/microblaze.md (eh_return): New pattern.
42025 2017-01-06  Jakub Jelinek  <jakub@redhat.com>
42027         * system.h (GCC_DIAGNOSTIC_PUSH_IGNORED, GCC_DIAGNOSTIC_POP,
42028         GCC_DIAGNOSTIC_STRINGIFY): Define.
42030         * read-rtl.c (rtx_reader::read_rtx_code): Avoid -Wsign-compare warning.
42032 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
42034         * config/arm/arm.md (<mcrr>): New.
42035         (<mrrc>): New.
42036         * config/arm/arm.c (arm_arch5te): New.
42037         (arm_option_override): Set arm_arch5te.
42038         (arm_coproc_builtin_available): Add support for mcrr, mcrr2, mrrc
42039         and mrrc2.
42040         * config/arm/arm-builtins.c (MCRR_QUALIFIERS): Define to...
42041         (arm_mcrr_qualifiers): ... this. New.
42042         (MRRC_QUALIFIERS): Define to...
42043         (arm_mrrc_qualifiers): ... this. New.
42044         * config/arm/arm_acle.h (__arm_mcrr, __arm_mcrr2, __arm_mrrc,
42045         __arm_mrrc2): New.
42046         * config/arm/arm_acle_builtins.def (mcrr, mcrr2, mrrc, mrrc2): New.
42047         * config/arm/iterators.md (MCRRI, mcrr, MCRR): New.
42048         (MRRCI, mrrc, MRRC): New.
42049         * config/arm/unspecs.md (VUNSPEC_MCRR, VUNSPEC_MCRR2, VUNSPEC_MRRC,
42050         VUNSPEC_MRRC2): New.
42052 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
42054         * config/arm/arm.md (<mcr>): New.
42055         (<mrc>): New.
42056         * config/arm/arm.c (arm_coproc_builtin_available): Add
42057         support for mcr, mrc, mcr2 and mrc2.
42058         * config/arm/arm-builtins.c (MCR_QUALIFIERS): Define to...
42059         (arm_mcr_qualifiers): ... this. New.
42060         (MRC_QUALIFIERS): Define to ...
42061         (arm_mrc_qualifiers): ... this. New.
42062         (MCR_QUALIFIERS): Define to ...
42063         (arm_mcr_qualifiers): ... this. New.
42064         * config/arm/arm_acle.h (__arm_mcr, __arm_mrc, __arm_mcr2,
42065         __arm_mrc2): New.
42066         * config/arm/arm_acle_builtins.def (mcr, mcr2, mrc, mrc2): New.
42067         * config/arm/iterators.md (MCRI, mcr, MCR, MRCI, mrc, MRC): New.
42068         * config/arm/unspecs.md (VUNSPEC_MCR, VUNSPEC_MCR2, VUNSPEC_MRC,
42069         VUNSPEC_MRC2): New.
42071 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
42073         * config/arm/arm.md (*ldc): New.
42074         (*stc): New.
42075         (<ldc>): New.
42076         (<stc>): New.
42077         * config/arm/arm.c (arm_coproc_builtin_available): Add
42078         support for ldc,ldcl,stc,stcl,ldc2,ldc2l,stc2 and stc2l.
42079         (arm_coproc_ldc_stc_legitimate_address): New.
42080         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
42081         'qualifier_const_pointer'.
42082         (LDC_QUALIFIERS): Define to...
42083         (arm_ldc_qualifiers): ... this. New.
42084         (STC_QUALIFIERS): Define to...
42085         (arm_stc_qualifiers): ... this. New.
42086         * config/arm/arm-protos.h
42087         (arm_coproc_ldc_stc_legitimate_address): New.
42088         * config/arm/arm_acle.h (__arm_ldc, __arm_ldcl, __arm_stc,
42089         __arm_stcl, __arm_ldc2, __arm_ldc2l, __arm_stc2, __arm_stc2l): New.
42090         * config/arm/arm_acle_builtins.def (ldc, ldc2, ldcl, ldc2l, stc,
42091         stc2, stcl, stc2l): New.
42092         * config/arm/constraints.md (Uz): New.
42093         * config/arm/iterators.md (LDCI, STCI, ldc, stc, LDC STC): New.
42094         * config/arm/unspecs.md (VUNSPEC_LDC, VUNSPEC_LDC2, VUNSPEC_LDCL,
42095         VUNSPEC_LDC2L, VUNSPEC_STC, VUNSPEC_STC2, VUNSPEC_STCL,
42096         VUNSPEC_STC2L): New.
42098 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
42100         * config/arm/arm.md (<cdp>): New.
42101         * config/arm/arm.c (neon_const_bounds): Rename this ...
42102         (arm_const_bounds): ... this.
42103         (arm_coproc_builtin_available): New.
42104         * config/arm/arm-builtins.c (SIMD_MAX_BUILTIN_ARGS): Increase.
42105         (arm_type_qualifiers): Add 'qualifier_unsigned_immediate'.
42106         (CDP_QUALIFIERS): Define to...
42107         (arm_cdp_qualifiers): ... this. New.
42108         (void_UP): Define.
42109         (arm_expand_builtin_args): Add case for 6 arguments.
42110         * config/arm/arm-protos.h (neon_const_bounds): Rename this ...
42111         (arm_const_bounds): ... this.
42112         (arm_coproc_builtin_available): New.
42113         * config/arm/arm_acle.h (__arm_cdp): New.
42114         (__arm_cdp2): New.
42115         * config/arm/arm_acle_builtins.def (cdp): New.
42116         (cdp2): New.
42117         * config/arm/iterators.md (CDPI,CDP,cdp): New.
42118         * config/arm/neon.md: Rename all 'neon_const_bounds' to
42119         'arm_const_bounds'.
42120         * config/arm/types.md (coproc): New.
42121         * config/arm/unspecs.md (VUNSPEC_CDP, VUNSPEC_CDP2): New.
42122         * doc/extend.texi (ACLE): Add a mention of Coprocessor intrinsics.
42123         * doc/sourcebuild.texi (arm_coproc1_ok, arm_coproc2_ok,
42124         arm_coproc3_ok, arm_coproc4_ok): Document new effective targets.
42126 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
42128         * config/arm/arm-builtins.c (arm_unsigned_binop_qualifiers): New.
42129         (UBINOP_QUALIFIERS): New.
42130         (si_UP): Define.
42131         (acle_builtin_data): New. Change comment.
42132         (arm_builtins): Remove ARM_BUILTIN_CRC32B, ARM_BUILTIN_CRC32H,
42133         ARM_BUILTIN_CRC32W, ARM_BUILTIN_CRC32CB, ARM_BUILTIN_CRC32CH,
42134         ARM_BUILTIN_CRC32CW. Add ARM_BUILTIN_ACLE_BASE and include
42135         arm_acle_builtins.def.
42136         (ARM_BUILTIN_ACLE_PATTERN_START): Define.
42137         (arm_init_acle_builtins): New.
42138         (CRC32_BUILTIN): Remove.
42139         (bdesc_2arg): Remove entries for crc32b, crc32h, crc32w,
42140         crc32cb, crc32ch and crc32cw.
42141         (arm_init_crc32_builtins): Remove.
42142         (arm_init_builtins): Use arm_init_acle_builtins rather
42143         than arm_init_crc32_builtins.
42144         (arm_expand_acle_builtin): New.
42145         (arm_expand_builtin): Use 'arm_expand_acle_builtin'.
42146         * config/arm/arm_acle_builtins.def: New.
42148 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
42150         * config/arm/arm-builtins.c (neon_builtin_datum): Rename to ..
42151         (arm_builtin_datum): ... this.
42152         (arm_init_neon_builtin): Rename to ...
42153         (arm_init_builtin): ... this. Add a new parameters PREFIX
42154         and USE_SIG_IN_NAME.
42155         (arm_init_neon_builtins): Replace 'arm_init_neon_builtin' with
42156         'arm_init_builtin'. Replace type 'neon_builtin_datum' with
42157         'arm_builtin_datum'.
42158         (arm_init_vfp_builtins): Likewise.
42159         (builtin_arg): Rename enum's replacing 'NEON_ARG' with
42160         'ARG_BUILTIN' and add a 'ARG_BUILTIN_NEON_MEMORY.
42161         (arm_expand_neon_args): Rename to ...
42162         (arm_expand_builtin_args): ... this. Rename builtin_arg
42163         enum values and differentiate between ARG_BUILTIN_MEMORY
42164         and ARG_BUILTIN_NEON_MEMORY.
42165         (arm_expand_neon_builtin_1): Rename to ...
42166         (arm_expand_builtin_1): ... this. Rename builtin_arg enum
42167         values, arm_expand_builtin_args and add bool parameter NEON.
42168         (arm_expand_neon_builtin): Use arm_expand_builtin_1.
42169         (arm_expand_vfp_builtin): Likewise.
42170         (NEON_MAX_BUILTIN_ARGS): Remove, it was unused.
42172 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
42174         PR middle-end/77484
42175         * predict.def (PRED_POLYMORPHIC_CALL): Set to 59.
42176         * predict.c (tree_estimate_probability_bb): Reverse direction of
42177         polymorphic call predictor.
42179 2017-01-06  David Malcolm  <dmalcolm@redhat.com>
42181         * passes.c (execute_one_pass): Split out pass-skipping logic into...
42182         (determine_pass_name_match): ...this new function and...
42183         (should_skip_pass_p): ...this new function.
42185 2017-01-06  Nathan Sidwell  <nathan@acm.org>
42187         * ipa-visibility.c (function_and_variable_visibility): Reformat
42188         comments and long lines.  Remove extrneous if.
42189         * symtab.c (symtab_node::make_decl_local): Fix code format.
42190         (symtab_node::set_section_for_node): Fix comment typo.
42192 2017-01-06  Martin Liska  <mliska@suse.cz>
42194         PR bootstrap/79003
42195         * lra-constraints.c: Rename invariant to lra_invariant.
42196         * predict.c (set_even_probabilities): Initialize e to NULL.
42198 2017-01-05  Martin Sebor  <msebor@redhat.com>
42200         PR tree-optimization/78910
42201         * gimple-ssa-sprintf.c (tree_digits): Add an argument.
42202         (format_integer): Correct off-by-one error in the handling
42203         of precision with negative numbers in signed conversions..
42205 2017-01-05  Eric Botcazou  <ebotcazou@adacore.com>
42207         * doc/invoke.texi (C Dialect Options): Adjust -fsso-struct entry.
42209 2017-01-05  Jakub Jelinek  <jakub@redhat.com>
42211         PR tree-optimization/71016
42212         * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Pass cond_stmt to
42213         factor_out_conditional_conversion.  Formatting fix.
42214         (factor_out_conditional_conversion): Add cond_stmt argument.
42215         If arg1 is INTEGER_CST, punt if new_arg0 is not any operand of
42216         cond_stmt and if arg0_def_stmt is not the only stmt in its bb.
42217         Formatting fix.
42219 2017-01-05  David Malcolm  <dmalcolm@redhat.com>
42221         * Makefile.in (OBJS): Add read-md.o, read-rtl.o,
42222         read-rtl-function.o, and selftest-rtl.o.
42223         * config/aarch64/aarch64.c: Include selftest.h and selftest-rtl.h.
42224         (selftest::aarch64_test_loading_full_dump): New function.
42225         (selftest::aarch64_run_selftests): New function.
42226         (TARGET_RUN_TARGET_SELFTESTS): Wire it up to
42227         selftest::aarch64_run_selftests.
42228         * config/i386/i386.c
42229         (selftest::ix86_test_loading_dump_fragment_1): New function.
42230         (selftest::ix86_test_loading_call_insn): New function.
42231         (selftest::ix86_test_loading_full_dump): New function.
42232         (selftest::ix86_test_loading_unspec): New function.
42233         (selftest::ix86_run_selftests): Call the new functions.
42234         * emit-rtl.c (maybe_set_max_label_num): New function.
42235         * emit-rtl.h (maybe_set_max_label_num): New decl.
42236         * function.c (instantiate_decls): Guard call to
42237         instantiate_decls_1 with if (DECL_INITIAL (fndecl)).
42238         * function-tests.c (selftest::verify_three_block_rtl_cfg): Remove
42239         "static".
42240         * gensupport.c (gen_reader::gen_reader): Pass "false"
42241         for new "compact" param of rtx_reader.
42242         * print-rtl.c (rtx_writer::print_rtx_operand): Print "(nil)"
42243         rather than an empty string for NULL strings.
42244         * read-md.c: Potentially include config.h rather than bconfig.h.
42245         Wrap include of errors.h with #ifdef GENERATOR_FILE.
42246         (have_error): New global, copied from errors.c.
42247         (md_reader::read_name): Rename to...
42248         (md_reader::read_name_1): ...this, adding "out_loc" param,
42249         and converting "missing name or number" to returning false, rather
42250         than failing.
42251         (md_reader::read_name): Reimplement in terms of read_name_1.
42252         (md_reader::read_name_or_nil): New function.
42253         (md_reader::read_string): Handle "(nil)" by returning NULL.
42254         (md_reader::md_reader): Add new param "compact".
42255         (md_reader::read_md_files): Wrap with #ifdef GENERATOR_FILE.
42256         (md_reader::read_file): New method.
42257         * read-md.h (md_reader::md_reader): Add new param "compact".
42258         (md_reader::read_file): New method.
42259         (md_reader::is_compact): New accessor.
42260         (md_reader::read_name): Convert return type from void to file_location.
42261         (md_reader::read_name_or_nil): New decl.
42262         (md_reader::read_name_1): New decl.
42263         (md_reader::m_compact): New field.
42264         (noop_reader::noop_reader): Pass "false" for new "compact" param
42265         of rtx_reader.
42266         (rtx_reader::rtx_reader): Add new "compact" param.
42267         (rtx_reader::read_rtx_operand): Make virtual and convert return
42268         type from void to rtx.
42269         (rtx_reader::read_until): New decl.
42270         (rtx_reader::handle_any_trailing_information): New virtual function.
42271         (rtx_reader::postprocess): New virtual function.
42272         (rtx_reader::finalize_string): New virtual function.
42273         (rtx_reader::m_in_call_function_usage): New field.
42274         (rtx_reader::m_reuse_rtx_by_id): New field.
42275         * read-rtl-function.c: New file.
42276         * selftest-rtl.c (selftest::assert_rtx_ptr_eq_at): New function.
42277         * selftest-rtl.h (ASSERT_RTX_PTR_EQ): New macro.
42278         (selftest::verify_three_block_rtl_cfg): New decl.
42279         * read-rtl-function.h: New file.
42280         * read-rtl.c: Potentially include config.h rather than bconfig.h.
42281         For host, include function.h, memmodel.h, and emit-rtl.h.
42282         (one_time_initialization): New function.
42283         (struct compact_insn_name): New struct.
42284         (compact_insn_names): New array.
42285         (find_code): Handle insn codes in compact dumps.
42286         (apply_subst_iterator): Wrap with #ifdef GENERATOR_FILE.
42287         (bind_subst_iter_and_attr): Likewise.
42288         (add_condition_to_string): Likewise.
42289         (add_condition_to_rtx): Likewise.
42290         (apply_attribute_uses): Likewise.
42291         (add_current_iterators): Likewise.
42292         (apply_iterators): Likewise.
42293         (initialize_iterators): Guard usage of apply_subst_iterator with
42294         #ifdef GENERATOR_FILE.
42295         (read_conditions): Wrap with #ifdef GENERATOR_FILE.
42296         (md_reader::read_mapping): Likewise.
42297         (add_define_attr_for_define_subst): Likewise.
42298         (add_define_subst_attr): Likewise.
42299         (read_subst_mapping): Likewise.
42300         (check_code_iterator): Likewise.
42301         (rtx_reader::read_rtx): Likewise.  Move one-time initialization
42302         logic to...
42303         (one_time_initialization): New function.
42304         (rtx_reader::read_until): New method.
42305         (read_flags): New function.
42306         (parse_reg_note_name): New function.
42307         (rtx_reader::read_rtx_code): Initialize "iterator" to NULL.
42308         Handle reuse_rtx ids.
42309         Wrap iterator lookup within #ifdef GENERATOR_FILE.
42310         Add parsing support for RTL dumps, mirroring the special-cases in
42311         print_rtx, by calling read_flags, reading REG_NOTE names, INSN_UID
42312         values, and calling handle_any_trailing_information.
42313         (rtx_reader::read_rtx_operand): Convert return type from void
42314         to rtx, returning return_rtx.  Handle case 'e'.  Call
42315         finalize_string on XSTR and XTMPL fields.
42316         (rtx_reader::read_nested_rtx):  Handle dumps in which trailing
42317          "(nil)" values were omitted.  Call the postprocess vfunc on the
42318         return_rtx.
42319         (rtx_reader::rtx_reader): Add new "compact" param and pass to base
42320         class ctor.  Initialize m_in_call_function_usage.  Call
42321         one_time_initialization.
42322         * rtl-tests.c (selftest::test_uncond_jump): Call
42323         set_new_first_and_last_insn.
42324         * rtl.h (read_rtx): Wrap decl with #ifdef GENERATOR_FILE.
42325         * selftest-rtl.c: New file.
42326         * selftest-rtl.h (class selftest::rtl_dump_test): New class.
42327         (selftest::get_insn_by_uid): New decl.
42328         * selftest-run-tests.c (selftest::run_tests): Call
42329         read_rtl_function_c_tests.
42330         * selftest.h  (selftest::read_rtl_function_c_tests): New decl.
42331         * tree-dfa.c (ssa_default_def): Return NULL_TREE for rtl function
42332         dumps.
42334 2017-01-05  Uros Bizjak  <ubizjak@gmail.com>
42336         * config/i386/i386.md (*testqi_ext_3): No need to handle memory
42337         operands in a special way.  Assert that pos+len <= mode precision.
42339 2017-01-05  Jakub Jelinek  <jakub@redhat.com>
42341         * common.opt (fvect-cost-model): Remove RejectNegative flag, use
42342         3 argument Alias with unlimited for the negative form.
42343         (fno-vect-cost-model): Removed.
42345 2017-01-05  Martin Liska  <mliska@suse.cz>
42347         * hsa-gen.c (gen_hsa_divmod): New function.
42348         (gen_hsa_insn_for_internal_fn_call): Use the function for IFN_DIVMOD.
42350 2017-01-05  Martin Liska  <mliska@suse.cz>
42352         PR pch/78970
42353         * gcc.c (lookup_compiler): Reject '-' filename for a precompiled
42354         header.
42356 2017-01-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
42358         * config/s390/s390.c (s390_expand_setmem): Unroll the loop for
42359         small constant length operands.
42361 2017-01-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
42363         * config/s390/s390.c (s390_expand_setmem): Avoid overlapping bytes
42364         between loop iterations.
42366 2017-01-05  Martin Liska  <mliska@suse.cz>
42368         PR sanitizer/78815
42369         * gimplify.c (gimplify_decl_expr): Compare to
42370         asan_poisoned_variables instread of checking flags.
42371         (gimplify_target_expr): Likewise.
42372         (gimplify_expr): Likewise.
42373         (gimplify_function_tree): Conditionally initialize
42374         asan_poisoned_variables.
42376 2017-01-04  Jeff Law  <law@redhat.com>
42378         PR tree-optimizatin/78812
42379         * rtl.h (contains_mem_rtx_p): Prototype.
42380         * ifcvt.c (containts_mem_rtx_p): Move from here to...
42381         * rtlanal.c (contains_mem_rtx_p): Here and remove static linkage.
42382         * gcse.c (prune_expressions): Use contains_mem_rtx_p to discover
42383         and prune MEMs that are not at the toplevel of a SET_SRC rtx.  Look
42384         through ZERO_EXTEND and SIGN_EXTEND when trying to avoid pruning MEMs.
42386 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42388         * input.c (assert_char_at_range): Default-initialize actual_range.
42390 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42392         * df-scan.c (df_ref_create_structure): Make regno unsigned,
42393         to match the caller.
42395 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42397         * cfgexpand.c (expand_gimple_basic_block): Disregard debug
42398         insns after final jump in test to emit dummy move.
42400 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42402         * gimple-iterator.h (gsi_one_nondebug_before_end_p): New.
42403         * tree-eh.c (cleanup_empty_eh): Skip more debug stmts.
42405 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42407         * multiple_target.c (create_dispatcher_calls): Init e_next.
42408         * tree-ssa-loop-split.c (split_loop): Init border.
42409         * tree-vect-loop.c (vect_determine_vectorization_factor): Init
42410         scalar_type.
42412 2017-01-04  Michael Meissner  <meissner@linux.vnet.ibm.com>
42414         PR target/71977
42415         PR target/70568
42416         PR target/78823
42417         * config/rs6000/predicates.md (sf_subreg_operand): New predicate.
42418         (altivec_register_operand): Do not return true if the operand
42419         contains a SUBREG mixing SImode and SFmode.
42420         (vsx_register_operand): Likewise.
42421         (vsx_reg_sfsubreg_ok): New predicate.
42422         (vfloat_operand): Do not return true if the operand contains a
42423         SUBREG mixing SImode and SFmode.
42424         (vint_operand): Likewise.
42425         (vlogical_operand): Likewise.
42426         (gpc_reg_operand): Likewise.
42427         (int_reg_operand): Likewise.
42428         * config/rs6000/rs6000-protos.h (valid_sf_si_move): Add declaration.
42429         * config/rs6000/rs6000.c (valid_sf_si_move): New function to
42430         determine if a MOVSI or MOVSF operation contains SUBREGs that mix
42431         SImode and SFmode.
42432         (rs6000_emit_move_si_sf_subreg): New helper function.
42433         (rs6000_emit_move): Call rs6000_emit_move_si_sf_subreg to possbily
42434         fixup SUBREGs involving SImode and SFmode.
42435         * config/rs6000/vsx.md (SFBOOL_*): New constants that are operand
42436         numbers for the new peephole2 optimization.
42437         (peephole2 for SFmode unions): New peephole2 to optimize cases in
42438         the GLIBC math library that do AND/IOR/XOR operations on single
42439         precision floating point.
42440         * config/rs6000/rs6000.h (TARGET_NO_SF_SUBREG): New internal
42441         target macros to say whether we need to avoid SUBREGs mixing
42442         SImode and SFmode.
42443         (TARGET_ALLOW_SF_SUBREG): Likewise.
42444         * config/rs6000/rs6000.md (UNSPEC_SF_FROM_SI): New unspecs.
42445         (UNSPEC_SI_FROM_SF): Likewise.
42446         (iorxor): Change spacing.
42447         (and_ior_xor): New iterator for AND, IOR, and XOR.
42448         (movsi_from_sf): New insns for SImode/SFmode SUBREG support.
42449         (movdi_from_sf_zero_ext): Likewise.
42450         (mov<mode>_hardfloat, FMOVE32 iterator): Use register_operand
42451         instead of gpc_reg_operand.  Add SImode/SFmode SUBREG support.
42452         (movsf_from_si): New insn for SImode/SFmode SUBREG support.
42453         (fma<mode>4): Use gpc_reg_operand instead of register_operand.
42454         (fms<mode>4): Likewise.
42455         (fnma<mode>4): Likewise.
42456         (fnms<mode>4): Likewise.
42457         (nfma<mode>4): Likewise.
42458         (nfms<mode>4): Likewise.
42460 2017-01-04  Marek Polacek  <polacek@redhat.com>
42462         PR c++/64767
42463         * doc/invoke.texi: Document -Wpointer-compare.
42465 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42467         * optc-gen.awk: Emit #error for -W*/-f*/-m* Enum without
42468         RejectNegative.
42470         * dwarf2out.c (output_loc_list): Don't throw away 64K+ location
42471         descriptions for -gdwarf-5 and emit them as uleb128 instead of
42472         2-byte data.
42474 2017-01-04  Kelvin Nilsen  <kelvin@gcc.gnu.org>
42476         PR target/78056
42477         * doc/sourcebuild.texi (PowerPC-specific attributes): Add
42478         documentation of the powerpc_popcntb_ok attribute.
42479         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
42480         code to issue warning messages if a requested CPU configuration is
42481         not supported by the binary (assembler and loader) toolchain.
42482         (spe_init_builtins): Add two assertions to prevent ICE if attempt is
42483         made to define a built-in function that has been disabled.
42484         (paired_init_builtins): Add assertion to prevent ICE if attempt is
42485         made to define a built-in function that has been disabled.
42486         (altivec_init_builtins): Add comment explaining why definition
42487         of the DST built-in functions is not preceded by an assertion
42488         check.  Add assertions to prevent ICE if attempts are made to
42489         define an altivec predicate or an abs* built-in function that has
42490         been disabled.
42491         (htm_init_builtins): Add comment explaining why definition of the
42492         htm built-in functions is not preceded by an assertion check.
42494 2017-01-04  Jeff Law  <law@redhat.com>
42496         PR tree-optimizatin/67955
42497         * tree-ssa-alias.c (same_addr_size_stores_p): Check offsets first.
42498         Allow any SSA_VAR_P as the base objects.  Use integer_zerop.  Verify
42499         the points-to solution does not include pt_null.  Use DECL_PT_UID
42500         unconditionally.
42502 2017-01-04  Uros Bizjak  <ubizjak@gmail.com>
42504         * config/i386/i386.md (HI/SImode test with imm to QImode splitters):
42505         Use gen_int_mode instead of gen_lopwart for const_int operands.
42507 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42509         PR tree-optimization/71563
42510         * match.pd: Simplify X << Y into X if Y is known to be 0 or
42511         out of range value - has low bits known to be zero.
42513 2017-01-04  Alan Modra  <amodra@gmail.com>
42515         * Makefile.in (aclocal_deps): Update and order as per aclocal.m4.
42516         * configure: Regenerate.
42517         * config.in: Regenerate.
42519 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42521         PR bootstrap/77569
42522         * input.c (ebcdic_execution_charset::on_error): Don't use strstr for
42523         a substring of the message, but strcmp with the whole message.  Ifdef
42524         ENABLE_NLS, translate the message first using dgettext.
42526 2017-01-03  Jeff Law  <law@redhat.com>
42528         PR tree-optimizatin/78856
42529         * tree-ssa-threadupdate.c: Include tree-vectorizer.h.
42530         (mark_threaded_blocks): Remove code to truncate thread paths that
42531         cross multiple loop headers.  Instead invalidate the cached loop
42532         iteration information and handle case of a thread path walking
42533         into an irreducible region.
42535 2017-01-03  Michael Meissner  <meissner@linux.vnet.ibm.com>
42537         PR target/78900
42538         * config/rs6000/rs6000.c (rs6000_split_signbit): Change some
42539         assertions.  Add support for doing the signbit if the IEEE 128-bit
42540         floating point value is in a GPR.
42541         * config/rs6000/rs6000.md (Fsignbit): Delete.
42542         (signbit<mode>2_dm): Delete using <Fsignbit> and just use "wa".
42543         Update the length attribute if the value is in a GPR.
42544         (signbit<mode>2_dm_<su>ext): Add combiner pattern to eliminate
42545         the sign or zero extension instruction, since the value is always 0/1.
42546         (signbit<mode>2_dm2): Delete using <Fsignbit>.
42548         PR target/78953
42549         * config/rs6000/vsx.md (vsx_extract_<mode>_store_p9): If we are
42550         extracting SImode to a GPR register so that we can generate a
42551         store, limit the vector to be in a traditional Altivec register
42552         for the vextuwrx instruction.
42554 2017-01-03  Ian Lance Taylor  <iant@google.com>
42556         * godump.c (go_format_type): Treat ENUMERAL_TYPE like INTEGER_TYPE.
42558 2017-01-03  Martin Sebor  <msebor@redhat.com>
42560         PR tree-optimization/78696
42561         * gimple-ssa-sprintf.c (format_floating): Correct handling of
42562         precision.  Use MPFR for %f for greater fidelity.  Correct handling
42563         of %g.
42564         (pass_sprintf_length::compute_format_length): Set width and precision
42565         specified by asrerisk to void_node for vararg functions.
42566         (try_substitute_return_value): Adjust dump output.
42568 2017-01-03  David Edelsohn  <dje.gcc@gmail.com>
42570         * doc/invoke.texi (RS6000 options): LRA is enabled by default.
42572 2017-01-03  Eric Botcazou  <ebotcazou@adacore.com>
42574         * doc/invoke.texi (SPARC options): Document -mlra as the default.
42575         * config/sparc/sparc.c (sparc_option_override): Force LRA unless
42576         -mlra/-mno-lra was passed to the compiler.
42578 2017-01-03  James Cowgill  <James.Cowgill@imgtec.com>
42580         PR rtl-optimization/65618
42581         * emit-rtl.c (try_split): Move initialization of "before" and
42582         "after" to just before the call to emit_insn_after_setloc.
42584 2017-01-03  Gerald Pfeifer  <gerald@pfeifer.com>
42586         * doc/md.texi (Standard Names): Remove reference to Java frontend.
42588 2017-01-03  Pierre-Marie de Rodat  <derodat@adacore.com>
42590         * dwarf2out.c (gen_enumeration_type_die): When
42591         -gno-strict-dwarf, add a DW_AT_encoding attribute.
42593 2017-01-03  Jakub Jelinek  <jakub@redhat.com>
42595         PR tree-optimization/78965
42596         * gimple-ssa-sprintf.c (pass_sprintf_length::compute_format_length):
42597         Change first argument from const call_info & to call_info &.  For %n
42598         set info.nowrite to false.
42600         PR middle-end/78901
42601         * gimple-ssa-sprintf.c (try_substitute_return_value): Don't change
42602         possibly throwing calls.
42604         * genmatch.c (dt_node::gen_kids_1): If generic_exprs include SSA_NAME
42605         and exprs_len || fns_len, emit the code for SSA_NAME next to the exprs
42606         and fns handling, rather than in a separate case SSA_NAME.
42608 2017-01-02  Jeff Law  <law@redhat.com>
42610         * config/darwin-driver.c (darwin_driver_init): Const-correctness
42611         fixes for first_period and second_period variables.
42613 2017-01-02  Uros Bizjak  <ubizjak@gmail.com>
42615         PR target/78967
42616         * config/i386/i386.md (UNSPEC_NOREX_MEM): New unspec.
42617         (*insvqi_1): New insn pattern.
42618         (*insvqi_1_mem_rex64): Ditto.
42619         (*insvqi_2): Ditto.
42620         (*insvqi_3): Rename from *insvqi.
42622         (*extzvqi_mem_rex64): Add UNSPEC_NOREX_MEM tag.
42624 2017-01-02  Gerald Pfeifer  <gerald@pfeifer.com>
42626         * doc/cfg.texi (Edges): Remove reference to Java.
42627         (Maintaining the CFG): Ditto.
42629 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
42631         PR middle-end/77674
42632         * symtab.c (symtab_node::binds_to_current_def_p): Fix handling of
42633         transparent aliases.
42635 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
42637         PR middle-end/77484
42638         * predict.def (PRED_CALL): Update hitrate.
42639         (PRED_INDIR_CALL, PRED_POLYMORPHIC_CALL): New predictors.
42640         * predict.c (tree_estimate_probability_bb): Split CALL predictor
42641         into direct/indirect/polymorphic variants.
42643 2017-01-01  Jakub Jelinek  <jakub@redhat.com>
42645         Update copyright years.
42647         * gcc.c (process_command): Update copyright notice dates.
42648         * gcov-dump.c (print_version): Ditto.
42649         * gcov.c (print_version): Ditto.
42650         * gcov-tool.c (print_version): Ditto.
42651         * gengtype.c (create_file): Ditto.
42652         * doc/cpp.texi: Bump @copying's copyright year.
42653         * doc/cppinternals.texi: Ditto.
42654         * doc/gcc.texi: Ditto.
42655         * doc/gccint.texi: Ditto.
42656         * doc/gcov.texi: Ditto.
42657         * doc/install.texi: Ditto.
42658         * doc/invoke.texi: Ditto.
42660 Copyright (C) 2017 Free Software Foundation, Inc.
42662 Copying and distribution of this file, with or without modification,
42663 are permitted in any medium without royalty provided the copyright
42664 notice and this notice are preserved.