poly_int: PUSH_ROUNDING
[official-gcc.git] / gcc / ChangeLog
blobd9428fe63c6a0f8425467db27587ef7d97ec7c48
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         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
6         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
7         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
8         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
9         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
10         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
11         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
13         * config/i386/i386.c (ix86_push_rounding): ...this new function.
14         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
15         a poly_int64.
16         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
17         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
18         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
19         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
20         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
21         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
22         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
23         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
24         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
25         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
26         function.
27         * expr.c (emit_move_resolve_push): Treat the input and result
28         of PUSH_ROUNDING as a poly_int64.
29         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
30         (emit_push_insn): Likewise.
31         * lra-eliminations.c (mark_not_eliminable): Likewise.
32         * recog.c (push_operand): Likewise.
33         * reload1.c (elimination_effects): Likewise.
34         * rtlanal.c (nonzero_bits1): Likewise.
35         * calls.c (store_one_arg): Likewise.  Require the padding to be
36         known at compile time.
38 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
39             Alan Hayward  <alan.hayward@arm.com>
40             David Sherwood  <david.sherwood@arm.com>
42         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
43         Use plus_constant instead of gen_rtx_PLUS.
45 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
46             Alan Hayward  <alan.hayward@arm.com>
47             David Sherwood  <david.sherwood@arm.com>
49         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
50         rather than an int.
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 (expand_expr_real_1): Use tree_to_poly_uint64
57         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
58         via stack temporaries.  Treat the mode size as polynomial too.
60 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
61             Alan Hayward  <alan.hayward@arm.com>
62             David Sherwood  <david.sherwood@arm.com>
64         * expr.c (expand_expr_real_2): When handling conversions involving
65         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
66         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
67         as a poly_uint64 too.
69 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
70             Alan Hayward  <alan.hayward@arm.com>
71             David Sherwood  <david.sherwood@arm.com>
73         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
75 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
76             Alan Hayward  <alan.hayward@arm.com>
77             David Sherwood  <david.sherwood@arm.com>
79         * combine.c (can_change_dest_mode): Handle polynomial
80         REGMODE_NATURAL_SIZE.
81         * expmed.c (store_bit_field_1): Likewise.
82         * expr.c (store_constructor): Likewise.
83         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
84         and polynomial REGMODE_NATURAL_SIZE.
85         (gen_lowpart_common): Likewise.
86         * reginfo.c (record_subregs_of_mode): Likewise.
87         * rtlanal.c (read_modify_subreg_p): Likewise.
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         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
94         numbers of elements.
96 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
97             Alan Hayward  <alan.hayward@arm.com>
98             David Sherwood  <david.sherwood@arm.com>
100         * match.pd: Cope with polynomial numbers of vector elements.
102 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
103             Alan Hayward  <alan.hayward@arm.com>
104             David Sherwood  <david.sherwood@arm.com>
106         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
107         in a POINTER_PLUS_EXPR.
109 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
110             Alan Hayward  <alan.hayward@arm.com>
111             David Sherwood  <david.sherwood@arm.com>
113         * omp-simd-clone.c (simd_clone_subparts): New function.
114         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
115         (ipa_simd_modify_function_body): Likewise.
117 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
118             Alan Hayward  <alan.hayward@arm.com>
119             David Sherwood  <david.sherwood@arm.com>
121         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
122         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
123         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
124         (expand_vector_condition, vector_element): Likewise.
125         (subparts_gt): New function.
126         (get_compute_type): Use subparts_gt.
127         (count_type_subparts): Delete.
128         (expand_vector_operations_1): Use subparts_gt instead of
129         count_type_subparts.
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-data-refs.c (vect_no_alias_p): Replace with...
136         (vect_compile_time_alias): ...this new function.  Do the calculation
137         on poly_ints rather than trees.
138         (vect_prune_runtime_alias_test_list): Update call accordingly.
140 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
141             Alan Hayward  <alan.hayward@arm.com>
142             David Sherwood  <david.sherwood@arm.com>
144         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
145         numbers of units.
146         (vect_schedule_slp_instance): Likewise.
148 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
149             Alan Hayward  <alan.hayward@arm.com>
150             David Sherwood  <david.sherwood@arm.com>
152         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
153         constant and extern definitions for variable-length vectors.
154         (vect_get_constant_vectors): Note that the number of units
155         is known to be constant.
157 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
158             Alan Hayward  <alan.hayward@arm.com>
159             David Sherwood  <david.sherwood@arm.com>
161         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
162         of units as polynomial.  Choose between WIDE and NARROW based
163         on multiple_p.
165 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
166             Alan Hayward  <alan.hayward@arm.com>
167             David Sherwood  <david.sherwood@arm.com>
169         * tree-vect-stmts.c (simd_clone_subparts): New function.
170         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
172 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
173             Alan Hayward  <alan.hayward@arm.com>
174             David Sherwood  <david.sherwood@arm.com>
176         * tree-vect-stmts.c (vectorizable_call): Treat the number of
177         vectors as polynomial.  Use build_index_vector for
178         IFN_GOMP_SIMD_LANE.
180 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
181             Alan Hayward  <alan.hayward@arm.com>
182             David Sherwood  <david.sherwood@arm.com>
184         * tree-vect-stmts.c (get_load_store_type): Treat the number of
185         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
186         for variable-length vectors.
187         (vectorizable_mask_load_store): Treat the number of units as
188         polynomial, asserting that it is constant if the condition has
189         already been enforced.
190         (vectorizable_store, vectorizable_load): Likewise.
192 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
193             Alan Hayward  <alan.hayward@arm.com>
194             David Sherwood  <david.sherwood@arm.com>
196         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
197         of units as polynomial.  Punt if we can't tell at compile time
198         which vector contains the final result.
200 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
201             Alan Hayward  <alan.hayward@arm.com>
202             David Sherwood  <david.sherwood@arm.com>
204         * tree-vect-loop.c (vectorizable_induction): Treat the number
205         of units as polynomial.  Punt on SLP inductions.  Use an integer
206         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
207         cast of such a series for variable-length floating-point
208         reductions.
210 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
211             Alan Hayward  <alan.hayward@arm.com>
212             David Sherwood  <david.sherwood@arm.com>
214         * tree.h (build_index_vector): Declare.
215         * tree.c (build_index_vector): New function.
216         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
217         of units as polynomial, forcibly converting it to a constant if
218         vectorizable_reduction has already enforced the condition.
219         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
220         to create a {1,2,3,...} vector.
221         (vectorizable_reduction): Treat the number of units as polynomial.
222         Choose vectype_in based on the largest scalar element size rather
223         than the smallest number of units.  Enforce the restrictions
224         relied on above.
226 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
227             Alan Hayward  <alan.hayward@arm.com>
228             David Sherwood  <david.sherwood@arm.com>
230         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
231         number of units as polynomial.
233 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
234             Alan Hayward  <alan.hayward@arm.com>
235             David Sherwood  <david.sherwood@arm.com>
237         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
238         * target.def (autovectorize_vector_sizes): Return the vector sizes
239         by pointer, using vector_sizes rather than a bitmask.
240         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
241         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
242         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
243         Likewise.
244         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
245         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
246         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
247         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
248         * omp-general.c (omp_max_vf): Likewise.
249         * omp-low.c (omp_clause_aligned_alignment): Likewise.
250         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
251         * tree-vect-loop.c (vect_analyze_loop): Likewise.
252         * tree-vect-slp.c (vect_slp_bb): Likewise.
253         * doc/tm.texi: Regenerate.
254         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
255         to a poly_uint64.
256         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
257         the vector size as a poly_uint64 rather than an unsigned int.
258         (current_vector_size): Change from an unsigned int to a poly_uint64.
259         (get_vectype_for_scalar_type): Update accordingly.
260         * tree.h (build_truth_vector_type): Take the size and number of
261         units as a poly_uint64 rather than an unsigned int.
262         (build_vector_type): Add a temporary overload that takes
263         the number of units as a poly_uint64 rather than an unsigned int.
264         * tree.c (make_vector_type): Likewise.
265         (build_truth_vector_type): Take the number of units as a poly_uint64
266         rather than an unsigned int.
268 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
269             Alan Hayward  <alan.hayward@arm.com>
270             David Sherwood  <david.sherwood@arm.com>
272         * target.def (get_mask_mode): Take the number of units and length
273         as poly_uint64s rather than unsigned ints.
274         * targhooks.h (default_get_mask_mode): Update accordingly.
275         * targhooks.c (default_get_mask_mode): Likewise.
276         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
277         * doc/tm.texi: Regenerate.
279 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
280             Alan Hayward  <alan.hayward@arm.com>
281             David Sherwood  <david.sherwood@arm.com>
283         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
284         * omp-general.c (omp_max_vf): Likewise.
285         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
286         (expand_omp_simd): Handle polynomial safelen.
287         * omp-low.c (omplow_simd_context): Add a default constructor.
288         (omplow_simd_context::max_vf): Change from int to poly_uint64.
289         (lower_rec_simd_input_clauses): Update accordingly.
290         (lower_rec_input_clauses): Likewise.
292 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
293             Alan Hayward  <alan.hayward@arm.com>
294             David Sherwood  <david.sherwood@arm.com>
296         * tree-vectorizer.h (vect_nunits_for_cost): New function.
297         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
298         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
299         (vect_analyze_slp_cost): Likewise.
300         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
301         (vect_model_load_cost): Likewise.
303 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
304             Alan Hayward  <alan.hayward@arm.com>
305             David Sherwood  <david.sherwood@arm.com>
307         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
308         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
309         from an unsigned int * to a poly_uint64_pod *.
310         (calculate_unrolling_factor): New function.
311         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
313 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
314             Alan Hayward  <alan.hayward@arm.com>
315             David Sherwood  <david.sherwood@arm.com>
317         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
318         from an unsigned int to a poly_uint64.
319         (_loop_vec_info::slp_unrolling_factor): Likewise.
320         (_loop_vec_info::vectorization_factor): Change from an int
321         to a poly_uint64.
322         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
323         (vect_get_num_vectors): New function.
324         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
325         (vect_get_num_copies): Use vect_get_num_vectors.
326         (vect_analyze_data_ref_dependences): Change max_vf from an int *
327         to an unsigned int *.
328         (vect_analyze_data_refs): Change min_vf from an int * to a
329         poly_uint64 *.
330         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
331         than an unsigned HOST_WIDE_INT.
332         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
333         (vect_analyze_data_ref_dependence): Change max_vf from an int *
334         to an unsigned int *.
335         (vect_analyze_data_ref_dependences): Likewise.
336         (vect_compute_data_ref_alignment): Handle polynomial vf.
337         (vect_enhance_data_refs_alignment): Likewise.
338         (vect_prune_runtime_alias_test_list): Likewise.
339         (vect_shift_permute_load_chain): Likewise.
340         (vect_supportable_dr_alignment): Likewise.
341         (dependence_distance_ge_vf): Take the vectorization factor as a
342         poly_uint64 rather than an unsigned HOST_WIDE_INT.
343         (vect_analyze_data_refs): Change min_vf from an int * to a
344         poly_uint64 *.
345         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
346         vfm1 as a poly_uint64 rather than an int.  Make the same change
347         for the returned bound_scalar.
348         (vect_gen_vector_loop_niters): Handle polynomial vf.
349         (vect_do_peeling): Likewise.  Update call to
350         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
351         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
352         be constant.
353         * tree-vect-loop.c (vect_determine_vectorization_factor)
354         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
355         (vect_get_known_peeling_cost): Likewise.
356         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
357         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
358         (vect_transform_loop): Likewise.  Use the lowest possible VF when
359         updating the upper bounds of the loop.
360         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
361         rather than an int.
362         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
363         polynomial unroll factors.
364         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
365         (vect_make_slp_decision): Likewise.
366         (vect_supported_load_permutation_p): Likewise, and polynomial
367         vf too.
368         (vect_analyze_slp_cost): Handle polynomial vf.
369         (vect_slp_analyze_node_operations): Likewise.
370         (vect_slp_analyze_bb_1): Likewise.
371         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
372         than an unsigned HOST_WIDE_INT.
373         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
374         (vectorizable_load): Handle polynomial vf.
375         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
376         a poly_uint64.
377         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
379 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
380             Alan Hayward  <alan.hayward@arm.com>
381             David Sherwood  <david.sherwood@arm.com>
383         * match.pd: Handle bit operations involving three constants
384         and try to fold one pair.
386 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
388         * tree-vect-loop-manip.c: Include gimple-fold.h.
389         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
390         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
391         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
392         Add a path that uses a step of VF instead of 1, but disable it
393         for now.
394         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
395         and niters_no_overflow parameters.  Update calls to
396         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
397         Create a new SSA name if the latter choses to use a ste other
398         than zero, and return it via niters_vector_mult_vf_var.
399         * tree-vect-loop.c (vect_transform_loop): Update calls to
400         vect_do_peeling, vect_gen_vector_loop_niters and
401         slpeel_make_loop_iterate_ntimes.
402         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
403         (vect_gen_vector_loop_niters): Update declarations after above changes.
405 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
407         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
408         128-bit round to integer instructions.
409         (ceil<mode>2): Likewise.
410         (btrunc<mode>2): Likewise.
411         (round<mode>2): Likewise.
413 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
415         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
416         unaligned VSX load/store on P8/P9.
417         (expand_block_clear): Allow the use of unaligned VSX
418         load/store on P8/P9.
420 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
422         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
423         New function.
424         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
425         swap associated with both a load and a store.
427 2018-01-02  Andrew Waterman  <andrew@sifive.com>
429         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
430         * config/riscv/riscv.md (clear_cache): Use it.
432 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
434         * web.c: Remove out-of-date comment.
436 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
438         * expr.c (fixup_args_size_notes): Check that any existing
439         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
440         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
441         (emit_single_push_insn): ...here.
443 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
445         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
446         (const_vector_encoded_nelts): New function.
447         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
448         (const_vector_int_elt, const_vector_elt): Declare.
449         * emit-rtl.c (const_vector_int_elt_1): New function.
450         (const_vector_elt): Likewise.
451         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
452         of CONST_VECTOR_ELT.
454 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
456         * expr.c: Include rtx-vector-builder.h.
457         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
458         directly on the tree encoding.
459         (const_vector_from_tree): Likewise.
460         * optabs.c: Include rtx-vector-builder.h.
461         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
462         sequence of "u" values.
463         * vec-perm-indices.c: Include rtx-vector-builder.h.
464         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
465         directly on the vec_perm_indices encoding.
467 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
469         * doc/rtl.texi (const_vector): Describe new encoding scheme.
470         * Makefile.in (OBJS): Add rtx-vector-builder.o.
471         * rtx-vector-builder.h: New file.
472         * rtx-vector-builder.c: Likewise.
473         * rtl.h (rtx_def::u2): Add a const_vector field.
474         (CONST_VECTOR_NPATTERNS): New macro.
475         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
476         (CONST_VECTOR_DUPLICATE_P): Likewise.
477         (CONST_VECTOR_STEPPED_P): Likewise.
478         (CONST_VECTOR_ENCODED_ELT): Likewise.
479         (const_vec_duplicate_p): Check for a duplicated vector encoding.
480         (unwrap_const_vec_duplicate): Likewise.
481         (const_vec_series_p): Check for a non-duplicated vector encoding.
482         Say that the function only returns true for integer vectors.
483         * emit-rtl.c: Include rtx-vector-builder.h.
484         (gen_const_vec_duplicate_1): Delete.
485         (gen_const_vector): Call gen_const_vec_duplicate instead of
486         gen_const_vec_duplicate_1.
487         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
488         (gen_const_vec_duplicate): Use rtx_vector_builder.
489         (gen_const_vec_series): Likewise.
490         (gen_rtx_CONST_VECTOR): Likewise.
491         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
492         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
493         Build a new vector rather than modifying a CONST_VECTOR in-place.
494         (handle_special_swappables): Update call accordingly.
495         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
496         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
497         Build a new vector rather than modifying a CONST_VECTOR in-place.
498         (handle_special_swappables): Update call accordingly.
500 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
502         * simplify-rtx.c (simplify_const_binary_operation): Use
503         CONST_VECTOR_ELT instead of XVECEXP.
505 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
507         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
508         the selector elements to be different from the data elements
509         if the selector is a VECTOR_CST.
510         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
511         ssizetype for the selector.
513 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
515         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
516         before testing each element individually.
517         * tree-vect-generic.c (lower_vec_perm): Likewise.
519 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
521         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
522         * selftest-run-tests.c (selftest::run_tests): Call it.
523         * vector-builder.h (vector_builder::operator ==): New function.
524         (vector_builder::operator !=): Likewise.
525         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
526         (vec_perm_indices::all_from_input_p): New function.
527         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
528         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
529         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
530         instead of reading the VECTOR_CST directly.  Detect whether both
531         vector inputs are the same before constructing the vec_perm_indices,
532         and update the number of inputs argument accordingly.  Use the
533         utility functions added above.  Only construct sel2 if we need to.
535 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
537         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
538         the broadcast of the low byte.
539         (expand_mult_highpart): Use an explicit encoding for the permutes.
540         * optabs-query.c (can_mult_highpart_p): Likewise.
541         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
542         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
543         (vectorizable_bswap): Likewise.
544         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
545         explicit encoding for the power-of-2 permutes.
546         (vect_permute_store_chain): Likewise.
547         (vect_grouped_load_supported): Likewise.
548         (vect_permute_load_chain): Likewise.
550 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
552         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
553         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
554         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
555         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
556         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
557         (vect_gen_perm_mask_any): Likewise.
559 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
561         * int-vector-builder.h: New file.
562         * vec-perm-indices.h: Include int-vector-builder.h.
563         (vec_perm_indices): Redefine as an int_vector_builder.
564         (auto_vec_perm_indices): Delete.
565         (vec_perm_builder): Redefine as a stand-alone class.
566         (vec_perm_indices::vec_perm_indices): New function.
567         (vec_perm_indices::clamp): Likewise.
568         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
569         (vec_perm_indices::new_vector): New function.
570         (vec_perm_indices::new_expanded_vector): Update for new
571         vec_perm_indices class.
572         (vec_perm_indices::rotate_inputs): New function.
573         (vec_perm_indices::all_in_range_p): Operate directly on the
574         encoded form, without computing elided elements.
575         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
576         encoding.  Update for new vec_perm_indices class.
577         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
578         the given vec_perm_builder.
579         (expand_vec_perm_var): Update vec_perm_builder constructor.
580         (expand_mult_highpart): Use vec_perm_builder instead of
581         auto_vec_perm_indices.
582         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
583         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
584         or double series encoding as appropriate.
585         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
586         vec_perm_indices instead of auto_vec_perm_indices.
587         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
588         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
589         (vect_permute_store_chain): Likewise.
590         (vect_grouped_load_supported): Likewise.
591         (vect_permute_load_chain): Likewise.
592         (vect_shift_permute_load_chain): Likewise.
593         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
594         (vect_transform_slp_perm_load): Likewise.
595         (vect_schedule_slp_instance): Likewise.
596         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
597         (vectorizable_mask_load_store): Likewise.
598         (vectorizable_bswap): Likewise.
599         (vectorizable_store): Likewise.
600         (vectorizable_load): Likewise.
601         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
602         vec_perm_indices instead of auto_vec_perm_indices.  Use
603         tree_to_vec_perm_builder to read the vector from a tree.
604         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
605         vec_perm_builder instead of a vec_perm_indices.
606         (have_whole_vector_shift): Use vec_perm_builder and
607         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
608         truncation to calc_vec_perm_mask_for_shift.
609         (vect_create_epilog_for_reduction): Likewise.
610         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
611         from auto_vec_perm_indices to vec_perm_indices.
612         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
613         instead of changing individual elements.
614         (aarch64_vectorize_vec_perm_const): Use new_vector to install
615         the vector in d.perm.
616         * config/arm/arm.c (expand_vec_perm_d::perm): Change
617         from auto_vec_perm_indices to vec_perm_indices.
618         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
619         instead of changing individual elements.
620         (arm_vectorize_vec_perm_const): Use new_vector to install
621         the vector in d.perm.
622         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
623         Update vec_perm_builder constructor.
624         (rs6000_expand_interleave): Likewise.
625         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
626         (rs6000_expand_interleave): Likewise.
628 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
630         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
631         to qimode could truncate the indices.
632         * optabs.c (expand_vec_perm_var): Likewise.
634 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
636         * Makefile.in (OBJS): Add vec-perm-indices.o.
637         * vec-perm-indices.h: New file.
638         * vec-perm-indices.c: Likewise.
639         * target.h (vec_perm_indices): Replace with a forward class
640         declaration.
641         (auto_vec_perm_indices): Move to vec-perm-indices.h.
642         * optabs.h: Include vec-perm-indices.h.
643         (expand_vec_perm): Delete.
644         (selector_fits_mode_p, expand_vec_perm_var): Declare.
645         (expand_vec_perm_const): Declare.
646         * target.def (vec_perm_const_ok): Replace with...
647         (vec_perm_const): ...this new hook.
648         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
649         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
650         * doc/tm.texi: Regenerate.
651         * optabs.def (vec_perm_const): Delete.
652         * doc/md.texi (vec_perm_const): Likewise.
653         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
654         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
655         expand_vec_perm for constant permutation vectors.  Assert that
656         the mode of variable permutation vectors is the integer equivalent
657         of the mode that is being permuted.
658         * optabs-query.h (selector_fits_mode_p): Declare.
659         * optabs-query.c: Include vec-perm-indices.h.
660         (selector_fits_mode_p): New function.
661         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
662         is defined, instead of checking whether the vec_perm_const_optab
663         exists.  Use targetm.vectorize.vec_perm_const instead of
664         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
665         fit in the vector mode before using a variable permute.
666         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
667         vec_perm_indices instead of an rtx.
668         (expand_vec_perm): Replace with...
669         (expand_vec_perm_const): ...this new function.  Take the selector
670         as a vec_perm_indices rather than an rtx.  Also take the mode of
671         the selector.  Update call to shift_amt_for_vec_perm_mask.
672         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
673         Use vec_perm_indices::new_expanded_vector to expand the original
674         selector into bytes.  Check whether the indices fit in the vector
675         mode before using a variable permute.
676         (expand_vec_perm_var): Make global.
677         (expand_mult_highpart): Use expand_vec_perm_const.
678         * fold-const.c: Includes vec-perm-indices.h.
679         * tree-ssa-forwprop.c: Likewise.
680         * tree-vect-data-refs.c: Likewise.
681         * tree-vect-generic.c: Likewise.
682         * tree-vect-loop.c: Likewise.
683         * tree-vect-slp.c: Likewise.
684         * tree-vect-stmts.c: Likewise.
685         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
686         Delete.
687         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
688         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
689         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
690         (aarch64_vectorize_vec_perm_const): ...this new function.
691         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
692         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
693         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
694         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
695         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
696         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
697         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
698         into...
699         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
700         check for NEON modes.
701         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
702         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
703         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
704         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
705         into...
706         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
707         the old VEC_PERM_CONST conditions.
708         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
709         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
710         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
711         (ia64_vectorize_vec_perm_const_ok): Merge into...
712         (ia64_vectorize_vec_perm_const): ...this new function.
713         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
714         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
715         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
716         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
717         * config/mips/mips.c (mips_expand_vec_perm_const)
718         (mips_vectorize_vec_perm_const_ok): Merge into...
719         (mips_vectorize_vec_perm_const): ...this new function.
720         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
721         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
722         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
723         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
724         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
725         (rs6000_expand_vec_perm_const): Delete.
726         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
727         Delete.
728         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
729         (altivec_expand_vec_perm_const_le): Take each operand individually.
730         Operate on constant selectors rather than rtxes.
731         (altivec_expand_vec_perm_const): Likewise.  Update call to
732         altivec_expand_vec_perm_const_le.
733         (rs6000_expand_vec_perm_const): Delete.
734         (rs6000_vectorize_vec_perm_const_ok): Delete.
735         (rs6000_vectorize_vec_perm_const): New function.
736         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
737         an element count and rtx array.
738         (rs6000_expand_extract_even): Update call accordingly.
739         (rs6000_expand_interleave): Likewise.
740         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
741         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
742         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
743         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
744         (rs6000_expand_vec_perm_const): Delete.
745         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
746         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
747         (altivec_expand_vec_perm_const_le): Take each operand individually.
748         Operate on constant selectors rather than rtxes.
749         (altivec_expand_vec_perm_const): Likewise.  Update call to
750         altivec_expand_vec_perm_const_le.
751         (rs6000_expand_vec_perm_const): Delete.
752         (rs6000_vectorize_vec_perm_const_ok): Delete.
753         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
754         reference to the SPE evmerge intructions.
755         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
756         an element count and rtx array.
757         (rs6000_expand_extract_even): Update call accordingly.
758         (rs6000_expand_interleave): Likewise.
759         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
760         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
761         new function.
762         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
764 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
766         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
767         vector mode and that that mode matches the mode of the data
768         being permuted.
769         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
770         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
771         directly using expand_vec_perm_1 when forcing selectors into
772         registers.
773         (expand_vec_perm_var): New function, split out from expand_vec_perm.
775 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
777         * optabs-query.h (can_vec_perm_p): Delete.
778         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
779         * optabs-query.c (can_vec_perm_p): Split into...
780         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
781         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
782         particular selector is valid.
783         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
784         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
785         (vect_grouped_load_supported): Likewise.
786         (vect_shift_permute_load_chain): Likewise.
787         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
788         (vect_transform_slp_perm_load): Likewise.
789         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
790         (vectorizable_bswap): Likewise.
791         (vect_gen_perm_mask_checked): Likewise.
792         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
793         implementations of variable permutation vectors into account
794         when deciding which selector to use.
795         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
796         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
797         with a false third argument.
798         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
799         to test whether the constant selector is valid and can_vec_perm_var_p
800         to test whether a variable selector is valid.
802 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
804         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
805         * optabs-query.c (can_vec_perm_p): Likewise.
806         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
807         instead of vec_perm_indices.
808         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
809         (vect_gen_perm_mask_checked): Likewise,
810         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
811         (vect_gen_perm_mask_checked): Likewise,
813 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
815         * optabs-query.h (qimode_for_vec_perm): Declare.
816         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
817         (qimode_for_vec_perm): ...this new function.
818         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
820 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
822         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
823         does not have a conditional at the top.
825 2018-01-02  Richard Biener  <rguenther@suse.de>
827         * ipa-inline.c (big_speedup_p): Fix expression.
829 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
831         PR target/81616
832         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
833         for generic 4->6.
835 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
837         PR target/81616
838         Generic tuning.
839         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
840         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
841         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
842         cond_taken_branch_cost 3->4.
844 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
846         PR tree-optimization/83581
847         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
848         TODO_cleanup_cfg if any changes have been made.
850         PR middle-end/83608
851         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
852         convert_modes if target mode has the right side, but different mode
853         class.
855         PR middle-end/83609
856         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
857         last argument when extracting from CONCAT.  If either from_real or
858         from_imag is NULL, use expansion through memory.  If result is not
859         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
860         the parts directly to inner mode, if even that fails, use expansion
861         through memory.
863         PR middle-end/83623
864         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
865         check for bswap in mode rather than HImode and use that in expand_unop
866         too.
868 2017-12-31  Jakub Jelinek  <jakub@redhat.com>
870         PR target/83536
871         * config/i386/i386.c (ix86_attribute_table): Remove excess
872         initializer for "shared" attribute.
874 2017-12-30  Tom de Vries  <tom@codesourcery.com>
876         PR libgomp/83046
877         * omp-expand.c (expand_omp_target): If in_lto_p, mark offload_funcs with
878         DECL_PRESERVE_P.
879         * lto-streamer-out.c (prune_offload_funcs): New function.  Remove
880         offload_funcs entries that no longer have a corresponding cgraph_node.
881         Mark the remaining ones as DECL_PRESERVE_P.
882         (output_lto): Call prune_offload_funcs.
884 2017-12-30  Jakub Jelinek  <jakub@redhat.com>
886         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
887         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>,
888         vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
889         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
890         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
891         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1, vpdpbusd_<mode>,
892         vpdpbusd_<mode>_mask, vpdpbusd_<mode>_maskz, vpdpbusd_<mode>_maskz_1,
893         vpdpbusds_<mode>, vpdpbusds_<mode>_mask, vpdpbusds_<mode>_maskz,
894         vpdpbusds_<mode>_maskz_1, vpdpwssd_<mode>, vpdpwssd_<mode>_mask,
895         vpdpwssd_<mode>_maskz, vpdpwssd_<mode>_maskz_1, vpdpwssds_<mode>,
896         vpdpwssds_<mode>_mask, vpdpwssds_<mode>_maskz,
897         vpdpwssds_<mode>_maskz_1, vaesdec_<mode>, vaesdeclast_<mode>,
898         vaesenc_<mode>, vpclmulqdq_<mode>,
899         avx512vl_vpshufbitqmb<mode><mask_scalar_merge_name>): Formatting fixes.
901 2017-12-28  Michael Meissner  <meissner@linux.vnet.ibm.com>
903         * builtins.def: (_Float<N> and _Float<N>X BUILT_IN_CEIL): Add
904         _Float<N> and _Float<N>X variants for rounding built-in
905         functions.
906         (_Float<N> and _Float<N>X BUILT_IN_FLOOR): Likewise.
907         (_Float<N> and _Float<N>X BUILT_IN_NEARBYINT): Likewise.
908         (_Float<N> and _Float<N>X BUILT_IN_RINT): Likewise.
909         (_Float<N> and _Float<N>X BUILT_IN_ROUND): Likewise.
910         (_Float<N> and _Float<N>X BUILT_IN_TRUNC): Likewise.
911         * builtins.c (mathfn_built_in_2): Likewise.
912         * internal-fn.def (CEIL): Likewise.
913         (FLOOR): Likewise.
914         (NEARBYINT): Likewise.
915         (RINT): Likewise.
916         (ROUND): Likewise.
917         (TRUNC): Likewise.
918         * convert.c (convert_to_integer_1): Likewise.
919         * fold-const.c (tree_call_nonnegative_warnv_p): Likewise.
920         (integer_valued_real_call_p): Likewise.
921         * fold-const-call.c (fold_const_call_ss): Likewise.
922         * gencfn-macros.c (print_case_cfn): Change CFN and operator
923         printers to take a const char * suffix instead of a bool.
924         (print_define_operator_list): Likewise.
925         (fltall_suffixes): New list of suffixes, that include the
926         traditional suffixes as well as all of the _Float<N> and
927         _Float<N>X suffixes.
928         (main): For _Float<N> and _Float<N>X functions, emit both
929         <name>_FN and <name>_ALL variants.  The <macro>_FN variant only
930         has the _Float<N> and _Float<N>X case names or operators.  The
931         <name>_ALL variant has both the traditional and the
932         _Float<N>/_Float<N>X case names or operators.
933         * match.pd (COPYSIGN optimizations): Provide optimizations for
934         _Float<N> and _Float<N>X types where possible.
935         (MIN/MAX optimizations): Likewise.
936         (sqrt optimizations): Likewise.
937         (rounding optimizations): Likewise.
939 2017-12-28  Richard Sandiford  <richard.sandiford@linaro.org>
941         * config/powerpcspe/powerpcspe.md (ieee_128bit_negative_zero): Use
942         gen_int_mode rather than GEN_INT.
943         * config/rs6000/rs6000.md (ieee_128bit_negative_zero): Likewise.
945 2017-12-28  Richard Sandiford  <richard.sandiford@linaro.org>
947         * emit-rtl.c (gen_const_vec_series): Use valid_for_const_vector_p
948         instead of CONSTANT_P.
949         (gen_vec_series): Likewise.
950         * simplify-rtx.c (simplify_binary_operation_1): Likewise.
952 2017-12-28  Andreas Schwab  <schwab@linux-m68k.org>
954         * config/m68k/m68k.md (ashrdi3_const1, lshrdi3_const1): Add
955         CC_STATUS_INIT.
957 2017-12-27  Kugan Vivekanandarajah  <kuganv@linaro.org>
959         * config/aarch64/aarch64-simd.md (aarch64_ld1x2<VQ:mode>): New.
960         (aarch64_ld1x2<VDC:mode>): Likewise.
961         (aarch64_simd_ld1<mode>_x2): Likewise.
962         (aarch64_simd_ld1<mode>_x2): Likewise.
963         * config/aarch64/arm_neon.h (vld1_u8_x2): New.
964         (vld1_s8_x2): Likewise.
965         (vld1_u16_x2): Likewise.
966         (vld1_s16_x2): Likewise.
967         (vld1_u32_x2): Likewise.
968         (vld1_s32_x2): Likewise.
969         (vld1_u64_x2): Likewise.
970         (vld1_s64_x2): Likewise.
971         (vld1_f16_x2): Likewise.
972         (vld1_f32_x2): Likewise.
973         (vld1_f64_x2): Likewise.
974         (vld1_p8_x2): Likewise.
975         (vld1_p16_x2): Likewise.
976         (vld1_p64_x2): Likewise.
977         (vld1q_u8_x2): Likewise.
978         (vld1q_s8_x2): Likewise.
979         (vld1q_u16_x2): Likewise.
980         (vld1q_s16_x2): Likewise.
981         (vld1q_u32_x2): Likewise.
982         (vld1q_s32_x2): Likewise.
983         (vld1q_u64_x2): Likewise.
984         (vld1q_s64_x2): Likewise.
985         (vld1q_f16_x2): Likewise.
986         (vld1q_f32_x2): Likewise.
987         (vld1q_f64_x2): Likewise.
988         (vld1q_p8_x2): Likewise.
989         (vld1q_p16_x2): Likewise.
990         (vld1q_p64_x2): Likewise.
992 2017-12-27  Martin Liska  <mliska@suse.cz>
994         PR tree-optimization/83552
995         * tree-ssa-strlen.c (fold_strstr_to_strncmp): Assign result
996         of get_string_lenth to a SSA_NAME if not a GIMPLE value.
998 2017-12-27  Tom de Vries  <tom@codesourcery.com>
1000         * config/nvptx/nvptx.c (nvptx_option_override): Disable
1001         -gstatement-frontiers.
1003 2017-12-26  Alexander Monakov  <amonakov@ispras.ru>
1005         PR rtl-optimization/83513
1006         * sel-sched.c (sel_rank_for_schedule): Order by non-zero usefulness
1007         before priority comparison.
1009 2017-12-25  Jakub Jelinek  <jakub@redhat.com>
1011         PR target/83488
1012         * config/i386/i386.opt (-mavx512vpopcntdq, -mmavx512bitalg): Move from
1013         ix86_isa_flags2 to ix86_isa_flags.
1014         * config/i386/i386-c.c (ix86_target_macros_internal): Test
1015         OPTION_MASK_ISA_AVX512BITALG and OPTION_MASK_ISA_AVX512VPOPCNTDQ in
1016         isa_flags rather than isa_flags2.
1017         * config/i386/i386.c (ix86_target_string): Move -mavx512vpopcntdq
1018         and -mavx512bitalg from isa2_opts to isa_opts.
1019         (ix86_option_override_internal): Test OPTION_MASK_ISA_AVX512VPOPCNTDQ
1020         in x_ix86_isa_flags_explicit rather than x_ix86_isa_flags2_explicit
1021         and set it in x_ix86_isa_flags rather than x_ix86_isa_flags2.
1022         Formatting fixes.
1023         (def_builtin): Treat OPTION_MASK_ISA_AVX512BW or
1024         OPTION_MASK_ISA_AVX512F ored with another option similarly to
1025         OPTION_MASK_ISA_AVX512VL.  Even for OPTION_MASK_ISA_AVX512VL don't
1026         clear it if mask is just OPTION_MASK_ISA_AVX512VL itself.
1027         (ix86_expand_builtin): Don't handle OPTION_MASK_ISA_GFNI and
1028         OPTION_MASK_ISA_VPCLMULQDQ specially, instead handle
1029         OPTION_MASK_ISA_AVX512BW and OPTION_MASK_ISA_AVX512F that way.
1030         * config/i386/i386-builtin.def: Move AVX512VPOPCNTDQ and AVX512BITALG
1031         builtins from bdesc_args2 to bdesc_args section.
1032         (__builtin_ia32_compressstoreuqi512_mask,
1033         __builtin_ia32_compressstoreuhi512_mask,
1034         __builtin_ia32_compressstoreuqi256_mask,
1035         __builtin_ia32_expandloadqi512_mask,
1036         __builtin_ia32_expandloadqi512_maskz,
1037         __builtin_ia32_expandloadhi512_mask,
1038         __builtin_ia32_expandloadhi512_maskz,
1039         __builtin_ia32_compressqi512_mask, __builtin_ia32_compresshi512_mask,
1040         __builtin_ia32_compressqi256_mask, __builtin_ia32_expandqi512_mask,
1041         __builtin_ia32_expandqi512_maskz, __builtin_ia32_expandhi512_mask,
1042         __builtin_ia32_expandhi512_maskz, __builtin_ia32_expandqi256_mask,
1043         __builtin_ia32_expandqi256_maskz, __builtin_ia32_vpshrd_v32hi_mask,
1044         __builtin_ia32_vpshld_v32hi_mask, __builtin_ia32_vpshrdv_v32hi_mask,
1045         __builtin_ia32_vpshrdv_v32hi_maskz, __builtin_ia32_vpshldv_v32hi_mask,
1046         __builtin_ia32_vpshldv_v32hi_maskz,
1047         __builtin_ia32_vpopcountb_v64qi_mask,
1048         __builtin_ia32_vpopcountw_v32hi_mask,
1049         __builtin_ia32_vpshufbitqmb512_mask,
1050         __builtin_ia32_vpshufbitqmb256_mask): Add
1051         " | OPTION_MASK_ISA_AVX512BW".
1052         (__builtin_ia32_expandloadqi256_mask,
1053         __builtin_ia32_expandloadqi256_maskz,
1054         __builtin_ia32_vpopcountb_v32qi_mask): Add
1055         " | OPTION_MASK_ISA_AVX512VL | OPTION_MASK_ISA_AVX512BW".
1056         (__builtin_ia32_expandloadhi256_mask,
1057         __builtin_ia32_expandloadhi256_maskz,
1058         __builtin_ia32_expandloadqi128_mask,
1059         __builtin_ia32_expandloadqi128_maskz,
1060         __builtin_ia32_expandloadhi128_mask,
1061         __builtin_ia32_expandloadhi128_maskz,
1062         __builtin_ia32_vpshrd_v16hi, __builtin_ia32_vpshrd_v16hi_mask,
1063         __builtin_ia32_vpshrd_v8hi, __builtin_ia32_vpshrd_v8hi_mask,
1064         __builtin_ia32_vpshrd_v8si, __builtin_ia32_vpshrd_v8si_mask,
1065         __builtin_ia32_vpshrd_v4si, __builtin_ia32_vpshrd_v4si_mask,
1066         __builtin_ia32_vpshrd_v4di, __builtin_ia32_vpshrd_v4di_mask,
1067         __builtin_ia32_vpshrd_v2di, __builtin_ia32_vpshrd_v2di_mask,
1068         __builtin_ia32_vpshld_v16hi, __builtin_ia32_vpshld_v16hi_mask,
1069         __builtin_ia32_vpshld_v8hi, __builtin_ia32_vpshld_v8hi_mask,
1070         __builtin_ia32_vpshld_v8si, __builtin_ia32_vpshld_v8si_mask,
1071         __builtin_ia32_vpshld_v4si, __builtin_ia32_vpshld_v4si_mask,
1072         __builtin_ia32_vpshld_v4di, __builtin_ia32_vpshld_v4di_mask,
1073         __builtin_ia32_vpshld_v2di, __builtin_ia32_vpshld_v2di_mask,
1074         __builtin_ia32_vpshrdv_v16hi, __builtin_ia32_vpshrdv_v16hi_mask,
1075         __builtin_ia32_vpshrdv_v16hi_maskz, __builtin_ia32_vpshrdv_v8hi,
1076         __builtin_ia32_vpshrdv_v8hi_mask, __builtin_ia32_vpshrdv_v8hi_maskz,
1077         __builtin_ia32_vpshrdv_v8si, __builtin_ia32_vpshrdv_v8si_mask,
1078         __builtin_ia32_vpshrdv_v8si_maskz, __builtin_ia32_vpshrdv_v4si,
1079         __builtin_ia32_vpshrdv_v4si_mask, __builtin_ia32_vpshrdv_v4si_maskz,
1080         __builtin_ia32_vpshrdv_v4di, __builtin_ia32_vpshrdv_v4di_mask,
1081         __builtin_ia32_vpshrdv_v4di_maskz, __builtin_ia32_vpshrdv_v2di,
1082         __builtin_ia32_vpshrdv_v2di_mask, __builtin_ia32_vpshrdv_v2di_maskz,
1083         __builtin_ia32_vpshldv_v16hi, __builtin_ia32_vpshldv_v16hi_mask,
1084         __builtin_ia32_vpshldv_v16hi_maskz, __builtin_ia32_vpshldv_v8hi,
1085         __builtin_ia32_vpshldv_v8hi_mask, __builtin_ia32_vpshldv_v8hi_maskz,
1086         __builtin_ia32_vpshldv_v8si, __builtin_ia32_vpshldv_v8si_mask,
1087         __builtin_ia32_vpshldv_v8si_maskz, __builtin_ia32_vpshldv_v4si,
1088         __builtin_ia32_vpshldv_v4si_mask, __builtin_ia32_vpshldv_v4si_maskz,
1089         __builtin_ia32_vpshldv_v4di, __builtin_ia32_vpshldv_v4di_mask,
1090         __builtin_ia32_vpshldv_v4di_maskz, __builtin_ia32_vpshldv_v2di,
1091         __builtin_ia32_vpshldv_v2di_mask, __builtin_ia32_vpshldv_v2di_maskz,
1092         __builtin_ia32_vpopcountb_v32qi, __builtin_ia32_vpopcountb_v16qi,
1093         __builtin_ia32_vpopcountb_v16qi_mask, __builtin_ia32_vpopcountw_v16hi,
1094         __builtin_ia32_vpopcountw_v16hi_mask, __builtin_ia32_vpopcountw_v8hi,
1095         __builtin_ia32_vpopcountw_v8hi_mask): Add
1096         " | OPTION_MASK_ISA_AVX512VL".
1097         * config/i386/avx512vbmi2intrin.h (_mm512_shrdi_epi16,
1098         _mm512_shrdi_epi32, _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
1099         _mm512_shrdi_epi64, _mm512_mask_shrdi_epi64, _mm512_maskz_shrdi_epi64,
1100         _mm512_shldi_epi16, _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
1101         _mm512_maskz_shldi_epi32, _mm512_shldi_epi64, _mm512_mask_shldi_epi64,
1102         _mm512_maskz_shldi_epi64, _mm512_shrdv_epi16, _mm512_shrdv_epi32,
1103         _mm512_mask_shrdv_epi32, _mm512_maskz_shrdv_epi32, _mm512_shrdv_epi64,
1104         _mm512_mask_shrdv_epi64, _mm512_maskz_shrdv_epi64, _mm512_shldv_epi16,
1105         _mm512_shldv_epi32, _mm512_mask_shldv_epi32, _mm512_maskz_shldv_epi32,
1106         _mm512_shldv_epi64, _mm512_mask_shldv_epi64,
1107         _mm512_maskz_shldv_epi64): Don't require avx512bw for these intrinsics.
1108         * config/i386/avx512bitalgintrin.h (_mm_bitshuffle_epi64_mask,
1109         _mm_mask_bitshuffle_epi64_mask): Likewise.
1110         * common/config/i386/i386-common.c
1111         (OPTION_MASK_ISA_AVX512VPOPCNTDQ_SET,
1112         OPTION_MASK_ISA_AVX512BITALG_SET): Or in OPTION_MASK_ISA_AVX512F_SET.
1113         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
1114         OPTION_MASK_ISA_AVX512VPOPCNTDQ_UNSET and
1115         OPTION_MASK_ISA_AVX512BITALG_UNSET.
1116         (OPTION_MASK_ISA2_AVX512F_UNSET,
1117         OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Define.
1118         (ix86_handle_option): For -mno-general-regs-only, clear from
1119         ix86_isa_flags2 OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET rather than
1120         just OPTION_MASK_ISA_MPX.  For -mno-sse{,2,3,4,4.1,4.2,avx,avx2} and
1121         -mno-ssse3 clear OPTION_MASK_ISA2_AVX512F_UNSET bits from
1122         ix86_isa_flags2.  For -mno-avx512f likewise, instead of masking
1123         individually listed ISAs.  For -m{,no-}avx512{vpopcntdq,bitalg} adjust
1124         for moving from ix86_isa_flags2 to ix86_isa_flags.
1126 2017-12-23  Jakub Jelinek  <jakub@redhat.com>
1128         PR c++/83553
1129         * fold-const.c (struct contains_label_data): New type.
1130         (contains_label_1): Return non-NULL even for CASE_LABEL_EXPR, unless
1131         inside of a SWITCH_BODY seen during the walk.
1132         (contains_label_p): Use walk_tree instead of
1133         walk_tree_without_duplicates, prepare data for contains_label_1 and
1134         provide own pset.
1136 2017-12-22  Jakub Jelinek  <jakub@redhat.com>
1138         PR debug/83547
1139         * tree-iterator.c (alloc_stmt_list): Start with cleared
1140         TREE_SIDE_EFFECTS regardless whether a new STATEMENT_LIST is allocated
1141         or old one reused.
1143         PR target/83488
1144         * config/i386/avx512vnniintrin.h: Don't check for __AVX512F__ nor
1145         enable avx512f explicitly in #pragma GCC target.
1146         * config/i386/i386-builtin.def (__builtin_ia32_vpdpbusd_v8si,
1147         __builtin_ia32_vpdpbusd_v8si_mask, __builtin_ia32_vpdpbusd_v8si_maskz,
1148         __builtin_ia32_vpdpbusd_v4si, __builtin_ia32_vpdpbusd_v4si_mask,
1149         __builtin_ia32_vpdpbusd_v4si_maskz, __builtin_ia32_vpdpbusds_v8si,
1150         __builtin_ia32_vpdpbusds_v8si_mask,
1151         __builtin_ia32_vpdpbusds_v8si_maskz, __builtin_ia32_vpdpbusds_v4si,
1152         __builtin_ia32_vpdpbusds_v4si_mask,
1153         __builtin_ia32_vpdpbusds_v4si_maskz, __builtin_ia32_vpdpwssd_v8si,
1154         __builtin_ia32_vpdpwssd_v8si_mask, __builtin_ia32_vpdpwssd_v8si_maskz,
1155         __builtin_ia32_vpdpwssd_v4si, __builtin_ia32_vpdpwssd_v4si_mask,
1156         __builtin_ia32_vpdpwssd_v4si_maskz, __builtin_ia32_vpdpwssds_v8si,
1157         __builtin_ia32_vpdpwssds_v8si_mask,
1158         __builtin_ia32_vpdpwssds_v8si_maskz, __builtin_ia32_vpdpwssds_v4si,
1159         __builtin_ia32_vpdpwssds_v4si_mask,
1160         __builtin_ia32_vpdpwssds_v4si_maskz): Use
1161         OPTION_MASK_ISA_AVX512VNNI | OPTION_MASK_ISA_AVX512VL instead of
1162         just OPTION_MASK_ISA_AVX512VNNI.
1164 2017-12-22  Martin Jambor  <mjambor@suse.cz>
1166         PR lto/82027
1167         * lto-cgraph.c (output_cgraph_opt_summary_p): Also check former
1168         clones.
1170 2017-12-22  Julia Koval  <julia.koval@intel.com>
1171             Sebastian Peryt  <sebastian.peryt@intel.com>
1173         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512BITALG_SET,
1174         OPTION_MASK_ISA_AVX512BITALG_UNSET): New.
1175         (ix86_handle_option): Handle -mavx512bitalg, fix 4VNNIW formatting.
1176         * config.gcc: Add avx512vpopcntdqvlintrin.h and avx512bitalgintrin.h.
1177         * config/i386/avx512bitalgintrin.h (_mm512_popcnt_epi8, _mm512_popcnt_epi16,
1178         _mm512_mask_popcnt_epi8, _mm512_maskz_popcnt_epi8, _mm512_mask_popcnt_epi16,
1179         _mm512_maskz_popcnt_epi16, _mm512_bitshuffle_epi64_mask, _mm256_popcnt_epi8,
1180         _mm512_mask_bitshuffle_epi64_mask, _mm256_mask_popcnt_epi8, _mm_popcnt_epi8,
1181         _mm256_maskz_popcnt_epi8, _mm_bitshuffle_epi64_mask, _mm256_popcnt_epi16,
1182         _mm_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
1183         _mm256_mask_bitshuffle_epi64_mask, _mm_popcnt_epi16, _mm_maskz_popcnt_epi8,
1184         _mm256_mask_popcnt_epi16, _mm256_maskz_popcnt_epi16, _mm_mask_popcnt_epi8,
1185         _mm_mask_popcnt_epi16, _mm_maskz_popcnt_epi16): New intrinsics.
1186         * config/i386/avx512vpopcntdqvlintrin.h (_mm_popcnt_epi32, _mm_popcnt_epi64,
1187         _mm_mask_popcnt_epi32, _mm_maskz_popcnt_epi32, _mm256_popcnt_epi32,
1188         _mm256_mask_popcnt_epi32, _mm256_maskz_popcnt_epi32, _mm_mask_popcnt_epi64,
1189         _mm_maskz_popcnt_epi64, _mm256_popcnt_epi64, _mm256_mask_popcnt_epi64,
1190         _mm256_maskz_popcnt_epi64): New intrinsics.
1191         * config/i386/cpuid.h (bit_AVX512BITALG): New bit.
1192         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mavx512bitalg.
1193         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI, V64QI_FTYPE_V64QI,
1194         V4DI_FTYPE_V4DI, UHI_FTYPE_V2DI_V2DI_UHI, USI_FTYPE_V4DI_V4DI_USI,
1195         V4SI_FTYPE_V4SI_V4SI_UHI, V8SI_FTYPE_V8SI_V8SI_UHI): New types.
1196         * config/i386/i386-builtin.def (__builtin_ia32_vpopcountq_v4di,
1197         __builtin_ia32_vpopcountq_v4di_mask, __builtin_ia32_vpopcountq_v2di,
1198         __builtin_ia32_vpopcountq_v2di_mask, __builtin_ia32_vpopcountd_v4si,
1199         __builtin_ia32_vpopcountd_v4si_mask, __builtin_ia32_vpopcountd_v8si,
1200         __builtin_ia32_vpopcountd_v8si_mask, __builtin_ia32_vpopcountb_v64qi,
1201         __builtin_ia32_vpopcountb_v64qi_mask, __builtin_ia32_vpopcountb_v32qi,
1202         __builtin_ia32_vpopcountb_v32qi_mask, __builtin_ia32_vpopcountb_v16qi,
1203         __builtin_ia32_vpopcountb_v16qi_mask, __builtin_ia32_vpopcountw_v32hi,
1204         __builtin_ia32_vpopcountw_v32hi_mask, __builtin_ia32_vpopcountw_v16hi,
1205         __builtin_ia32_vpopcountw_v16hi_mask, __builtin_ia32_vpopcountw_v8hi,
1206         __builtin_ia32_vpopcountw_v8hi_mask, __builtin_ia32_vpshufbitqmb128_mask,
1207         __builtin_ia32_vpshufbitqmb256_mask,
1208         __builtin_ia32_vpshufbitqmb512_mask): New builtins.
1209         * config/i386/i386-c.c (__AVX512BITALG__): New.
1210         * config/i386/i386.c (isa2_opts): Add -mavx512bitalg.
1211         (ix86_valid_target_attribute_inner_p): Ditto.
1212         (ix86_expand_args_builtin): Handle new types.
1213         * config/i386/i386.h (TARGET_AVX512BITALG, TARGET_AVX512BITALG_P): New.
1214         * config/i386/i386.opt: Add -mavx512bitalg.
1215         * config/i386/immintrin.h: Add avx512vpopcntdqvlintrin.h and
1216         avx512bitalgintrin.h.
1217         * config/i386/sse.md (VI48_AVX512VLBW): New iterator.
1218         (vpopcount<mode><mask_name>): Add more types.
1219         (avx512vl_vpshufbitqmb<mode><mask_scalar_merge_name>): New.
1220         * doc/invoke.texi: Add -mavx512bitalg and -mavx512vpopcntdq.
1222 2017-12-22  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
1224         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VNNI_SET):
1225         Or in OPTION_MASK_ISA_AVX512F_SET.
1226         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
1227         OPTION_MASK_ISA_AVX512VNNI_UNSET.
1228         (ix86_handle_option): Adjust for
1229         OPTION_MASK_ISA_AVX512VNNI_*SET being in ix86_isa_flags.
1230         * config/i386/i386-builtin.def: Move VNNI builtins from ARGS2
1231         section to ARGS.
1232         * config/i386/i386-c.c: Check for OPTION_MASK_ISA_AVX512VNNI in
1233         isa_flag instead of isa_flag2.
1234         * config/i386/i386.c (ix86_target_string): Move -mavx512vnni from
1235         isa_opts2 to isa_opts.
1236         * config/i386/i386.opt (mavx512vnni): Move from ix86_isa_flags2
1237         to ix86_isa_flags.
1239 2017-12-22  Mike Stump  <mikestump@comcast.net>
1240             Eric Botcazou  <ebotcazou@adacore.com>
1242         * doc/extend.texi (Loop-Specific Pragmas): Document pragma GCC unroll.
1244 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1246         PR middle-end/83487
1247         * config/i386/i386.c (ix86_function_arg_boundary): Return
1248         PARM_BOUNDARY for TYPE_EMPTY_P types.
1250         PR c/83448
1251         * gimple-ssa-sprintf.c (maybe_warn): Don't call set_caret_index
1252         if navail is >= dir.len.
1254 2017-12-21  Steve Ellcey  <sellcey@cavium.com>
1256         * config/aarch64/t-aarch64-linux (MULTILIB_OSDIRNAMES): Fix
1257         triplet for ilp32.
1259 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1261         PR rtl-optimization/80747
1262         PR rtl-optimization/83512
1263         * cfgrtl.c (force_nonfallthru_and_redirect): When splitting
1264         succ edge from ENTRY, copy partition from e->dest to the newly
1265         created bb.
1266         * bb-reorder.c (reorder_basic_blocks_simple): If last_tail is
1267         ENTRY, use BB_PARTITION of its successor block as current_partition.
1268         Don't copy partition when splitting succ edge from ENTRY.
1270         PR tree-optimization/83523
1271         * tree-ssa-math-opts.c (is_widening_mult_p): Return false if
1272         for INTEGER_TYPE TYPE_OVERFLOW_TRAPS.
1273         (convert_mult_to_fma): Likewise.
1275         PR tree-optimization/83521
1276         * tree-ssa-phiopt.c (factor_out_conditional_conversion): Use
1277         gimple_build_assign without code on result of
1278         fold_build1 (VIEW_CONVERT_EXPR, ...), as it might not create
1279         a VIEW_CONVERT_EXPR.
1281 2017-12-21  Andrew Pinski  <apinski@cavium.com>
1282             Steve Ellcey  <sellcey@cavium.com>
1284         * config/aarch64/t-aarch64-linux (MULTILIB_OSDIRNAMES): Handle
1285         multi-arch for ilp32.
1287 2017-12-21  Uros Bizjak  <ubizjak@gmail.com>
1289         PR target/83467
1290         * config/i386/i386.md (*ashl<mode>3_mask): Add operand
1291         constraints to operand 2.
1292         (*ashl<mode>3_mask_1): Ditto.
1293         (*<shift_insn><mode>3_mask): Ditto.
1294         (*<shift_insn><mode>3_mask_1): Ditto.
1295         (*<rotate_insn><mode>3_mask): Ditto.
1296         (*<rotate_insn><mode>3_mask_1): Ditto.
1298 2017-12-21  Alexandre Oliva  <aoliva@redhat.com>
1300         * reorg.c (make_return_insns): Reemit each insn with its own location.
1302 2017-12-21  James Greenhalgh  <james.greenhalgh@arm.com>
1304         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
1305         generation for cases where splatting a value is not useful.
1306         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
1307         across a vec_duplicate and a paradoxical subreg forming a vector
1308         mode to a vec_concat.
1310 2017-12-21  James Greenhalgh  <james.greenhalgh@arm.com>
1312         * combine.c (simplify_set): Do not transform subregs to zero_extends
1313         if the destination is not a scalar int mode.
1315 2017-12-21  Eric Botcazou  <ebotcazou@adacore.com>
1317         PR c++/82872
1318         * convert.c (convert_to_integer_1) <POINTER_TYPE>: Do not return the
1319         shared zero if the input has overflowed.
1321 2017-12-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1323         * config/arm/driver-arm.c (arm_cpu_table): Specify dotprod
1324         support for Cortex-A55 and Cortex-A75.
1326 2017-12-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1328         * common/config/arm/arm-common.c (compare_opt_names): Add function
1329         comment.  Use strcmp instead of manual loop.
1331 2017-12-21  Martin Liska  <mliska@suse.cz>
1333         PR gcov-profile/83509
1334         * gcov-dump.c (dump_gcov_file): Do not read info about
1335         support_unexecuted_blocks for gcda files.
1337 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1339         PR rtl-optimization/82973
1340         * emit-rtl.h (valid_for_const_vec_duplicate_p): Rename to ...
1341         (valid_for_const_vector_p): ... this.
1342         * emit-rtl.c (valid_for_const_vec_duplicate_p): Rename to ...
1343         (valid_for_const_vector_p): ... this.  Adjust function comment.
1344         (gen_vec_duplicate): Adjust caller.
1345         * optabs.c (expand_vector_broadcast): Likewise.
1346         * simplify-rtx.c (simplify_const_unary_operation): Don't optimize into
1347         CONST_VECTOR if some element isn't simplified valid_for_const_vector_p
1348         constant.
1349         (simplify_const_binary_operation): Likewise.  Use CONST_FIXED_P macro
1350         instead of GET_CODE == CONST_FIXED.
1351         (simplify_subreg): Use CONST_FIXED_P macro instead of
1352         GET_CODE == CONST_FIXED.
1354         PR target/83488
1355         * config/i386/i386.c (ix86_target_string): Move -mavx512vbmi2 and
1356         -mshstk entries from isa_opts2 to isa_opts and -mhle, -mmovbe,
1357         -mclzero and -mmwaitx entries from isa_opts to isa_opts2.
1358         (ix86_option_override_internal): Adjust for
1359         OPTION_MASK_ISA_{HLE,MOVBE,CLZERO,MWAITX} moving to ix86_isa_flags2
1360         and OPTION_MASK_ISA_SHSTK moving to ix86_isa_flags.
1361         (BDESC_VERIFYS): Remove SPECIAL_ARGS2 related checks.
1362         (ix86_init_mmx_sse_builtins): Remove bdesc_special_args2 handling.
1363         Use def_builtin2 instead of def_builtin for OPTION_MASK_ISA_MWAITX
1364         and OPTION_MASK_ISA_CLZERO builtins.  Use def_builtin instead of
1365         def_builtin2 for CET builtins.
1366         (ix86_expand_builtin): Remove bdesc_special_args2 handling.  Fix
1367         up formatting in IX86_BUILTIN_RDPID code.
1368         * config/i386/i386-builtin.def: Move VBMI2 builtins from SPECIAL_ARGS2
1369         section to SPECIAL_ARGS and from ARGS2 section to ARGS.
1370         * config/i386/i386.opt (mavx512vbmi2, mshstk): Move from
1371         ix86_isa_flags2 to ix86_isa_flags.
1372         (mhle, mmovbe, mclzero, mmwaitx): Move from ix86_isa_flags to
1373         ix86_isa_flags2.
1374         * config/i386/i386-c.c (ix86_target_macros_internal): Check for
1375         OPTION_MASK_ISA_{CLZERO,MWAITX} in isa_flag2 instead of isa_flag.
1376         Check for OPTION_MASK_ISA_{SHSTK,AVX512VBMI2} in isa_flag instead
1377         of isa_flag2.
1378         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VBMI2_SET):
1379         Or in OPTION_MASK_ISA_AVX512F_SET.
1380         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
1381         OPTION_MASK_ISA_AVX512VBMI2_UNSET.
1382         (ix86_handle_option): Adjust for
1383         OPTION_MASK_ISA_{SHSTK,AVX512VBMI2}_*SET being in ix86_isa_flags
1384         and OPTION_MASK_ISA_{MOVBE,MWAITX,CLZERO}_*SET in ix86_isa_flags2.
1386 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1387             Alan Hayward  <alan.hayward@arm.com>
1388             David Sherwood  <david.sherwood@arm.com>
1390         * tree-data-ref.h (prune_runtime_alias_test_list): Take the
1391         factor as a poly_uint64 rather than an unsigned HOST_WIDE_INT.
1392         * tree-data-ref.c (prune_runtime_alias_test_list): Likewise.
1393         Track polynomial offsets.
1395 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1396             Alan Hayward  <alan.hayward@arm.com>
1397             David Sherwood  <david.sherwood@arm.com>
1399         * tree-vect-data-refs.c (vect_compute_data_ref_alignment):
1400         Treat drb->init as a poly_int.  Fail if its misalignment wrt
1401         vector_alignment isn't known.
1403 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1404             Alan Hayward  <alan.hayward@arm.com>
1405             David Sherwood  <david.sherwood@arm.com>
1407         * tree-vectorizer.h (_loop_vec_info): Add a versioning_threshold
1408         field.
1409         (LOOP_VINFO_VERSIONING_THRESHOLD): New macro
1410         (vect_loop_versioning): Take the loop versioning threshold as a
1411         separate parameter.
1412         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
1413         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
1414         versioning_threshold.
1415         (vect_analyze_loop_2): Compute the loop versioning threshold
1416         whenever loop versioning is needed, and store it in the new
1417         field rather than combining it with the cost model threshold.
1418         (vect_transform_loop): Update call to vect_loop_versioning.
1419         Try to combine the loop versioning and cost thresholds here.
1421 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1422             Alan Hayward  <alan.hayward@arm.com>
1423             David Sherwood  <david.sherwood@arm.com>
1425         * tree-ssa-loop-ivopts.h (strip_offset): Return the offset as
1426         poly_uint64_pod rather than an unsigned HOST_WIDE_INT.
1427         * tree-loop-distribution.c (classify_builtin_st): Update accordingly.
1428         * tree-ssa-loop-ivopts.c (iv_use::addr_offset): Change from
1429         an unsigned HOST_WIDE_INT to a poly_uint64_pod.
1430         (group_compare_offset): Update accordingly.
1431         (split_small_address_groups_p): Likewise.
1432         (record_use): Take addr_offset as a poly_uint64 rather than
1433         an unsigned HOST_WIDE_INT.
1434         (strip_offset): Return the offset as a poly_uint64 rather than
1435         an unsigned HOST_WIDE_INT.
1436         (record_group_use, split_address_groups): Track polynomial offsets.
1437         (add_iv_candidate_for_use): Likewise.
1438         (addr_offset_valid_p): Take the offset as a poly_int64 rather
1439         than a HOST_WIDE_INT.
1440         (strip_offset_1): Return the offset as a poly_int64 rather than
1441         a HOST_WIDE_INT.
1443 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1444             Alan Hayward  <alan.hayward@arm.com>
1445             David Sherwood  <david.sherwood@arm.com>
1447         * tree.h (get_binfo_at_offset): Take the offset as a poly_int64
1448         rather than a HOST_WIDE_INT.
1449         * tree.c (get_binfo_at_offset): Likewise.
1451 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1452             Alan Hayward  <alan.hayward@arm.com>
1453             David Sherwood  <david.sherwood@arm.com>
1455         * ipa-prop.h (build_ref_for_offset): Take the offset as a poly_int64
1456         rather than a HOST_WIDE_INT.
1457         * tree-sra.c (build_ref_for_offset): Likewise.
1459 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1460             Alan Hayward  <alan.hayward@arm.com>
1461             David Sherwood  <david.sherwood@arm.com>
1463         * fold-const.h (mem_ref_offset): Return a poly_offset_int rather
1464         than an offset_int.
1465         * tree.c (mem_ref_offset): Likewise.
1466         (build_simple_mem_ref_loc): Treat MEM_REF offsets as poly_ints.
1467         * builtins.c (get_object_alignment_2): Likewise.
1468         * expr.c (get_inner_reference, expand_expr_real_1): Likewise.
1469         * gimple-fold.c (get_base_constructor): Likewise.
1470         * gimple-ssa-strength-reduction.c (restructure_reference): Likewise.
1471         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
1472         Likewise.
1473         * ipa-polymorphic-call.c
1474         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Likewise.
1475         * ipa-prop.c (compute_complex_assign_jump_func): Likewise.
1476         (get_ancestor_addr_info): Likewise.
1477         * ipa-param-manipulation.c (ipa_get_adjustment_candidate): Likewise.
1478         * match.pd: Likewise.
1479         * tree-data-ref.c (dr_analyze_innermost): Likewise.
1480         * tree-dfa.c (get_addr_base_and_unit_offset_1): Likewise.
1481         * tree-eh.c (tree_could_trap_p): Likewise.
1482         * tree-object-size.c (addr_object_size): Likewise.
1483         * tree-ssa-address.c (copy_ref_info): Likewise.
1484         * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Likewise.
1485         (indirect_refs_may_alias_p): Likewise.
1486         * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Likewise.
1487         * tree-ssa.c (maybe_rewrite_mem_ref_base): Likewise.
1488         (non_rewritable_mem_ref_base): Likewise.
1489         * tree-vect-data-refs.c (vect_check_gather_scatter): Likewise.
1490         * tree-vrp.c (vrp_prop::check_array_ref): Likewise.
1491         * varasm.c (decode_addr_const): Likewise.
1493 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1494             Alan Hayward  <alan.hayward@arm.com>
1495             David Sherwood  <david.sherwood@arm.com>
1497         * gimple-ssa-stor-merging.c (find_bswap_or_nop_load): Track polynomial
1498         offsets for MEM_REFs.
1500 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1501             Alan Hayward  <alan.hayward@arm.com>
1502             David Sherwood  <david.sherwood@arm.com>
1504         * tree-ssanames.h (adjust_ptr_info_misalignment): Take the increment
1505         as a poly_uint64 rather than an unsigned int.
1506         * tree-ssanames.c (adjust_ptr_info_misalignment): Likewise.
1508 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1509             Alan Hayward  <alan.hayward@arm.com>
1510             David Sherwood  <david.sherwood@arm.com>
1512         * varasm.c (addr_const::offset): Change from HOST_WIDE_INT
1513         to poly_int64.
1514         (decode_addr_const): Update accordingly.
1516 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1517             Alan Hayward  <alan.hayward@arm.com>
1518             David Sherwood  <david.sherwood@arm.com>
1520         * tree.h (bit_field_size, bit_field_offset): New functions.
1521         * hsa-gen.c (gen_hsa_addr): Use them.
1522         * tree-ssa-forwprop.c (simplify_bitfield_ref): Likewise.
1523         (simplify_vector_constructor): Likewise.
1524         * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Likewise.
1525         * tree-cfg.c (verify_expr): Require the sizes and offsets of a
1526         BIT_FIELD_REF to be poly_uint64s rather than uhwis.
1527         * fold-const.c (fold_ternary_loc): Protect tree_to_uhwi with
1528         tree_fits_uhwi_p.
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         * expr.h (emit_group_load, emit_group_load_into_temps)
1535         (emit_group_store): Take the size as a poly_int64 rather than an int.
1536         * expr.c (emit_group_load_1, emit_group_load): Likewise.
1537         (emit_group_load_into_temp, emit_group_store): Likewise.
1539 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1540             Alan Hayward  <alan.hayward@arm.com>
1541             David Sherwood  <david.sherwood@arm.com>
1543         * ira-int.h (ira_spilled_reg_stack_slot::width): Change from
1544         an unsigned int to a poly_uint64.
1545         * ira.h (ira_reuse_stack_slot, ira_mark_new_stack_slot): Take the
1546         sizes as poly_uint64s rather than unsigned ints.
1547         * ira-color.c (ira_reuse_stack_slot, ira_mark_new_stack_slot):
1548         Likewise.
1550 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1551             Alan Hayward  <alan.hayward@arm.com>
1552             David Sherwood  <david.sherwood@arm.com>
1554         * lra-constraints.c (emit_inc): Change inc_amount from an int
1555         to a poly_int64.
1557 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1558             Alan Hayward  <alan.hayward@arm.com>
1559             David Sherwood  <david.sherwood@arm.com>
1561         * cfgexpand.c (stack_var::size): Change from a HOST_WIDE_INT
1562         to a poly_uint64.
1563         (add_stack_var, stack_var_cmp, partition_stack_vars)
1564         (dump_stack_var_partition): Update accordingly.
1565         (alloc_stack_frame_space): Take the size as a poly_int64 rather
1566         than a HOST_WIDE_INT.
1567         (expand_stack_vars, expand_one_stack_var_1): Handle polynomial sizes.
1568         (defer_stack_allocation, estimated_stack_frame_size): Likewise.
1569         (account_stack_vars, expand_one_var): Likewise.  Return a poly_uint64
1570         rather than a HOST_WIDE_INT.
1572 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1573             Alan Hayward  <alan.hayward@arm.com>
1574             David Sherwood  <david.sherwood@arm.com>
1576         * target.def (return_pops_args): Treat both the input and output
1577         sizes as poly_int64s rather than HOST_WIDE_INTS.
1578         * targhooks.h (default_return_pops_args): Update accordingly.
1579         * targhooks.c (default_return_pops_args): Likewise.
1580         * doc/tm.texi: Regenerate.
1581         * emit-rtl.h (incoming_args): Change pops_args, size and
1582         outgoing_args_size from int to poly_int64_pod.
1583         * function.h (expr_status): Change x_pending_stack_adjust and
1584         x_stack_pointer_delta from int to poly_int64.
1585         (args_size::constant): Change from HOST_WIDE_INT to poly_int64.
1586         (ARGS_SIZE_RTX): Update accordingly.
1587         * calls.c (highest_outgoing_arg_in_use): Change from int to
1588         unsigned int.
1589         (stack_usage_watermark, stored_args_watermark): New variables.
1590         (stack_region_maybe_used_p, mark_stack_region_used): New functions.
1591         (emit_call_1): Change the stack_size and rounded_stack_size
1592         parameters from HOST_WIDE_INT to poly_int64.  Track n_popped
1593         as a poly_int64.
1594         (save_fixed_argument_area): Check stack_usage_watermark.
1595         (initialize_argument_information): Change old_pending_adj from
1596         a HOST_WIDE_INT * to a poly_int64_pod *.
1597         (compute_argument_block_size): Return the size as a poly_int64
1598         rather than an int.
1599         (finalize_must_preallocate): Track polynomial argument sizes.
1600         (compute_argument_addresses): Likewise.
1601         (internal_arg_pointer_based_exp): Track polynomial offsets.
1602         (mem_overlaps_already_clobbered_arg_p): Rename to...
1603         (mem_might_overlap_already_clobbered_arg_p): ...this and take the
1604         size as a poly_uint64 rather than an unsigned HOST_WIDE_INT.
1605         Check stored_args_used_watermark.
1606         (load_register_parameters): Update accordingly.
1607         (check_sibcall_argument_overlap_1): Likewise.
1608         (combine_pending_stack_adjustment_and_call): Take the unadjusted
1609         args size as a poly_int64 rather than an int.  Return a bool
1610         indicating whether the optimization was possible and return
1611         the new adjustment by reference.
1612         (check_sibcall_argument_overlap): Track polynomail argument sizes.
1613         Update stored_args_watermark.
1614         (can_implement_as_sibling_call_p): Handle polynomial argument sizes.
1615         (expand_call): Likewise.  Maintain stack_usage_watermark and
1616         stored_args_watermark.  Update calls to
1617         combine_pending_stack_adjustment_and_call.
1618         (emit_library_call_value_1): Handle polynomial argument sizes.
1619         Call stack_region_maybe_used_p and mark_stack_region_used.
1620         Maintain stack_usage_watermark.
1621         (store_one_arg): Likewise.  Update call to
1622         mem_overlaps_already_clobbered_arg_p.
1623         * config/arm/arm.c (arm_output_function_prologue): Add a cast to
1624         HOST_WIDE_INT.
1625         * config/avr/avr.c (avr_outgoing_args_size): Likewise.
1626         * config/microblaze/microblaze.c (microblaze_function_prologue):
1627         Likewise.
1628         * config/cr16/cr16.c (cr16_return_pops_args): Update for new
1629         TARGET_RETURN_POPS_ARGS interface.
1630         (cr16_compute_frame, cr16_initial_elimination_offset): Add casts
1631         to HOST_WIDE_INT.
1632         * config/ft32/ft32.c (ft32_compute_frame): Likewise.
1633         * config/i386/i386.c (ix86_return_pops_args): Update for new
1634         TARGET_RETURN_POPS_ARGS interface.
1635         (ix86_expand_split_stack_prologue): Add a cast to HOST_WIDE_INT.
1636         * config/moxie/moxie.c (moxie_compute_frame): Likewise.
1637         * config/m68k/m68k.c (m68k_return_pops_args): Update for new
1638         TARGET_RETURN_POPS_ARGS interface.
1639         * config/vax/vax.c (vax_return_pops_args): Likewise.
1640         * config/pa/pa.h (STACK_POINTER_OFFSET): Add a cast to poly_int64.
1641         (EXIT_IGNORE_STACK): Update reference to crtl->outgoing_args_size.
1642         * config/arm/arm.h (CALLER_INTERWORKING_SLOT_SIZE): Likewise.
1643         * config/powerpcspe/aix.h (STACK_DYNAMIC_OFFSET): Likewise.
1644         * config/powerpcspe/darwin.h (STACK_DYNAMIC_OFFSET): Likewise.
1645         * config/powerpcspe/powerpcspe.h (STACK_DYNAMIC_OFFSET): Likewise.
1646         * config/rs6000/aix.h (STACK_DYNAMIC_OFFSET): Likewise.
1647         * config/rs6000/darwin.h (STACK_DYNAMIC_OFFSET): Likewise.
1648         * config/rs6000/rs6000.h (STACK_DYNAMIC_OFFSET): Likewise.
1649         * dojump.h (saved_pending_stack_adjust): Change x_pending_stack_adjust
1650         and x_stack_pointer_delta from int to poly_int64.
1651         * dojump.c (do_pending_stack_adjust): Update accordingly.
1652         * explow.c (allocate_dynamic_stack_space): Handle polynomial
1653         stack_pointer_deltas.
1654         * function.c (STACK_DYNAMIC_OFFSET): Add a cast to poly_int64.
1655         (pad_to_arg_alignment): Track polynomial offsets.
1656         (assign_parm_find_stack_rtl): Likewise.
1657         (assign_parms, locate_and_pad_parm): Handle polynomial argument sizes.
1658         * toplev.c (output_stack_usage): Update reference to
1659         current_function_pushed_stack_size.
1661 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1662             Alan Hayward  <alan.hayward@arm.com>
1663             David Sherwood  <david.sherwood@arm.com>
1665         * function.c (in_arg_offset, var_offset, dynamic_offset)
1666         (out_arg_offset, cfa_offset): Change from int to poly_int64.
1667         (instantiate_new_reg): Return the new offset as a poly_int64_pod
1668         rather than a HOST_WIDE_INT.
1669         (instantiate_virtual_regs_in_rtx): Track polynomial offsets.
1670         (instantiate_virtual_regs_in_insn): Likewise.
1672 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1673             Alan Hayward  <alan.hayward@arm.com>
1674             David Sherwood  <david.sherwood@arm.com>
1676         * rtl.h (get_args_size, add_args_size_note): New functions.
1677         (find_args_size_adjust): Return a poly_int64 rather than a
1678         HOST_WIDE_INT.
1679         (fixup_args_size_notes): Likewise.  Make the same change to the
1680         end_args_size parameter.
1681         * rtlanal.c (get_args_size, add_args_size_note): New functions.
1682         * builtins.c (expand_builtin_trap): Use add_args_size_note.
1683         * calls.c (emit_call_1): Likewise.
1684         * explow.c (adjust_stack_1): Likewise.
1685         * cfgcleanup.c (old_insns_match_p): Update use of
1686         find_args_size_adjust.
1687         * combine.c (distribute_notes): Track polynomial arg sizes.
1688         * dwarf2cfi.c (dw_trace_info): Change beg_true_args_size,
1689         end_true_args_size, beg_delay_args_size and end_delay_args_size
1690         from HOST_WIDE_INT to poly_int64.
1691         (add_cfi_args_size): Take the args_size as a poly_int64 rather
1692         than a HOST_WIDE_INT.
1693         (notice_args_size, notice_eh_throw, maybe_record_trace_start)
1694         (maybe_record_trace_start_abnormal, scan_trace, connect_traces): Track
1695         polynomial arg sizes.
1696         * emit-rtl.c (try_split): Use get_args_size.
1697         * recog.c (peep2_attempt): Likewise.
1698         * reload1.c (reload_as_needed): Likewise.
1699         * expr.c (find_args_size_adjust): Return the adjustment as a
1700         poly_int64 rather than a HOST_WIDE_INT.
1701         (fixup_args_size_notes): Change end_args_size from a HOST_WIDE_INT
1702         to a poly_int64 and change the return type in the same way.
1703         (emit_single_push_insn): Track polynomial arg sizes.
1705 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1706             Alan Hayward  <alan.hayward@arm.com>
1707             David Sherwood  <david.sherwood@arm.com>
1709         * expr.h (push_block, emit_push_insn): Change the "extra" parameter
1710         from HOST_WIDE_INT to poly_int64.
1711         * expr.c (push_block, emit_push_insn): Likewise.
1713 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1714             Alan Hayward  <alan.hayward@arm.com>
1715             David Sherwood  <david.sherwood@arm.com>
1717         * function.h (frame_space): Change start and length from HOST_WIDE_INT
1718         to poly_int64.
1719         (get_frame_size): Return the size as a poly_int64 rather than a
1720         HOST_WIDE_INT.
1721         (frame_offset_overflow): Take the offset as a poly_int64 rather
1722         than a HOST_WIDE_INT.
1723         (assign_stack_local_1, assign_stack_local, assign_stack_temp_for_type)
1724         (assign_stack_temp): Likewise for the size.
1725         * function.c (get_frame_size): Return a poly_int64 rather than
1726         a HOST_WIDE_INT.
1727         (frame_offset_overflow): Take the offset as a poly_int64 rather
1728         than a HOST_WIDE_INT.
1729         (try_fit_stack_local): Take the start, length and size as poly_int64s
1730         rather than HOST_WIDE_INTs.  Return the offset as a poly_int64_pod
1731         rather than a HOST_WIDE_INT.
1732         (add_frame_space): Take the start and end as poly_int64s rather than
1733         HOST_WIDE_INTs.
1734         (assign_stack_local_1, assign_stack_local, assign_stack_temp_for_type)
1735         (assign_stack_temp): Likewise for the size.
1736         (temp_slot): Change size, base_offset and full_size from HOST_WIDE_INT
1737         to poly_int64.
1738         (find_temp_slot_from_address): Handle polynomial offsets.
1739         (combine_temp_slots): Likewise.
1740         * emit-rtl.h (rtl_data::x_frame_offset): Change from HOST_WIDE_INT
1741         to poly_int64.
1742         * cfgexpand.c (alloc_stack_frame_space): Return the offset as a
1743         poly_int64 rather than a HOST_WIDE_INT.
1744         (expand_one_stack_var_at): Take the offset as a poly_int64 rather
1745         than a HOST_WIDE_INT.
1746         (expand_stack_vars, expand_one_stack_var_1, expand_used_vars): Handle
1747         polynomial frame offsets.
1748         * config/m32r/m32r-protos.h (m32r_compute_frame_size): Take the size
1749         as a poly_int64 rather than an int.
1750         * config/m32r/m32r.c (m32r_compute_frame_size): Likewise.
1751         * config/v850/v850-protos.h (compute_frame_size): Likewise.
1752         * config/v850/v850.c (compute_frame_size): Likewise.
1753         * config/xtensa/xtensa-protos.h (compute_frame_size): Likewise.
1754         * config/xtensa/xtensa.c (compute_frame_size): Likewise.
1755         * config/pa/pa-protos.h (pa_compute_frame_size): Likewise.
1756         * config/pa/pa.c (pa_compute_frame_size): Likewise.
1757         * explow.h (get_dynamic_stack_base): Take the offset as a poly_int64
1758         rather than a HOST_WIDE_INT.
1759         * explow.c (get_dynamic_stack_base): Likewise.
1760         * final.c (final_start_function): Use the constant lower bound
1761         of the frame size for -Wframe-larger-than.
1762         * ira.c (do_reload): Adjust for new get_frame_size return type.
1763         * lra.c (lra): Likewise.
1764         * reload1.c (reload): Likewise.
1765         * config/avr/avr.c (avr_asm_function_end_prologue): Likewise.
1766         * config/pa/pa.h (EXIT_IGNORE_STACK): Likewise.
1767         * rtlanal.c (get_initial_register_offset): Return the offset as
1768         a poly_int64 rather than a HOST_WIDE_INT.
1770 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1771             Alan Hayward  <alan.hayward@arm.com>
1772             David Sherwood  <david.sherwood@arm.com>
1774         * reload1.c (elim_table): Change initial_offset, offset and
1775         previous_offset from HOST_WIDE_INT to poly_int64_pod.
1776         (offsets_at): Change the target array's element type from
1777         HOST_WIDE_INT to poly_int64_pod.
1778         (set_label_offsets, eliminate_regs_1, eliminate_regs_in_insn)
1779         (elimination_costs_in_insn, update_eliminable_offsets)
1780         (verify_initial_elim_offsets, set_offsets_for_label)
1781         (init_eliminable_invariants): Update after above changes.
1783 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1784             Alan Hayward  <alan.hayward@arm.com>
1785             David Sherwood  <david.sherwood@arm.com>
1787         * reload.h (reload::inc): Change from an int to a poly_int64_pod.
1788         * reload.c (combine_reloads, debug_reload_to_stream): Likewise.
1789         (decomposition): Change start and end from HOST_WIDE_INT
1790         to poly_int64_pod.
1791         (decompose, immune_p): Update accordingly.
1792         (find_inc_amount): Return a poly_int64 rather than an int.
1793         * reload1.c (inc_for_reload): Take the inc_amount as a poly_int64
1794         rather than an int.
1796 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1797             Alan Hayward  <alan.hayward@arm.com>
1798             David Sherwood  <david.sherwood@arm.com>
1800         * tree.h (get_inner_reference): Return the bitsize and bitpos
1801         as poly_int64_pods rather than HOST_WIDE_INT.
1802         * fold-const.h (ptr_difference_const): Return the pointer difference
1803         as a poly_int64_pod rather than a HOST_WIDE_INT.
1804         * expr.c (get_inner_reference): Return the bitsize and bitpos
1805         as poly_int64_pods rather than HOST_WIDE_INT.
1806         (expand_expr_addr_expr_1, expand_expr_real_1): Track polynomial
1807         offsets and sizes.
1808         * fold-const.c (make_bit_field_ref): Take the bitpos as a poly_int64
1809         rather than a HOST_WIDE_INT.  Update call to get_inner_reference.
1810         (optimize_bit_field_compare): Update call to get_inner_reference.
1811         (decode_field_reference): Likewise.
1812         (fold_unary_loc): Track polynomial offsets and sizes.
1813         (split_address_to_core_and_offset): Return the bitpos as a
1814         poly_int64_pod rather than a HOST_WIDE_INT.
1815         (ptr_difference_const): Likewise for the pointer difference.
1816         * asan.c (instrument_derefs): Track polynomial offsets and sizes.
1817         * config/mips/mips.c (r10k_safe_mem_expr_p): Likewise.
1818         * dbxout.c (dbxout_expand_expr): Likewise.
1819         * dwarf2out.c (loc_list_for_address_of_addr_expr_of_indirect_ref)
1820         (loc_list_from_tree_1, fortran_common): Likewise.
1821         * gimple-laddress.c (pass_laddress::execute): Likewise.
1822         * gimple-ssa-store-merging.c (find_bswap_or_nop_load): Likewise.
1823         * gimplify.c (gimplify_scan_omp_clauses): Likewise.
1824         * simplify-rtx.c (delegitimize_mem_from_attrs): Likewise.
1825         * tree-affine.c (tree_to_aff_combination): Likewise.
1826         (get_inner_reference_aff): Likewise.
1827         * tree-data-ref.c (split_constant_offset_1): Likewise.
1828         (dr_analyze_innermost): Likewise.
1829         * tree-scalar-evolution.c (interpret_rhs_expr): Likewise.
1830         * tree-sra.c (ipa_sra_check_caller): Likewise.
1831         * tree-vect-data-refs.c (vect_check_gather_scatter): Likewise.
1832         * ubsan.c (maybe_instrument_pointer_overflow): Likewise.
1833         (instrument_bool_enum_load, instrument_object_size): Likewise.
1834         * gimple-ssa-strength-reduction.c (slsr_process_ref): Update call
1835         to get_inner_reference.
1836         * hsa-gen.c (gen_hsa_addr): Likewise.
1837         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Likewise.
1838         * tsan.c (instrument_expr): Likewise.
1839         * match.pd: Update call to ptr_difference_const.
1841 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1842             Alan Hayward  <alan.hayward@arm.com>
1843             David Sherwood  <david.sherwood@arm.com>
1845         * fold-const.c (fold_comparison): Track sizes and offsets as
1846         poly_int64s rather than HOST_WIDE_INTs when folding address
1847         comparisons.
1849 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1850             Alan Hayward  <alan.hayward@arm.com>
1851             David Sherwood  <david.sherwood@arm.com>
1853         * expr.h (get_bit_range): Return the bitstart and bitend as
1854         poly_uint64s rather than unsigned HOST_WIDE_INTs.  Return the bitpos
1855         as a poly_int64 rather than a HOST_WIDE_INT.
1856         * expr.c (get_bit_range): Likewise.
1857         (expand_assignment): Update call accordingly.
1858         * fold-const.c (optimize_bit_field_compare): Likewise.
1860 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1861             Alan Hayward  <alan.hayward@arm.com>
1862             David Sherwood  <david.sherwood@arm.com>
1864         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): New enum.
1865         (aarch64_legitimate_address_p): Use it instead of an rtx code,
1866         as an optional final parameter.
1867         * config/aarch64/aarch64.c (aarch64_classify_address): Likewise.
1868         (aarch64_legitimate_address_p): Likewise.
1869         (aarch64_print_address_internal): Take an aarch64_addr_query_type
1870         instead of an rtx code.
1871         (aarch64_address_valid_for_prefetch_p): Update calls accordingly.
1872         (aarch64_legitimate_address_hook_p): Likewise.
1873         (aarch64_print_ldpstp_address): Likewise.
1874         (aarch64_print_operand_address): Likewise.
1875         (aarch64_address_cost): Likewise.
1876         * config/aarch64/constraints.md (Uml, Umq, Ump, Utq): Likewise.
1877         * config/aarch64/predicates.md (aarch64_mem_pair_operand): Likewise.
1878         (aarch64_mem_pair_lanes_operand): Likewise.
1880 2017-12-20  Richard Biener  <rguenther@suse.de>
1882         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Call
1883         update_stmt_if_modified.
1885 2017-12-20  Wilco Dijkstra  <wdijkstr@arm.com>
1887         PR tree-optimization/83491
1888         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Check for SSA_NAME
1889         before walking uses.  Improve coding style and comments.
1891 2017-12-20  Tom de Vries  <tom@codesourcery.com>
1893         * gimple-fold.c (fold_internal_goacc_dim): Simplify.
1895 2017-12-20  Jakub Jelinek  <jakub@redhat.com>
1897         PR ipa/83506
1898         * ipa-fnsummary.c (pass_data_ipa_free_fn_summary): Use 0 for
1899         todo_flags_finish.
1900         (pass_ipa_free_fn_summary): Add small_p private data member,
1901         initialize to false in the ctor.
1902         (pass_ipa_free_fn_summary::clone,
1903         pass_ipa_free_fn_summary::set_pass_param,
1904         pass_ipa_free_fn_summary::gate): New methods.
1905         (pass_ipa_free_fn_summary::execute): Return TODO_remove_functions
1906         | TODO_dump_symtab if small_p.
1907         * passes.def: Add true parm for the existing pass_ipa_free_fn_summary
1908         entry and add another instance of the pass with false parm after
1909         ipa-pure-const.
1910         * ipa-pure-const.c (pass_ipa_pure_const): Don't call
1911         ipa_free_fn_summary here.
1913 2017-12-20  Paolo Carlini  <paolo.carlini@oracle.com>
1915         * gimplify.c (gimplify_return_expr): Remove dead error_mark_node check.
1917 2017-12-20  Martin Sebor  <msebor@redhat.com>
1919         PR testsuite/83131
1920         * builtins.c (expand_builtin_strlen): Use get_callee_fndecl.
1921         (expand_builtin_strcmp): Call maybe_warn_nonstring_arg.
1922         (expand_builtin_strncmp): Same.
1924 2017-12-20  Alexandre Oliva  <aoliva@redhat.com>
1926         PR bootstrap/83396
1927         * cfgexpand.c (label_rtx_for_bb): Revert SFN changes that
1928         allowed debug stmts before labels.
1929         (expand_gimple_basic_block): Likewise.
1930         * gimple-iterator.c (gimple_find_edge_insert_loc): Likewise.
1931         * gimple-iterator.h (gsi_after_labels): Likewise.
1932         * tree-cfgcleanup (remove_forwarder_block): Likewise, but
1933         rename reused variable, and simplify using gsi_move_before.
1934         * tree-ssa-tail-merge.c (find_duplicate): Likewise.
1935         * tree-cfg.c (make_edges, cleanup_dead_labels): Likewise.
1936         (gimple_can_merge_blocks_p, verify_gimple_in_cfg): Likewise.
1937         (gimple_verify_flow_info, gimple_block_label): Likewise.
1938         (make_blocks): Move debug markers after adjacent labels.
1939         * cfgrtl.c (skip_insns_after_block): Revert SFN changes that
1940         allowed debug insns outside blocks.
1941         * df-scan.c (df_insn_delete): Likewise.
1942         * lra-constraints.c (update_ebb_live_info): Likewise.
1943         * var-tracking.c (get_first_insn, vt_emit_notes): Likewise.
1944         (vt_initialize, delete_vta_debug_insns): Likewise.
1945         (reemit_marker_as_note): Drop BB parm.  Adjust callers.
1947 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1948             Alan Hayward  <alan.hayward@arm.com>
1949             David Sherwood  <david.sherwood@arm.com>
1951         * poly-int-types.h (round_down_to_byte_boundary): New macro.
1952         (round_up_to_byte_boundary): Likewise.
1953         * expr.h (get_bit_range): Add temporary shim.
1954         * gimple-ssa-store-merging.c (store_operand_info): Change the
1955         bitsize, bitpos, bitregion_start and bitregion_end fields from
1956         unsigned HOST_WIDE_INT to poly_uint64.
1957         (merged_store_group): Likewise load_align_base.
1958         (compatible_load_p, compatible_load_p): Update accordingly.
1959         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
1960         (split_group, imm_store_chain_info::output_merged_store): Likewise.
1961         (mem_valid_for_store_merging): Return the bitsize, bitpos,
1962         bitregion_start and bitregion_end as poly_uint64s rather than
1963         unsigned HOST_WIDE_INTs.  Track polynomial offsets internally.
1964         (handled_load): Take the bitsize, bitpos,
1965         bitregion_start and bitregion_end as poly_uint64s rather than
1966         unsigned HOST_WIDE_INTs.
1967         (pass_store_merging::process_store): Update call to
1968         mem_valid_for_store_merging.
1970 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1971             Alan Hayward  <alan.hayward@arm.com>
1972             David Sherwood  <david.sherwood@arm.com>
1974         * builtins.c (get_object_alignment_2): Track polynomial offsets
1975         and sizes.  Update the alignment handling.
1977 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1978             Alan Hayward  <alan.hayward@arm.com>
1979             David Sherwood  <david.sherwood@arm.com>
1981         * tree.h (get_inner_reference): Add a version that returns the
1982         offset and size as poly_int64_pods rather than HOST_WIDE_INTs.
1983         * cfgexpand.c (expand_debug_expr): Track polynomial offsets.  Simply
1984         the case in which bitpos is not associated with the first byte.
1986 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1987             Alan Hayward  <alan.hayward@arm.com>
1988             David Sherwood  <david.sherwood@arm.com>
1990         * tree-affine.h (get_inner_reference_aff): Return the size as a
1991         poly_widest_int.
1992         * tree-affine.c (get_inner_reference_aff): Likewise.
1993         * tree-data-ref.c (dr_may_alias_p): Update accordingly.
1994         * tree-ssa-loop-im.c (mem_refs_may_alias_p): Likewise.
1996 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1997             Alan Hayward  <alan.hayward@arm.com>
1998             David Sherwood  <david.sherwood@arm.com>
2000         * fold-const.c (pointer_may_wrap_p): Take the offset as a
2001         HOST_WIDE_INT rather than a poly_int64.
2003 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2004             Alan Hayward  <alan.hayward@arm.com>
2005             David Sherwood  <david.sherwood@arm.com>
2007         * gimple-ssa-store-merging.c (symbolic_number::bytepos): Change from
2008         HOST_WIDE_INT to poly_int64_pod.
2009         (perform_symbolic_merge): Update accordingly.
2010         (bswap_replace): Likewise.
2012 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2013             Alan Hayward  <alan.hayward@arm.com>
2014             David Sherwood  <david.sherwood@arm.com>
2016         * tree-affine.h (aff_tree::offset): Change from widest_int
2017         to poly_widest_int.
2018         (wide_int_ext_for_comb): Delete.
2019         (aff_combination_const, aff_comb_cannot_overlap_p): Take the
2020         constants as poly_widest_int rather than widest_int.
2021         (aff_combination_constant_multiple_p): Return the multiplier
2022         as a poly_widest_int.
2023         (aff_combination_zero_p, aff_combination_singleton_var_p): Handle
2024         polynomial offsets.
2025         * tree-affine.c (wide_int_ext_for_comb): Make original widest_int
2026         version static and add an overload for poly_widest_int.
2027         (aff_combination_const, aff_combination_add_cst)
2028         (wide_int_constant_multiple_p, aff_comb_cannot_overlap_p): Take
2029         the constants as poly_widest_int rather than widest_int.
2030         (tree_to_aff_combination): Generalize INTEGER_CST case to
2031         poly_int_tree_p.
2032         (aff_combination_to_tree): Track offsets as poly_widest_ints.
2033         (aff_combination_add_product, aff_combination_mult): Handle
2034         polynomial offsets.
2035         (aff_combination_constant_multiple_p): Return the multiplier
2036         as a poly_widest_int.
2037         * tree-predcom.c (determine_offset): Return the offset as a
2038         poly_widest_int.
2039         (split_data_refs_to_components, suitable_component_p): Update
2040         accordingly.
2041         (valid_initializer_p): Update call to
2042         aff_combination_constant_multiple_p.
2043         * tree-ssa-address.c (addr_to_parts): Handle polynomial offsets.
2044         * tree-ssa-loop-ivopts.c (get_address_cost_ainc): Take the step
2045         as a poly_int64 rather than a HOST_WIDE_INT.
2046         (get_address_cost): Handle polynomial offsets.
2047         (iv_elimination_compare_lt): Likewise.
2048         (rewrite_use_nonlinear_expr): Likewise.
2050 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2051             Alan Hayward  <alan.hayward@arm.com>
2052             David Sherwood  <david.sherwood@arm.com>
2054         * tree-dfa.h (get_addr_base_and_unit_offset_1): Return the offset
2055         as a poly_int64_pod rather than a HOST_WIDE_INT.
2056         (get_addr_base_and_unit_offset): Likewise.
2057         * tree-dfa.c (get_addr_base_and_unit_offset_1): Likewise.
2058         (get_addr_base_and_unit_offset): Likewise.
2059         * doc/match-and-simplify.texi: Change off from HOST_WIDE_INT
2060         to poly_int64 in example.
2061         * fold-const.c (fold_binary_loc): Update call to
2062         get_addr_base_and_unit_offset.
2063         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
2064         (maybe_canonicalize_mem_ref_addr): Likewise.
2065         (gimple_fold_stmt_to_constant_1): Likewise.
2066         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
2067         Likewise.
2068         * ipa-param-manipulation.c (ipa_modify_call_arguments): Likewise.
2069         * match.pd: Likewise.
2070         * omp-low.c (lower_omp_target): Likewise.
2071         * tree-sra.c (build_ref_for_offset): Likewise.
2072         (build_debug_ref_for_model): Likewise.
2073         * tree-ssa-address.c (maybe_fold_tmr): Likewise.
2074         * tree-ssa-alias.c (ao_ref_init_from_ptr_and_size): Likewise.
2075         * tree-ssa-ccp.c (optimize_memcpy): Likewise.
2076         * tree-ssa-forwprop.c (forward_propagate_addr_expr_1): Likewise.
2077         (constant_pointer_difference): Likewise.
2078         * tree-ssa-loop-niter.c (expand_simple_operations): Likewise.
2079         * tree-ssa-phiopt.c (jump_function_from_stmt): Likewise.
2080         * tree-ssa-pre.c (create_component_ref_by_pieces_1): Likewise.
2081         * tree-ssa-sccvn.c (vn_reference_fold_indirect): Likewise.
2082         (vn_reference_maybe_forwprop_address, vn_reference_lookup_3): Likewise.
2083         (set_ssa_val_to): Likewise.
2084         * tree-ssa-strlen.c (get_addr_stridx, addr_stridxptr)
2085         (maybe_diag_stxncpy_trunc): Likewise.
2086         * tree-vrp.c (vrp_prop::check_array_ref): Likewise.
2087         * tree.c (build_simple_mem_ref_loc): Likewise.
2088         (array_at_struct_end_p): Likewise.
2090 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2091             Alan Hayward  <alan.hayward@arm.com>
2092             David Sherwood  <david.sherwood@arm.com>
2094         * tree-dfa.h (get_ref_base_and_extent): Return the base, size and
2095         max_size as poly_int64_pods rather than HOST_WIDE_INTs.
2096         (get_ref_base_and_extent_hwi): Declare.
2097         * tree-dfa.c (get_ref_base_and_extent): Return the base, size and
2098         max_size as poly_int64_pods rather than HOST_WIDE_INTs.
2099         (get_ref_base_and_extent_hwi): New function.
2100         * cfgexpand.c (expand_debug_expr): Update call to
2101         get_ref_base_and_extent.
2102         * dwarf2out.c (add_var_loc_to_decl): Likewise.
2103         * gimple-fold.c (get_base_constructor): Return the offset as a
2104         poly_int64_pod rather than a HOST_WIDE_INT.
2105         (fold_const_aggregate_ref_1): Track polynomial sizes and offsets.
2106         * ipa-polymorphic-call.c
2107         (ipa_polymorphic_call_context::set_by_invariant)
2108         (extr_type_from_vtbl_ptr_store): Track polynomial offsets.
2109         (ipa_polymorphic_call_context::ipa_polymorphic_call_context)
2110         (check_stmt_for_type_change): Use get_ref_base_and_extent_hwi
2111         rather than get_ref_base_and_extent.
2112         (ipa_polymorphic_call_context::get_dynamic_type): Likewise.
2113         * ipa-prop.c (ipa_load_from_parm_agg, compute_complex_assign_jump_func)
2114         (get_ancestor_addr_info, determine_locally_known_aggregate_parts):
2115         Likewise.
2116         * ipa-param-manipulation.c (ipa_get_adjustment_candidate): Update
2117         call to get_ref_base_and_extent.
2118         * tree-sra.c (create_access, get_access_for_expr): Likewise.
2119         * tree-ssa-alias.c (ao_ref_base, aliasing_component_refs_p)
2120         (stmt_kills_ref_p): Likewise.
2121         * tree-ssa-dce.c (mark_aliased_reaching_defs_necessary_1): Likewise.
2122         * tree-ssa-scopedtables.c (avail_expr_hash, equal_mem_array_ref_p):
2123         Likewise.
2124         * tree-ssa-sccvn.c (vn_reference_lookup_3): Likewise.
2125         Use get_ref_base_and_extent_hwi rather than get_ref_base_and_extent
2126         when calling native_encode_expr.
2127         * tree-ssa-structalias.c (get_constraint_for_component_ref): Update
2128         call to get_ref_base_and_extent.
2129         (do_structure_copy): Use get_ref_base_and_extent_hwi rather than
2130         get_ref_base_and_extent.
2131         * var-tracking.c (track_expr_p): Likewise.
2133 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2134             Alan Hayward  <alan.hayward@arm.com>
2135             David Sherwood  <david.sherwood@arm.com>
2137         * ipa-param-manipulation.h (ipa_parm_adjustment::offset): Change from
2138         HOST_WIDE_INT to poly_int64_pod.
2139         * ipa-param-manipulation.c (ipa_modify_call_arguments): Track
2140         polynomail parameter offsets.
2142 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2143             Alan Hayward  <alan.hayward@arm.com>
2144             David Sherwood  <david.sherwood@arm.com>
2146         * gengtype.c (main): Handle poly_int64_pod.
2147         * dwarf2out.h (dw_cfi_oprnd_cfa_loc): New dw_cfi_oprnd_type.
2148         (dw_cfi_oprnd::dw_cfi_cfa_loc): New field.
2149         (dw_cfa_location::offset, dw_cfa_location::base_offset): Change
2150         from HOST_WIDE_INT to poly_int64_pod.
2151         * dwarf2cfi.c (queued_reg_save::cfa_offset): Likewise.
2152         (copy_cfa): New function.
2153         (lookup_cfa_1): Use the cached dw_cfi_cfa_loc, if it exists.
2154         (cfi_oprnd_equal_p): Handle dw_cfi_oprnd_cfa_loc.
2155         (cfa_equal_p, dwarf2out_frame_debug_adjust_cfa)
2156         (dwarf2out_frame_debug_cfa_offset, dwarf2out_frame_debug_expr)
2157         (initial_return_save): Treat offsets as poly_ints.
2158         (def_cfa_0): Likewise.  Cache the CFA in dw_cfi_cfa_loc if either
2159         offset is nonconstant.
2160         (reg_save): Take the offset as a poly_int64.  Fall back to
2161         DW_CFA_expression for nonconstant offsets.
2162         (queue_reg_save): Take the offset as a poly_int64.
2163         * dwarf2out.c (dw_cfi_oprnd2_desc): Handle DW_CFA_def_cfa_expression.
2165 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2166             Alan Hayward  <alan.hayward@arm.com>
2167             David Sherwood  <david.sherwood@arm.com>
2169         * rtl.h (operand_subword, operand_subword_force): Take the offset
2170         as a poly_uint64 an unsigned int.
2171         * emit-rtl.c (operand_subword, operand_subword_force): Likewise.
2173 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2174             Alan Hayward  <alan.hayward@arm.com>
2175             David Sherwood  <david.sherwood@arm.com>
2177         * doc/rtl.texi: Update documentation of SUBREG_BYTE.  Document the
2178         'p' format code.  Use INT_LIST rather than SUBREG as the example of
2179         a code with an XINT and an XEXP.  Remove the implication that
2180         accessing an rtx field using XINT is expected to work.
2181         * rtl.def (SUBREG): Change format from "ei" to "ep".
2182         * rtl.h (rtunion::rt_subreg): New field.
2183         (XCSUBREG): New macro.
2184         (SUBREG_BYTE): Use it.
2185         (subreg_shape): Change offset from an unsigned int to a poly_uint16.
2186         Update constructor accordingly.
2187         (subreg_shape::operator ==): Update accordingly.
2188         (subreg_shape::unique_id): Return an unsigned HOST_WIDE_INT rather
2189         than an unsigned int.
2190         (subreg_lsb, subreg_lowpart_offset, subreg_highpart_offset): Return
2191         a poly_uint64 rather than an unsigned int.
2192         (subreg_lsb_1): Likewise.  Take the offset as a poly_uint64 rather
2193         than an unsigned int.
2194         (subreg_size_offset_from_lsb, subreg_size_lowpart_offset)
2195         (subreg_size_highpart_offset): Return a poly_uint64 rather than
2196         an unsigned int.  Take the sizes as poly_uint64s.
2197         (subreg_offset_from_lsb): Return a poly_uint64 rather than
2198         an unsigned int.  Take the shift as a poly_uint64 rather than
2199         an unsigned int.
2200         (subreg_regno_offset, subreg_offset_representable_p): Take the offset
2201         as a poly_uint64 rather than an unsigned int.
2202         (simplify_subreg_regno): Likewise.
2203         (byte_lowpart_offset): Return the memory offset as a poly_int64
2204         rather than an int.
2205         (subreg_memory_offset): Likewise.  Take the subreg offset as a
2206         poly_uint64 rather than an unsigned int.
2207         (simplify_subreg, simplify_gen_subreg, subreg_get_info)
2208         (gen_rtx_SUBREG, validate_subreg): Take the subreg offset as a
2209         poly_uint64 rather than an unsigned int.
2210         * rtl.c (rtx_format): Describe 'p' in comment.
2211         (copy_rtx, rtx_equal_p_cb, rtx_equal_p): Handle 'p'.
2212         * emit-rtl.c (validate_subreg, gen_rtx_SUBREG): Take the subreg
2213         offset as a poly_uint64 rather than an unsigned int.
2214         (byte_lowpart_offset): Return the memory offset as a poly_int64
2215         rather than an int.
2216         (subreg_memory_offset): Likewise.  Take the subreg offset as a
2217         poly_uint64 rather than an unsigned int.
2218         (subreg_size_lowpart_offset, subreg_size_highpart_offset): Take the
2219         mode sizes as poly_uint64s rather than unsigned ints.  Return a
2220         poly_uint64 rather than an unsigned int.
2221         (subreg_lowpart_p): Treat subreg offsets as poly_ints.
2222         (copy_insn_1): Handle 'p'.
2223         * rtlanal.c (set_noop_p): Treat subregs offsets as poly_uint64s.
2224         (subreg_lsb_1): Take the subreg offset as a poly_uint64 rather than
2225         an unsigned int.  Return the shift in the same way.
2226         (subreg_lsb): Return the shift as a poly_uint64 rather than an
2227         unsigned int.
2228         (subreg_size_offset_from_lsb): Take the sizes and shift as
2229         poly_uint64s rather than unsigned ints.  Return the offset as
2230         a poly_uint64.
2231         (subreg_get_info, subreg_regno_offset, subreg_offset_representable_p)
2232         (simplify_subreg_regno): Take the offset as a poly_uint64 rather than
2233         an unsigned int.
2234         * rtlhash.c (add_rtx): Handle 'p'.
2235         * genemit.c (gen_exp): Likewise.
2236         * gengenrtl.c (type_from_format, gendef): Likewise.
2237         * gensupport.c (subst_pattern_match, get_alternatives_number)
2238         (collect_insn_data, alter_predicate_for_insn, alter_constraints)
2239         (subst_dup): Likewise.
2240         * gengtype.c (adjust_field_rtx_def): Likewise.
2241         * genrecog.c (find_operand, find_matching_operand, validate_pattern)
2242         (match_pattern_2): Likewise.
2243         (rtx_test::SUBREG_FIELD): New rtx_test::kind_enum.
2244         (rtx_test::subreg_field): New function.
2245         (operator ==, safe_to_hoist_p, transition_parameter_type)
2246         (print_nonbool_test, print_test): Handle SUBREG_FIELD.
2247         * genattrtab.c (attr_rtx_1): Say that 'p' is deliberately not handled.
2248         * genpeep.c (match_rtx): Likewise.
2249         * print-rtl.c (print_poly_int): Include if GENERATOR_FILE too.
2250         (rtx_writer::print_rtx_operand): Handle 'p'.
2251         (print_value): Handle SUBREG.
2252         * read-rtl.c (apply_int_iterator): Likewise.
2253         (rtx_reader::read_rtx_operand): Handle 'p'.
2254         * alias.c (rtx_equal_for_memref_p): Likewise.
2255         * cselib.c (rtx_equal_for_cselib_1, cselib_hash_rtx): Likewise.
2256         * caller-save.c (replace_reg_with_saved_mem): Treat subreg offsets
2257         as poly_ints.
2258         * calls.c (expand_call): Likewise.
2259         * combine.c (combine_simplify_rtx, expand_field_assignment): Likewise.
2260         (make_extraction, gen_lowpart_for_combine): Likewise.
2261         * loop-invariant.c (hash_invariant_expr_1, invariant_expr_equal_p):
2262         Likewise.
2263         * cse.c (remove_invalid_subreg_refs): Take the offset as a poly_uint64
2264         rather than an unsigned int.  Treat subreg offsets as poly_ints.
2265         (exp_equiv_p): Handle 'p'.
2266         (hash_rtx_cb): Likewise.  Treat subreg offsets as poly_ints.
2267         (equiv_constant, cse_insn): Treat subreg offsets as poly_ints.
2268         * dse.c (find_shift_sequence): Likewise.
2269         * dwarf2out.c (rtl_for_decl_location): Likewise.
2270         * expmed.c (extract_low_bits): Likewise.
2271         * expr.c (emit_group_store, undefined_operand_subword_p): Likewise.
2272         (expand_expr_real_2): Likewise.
2273         * final.c (alter_subreg): Likewise.
2274         (leaf_renumber_regs_insn): Handle 'p'.
2275         * function.c (assign_parm_find_stack_rtl, assign_parm_setup_stack):
2276         Treat subreg offsets as poly_ints.
2277         * fwprop.c (forward_propagate_and_simplify): Likewise.
2278         * ifcvt.c (noce_emit_move_insn, noce_emit_cmove): Likewise.
2279         * ira.c (get_subreg_tracking_sizes): Likewise.
2280         * ira-conflicts.c (go_through_subreg): Likewise.
2281         * ira-lives.c (process_single_reg_class_operands): Likewise.
2282         * jump.c (rtx_renumbered_equal_p): Likewise.  Handle 'p'.
2283         * lower-subreg.c (simplify_subreg_concatn): Take the subreg offset
2284         as a poly_uint64 rather than an unsigned int.
2285         (simplify_gen_subreg_concatn, resolve_simple_move): Treat
2286         subreg offsets as poly_ints.
2287         * lra-constraints.c (operands_match_p): Handle 'p'.
2288         (match_reload, curr_insn_transform): Treat subreg offsets as poly_ints.
2289         * lra-spills.c (assign_mem_slot): Likewise.
2290         * postreload.c (move2add_valid_value_p): Likewise.
2291         * recog.c (general_operand, indirect_operand): Likewise.
2292         * regcprop.c (copy_value, maybe_mode_change): Likewise.
2293         (copyprop_hardreg_forward_1): Likewise.
2294         * reginfo.c (simplifiable_subregs_hasher::hash, simplifiable_subregs)
2295         (record_subregs_of_mode): Likewise.
2296         * rtlhooks.c (gen_lowpart_general, gen_lowpart_if_possible): Likewise.
2297         * reload.c (operands_match_p): Handle 'p'.
2298         (find_reloads_subreg_address): Treat subreg offsets as poly_ints.
2299         * reload1.c (alter_reg, choose_reload_regs): Likewise.
2300         (compute_reload_subreg_offset): Likewise, and return an poly_int64.
2301         * simplify-rtx.c (simplify_truncation, simplify_binary_operation_1):
2302         (test_vector_ops_duplicate): Treat subreg offsets as poly_ints.
2303         (simplify_const_poly_int_tests<N>::run): Likewise.
2304         (simplify_subreg, simplify_gen_subreg): Take the subreg offset as
2305         a poly_uint64 rather than an unsigned int.
2306         * valtrack.c (debug_lowpart_subreg): Likewise.
2307         * var-tracking.c (var_lowpart): Likewise.
2308         (loc_cmp): Handle 'p'.
2310 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2311             Alan Hayward  <alan.hayward@arm.com>
2312             David Sherwood  <david.sherwood@arm.com>
2314         * ira.c (get_subreg_tracking_sizes): New function.
2315         (init_live_subregs): Take an integer size rather than a register.
2316         (build_insn_chain): Use get_subreg_tracking_sizes.  Update calls
2317         to init_live_subregs.
2319 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2320             Alan Hayward  <alan.hayward@arm.com>
2321             David Sherwood  <david.sherwood@arm.com>
2323         * expr.c (store_constructor_field): Change bitsize from a
2324         unsigned HOST_WIDE_INT to a poly_uint64 and bitpos from a
2325         HOST_WIDE_INT to a poly_int64.
2326         (store_constructor): Change size from a HOST_WIDE_INT to
2327         a poly_int64.
2328         (store_field): Likewise bitsize and bitpos.
2330 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2331             Alan Hayward  <alan.hayward@arm.com>
2332             David Sherwood  <david.sherwood@arm.com>
2334         * expmed.h (store_bit_field): Change bitregion_start and
2335         bitregion_end from unsigned HOST_WIDE_INT to poly_uint64.
2336         * expmed.c (adjust_bit_field_mem_for_reg, strict_volatile_bitfield_p)
2337         (store_bit_field_1, store_integral_bit_field, store_bit_field)
2338         (store_fixed_bit_field, store_split_bit_field): Likewise.
2339         * expr.c (store_constructor_field, store_field): Likewise.
2340         (optimize_bitfield_assignment_op): Likewise.  Make the same change
2341         to bitsize and bitpos.
2342         * machmode.h (bit_field_mode_iterator): Change m_bitregion_start
2343         and m_bitregion_end from HOST_WIDE_INT to poly_int64.  Make the
2344         same change in the constructor arguments.
2345         (get_best_mode): Change bitregion_start and bitregion_end from
2346         unsigned HOST_WIDE_INT to poly_uint64.
2347         * stor-layout.c (bit_field_mode_iterator::bit_field_mode_iterator):
2348         Change bitregion_start and bitregion_end from HOST_WIDE_INT to
2349         poly_int64.
2350         (bit_field_mode_iterator::next_mode): Update for new types
2351         of m_bitregion_start and m_bitregion_end.
2352         (get_best_mode): Change bitregion_start and bitregion_end from
2353         unsigned HOST_WIDE_INT to poly_uint64.
2355 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2356             Alan Hayward  <alan.hayward@arm.com>
2357             David Sherwood  <david.sherwood@arm.com>
2359         * rtl.h (simplify_gen_subreg): Add a temporary overload that
2360         accepts poly_uint64 offsets.
2361         * expmed.h (extract_bit_field): Take bitsize and bitnum as
2362         poly_uint64s rather than unsigned HOST_WIDE_INTs.
2363         * expmed.c (lowpart_bit_field_p): Likewise.
2364         (extract_bit_field_as_subreg): New function, split out from...
2365         (extract_bit_field_1): ...here.  Take bitsize and bitnum as
2366         poly_uint64s rather than unsigned HOST_WIDE_INTs.  For vector
2367         extractions, check that BITSIZE matches the size of the extracted
2368         value and that BITNUM is an exact multiple of that size.
2369         If all else fails, try forcing the value into memory if
2370         BITNUM is variable, and adjusting the address so that the
2371         offset is constant.  Split the part that can only handle constant
2372         bitsize and bitnum out into...
2373         (extract_integral_bit_field): ...this new function.
2374         (extract_bit_field): Take bitsize and bitnum as poly_uint64s
2375         rather than unsigned HOST_WIDE_INTs.
2377 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2378             Alan Hayward  <alan.hayward@arm.com>
2379             David Sherwood  <david.sherwood@arm.com>
2381         * expmed.h (store_bit_field): Take bitsize and bitnum as
2382         poly_uint64s rather than unsigned HOST_WIDE_INTs.
2383         * expmed.c (simple_mem_bitfield_p): Likewise.  Add a parameter
2384         that returns the byte size.
2385         (store_bit_field_1): Take bitsize and bitnum as
2386         poly_uint64s rather than unsigned HOST_WIDE_INTs.  Update call
2387         to simple_mem_bitfield_p.  Split the part that can only handle
2388         constant bitsize and bitnum out into...
2389         (store_integral_bit_field): ...this new function.
2390         (store_bit_field): Take bitsize and bitnum as poly_uint64s rather
2391         than unsigned HOST_WIDE_INTs.
2392         (extract_bit_field_1): Update call to simple_mem_bitfield_p.
2394 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2395             Alan Hayward  <alan.hayward@arm.com>
2396             David Sherwood  <david.sherwood@arm.com>
2398         * lra-int.h (lra_reg): Change offset from int to poly_int64.
2399         (lra_insn_recog_data): Change sp_offset from HOST_WIDE_INT
2400         to poly_int64.
2401         (lra_eliminate_regs_1, eliminate_regs_in_insn): Change
2402         update_sp_offset from a HOST_WIDE_INT to a poly_int64.
2403         (lra_update_reg_val_offset, lra_reg_val_equal_p): Take the
2404         offset as a poly_int64 rather than an int.
2405         * lra-assigns.c (find_hard_regno_for_1): Handle poly_int64 offsets.
2406         (setup_live_pseudos_and_spill_after_risky_transforms): Likewise.
2407         * lra-constraints.c (equiv_address_substitution): Track offsets
2408         as poly_int64s.
2409         (emit_inc): Check poly_int_rtx_p instead of CONST_INT_P.
2410         (curr_insn_transform): Handle the new form of sp_offset.
2411         * lra-eliminations.c (lra_elim_table): Change previous_offset
2412         and offset from HOST_WIDE_INT to poly_int64.
2413         (print_elim_table, update_reg_eliminate): Update accordingly.
2414         (self_elim_offsets): Change from HOST_WIDE_INT to poly_int64_pod.
2415         (get_elimination): Update accordingly.
2416         (form_sum): Check poly_int_rtx_p instead of CONST_INT_P.
2417         (lra_eliminate_regs_1, eliminate_regs_in_insn): Change
2418         update_sp_offset from a HOST_WIDE_INT to a poly_int64.  Handle
2419         poly_int64 offsets generally.
2420         (curr_sp_change): Change from HOST_WIDE_INT to poly_int64.
2421         (mark_not_eliminable, init_elimination): Update accordingly.
2422         (remove_reg_equal_offset_note): Return a bool and pass the new
2423         offset back by pointer as a poly_int64.
2424         * lra-remat.c (change_sp_offset): Take sp_offset as a poly_int64
2425         rather than a HOST_WIDE_INT.
2426         (do_remat): Track offsets poly_int64s.
2427         * lra.c (lra_update_insn_recog_data, setup_sp_offset): Likewise.
2429 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2430             Alan Hayward  <alan.hayward@arm.com>
2431             David Sherwood  <david.sherwood@arm.com>
2433         * rtl.h (mem_attrs): Add a default constructor.  Change size and
2434         offset from HOST_WIDE_INT to poly_int64.
2435         * emit-rtl.h (set_mem_offset, set_mem_size, adjust_address_1)
2436         (adjust_automodify_address_1, set_mem_attributes_minus_bitpos)
2437         (widen_memory_access): Take the sizes and offsets as poly_int64s
2438         rather than HOST_WIDE_INTs.
2439         * alias.c (ao_ref_from_mem): Handle the new form of MEM_OFFSET.
2440         (offset_overlap_p): Take poly_int64s rather than HOST_WIDE_INTs
2441         and ints.
2442         (adjust_offset_for_component_ref): Change the offset from a
2443         HOST_WIDE_INT to a poly_int64.
2444         (nonoverlapping_memrefs_p): Track polynomial offsets and sizes.
2445         * cfgcleanup.c (merge_memattrs): Update after mem_attrs changes.
2446         * dce.c (find_call_stack_args): Likewise.
2447         * dse.c (record_store): Likewise.
2448         * dwarf2out.c (tls_mem_loc_descriptor, dw_sra_loc_expr): Likewise.
2449         * print-rtl.c (rtx_writer::print_rtx): Likewise.
2450         * read-rtl-function.c (test_loading_mem): Likewise.
2451         * rtlanal.c (may_trap_p_1): Likewise.
2452         * simplify-rtx.c (delegitimize_mem_from_attrs): Likewise.
2453         * var-tracking.c (int_mem_offset, track_expr_p): Likewise.
2454         * emit-rtl.c (mem_attrs_eq_p, get_mem_align_offset): Likewise.
2455         (mem_attrs::mem_attrs): New function.
2456         (set_mem_attributes_minus_bitpos): Change bitpos from a
2457         HOST_WIDE_INT to poly_int64.
2458         (set_mem_alias_set, set_mem_addr_space, set_mem_align, set_mem_expr)
2459         (clear_mem_offset, clear_mem_size, change_address)
2460         (get_spill_slot_decl, set_mem_attrs_for_spill): Directly
2461         initialize mem_attrs.
2462         (set_mem_offset, set_mem_size, adjust_address_1)
2463         (adjust_automodify_address_1, offset_address, widen_memory_access):
2464         Likewise.  Take poly_int64s rather than HOST_WIDE_INT.
2466 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2467             Alan Hayward  <alan.hayward@arm.com>
2468             David Sherwood  <david.sherwood@arm.com>
2470         * rtlanal.c (rtx_addr_can_trap_p_1): Take the offset and size
2471         as poly_int64s rather than HOST_WIDE_INTs.  Use a size of -1
2472         rather than 0 to represent an unknown size.  Assert that the size
2473         is known when the mode isn't BLKmode.
2474         (may_trap_p_1): Use -1 for unknown sizes.
2475         (rtx_addr_can_trap_p): Likewise.  Pass BLKmode rather than VOIDmode.
2477 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2478             Alan Hayward  <alan.hayward@arm.com>
2479             David Sherwood  <david.sherwood@arm.com>
2481         * dse.c (store_info): Change offset and width from HOST_WIDE_INT
2482         to poly_int64.  Update commentary for positions_needed.large.
2483         (read_info_type): Change offset and width from HOST_WIDE_INT
2484         to poly_int64.
2485         (set_usage_bits): Likewise.
2486         (canon_address): Return the offset as a poly_int64 rather than
2487         a HOST_WIDE_INT.  Use strip_offset_and_add.
2488         (set_all_positions_unneeded, any_positions_needed_p): Use
2489         positions_needed.large to track stores with non-constant widths.
2490         (all_positions_needed_p): Likewise.  Take the offset and width
2491         as poly_int64s rather than ints.  Assert that rhs is nonnull.
2492         (record_store): Cope with non-constant offsets and widths.
2493         Nullify the rhs of an earlier store if we can't tell which bytes
2494         of it are needed.
2495         (find_shift_sequence): Take the access_size and shift as poly_int64s
2496         rather than ints.
2497         (get_stored_val): Take the read_offset and read_width as poly_int64s
2498         rather than HOST_WIDE_INTs.
2499         (check_mem_read_rtx, scan_stores, scan_reads, dse_step5): Handle
2500         non-constant offsets and widths.
2502 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2503             Alan Hayward  <alan.hayward@arm.com>
2504             David Sherwood  <david.sherwood@arm.com>
2506         * inchash.h (inchash::hash::add_poly_int): New function.
2507         * tree-ssa-alias.h (ao_ref::offset, ao_ref::size, ao_ref::max_size):
2508         Use poly_int64 rather than HOST_WIDE_INT.
2509         (ao_ref::max_size_known_p): New function.
2510         * tree-ssa-sccvn.h (vn_reference_op_struct::off): Use poly_int64_pod
2511         rather than HOST_WIDE_INT.
2512         * tree-ssa-alias.c (ao_ref_base): Apply get_ref_base_and_extent
2513         to temporaries until its interface is adjusted to match.
2514         (ao_ref_init_from_ptr_and_size): Handle polynomial offsets and sizes.
2515         (aliasing_component_refs_p, decl_refs_may_alias_p)
2516         (indirect_ref_may_alias_decl_p, indirect_refs_may_alias_p): Take
2517         the offsets and max_sizes as poly_int64s instead of HOST_WIDE_INTs.
2518         (refs_may_alias_p_1, stmt_kills_ref_p): Adjust for changes to
2519         ao_ref fields.
2520         * alias.c (ao_ref_from_mem): Likewise.
2521         * tree-ssa-dce.c (mark_aliased_reaching_defs_necessary_1): Likewise.
2522         * tree-ssa-dse.c (valid_ao_ref_for_dse, normalize_ref)
2523         (clear_bytes_written_by, setup_live_bytes_from_ref, compute_trims)
2524         (maybe_trim_complex_store, maybe_trim_constructor_store)
2525         (live_bytes_read, dse_classify_store): Likewise.
2526         * tree-ssa-sccvn.c (vn_reference_compute_hash, vn_reference_eq):
2527         (copy_reference_ops_from_ref, ao_ref_init_from_vn_reference)
2528         (fully_constant_vn_reference_p, valueize_refs_1): Likewise.
2529         (vn_reference_lookup_3): Likewise.
2530         * tree-ssa-uninit.c (warn_uninitialized_vars): Likewise.
2532 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2533             Alan Hayward  <alan.hayward@arm.com>
2534             David Sherwood  <david.sherwood@arm.com>
2536         * tree-ssa-alias.c (indirect_ref_may_alias_decl_p)
2537         (indirect_refs_may_alias_p): Use ranges_may_overlap_p
2538         instead of ranges_overlap_p.
2540 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2541             Alan Hayward  <alan.hayward@arm.com>
2542             David Sherwood  <david.sherwood@arm.com>
2544         * tree-ssa-alias.c (same_addr_size_stores_p): Take the offsets and
2545         sizes as poly_int64s rather than HOST_WIDE_INTs.
2547 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2548             Alan Hayward  <alan.hayward@arm.com>
2549             David Sherwood  <david.sherwood@arm.com>
2551         * gimple-fold.h (fold_ctor_reference): Take the offset and size
2552         as poly_uint64 rather than unsigned HOST_WIDE_INT.
2553         * gimple-fold.c (fold_ctor_reference): Likewise.
2555 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2556             Alan Hayward  <alan.hayward@arm.com>
2557             David Sherwood  <david.sherwood@arm.com>
2559         * target.def (dwarf_poly_indeterminate_value): New hook.
2560         * targhooks.h (default_dwarf_poly_indeterminate_value): Declare.
2561         * targhooks.c (default_dwarf_poly_indeterminate_value): New function.
2562         * doc/tm.texi.in (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Document.
2563         * doc/tm.texi: Regenerate.
2564         * dwarf2out.h (build_cfa_loc, build_cfa_aligned_loc): Take the
2565         offset as a poly_int64.
2566         * dwarf2out.c (new_reg_loc_descr): Move later in file.  Take the
2567         offset as a poly_int64.
2568         (loc_descr_plus_const, loc_list_plus_const, build_cfa_aligned_loc):
2569         Take the offset as a poly_int64.
2570         (build_cfa_loc): Likewise.  Use loc_descr_plus_const.
2571         (frame_pointer_fb_offset): Change to a poly_int64.
2572         (int_loc_descriptor): Take the offset as a poly_int64.  Use
2573         targetm.dwarf_poly_indeterminate_value for polynomial offsets.
2574         (based_loc_descr): Take the offset as a poly_int64.
2575         Use strip_offset_and_add to handle (plus X (const)).
2576         Use new_reg_loc_descr instead of an open-coded version of the
2577         previous implementation.
2578         (mem_loc_descriptor): Handle CONST_POLY_INT.
2579         (compute_frame_pointer_to_fb_displacement): Take the offset as a
2580         poly_int64.  Use strip_offset_and_add to handle (plus X (const)).
2582 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2583             Alan Hayward  <alan.hayward@arm.com>
2584             David Sherwood  <david.sherwood@arm.com>
2586         * rtl.h (reg_attrs::offset): Change from HOST_WIDE_INT to poly_int64.
2587         (gen_rtx_REG_offset): Take the offset as a poly_int64.
2588         * inchash.h (inchash::hash::add_poly_hwi): New function.
2589         * gengtype.c (main): Register poly_int64.
2590         * emit-rtl.c (reg_attr_hasher::hash): Use inchash.  Treat the
2591         offset as a poly_int.
2592         (reg_attr_hasher::equal): Use must_eq to compare offsets.
2593         (get_reg_attrs, update_reg_offset, gen_rtx_REG_offset): Take the
2594         offset as a poly_int64.
2595         (set_reg_attrs_from_value): Treat the offset as a poly_int64.
2596         * print-rtl.c (print_poly_int): New function.
2597         (rtx_writer::print_rtx_operand_code_r): Treat REG_OFFSET as
2598         a poly_int.
2599         * var-tracking.c (track_offset_p, get_tracked_reg_offset): New
2600         functions.
2601         (var_reg_set, var_reg_delete_and_set, var_reg_delete): Use them.
2602         (same_variable_part_p, track_loc_p): Take the offset as a poly_int64.
2603         (vt_get_decl_and_offset): Return the offset as a poly_int64.
2604         Enforce track_offset_p for parts of a PARALLEL.
2605         (vt_add_function_parameter): Use const_offset for the final
2606         offset to track.  Use get_tracked_reg_offset for the parts
2607         of a PARALLEL.
2609 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2610             Alan Hayward  <alan.hayward@arm.com>
2611             David Sherwood  <david.sherwood@arm.com>
2613         * target.def (truly_noop_truncation): Take poly_uint64s instead of
2614         unsigned ints.  Change default to hook_bool_puint64_puint64_true.
2615         * doc/tm.texi: Regenerate.
2616         * hooks.h (hook_bool_uint_uint_true): Delete.
2617         (hook_bool_puint64_puint64_true): Declare.
2618         * hooks.c (hook_bool_uint_uint_true): Delete.
2619         (hook_bool_puint64_puint64_true): New function.
2620         * config/mips/mips.c (mips_truly_noop_truncation): Take poly_uint64s
2621         instead of unsigned ints.
2622         * config/spu/spu.c (spu_truly_noop_truncation): Likewise.
2623         * config/tilegx/tilegx.c (tilegx_truly_noop_truncation): Likewise.
2625 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2626             Alan Hayward  <alan.hayward@arm.com>
2627             David Sherwood  <david.sherwood@arm.com>
2629         * optabs.h (expand_operand): Add an int_value field.
2630         (create_expand_operand): Add an int_value parameter and use it
2631         to initialize the new expand_operand field.
2632         (create_integer_operand): Replace with a declaration of a function
2633         that accepts poly_int64s.  Move the implementation to...
2634         * optabs.c (create_integer_operand): ...here.
2635         (maybe_legitimize_operand): For EXPAND_INTEGER, check whether
2636         the mode preserves the value of int_value, instead of calling
2637         const_int_operand on the rtx.  Use gen_int_mode to generate
2638         the new rtx.
2640 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2641             Alan Hayward  <alan.hayward@arm.com>
2642             David Sherwood  <david.sherwood@arm.com>
2644         * dumpfile.h (dump_dec): Declare.
2645         * dumpfile.c (dump_dec): New function.
2646         * pretty-print.h (pp_wide_integer): Turn into a function and
2647         declare a poly_int version.
2648         * pretty-print.c (pp_wide_integer): New function for poly_ints.
2650 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2651             Alan Hayward  <alan.hayward@arm.com>
2652             David Sherwood  <david.sherwood@arm.com>
2654         * doc/generic.texi (POLY_INT_CST): Document.
2655         * tree.def (POLY_INT_CST): New tree code.
2656         * treestruct.def (TS_POLY_INT_CST): New tree layout.
2657         * tree-core.h (tree_poly_int_cst): New struct.
2658         (tree_node): Add a poly_int_cst field.
2659         * tree.h (POLY_INT_CST_P, POLY_INT_CST_COEFF): New macros.
2660         (wide_int_to_tree, force_fit_type): Take a poly_wide_int_ref
2661         instead of a wide_int_ref.
2662         (build_int_cst, build_int_cst_type): Take a poly_int64 instead
2663         of a HOST_WIDE_INT.
2664         (build_int_cstu, build_array_type_nelts): Take a poly_uint64
2665         instead of an unsigned HOST_WIDE_INT.
2666         (build_poly_int_cst, tree_fits_poly_int64_p, tree_fits_poly_uint64_p)
2667         (ptrdiff_tree_p): Declare.
2668         (tree_to_poly_int64, tree_to_poly_uint64): Likewise.  Provide
2669         extern inline implementations if the target doesn't use POLY_INT_CST.
2670         (poly_int_tree_p): New function.
2671         (wi::unextended_tree): New class.
2672         (wi::int_traits <unextended_tree>): New override.
2673         (wi::extended_tree): Add a default constructor.
2674         (wi::extended_tree::get_tree): New function.
2675         (wi::widest_extended_tree, wi::offset_extended_tree): New typedefs.
2676         (wi::tree_to_widest_ref, wi::tree_to_offset_ref): Use them.
2677         (wi::tree_to_poly_widest_ref, wi::tree_to_poly_offset_ref)
2678         (wi::tree_to_poly_wide_ref): New typedefs.
2679         (wi::ints_for): Provide overloads for extended_tree and
2680         unextended_tree.
2681         (poly_int_cst_value, wi::to_poly_widest, wi::to_poly_offset)
2682         (wi::to_wide): New functions.
2683         (wi::fits_to_boolean_p, wi::fits_to_tree_p): Handle poly_ints.
2684         * tree.c (poly_int_cst_hasher): New struct.
2685         (poly_int_cst_hash_table): New variable.
2686         (tree_node_structure_for_code, tree_code_size, simple_cst_equal)
2687         (valid_constant_size_p, add_expr, drop_tree_overflow): Handle
2688         POLY_INT_CST.
2689         (initialize_tree_contains_struct): Handle TS_POLY_INT_CST.
2690         (init_ttree): Initialize poly_int_cst_hash_table.
2691         (build_int_cst, build_int_cst_type, build_invariant_address): Take
2692         a poly_int64 instead of a HOST_WIDE_INT.
2693         (build_int_cstu, build_array_type_nelts): Take a poly_uint64
2694         instead of an unsigned HOST_WIDE_INT.
2695         (wide_int_to_tree): Rename to...
2696         (wide_int_to_tree_1): ...this.
2697         (build_new_poly_int_cst, build_poly_int_cst): New functions.
2698         (force_fit_type): Take a poly_wide_int_ref instead of a wide_int_ref.
2699         (wide_int_to_tree): New function that takes a poly_wide_int_ref.
2700         (ptrdiff_tree_p, tree_to_poly_int64, tree_to_poly_uint64)
2701         (tree_fits_poly_int64_p, tree_fits_poly_uint64_p): New functions.
2702         * lto-streamer-out.c (DFS::DFS_write_tree_body, hash_tree): Handle
2703         TS_POLY_INT_CST.
2704         * tree-streamer-in.c (lto_input_ts_poly_tree_pointers): Likewise.
2705         (streamer_read_tree_body): Likewise.
2706         * tree-streamer-out.c (write_ts_poly_tree_pointers): Likewise.
2707         (streamer_write_tree_body): Likewise.
2708         * tree-streamer.c (streamer_check_handled_ts_structures): Likewise.
2709         * asan.c (asan_protect_global): Require the size to be an INTEGER_CST.
2710         * cfgexpand.c (expand_debug_expr): Handle POLY_INT_CST.
2711         * expr.c (expand_expr_real_1, const_vector_from_tree): Likewise.
2712         * gimple-expr.h (is_gimple_constant): Likewise.
2713         * gimplify.c (maybe_with_size_expr): Likewise.
2714         * print-tree.c (print_node): Likewise.
2715         * tree-data-ref.c (data_ref_compare_tree): Likewise.
2716         * tree-pretty-print.c (dump_generic_node): Likewise.
2717         * tree-ssa-address.c (addr_for_mem_ref): Likewise.
2718         * tree-vect-data-refs.c (dr_group_sort_cmp): Likewise.
2719         * tree-vrp.c (compare_values_warnv): Likewise.
2720         * tree-ssa-loop-ivopts.c (determine_base_object, constant_multiple_of)
2721         (get_loop_invariant_expr, add_candidate_1, get_computation_aff_1)
2722         (force_expr_to_var_cost): Likewise.
2723         * tree-ssa-loop.c (for_each_index): Likewise.
2724         * fold-const.h (build_invariant_address, size_int_kind): Take a
2725         poly_int64 instead of a HOST_WIDE_INT.
2726         * fold-const.c (fold_negate_expr_1, const_binop, const_unop)
2727         (fold_convert_const, multiple_of_p, fold_negate_const): Handle
2728         POLY_INT_CST.
2729         (size_binop_loc): Likewise.  Allow int_const_binop_1 to fail.
2730         (int_const_binop_2): New function, split out from...
2731         (int_const_binop_1): ...here.  Handle POLY_INT_CST.
2732         (size_int_kind): Take a poly_int64 instead of a HOST_WIDE_INT.
2733         * expmed.c (make_tree): Handle CONST_POLY_INT_P.
2734         * gimple-ssa-strength-reduction.c (slsr_process_add)
2735         (slsr_process_mul): Check for INTEGER_CSTs before using them
2736         as candidates.
2737         * stor-layout.c (bits_from_bytes): New function.
2738         (bit_from_pos): Use it.
2739         (layout_type): Likewise.  For vectors, multiply the TYPE_SIZE_UNIT
2740         by BITS_PER_UNIT to get the TYPE_SIZE.
2741         * tree-cfg.c (verify_expr, verify_types_in_gimple_reference): Allow
2742         MEM_REF and TARGET_MEM_REF offsets to be a POLY_INT_CST.
2744 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2745             Alan Hayward  <alan.hayward@arm.com>
2746             David Sherwood  <david.sherwood@arm.com>
2748         * doc/rtl.texi (const_poly_int): Document.  Also document the
2749         rtl sharing behavior.
2750         * gengenrtl.c (excluded_rtx): Return true for CONST_POLY_INT.
2751         * rtl.h (const_poly_int_def): New struct.
2752         (rtx_def::u): Add a cpi field.
2753         (CASE_CONST_UNIQUE, CASE_CONST_ANY): Add CONST_POLY_INT.
2754         (CONST_POLY_INT_P, CONST_POLY_INT_COEFFS): New macros.
2755         (wi::rtx_to_poly_wide_ref): New typedef
2756         (const_poly_int_value, wi::to_poly_wide, rtx_to_poly_int64)
2757         (poly_int_rtx_p): New functions.
2758         (trunc_int_for_mode): Declare a poly_int64 version.
2759         (plus_constant): Take a poly_int64 instead of a HOST_WIDE_INT.
2760         (immed_wide_int_const): Take a poly_wide_int_ref rather than
2761         a wide_int_ref.
2762         (strip_offset): Declare.
2763         (strip_offset_and_add): New function.
2764         * rtl.def (CONST_POLY_INT): New rtx code.
2765         * rtl.c (rtx_size): Handle CONST_POLY_INT.
2766         (shared_const_p): Use poly_int_rtx_p.
2767         * emit-rtl.h (gen_int_mode): Take a poly_int64 instead of a
2768         HOST_WIDE_INT.
2769         (gen_int_shift_amount): Likewise.
2770         * emit-rtl.c (const_poly_int_hasher): New class.
2771         (const_poly_int_htab): New variable.
2772         (init_emit_once): Initialize it when NUM_POLY_INT_COEFFS > 1.
2773         (const_poly_int_hasher::hash): New function.
2774         (const_poly_int_hasher::equal): Likewise.
2775         (gen_int_mode): Take a poly_int64 instead of a HOST_WIDE_INT.
2776         (immed_wide_int_const): Rename to...
2777         (immed_wide_int_const_1): ...this and make static.
2778         (immed_wide_int_const): New function, taking a poly_wide_int_ref
2779         instead of a wide_int_ref.
2780         (gen_int_shift_amount): Take a poly_int64 instead of a HOST_WIDE_INT.
2781         (gen_lowpart_common): Handle CONST_POLY_INT.
2782         * cse.c (hash_rtx_cb, equiv_constant): Likewise.
2783         * cselib.c (cselib_hash_rtx): Likewise.
2784         * dwarf2out.c (const_ok_for_output_1): Likewise.
2785         * expr.c (convert_modes): Likewise.
2786         * print-rtl.c (rtx_writer::print_rtx, print_value): Likewise.
2787         * rtlhash.c (add_rtx): Likewise.
2788         * explow.c (trunc_int_for_mode): Add a poly_int64 version.
2789         (plus_constant): Take a poly_int64 instead of a HOST_WIDE_INT.
2790         Handle existing CONST_POLY_INT rtxes.
2791         * expmed.h (expand_shift): Take a poly_int64 instead of a
2792         HOST_WIDE_INT.
2793         * expmed.c (expand_shift): Likewise.
2794         * rtlanal.c (strip_offset): New function.
2795         (commutative_operand_precedence): Give CONST_POLY_INT the same
2796         precedence as CONST_DOUBLE and put CONST_WIDE_INT between that
2797         and CONST_INT.
2798         * rtl-tests.c (const_poly_int_tests): New struct.
2799         (rtl_tests_c_tests): Use it.
2800         * simplify-rtx.c (simplify_const_unary_operation): Handle
2801         CONST_POLY_INT.
2802         (simplify_const_binary_operation): Likewise.
2803         (simplify_binary_operation_1): Fold additions of symbolic constants
2804         and CONST_POLY_INTs.
2805         (simplify_subreg): Handle extensions and truncations of
2806         CONST_POLY_INTs.
2807         (simplify_const_poly_int_tests): New struct.
2808         (simplify_rtx_c_tests): Use it.
2809         * wide-int.h (storage_ref): Add default constructor.
2810         (wide_int_ref_storage): Likewise.
2811         (trailing_wide_ints): Use GTY((user)).
2812         (trailing_wide_ints::operator[]): Add a const version.
2813         (trailing_wide_ints::get_precision): New function.
2814         (trailing_wide_ints::extra_size): Likewise.
2816 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2817             Alan Hayward  <alan.hayward@arm.com>
2818             David Sherwood  <david.sherwood@arm.com>
2820         * emit-rtl.h (gen_int_shift_amount): Declare.
2821         * emit-rtl.c (gen_int_shift_amount): New function.
2822         * asan.c (asan_emit_stack_protection): Use gen_int_shift_amount
2823         instead of GEN_INT.
2824         * calls.c (shift_return_value): Likewise.
2825         * cse.c (fold_rtx): Likewise.
2826         * dse.c (find_shift_sequence): Likewise.
2827         * expmed.c (init_expmed_one_mode, store_bit_field_1, expand_shift_1)
2828         (expand_shift, expand_smod_pow2): Likewise.
2829         * lower-subreg.c (shift_cost): Likewise.
2830         * optabs.c (expand_superword_shift, expand_doubleword_mult)
2831         (expand_unop, expand_binop, shift_amt_for_vec_perm_mask)
2832         (expand_vec_perm_var): Likewise.
2833         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
2834         (simplify_binary_operation_1): Likewise.
2835         * combine.c (try_combine, find_split_point, force_int_to_mode)
2836         (simplify_shift_const_1, simplify_shift_const): Likewise.
2837         (change_zero_ext): Likewise.  Use simplify_gen_binary.
2839 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2841         * poly-int.h (multiple_p): Fix handling of two non-poly_ints.
2843 2017-12-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2845         * doc/invoke.texi (ARM Options): Document accepted extension options
2846         for -march=armv8.3-a.
2848 2017-12-20  Richard Earnshaw  <rearnsha@arm.com>
2850         PR target/83105
2851         * config.gcc (arm*-*-linux*): When configured with --with-float=hard
2852         or --with-float=softfp, set the default CPU to arm10e.
2854 2017-12-20  Eric Botcazou  <ebotcazou@adacore.com>
2856         * config/visium/constraints.md (J, K, L): Use IN_RANGE macro.
2857         * config/visium/predicates.md (const_shift_operand): Likewise.
2858         * config/visium/visium.c (visium_legitimize_address): Fix oversight.
2859         (visium_legitimize_reload_address): Likewise.
2861 2017-12-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2863         PR target/82975
2864         * config/arm/arm.h (TEST_REGNO): Adjust comment as expected in r255830.
2866 2017-12-20  Jakub Jelinek  <jakub@redhat.com>
2868         PR c++/83490
2869         * calls.c (compute_argument_addresses): Ignore TYPE_EMPTY_P arguments.
2871 2017-12-20  Julia Koval  <julia.koval@intel.com>
2873         * common/config/i386/i386-common.c (OPTION_MASK_ISA_VPCLMULQDQ_SET,
2874         OPTION_MASK_ISA_VPCLMULQDQ_UNSET): New.
2875         (ix86_handle_option): Handle -mvpclmulqdq, move cx6 to flags2.
2876         * config.gcc: Include vpclmulqdqintrin.h.
2877         * config/i386/cpuid.h: Handle bit_VPCLMULQDQ.
2878         * config/i386/driver-i386.c (host_detect_local_cpu): Handle
2879         -mvpclmulqdq.
2880         * config/i386/i386-builtin.def (__builtin_ia32_vpclmulqdq_v2di,
2881         __builtin_ia32_vpclmulqdq_v4di, __builtin_ia32_vpclmulqdq_v8di): New.
2882         * config/i386/i386-c.c (__VPCLMULQDQ__): New.
2883         * config/i386/i386.c (isa2_opts): Add -mcx16.
2884         (isa_opts): Add -mpclmulqdq, remove -mcx16.
2885         (ix86_option_override_internal): Move mcx16 to flags2.
2886         (ix86_valid_target_attribute_inner_p): Add vpclmulqdq.
2887         (ix86_expand_builtin): Handle OPTION_MASK_ISA_VPCLMULQDQ.
2888         * config/i386/i386.h (TARGET_VPCLMULQDQ, TARGET_VPCLMULQDQ_P): New.
2889         * config/i386/i386.opt: Add mvpclmulqdq, move mcx16 to flags2.
2890         * config/i386/immintrin.h: Include vpclmulqdqintrin.h.
2891         * config/i386/sse.md (vpclmulqdq_<mode>): New pattern.
2892         * config/i386/vpclmulqdqintrin.h (_mm512_clmulepi64_epi128,
2893         _mm_clmulepi64_epi128, _mm256_clmulepi64_epi128): New intrinsics.
2894         * doc/invoke.texi: Add -mvpclmulqdq.
2896 2017-12-20  Tom de Vries  <tom@codesourcery.com>
2898         PR middle-end/83423
2899         * config/i386/i386.c (ix86_static_chain): Move
2900         DECL_STATIC_CHAIN test ...
2901         * calls.c (rtx_for_static_chain): ... here.  New function.
2902         * calls.h (rtx_for_static_chain): Declare.
2903         * builtins.c (expand_builtin_setjmp_receiver): Use rtx_for_static_chain
2904         instead of targetm.calls.static_chain.
2905         * df-scan.c (df_get_entry_block_def_set): Same.
2907 2017-12-19  Tom de Vries  <tom@codesourcery.com>
2909         PR tree-optimization/83493
2910         * graphite-isl-ast-to-gimple.c (translate_isl_ast_node_for): Unshare ub
2911         and lb.
2913 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2915         * gimple-ssa-sprintf.c (format_directive): Use inform_n instead of
2916         inform with hardcoded english plural handling.
2918 2017-12-18  Jeff Law  <law@redhat.com>
2920         PR tree-optimization/83477
2921         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis): For
2922         a non-virtual PHI, always push a new range.
2924 2017-12-19  Martin Sebor  <msebor@redhat.com>
2926         PR middle-end/77608
2927         * builtins.c (compute_objsize): Handle non-constant offsets.
2929 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2931         PR tree-optimization/83444
2932         * tree-ssa-strlen.c (strlen_check_and_optimize_stmt): For the
2933         character load case, if get_stridx on MEM_REF's operand doesn't
2934         look usable, retry with get_addr_stridx.
2936 2017-12-19  Alexandre Oliva  <aoliva@redhat.com>
2938         PR debug/83422
2939         * var-tracking.c (vt_debug_insns_local): Do not drop markers.
2940         (variable_tracking_main_1): Keep markers even when VTA fails.
2942         PR bootstrap/83396
2943         * cfgexpand.c (expand_gimple_basic_block): Expand label first,
2944         even if there are markers before it.
2945         * cfgrtl.c (rtl_verify_bb_layout): Reject DEBUG_INSNs outside BBs.
2947 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2949         * read-rtl.c (parse_reg_note_name): Replace Yoda conditions with
2950         typical order conditions.
2951         * sel-sched.c (extract_new_fences_from): Likewise.
2952         * config/visium/constraints.md (J, K, L): Likewise.
2953         * config/visium/predicates.md (const_shift_operand): Likewise.
2954         * config/visium/visium.c (visium_legitimize_address,
2955         visium_legitimize_reload_address): Likewise.
2956         * config/m68k/m68k.c (output_reg_adjust, emit_reg_adjust): Likewise.
2957         * config/arm/arm.c (arm_block_move_unaligned_straight): Likewise.
2958         * config/avr/constraints.md (Y01, Ym1, Y02, Ym2): Likewise.
2959         * config/avr/avr-log.c (avr_vdump, avr_log_set_avr_log,
2960         SET_DUMP_DETAIL): Likewise.
2961         * config/avr/predicates.md (const_8_16_24_operand): Likewise.
2962         * config/avr/avr.c (STR_PREFIX_P, avr_popcount_each_byte,
2963         avr_is_casesi_sequence, avr_casei_sequence_check_operands,
2964         avr_set_core_architecture, avr_set_current_function,
2965         avr_legitimize_reload_address, avr_asm_len, avr_print_operand,
2966         output_movqi, output_movsisf, avr_out_plus, avr_out_bitop,
2967         avr_out_fract, avr_adjust_insn_length, avr_encode_section_info,
2968         avr_2word_insn_p, output_reload_in_const, avr_has_nibble_0xf,
2969         avr_map_decompose, avr_fold_builtin): Likewise.
2970         * config/avr/driver-avr.c (avr_devicespecs_file): Likewise.
2971         * config/avr/gen-avr-mmcu-specs.c (str_prefix_p, print_mcu): Likewise.
2972         * config/i386/i386.c (ix86_parse_stringop_strategy_string): Likewise.
2973         * config/m32c/m32c-pragma.c (m32c_pragma_memregs): Likewise.
2974         * config/m32c/m32c.c (m32c_conditional_register_usage,
2975         m32c_address_cost): Likewise.
2976         * config/m32c/predicates.md (shiftcount_operand,
2977         longshiftcount_operand): Likewise.
2978         * config/iq2000/iq2000.c (iq2000_expand_prologue): Likewise.
2979         * config/nios2/nios2.c (nios2_handle_custom_fpu_insn_option,
2980         can_use_cdx_ldstw): Likewise.
2981         * config/nios2/nios2.h (CDX_REG_P): Likewise.
2982         * config/cr16/cr16.h (RETURN_ADDR_RTX, REGNO_MODE_OK_FOR_BASE_P):
2983         Likewise.
2984         * config/cr16/cr16.md (*mov<mode>_double): Likewise.
2985         * config/cr16/cr16.c (cr16_create_dwarf_for_multi_push): Likewise.
2986         * config/h8300/h8300.c (h8300_rtx_costs, get_shift_alg): Likewise.
2987         * config/vax/constraints.md (U06, U08, U16, CN6, S08, S16): Likewise.
2988         * config/vax/vax.c (adjacent_operands_p): Likewise.
2989         * config/ft32/constraints.md (L, b, KA): Likewise.
2990         * config/ft32/ft32.c (ft32_load_immediate, ft32_expand_prologue):
2991         Likewise.
2992         * cfgexpand.c (expand_stack_alignment): Likewise.
2993         * gcse.c (insert_expr_in_table): Likewise.
2994         * print-rtl.c (rtx_writer::print_rtx_operand_codes_E_and_V): Likewise.
2995         * cgraphunit.c (cgraph_node::expand): Likewise.
2996         * ira-build.c (setup_min_max_allocno_live_range_point): Likewise.
2997         * emit-rtl.c (add_insn): Likewise.
2998         * input.c (dump_location_info): Likewise.
2999         * passes.c (NEXT_PASS): Likewise.
3000         * read-rtl-function.c (parse_note_insn_name,
3001         function_reader::read_rtx_operand_r, function_reader::parse_mem_expr):
3002         Likewise.
3003         * sched-rgn.c (sched_rgn_init): Likewise.
3004         * diagnostic-show-locus.c (layout::show_ruler): Likewise.
3005         * combine.c (find_split_point, simplify_if_then_else, force_to_mode,
3006         if_then_else_cond, simplify_shift_const_1, simplify_comparison):
3007         Likewise.
3008         * explow.c (eliminate_constant_term): Likewise.
3009         * final.c (leaf_renumber_regs_insn): Likewise.
3010         * cfgrtl.c (print_rtl_with_bb): Likewise.
3011         * genhooks.c (emit_init_macros): Likewise.
3012         * poly-int.h (maybe_ne, maybe_le, maybe_lt): Likewise.
3013         * tree-data-ref.c (conflict_fn): Likewise.
3014         * selftest.c (assert_streq): Likewise.
3015         * expr.c (store_constructor_field, expand_expr_real_1): Likewise.
3016         * fold-const.c (fold_range_test, extract_muldiv_1, fold_truth_andor,
3017         fold_binary_loc, multiple_of_p): Likewise.
3018         * reload.c (push_reload, find_equiv_reg): Likewise.
3019         * et-forest.c (et_nca, et_below): Likewise.
3020         * dbxout.c (dbxout_symbol_location): Likewise.
3021         * reorg.c (relax_delay_slots): Likewise.
3022         * dojump.c (do_compare_rtx_and_jump): Likewise.
3023         * gengtype-parse.c (type): Likewise.
3024         * simplify-rtx.c (simplify_gen_ternary, simplify_gen_relational,
3025         simplify_const_relational_operation): Likewise.
3026         * reload1.c (do_output_reload): Likewise.
3027         * dumpfile.c (get_dump_file_info_by_switch): Likewise.
3028         * gengtype.c (type_for_name): Likewise.
3029         * gimple-ssa-sprintf.c (format_directive): Likewise.
3031 2017-12-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3033         PR target/82975
3034         * config/arm/arm.h (TEST_REGNO): Check reg_renumber is set before
3035         accessing it.  Adjust comment.
3037 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
3039         PR middle-end/81914
3040         * predict.c (zero_one_minusone): New function.
3041         (apply_return_prediction): Avoid return prediction for functions
3042         returning only -1, 0 and 1 values, unless they only return -1 and 0
3043         or 0 and 1.
3045 2017-12-19  Claudiu Zissulescu  <claziss@synopsys.com>
3047         * config/arc/arc.c (legitimate_scaled_address_p): Clean
3048         fall-through warning.
3049         (arc_compute_frame_size): Remove unused variables.
3050         (arc_print_operand): Fix fprintif format.
3051         (arc_can_follow_jump): Clean fall-through warning.
3053 2017-12-19  Marek Polacek  <polacek@redhat.com>
3055         PR c++/83489
3056         * config/i386/i386.c (init_cumulative_args): Don't check TYPE_EMPTY_P
3057         on an error node.
3059 2017-12-19  Claudiu Zissulescu  <claziss@synopsys.com>
3061         * config/arc/arc.c (overriderregs): New variable.
3062         (arc_override_options): Track fixed/call saved/call options.
3063         (arc_conditional_register_usage): Check against overrideregs
3064         variable whenever we change register properties.
3066 2017-12-19  Nathan Sidwell  <nathan@acm.org>
3068         * opts.c (finish_options): Don't prefix dump_base_name if it
3069         already contains directories.
3071 2017-12-19  Martin Liska  <mliska@suse.cz>
3073         PR rtl-optimization/82675
3074         * loop-unroll.c (unroll_loop_constant_iterations): Allocate one
3075         more element in sbitmap.
3077 2017-12-19  Paolo Carlini  <paolo.carlini@oracle.com>
3079         * gimplify.c (gimplify_expr): Use error_operand_p.
3081 2017-12-19  Sebastian Huber  <sebastian.huber@embedded-brains.de>
3083         PR target/83387
3084         * config/rs6000/t-rtems (MULTILIB_REQUIRED): Remove 64-bit soft-float
3085         multilib.
3087 2017-12-19  Daniel Cederman  <cederman@gaisler.com>
3089         * config/sparc/sparc.c (sparc_do_work_around_errata): Make sure
3090         the jump is to a label.
3092 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
3094         PR tree-optimization/83444
3095         * tree-ssa-strlen.c (strlen_check_and_optimize_stmt): Optimize
3096         character loads.
3098         PR ipa/82801
3099         PR ipa/83346
3100         * ipa-inline.c (flatten_remove_node_hook): New function.
3101         (ipa_inline): Keep only nodes with flatten attribute at the end of
3102         the array in the order from ipa_reverse_postorder, only walk that
3103         portion of array for flattening, if there is more than one such
3104         node, temporarily register a removal hook and ignore removed nodes.
3106         PR tree-optimization/80631
3107         * tree-vect-loop.c (vect_create_epilog_for_reduction): Compare
3108         induc_code against MAX_EXPR or MIN_EXPR instead of reduc_fn against
3109         IFN_REDUC_MAX or IFN_REDUC_MIN.
3111 2017-12-18  Jeff Law  <law@redhat.com>
3113         * tree-ssa-dom.c (record_equivalences_from_phis): Do not
3114         record symbolic equivalences from backedges in the CFG.
3116         Revert
3117         2017-11-19  Jeff Law  <law@redhat.com>
3119         * tree-ssa-dom.c (record_equivalences_from_phis): Fix handling
3120         of degenerates resulting from ignoring an edge.
3122 2017-12-18  Martin Sebor  <msebor@redhat.com>
3124         PR middle-end/83373
3125         PR tree-optimization/78450
3126         * tree-ssa-strlen.c (maybe_set_strlen_range): New function.
3127         (handle_builtin_strlen): Call it.
3129 2017-12-18  Segher Boessenkool  <segher@kernel.crashing.org>
3131         PR rtl-optimization/83424
3132         * rtlanal.c (dead_or_set_regno_p): Handle CLOBBER just like SET.
3134 2017-12-18  Marek Polacek  <polacek@redhat.com>
3136         PR middle-end/83463
3137         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
3138         Check if TYPE is INTEGRAL_TYPE_P before accessing its min/max
3139         values.
3141 2017-12-18  Claudiu Zissulescu  <claziss@synopsys.com>
3143         * config/arc/arc.md (maddsidi4, maddsidi4_split): Update pattern.
3144         (umaddsidi4, umaddsidi_split): Likewise.
3146 2017-12-18  Claudiu Zissulescu  <claziss@synopsys.com>
3148         * config/arc/arc.c (arc_legitimate_constant_p): Always check all
3149         constants.
3151 2017-12-18  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
3153         PR target/83420
3154         * config/s390/s390.c (s390_option_override): Avoid strncpy.
3156 2017-12-18  Richard Biener  <rguenther@suse.de>
3158         PR tree-optimization/81877
3159         * tree-ssa-loop-im.c (ref_indep_loop_p): Remove safelen parameters.
3160         (outermost_indep_loop): Adjust.
3161         (ref_indep_loop_p_1): Likewise.  Remove safelen handling again.
3162         (can_sm_ref_p): Adjust.
3164 2017-12-18  Richard Biener  <rguenther@suse.de>
3166         PR middle-end/77291
3167         * tree.c (array_at_struct_end_p): Return true if the underlying
3168         object has space for at least one element in excess of what
3169         the array domain specifies.
3171 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
3173         * doc/extend.texi (x86 Function Attributes): Reformat nocf_check
3174         example to avoid overfull hbox.
3175         * doc/invoke.texi (Option Summary): Add missing @gol.
3176         (C++ Dialect Options): Reformat -Wnoexcept-type example to avoid
3177         overfull hbox.
3179 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
3180             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3182         * doc/invoke.texi (Option Summary): Add -mverbose-cost-dump
3183         to AArch64 and ARM lists, plus missing -mflip-thumb for ARM.
3184         (AArch64 Options): Document -mverbose-cost-dump.
3185         (ARM Options): Likewise, plus -mflip-thumb.
3187 2017-12-17  Martin Sebor  <msebor@redhat.com>
3189         PR bootstrap/83446
3190         * gimple-ssa-warn-restrict.c
3191         (builtin_memref::offset_out_of_bounds): Correct the handling of
3192         anti-ranges.
3194 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
3196         * doc/invoke.texi (ARC Options): Add missing -mlra entry.
3198 2017-12-17  John David Anglin  <danglin@gcc.gnu.org>
3200         * config/pa/pa.c (pa_som_asm_init_sections): Fix comment.
3202 2017-12-17  Markus Trippelsdorf  <markus@trippelsdorf.de>
3204         * x86-tune-costs.h (skylake_cost, core_cost): Decrease r64 multiply
3205         latencies.
3207 2017-12-16  Sandra Loosemore  <sandra@codesourcery.com>
3209         * doc/invoke.texi: Fix some typos.
3211 2017-12-16  Martin Sebor  <msebor@redhat.com>
3213         PR tree-optimization/78918
3214         * Makefile.in (OBJS): Add gimple-ssa-warn-restrict.o.
3215         * builtins.c (check_sizes): Rename...
3216         (check_access): ...to this.  Rename function arguments for clarity.
3217         (check_memop_sizes): Adjust names.
3218         (expand_builtin_memchr, expand_builtin_memcpy): Same.
3219         (expand_builtin_memmove, expand_builtin_mempcpy): Same.
3220         (expand_builtin_strcat, expand_builtin_stpncpy): Same.
3221         (check_strncat_sizes, expand_builtin_strncat): Same.
3222         (expand_builtin_strncpy, expand_builtin_memset): Same.
3223         (expand_builtin_bzero, expand_builtin_memcmp): Same.
3224         (expand_builtin_memory_chk, maybe_emit_chk_warning): Same.
3225         (maybe_emit_sprintf_chk_warning): Same.
3226         (expand_builtin_strcpy): Adjust.
3227         (expand_builtin_stpcpy): Same.
3228         (expand_builtin_with_bounds): Detect out-of-bounds accesses
3229         in pointer-checking forms of memcpy, memmove, and mempcpy.
3230         (gcall_to_tree_minimal, max_object_size): Define new functions.
3231         * builtins.h (max_object_size): Declare.
3232         * calls.c (alloc_max_size): Call max_object_size instead of
3233         hardcoding ssizetype limit.
3234         (get_size_range): Handle new argument.
3235         * calls.h (get_size_range): Add a new argument.
3236         * cfgexpand.c (expand_call_stmt): Propagate no-warning bit.
3237         * doc/invoke.texi (-Wrestrict): Adjust, add example.
3238         * gimple-fold.c (gimple_fold_builtin_memory_op): Detect overlapping
3239         operations.
3240         (gimple_fold_builtin_memory_chk): Same.
3241         (gimple_fold_builtin_stxcpy_chk): New function.
3242         * gimple-ssa-warn-restrict.c: New source.
3243         * gimple-ssa-warn-restrict.h: New header.
3244         * gimple.c (gimple_build_call_from_tree): Propagate location.
3245         * passes.def (pass_warn_restrict): Add new pass.
3246         * tree-pass.h (make_pass_warn_restrict): Declare.
3247         * tree-ssa-strlen.c (handle_builtin_strcpy): Detect overlapping
3248         operations.
3249         (handle_builtin_strcat): Same.
3250         (strlen_optimize_stmt): Rename...
3251         (strlen_check_and_optimize_stmt): ...to this.  Handle strncat,
3252         stpncpy, strncpy, and their checking forms.
3254 2017-12-16  Jan Hubicka  <hubicka@ucw.cz>
3256         PR rtl-optimization/82849
3257         * modulo-sched.c (sms_schedule): Use get_estimated_loop_iterations_int
3258         and get_max_loop_iterations_int.
3260 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3261             Alan Hayward  <alan.hayward@arm.com>
3262             David Sherwood  <david.sherwood@arm.com>
3264         * machmode.h (mode_for_size, int_mode_for_size, float_mode_for_size)
3265         (smallest_mode_for_size, smallest_int_mode_for_size): Take the mode
3266         size as a poly_uint64.
3267         (mode_for_vector, mode_for_int_vector): Take the number of vector
3268         elements as a poly_uint64.
3269         * stor-layout.c (mode_for_size, smallest_mode_for_size): Take the mode
3270         size as a poly_uint64.
3271         (mode_for_vector, mode_for_int_vector): Take the number of vector
3272         elements as a poly_uint64.
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 (MACRO_MODE): New macro.
3279         * addresses.h (base_reg_class, ok_for_base_p_1): Use it.
3280         * targhooks.c (default_libcall_value, default_secondary_reload)
3281         (default_memory_move_cost, default_register_move_cost)
3282         (default_class_max_nregs): Likewise.
3284 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3285             Alan Hayward  <alan.hayward@arm.com>
3286             David Sherwood  <david.sherwood@arm.com>
3288         * doc/sourcebuild.texi: Document IN_TARGET_CODE.
3289         * genattrtab.c (write_header): Define IN_TARGET_CODE to 1 in the
3290         target C file.
3291         * genautomata.c (main): Likewise.
3292         * genconditions.c (write_header): Likewise.
3293         * genemit.c (main): Likewise.
3294         * genextract.c (print_header): Likewise.
3295         * genopinit.c (main): Likewise.
3296         * genoutput.c (output_prologue): Likewise.
3297         * genpeep.c (main): Likewise.
3298         * genpreds.c (write_insn_preds_c): Likewise.
3299         * genrecog.c (writer_header): Likewise.
3300         * config/aarch64/aarch64-builtins.c (IN_TARGET_CODE): Define.
3301         * config/aarch64/aarch64-c.c (IN_TARGET_CODE): Likewise.
3302         * config/aarch64/aarch64.c (IN_TARGET_CODE): Likewise.
3303         * config/aarch64/cortex-a57-fma-steering.c (IN_TARGET_CODE): Likewise.
3304         * config/aarch64/driver-aarch64.c (IN_TARGET_CODE): Likewise.
3305         * config/alpha/alpha.c (IN_TARGET_CODE): Likewise.
3306         * config/alpha/driver-alpha.c (IN_TARGET_CODE): Likewise.
3307         * config/arc/arc-c.c (IN_TARGET_CODE): Likewise.
3308         * config/arc/arc.c (IN_TARGET_CODE): Likewise.
3309         * config/arc/driver-arc.c (IN_TARGET_CODE): Likewise.
3310         * config/arm/aarch-common.c (IN_TARGET_CODE): Likewise.
3311         * config/arm/arm-builtins.c (IN_TARGET_CODE): Likewise.
3312         * config/arm/arm-c.c (IN_TARGET_CODE): Likewise.
3313         * config/arm/arm.c (IN_TARGET_CODE): Likewise.
3314         * config/arm/driver-arm.c (IN_TARGET_CODE): Likewise.
3315         * config/avr/avr-c.c (IN_TARGET_CODE): Likewise.
3316         * config/avr/avr-devices.c (IN_TARGET_CODE): Likewise.
3317         * config/avr/avr-log.c (IN_TARGET_CODE): Likewise.
3318         * config/avr/avr.c (IN_TARGET_CODE): Likewise.
3319         * config/avr/driver-avr.c (IN_TARGET_CODE): Likewise.
3320         * config/avr/gen-avr-mmcu-specs.c (IN_TARGET_CODE): Likewise.
3321         * config/bfin/bfin.c (IN_TARGET_CODE): Likewise.
3322         * config/c6x/c6x.c (IN_TARGET_CODE): Likewise.
3323         * config/cr16/cr16.c (IN_TARGET_CODE): Likewise.
3324         * config/cris/cris.c (IN_TARGET_CODE): Likewise.
3325         * config/darwin.c (IN_TARGET_CODE): Likewise.
3326         * config/epiphany/epiphany.c (IN_TARGET_CODE): Likewise.
3327         * config/epiphany/mode-switch-use.c (IN_TARGET_CODE): Likewise.
3328         * config/epiphany/resolve-sw-modes.c (IN_TARGET_CODE): Likewise.
3329         * config/fr30/fr30.c (IN_TARGET_CODE): Likewise.
3330         * config/frv/frv.c (IN_TARGET_CODE): Likewise.
3331         * config/ft32/ft32.c (IN_TARGET_CODE): Likewise.
3332         * config/h8300/h8300.c (IN_TARGET_CODE): Likewise.
3333         * config/i386/djgpp.c (IN_TARGET_CODE): Likewise.
3334         * config/i386/driver-i386.c (IN_TARGET_CODE): Likewise.
3335         * config/i386/driver-mingw32.c (IN_TARGET_CODE): Likewise.
3336         * config/i386/host-cygwin.c (IN_TARGET_CODE): Likewise.
3337         * config/i386/host-i386-darwin.c (IN_TARGET_CODE): Likewise.
3338         * config/i386/host-mingw32.c (IN_TARGET_CODE): Likewise.
3339         * config/i386/i386-c.c (IN_TARGET_CODE): Likewise.
3340         * config/i386/i386.c (IN_TARGET_CODE): Likewise.
3341         * config/i386/intelmic-mkoffload.c (IN_TARGET_CODE): Likewise.
3342         * config/i386/msformat-c.c (IN_TARGET_CODE): Likewise.
3343         * config/i386/winnt-cxx.c (IN_TARGET_CODE): Likewise.
3344         * config/i386/winnt-stubs.c (IN_TARGET_CODE): Likewise.
3345         * config/i386/winnt.c (IN_TARGET_CODE): Likewise.
3346         * config/i386/x86-tune-sched-atom.c (IN_TARGET_CODE): Likewise.
3347         * config/i386/x86-tune-sched-bd.c (IN_TARGET_CODE): Likewise.
3348         * config/i386/x86-tune-sched-core.c (IN_TARGET_CODE): Likewise.
3349         * config/i386/x86-tune-sched.c (IN_TARGET_CODE): Likewise.
3350         * config/ia64/ia64-c.c (IN_TARGET_CODE): Likewise.
3351         * config/ia64/ia64.c (IN_TARGET_CODE): Likewise.
3352         * config/iq2000/iq2000.c (IN_TARGET_CODE): Likewise.
3353         * config/lm32/lm32.c (IN_TARGET_CODE): Likewise.
3354         * config/m32c/m32c-pragma.c (IN_TARGET_CODE): Likewise.
3355         * config/m32c/m32c.c (IN_TARGET_CODE): Likewise.
3356         * config/m32r/m32r.c (IN_TARGET_CODE): Likewise.
3357         * config/m68k/m68k.c (IN_TARGET_CODE): Likewise.
3358         * config/mcore/mcore.c (IN_TARGET_CODE): Likewise.
3359         * config/microblaze/microblaze-c.c (IN_TARGET_CODE): Likewise.
3360         * config/microblaze/microblaze.c (IN_TARGET_CODE): Likewise.
3361         * config/mips/driver-native.c (IN_TARGET_CODE): Likewise.
3362         * config/mips/frame-header-opt.c (IN_TARGET_CODE): Likewise.
3363         * config/mips/mips.c (IN_TARGET_CODE): Likewise.
3364         * config/mmix/mmix.c (IN_TARGET_CODE): Likewise.
3365         * config/mn10300/mn10300.c (IN_TARGET_CODE): Likewise.
3366         * config/moxie/moxie.c (IN_TARGET_CODE): Likewise.
3367         * config/msp430/driver-msp430.c (IN_TARGET_CODE): Likewise.
3368         * config/msp430/msp430-c.c (IN_TARGET_CODE): Likewise.
3369         * config/msp430/msp430.c (IN_TARGET_CODE): Likewise.
3370         * config/nds32/nds32-cost.c (IN_TARGET_CODE): Likewise.
3371         * config/nds32/nds32-fp-as-gp.c (IN_TARGET_CODE): Likewise.
3372         * config/nds32/nds32-intrinsic.c (IN_TARGET_CODE): Likewise.
3373         * config/nds32/nds32-isr.c (IN_TARGET_CODE): Likewise.
3374         * config/nds32/nds32-md-auxiliary.c (IN_TARGET_CODE): Likewise.
3375         * config/nds32/nds32-memory-manipulation.c (IN_TARGET_CODE): Likewise.
3376         * config/nds32/nds32-pipelines-auxiliary.c (IN_TARGET_CODE): Likewise.
3377         * config/nds32/nds32-predicates.c (IN_TARGET_CODE): Likewise.
3378         * config/nds32/nds32.c (IN_TARGET_CODE): Likewise.
3379         * config/nios2/nios2.c (IN_TARGET_CODE): Likewise.
3380         * config/nvptx/mkoffload.c (IN_TARGET_CODE): Likewise.
3381         * config/nvptx/nvptx.c (IN_TARGET_CODE): Likewise.
3382         * config/pa/pa.c (IN_TARGET_CODE): Likewise.
3383         * config/pdp11/pdp11.c (IN_TARGET_CODE): Likewise.
3384         * config/powerpcspe/driver-powerpcspe.c (IN_TARGET_CODE): Likewise.
3385         * config/powerpcspe/host-darwin.c (IN_TARGET_CODE): Likewise.
3386         * config/powerpcspe/host-ppc64-darwin.c (IN_TARGET_CODE): Likewise.
3387         * config/powerpcspe/powerpcspe-c.c (IN_TARGET_CODE): Likewise.
3388         * config/powerpcspe/powerpcspe-linux.c (IN_TARGET_CODE): Likewise.
3389         * config/powerpcspe/powerpcspe.c (IN_TARGET_CODE): Likewise.
3390         * config/riscv/riscv-builtins.c (IN_TARGET_CODE): Likewise.
3391         * config/riscv/riscv-c.c (IN_TARGET_CODE): Likewise.
3392         * config/riscv/riscv.c (IN_TARGET_CODE): Likewise.
3393         * config/rl78/rl78-c.c (IN_TARGET_CODE): Likewise.
3394         * config/rl78/rl78.c (IN_TARGET_CODE): Likewise.
3395         * config/rs6000/driver-rs6000.c (IN_TARGET_CODE): Likewise.
3396         * config/rs6000/host-darwin.c (IN_TARGET_CODE): Likewise.
3397         * config/rs6000/host-ppc64-darwin.c (IN_TARGET_CODE): Likewise.
3398         * config/rs6000/rs6000-c.c (IN_TARGET_CODE): Likewise.
3399         * config/rs6000/rs6000-linux.c (IN_TARGET_CODE): Likewise.
3400         * config/rs6000/rs6000-p8swap.c (IN_TARGET_CODE): Likewise.
3401         * config/rs6000/rs6000-string.c (IN_TARGET_CODE): Likewise.
3402         * config/rs6000/rs6000.c (IN_TARGET_CODE): Likewise.
3403         * config/rx/rx.c (IN_TARGET_CODE): Likewise.
3404         * config/s390/driver-native.c (IN_TARGET_CODE): Likewise.
3405         * config/s390/s390-c.c (IN_TARGET_CODE): Likewise.
3406         * config/s390/s390.c (IN_TARGET_CODE): Likewise.
3407         * config/sh/sh-c.c (IN_TARGET_CODE): Likewise.
3408         * config/sh/sh-mem.cc (IN_TARGET_CODE): Likewise.
3409         * config/sh/sh.c (IN_TARGET_CODE): Likewise.
3410         * config/sh/sh_optimize_sett_clrt.cc (IN_TARGET_CODE): Likewise.
3411         * config/sh/sh_treg_combine.cc (IN_TARGET_CODE): Likewise.
3412         * config/sparc/driver-sparc.c (IN_TARGET_CODE): Likewise.
3413         * config/sparc/sparc-c.c (IN_TARGET_CODE): Likewise.
3414         * config/sparc/sparc.c (IN_TARGET_CODE): Likewise.
3415         * config/spu/spu-c.c (IN_TARGET_CODE): Likewise.
3416         * config/spu/spu.c (IN_TARGET_CODE): Likewise.
3417         * config/stormy16/stormy16.c (IN_TARGET_CODE): Likewise.
3418         * config/tilegx/mul-tables.c (IN_TARGET_CODE): Likewise.
3419         * config/tilegx/tilegx-c.c (IN_TARGET_CODE): Likewise.
3420         * config/tilegx/tilegx.c (IN_TARGET_CODE): Likewise.
3421         * config/tilepro/mul-tables.c (IN_TARGET_CODE): Likewise.
3422         * config/tilepro/tilepro-c.c (IN_TARGET_CODE): Likewise.
3423         * config/tilepro/tilepro.c (IN_TARGET_CODE): Likewise.
3424         * config/v850/v850-c.c (IN_TARGET_CODE): Likewise.
3425         * config/v850/v850.c (IN_TARGET_CODE): Likewise.
3426         * config/vax/vax.c (IN_TARGET_CODE): Likewise.
3427         * config/visium/visium.c (IN_TARGET_CODE): Likewise.
3428         * config/vms/vms-c.c (IN_TARGET_CODE): Likewise.
3429         * config/vms/vms-f.c (IN_TARGET_CODE): Likewise.
3430         * config/vms/vms.c (IN_TARGET_CODE): Likewise.
3431         * config/xtensa/xtensa.c (IN_TARGET_CODE): Likewise.
3433 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3434             Alan Hayward  <alan.hayward@arm.com>
3435             David Sherwood  <david.sherwood@arm.com>
3437         * tree-ssa-sccvn.c (vn_reference_lookup_3): Avoid repeated
3438         checks for MEM_REF.
3440 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3441             Alan Hayward  <alan.hayward@arm.com>
3442             David Sherwood  <david.sherwood@arm.com>
3444         * doc/generic.texi (VEC_SERIES_EXPR): Document.
3445         * doc/md.texi (vec_series@var{m}): Document.
3446         * tree.def (VEC_SERIES_EXPR): New tree code.
3447         * tree.h (build_vec_series): Declare.
3448         * tree.c (build_vec_series): New function.
3449         * cfgexpand.c (expand_debug_expr): Handle VEC_SERIES_EXPR.
3450         * tree-pretty-print.c (dump_generic_node): Likewise.
3451         * gimple-pretty-print.c (dump_binary_rhs): Likewise.
3452         * tree-inline.c (estimate_operator_cost): Likewise.
3453         * expr.c (expand_expr_real_2): Likewise.
3454         * optabs-tree.c (optab_for_tree_code): Likewise.
3455         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
3456         * fold-const.c (const_binop): Fold VEC_SERIES_EXPRs of constants.
3457         * expmed.c (make_tree): Handle VEC_SERIES.
3458         * optabs.def (vec_series_optab): New optab.
3459         * optabs.h (expand_vec_series_expr): Declare.
3460         * optabs.c (expand_vec_series_expr): New function.
3461         * tree-vect-generic.c (expand_vector_operations_1): Check that
3462         the operands also have vector type.
3464 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3465             Alan Hayward  <alan.hawyard@arm.com>
3466             David Sherwood  <david.sherwood@arm.com>
3468         * doc/generic.texi (VEC_DUPLICATE_EXPR): Document.
3469         (VEC_COND_EXPR): Add missing @tindex.
3470         * doc/md.texi (vec_duplicate@var{m}): Document.
3471         * tree.def (VEC_DUPLICATE_EXPR): New tree codes.
3472         * tree.c (build_vector_from_val): Add stubbed-out handling of
3473         variable-length vectors, using VEC_DUPLICATE_EXPR.
3474         (uniform_vector_p): Handle VEC_DUPLICATE_EXPR.
3475         * cfgexpand.c (expand_debug_expr): Likewise.
3476         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
3477         * tree-inline.c (estimate_operator_cost): Likewise.
3478         * tree-pretty-print.c (dump_generic_node): Likewise.
3479         * tree-vect-generic.c (ssa_uniform_vector_p): Likewise.
3480         * fold-const.c (const_unop): Fold VEC_DUPLICATE_EXPRs of a constant.
3481         (test_vec_duplicate_folding): New function.
3482         (fold_const_c_tests): Call it.
3483         * optabs.def (vec_duplicate_optab): New optab.
3484         * optabs-tree.c (optab_for_tree_code): Handle VEC_DUPLICATE_EXPR.
3485         * optabs.h (expand_vector_broadcast): Declare.
3486         * optabs.c (expand_vector_broadcast): Make non-static.  Try using
3487         vec_duplicate_optab.
3488         * expr.c (store_constructor): Try using vec_duplicate_optab for
3489         uniform vectors.
3490         (expand_expr_real_2): Handle VEC_DUPLICATE_EXPR.
3492 2017-12-15  Markus Trippelsdorf  <markus@trippelsdorf.de>
3494         PR target/83358
3495         * config/i386/x86-tune-costs.h (skylake_cost, core_cost): Increase
3496         div/mod latencies a bit.
3498 2017-12-15  Jeff Law  <law@redhat.com>
3500         PR tree-optimization/36550
3501         * tree-ssa-threadupdate.c (count_stmts_and_phis_in_block): New.
3502         (mark_threaded_blocks): Rewrite code to avoid block copying when
3503         optimizing for size.  Don't pessimize blocks which will be
3504         copied, but all the statements will be dead.
3506 2017-12-15  Alexandre Oliva  <aoliva@redhat.com>
3508         PR tree-optimization/81165
3509         * tree-ssa-threadupdate.c (uses_in_bb): New.
3510         (estimate_threading_killed_stmts): New.
3511         * tree-ssa-threadupdate.h (estimate_threading_killed_stmts): Prototype.
3512         * tree-ssa-threadedge.c
3513         (record_temporary_equivalences_from_stmts_at_dest): Expand limit
3514         when its hit.
3516 2017-12-15  Jeff Law  <law@redhat.com>
3518         PR tree-optimization/83410
3519         * tree-ssa-threadupdate.c (thread_block_1): Avoid certain jump
3520         threads when parallelizing loops.
3522 2017-12-15  Jakub Jelinek  <jakub@redhat.com>
3524         * tree-core.h (struct attribute_spec): Swap affects_type_identity and
3525         handler fields.
3526         * config/alpha/alpha.c (vms_attribute_table): Swap
3527         affects_type_identity and handler fields, adjust comments.
3528         * config/mips/mips.c (mips_attribute_table): Likewise.
3529         * config/visium/visium.c (visium_attribute_table): Likewise.
3530         * config/epiphany/epiphany.c (epiphany_attribute_table): Likewise.
3531         * config/microblaze/microblaze.c (microblaze_attribute_table):
3532         Likewise.
3533         * config/spu/spu.c (spu_attribute_table): Likewise.
3534         * config/mcore/mcore.c (mcore_attribute_table): Likewise.
3535         * config/arc/arc.c (arc_attribute_table): Likewise.
3536         * config/m68k/m68k.c (m68k_attribute_table): Likewise.
3537         * config/v850/v850.c (v850_handle_interrupt_attribute,
3538         v850_handle_data_area_attribute): Formatting fixes.
3539         (v850_attribute_table): Swap affects_type_identity and handler
3540         fields, adjust comments.
3541         * config/m32r/m32r.c (m32r_attribute_table): Likewise.
3542         * config/arm/arm.c (arm_attribute_table): Likewise.
3543         * config/avr/avr.c (avr_attribute_table): Likewise.
3544         * config/s390/s390.c (s390_attribute_table): Likewise.
3545         * config/sh/sh.c (sh_attribute_table): Likewise.
3546         * config/i386/i386.c (ix86_handle_cconv_attribute,
3547         ix86_handle_callee_pop_aggregate_return): Formatting fixes.
3548         (ix86_attribute_table): Swap affects_type_identity and handler
3549         fields, adjust comments.
3550         * config/i386/cygming.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
3551         * config/sparc/sparc.c (sparc_attribute_table): Likewise.
3552         * config/m32c/m32c.c (m32c_attribute_table): Likewise.
3553         * config/sol2.h (SOLARIS_ATTRIBUTE_TABLE): Likewise.
3554         * config/ia64/ia64.c (ia64_attribute_table): Likewise.
3555         * config/msp430/msp430.c (msp430_attribute_table): Likewise.
3556         * config/rx/rx.c (rx_attribute_table): Likewise.
3557         * config/cr16/cr16.c (cr16_attribute_table): Likewise.
3558         * config/h8300/h8300.c (h8300_attribute_table): Likewise.
3559         * config/nvptx/nvptx.c (nvptx_attribute_table): Likewise.
3560         * config/powerpcspe/powerpcspe.c (rs6000_attribute_table): Likewise.
3561         * config/darwin.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
3562         * config/stormy16/stormy16.c (xstormy16_attribute_table): Likewise.
3563         * config/bfin/bfin.c (bfin_attribute_table): Likewise.
3564         * config/rs6000/rs6000.c (rs6000_attribute_table): Likewise.
3565         * config/rl78/rl78.c (rl78_attribute_table): Likewise.
3566         * config/nds32/nds32.c (nds32_attribute_table): Likewise.
3567         * doc/plugins.texi (user_attr): Likewise.  Add NULL for
3568         exclude.
3569         * attribs.c (empty_attribute_table): Swap affects_type_identity and
3570         handler fields.
3571         (register_scoped_attributes, decl_attributes): Formatting fixes.
3573         PR tree-optimization/83269
3574         * fold-const.c (fold_binary_loc): Perform (-A) - B -> (-B) - A
3575         subtraction in arg0's type if type is signed and arg0 is unsigned.
3576         Formatting fix.
3578         PR sanitizer/81281
3579         * match.pd ((T)(P + A) - (T)P -> (T) A): Use @@0 instead of @0 and
3580         convert? on @0 instead of convert.  Check type of @1, not @0.
3581         ((T)P - (T)(P + A) -> -(T) A): Use @@0 instead of @0 and
3582         convert? on @0 instead of convert.  Check type of @1, not @0.
3583         ((T)(P + A) - (T)(P + B) -> (T)A - (T)B): Use @@0 instead of @0,
3584         only optimize if either both @1 and @2 types are narrower
3585         precision, or both are wider or equal precision, and in the former
3586         case only if both have undefined overflow.
3588 2017-12-15  Richard Biener  <rguenther@suse.de>
3590         PR lto/83388
3591         * internal-fn.def (IFN_NOP): Add.
3592         * internal-fn.c (expand_NOP): Do nothing.
3593         * lto-streamer-in.c (input_function): Instead of removing
3594         sanitizer calls replace them with IFN_NOP calls.
3596 2017-12-15  Richard Sandiford  <richard.sandiford@linaro.org>
3597             Alan Hayward  <alan.hayward@arm.com>
3598             David Sherwood  <david.sherwood@arm.com>
3600         * dse.c (store_info, read_info_type): Replace begin and end with
3601         offset and width.
3602         (print_range): New function.
3603         (set_all_positions_unneeded, any_positions_needed_p)
3604         (check_mem_read_rtx, scan_stores, scan_reads, dse_step5): Update
3605         accordingly.
3606         (record_store): Likewise.  Optimize the case in which all positions
3607         are unneeded.
3608         (get_stored_val): Replace read_begin and read_end with read_offset
3609         and read_width.
3610         (replace_read): Update call accordingly.
3612 2017-12-15  Bin Cheng  <bin.cheng@arm.com>
3614         * gimple-loop-interchange.cc (STMT_COST_RATIO): New macro.
3615         (loop_cand::m_num_stmts, loop_cand::m_const_init_reduc): New members.
3616         (loop_cand::loop_cand): Initialize above members.
3617         (loop_cand::supported_operations): Delete.
3618         (loop_cand::can_interchange_p): Inline above function.
3619         (loop_cand::classify_simple_reduction): Record number of constant
3620         initialized simple reductions.
3621         (should_interchange_loops): New parameters.  Check stmt cost of loops
3622         to be interchange.
3623         (tree_loop_interchange::interchange): Prepare stmt cost of outer loop.
3624         Update call to should_interchange_loops.
3625         (should_interchange_loop_nest): Update call to
3626         should_interchange_loops.
3628 2017-12-15  Eric Botcazou  <ebotcazou@adacore.com>
3630         PR target/66488
3631         * ggc-page.c (HOST_BITS_PER_PTR): Do not define here...
3632         * hwint.h (HOST_BITS_PER_PTR): ...but here instead.
3633         * config/i386/xm-mingw32.h (HOST_BITS_PER_PTR): Delete.
3635 2017-12-15  Julia Koval  <julia.koval@intel.com>
3637         * config/i386/i386-builtin.def (__builtin_ia32_vaesenclast_v16qi,
3638         __builtin_ia32_vaesenclast_v32qi, __builtin_ia32_vaesenclast_v64qi):
3639         New.
3640         * config/i386/sse.md (vaesenclast_<mode>): New pattern.
3641         * config/i386/vaesintrin.h (_mm256_aesenclast_epi128,
3642         _mm512_aesenclast_epi128, _mm_aesenclast_epi128): New intrinsics.
3644 2017-12-15  Julia Koval  <julia.koval@intel.com>
3646         * config/i386/i386-builtin.def (__builtin_ia32_vaesenc_v16qi,
3647         __builtin_ia32_vaesenc_v32qi, __builtin_ia32_vaesenc_v64qi): New.
3648         * config/i386/sse.md (vaesenc_<mode>): New pattern.
3649         * config/i386/vaesintrin.h (_mm256_aesenc_epi128, _mm512_aesenc_epi128,
3650         _mm_aesenc_epi128): New intrinsics.
3652 2017-12-15  Julia Koval  <julia.koval@intel.com>
3654         * config/i386/i386-builtin.def (__builtin_ia32_vaesdeclast_v16qi,
3655         __builtin_ia32_vaesdeclast_v32qi, __builtin_ia32_vaesdeclast_v64qi):
3656         New.
3657         * config/i386/sse.md (vaesdeclast_<mode>): New pattern.
3658         * config/i386/vaesintrin.h (_mm256_aesdeclast_epi128,
3659         _mm512_aesdeclast_epi128, _mm_aesdeclast_epi128): New intrinsics.
3661 2017-12-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3663         * gimple-ssa-strength-reduction.c (analyze_increments):
3664         Distinguish replacement costs for constant strides from those for
3665         unknown strides.
3667 2017-12-14  Jakub Jelinek  <jakub@redhat.com>
3669         * var-tracking.c (variable_tracking_main_1): Formatting fix.
3671 2017-12-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3673         * doc/invoke.texi: Document -Wcast-function-type.
3674         * recog.h (stored_funcptr): Change signature.
3675         * tree-dump.c (dump_node): Avoid warning.
3676         * typed-splay-tree.h (typed_splay_tree): Avoid warning.
3678 2017-12-14  Qing Zhao  <qing.zhao@oracle.com>
3680         PR middle_end/79538
3681         * gimple-fold.c (get_range_strlen): Add the handling of non-member 
3682         array.
3684 2017-12-14  David Malcolm  <dmalcolm@redhat.com>
3686         PR tree-optimization/83312
3687         * domwalk.h (dom_walker::dom_walker): Fix typo in comment.
3688         * tree-cfg.c (find_taken_edge): Update to handle NULL_TREE for
3689         "val" param, and to cope with arbitrary basic blocks.
3690         (find_taken_edge_cond_expr): Add "cond_stmt" param and use it to
3691         handle NULL_TREE for "val", dropping "bb" param.
3692         (find_taken_edge_switch_expr): Make "switch_stmt" param const and
3693         drop "bb" param.  Handle NULL_TREE for "val".
3694         (find_case_label_for_value): Make "switch_stmt" param const.
3695         * tree-vrp.c (class check_array_bounds_dom_walker): New subclass
3696         of dom_walker.
3697         (vrp_prop::check_all_array_refs): Reimplement as...
3698         (check_array_bounds_dom_walker::before_dom_children): ...this new
3699         vfunc.  Replace linear search through BB block list, excluding
3700         those with non-executable in-edges via dominator walk.
3702 2017-12-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3704         * config/arm/arm.opt (mverbose-cost-dump): New option.
3705         * config/arm/arm.c (arm_rtx_costs): Use it.
3707 2017-12-14  Andreas Schwab  <schwab@linux-m68k.org>
3709         PR bootstrap/83396
3710         * reload1.c (emit_input_reload_insns): Skip debug markers.
3712 2017-12-14  Alexandre Oliva  <aoliva@redhat.com>
3714         * config/i386/i386.c (rest_of_insert_endbranch): Use call loc
3715         for its nop_endbr.
3717         PR bootstrap/83396
3718         * config/arc/arc.c (hwloop_optimize): Skip debug insns.
3719         * config/sh/sh-protos.h (sh_find_set_of_reg): Adjust.
3720         * config/sh/sh.c: Skip debug insns besides notes.
3721         * config/sh/sh.md: Likewise.
3722         * config/sh/sh_treg_combine.cc: Likewise.
3723         * config/sh/sync.md: Likewise.
3725 2017-12-14  Tom de Vries  <tom@codesourcery.com>
3727         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
3728         Add item for weak_undefined.
3730 2017-12-14  Richard Biener  <rguenther@suse.de>
3732         PR tree-optimization/67842
3733         * sese.h (bb_in_region): Remove #if 0'ed code.
3735 2017-12-14  Richard Biener  <rguenther@suse.de>
3737         PR tree-optimization/83326
3738         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Add
3739         may_be_zero parameter and handle it by not marking the first
3740         peeled copy as not exiting the loop.
3741         (try_peel_loop): Likewise.
3742         (canonicalize_loop_induction_variables): Use number_of_iterations_exit
3743         to handle the case of constant or zero iterations and perform
3744         loop header copying on-the-fly.
3746 2017-12-14  Richard Biener  <rguenther@suse.de>
3748         PR tree-optimization/83418
3749         * vr-values.c (vr_values::extract_range_for_var_from_comparison_expr):
3750         Instead of asserting we don't get unfolded comparisons deal with them.
3752 2017-12-14  Jakub Jelinek  <jakub@redhat.com>
3754         PR bootstrap/83396
3755         * var-tracking.c (vt_initialize): Ignore non-DEBUG_INSNs outside of
3756         basic blocks.  Assert debug bind insns don't appear outside of bbs,
3757         don't reset them.  Assert insns without BLOCK_FOR_INSN are outside of
3758         bb.  Simplify.
3760         PR tree-optimization/83198
3761         * gimple-ssa-sprintf.c (format_floating): Set type solely based on
3762         dir.modifier, regardless of TREE_TYPE (arg).  Assume non-REAL_CST
3763         value if arg is a REAL_CST with incompatible type.
3765 2017-12-14  Sudakshina Das  <sudi.das@arm.com>
3766             Bin Cheng  <bin.cheng@arm.com>
3768         PR target/81228
3769         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Move LTGT
3770         to CCFPEmode.
3771         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Add
3772         LTGT.
3774 2017-12-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3776         * config/aarch64/aarch64-cores.def (cortex-a55, cortex-a75,
3777         cortex-a75.cortex-a55): Specify AARCH64_FL_F16 in the arch features.
3779 2017-12-14  Richard Sandiford  <richard.sandiford@linaro.org>
3780             Alan Hayward  <alan.hayward@arm.com>
3781             David Sherwood  <david.sherwood@arm.com>
3783         * poly-int.h: New file.
3784         * poly-int-types.h: Likewise.
3785         * coretypes.h: Include them.
3786         (POLY_INT_CONVERSION): Define.
3787         * target.def (estimated_poly_value): New hook.
3788         * doc/tm.texi.in (TARGET_ESTIMATED_POLY_VALUE): New hook.
3789         * doc/tm.texi: Regenerate.
3790         * doc/poly-int.texi: New file.
3791         * doc/gccint.texi: Include it.
3792         * doc/rtl.texi: Describe restrictions on subreg modes.
3793         * Makefile.in (TEXI_GCCINT_FILES): Add poly-int.texi.
3794         * genmodes.c (NUM_POLY_INT_COEFFS): Provide a default definition.
3795         (emit_insn_modes_h): Emit a definition of NUM_POLY_INT_COEFFS.
3796         * targhooks.h (default_estimated_poly_value): Declare.
3797         * targhooks.c (default_estimated_poly_value): New function.
3798         * target.h (estimated_poly_value): Likewise.
3799         * wide-int.h (WI_UNARY_RESULT): Use wi::binary_traits.
3800         (wi::unary_traits): Delete.
3801         (wi::binary_traits::signed_shift_result_type): Define for
3802         offset_int << HOST_WIDE_INT, etc.
3803         (generic_wide_int::operator <<=): Define for all types and use
3804         wi::lshift instead of <<.
3805         (wi::hwi_with_prec): Add a default constructor.
3806         (wi::ints_for): New class.
3807         (operator <<): Define for all wide-int types.
3808         (operator /): New function.
3809         (operator %): Likewise.
3810         * selftest.h (ASSERT_KNOWN_EQ, ASSERT_KNOWN_EQ_AT, ASSERT_MAYBE_NE)
3811         (ASSERT_MAYBE_NE_AT): New macros.
3813 2017-12-13  Eric Botcazou  <ebotcazou@adacore.com>
3814             Dominik Vogt  <vogt@linux.vnet.ibm.com>
3816         PR middle-end/78468
3817         * emit-rtl.c (init_emit): Remove ??? comment.
3818         * explow.c (get_dynamic_stack_size): Take known alignment of stack
3819         pointer + STACK_DYNAMIC_OFFSET into account in lieu of STACK_BOUNDARY.
3820         * config/sparc/sparc.h (INIT_EXPANDERS): In 32-bit mode, lower the
3821         alignment of 3 virtual registers to BITS_PER_WORD.
3823         * config/sparc/sparc.c (sparc_compute_frame_size): Simplify.
3825 2017-12-13  Peter Bergner  <bergner@vnet.ibm.com>
3827         * config/rs6000/ppc-auxv.h (PPC_FEATURE2_HTM_NO_SUSPEND): New define.
3828         * config/rs6000/rs6000.c (cpu_supports_info): Use it.
3830 2017-12-13  Alexandre Oliva  <aoliva@redhat.com>
3832         PR bootstrap/83396
3833         * reload1.c (eliminate_regs_in_insn): Skip debug markers.
3835 2017-12-13  Jakub Jelinek  <jakub@redhat.com>
3837         * tree-cfg.c (verify_gimple_in_cfg): Verify no non-label stmts
3838         with the exception of debug begin stmt markers appear before
3839         labels.
3841         PR bootstrap/83396
3842         * final.c (rest_of_handle_final): Call variable_tracking_main only
3843         if !flag_var_tracking.
3845 2017-12-13  Alexandre Oliva  <aoliva@redhat.com>
3846             Jakub Jelinek  <jakub@redhat.com>
3848         PR bootstrap/83396
3849         PR debug/83391
3850         * tree-cfgcleanup.c (remove_forwarder_block): Keep after
3851         labels debug stmts that can only appear after labels.
3853 2017-12-13  Alexander Monakov  <amonakov@ispras.ru>
3855         PR rtl-optimization/82398
3856         * sel-sched.c (sel_rank_for_schedule): Fix check for zero
3857         EXPR_USEFULNESS in priority comparison.
3859 2017-12-13  Segher Boessenkool  <segher@kernel.crashing.org>
3861         PR rtl-optimization/83393
3862         * combine.c (move_deaths): If reg_stat points to a too new insn in
3863         last_death, do not use it: find the proper insn instead.
3865 2017-12-12  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
3867         PR target/82190
3868         * config/rs6000/rs6000-string.c (expand_block_compare,
3869         expand_strn_compare): Fix set_mem_size() calls.
3871 2017-12-12  Jeff Law  <law@redhat.com>
3873         PR tree-optimization/83298
3874         PR tree-optimization/83362
3875         PR tree-optimization/83383
3876         * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Make
3877         push_value_range a public interface.  Add new argument to
3878         record_ranges_from_stmt.
3879         * gimple-ssa-evrp-analyze.c
3880         (evrp_range_analyzer::record_ranges_from_stmt): Add new argument.
3881         Update comments.  Handle recording temporary equivalences.
3882         * tree-ssa-dom.c (dom_opt_opt_walker::before_dom_children): Add
3883         new argument to call to evrp_range_analyzer::record_ranges_from_stmt.
3884         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Likewise.
3885         * tree-ssa-threadedge.c: Include alloc-pool.h, vr-values.h and
3886         gimple-ssa-evrp-analyze.h.
3887         (record_temporary_equivalences_from_phis): Add new argument.  When
3888         the PHI arg is an SSA_NAME, set the result's range to the range
3889         of the PHI arg.
3890         (record_temporary_equivalences_from_stmts_at_dest): Record ranges
3891         from statements too.
3892         (thread_through_normal_block): Accept new argument, evrp_range_analyzer.
3893         Pass it down to children as needed.
3894         (thread_outgoing_edges): Likewise.
3895         (thread_across_edge): Likewise.   Push/pop range state as needed.
3896         * tree-ssa-threadedge.h (thread_outgoing_edges): Update prototype.
3898 2017-12-12  Julia Koval  <julia.koval@intel.com>
3900         * config/i386/i386.c (PTA_SKYLAKE_AVX512): Add PTA_CLWB.
3901         (PTA_CANNONLAKE): Remove PTA_CLWB.
3903 2017-12-12  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3905         PR target/83332
3906         * config/rs6000/vector.md (vcondv2dfv2di): New define_expand.
3907         (vcondv2div2df): Likewise.
3908         (vconduv2dfv2di): Likewise.
3910 2017-12-12  Jakub Jelinek  <jakub@redhat.com>
3912         * builtins.def (BUILT_IN_NEXTAFTER, BUILT_IN_NEXTAFTERF,
3913         BUILT_IN_NEXTAFTERL, BUILT_IN_NEXTTOWARD, BUILT_IN_NEXTTOWARDF,
3914         BUILT_IN_NEXTTOWARDL): Use ATTR_MATHFN_ERRNO instead of
3915         ATTR_MATHFN_FPROUNDING_ERRNO.
3917 2017-12-12  Richard Biener  <rguenther@suse.de>
3919         PR tree-optimization/83385
3920         * graphite-scop-detection.c (get_order, order): Remove.
3921         (bb_to_rpo): New global.
3922         (cmp_pbbs): Adjust.
3923         (build_scops): Sort pbbs in RPO order.
3925 2017-12-12  James Greenhalgh  <james.greenhalgh@arm.com>
3927         * combine.c (simplify_set): Do not transform subregs to zero_extends
3928         if the destination mode is a vector mode.
3930 2017-12-12  Jakub Jelinek  <jakub@redhat.com>
3932         PR tree-optimization/83359
3933         * tree-cfg.h (fold_loop_internal_call): Declare.
3934         * tree-vectorizer.c (fold_loop_internal_call): Moved to ...
3935         * tree-cfg.c (fold_loop_internal_call): ... here.  No longer static.
3936         (find_loop_dist_alias): New function.
3937         (move_sese_region_to_fn): If any dloop->orig_loop_num value is
3938         updated, also adjust any corresponding LOOP_DIST_ALIAS internal
3939         calls.
3941         PR tree-optimization/80631
3942         * tree-vect-loop.c (get_initial_def_for_reduction): Fix comment typo.
3943         (vect_create_epilog_for_reduction): Add INDUC_VAL and INDUC_CODE
3944         arguments, for INTEGER_INDUC_COND_REDUCTION use INDUC_VAL instead of
3945         hardcoding zero as the value if COND_EXPR is never true.  For
3946         INTEGER_INDUC_COND_REDUCTION don't emit the final COND_EXPR if
3947         INDUC_VAL is equal to INITIAL_DEF, and use INDUC_CODE instead of
3948         hardcoding MAX_EXPR as the reduction operation.
3949         (is_nonwrapping_integer_induction): Allow negative step.
3950         (vectorizable_reduction): Compute INDUC_VAL and INDUC_CODE for
3951         vect_create_epilog_for_reduction, if no value is suitable, don't
3952         use INTEGER_INDUC_COND_REDUCTION for now.  Formatting fixes.
3954 2017-12-12  Richard Biener  <rguenther@suse.de>
3956         PR tree-optimization/81889
3957         * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness): Use
3958         range info from the non-wrapping IV instead of just the range
3959         of the type.
3961 2017-12-12  Julia Koval  <julia.koval@intel.com>
3963         * config.gcc: Add vaesintrin.h.
3964         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI_V64QI):
3965         New type.
3966         * config/i386/i386-builtin.def (__builtin_ia32_vaesdec_v16qi,
3967         __builtin_ia32_vaesdec_v32qi, __builtin_ia32_vaesdec_v64qi):
3968         New builtins.
3969         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
3970         * config/i386/immintrin.h: Include vaesintrin.h.
3971         * config/i386/sse.md (vaesdec_<mode>): New pattern.
3972         * config/i386/vaesintrin.h (_mm256_aesdec_epi128, _mm512_aesdec_epi128,
3973         _mm_aesdec_epi128): New intrinsics.
3975 2017-12-12  Julia Koval  <julia.koval@intel.com>
3977         * common/config/i386/i386-common.c (OPTION_MASK_ISA_VAES_SET,
3978         OPTION_MASK_ISA_VAES_UNSET): New.
3979         (ix86_handle_option): Handle -mvaes.
3980         * config/i386/cpuid.h: Define bit_VAES.
3981         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mvaes.
3982         * config/i386/i386-c.c (__VAES__): New.
3983         * config/i386/i386.c (ix86_target_string): Add -mvaes.
3984         (ix86_valid_target_attribute_inner_p): Ditto.
3985         * config/i386/i386.h (TARGET_VAES, TARGET_VAES_P): New.
3986         * config/i386/i386.opt: Add -mvaes.
3987         * doc/invoke.texi: Ditto.
3989 2017-12-12  Alexandre Oliva  <aoliva@redhat.com>
3991         * debug.h (gcc_debug_hooks): Add inline_entry.
3992         * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Likewise.
3993         * debug.c (do_nothing_debug_hooks): Likewise.
3994         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
3995         * dwarf2out.c (dwarf2_debug_hooks): Likewise.
3996         (dwarf2_lineno_debug_hooks): Likewise.
3998         * common.opt (gstatement-frontiers): New, setting
3999         debug_nonbind_markers_p.
4000         * rtl.h (MAY_HAVE_DEBUG_MARKER_INSNS): Activate.
4001         * toplev.c (process_options): Autodetect value for debug statement
4002         frontiers option.
4003         * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): Activate.
4004         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers): New.
4006         * cfgexpand.c (expand_gimple_basic_block): Handle begin stmt
4007         markers.  Integrate source bind into debug stmt expand loop.
4008         (pass_expand::execute): Check debug marker limit.  Avoid deep
4009         TER and expand debug locations for debug bind insns only.
4010         * cse.c (insn_live_p): Keep nonbind markers and debug bindings
4011         followed by them.
4012         * df-scan.c (df_insn_delete): Accept out-of-block debug insn.
4013         * final.c (reemit_insn_block_notes): Take current block from
4014         nonbind markers.  Declare note where it's first set.
4015         (final_scan_insn): Handle begin stmt notes.  Emit is_stmt according to
4016         begin stmt markers if enabled.
4017         (notice_source_line): Handle nonbind markers.  Fail if their
4018         location is unknown or that of builtins.
4019         (rest_of_handle_final): Convert begin stmt markers to notes if
4020         var-tracking didn't run.
4021         (rest_of_clean_state): Skip begin stmt markers.
4022         * gimple-pretty-print.c (dump_gimple_debug): Handle begin stmt
4023         markers.
4024         * function.c (allocate_struct_function): Set begin_stmt_markers.
4025         * function.h (struct function): Add debug_marker_count counter
4026         and debug_nonbind_markers flag.
4027         * gimple-iterator.c (gsi_remove): Adjust debug_marker_count.
4028         * gimple-low.c (lower_function_body): Adjust
4029         debug_nonbind_markers.
4030         (lower_stmt): Drop or skip gimple debug stmts.
4031         (lower_try_catch): Skip debug stmts.
4032         * gimple.c (gimple_build_debug_begin_stmt): New.
4033         (gimple_copy): Increment debug_marker_count if copying one.
4034         * gimple.h (gimple_build_debug_begin_stmt): Declare.
4035         * gimplify.c (rexpr_location): New.
4036         (rexpr_has_location): New.
4037         (warn_switch_unreachable_r): Handle gimple debug stmts.
4038         (shortcut_cond_r): Call expr_location.
4039         (find_goto): New.
4040         (find_goto_label): New.
4041         (shortcut_cond_expr): Call expr_has_location, expr_location, and
4042         find_goto_label.
4043         (gimplify_cond_expr): Call find_goto_label, expr_has_location, and
4044         expr_location.
4045         (gimplify_expr): Handle begin stmt markers.  Reject debug expr decls.
4046         * langhooks-def.h (LANG_HOOKS_EMITS_BEGIN_STMT): New.  Add to...
4047         (LANG_HOOKS_INITIALIZER): ... this.
4048         * langhooks.h (struct lang_hooks): Add emits_begin_stmt.
4049         * lra-contraints.c (inherit_reload_reg): Tolerate between-blocks
4050         debug insns.
4051         (update_ebb_live_info): Skip debug insn markers.
4052         * lra.c (debug_insn_static_data): Rename to...
4053         (debug_bind_static_data): ... this.
4054         (debug_marker_static_data): New.
4055         (lra_set_insn_recog_data): Select one of the above depending
4056         on debug insn kind.
4057         (lra_update_isn_regno_info): Don't assume debug insns have
4058         freqs.
4059         (push_insns): Skip debug insns.
4060         * lto-streamer-in.c (input_function): Drop debug stmts
4061         depending on active options.  Adjust debug_nonbind_markers.
4062         * params.def (PARAM_MAX_DEBUG_MARKER_COUNT): New.
4063         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
4064         begin stmt marker notes.
4065         (print_insn): Likewise.
4066         * recog.c (extract_insn): Recognize rtl for debug markers.
4067         * rtl.def (DEBUG_MARKER): New.
4068         * tree-inline.c: Include params.h.
4069         (remap_gimple_stmt): Handle nonbind markers.
4070         (maybe_move_debug_stmts_to_successors): Likewise.
4071         (copy_debug_stmt): Likewise.
4072         * tree-iterator.c (append_to_statement_list_1): Append begin stmt
4073         markers regardless of no side effects.
4074         (tsi_link_before): Don't update container's side effects when adding
4075         a begin stmt marker.
4076         (tsi_link_after): Likewise.
4077         (expr_first): Skip begin stmt markers.
4078         (expr_last): Likewise.
4079         * tree-pretty-print (dump_generic_node): Handle begin stmt markers.
4080         * tree-ssa-threadedge.c (propagate_threaded_block_debug_info):
4081         Disregard nonbind markers.
4082         * tree.c (make_node_stat): Don't set side effects for begin stmt
4083         markers.
4084         (build1_stat): Likewise.
4085         * tree.def (DEBUG_BEGIN_STMT): New.
4086         * tree.h (GOTO_DESTINATION): Require a GOTO_EXPR.
4087         * var-tracking.c (delete_debug_insns): Renamed to...
4088         (delete_vta_debug_insns): ... this.
4089         (reemit_marker_as_note): New.
4090         (vt_initialize): Reemit markers.
4091         (delete_vta_debug_insns): Likewise.
4092         (vt_debug_insns_local): Reemit or delete markers.
4093         (variable_tracking_main_1): Likewise.
4094         * doc/generic.texi (DEBUG_BEGIN_STMT): Document.
4095         * doc/gimple.texi (gimple_debug_begin_stmt_p): New.
4096         (gimple_debug_nonbind_marker_p): New.
4097         (gimple_build_debug_bind): Adjust.
4098         (gimple_build_debug_begin_stmt): New.
4099         * doc/invoke.texi (max-debug-marker-count): New param.
4100         * doc/rtl.texi (debug_implicit_ptr, entry_value): New.
4101         (debug_parameter_ref, debug_marker): New.
4102         (NOTE_INSN_BEGIN_STMT): New.
4103         (DEBUG_INSN): Describe begin stmt markers.
4105         * cfgbuild.c (find_bb_boundaries): Don't purge dead edges if,
4106         without debug insns, we wouldn't, but clean up debug insns
4107         after a control flow insn nevertheless.
4109         * cfgcleanup.c (delete_unreachable_blocks): Use alternate
4110         block removal order if MAY_HAVE_DEBUG_BIND_INSNS.
4111         * cfgexpand.c (label_rtx_for_bb): Skip debug insns.
4112         * cfgrtl.c (try_redirect_by_replacing_jump): Skip debug insns.
4113         (rtl_tidy_fallthru_edge): Likewise.
4114         (rtl_verify_fallthru): Likewise.
4115         (rtl_verify_bb_layout): Likewise.
4116         (skip_insns_after_block): Likewise.
4117         (duplicate_insn_chain): Use DEBUG_BIND_INSN_P.
4118         * dwarf2out.c: Include print-rtl.h.
4119         (dwarf2out_next_real_insn): New.
4120         (dwarf2out_var_location): Call it.  Disregard begin stmt markers.
4121         Dump debug binds in asm comments.
4122         * gimple-iterator.c (gimple_find_edge_insert_loc): Skip debug stmts.
4123         * gimple-iterator.h (gsi_start_bb_nondebug): Remove; adjust
4124         callers to use gsi_start_nondebug_bb instead.
4125         (gsi_after_labels): Skip gimple debug stmts.
4126         (gsi_start_nondebug): New.
4127         * gimple-loop-interchange.c (find_deps_in_bb_for_stmt): Adjust.
4128         (proper_loop_form_for_interchange): Adjust.
4129         * gimple-low.c (gimple_seq_may_fallthru): Take last nondebug stmt.
4130         * gimple.h (gimple_seq_last_nondebug_stmt): New.
4131         * gimplify.c (last_stmt_in_scope): Skip debug stmts.
4132         (collect_fallthrough_labels): Likewise.
4133         (should_warn_for_implicit_fallthrough): Likewise.
4134         (warn_implicit_fallthrough_r): Likewise.
4135         (expand_FALLTHROUGH_r): Likewise.
4136         * graphite-isl-ast-to-gimple.c (gsi_insert_earliest): Adjust.
4137         (graphite_copy_stmts_from_block): Skip nonbind markers.
4138         * haifa-sched.c (sched_extend_bb): Skip debug insns.
4139         * ipa-icf-gimple.c (func_checker::compare_bb): Adjust.
4140         * jump.c (clean_barriers): Skip debug insns.
4141         * omp-expand.c (expand_parallel_call): Skip debug insns.
4142         (expand_task_call): Likewise.
4143         (remove_exit_barrier): Likewise.
4144         (expand_omp_taskreg): Likewise.
4145         (expand_omp_for_init_counts): Likewise.
4146         (expand_omp_for_generic): Likewise.
4147         (expand_omp_for_static_nochunk): Likewise.
4148         (expand_omp_for_static_chunk): Likewise.
4149         (expand_omp_simd): Likewise.
4150         (expand_omp_taskloop_for_outer): Likewise.
4151         (expand_omp_taskloop_for_inner): Likewise.
4152         (expand_oacc_for): Likewise.
4153         (expand_omp_sections): Likewise.
4154         (expand_omp_single): Likewise.
4155         (expand_omp_synch): Likewise.
4156         (expand_omp_atomic_load): Likewise.
4157         (expand_omp_atomic_store): Likewise.
4158         (expand_omp_atomic_fetch_op): Likewise.
4159         (expand_omp_atomic_pipeline): Likewise.
4160         (expand_omp_atomic_mutex): Likewise.
4161         (expand_omp_target): Likewise.
4162         (grid_expand_omp_for_loop): Likewise.
4163         (grid_expand_target_grid_body): Likewise.
4164         (build_omp_regions_1): Likewise.
4165         * omp-low.c (check_combined_parallel): Skip debug stmts.
4166         * postreload.c (fixup_debug_insns): Skip nonbind debug insns.
4167         * regcprop.c (find_oldest_value_reg): Ensure REGNO is not a pseudo.
4168         * sese.c (sese_trivially_empty_bb_p): Call is_gimple_debug in
4169         test.
4170         * tree-cfg.c (make_blobs_1): Skip debug stmts.
4171         (make_edges): Likewise.
4172         (cleanup_dead_labels): Likewise.
4173         (gimple_can_merge_blocks_p): Likewise.
4174         (stmt_starts_bb_p): Likewise.
4175         (gimple_block_label): Likewise.
4176         (gimple_redirect_edge_and_branch): Likewise.
4177         * tree-cfgcleanup.c (remove_forwarder_block): Rearrange skipping
4178         of debug stmts.
4179         (execute_cleanup_cfg_post_optimizing): Dump enumerated decls with
4180         TDF_SLIM.
4181         * tree-pretty-print (print_declaration): Omit initializer in slim
4182         dumps.
4183         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Mark begin stmt
4184         markers.
4185         (eliminate_unnecessary_stmts): Stabilize block removal order.
4186         * tree-ssa-tail-merge.c (find_duplicate): Skip debug stmts.
4187         * var-tracking.c (get_first_insn): New.
4188         (vt_emit_notes): Call it.
4189         (vt_initialize): Walk any insns before the first BB.
4190         (delete_debug_insns): Likewise.
4192         * gimple.h (enum gimple_debug_subcode): Add
4193         GIMPLE_DEBUG_BEGIN_STMT.
4194         (gimple_debug_begin_stmt_p): New.
4195         (gimple_debug_nonbind_marker_p): New.
4196         * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): New.
4197         (MAY_HAVE_DEBUG_BIND_STMTS): Renamed from....
4198         (MAY_HAVE_DEBUG_STMTS): ... this.  Check both.
4199         * insn-notes.def (BEGIN_STMT): New.
4200         * rtl.h (MAY_HAVE_DEBUG_MARKER_INSNS): New.
4201         (MAY_HAVE_DEBUG_BIND_INSNS): Renamed from....
4202         (MAY_HAVE_DEBUG_INSNS): ... this.  Check both.
4203         (NOTE_MARKER_LOCATION, NOTE_MARKER_P): New.
4204         (DEBUG_BIND_INSN_P, DEBUG_MARKER_INSN_P): New.
4205         (INSN_DEBUG_MARKER_KIND): New.
4206         (GEN_RTX_DEBUG_MARKER_BEGIN_STMT_PAT): New.
4207         (INSN_VAR_LOCATION): Check for VAR_LOCATION.
4208         (INSN_VAR_LOCATION_PTR): New.
4209         * cfgexpand.c (expand_debug_locations): Handle debug bind insns
4210         only.
4211         (expand_gimple_basic_block): Likewise.  Emit debug temps for TER
4212         deps only if debug bind insns are enabled.
4213         (pass_expand::execute): Avoid deep TER and expand
4214         debug locations for debug bind insns only.
4215         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Narrow
4216         debug stmts special handling down to debug bind stmts.
4217         * combine.c (try_combine): Narrow debug insns special handling
4218         down to debug bind insns.
4219         * cse.c (delete_trivially_dead_insns): Handle debug bindings.
4220         Narrow debug insns preexisting special handling down to debug
4221         bind insns.
4222         * dce.c (rest_of_handle_ud_dce): Narrow debug insns special
4223         handling down to debug bind insns.
4224         * function.c (instantiate_virtual_regs): Skip debug markers,
4225         adjust handling of debug binds.
4226         * gimple-ssa-backprop.c (backprop::prepare_change): Try debug
4227         temp insertion iff MAY_HAVE_DEBUG_BIND_STMTS.
4228         * haifa-sched.c (schedule_insn): Narrow special handling of debug
4229         insns to debug bind insns.
4230         * ipa-param-manipulation.c (ipa_modify_call_arguments): Narrow
4231         special handling of debug stmts to debug bind stmts.
4232         * ipa-split.c (split_function): Likewise.
4233         * ira.c (combine_and_move_insns): Adjust debug bind insns only.
4234         * loop-unroll.c (apply_opt_in_copies): Adjust tests on bind
4235         debug insns.
4236         * reg-stack.c (convert_regs_1): Use DEBUG_BIND_INSN_P.
4237         * regrename.c (build_def_use): Likewise.
4238         * regcprop.c (copyprop_hardreg_forward_1): Likewise.
4239         (pass_cprop_hardreg): Narrow special casing of debug insns to
4240         debug bind insns.
4241         * regstat.c (regstat_init_n_sets_and_refs): Likewise.
4242         * reload1.c (reload): Likewise.
4243         * sese.c (sese_insert_phis_for_liveouts): Narrow special
4244         casing of debug stmts to debug bind stmts.
4245         * shrink-wrap.c (move_insn_for_shrink_wrap): Likewise.
4246         * ssa-iterators.h (num_imm_uses): Likewise.
4247         * tree-cfg.c (gimple_merge_blocks): Narrow special casing of
4248         debug stmts to debug bind stmts.
4249         * tree-inline.c (tree_function_versioning): Narrow special casing
4250         of debug stmts to debug bind stmts.
4251         * tree-loop-distribution.c (generate_loops_for_partition):
4252         Narrow special casing of debug stmts to debug bind stmts.
4253         * tree-sra.c (analyze_access_subtree): Narrow special casing
4254         of debug stmts to debug bind stmts.
4255         * tree-ssa-dce.c (remove_dead_stmt): Narrow special casing of debug
4256         stmts to debug bind stmts.
4257         * tree-ssa-loop-ivopt.c (remove_unused_ivs): Narrow special
4258         casing of debug stmts to debug bind stmts.
4259         * tree-ssa-reassoc.c (reassoc_remove_stmt): Likewise.
4260         * tree-ssa-tail-merge.c (tail_merge_optimize): Narrow special
4261         casing of debug stmts to debug bind stmts.
4262         * tree-ssa-threadedge.c (propagate_threaded_block_debug_info):
4263         Likewise.
4264         * tree-ssa.c (flush_pending_stmts): Narrow special casing of
4265         debug stmts to debug bind stmts.
4266         (gimple_replace_ssa_lhs): Likewise.
4267         (insert_debug_temp_for_var_def): Likewise.
4268         (insert_debug_temps_for_defs): Likewise.
4269         (reset_debug_uses): Likewise.
4270         * tree-ssanames.c (release_ssa_name_fn): Likewise.
4271         * tree-vect-loop-manip.c (adjust_debug_stmts_now): Likewise.
4272         (adjust_debug_stmts): Likewise.
4273         (adjust_phi_and_debug_stmts): Likewise.
4274         (vect_do_peeling): Likewise.
4275         * tree-vect-loop.c (vect_transform_loop): Likewise.
4276         * valtrack.c (propagate_for_debug): Use BIND_DEBUG_INSN_P.
4277         * var-tracking.c (adjust_mems): Narrow special casing of debug
4278         insns to debug bind insns.
4279         (dv_onepart_p, dataflow_set_clar_at_call, use_type): Likewise.
4280         (compute_bb_dataflow, vt_find_locations): Likewise.
4281         (vt_expand_loc, emit_notes_for_changes): Likewise.
4282         (vt_init_cfa_base): Likewise.
4283         (vt_emit_notes): Likewise.
4284         (vt_initialize): Likewise.
4285         (vt_finalize): Likewise.
4287         * emit-rtl.c (next_nondebug_insn, prev_nondebug_insn): Reorder.
4288         (next_nonnote_nondebug_insn, prev_nonnote_nondebug_insn): Reorder.
4289         (next_nonnote_nondebug_insn_bb): New.
4290         (prev_nonnote_nondebug_insn_bb): New.
4291         (prev_nonnote_insn_bb, next_nonnote_insn_bb): Remove.
4292         * rtl.h (prev_nonnote_insn_bb, next_nonnote_insn_bb): Remove decls.
4293         (prev_nonnote_nondebug_insn_bb): Declare.
4294         (next_nonnote_nondebug_insn_bb): Declare.
4295         * cfgbuild.c (find_bb_boundaries): Adjust to skip debug insns.
4296         * cfgrtl.c (get_last_bb_insn): Likewise.
4297         * lra.c (push_insns): Likewise.
4299 2017-12-11  David Malcolm  <dmalcolm@redhat.com>
4301         PR c/82050
4302         * selftest-run-tests.c (selftest::run_tests): Move start/finish code
4303         to...
4304         * selftest.c (selftest::test_runner::test_runner): New ctor.
4305         (selftest::test_runner::~test_runner): New dtor.
4306         * selftest.h (class selftest::test_runner): New class.
4308 2017-12-11  Carl Love  <cel@us.ibm.com>
4310         * config/rs6000/altivec.h (vec_extract_fp32_from_shorth,
4311         vec_extract_fp32_from_shortl]): Add #defines.
4312         * config/rs6000/rs6000-builtin.def (VSLDOI_2DI): Add macro expansion.
4313         * config/rs6000/rs6000-c.c (ALTIVEC_BUILTIN_VEC_UNPACKH,
4314         ALTIVEC_BUILTIN_VEC_UNPACKL, ALTIVEC_BUILTIN_VEC_AND,
4315         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SRL,
4316         ALTIVEC_BUILTIN_VEC_SRO, ALTIVEC_BUILTIN_VEC_SLD,
4317         ALTIVEC_BUILTIN_VEC_SLL): Add expansions.
4318         * doc/extend.texi: Add documentation for the added builtins.
4320 2017-12-11  Alexandre Oliva  <aoliva@redhat.com>
4322         PR rtl-optimization/80693
4323         PR rtl-optimization/81019
4324         PR rtl-optimization/81020
4325         * combine.c (distribute_notes): Reset any REG_UNUSED REGs that
4326         are not mentioned in i3.  Place the REG_UNUSED note on i2,
4327         possibly modified to REG_DEAD, if it did not originate in i3.
4329 2017-12-11  Jakub Jelinek  <jakub@redhat.com>
4331         * recog.c (store_data_bypass_p_1): New function.
4332         (store_data_bypass_p): Handle USE in a PARALLEL like CLOBBER.  Use
4333         store_data_bypass_p_1 to avoid code duplication.  Formatting fixes.
4335 2017-12-11  Segher Boessenkool  <segher@kernel.crashing.org>
4337         PR rtl-optimization/83361
4338         * ifcvt.c (if_convert): Call fixup_partitions.
4340 2017-12-11  Will Schmidt  <will_schmidt@vnet.ibm.com>
4342         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
4343         early folding of splat_u{8,16,32}.
4345 2017-12-11  Jakub Jelinek  <jakub@redhat.com>
4347         * config/aarch64/aarch64.c (aarch64_print_operand): Don't start
4348         output_operand_lossage first argument with capital letter.
4349         (aarch64_override_options): Don't start error and sorry first argument
4350         with capital letter.
4352 2017-12-11  Andi Kleen  <ak@linux.intel.com>
4354         PR gcov-profile/83355
4355         * auto-profile.c (string_table::get_index_by_decl): Don't
4356         recurse when abstract origin points to itself.
4358 2017-12-11  Bin Cheng  <bin.cheng@arm.com>
4360         PR tree-optimization/83320
4361         * gimple-loop-interchange.cc (free_data_refs_with_aux): Use delete.
4362         (prune_datarefs_not_in_loop): Ditto.
4364 2017-12-10  Gerald Pfeifer  <gerald@pfeifer.com>
4366         * doc/install.texi (Specific): Tweak link to mkssoftware.com.
4368 2017-12-10  Jakub Jelinek  <jakub@redhat.com>
4370         PR tree-optimization/83337
4371         * gimple-loop-interchange.cc (compute_access_stride): Handle
4372         bitfield DRs properly.
4374 2017-12-09  Jakub Jelinek  <jakub@redhat.com>
4376         PR tree-optimization/83338
4377         * tree-vect-stmts.c (vectorizable_operation): Handle POINTER_DIFF_EXPR
4378         vectorization as MINUS_EXPR with a subsequent VIEW_CONVERT_EXPR from
4379         vector of unsigned integers to vector of signed integers.
4381 2017-12-08  Vladimir Makarov  <vmakarov@redhat.com>
4383         PR rtl-optimization/83317
4384         * lra-constraints.c (process_address_1): Add insn code check.
4386 2017-12-08  Michael Matz  <matz@suse.de>
4388         Fix PR tree-optimization/83323
4389         * gimple-loop-jam (unroll_jam_possible_p): Correct test for
4390         head-controlled loops and loop BBs.
4391         * common.opt (funroll-and-jam): Remove, instead ...
4392         (floop-unroll-and-jam): ... reuse this option.
4393         * opts.c (default_options_table): Use OPT_floop_unroll_and_jam.
4394         * doc/invoke.texi (-funroll-and-jam): Move docu to ...
4395         (-floop-unroll-and-jam): ... this option.
4397 2017-12-08  Jakub Jelinek  <jakub@redhat.com>
4399         * ipa-polymorphic-call.c (noncall_stmt_may_be_vtbl_ptr_store): Fix
4400         a comment typo, get_base_ref_and_offset -> get_ref_base_and_extent.
4401         * ipa-prop.c (stmt_may_be_vtbl_ptr_store): Likewise.
4403 2017-12-08  Richard Biener  <rguenther@suse.de>
4405         PR middle-end/81782
4406         * tree-ssa-uninit.c (warn_uninitialized_vars): Properly
4407         handle accesses outside of zero-sized vars.
4409 2017-12-08  Martin Jambor  <mjambor@suse.cz>
4411         PR tree-optimization/83141
4412         * tree-sra.c (contains_vce_or_bfcref_p): Move up in the file, also
4413         test for MEM_REFs implicitely changing types with padding.  Remove
4414         inline keyword.
4415         (build_accesses_from_assign): Added contains_vce_or_bfcref_p checks.
4417 2017-12-08  Jakub Jelinek  <jakub@redhat.com>
4419         * config/arc/arc.c (arc_attribute_table): Add exclusions to
4420         the comment.
4421         * config/avr/avr.c (avr_attribute_table): Likewise.
4422         * config/msp430/msp430.c (msp430_attribute_table): Likewise.
4423         * config/rl78/rl78.c (rl78_attribute_table): Likewise.
4424         * config/nds32/nds32.c (nds32_attribute_table): Likewise.
4425         * config/darwin.h (SUBTARGET_ATTRIBUTE_TABLE): Initialize new member
4426         of struct attribute_spec.
4427         * config/i386/cygming.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
4429 2017-12-08  Ulrich Weigand  <uweigand@de.ibm.com>
4431         PR target/82960
4432         * config/spu/spu.c (pad_bb): Only check INSN_CODE when INSN_P is true.
4434 2017-12-08  Jan Hubicka  <hubicka@ucw.cz>
4436         PR middle-end/83609
4437         * profile-count.c (profile_count::from_gcov_type): Move from
4438         profile-count.h; handle overflow.
4439         * profile-count.h (profile_count::from_gcov_type): Move offline.
4441 2017-12-08  Segher Boessenkool  <segher@kernel.crashing.org>
4443         PR rtl-optimization/83304
4444         * combine.c (move_deaths): If we do not know where a register died,
4445         search for it.
4447 2017-12-08  Richard Biener  <rguenther@suse.de>
4449         * gimple-loop-interchange.cc (tree_loop_interchange::interchange):
4450         Provide -fopt-info-loop feedback when we interchange in a nest.
4452 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4454         * config/arm/driver-arm.c (arm_cpu_table): Use fp-capable product names
4455         for armv6 ARM CPU IDs.
4457 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4459         * common/config/arm/arm-common.c: Include <algorithm>.
4460         (INCLUDE_VECTOR): Define.
4461         (compare_opt_names): New function.
4462         (arm_rewrite_selected_arch): Only strip out extensions that can be
4463         expressed through -mfpu.  Sort the remaining extensions
4464         alphabetically.
4466 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4468         * config/arm/arm.h (arm_asm_auto_mfpu): Declare.
4469         (ASM_CPU_SPEC_FUNCTIONS): Add new rule asm_auto_mfpu.
4470         (ASM_CPU_SPEC): Use it if -mfpu is set to auto.
4471         * common/config/arm/arm-common.c (arm_asm_auto_mfpu): New function.
4473 2017-06-08  Tristan Gingold  <gindold@adacore.com>
4475         PR ada/81470
4476         * dwarf2out.c (dwarf2out_do_cfi_startproc): Only emit
4477         .cfi_personality or .cfi_lsda if the eh data format is dwarf2.
4479 2017-12-08  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4481         * config/sol2.h (SOLARIS_ATTRIBUTE_TABLE): Initialize new member
4482         of struct attribute_spec.
4484 2017-12-08  Julia Koval  <julia.koval@intel.com>
4486         * config/i386/avx512vnniintrin.h (_mm512_dpwssds_epi32,
4487         _mm512_mask_dpwssds_epi32, _mm512_maskz_dpwssds_epi32): New intrinsics.
4488         * config/i386/avx512vnnivlintrin.h (_mm256_dpwssds_epi32,
4489         _mm256_mask_dpwssds_epi32, _mm256_maskz_dpwssds_epi32,
4490         _mm_dpwssds_epi32, _mm_mask_dpwssds_epi32,
4491         _mm_maskz_dpwssds_epi32): Ditto.
4493 2017-12-08  Richard Biener  <rguenther@suse.de>
4495         PR tree-optimization/81303
4496         * tree-vect-stmts.c (vect_is_simple_cond): For invariant
4497         conditions try to create a comparison vector type matching
4498         the data vector type.
4499         (vectorizable_condition): Adjust.
4500         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
4501         Leave invariant conditions alone in case we can vectorize those.
4503 2017-12-08  Julia Koval  <julia.koval@intel.com>
4505         * config/i386/avx512vnniintrin.h (_mm512_dpwssd_epi32,
4506         _mm512_mask_dpwssd_epi32, _mm512_maskz_dpwssd_epi32): New intrinsics.
4507         * config/i386/avx512vnnivlintrin.h (_mm256_dpwssd_epi32,
4508         _mm256_mask_dpwssd_epi32, _mm256_maskz_dpwssd_epi32, _mm_dpwssd_epi32,
4509         _mm_mask_dpwssd_epi32, _mm_maskz_dpwssd_epi32): Ditto.
4511 2017-12-08  Julia Koval  <julia.koval@intel.com>
4513         * config/i386/avx512vnniintrin.h (_mm512_dpbusds_epi32,
4514         _mm512_mask_dpbusds_epi32, _mm512_maskz_dpbusds_epi32): New.
4515         * config/i386/avx512vnnivlintrin.h (_mm256_dpbusds_epi32,
4516         _mm256_mask_dpbusds_epi32, _mm256_maskz_dpbusds_epi32,
4517         _mm_dpbusds_epi32, _mm_mask_dpbusds_epi32,
4518         _mm_maskz_dpbusds_epi32): New intrinsics.
4520 2017-12-07  Sandra Loosemore  <sandra@codesourcery.com>
4522         * config/nios2/nios2.md (ld<bhw_uns>io): Add splitter for memory
4523         operand.
4524         (ld<bh>io_signed): Likewise.
4525         (st<bhw>io): Likewise.
4526         * config/nios2/predicates.md (ldstio_memory_operand): Allow
4527         SMALL_INT12 constant integer operand.
4529 2017-12-07  Sandra Loosemore  <sandra@codesourcery.com>
4531         * config/nios2/nios2.c (nios2_symbolic_constant_allowed):
4532         Rename to...
4533         (nios2_large_constant_allowed): ...this.  Adjust uses.
4534         (nios2_plus_symbolic_constant_p): Rename to...
4535         (nios2_plus_large_constant_p): ...this.  Adjust uses.
4536         (nios2_legitimate_address_p): Correct CONST_INT handling.
4537         (nios2_symbolic_memory_operand_p): Rename to...
4538         (nios2_large_constant_memory_operand_p): ...this.  Adjust uses.
4539         (nios2_large_constant_p): Check for large constant integers too.
4540         (nios2_split_large_constant): Handle constant integers.
4541         (nios2_split_symbolic_memory_operand): Rename to...
4542         (nios2_split_large_constant_memory_operand): ...this.  Adjust uses.
4543         (nios2_legitimize_constant_address): Handle constant integers.
4544         (r0rel_constant_p): Handle small constant integers.
4545         (nios2_print_operand_address): Handle r0-relative integer addresses.
4546         * config/nios2/nios2-protos.h: Adjust for renamed functions.
4547         * config/nios2/nios2.md: Adjust for renamed functions.
4549 2017-12-07  Andrew Waterman  <andrew@sifive.com>
4551         * config/riscv/riscv.c (TARGET_ASM_SELECT_SECTION): New define.
4552         (TARGET_HAVE_SRODATA_SECTION): New define.
4553         (riscv_select_section): New function.
4555 2017-12-08  Joseph Myers  <joseph@codesourcery.com>
4556             Alexander Monakov  <amonakov@ispras.ru>
4557             Jakub Jelinek  <jakub@redhat.com>
4559         PR target/81906
4560         * config/i386/i386.c (ix86_expand_rint): Handle flag_rounding_math.
4562 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4564         * config/aarch64/aarch64.c (aarch64_print_address_internal): Return
4565         a bool success value.  Don't call output_operand_lossage here.
4566         (aarch64_print_ldpstp_address): Return a bool success value.
4567         (aarch64_print_operand_address): Call output_addr_const if
4568         aarch64_print_address_internal fails.
4569         (aarch64_print_operand): Don't assert that the mode is 16 bytes for
4570         'y'; call output_operand_lossage instead.  Call output_operand_lossage
4571         if aarch64_print_ldpstp_address fails.
4573 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4575         * tree-vector-builder.h
4576         (tree_vector_builder::binary_encoded_nelts): Declare.
4577         * tree-vector-builder.c
4578         (tree_vector_builder::binary_encoded_nelts): New function.
4579         * fold-const.c (negate_expr_p): Likewise.
4580         (operand_equal_p, fold_checksum_tree): Likewise.
4581         * tree-loop-distribution.c (const_with_all_bytes_same): Likewise.
4582         * tree.c (integer_zerop, integer_onep, integer_all_onesp, real_zerop)
4583         (real_onep, real_minus_onep, add_expr, initializer_zerop): Likewise.
4584         (uniform_vector_p): Likewise.
4585         * varasm.c (const_hash_1, compare_constant): Likewise.
4586         * tree-ssa-ccp.c: Include tree-vector-builder.h.
4587         (valid_lattice_transition): Operate directly on the VECTOR_CST
4588         encoding.
4589         * ipa-icf.c: Include tree-vector-builder.h.
4590         (sem_variable::equals): Operate directly on the VECTOR_CST encoding.
4591         * print-tree.c (print_node): Print encoding of VECTOR_CSTs.
4593 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4595         * tree.c (build_vector): Delete.
4596         * tree.h (build_vector): Make static and move into the self-testing
4597         block.
4599 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4601         * vector-builder.h (vector_builder::derived): New const overload.
4602         (vector_builder::elt): New function.
4603         * tree-vector-builder.h (tree_vector_builder::type): New function.
4604         (tree_vector_builder::apply_step): Declare.
4605         * tree-vector-builder.c (tree_vector_builder::apply_step): New
4606         function.
4607         * gimple-fold.h (tree_vector_builder): Declare.
4608         (gimple_build_vector): Take a tree_vector_builder instead of a
4609         type and vector of elements.
4610         * gimple-fold.c (gimple_build_vector): Likewise.
4611         * tree-vect-loop.c (get_initial_def_for_reduction): Update call
4612         accordingly.
4613         (get_initial_defs_for_reduction): Likewise.
4614         (vectorizable_induction): Likewise.
4616 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4618         * tree-vector-builder.h
4619         (tree_vector_builder::new_binary_operation): Declare.
4620         * tree-vector-builder.c
4621         (tree_vector_builder::new_binary_operation): New function.
4622         * fold-const.c (fold_relational_const): Use it.
4623         (const_binop): Likewise.  Check that both input vectors have
4624         the same number of elements, thus excluding things like WIDEN_SUM.
4625         Check whether it is possible to operate directly on the encodings
4626         of stepped inputs.
4628 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4630         * fold-const.c (fold_negate_expr_1): Use tree_vector_builder and
4631         new_unary_operation, operating only on the encoded elements.
4632         (const_unop): Likewise.
4633         (exact_inverse): Likewise.
4634         (distributes_over_addition_p): New function.
4635         (const_binop): Use tree_vector_builder and new_unary_operation
4636         for combinations of VECTOR_CST and INTEGER_CST.  Operate only
4637         on the encoded elements unless the encoding is strided and the
4638         operation does not distribute over addition.
4639         (fold_convert_const):  Use tree_vector_builder and
4640         new_unary_operation.  Operate only on the encoded elements
4641         for truncating integer conversions, or for non-stepped encodings.
4643 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4645         * config/sparc/sparc.c: Include tree-vector-builder.h.
4646         (sparc_fold_builtin): Use tree_vector_builder instead of build_vector.
4647         * expmed.c: Include tree-vector-builder.h.
4648         (make_tree): Use tree_vector_builder instead of build_vector.
4649         * fold-const.c: Include tree-vector-builder.h.
4650         (const_binop): Use tree_vector_builder instead of build_vector.
4651         (const_unop): Likewise.
4652         (native_interpret_vector): Likewise.
4653         (fold_vec_perm): Likewise.
4654         (fold_ternary_loc): Likewise.
4655         * gimple-fold.c: Include tree-vector-builder.h.
4656         (gimple_fold_stmt_to_constant_1): Use tree_vector_builder instead
4657         of build_vector.
4658         * tree-ssa-forwprop.c: Include tree-vector-builder.h.
4659         (simplify_vector_constructor): Use tree_vector_builder instead
4660         of build_vector.
4661         * tree-vect-generic.c: Include tree-vector-builder.h.
4662         (add_rshift): Use tree_vector_builder instead of build_vector.
4663         (expand_vector_divmod): Likewise.
4664         (optimize_vector_constructor): Likewise.
4665         * tree-vect-loop.c: Include tree-vector-builder.h.
4666         (vect_create_epilog_for_reduction): Use tree_vector_builder instead
4667         of build_vector.  Explicitly use a stepped encoding for
4668         { 1, 2, 3, ... }.
4669         * tree-vect-slp.c: Include tree-vector-builder.h.
4670         (vect_get_constant_vectors): Use tree_vector_builder instead
4671         of build_vector.
4672         (vect_transform_slp_perm_load): Likewise.
4673         (vect_schedule_slp_instance): Likewise.
4674         * tree-vect-stmts.c: Include tree-vector-builder.h.
4675         (vectorizable_bswap): Use tree_vector_builder instead of build_vector.
4676         (vect_gen_perm_mask_any): Likewise.
4677         (vectorizable_call): Likewise.  Explicitly use a stepped encoding.
4678         * tree.c: (build_vector_from_ctor): Use tree_vector_builder instead
4679         of build_vector.
4680         (build_vector_from_val): Likewise.  Explicitly use a duplicate
4681         encoding.
4683 2017-12-07  Richard Sandiford  <richard.sandiford@arm.com>
4685         * doc/generic.texi (VECTOR_CST): Describe new representation of
4686         vector constants.
4687         * vector-builder.h: New file.
4688         * tree-vector-builder.h: Likewise.
4689         * tree-vector-builder.c: Likewise.
4690         * Makefile.in (OBJS): Add tree-vector-builder.o.
4691         * tree.def (VECTOR_CST): Update comment to refer to generic.texi.
4692         * tree-core.h (tree_base): Add a vector_cst field to the u union.
4693         (tree_vector): Change the number of elements to
4694         vector_cst_encoded_nelts.
4695         * tree.h (VECTOR_CST_NELTS): Redefine using TYPE_VECTOR_SUBPARTS.
4696         (VECTOR_CST_ELTS): Delete.
4697         (VECTOR_CST_ELT): Redefine using vector_cst_elt.
4698         (VECTOR_CST_LOG2_NPATTERNS, VECTOR_CST_NPATTERNS): New macros.
4699         (VECTOR_CST_NELTS_PER_PATTERN, VECTOR_CST_DUPLICATE_P): Likewise.
4700         (VECTOR_CST_STEPPED_P, VECTOR_CST_ENCODED_ELTS): Likewise.
4701         (VECTOR_CST_ENCODED_ELT): Likewise.
4702         (vector_cst_encoded_nelts): New function.
4703         (make_vector): Take the values of VECTOR_CST_LOG2_NPATTERNS and
4704         VECTOR_CST_NELTS_PER_PATTERN as arguments.
4705         (vector_cst_int_elt, vector_cst_elt): Declare.
4706         * tree.c: Include tree-vector-builder.h.
4707         (tree_code_size): Abort if passed VECTOR_CST.
4708         (tree_size): Update for new VECTOR_CST layout.
4709         (make_vector): Take the values of VECTOR_CST_LOG2_NPATTERNS and
4710         VECTOR_CST_NELTS_PER_PATTERN as arguments.
4711         (build_vector): Use tree_vector_builder.
4712         (vector_cst_int_elt, vector_cst_elt): New functions.
4713         (drop_tree_overflow): For VECTOR_CST, drop the TREE_OVERFLOW from the
4714         encoded elements and then create the vector in the canonical form.
4715         (check_vector_cst, check_vector_cst_duplicate, check_vector_cst_fill)
4716         (check_vector_cst_stepped, test_vector_cst_patterns): New functions.
4717         (tree_c_tests): Call test_vector_cst_patterns.
4718         * lto-streamer-out.c (DFS::DFS_write_tree_body): Handle the new
4719         VECTOR_CST fields.
4720         (hash_tree): Likewise.
4721         * tree-streamer-out.c (write_ts_vector_tree_pointers): Likewise.
4722         (streamer_write_tree_header): Likewise.
4723         * tree-streamer-in.c (lto_input_ts_vector_tree_pointers): Likewise.
4724         (streamer_alloc_tree): Likewise.  Update call to make_vector.
4725         * fold-const.c (fold_ternary_loc): Avoid using VECTOR_CST_ELTS.
4727 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4729         * selftest.h (ASSERT_TRUE_AT, ASSERT_FALSE_AT, ASSERT_EQ_AT)
4730         (ASSERT_NE, ASSERT_PRED1): Add underscores to local variable names
4731         * selftest-rtl.h (ASSERT_RTX_EQ, ASSERT_RTX_PTR_EQ): Likewise.
4733 2017-12-07  Bin Cheng  <bin.cheng@arm.com>
4734             Richard Biener  <rguenther@suse.de>
4736         PR tree-optimization/81303
4737         * Makefile.in (gimple-loop-interchange.o): New object file.
4738         * common.opt (floop-interchange): Reuse the option from graphite.
4739         * doc/invoke.texi (-floop-interchange): Ditto.  New document for
4740         -floop-interchange and mention it for -O3.
4741         * opts.c (default_options_table): Enable -floop-interchange at -O3.
4742         * gimple-loop-interchange.cc: New file.
4743         * params.def (PARAM_LOOP_INTERCHANGE_MAX_NUM_STMTS): New parameter.
4744         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): New parameter.
4745         * passes.def (pass_linterchange): New pass.
4746         * timevar.def (TV_LINTERCHANGE): New time var.
4747         * tree-pass.h (make_pass_linterchange): New declaration.
4748         * tree-ssa-loop-ivcanon.c (create_canonical_iv): Change to external
4749         interchange.  Record IV before/after increment in new parameters.
4750         * tree-ssa-loop-ivopts.h (create_canonical_iv): New declaration.
4751         * tree-vect-loop.c (vect_is_simple_reduction): Factor out reduction
4752         path check into...
4753         (check_reduction_path): ...New function here.
4754         * tree-vectorizer.h (check_reduction_path): New declaration.
4756 2017-12-07  Vladimir Makarov  <vmakarov@redhat.com>
4758         PR target/83252
4759         PR rtl-optimization/80818
4760         * lra.c (add_regs_to_insn_regno_info): Make a hard reg in CLOBBER
4761         always early clobbered.
4762         * lra-lives.c (process_bb_lives): Check input hard regs for early
4763         clobbered non-operand hard reg.
4765 2017-12-07  Jakub Jelinek  <jakub@redhat.com>
4767         PR middle-end/83164
4768         * tree-cfg.c (verify_gimple_assign_binary): Don't require
4769         types_compatible_p, just that TYPE_MODE is the same.
4771 2017-12-07  Martin Sebor  <msebor@redhat.com>
4773         PR c/81544
4774         * attribs.c (empty_attribute_table): Initialize new member of
4775         struct attribute_spec.
4776         (decl_attributes): Add argument.  Handle mutually exclusive
4777         combinations of attributes.
4778         (selftests::test_attribute_exclusions): New function.
4779         (selftests::attribute_c_tests): Ditto.
4780         * attribs.h (decl_attributes): Add default argument.
4781         * selftest.h (attribute_c_tests): Declare.
4782         * selftest-run-tests.c (selftest::run_tests): Call attribute_c_tests.
4783         * tree-core.h (attribute_spec::exclusions, exclude): New type and
4784         member.
4785         * doc/extend.texi (Common Function Attributes): Update const and pure.
4786         * config/alpha/alpha.c (vms_attribute_table): Initialize new member
4787         of struct attribute_spec.
4788         * config/arc/arc.c (arc_attribute_table): Same.
4789         * config/arm/arm.c (arm_attribute_table): Same.
4790         * config/avr/avr.c ( avr_attribute_table): Same.
4791         * config/bfin/bfin.c (bfin_attribute_table): Same.
4792         * config/cr16/cr16.c (cr16_attribute_table): Same.
4793         * config/epiphany/epiphany.c (epiphany_attribute_table): Same.
4794         * config/h8300/h8300.c (h8300_attribute_table): Same.
4795         * config/i386/i386.c (ix86_attribute_table): Same.
4796         * config/ia64/ia64.c (ia64_attribute_table): Same.
4797         * config/m32c/m32c.c (m32c_attribute_table): Same.
4798         * config/m32r/m32r.c (m32r_attribute_table): Same.
4799         * config/m68k/m68k.c (m68k_attribute_table): Same.
4800         * config/mcore/mcore.c (mcore_attribute_table): Same.
4801         * config/microblaze/microblaze.c (microblaze_attribute_table): Same.
4802         * config/mips/mips.c (mips_attribute_table): Same.
4803         * config/msp430/msp430.c (msp430_attribute_table): Same.
4804         * config/nds32/nds32.c (nds32_attribute_table): Same.
4805         * config/nvptx/nvptx.c (nvptx_attribute_table): Same.
4806         * config/powerpcspe/powerpcspe.c (rs6000_attribute_table): Same.
4807         * config/rl78/rl78.c (rl78__attribute_table): Same.
4808         * config/rs6000/rs6000.c (rs6000_attribute_table): Same.
4809         * onfig/rx/rx.c (rx_attribute_table): Same.
4810         * config/s390/s390.c (s390_handle_vectorbool_attribute): Same.
4811         * config/sh/sh.c (sh_attribute_table): Same.
4812         * config/sparc/sparc.c (sparc_attribute_table): Same.
4813         * config/spu/spu.c (spu_attribute_table): Same.
4814         * config/stormy16/stormy16.c (xstormy16_attribute_table): Same.
4815         * config/v850/v850.c (v850_attribute_table): Same.
4816         * config/visium/visium.c (visium_attribute_table): Same.
4818 2017-12-07  Tamar Christina  <tamar.christina@arm.com>
4820         PR target/82641
4821         * config/arm/arm.c (INCLUDE_STRING): Define.
4822         (arm_last_printed_arch_string, arm_last_printed_fpu_string): New.
4823         (arm_declare_function_name): Conservatively emit .arch, .arch_extensions
4824         and .fpu.
4826 2017-12-07  Michael Matz  <matz@suse.de>
4828         Add unroll and jam pass
4830         * gimple-loop-jam.c: New file.
4831         * Makefile.in (OBJS): Add gimple-loop-jam.o.
4832         * common.opt (funroll-and-jam): New option.
4833         * opts.c (default_options_table): Add unroll-and-jam at -O3.
4834         * params.def (PARAM_UNROLL_JAM_MIN_PERCENT): New param.
4835         (PARAM_UNROLL_JAM_MAX_UNROLL): Ditto.
4836         * passes.def: Add pass_loop_jam.
4837         * timevar.def (TV_LOOP_JAM): Add.
4838         * tree-pass.h (make_pass_loop_jam): Declare.
4839         * cfgloop.c (flow_loop_tree_node_add): Add AT argument.
4840         * cfgloop.h (flow_loop_tree_node_add): Adjust declaration.
4841         * cfgloopmanip.c (duplicate_loop): Add AT argument, adjust call
4842         to flow_loop_tree_node_add.
4843         (duplicate_subloops, copy_loops_to): Append to sibling list.
4844         * cfgloopmanip.h: (duplicate_loop): Adjust declaration.
4845         * doc/invoke.texi (-funroll-and-jam): Document new option.
4846         (unroll-jam-min-percent, unroll-jam-max-unroll): Document new params.
4848 2017-12-07  Richard Biener  <rguenther@suse.de>
4850         PR tree-optimization/83296
4851         PR tree-optimization/67769
4852         * tree-ssa-phiopt.c (conditional_replacement): Do not reset
4853         flow sensitive info in an unrelated BB.
4854         (value_replacement): Use reset_flow_sensitive_info.
4855         (minmax_replacement): Reset flow sensitive info on the def
4856         we move.  Do not reset flow sensitive info in the whole BB
4857         we move the stmt to.
4858         (abs_replacement): Likewise.
4860 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4862         PR target/43871
4863         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
4864         rs6000_cpu to the given -mcpu=, or to the default processor.
4866 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4868         * config/rs6000/rs6000.h (rs6000_cpu_attr): Delete.
4869         * config/rs6000/rs6000.c (rs6000_variable_issue_1): Use rs6000_tune
4870         instead of rs6000_cpu_attr.
4871         (rs6000_adjust_cost): Ditto.
4872         (is_microcoded_insn): Ditto.
4873         (rs6000_adjust_priority): Ditto.
4874         (rs6000_issue_rate): Ditto.
4875         (rs6000_use_sched_lookahead): Ditto.
4876         (rs6000_use_sched_lookahead_guard): Ditto.
4877         (rs6000_sched_reorder): Ditto.
4878         (force_new_group): Ditto.
4879         * config/rs6000/rs6000.md (cpu attribute): Ditto.
4880         (group_ending_nop): Ditto.
4882 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4884         * config/rs6000/rs6000.opt (rs6000_tune): New variable.
4885         * config/rs6000/rs6000.c (rs6000_option_override_internal): Also set
4886         rs6000_tune.  Use rs6000_tune instead of rs6000_cpu where appropriate.
4887         (rs6000_loop_align): Use rs6000_tune instead of rs6000_cpu where
4888         appropriate.
4889         (rs6000_reassociation_width): Ditto.
4890         (rs6000_emit_epilogue): Ditto.
4891         (rs6000_adjust_cost): Ditto.
4892         (is_microcoded_insn): Ditto.
4893         (is_cracked_insn): Ditto.
4894         (rs6000_adjust_priority): Ditto.
4895         (rs6000_sched_reorder): Ditto.
4896         (rs6000_sched_reorder2): Ditto.
4897         (insn_must_be_first_in_group): Ditto.
4898         (insn_must_be_last_in_group): Ditto.
4899         (rs6000_register_move_cost): Ditto.
4900         * config/rs6000/rs6000.h (rs6000_cpu_attr): Use rs6000_tune instead of
4901         rs6000_cpu.
4903 2017-12-07  Julia Koval  <julia.koval@intel.com>
4905         * config.gcc: Add vaesintrin.h.
4906         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI_V64QI):
4907         New type.
4908         * config/i386/i386-builtin.def (__builtin_ia32_vaesdec_v16qi,
4909         __builtin_ia32_vaesdec_v32qi, __builtin_ia32_vaesdec_v64qi):
4910         New builtins.
4911         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
4912         * config/i386/immintrin.h: Include vaesintrin.h.
4913         * config/i386/sse.md (vaesdec_<mode>): New pattern.
4914         * config/i386/vaesintrin.h (_mm256_aesdec_epi128, _mm512_aesdec_epi128,
4915         _mm_aesdec_epi128): New intrinsics.
4917 2017-12-06  David Malcolm  <dmalcolm@redhat.com>
4919         * Makefile.in (C_COMMON_OBJS): Add c-family/c-spellcheck.o.
4920         * spellcheck-tree.c (find_closest_macro_cpp_cb): Move to
4921         c-family/c-spellcheck.cc.
4922         (best_macro_match::best_macro_match): Likewise.
4923         * spellcheck-tree.h
4924         (struct edit_distance_traits<cpp_hashnode *>): Move to
4925         c-family/c-spellcheck.h.
4926         (class best_macro_match): Likewise.
4928 2017-12-06  Jakub Jelinek  <jakub@redhat.com>
4930         PR tree-optimization/83293
4931         * gimple-ssa-strength-reduction.c (insert_initializers): Use
4932         GSI_NEW_STMT instead of GSI_SAME_STMT in gsi_insert_after that
4933         might insert into empty bb.
4935         PR sanitizer/81281
4936         * match.pd ((T)(P + A) - (T)P -> (T) A): Split into separate
4937         simplify for plus with :c added, and pointer_plus without that.
4938         ((T)P - (T)(P + A) -> -(T) A): Likewise.  If type is integral
4939         with undefined overflow and the conversion is not widening,
4940         perform negation in utype and only convert to type afterwards.
4941         ((T)(P + A) - (T)(P + B) -> (T)A - (T)B): Split into separate
4942         simplify for plus with :c added, and pointer_plus without that.
4943         If type is integral with undefined overflow and the conversion is
4944         not widening, perform minus in utype and only convert to type
4945         afterwards.  Move the last pointer_diff_expr simplify into the
4946         two outermost ifs.
4948 2017-12-06  Martin Sebor  <msebor@redhat.com>
4950         PR tree-optimization/82646
4951         * builtins.c (maybe_emit_chk_warning): Use size as the bound for
4952         strncpy, not maxlen.
4954 2017-12-06  Martin Sebor  <msebor@redhat.com>
4956         * doc/invoke.texi (-Wstringop-truncation): Mention attribute
4957         nonstring.
4959         PR tree-optimization/83075
4960         * tree-ssa-strlen.c (handle_builtin_stxncpy): Avoid assuming
4961         strncat/strncpy don't change length of source string.
4963 2017-12-06  Eric Botcazou  <ebotcazou@adacore.com>
4965         Revert
4966         2017-11-29  Martin Aberg  <maberg@gaisler.com>
4968         * config/sparc/sparc.md (divdf3_fix): Add NOP and adjust length
4969         to prevent b2bst errata sequence.
4970         (sqrtdf2_fix): Likewise.
4972 2017-12-06  Jakub Jelinek  <jakub@redhat.com>
4974         PR tree-optimization/81945
4975         * cfgloop.h (FOR_EACH_LOOP_FN): Use FN instead of hardcoding fn.
4976         * tree-cfg.c (move_sese_region_to_fn): If any of the loops moved
4977         to dest_cfun has orig_loop_num set, either remap it to the new
4978         loop number if the loop got moved too, or clear it.
4980 2017-12-05  Steve Ellcey  <sellcey@cavium.com>
4982         * config/aarch64/thunderx2-t99.md (thunderx2t99_branch): Add trap
4983         to reservation.
4984         (thunderx2t99_nothing): New insn reservation.
4985         (thunderx2t99_mrs): New insn reservation.
4986         (thunderx2t99_multiple): New insn reservation.
4987         (thunderx2t99_alu_basi): Add bfx to reservation.
4988         (thunderx2t99_fp_cmp): Add fccmps and fccmpd to reservation.
4990 2017-12-05  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
4992         PR target/82248
4993         * config/arm/arm.md (probe_stack) : Use the 'o' constraint.
4995 2017-12-05  Bin Cheng  <bin.cheng@arm.com>
4997         * tree-ssa-dce.c (simple_dce_from_worklist): Move and rename from
4998         tree-ssa-pre.c::remove_dead_inserted_code.
4999         * tree-ssa-dce.h: New file.
5000         * tree-ssa-pre.c (tree-ssa-dce.h): Include new header file.
5001         (remove_dead_inserted_code): Move and rename to function
5002         tree-ssa-dce.c::simple_dce_from_worklist.
5003         (pass_pre::execute): Update use.
5005 2017-12-05  Richard Biener  <rguenther@suse.de>
5007         PR tree-optimization/83277
5008         * graphite-isl-ast-to-gimple.c (should_copy_to_new_region): Make sure
5009         to code-gen liveout vars.
5011 2017-12-05  Richard Sandiford  <richard.sandiford@linaro.org>
5013         * config/aarch64/aarch64-simd.md (aarch64_simd_bsldi_internal)
5014         (aarch64_simd_bsldi_alt): Check REG_P before GP_REGNUM_P.
5015         (aarch64_cm<optab>di, aarch64_cmtstdi): Add leading "&&" to
5016         split condition.
5018 2017-12-05  Max Filippov  <jcmvbkbc@gmail.com>
5020         * config/xtensa/xtensa.c (xtensa_asan_shadow_offset): New
5021         function.
5022         (TARGET_ASAN_SHADOW_OFFSET): New macro definition.
5023         * config/xtensa/xtensa.h (FRAME_GROWS_DOWNWARD): Set to 1 if
5024         ASAN is enabled.
5026 2017-12-05  Richard Biener  <rguenther@suse.de>
5028         * timevar.def (TV_TREE_RECIP, TV_TREE_SINCOS, TV_TREE_WIDEN_MUL): Add.
5029         * tree-ssa-math-opts.c (pass_data_cse_reciprocal): Use TV_TREE_RECIP.
5030         (pass_data_cse_sincos): Use TV_TREE_SINCOS.
5031         (pass_data_optimize_widening_mul): Use TV_TREE_WIDEN_MUL.
5033 2017-12-05  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
5035         * dbxout.c (dbxout_block): Grow buf to 30 bytes.
5037 2017-12-05  Martin Liska  <mliska@suse.cz>
5038             Jakub Jelinek  <jakub@redhat.com>
5040         * doc/invoke.texi: Document the options.
5041         * flag-types.h (enum sanitize_code): Add
5042         SANITIZE_POINTER_COMPARE and SANITIZE_POINTER_SUBTRACT.
5043         * ipa-inline.c (sanitize_attrs_match_for_inline_p): Add handling
5044         of SANITIZE_POINTER_COMPARE and SANITIZE_POINTER_SUBTRACT.
5045         * opts.c: Define new sanitizer options.
5046         * sanitizer.def (BUILT_IN_ASAN_POINTER_COMPARE): Likewise.
5047         (BUILT_IN_ASAN_POINTER_SUBTRACT): Likewise.
5049 2017-12-05  Julia Koval  <julia.koval@intel.com>
5051         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VNNI_SET,
5052         OPTION_MASK_ISA_AVX512VNNI_UNSET): New.
5053         (ix86_handle_option): Handle -mavx512vnni.
5054         * config/i386/cpuid.h (bit_AVX512VNNI): New bit.
5055         * config/i386/driver-i386.c (host_detect_local_cpu): Handle new bit.
5056         * config/i386/i386-c (__AVX512VNNI__): New.
5057         * config/i386/i386.c (ix86_target_string): Handle new option.
5058         (ix86_valid_target_attribute_inner_p): Handle new option.
5059         * config/i386/i386.h (TARGET_AVX512VNNI, TARGET_AVX512VNNI_P): New.
5060         * config/i386/i386.opt (mavx512vnni): New option.
5062 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
5064         PR target/81616
5065         * athlon.md: Disable for generic.
5066         * haswell.md: Enable for generic.
5067         * i386.c (ix86_sched_init_global): Add core hooks for generic.
5068         * x86-tune-sched.c (ix86_issue_rate): Increase issue rate for generic
5069         to 4.
5070         (ix86_adjust_cost): Move generic to haswell path.
5072 2017-12-04  Eric Botcazou  <ebotcazou@adacore.com>
5074         * config/sparc/sparc.c (sparc_do_work_around_errata): Use mem_ref
5075         instead of MEM_P in a couple more places.  Fix formatting issues.
5077 2017-12-04  Jim Wilson  <jimw@sifive.com>
5079         * config/riscv/riscv.c (riscv_for_each_saved_reg): Use GP_REG_LAST
5080         instead of GP_REG_LAST-1.
5081         (riscv_adjust_libcall_cfi_prologue): Likewise.
5082         (riscv_adjust_libcall_cri_epilogue): Likewise.
5083         * config/riscv/riscv.h (CALL_USED_REGISTERS): Change a6 to t6 in
5084         comment.
5086 2017-12-04  Luis Machado  <luis.machado@linaro.org>
5088         * ipa-pure-const.c (check_decl): Add missing newline.
5089         (state_from_flags): Likewise.
5091 2017-12-04  Jeff Law  <law@redhat.com>
5093         PR tree-optimizatin/78496
5094         * gimple-ssa-evrp-analyze.h
5095         (evrp_range_analyzer::get_vr_values): Simplify.
5096         * gimple-ssa-evrp-analyze.c: Corresponding changes.
5097         * tree-ssa-dom.c: Include alloc-pool.h, tree-vrp.h, vr-values.h
5098         and gimple-ssa-evrp-analyze.h.
5099         (dom_opt_dom_walker class): Add evrp_range_analyzer member.
5100         (simplify_stmt_for_jump_threading): Copy a blob of code from
5101         tree-vrp.c to use ranges to simplify statements.
5102         (dom_opt_dom_walker::before_dom_children): Call
5103         evrp_range_analyzer::{enter,record_ranges_from_stmt} methods.
5104         (dom_opt_dom_walker::after_dom_children): Similarly for
5105         evrp_range_analyzer::leave.
5106         (dom_opt_dom_walker::optimize_stmt): Use EVRP ranges to optimize
5107         conditionals.
5109         * gimple-ssa-evrp-analyze.c
5110         (evrp_range_analyzer::extract_range_from_stmt):  Always use
5111         vr_values::update_value_range so preexisting range info is
5112         medged with new range info, even if the new range is VR_VARYING.
5114 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
5116         * combine.c: Adjust comment.
5117         (use_crosses_set_p): Delete.
5118         (can_combine_p): Use modified_between_p instead of use_crosses_set_p.
5119         (try_combine): Ditto.
5121 2017-12-04  Richard Biener  <rguenther@suse.de>
5123         PR tree-optimization/83255
5124         * graphite-isl-ast-to-gimple.c (translate_isl_ast_node_for):
5125         Re-add zero-iteration check.
5127 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
5129         PR rtl-optimization/83245
5130         * lra.c (collect_non_operand_hard_regs): Treat clobbers of non-operand
5131         hard registers as earlyclobber, also if not in an asm.
5133 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
5135         PR bootstrap/83265
5136         Revert
5137         2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5139         PR target/43871
5140         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
5141         rs6000_cpu based on cpu_index, not tune_index.
5143 2017-12-04  Richard Biener  <rguenther@suse.de>
5145         PR tree-optimization/83238
5146         * graphite-scop-detection.c (scop_detection::merge_sese): Make
5147         code match comment, rejecting invalid SESE regions.
5149 2017-12-03  John David Anglin  <danglin@gcc.gnu.org>
5151         * config/pa/pa.c (pa_legitimate_address_p): For scaled indexing,
5152         require base operand is a REG_POINTER prior to reload on targets
5153         with non-equivalent space registers.
5155 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
5157         * ipa-cp.c (ipcp_lattice<valtype>::print): Update dumping.
5158         (update_specialized_profile): Fix updating of counts.
5159         (perhaps_add_new_callers): Likewise.
5161 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
5163         PR target/81616
5164         * x86-tune.def: Remove obsolette FIXMEs.
5165         (X86_TUNE_PARTIAL_FLAG_REG_STALL): Disable for generic
5166         (X86_TUNE_FUSE_CMP_AND_BRANCH_32, X86_TUNE_FUSE_CMP_AND_BRANCH_64,
5167         X86_TUNE_FUSE_CMP_AND_BRANCH_SOFLAGS, X86_TUNE_FUSE_ALU_AND_BRANCH):
5168         Enable for generic.
5169         (X86_TUNE_PAD_RETURNS): Disable for generic.
5171 2017-12-02  Jakub Jelinek  <jakub@redhat.com>
5173         PR tree-optimization/83170
5174         PR tree-optimization/83241
5175         * gimple-ssa-store-merging.c
5176         (imm_store_chain_info::try_coalesce_bswap): Update vuse field from
5177         gimple_vuse (ins_stmt) in case it has changed.
5178         (imm_store_chain_info::output_merged_store): Likewise.
5180         * tree-chkp.c (chkp_compute_bounds_for_assignment): Handle
5181         POINTER_DIFF_EXPR.
5183         PR c++/81212
5184         * tree-cfg.c (pass_warn_function_return::execute): Handle
5185         __builtin_ubsan_handle_missing_return like __builtin_unreachable
5186         with BUILTINS_LOCATION.
5188         PR target/78643
5189         PR target/80583
5190         * expr.c (get_inner_reference): If DECL_MODE of a non-bitfield
5191         is BLKmode for vector field with vector raw mode, use TYPE_MODE
5192         instead of DECL_MODE.
5194         * config/i386/i386-protos.h (standard_sse_constant_opcode): Change
5195         last argument to rtx pointer.
5196         * config/i386/i386.c (standard_sse_constant_opcode): Replace X argument
5197         with OPERANDS.  For AVX+ 128-bit VEX encoded instructions over 256-bit
5198         or 512-bit.  If setting EXT_REX_SSE_REG_P, use EVEX encoded insn
5199         depending on the chosen ISAs.
5200         * config/i386/i386.md (*movxi_internal_avx512f, *movoi_internal_avx,
5201         *movti_internal, *movdi_internal, *movsi_internal, *movtf_internal,
5202         *movdf_internal, *movsf_internal): Adjust standard_sse_constant_opcode
5203         callers.
5204         * config/i386/sse.md (mov<mode>_internal): Likewise.
5205         * config/i386/mmx.md (*mov<mode>_internal): Likewise.
5207 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5209         * doc/invoke.texi (-dp): Say that instruction cost is printed as well.
5211 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5213         * config/rs6000/rs6000-protos.h (emit_fusion_addis): Remove last two
5214         parameters from prototype.
5215         * config/rs6000/rs6000.c (emit_fusion_addis): Remove last two
5216         parameters.  Don't print a comment.
5217         (emit_fusion_gpr_load): Adjust.
5218         (emit_fusion_load_store): Adjust.
5219         * config/rs6000/rs6000.md (*fusion_p9_<mode>_constant): Adjust.
5220         * config/rs6000/vsx.md (two peepholes): Print the "vector load fusion"
5221         comment on the second line.
5223 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5225         PR target/43871
5226         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
5227         rs6000_cpu based on cpu_index, not tune_index.
5229 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5231         * final.c (output_asm_name): Print insn_cost.  Shorten output.  Print
5232         which_alternative instead of which_alternative + 1.
5233         (output_asm_insn): Print an extra tab if the template is short.
5235 2017-12-01  Jim Wilson  <jimw@sifive.com>
5237         * common.opt (use_gnu_debug_info_extensions): Delete DWARF_DEBUG from
5238         comment.
5239         * config/vx-common.h (DWARF_DEBUGGING_INFO): Delete undef.
5240         * doc/tm.texi.in (PREFERRED_DEBUGGING_TYPE): Delete DWARF_DEBUG
5241         reference.
5242         * doc/tm.texi: Regenerate.
5244 2017-12-01  Michael Meissner  <meissner@linux.vnet.ibm.com>
5246         PR target/81959
5247         * config/rs6000/rs6000.c (rs6000_address_for_fpconvert): Check for
5248         whether we can allocate pseudos before trying to fix an address.
5249         * config/rs6000/rs6000.md (float_<mode>si2_hw): Make sure the
5250         memory address is indexed or indirect.
5251         (floatuns_<mode>si2_hw2): Likewise.
5253 2017-12-01  Jason Merrill  <jason@redhat.com>
5255         * Makefile.in (TAGS): Add c-family/*.cc.
5257 2017-12-01  Wilco Dijkstra  <wdijkstr@arm.com>
5259         * config/aarch64/aarch64.md (call_insn): Use %c rather than %a.
5260         (call_value_insn): Likewise.
5261         (sibcall_insn): Likewise.
5262         (sibcall_value_insn): Likewise.
5263         (movsi_aarch64): Likewise.
5264         (movdi_aarch64): Likewise.
5265         (add_losym_): Likewise.
5266         (ldr_got_small_): Likewise.
5267         (ldr_got_small_sidi): Likewise.
5268         (ldr_got_small_28k_): Likewise.
5269         (ldr_got_small_28k_sidi): Likewise.
5270         * config/aarch64/aarch64.c (aarch64_print_address_internal):
5271         Move output_addr_const to symbolic case. Add error check.
5273 2017-12-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
5275         * config/s390/predicates.md (plus16_Q_operand): New predicate.
5276         * config/s390/s390.md: Disable MVC merging peephole if it would
5277         disable operand forwarding.
5278         (new peephole2): Split MVCs if it would turn them into up to 2
5279         forwardable MVCs.
5281 2017-12-01  Richard Biener  <rguenther@suse.de>
5283         PR tree-optimization/83232
5284         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Fix
5285         detection of same access. Instead of breaking the group here
5286         do not consider the duplicate.  Add comment explaining real fix.
5288 2017-12-01  Jonathan Wakely  <jwakely@redhat.com>
5290         * doc/md.texi (Insn Splitting): Fix "central flowgraph" typo.
5292 2017-12-01  Sudakshina Das  <sudi.das@arm.com>
5294         * config/arm/vfp.md (*movhf_vfp_fp16): Add conds attribute.
5296 2017-12-01  Jakub Jelinek  <jakub@redhat.com>
5298         * function.h (struct function): Remove cilk_frame_decl,
5299         is_cilk_function and calls_cilk_spawn fields.
5300         * tree-inline.h (struct copy_body_data): Remove remap_var_for_cilk
5301         field.
5302         * omp-simd-clone.c (simd_clone_clauses_extract): Don't clear
5303         cilk_elemental field.
5304         * cgraph.h (struct cgraph_simd_clone): Remove cilk_elemental field.
5305         * target.def: Adjust comment.
5306         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
5307         Don't test cilk_elemental.
5309         PR tree-optimization/83233
5310         * gimple-ssa-store-merging.c (nop_stats, bswap_stats): Use
5311         bswap_stat name for the struct.
5313         PR c/79153
5314         * tree.h (SWITCH_BREAK_LABEL_P): Define.
5315         * gimplify.c (collect_fallthrough_labels): Handle GIMPLE_BIND
5316         starting with a GIMPLE_SWITCH and ending with GIMPLE_LABEL with
5317         SWITCH_BREAK_LABEL_P set on the label.
5318         (gimplify_switch_expr): Set SWITCH_BREAK_LABEL_P on the label
5319         added for default case if it was missing and not all cases covered.
5320         Wrap GIMPLE_SWITCH and the switch_body_seq into a GIMPLE_BIND if
5321         switch_body_seq ends with a GIMPLE_LABEL with SWITCH_BREAK_LABEL_P
5322         set on the label.
5323         * tree-chrec.c (evolution_function_is_univariate_p): Add return true;
5324         to avoid -Wimplicit-fallthrough warning.
5325         * config/i386/i386.c (ix86_expand_special_args_builtin): Add
5326         FALLTHRU comment to avoid -Wimplicit-fallthrough warning.
5328         PR tree-optimization/83221
5329         * tree-ssa-reassoc.c (sort_by_operand_rank): Shift bb_rank
5330         down by 16.
5331         (init_reassoc): Formatting fix.
5333         PR sanitizer/81275
5334         * tree-cfg.c (group_case_labels_stmt): Don't optimize away
5335         C++ FE implicitly added __builtin_unreachable () until -Wreturn-type
5336         is diagnosed.
5338         PR sanitizer/83219
5339         * tree-cfg.c: Include asan.h.
5340         (gimple_seq_unreachable_p): Return false for -fsanitize=unreachable.
5342 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5344         * config/i386/i386.md: Fix AVX512 register width in AVX512 instruction.
5346 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5348         * config/i386/i386.c (standard_sse_constant_opcode): Fix wrong form for
5349         vpcmpeqd instruction.
5351 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5353         * config/i386/i386.c (standard_sse_constant_opcode): Fix
5354         registers type for 128bit mode.
5356 2017-11-30  Jakub Jelinek  <jakub@redhat.com>
5358         * spellcheck-tree.c (test_find_closest_identifier): Use ; instead
5359         of ;;.
5360         * gengtype-state.c (read_state_pair): Likewise.
5361         * gimple-fold.c (gimple_fold_builtin_string_compare): Likewise.
5362         * sel-sched-dump.c (dump_insn_rtx_1): Likewise.
5363         * ipa-cp.c (intersect_aggregates_with_edge): Likewise.
5364         * ifcvt.c (noce_try_store_flag_constants): Likewise.
5365         * tree-ssa-ccp.c (ccp_finalize): Likewise.
5366         * omp-grid.c (grid_process_kernel_body_copy): Likewise.
5367         * builtins.c (fold_builtin_3): Likewise.
5368         * graphite-scop-detection.c
5369         (scop_detection::stmt_has_simple_data_refs_p): Likewise.
5370         * hsa-gen.c (hsa_function_representation::hsa_function_representation):
5371         Likewise.
5373 2017-12-01  Maxim Ostapenko  <m.ostapenko@samsung.com>
5375         PR sanitizer/81697
5376         * asan.c (asan_protect_global): Add new ignore_decl_rtl_set_p
5377         parameter. Return true if ignore_decl_rtl_set_p is true and other
5378         conditions are satisfied.
5379         * asan.h (asan_protect_global): Add new parameter.
5380         * varasm.c (categorize_decl_for_section): Pass true as second parameter
5381         to asan_protect_global calls.
5383 2017-11-30  Jim Wilson  <jimw@sifive.com>
5385         * doc/invoke.texi (RISC-V Options): Delete nonexistent -mmemcpy and
5386         -mno-memcpy options.  For -mplt, -mfdiv, -mdiv, -msave-restore, and
5387         -mstrict-align, add info on default value.  Delete redundant lines for
5388         -mabi.  Add missing -mexplicit-relocs docs.
5390 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5392         * config/arc/arc.md (trap): New pattern.
5394 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5396         * config/arc/arc.c (hwloop_optimize): Prevent the last ZOL
5397         instruction to end into a delay slot.
5398         * config/arc/arc.md (cond_delay_insn): Check if the instruction
5399         can be placed into a delay slot against reg_note.
5401 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5403         * config/arc/arc.c (hwloop_optimize): Update hw-loop's end/start
5404         labels number of usages.
5406 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5408         * config/arc/arc.c (arc_cannot_substitue_mem_equiv_p): New
5409         function.
5410         (TARGET_CANNOT_SUBSTITUTE_MEM_EQUIV_P): Define.
5412 2017-11-30  Jakub Jelinek  <jakub@redhat.com>
5414         PR target/83210
5415         * internal-fn.c (expand_mul_overflow): Optimize unsigned
5416         multiplication by power of 2 constant into two shifts + comparison.
5418 2017-11-30  Jan Hubicka  <hubicka@ucw.cz>
5420         PR target/81616
5421         * config/i386/x86-tune-costs.h (generic_cost): Revise for modern CPUs.
5423 2017-11-30  Richard Biener  <rguenther@suse.de>
5425         PR tree-optimization/83202
5426         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Add
5427         allow_peel argument and guard peeling.
5428         (canonicalize_loop_induction_variables): Likewise.
5429         (canonicalize_induction_variables): Pass false.
5430         (tree_unroll_loops_completely_1): Pass unroll_outer to disallow
5431         peeling from cunrolli.
5433 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5435         * combine.c (try_combine): Print a message to dump file whenever
5436         I0, I1, or I2 cannot be combined into I3.
5438 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5440         PR rtl-optimization/83156
5441         PR rtl-optimization/82621
5442         * combine.c (try_combine): Don't split an I2 if one of the dests is
5443         set again before I3.  Allow unused dests.
5445 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5447         * config/rs6000/rs6000.md (*add<mode>3_carry_in_internal2): New.
5449 2017-11-29  Vladimir Makarov  <vmakarov@redhat.com>
5451         PR rtl-optimization/80818
5452         * lra.c (collect_non_operand_hard_regs): New arg insn.  Pass it
5453         recursively.  Use insn code for clobber.
5454         (lra_set_insn_recog_data): Pass the new arg to
5455         collect_non_operand_hard_regs.
5456         (add_regs_to_insn_regno_info): Pass insn instead of uid.  Use insn
5457         code for clobber.
5458         (lra_update_insn_regno_info): Pass insn to
5459         add_regs_to_insn_regno_info.
5461 2017-11-29  Jim Wilson  <jimw@sifive.com>
5462             Andrew Waterman  <andrew@sifive.com>
5464         * config/riscv/riscv.c (SINGLE_SHIFT_COST): New.
5465         (riscv_rtx_costs): Case ZERO_EXTRACT, match new pattern, and return
5466         SINGLE_SHIFT_COST.  Case LT and ZERO_EXTEND, likewise.  Case ASHIFT,
5467         use SINGLE_SHIFT_COST.
5468         * config/riscv/riscv.md (lshrsi3_zero_extend_1): New.
5469         (lshrsi3_zero_extend_2, lshrsi3_zero_extend_3): New.
5471 2017-11-29  Julia Koval  <julia.koval@intel.com>
5473         * config/i386/avx512vbmi2intrin.h (_mm512_shldv_epi16,
5474         _mm512_mask_shldv_epi16, _mm512_maskz_shldv_epi16, _mm512_shldv_epi32,
5475         _mm512_mask_shldv_epi32, _mm512_maskz_shldv_epi32, _mm512_shldv_epi64,
5476         _mm512_mask_shldv_epi64, _mm512_maskz_shldv_epi64): New intrinsics.
5477         * config/i386/avx512vbmi2vlintrin.h (_mm256_shldv_epi16,
5478         _mm256_mask_shldv_epi16, _mm256_maskz_shldv_epi16, _mm256_shldv_epi32,
5479         _mm256_mask_shldv_epi32, _mm256_maskz_shldv_epi32, _mm256_shldv_epi64,
5480         _mm256_mask_shldv_epi64, _mm256_maskz_shldv_epi64, _mm_shldv_epi16,
5481         _mm_mask_shldv_epi16, _mm_maskz_shldv_epi16, _mm_shldv_epi32,
5482         _mm_mask_shldv_epi32, _mm_maskz_shldv_epi32, _mm_shldv_epi64,
5483         _mm_mask_shldv_epi64, _mm_maskz_shldv_epi64): Ditto.
5484         * config/i386/i386-builtin.def (__builtin_ia32_vpshldv_v32hi,
5485         __builtin_ia32_vpshldv_v32hi_mask, __builtin_ia32_vpshldv_v32hi_maskz,
5486         __builtin_ia32_vpshldv_v16hi, __builtin_ia32_vpshldv_v16hi_mask,
5487         __builtin_ia32_vpshldv_v16hi_maskz, __builtin_ia32_vpshldv_v8hi,
5488         __builtin_ia32_vpshldv_v8hi_mask, __builtin_ia32_vpshldv_v8hi_maskz,
5489         __builtin_ia32_vpshldv_v16si, __builtin_ia32_vpshldv_v16si_mask,
5490         __builtin_ia32_vpshldv_v16si_maskz, __builtin_ia32_vpshldv_v8si,
5491         __builtin_ia32_vpshldv_v8si_mask, __builtin_ia32_vpshldv_v8si_maskz,
5492         __builtin_ia32_vpshldv_v4si, __builtin_ia32_vpshldv_v4si_mask,
5493         __builtin_ia32_vpshldv_v4si_maskz, __builtin_ia32_vpshldv_v8di,
5494         __builtin_ia32_vpshldv_v8di_mask, __builtin_ia32_vpshldv_v8di_maskz,
5495         __builtin_ia32_vpshldv_v4di, __builtin_ia32_vpshldv_v4di_mask,
5496         __builtin_ia32_vpshldv_v4di_maskz, __builtin_ia32_vpshldv_v2di,
5497         __builtin_ia32_vpshldv_v2di_mask,
5498         __builtin_ia32_vpshldv_v2di_maskz): New builtins.
5499         * config/i386/sse.md (vpshldv_<mode>, vpshldv_<mode>_mask,
5500         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): New patterns.
5502 2017-11-29  Julia Koval  <julia.koval@intel.com>
5504         * config/i386/avx512vbmi2intrin.h (_mm512_shrdv_epi16,
5505         _mm512_mask_shrdv_epi16, _mm512_maskz_shrdv_epi16, _mm512_shrdv_epi32,
5506         _mm512_mask_shrdv_epi32, _mm512_maskz_shrdv_epi32, _mm512_shrdv_epi64,
5507         _mm512_mask_shrdv_epi64, _mm512_maskz_shrdv_epi64): New intrinsics.
5508         * config/i386/avx512vbmi2vlintrin.h (_mm256_shrdv_epi16,
5509         _mm256_mask_shrdv_epi16, _mm256_maskz_shrdv_epi16, _mm256_shrdv_epi32,
5510         _mm256_mask_shrdv_epi32, _mm256_maskz_shrdv_epi32, _mm256_shrdv_epi64,
5511         _mm256_mask_shrdv_epi64, _mm256_maskz_shrdv_epi64, _mm_shrdv_epi16,
5512         _mm_mask_shrdv_epi16, _mm_maskz_shrdv_epi16, _mm_shrdv_epi32,
5513         _mm_mask_shrdv_epi32, _mm_maskz_shrdv_epi32, _mm_shrdv_epi64,
5514         _mm_mask_shrdv_epi64, _mm_maskz_shrdv_epi64): Ditto.
5515         * config/i386/i386-builtin-types.def (V32HI_FTYPE_V32HI_V32HI_V32HI,
5516         V32HI_FTYPE_V32HI_V32HI_V32HI_INT, V16HI_FTYPE_V16HI_V16HI_V16HI_INT,
5517         V8HI_FTYPE_V8HI_V8HI_V8HI_INT, V8SI_FTYPE_V8SI_V8SI_V8SI_INT,
5518         V4SI_FTYPE_V4SI_V4SI_V4SI_INT, V8DI_FTYPE_V8DI_V8DI_V8DI,
5519         V8DI_FTYPE_V8DI_V8DI_V8DI_INT, V4DI_FTYPE_V4DI_V4DI_V4DI_INT,
5520         V16SI_FTYPE_V16SI_V16SI_V16SI, V16SI_FTYPE_V16SI_V16SI_V16SI_INT,
5521         V2DI_FTYPE_V2DI_V2DI_V2DI_INT): New types.
5522         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
5523         * config/i386/sse.md (vpshrdv_<mode>, vpshrdv_<mode>_mask,
5524         vpshrdv_<mode>_maskz, vpshrdv_<mode>_maskz_1): New pattern.
5526 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5528         * config/sparc/sparc.c (sparc_do_work_around_errata): Treat the
5529         movsi_pic_gotdata_op instruction as a load for the UT699 errata
5530         workaround.
5532 2017-11-29  Martin Aberg  <maberg@gaisler.com>
5534         * config/sparc/sparc.md (divdf3_fix): Add NOP and adjust length
5535         to prevent b2bst errata sequence.
5536         (sqrtdf2_fix): Likewise.
5538 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5540         * config/sparc/sparc.c (fpop_reg_depend_p): New function.
5541         (div_sqrt_insn_p): New function.
5542         (sparc_do_work_around_errata): Insert NOP instructions to
5543         prevent sequences that could trigger the TN-0013 errata for
5544         certain LEON3 processors.
5545         (pass_work_around_errata::gate): Also test sparc_fix_lost_divsqrt.
5546         (sparc_option_override): Set sparc_fix_lost_divsqrt appropriately.
5547         * config/sparc/sparc.md (fix_lost_divsqrt): New attribute.
5548         (in_branch_delay): Prevent div and sqrt in delay slot if
5549         fix_lost_divsqrt.
5550         * config/sparc/sparc.opt (sparc_fix_lost_divsqrt): New variable.
5552 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5554         * config/sparc/sparc.c (atomic_insn_p): New function.
5555         (sparc_do_work_around_errata): Insert NOP instructions to
5556         prevent sequences that could trigger the TN-0010 errata for
5557         UT700.
5558         * config/sparc/sync.md (atomic_compare_and_swap_leon3_1): Make
5559         instruction referable in atomic_insns_p.
5561 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5563         * config/sparc/sync.md (swapsi): 16-byte align if sparc_fix_gr712rc.
5564         (atomic_compare_and_swap_leon3_1): Likewise.
5565         (ldstub): Likewise.
5567 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5569         * config/sparc/sparc.c (fpop_insn_p): New function.
5570         (sparc_do_work_around_errata): Insert NOP instructions to
5571         prevent sequences that could trigger the TN-0012 errata for
5572         GR712RC.
5573         (pass_work_around_errata::gate): Also test sparc_fix_gr712rc.
5574         * config/sparc/sparc.md (fix_gr712rc): New attribute.
5575         (in_branch_annul_delay): Prevent floating-point instructions
5576         in delay slot of annulled integer branch.
5578 2017-11-29  Richard Biener  <rguenther@suse.de>
5580         PR tree-optimization/83202
5581         * tree-vect-slp.c (scalar_stmts_set_t): New typedef.
5582         (bst_fail): Use it.
5583         (vect_analyze_slp_cost_1): Add visited set, do not account SLP
5584         nodes vectorized to the same stmts multiple times.
5585         (vect_analyze_slp_cost): Allocate a visited set and pass it down.
5586         (vect_analyze_slp_instance): Adjust.
5587         (scalar_stmts_to_slp_tree_map_t): New typedef.
5588         (vect_schedule_slp_instance): Add a map recording the SLP node
5589         representing the vectorized stmts for a set of scalar stmts.
5590         Avoid code-generating redundancies.
5591         (vect_schedule_slp): Allocate map and pass it down.
5593 2017-11-29  Nathan Sidwell  <nathan@acm.org>
5595         PR c++/83187
5596         * tree.c (build_complex_type): Fix canonicalization.  Only fill in
5597         type if it is new.
5599 2017-11-29  Wilco Dijkstra  <wdijkstr@arm.com>
5601         * config/aarch64/aarch64.c (aarch64_print_operand): Add new
5602         cases for printing LDP/STP memory addresses.
5603         (aarch64_print_address_internal): Renamed from
5604         aarch64_print_operand_address, added parameter, add Pmode check.
5605         (aarch64_print_ldpstp_address): New function for LDP/STP addresses.
5606         (aarch64_print_operand_address): Indirect to
5607         aarch64_print_address_internal.
5608         * config/aarch64/aarch64-simd.md (store_pair_lanes): Use new
5609         'y' operand output specifier.
5611 2017-11-29  Jakub Jelinek  <jakub@redhat.com>
5613         PR middle-end/83185
5614         * tree.c (build_simple_mem_ref_loc): Handle
5615         get_addr_base_and_unit_offset returning a MEM_REF.
5617         PR middle-end/80929
5618         * rtlanal.c (seq_cost): For non-single_set insns try to use insn_cost.
5620         PR target/80819
5621         * config/i386/sse.md (vec_concatv2di): Remove * from (=Yr,0,*rm)
5622         alternative.
5624 2017-11-28  David Malcolm  <dmalcolm@redhat.com>
5626         * diagnostic-show-locus.c (layout::print_trailing_fixits): Handle
5627         m_x_offset.
5628         (layout::move_to_column): Likewise.
5630 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5632         PR sanitizer/81275
5633         * tree.c (block_may_fallthru): Return false if SWITCH_ALL_CASES_P
5634         is set on SWITCH_EXPR and !block_may_fallthru (SWITCH_BODY ()).
5636 2017-11-28  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
5637             Martin Jambor  <mjambor@suse.cz>
5639         PR ipa/82808
5640         * tree.h (expr_type_first_operand_type_p): Declare
5641         * tree.c (expr_type_first_operand_type_p): New function.
5642         * ipa-prop.h (ipa_get_type): Allow i to be out of bounds.
5643         (ipa_value_from_jfunc): Adjust declaration.
5644         * ipa-cp.c (ipa_get_jf_pass_through_result): New parameter RES_TYPE.
5645         Use it as result type for arithmetics, unless it is NULL in which case
5646         be more conservative.
5647         (ipa_value_from_jfunc): New parameter PARM_TYPE, pass it to
5648         ipa_get_jf_pass_through_result.
5649         (propagate_vals_across_pass_through): Likewise.
5650         (propagate_scalar_across_jump_function): New parameter PARM_TYPE, pass
5651         is to propagate_vals_across_pass_through.
5652         (propagate_constants_across_call): Pass PARM_TYPE to
5653         propagate_scalar_across_jump_function.
5654         (find_more_scalar_values_for_callers_subset): Pass parameter type to
5655         ipa_value_from_jfunc.
5656         (cgraph_edge_brings_all_scalars_for_node): Likewise.
5657         * ipa-fnsummary.c (evaluate_properties_for_edge): Renamed parms_info
5658         to caller_parms_info, pass parameter type to ipa_value_from_jfunc.
5659         * ipa-prop.c (try_make_edge_direct_simple_call): New parameter
5660         target_type, pass it to ipa_value_from_jfunc.
5661         (update_indirect_edges_after_inlining): Pass parameter type to
5662         try_make_edge_direct_simple_call.
5664 2017-11-28  Jeff Law  <law@redhat.com>
5666         * gimple-ssa-evrp-analyze.c
5667         (evrp_range_analyzer::record_ranges_from_phis): Only use SCEV to
5668         refine ranges if scev_initialized_p returns true.
5669         * vr-values.c (vr_values::extract_range_from_phi_node): Likewise.
5671 2017-11-28  Julia Koval  <julia.koval@intel.com>
5673         * config/i386/avx512vbmi2intrin.h (_mm512_shrdi_epi16,
5674         _mm512_mask_shrdi_epi16, _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
5675         _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32, _mm512_shrdi_epi64,
5676         _mm512_mask_shrdi_epi64, _mm512_maskz_shrdi_epi64): New intrinsics.
5677         * config/i386/avx512vbmi2vlintrin.h (_mm256_shrdi_epi16,
5678         _mm256_mask_shrdi_epi16, _mm256_maskz_shrdi_epi16,
5679         _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32, _mm256_shrdi_epi32,
5680         _mm256_mask_shrdi_epi64, _mm256_maskz_shrdi_epi64, _mm256_shrdi_epi64,
5681         _mm_mask_shrdi_epi16, _mm_maskz_shrdi_epi16, _mm_shrdi_epi16,
5682         _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32, _mm_shrdi_epi32,
5683         _mm_mask_shrdi_epi64, _mm_maskz_shrdi_epi64, _mm_shrdi_epi64): Ditto.
5684         * config/i386/i386-builtin.def (__builtin_ia32_vpshrd_v32hi,
5685         __builtin_ia32_vpshrd_v32hi_mask, __builtin_ia32_vpshrd_v16hi,
5686         __builtin_ia32_vpshrd_v16hi_mask, __builtin_ia32_vpshrd_v8hi,
5687         __builtin_ia32_vpshrd_v8hi_mask, __builtin_ia32_vpshrd_v16si,
5688         __builtin_ia32_vpshrd_v16si_mask, __builtin_ia32_vpshrd_v8si,
5689         __builtin_ia32_vpshrd_v8si_mask, __builtin_ia32_vpshrd_v4si,
5690         __builtin_ia32_vpshrd_v4si_mask, __builtin_ia32_vpshrd_v8di,
5691         __builtin_ia32_vpshrd_v8di_mask, __builtin_ia32_vpshrd_v4di,
5692         __builtin_ia32_vpshrd_v4di_mask, __builtin_ia32_vpshrd_v2di,
5693         __builtin_ia32_vpshrd_v2di_mask): New builtins.
5694         * config/i386/sse.md (vpshrd_<mode><mask_name>): New pattern.
5696 2017-11-28  Julia Koval  <julia.koval@intel.com>
5698         * config/i386/avx512vbmi2intrin.h (_mm512_shldi_epi16,
5699         _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16, _mm512_shldi_epi32,
5700         _mm512_mask_shldi_epi32, _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
5701         _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): New intrinsics.
5702         * config/i386/avx512vbmi2vlintrin.h (_mm256_shldi_epi16,
5703         _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
5704         _mm256_mask_shldi_epi32, _mm256_maskz_shldi_epi32, _mm256_shldi_epi32,
5705         _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64, _mm256_shldi_epi64,
5706         _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16, _mm_shldi_epi16,
5707         _mm_mask_shldi_epi32, _mm_maskz_shldi_epi32, _mm_shldi_epi32,
5708         _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64, _mm_shldi_epi64): Ditto.
5709         * config/i386/i386-builtin-types.def (V32HI_FTYPE_V32HI_V32HI_INT,
5710         V32HI_FTYPE_V32HI_V32HI_INT_V32HI_INT, V16SI_FTYPE_V16SI_V16SI_INT,
5711         V16SI_FTYPE_V16SI_V16SI_INT_V16SI_INT,
5712         V8DI_FTYPE_V8DI_V8DI_INT_V8DI_INT, V8SI_FTYPE_V8SI_V8SI_INT_V8SI_INT,
5713         V16HI_FTYPE_V16HI_V16HI_INT_V16HI_INT,
5714         V4DI_FTYPE_V4DI_V4DI_INT_V4DI_INT,
5715         V8HI_FTYPE_V8HI_V8HI_INT_V8HI_INT,
5716         V4SI_FTYPE_V4SI_V4SI_INT_V4SI_INT,
5717         V2DI_FTYPE_V2DI_V2DI_INT_V2DI_INT): New types.
5718         * config/i386/i386-builtin.def (__builtin_ia32_vpshld_v32hi,
5719         __builtin_ia32_vpshld_v32hi_mask, __builtin_ia32_vpshld_v16hi,
5720         __builtin_ia32_vpshld_v16hi_mask, __builtin_ia32_vpshld_v8hi,
5721         __builtin_ia32_vpshld_v8hi_mask, __builtin_ia32_vpshld_v16si,
5722         __builtin_ia32_vpshld_v16si_mask, __builtin_ia32_vpshld_v8si,
5723         __builtin_ia32_vpshld_v8si_mask, __builtin_ia32_vpshld_v4si,
5724         __builtin_ia32_vpshld_v4si_mask, __builtin_ia32_vpshld_v8di,
5725         __builtin_ia32_vpshld_v8di_mask, __builtin_ia32_vpshld_v4di,
5726         __builtin_ia32_vpshld_v4di_mask, __builtin_ia32_vpshld_v2di,
5727         __builtin_ia32_vpshld_v2di_mask): New builtins.
5728         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
5729         * config/i386/sse.md (vpshld_<mode><mask_name>): New pattern.
5731 2017-11-28  Richard Biener  <rguenther@suse.de>
5733         PR tree-optimization/80776
5734         * gimple-ssa-evrp-analyze.h (evrp_range_analyzer::set_ssa_range_info):
5735         Declare.
5736         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
5737         New function.
5738         (evrp_range_analyzer::record_ranges_from_incoming_edges):
5739         If the incoming edge is an effective fallthru because the other
5740         edge only reaches a __builtin_unreachable () then record ranges
5741         derived from the controlling condition in SSA info.
5742         (evrp_range_analyzer::record_ranges_from_phis): Use set_ssa_range_info.
5743         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5745 2017-11-28  Olivier Hainque  <hainque@adacore.com>
5747         * Makefile.in (SELFTEST_FLAGS): Use nul instead of /dev/null
5748         on mingw build hosts.
5750 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5752         PR debug/81307
5753         * dbxout.c (dbx_block_with_cold_children): Fix function comment.
5755 2017-11-28  Richard Biener  <rguenther@suse.de>
5757         PR middle-end/83141
5758         * gimple-fold.c (gimple_fold_builtin_memory_op): For aggregate
5759         copies generated from memcpy use a character array as reference
5760         type.
5762 2017-11-28  Julia Koval  <julia.koval@intel.com>
5763             Sebastian Peryt  <sebastian.peryt@intel.com>
5765         * Makefile.in (cilkplus.def, cilk-builtins.def, c-family/cilk.o,
5766         c-family/c-cilkplus.o, c-family/array-notation-common.o,
5767         cilk-common.o, cilk.h, cilk-common.c): Remove.
5768         * builtin-types.def
5769         (BT_FN_INT_PTR_PTR_PTR_FTYPE_BT_INT_BT_PTR_BT_PTR_BT_PTR): Remove.
5770         * builtins.c (is_builtin_name): Remove cilkplus condition.
5771         (BUILT_IN_CILK_DETACH, BUILT_IN_CILK_POP_FRAME): Remove.
5772         * builtins.def (DEF_CILK_BUILTIN_STUB, DEF_CILKPLUS_BUILTIN,
5773         cilk-builtins.def, cilkplus.def): Remove.
5774         * cif-code.def (CILK_SPAWN): Remove.
5775         * cilk-builtins.def: Delete.
5776         * cilk-common.c: Ditto.
5777         * cilk.h: Ditto.
5778         * cilkplus.def: Ditto.
5779         * config/darwin.h (fcilkplus): Delete.
5780         * cppbuiltin.c: Ditto.
5781         * doc/extend.texi: Remove cilkplus doc.
5782         * doc/generic.texi: Ditto.
5783         * doc/invoke.texi: Ditto.
5784         * doc/passes.texi: Ditto.
5785         * gcc.c (fcilkplus): Remove.
5786         * gengtype.c (cilk.h): Remove.
5787         * gimple-pretty-print.c (dump_gimple_omp_for): Remove cilkplus
5788         support.
5789         * gimple.h (GF_OMP_FOR_KIND_CILKFOR, GF_OMP_FOR_KIND_CILKSIMD):
5790         Remove.
5791         * gimplify.c (gimplify_return_expr, maybe_fold_stmt,
5792         gimplify_call_expr,
5793         is_gimple_stmt, gimplify_modify_expr, gimplify_scan_omp_clauses,
5794         gimplify_adjust_omp_clauses, gimplify_omp_for, gimplify_expr): Remove
5795         cilkplus conditions.
5796         * ipa-fnsummary.c (ipa_dump_fn_summary, compute_fn_summary,
5797         inline_read_section): Ditto.
5798         * ipa-inline-analysis.c (cilk.h): Remove.
5799         * ira.c (ira_setup_eliminable_regset): Remove cilkplus support.
5800         * lto-wrapper.c (merge_and_complain, append_compiler_options,
5801         append_linker_options): Remove condition for fcilkplus.
5802         * lto/lto-lang.c (cilk.h): Remove.
5803         (lto_init): Remove condition for fcilkplus.
5804         * omp-expand.c (expand_cilk_for_call): Delete.
5805         (expand_omp_taskreg, expand_omp_for_static_chunk,
5806         expand_omp_for): Remove cilkplus
5807         conditions.
5808         (expand_cilk_for): Delete.
5809         * omp-general.c (omp_extract_for_data): Remove cilkplus support.
5810         * omp-low.c (scan_sharing_clauses, create_omp_child_function,
5811         execute_lower_omp, diagnose_sb_0): Ditto.
5812         * omp-simd-clone.c (simd_clone_clauses_extract): Ditto.
5813         * tree-core.h (OMP_CLAUSE__CILK_FOR_COUNT_): Delete.
5814         * tree-nested.c: Ditto.
5815         * tree-pretty-print.c (dump_omp_clause): Remove cilkplus support.
5816         (dump_generic_node): Ditto.
5817         * tree.c (OMP_CLAUSE__CILK_FOR_COUNT_): Delete.
5818         * tree.def (cilk_simd, cilk_for, cilk_spawn_stmt, cilk_sync_stmt):
5819         Delete.
5820         * tree.h (CILK_SPAWN_FN, EXPR_CILK_SPAWN): Delete.
5822 2017-11-28  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
5824         * config/aarch64/aarch64.md (div<mode>3): Change check to TARGET_FLOAT.
5825         * config/aarch64/aarch64.c (aarch64_emit_approx_div): Add early exit
5826         for vector mode and !TARGET_SIMD.
5828 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5830         * tree.def (SWITCH_EXPR): Change from 3 operand to 2 operand tree.
5831         Adjust comment.
5832         * tree.h (SWITCH_LABELS): Remove.
5833         * gimplify.c (gimplify_switch_expr): Don't test SWITCH_LABELS,
5834         assert SWITCH_BODY is non-NULL.
5835         * tree-pretty-print.c (dump_generic_node): Remove SWITCH_LABELS
5836         handling.
5837         * tree.c (block_may_fallthru): Always return true; for SWITCH_EXPR.
5839         PR tree-optimization/80788
5840         * match.pd (X +- C1 CMP C2 -> X CMP C2 -+ C1): If res
5841         has TREE_OVERFLOW set, call drop_tree_overflow.
5843 2017-11-28  Richard Biener  <rguenther@suse.de>
5845         PR tree-optimization/83158
5846         * tree-vrp.c (intersect_ranges): Prefer ~[0, 0] in a few more cases.
5848 2017-11-28  Segher Boessenkool  <segher@kernel.crashing.org>
5850         PR 81288/target
5851         * config/rs6000/rs6000.c (rs6000_rtx_costs): Do not handle
5852         TARGET_ISEL && !TARGET_MFCRF differently.  Simplify code.
5854 2017-11-27  Segher Boessenkool  <segher@kernel.crashing.org>
5856         * config/rs6000/rs6000.md (<code><GPR:mode><GPR2:mode>2_isel): Change
5857         LT/GT/LTU/GTU to LE/GE/LEU/GEU where possible.
5859 2017-11-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
5861         PR middle_end/82333
5862         * varasm.c (compare_constant): Take the mode of the constants into
5863         account when comparing floating point constants.
5865 2017-11-27  Gerald Pfeifer  <gerald@pfeifer.com>
5867         * hash-set.h (DEFINE_DEBUG_HASH_SET): Remove static qualifier
5868         from explicit instantiation of debug_helper.
5869         * vec.h (DEFINE_DEBUG_VEC): Ditto.
5871 2017-11-27  Richard Biener  <rguenther@suse.de>
5873         * gimple-fold.c (gimple_fold_builtin_memory_op): Remove dead code,
5874         refactor a bit.
5876 2017-11-27  Richard Biener  <rguenther@suse.de>
5878         * tree.c (wide_int_to_tree): Free discarded INTEGER_CST.
5879         (type_hash_canon): Also clear int_cst_hash_table entry for
5880         TYPE_MIN/MAX_VALUE.
5881         (build_nonstandard_integer_type): Hash all TYPE_MAX_VALUEs.
5883 2017-11-27  Tamar Christina  <tamar.christina@arm.com>
5885         * doc/extend.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5886         * doc/invoke.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5887         * doc/sourcebuild.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5889 2017-11-27  Markus Trippelsdorf  <markus@trippelsdorf.de>
5891         * hash-map.h (gt_cleare_cache): Avoid UB.
5893 2017-11-27  Eric Botcazou  <ebotcazou@adacore.com>
5895         * cfgloop.h (struct loop): Document usage of USHRT_MAX for unroll.
5896         * loop-unroll.c (decide_unroll_constant_iterations): Implement it.
5897         (decide_unroll_runtime_iterations): Likewise.
5898         (decide_unroll_stupid): Likewise.
5900 2017-11-27  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
5902         PR target/83109
5903         * config/i386/i386.md: Add a loop with incssp.
5905 2017-11-27  Martin Jambor  <mjambor@suse.cz>
5907         PR tree-optimization/81248
5908         * tree-sra.c (splice_param_accesses): Remove size check.
5909         (decide_one_param_reduction): Fix size check.
5910         * gimple-pretty-print.c (dump_profile): Silence warning.
5911         * params.def (PARAM_IPA_SRA_PTR_GROWTH_FACTOR): Adjust description.
5913 2017-11-27  Jakub Jelinek  <jakub@redhat.com>
5915         PR debug/81307
5916         * dbxout.c (lastlineno): New variable.
5917         (dbx_debug_hooks): Use dbxout_switch_text_section as
5918         switch_text_section debug hook.
5919         (dbxout_function_end): Switch to current_function_section
5920         rather than function_section.  If crtl->has_bb_partition,
5921         output just one N_FUN, depending on in_cold_section_p.
5922         (dbxout_source_line): Remember last lineno in lastlineno.
5923         (dbxout_switch_text_section): New function.
5924         (dbxout_function_decl): Adjust dbxout_block caller.
5925         (dbx_block_with_cold_children): New function.
5926         (dbxout_block): Return true if any LBRAC/RBRAC have been
5927         emitted.  Use dbx_block_with_cold_children at depth == 0
5928         in second partition.  Add PARENT_BLOCKNUM argument, pass
5929         it optionally adjusted to children.  Output LBRAC/RBRAC
5930         around recursive call only if the block is in the current
5931         partition, if not and anything was output, emit empty
5932         range LBRAC/RBRAC.
5933         * final.c (final_scan_insn): Compute cold_function_name
5934         before calling switch_text_section debug hook.  Call
5935         that hook even if dwarf2out_do_frame if not emitting
5936         dwarf debug info.
5938         PR target/83100
5939         * varasm.c (bss_initializer_p): Return true for DECL_COMMON
5940         TREE_READONLY decls.
5942 2017-11-27  Markus Trippelsdorf  <markus@trippelsdorf.de>
5944         PR rtl-optimization/82488
5945         * expr.c (fixup_args_size_notes): Avoid signed integer overflow.
5947 2017-11-26  Julia Koval  <julia.koval@intel.com>
5949         * config/i386/i386.c (processor_target_table): Add skylake_cost for
5950         skylake-avx512.
5951         * config/i386/x86-tune-costs.h (skylake_memcpy, skylake_memset,
5952         skylake_cost): New.
5954 2017-11-26  Julia Koval  <julia.koval@intel.com>
5956         * config/i386/driver-i386.c (host_detect_local_cpu):
5957         Detect skylake-avx512.
5959 2017-11-26  Julia Koval  <julia.koval@intel.com>
5961         * config.gcc: Add -march=cannonlake.
5962         * config/i386/driver-i386.c (host_detect_local_cpu): Detect cannonlake.
5963         * config/i386/i386-c.c (ix86_target_macros_internal): Handle cannonlake.
5964         * config/i386/i386.c (processor_costs): Add m_CANNONLAKE.
5965         (PTA_CANNONLAKE): New.
5966         (processor_target_table): Add cannonlake.
5967         (ix86_option_override_internal): Ditto.
5968         (fold_builtin_cpu): Ditto.
5969         (get_builtin_code_for_version): Handle cannonlake.
5970         (M_INTEL_COREI7_CANNONLAKE): New.
5971         * config/i386/i386.h (TARGET_CANNONLAKE, PROCESSOR_CANNONLAKE): New.
5972         * doc/invoke.texi: Add -march=cannonlake.
5974 2017-11-14  Boris Kolpackov  <boris@codesynthesis.com>
5976         * plugin.c (add_new_plugin): Use platform-specific library extensions.
5977         (try_init_one_plugin): Alternative implementation for MinGW.
5978         * Makefile.in (plugin_implib): New.
5979         (gengtype-lex.c): Fix broken AIX workaround.
5980         * configure: Regenerate.
5981         * doc/plugins.texi: Document support for MinGW.
5983 2017-11-25  Jakub Jelinek  <jakub@redhat.com>
5985         PR rtl-optimization/81553
5986         * combine.c (simplify_if_then_else): In (if_then_else COND (OP Z C1) Z)
5987         to (OP Z (mult COND (C1 * STORE_FLAG_VALUE))) optimization, if OP
5988         is a shift where C1 has different mode than the whole shift, use C1's
5989         mode for MULT rather than the shift's mode.
5991         PR target/82848
5992         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Don't fold
5993         builtins not enabled in the currently selected ISA.
5995 2017-11-24  Jackson Woodruff  <jackson.woodruff@arm.com>
5997         PR tree-optimization/71026
5998         * tree-ssa-math-opts (is_division_by_square, is_square_of): New.
5999         (insert_reciprocals): Change to insert reciprocals before a division
6000         by a square and to insert the square of a reciprocal.
6001         (execute_cse_reciprocals_1): Change to consider division by a square.
6002         (register_division_in): Add importance parameter.
6004 2017-11-24  Richard Biener  <rguenther@suse.de>
6006         PR tree-optimization/82402
6007         * tree-vect-loop-manip.c (create_lcssa_for_virtual_phi): Properly
6008         set SSA_NAME_OCCURS_IN_ABNORMAL_PHI.
6010 2017-11-24  Marc Glisse  <marc.glisse@inria.fr>
6012         * match.pd (0-ptr): New transformation.
6014 2017-11-24  Jan Hubicka  <hubicka@ucw.cz>
6016         PR bootstrap/83015
6017         * ipa-inline.c (inline_small_functions): Set current badnes correctly
6018         when skipping checking.
6020 2017-11-24  Richard Biener  <rguenther@suse.de>
6022         PR tree-optimization/83128
6023         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Handle STRING_CSTs.
6024         (vn_reference_lookup_3): Likewise.
6026 2017-11-24  Jakub Jelinek  <jakub@redhat.com>
6028         PR sanitizer/83014
6029         * ubsan.c (ubsan_type_descriptor): Use pp_unsigned_wide_integer
6030         instead of pp_printf with HOST_WIDE_INT_PRINT_DEC.  Avoid calling
6031         tree_to_uhwi twice.
6033         * tree-object-size.c (pass_through_call): Use gimple_call_return_flags
6034         ERF_RETURN*ARG* for builtins other than BUILT_IN_ASSUME_ALIGNED,
6035         check for the latter with gimple_call_builtin_p.  Do not handle
6036         BUILT_IN_STPNCPY_CHK which is not a pass through call.
6038 2017-11-24  Christophe Lyon  <christophe.lyon@linaro.org>
6040         * config/arm/arm_neon.h: Fix pragma GCC push_options before
6041         vdot_u32.
6043 2017-11-23  Julia Koval  <julia.koval@intel.com>
6045         * config/i386/avx512vbmi2intrin.h (_mm512_mask_expand_epi8,
6046         _mm512_maskz_expand_epi8, _mm512_mask_expandloadu_epi8,
6047         _mm512_maskz_expandloadu_epi8, _mm512_mask_expand_epi16,
6048         _mm512_maskz_expand_epi16, _mm512_mask_expandloadu_epi16,
6049         _mm512_maskz_expandloadu_epi16): New intrinsics.
6050         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_expand_epi8,
6051         _mm_maskz_expand_epi8, _mm_mask_expandloadu_epi8,
6052         _mm_maskz_expandloadu_epi8, _mm_mask_expand_epi16,
6053         _mm_maskz_expand_epi16, _mm_mask_expandloadu_epi16,
6054         _mm_maskz_expandloadu_epi16, _mm256_mask_expand_epi16,
6055         _mm256_maskz_expand_epi16, _mm256_mask_expandloadu_epi16,
6056         _mm256_maskz_expandloadu_epi16, _mm256_mask_expand_epi8,
6057         _mm256_maskz_expand_epi8, _mm256_mask_expandloadu_epi8,
6058         _mm256_maskz_expandloadu_epi8): New intrinsics.
6059         * config/i386/i386-builtin-types.def (V64QI_FTYPE_PCV64QI_V64QI_UDI,
6060         V32HI_FTYPE_PCV32HI_V32HI_USI, V32QI_FTYPE_PCV32QI_V32QI_USI,
6061         V16HI_FTYPE_PCV16HI_V16HI_UHI, V16QI_FTYPE_PCV16QI_V16QI_UHI,
6062         V8HI_FTYPE_PCV8HI_V8HI_UQI): New types.
6063         * config/i386/i386.c (ix86_expand_special_args_builtin): Use new types.
6064         * config/i386/sse.md (VI248_VLBW): New iterator.
6065         (expand<mode>_mask, expand<mode>_maskz): New patterns.
6067 2017-11-23  Julia Koval  <julia.koval@intel.com>
6069         * config.gcc (avx512vbmi2intrin.h, avx512vbmi2vlintrin): New headers.
6070         * config/i386/avx512vbmi2intrin.h (_mm512_mask_compress_epi8,
6071         _mm512_maskz_compress_epi8, _mm512_mask_compressstoreu_epi8,
6072         _mm512_mask_compress_epi16, _mm512_maskz_compress_epi16,
6073         _mm512_mask_compressstoreu_epi16): New.
6074         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_compress_epi8,
6075         _mm_maskz_compress_epi8, _mm256_mask_compressstoreu_epi16,
6076         _mm_mask_compress_epi16, _mm_maskz_compress_epi16,
6077         _mm256_mask_compress_epi16, _mm256_maskz_compress_epi16,
6078         _mm_mask_compressstoreu_epi8, _mm_mask_compressstoreu_epi16,
6079         _mm256_mask_compress_epi8, _mm256_maskz_compress_epi8,
6080         _mm256_mask_compressstoreu_epi8): New.
6081         * config/i386/i386-builtin-types.def (VOID_FTYPE_PV64QI_V64QI_UDI,
6082         VOID_FTYPE_PV32HI_V32HI_USI, VOID_FTYPE_PV32QI_V32QI_USI,
6083         VOID_FTYPE_PV16QI_V16QI_UHI, VOID_FTYPE_PV16HI_V16HI_UHI,
6084         VOID_FTYPE_PV8HI_V8HI_UQI): New types.
6085         * config/i386/i386-builtin.def (__builtin_ia32_compressqi512_mask,
6086         __builtin_ia32_compresshi512_mask, __builtin_ia32_compressqi256_mask,
6087         __builtin_ia32_compressqi128_mask, __builtin_ia32_compresshi256_mask,
6088         __builtin_ia32_compresshi128_mask,
6089         __builtin_ia32_compressstoreuqi512_mask,
6090         __builtin_ia32_compressstoreuhi512_mask,
6091         __builtin_ia32_compressstoreuqi256_mask,
6092         __builtin_ia32_compressstoreuqi128_mask,
6093         __builtin_ia32_compressstoreuhi256_mask,
6094         __builtin_ia32_compressstoreuhi128_mask): New builtins.
6095         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Create special args
6096         array for flags2.
6097         (ix86_expand_special_args_builtin): Handle new types.
6098         (s4fma_expand): Handle new builtin array.
6099         * config/i386/immintrin.h: Include new headers.
6100         * config/i386/sse.md (VI12_AVX512VLBW): New iterator.
6101         (compress<mode>_mask, compressstore<mode>_mask): New patterns.
6103 2017-11-23  Jakub Jelinek  <jakub@redhat.com>
6105         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Formatting
6106         fixes.  Declare temp and g variables at the top in order to avoid
6107         {} in most of the cases.
6109 2017-11-23  Marc Glisse  <marc.glisse@inria.fr>
6111         * match.pd (ptr-0): New transformation.
6113 2017-11-23  Charles Baylis  <charles.baylis@linaro.org>
6115         * config/arm/arm-protos.h (enum arm_addr_mode_op): New.
6116         (struct addr_mode_cost_table): New.
6117         (struct tune_params): Add field addr_mode_costs.
6118         * config/arm/arm.c (generic_addr_mode_costs): New.
6119         (arm_slowmul_tune): Initialise addr_mode_costs field.
6120         (arm_fastmul_tune): Likewise.
6121         (arm_strongarm_tune): Likewise.
6122         (arm_xscale_tune): Likewise.
6123         (arm_9e_tune): Likewise.
6124         (arm_marvell_pj4_tune): Likewise.
6125         (arm_v6t2_tune): Likewise.
6126         (arm_cortex_tune): Likewise.
6127         (arm_cortex_a8_tune): Likewise.
6128         (arm_cortex_a7_tune): Likewise.
6129         (arm_cortex_a15_tune): Likewise.
6130         (arm_cortex_a35_tune): Likewise.
6131         (arm_cortex_a53_tune): Likewise.
6132         (arm_cortex_a57_tune): Likewise.
6133         (arm_exynosm1_tune): Likewise.
6134         (arm_xgene1_tune): Likewise.
6135         (arm_cortex_a5_tune): Likewise.
6136         (arm_cortex_a9_tune): Likewise.
6137         (arm_cortex_a12_tune): Likewise.
6138         (arm_cortex_a73_tune): Likewise.
6139         (arm_v7m_tune): Likewise.
6140         (arm_cortex_m7_tune): Likewise.
6141         (arm_v6m_tune): Likewise.
6142         (arm_fa726te_tune): Likewise.
6143         (arm_mem_costs): Use table lookup to calculate cost of addressing mode.
6145 2017-11-23  Charles Baylis  <charles.baylis@linaro.org>
6147         * config/arm/arm.c (arm_mem_costs): New function.
6148         (arm_rtx_costs_internal): Use arm_mem_costs.
6150 2017-11-23  Mark Wielaard  <mark@klomp.org>
6152         * dwarf2out.c (init_sections_and_labels): Use generation to create
6153         unique ranges_section_label and ranges_base_label. Return generation.
6154         (output_rnglists): Add generation argument. Use generation to create
6155         unique ranges labels.
6156         (dwarf2out_finish): Get generation from init_sections_and_labels
6157         and pass generation to output_rnglists.
6159 2017-11-23  Mike Stump  <mikestump@comcast.net>
6160             Eric Botcazou  <ebotcazou@adacore.com>
6162         * doc/generic.texi (ANNOTATE_EXPR): Document 3rd operand.
6163         * cfgloop.h (struct loop): Add unroll field.
6164         * function.h (struct function): Add has_unroll bitfield.
6165         * gimplify.c (gimple_boolify) <ANNOTATE_EXPR>: Deal with unroll kind.
6166         (gimplify_expr) <ANNOTATE_EXPR>: Propagate 3rd operand.
6167         * loop-init.c (pass_loop2::gate): Return true if cfun->has_unroll.
6168         (pass_rtl_unroll_loops::gate): Likewise.
6169         * loop-unroll.c (decide_unrolling): Tweak note message.  Skip loops
6170         for which loop->unroll==1.
6171         (decide_unroll_constant_iterations): Use note for consistency and
6172         take loop->unroll into account.  Return early if loop->unroll is set.
6173         Fix thinko in existing test.
6174         (decide_unroll_runtime_iterations): Use note for consistency and
6175         take loop->unroll into account.
6176         (decide_unroll_stupid): Likewise.
6177         * lto-streamer-in.c (input_cfg): Read loop->unroll.
6178         * lto-streamer-out.c (output_cfg): Write loop->unroll.
6179         * tree-cfg.c (replace_loop_annotate_in_block) <annot_expr_unroll_kind>:
6180         New case.
6181         (replace_loop_annotate) <annot_expr_unroll_kind>: Likewise.
6182         (print_loop): Print loop->unroll if set.
6183         * tree-core.h (enum annot_expr_kind): Add annot_expr_unroll_kind.
6184         * tree-inline.c (copy_loops): Copy unroll and set cfun->has_unroll.
6185         * tree-pretty-print.c (dump_generic_node) <annot_expr_unroll_kind>:
6186         New case.
6187         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Bail out if
6188         loop->unroll is set and smaller than the trip count.  Otherwise bypass
6189         entirely the heuristics if loop->unroll is set.  Remove dead note.
6190         Fix off-by-one bug in other note.
6191         (try_peel_loop): Bail out if loop->unroll is set.  Fix formatting.
6192         (tree_unroll_loops_completely_1): Force unrolling if loop->unroll
6193         is greater than 1.
6194         (tree_unroll_loops_completely): Make static.
6195         (pass_complete_unroll::execute): Use correct type for variable.
6196         (pass_complete_unrolli::execute): Fix formatting.
6197         * tree.def (ANNOTATE_EXPR): Add 3rd operand.
6199 2017-11-23  Sergey Shalnov  <Sergey.Shalnov@intel.com>
6201         * config/i386/i386.h (TARGET_PREFER_AVX256): Also
6202         enable when TARGET_PREFER_AVX128 is set.
6204 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
6206         * ipa-profile.c (ipa_propagate_frequency_1): Use count instead of
6207         frequency.
6208         * cgraph.c (cgraph_edge::dump_edge_flags): Dump sreal frequencies.
6209         (cgraph_edge::maybe_hot_p): Use sreal frequencies.
6211 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
6213         * ipa-fnsummary.c (record_modified_bb_info): Use sreal
6214         frequencies. Fix estimation of aggregate parameters.
6216 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
6218         * cgraphclones.c (cgraph_node::create_clone): Fix updating of profile
6219         when inlining.
6221 2017-11-23  Tom de Vries  <tom@codesourcery.com>
6223         * config/sh/sh.h (ASM_OUTPUT_ADDR_VEC_ELT): Wrap in "do {} while (0)".
6225 2017-11-23  Tom de Vries  <tom@codesourcery.com>
6227         * config/ft32/ft32.h (ASM_OUTPUT_ADDR_VEC_ELT): Remove semicolon after
6228         macro.
6230 2017-11-23  Oleg Endo  <olegendo@gcc.gnu.org>
6232         PR target/83111
6233         * config/sh/sh.md (udivsi3, divsi3, sibcall_value_pcrel,
6234         sibcall_value_pcrel_fdpic): Use local variable instead of
6235         operands[3].
6236         (calli_tbr_rel): Add missing operand 2.
6237         (call_valuei_tbr_rel): Add missing operand 3.
6239 2017-11-23  Jakub Jelinek  <jakub@redhat.com>
6241         PR middle-end/82253
6242         * expr.c (expand_assignment): For CONCAT to_rtx, complex type from and
6243         bitpos/bitsize covering the whole destination, use store_expr only if
6244         the complex mode is the same.  Otherwise, use expand_normal and if
6245         it returns CONCAT, subreg each part separately instead of trying to
6246         subreg the whole result.
6248 2017-11-23  Richard Biener  <rguenther@suse.de>
6250         PR tree-optimization/23094
6251         * tree-ssa-sccvn.c (vuse_ssa_val): Handle VN_TOP when we
6252         come here from walking over backedges in the first iteration.
6253         (vn_reference_lookup_3): Skip clobbers that store the same value.
6255 2017-11-23  Richard Biener  <rguenther@suse.de>
6257         PR tree-optimization/81403
6258         * tree-ssa-pre.c (get_representative_for): Add parameter specifying
6259         a block we need a leader relative to.
6260         (phi_translate_1): For nary processing require a leader from
6261         get_representative_for given we run expression simplification
6262         using match-and-simplify.  Remove previous fix.
6264 2017-11-22  Jeff Law  <law@redhat.com>
6266         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::try_find_new_range):
6267         Use new method allocate_value_range rather than accessing the
6268         vrp_value_range_pool data member directly.
6269         * tree-vrp.c (simplify_stmt_for_jump_threading): Tweak slightly
6270         to use extract_range_from_stmt method to avoid need for
6271         extract_range_from_assignment method.
6272         (vrp_prop::vrp_finalize): Use set_lattice_propagation_complete
6273         method rather than setting values_propgated data member directly.
6274         * vr-values.h (class vr_values): Privatize vrp_value_range_pool,
6275         and values propagated data members and extract_range_from_assignment
6276         method.  Reorder private data members to conform to standards.
6277         Add new methods set_lattice_propagation_complete and
6278         allocate_value_range.
6280 2017-11-22  Eric Botcazou  <ebotcazou@adacore.com>
6282         PR rtl-optimization/83030
6283         * doc/rtl.texi (Flags in an RTL Expression): Alphabetize, add entry
6284         for CROSSING_JUMP_P and mention usage of 'jump' for JUMP_INSNs.
6285         (Insns): Delete entry for REG_CROSSING_JUMP in register notes.
6286         * bb-reorder.c (update_crossing_jump_flags): Do not test whether the
6287         CROSSING_JUMP_P flag is already set before setting it.
6288         * cfgrtl.c (fixup_partition_crossing): Likewise.
6289         * reorg.c (relax_delay_slots): Do not consider a CROSSING_JUMP_P
6290         insn as useless.
6292 2017-11-22  Jakub Jelinek  <jakub@redhat.com>
6294         * simplify-rtx.c (simplify_binary_operation_1) <case VEC_SERIES>:
6295         Handle the case where both arguments are using gen_const_vec_series.
6297 2017-11-22  David Malcolm  <dmalcolm@redhat.com>
6299         PR c++/62170
6300         * pretty-print.c (pp_format): Move quoting implementation to
6301         pp_begin_quote and pp_end_quote.  Update pp_format_decoder call
6302         to pass address of "quote" local.
6303         (pp_begin_quote): New function.
6304         (pp_end_quote): New function.
6305         * pretty-print.h (printer_fn): Convert penultimate param from bool
6306         to bool *.
6307         (pp_begin_quote): New decl.
6308         (pp_end_quote): New decl.
6309         * tree-diagnostic.c (default_tree_printer): Convert penultimate
6310         param from bool to bool *.
6311         * tree-diagnostic.h (default_tree_printer): Likewise.
6313 2017-11-22  Jeff Law  <law@redhat.com>
6315         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::evrp_range_analyzer)
6316         Initialize vr_values.
6317         (evrp_range_analyzer::try_find_new_range): Call methods attached to
6318         vr_values via vr_values class instance rather than delegators.
6319         (evrp_range_analyzer::record_ranges_from_phis): Likewise.
6320         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6321         (evrp_range_analyzer::push_value_range): Likewise.
6322         (evrp_range_analyzer::pop_value_range): Likewise.
6323         * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
6324         most delegators.  Those remaining are exposed as public interfaces.
6325         Make vr_values a pointer and private.
6326         (evrp_range_analyzer::~evrp_range_analyzer): Delete the attached
6327         vr_values.
6328         (evrp_range_analyzer::get_vr_value): New method.
6329         * gimple-ssa-evrp.c (class evrp_folder): Use DISABLE_COPY_AND_ASSIGN.
6330         (evrp_folder::evrp_folder): New ctor to initialize vr_values.
6331         (class evrp_dom_walker): Attach evrp_folder class, initialize
6332         it in the ctor.  Remove temporary delegators.
6333         (evrp_dom_walker::before_dom_children): Call methods in attached
6334         evrp_range_analyzer class via class instance pointer.  Use
6335         free value_range_constant_singleton to remove need for
6336         op_with_constant_singleton_value delegator method.  Do not
6337         create a vrp_prop class instance for every call!  Narrow
6338         scope of a couple variables.
6339         (evrp_dom_walker::cleanup): Call methods in attached
6340         evrp_range_analyzer class via class instance pointer.
6341         * vr-values.h (class vr_values): Privatize many methods and
6342         data members.
6344 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6346         * tree.def (REDUC_MAX_EXPR, REDUC_MIN_EXPR, REDUC_PLUS_EXPR): Delete.
6347         * cfgexpand.c (expand_debug_expr): Remove handling for them.
6348         * expr.c (expand_expr_real_2): Likewise.
6349         * fold-const.c (const_unop): Likewise.
6350         * optabs-tree.c (optab_for_tree_code): Likewise.
6351         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
6352         * tree-inline.c (estimate_operator_cost): Likewise.
6353         * tree-pretty-print.c (dump_generic_node): Likewise.
6354         (op_code_prio): Likewise.
6355         (op_symbol_code): Likewise.
6356         * internal-fn.def (DEF_INTERNAL_SIGNED_OPTAB_FN): Define.
6357         (IFN_REDUC_PLUS, IFN_REDUC_MAX, IFN_REDUC_MIN): New internal functions.
6358         * internal-fn.c (direct_internal_fn_optab): New function.
6359         (direct_internal_fn_array, direct_internal_fn_supported_p
6360         (internal_fn_expanders): Handle DEF_INTERNAL_SIGNED_OPTAB_FN.
6361         * fold-const-call.c (fold_const_reduction): New function.
6362         (fold_const_call): Handle CFN_REDUC_PLUS, CFN_REDUC_MAX and
6363         CFN_REDUC_MIN.
6364         * tree-vect-loop.c: Include internal-fn.h.
6365         (reduction_code_for_scalar_code): Rename to...
6366         (reduction_fn_for_scalar_code): ...this and return an internal
6367         function.
6368         (vect_model_reduction_cost): Take an internal_fn rather than
6369         a tree_code.
6370         (vect_create_epilog_for_reduction): Likewise.  Build calls rather
6371         than assignments.
6372         (vectorizable_reduction): Use internal functions rather than tree
6373         codes for the reduction operation.  Update calls to the functions
6374         above.
6375         * config/aarch64/aarch64-builtins.c (aarch64_gimple_fold_builtin):
6376         Use calls to internal functions rather than REDUC tree codes.
6377         * config/aarch64/aarch64-simd.md: Update comment accordingly.
6379 2017-11-22  Olivier Hainque  <hainque@adacore.com>
6381         * config/vxworks.c (vxworks_override_options): Pick default
6382         dwarf version from macro value, VXWORKS_DWARF_VERSION_DEFAULT.
6383         * config/vxworks.h: Define VXWORKS_DWARF_VERSION_DEFAULT and
6384         DWARF_GNAT_ENCODINGS_DEFAULT.
6385         * config/vxworksae.h: Likewise.
6387 2017-11-22  Marc Glisse  <marc.glisse@inria.fr>
6389         PR tree-optimization/83104
6390         * vr-values.c (simplify_stmt_using_ranges): Check integral argument,
6391         not result.
6393 2017-11-22  Marek Polacek  <polacek@redhat.com>
6394             H.J. Lu  <hongjiu.lu@intel.com>
6395             Jason Merrill  <jason@redhat.com>
6397         PR c++/60336
6398         PR middle-end/67239
6399         PR target/68355
6400         * calls.c (initialize_argument_information): Call
6401         warn_parameter_passing_abi target hook.
6402         (store_one_arg): Use 0 for empty record size.  Don't push 0 size
6403         argument onto stack.
6404         (must_pass_in_stack_var_size_or_pad): Return false for empty types.
6405         * common.opt: Update -fabi-version description.
6406         * config/i386/i386.c (init_cumulative_args): Set cum->warn_empty.
6407         (ix86_gimplify_va_arg): Call arg_int_size_in_bytes instead of
6408         int_size_in_bytes.
6409         (ix86_is_empty_record): New function.
6410         (ix86_warn_parameter_passing_abi): New function.
6411         (TARGET_EMPTY_RECORD_P): Redefine.
6412         (TARGET_WARN_PARAMETER_PASSING_ABI): Redefine.
6413         * config/i386/i386.h (CUMULATIVE_ARGS): Add warn_empty.
6414         * doc/tm.texi: Regenerated.
6415         * doc/tm.texi.in (TARGET_EMPTY_RECORD_P,
6416         TARGET_WARN_PARAMETER_PASSING_ABI): Add.
6417         * dwarf2out.c (get_ultimate_context): Move to tree.c.
6418         * explow.c (hard_function_value): Call arg_int_size_in_bytes
6419         instead of int_size_in_bytes.
6420         * expr.c (copy_blkmode_to_reg): Likewise.
6421         * function.c (aggregate_value_p): Return 0 for empty types.
6422         (assign_parm_find_entry_rtl): Call warn_parameter_passing_abi
6423         target hook.
6424         (locate_and_pad_parm): Call arg size_in_bytes instead
6425         size_in_bytes.
6426         * lto-streamer-out.c (hash_tree): Hash TYPE_EMPTY_P and DECL_PADDING_P.
6427         * stor-layout.c (finalize_type_size): Set TYPE_EMPTY_P.
6428         * target.def (empty_record_p, warn_parameter_passing_abi): New target
6429         hooks.
6430         * targhooks.c (hook_void_CUMULATIVE_ARGS_tree): New hook.
6431         (std_gimplify_va_arg_expr): Skip empty records.  Call
6432         arg_size_in_bytes instead size_in_bytes.
6433         * targhooks.h (hook_void_CUMULATIVE_ARGS_tree): Declare.
6434         * tree-core.h (tree_type_common): Add empty_flag.
6435         (tree_decl_common): Update comments.
6436         * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Stream
6437         DECL_PADDING_P.
6438         (unpack_ts_type_common_value_fields): Stream TYPE_EMPTY_P.
6439         * tree-streamer-out.c (pack_ts_decl_common_value_fields): Stream
6440         DECL_PADDING_P.
6441         (pack_ts_type_common_value_fields): Stream TYPE_EMPTY_P.
6442         * tree.c (default_is_empty_type): New function.
6443         (default_is_empty_record): New function.
6444         (arg_int_size_in_bytes): New function.
6445         (arg_size_in_bytes): New function.
6446         (get_ultimate_context): New function.
6447         * tree.h: Define TYPE_EMPTY_P, DECL_PADDING_P and
6448         TRANSLATION_UNIT_WARN_EMPTY_P.
6449         (default_is_empty_record, arg_int_size_in_bytes,
6450         arg_size_in_bytes, get_ultimate_context): Declare.
6452 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6454         * config/arm/arm.c (cmse_clear_registers): New function.
6455         (cmse_nonsecure_call_clear_caller_saved): Replace register clearing
6456         code by call to cmse_clear_registers.
6457         (cmse_nonsecure_entry_clear_before_return): Likewise.
6459 2017-11-22  Tamar Christina  <tamar.christina@arm.com>
6461         * config/arm/arm_neon.h (vdot_u32, vdotq_u32)
6462         (vdot_s32, vdotq_s32): New.
6463         (vdot_lane_u32, vdotq_lane_u32): New.
6464         (vdot_lane_s32, vdotq_lane_s32): New.
6467 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6469         PR middle-end/82547
6470         * wide-int.cc (wi::add_large, wi::sub_large): Fix overflow detection
6471         for unsigned values with fewer HWIs than the precision.
6472         (test_overflow): New function.
6473         (wide_int_cc_tests): Call it.
6475 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6477         * emit-rtl.c (init_derived_machine_modes): Make sure ptr_mode
6478         has the same mode class as Pmode.
6480 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6482         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Get rid of
6483         padding_bits_to_clear_ptr.
6484         (cmse_nonsecure_entry_clear_before_return): Likewise.
6486 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6488         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Use
6489         auto_sbitap instead of integer bitfield to control register needing
6490         clearing.
6492 2017-11-22  Jakub Jelinek  <jakub@redhat.com>
6494         PR tree-optimization/83044
6495         * tree-vrp.c (vrp_prop::check_array_ref): If eltsize is not
6496         INTEGER_CST or is 0, clear up_bound{,_p1} and later ignore tests
6497         that need the upper bound.  Subtract offset from
6498         get_addr_base_and_unit_offset only if positive and subtract it
6499         before division by eltsize rather than after it.
6501         PR debug/83084
6502         * valtrack.c (propagate_for_debug_subst, propagate_for_debug): Reset
6503         debug insns if they would contain UNSPEC_VOLATILE or volatile asm.
6504         (dead_debug_insert_temp): Likewise, but also ignore even non-volatile
6505         asm.
6507         PR middle-end/82875
6508         * optabs.c (expand_doubleword_mult, expand_binop): Before calling
6509         expand_binop with *mul_widen_optab, make sure at least one of the
6510         operands doesn't have VOIDmode.
6512         PR debug/83034
6513         * dwarf2out.c (mem_loc_descriptor): Handle VEC_SERIES.
6515         PR rtl-optimization/82044
6516         PR tree-optimization/82042
6517         * dse.c (record_store): Check for overflow.
6518         (check_mem_read_rtx): Properly check for overflow if width == -1, call
6519         add_wild_read instead of clear_rhs_from_active_local_stores on
6520         overflow and log it into dump_file.
6522 2017-11-22  Richard Biener  <rguenther@suse.de>
6524         * gimple-iterator.c (gimple_find_edge_insert_loc): Ignore
6525         fake edges to exit when looking for a place to insert.
6526         * tree-ssa-pre.c (clear_expression_ids): Inline into callers
6527         and remove.
6528         (insert_into_preds_of_block): Commit edge insertion immediately,
6529         assert that doesn't require new BBs.
6530         (fini_pre): Release expressions.
6531         (pass_pre::execute): Shuffle things around a bit, if the fn
6532         is too large do not compute AVAIL either as this is really the
6533         quadratic bit.
6535 2017-11-22  Richard Biener  <rguenther@suse.de>
6537         PR tree-optimization/83089
6538         * tree-if-conv.c (pass_if_conversion::execute): If anything
6539         changed reset SCEV and free the number of iteration estimates.
6541 2017-11-21  Martin Sebor  <msebor@redhat.com>
6543         PR tree-optimization/82945
6544         * calls.h (warn_nonstring_bound): Remove unused function.
6546 2017-11-21  Martin Sebor  <msebor@redhat.com>
6548         PR tree-optimization/82945
6549         * builtins.c (expand_builtin_strlen): Call maybe_warn_nonstring_arg.
6550         * calls.h (maybe_warn_nonstring_arg): Declare new function.
6551         * calls.c (get_attr_nonstring_decl, maybe_warn_nonstring_arg): New
6552         functions.
6553         (initialize_argument_information): Call maybe_warn_nonstring_arg.
6554         * calls.h (get_attr_nonstring_decl): Declare new function.
6555         * doc/extend.texi (attribute nonstring): Update.
6556         * gimple-fold.c (gimple_fold_builtin_strncpy): Call
6557         get_attr_nonstring_decl and handle it.
6558         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Same.  Improve
6559         detection of nul-termination.
6560         (strlen_to_stridx): Change to a pointer.
6561         (handle_builtin_strlen, handle_builtin_stxncpy): Adjust.
6562         (pass_strlen::execute): Same.
6564 2017-11-21  Sergey Shalnov  <Sergey.Shalnov@intel.com>
6566         * config/i386/i386-opts.h (enum prefer_vector_width): Added new enum
6567         for the new option -mprefer-vector-width=[none|128|256|512].
6568         * config/i386/i386.c (ix86_target_string): remove old style options
6569         -mprefer-avx256 and make -mprefer-avx128 as alias.
6570         (ix86_option_override_internal):  Apply defaults for the
6571         -mprefer-vector-width=[128|256] option.
6572         * config/i386/i386.h (TARGET_PREFER_AVX128, TARGET_PREFER_AVX256):
6573         Implement macros to work with -mprefer-vector-width=.
6574         * config/i386/i386.opt: Implemented option
6575         -mprefer-vector-width=[none|128|256|512].
6576         * doc/invoke.texi: Documentation for
6577         -mprefer-vector-width=[none|128|256|512].
6579 2017-11-21  Pat Haugen  <pthaugen@us.ibm.com>
6581         * config/rs6000/ppc-asm.h (f50, vs50): Fix values.
6583 2017-11-21  Jonathan Wakely  <jwakely@redhat.com>
6585         * doc/invoke.texi (-Wmaybe-uninitialized): Rephrase for clarity.
6587 2017-11-21  Marc Glisse  <marc.glisse@inria.fr>
6589         * doc/generic.texi: Document POINTER_DIFF_EXPR, update
6590         POINTER_PLUS_EXPR.
6591         * cfgexpand.c (expand_debug_expr): Handle POINTER_DIFF_EXPR.
6592         * expr.c (expand_expr_real_2): Likewise.
6593         * fold-const.c (const_binop, fold_addr_of_array_ref_difference,
6594         fold_binary_loc): Likewise.
6595         * match.pd (X-X, P+(Q-P), &D-P, (P+N)-P, P-(P+N), (P+M)-(P+N),
6596         P-Q==0, -(A-B), X-Z<Y-Z, (X-Z)-(Y-Z), Z-X<Z-Y, (Z-X)-(Z-Y),
6597         (A-B)+(C-A)): New transformations for POINTER_DIFF_EXPR, based on
6598         MINUS_EXPR transformations.
6599         * optabs-tree.c (optab_for_tree_code): Handle POINTER_DIFF_EXPR.
6600         * tree-cfg.c (verify_expr, verify_gimple_assign_binary): Likewise.
6601         * tree-inline.c (estimate_operator_cost): Likewise.
6602         * tree-pretty-print.c (dump_generic_node, op_code_prio,
6603         op_symbol_code): Likewise.
6604         * tree-vect-stmts.c (vectorizable_operation): Likewise.
6605         * vr-values.c (extract_range_from_binary_expr): Likewise.
6606         * varasm.c (initializer_constant_valid_p_1): Likewise.
6607         * tree.def: New tree code POINTER_DIFF_EXPR.
6609 2017-11-21  Uros Bizjak  <ubizjak@gmail.com>
6611         * config/i386/i386.md (*bswap<mode>2_movbe): Add
6612         integer suffix to movbe mnemonic.
6613         (*bswaphi2_movbe): Ditto.
6614         (bswaphi_lowpart): Merge with *bswaphi_lowpart_1.
6616 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6618         PR c++/83045
6619         * tree-cfg.c (pass_warn_function_return::execute): Formatting fix.
6620         Also warn if seen __builtin_unreachable () call with BUILTINS_LOCATION.
6621         Use LOCATION_LOCUS when comparing against UNKNOWN_LOCATION.
6623 2017-11-21  Martin Liska  <mliska@suse.cz>
6625         * tree-inline.c (expand_call_inline): Remove not needed
6626         xstrdup_for_dump.
6628 2017-11-21  James Cowgill  <James.Cowgill@imgtec.com>
6629             Jakub Jelinek  <jakub@redhat.com>
6631         PR target/82880
6632         * config/mips/frame-header-opt.c (mips_register_frame_header_opt):
6633         Remove static keyword from f variable.
6635 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6637         PR tree-optimization/83086
6638         * gimple-ssa-store-merging.c
6639         (imm_store_chain_info::try_coalesce_bswap): Test this_n.base_addr
6640         rather than n.base_addr.
6642 2017-11-21  Martin Liska  <mliska@suse.cz>
6644         PR rtl-optimization/82044
6645         PR tree-optimization/82042
6646         * dse.c (check_mem_read_rtx): Check for overflow.
6648 2017-11-21  Aldy Hernandez  <aldyh@redhat.com>
6650         * tree-ssa-threadbackward.c (find_jump_threads_backwards): Fix
6651         typo in comment.
6653 2017-11-21  Claudiu Zissulescu  <claziss@synopsys.com>
6655         * cfgrtl.c (force_nonfallthru_and_redirect): Don't split a call
6656         and its corresponding call arg location note.
6658 2017-11-21  Claudiu Zissulescu  <claziss@synopsys.com>
6659             Andrew Burgess  <andrew.burgess@embecosm.com>
6661         * config/arc/arc-protos.h (arc_compute_frame_size): Delete
6662         declaration.
6663         (arc_return_slot_offset): Likewise.
6664         (arc_eh_return_address_location): New declaration.
6665         * config/arc/arc.c (TARGET_BUILTIN_SETJMP_FRAME_VALUE): Define.
6666         (MUST_SAVE_REGISTER): Add exception handler case.
6667         (MUST_SAVE_RETURN_ADDR): Likewise.
6668         (arc_frame_pointer_required): Likewise.
6669         (arc_frame_pointer_needed): New function.
6670         (arc_compute_frame_size): Changed.
6671         (arc_expand_prologue): Likewise.
6672         (arc_expand_epilogue): Likewise.
6673         (arc_initial_elimination_offset): Likewise.
6674         (arc_return_slot_offset): Delete.
6675         (arc_eh_return_address_location): New function.
6676         (arc_builtin_setjmp_frame_value): Likewise.
6677         * config/arc/arc.h (EH_RETURN_DATA_REGNO): Use 2 registers.
6678         (EH_RETURN_STACKADJ_RTX): Define.
6679         (EH_RETURN_HANDLER_RTX): Likewise.
6680         * config/arc/arc.md (eh_return): Delete.
6682 2017-11-21  Aldy Hernandez  <aldyh@redhat.com>
6684         * print-tree.h (debug_vec_tree): Remove prototype.
6685         * gdbinit.in (pvt): Remove macro.
6687 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6689         PR tree-optimization/83047
6690         * gimple-ssa-store-merging.c
6691         (imm_store_chain_info::output_merged_store): If the loads with the
6692         same vuse are in different basic blocks, for load_gsi pick a load
6693         location that is dominated by the other loads.
6695         PR c++/83059
6696         * config/i386/i386.c (ix86_memmodel_check): Start
6697         -Winvalid-memory-model diagnostics with lowercase letter.
6699         PR debug/82718
6700         * dwarf2out.c (dw_loc_list): If crtl->has_bb_partition, temporarily
6701         set in_cold_section_p to the partition containing loc_list->first.
6702         When seeing loc_list->last_before_switch node, update secname and
6703         perform range_across_switch second partition handling only after that.
6705         PR debug/82933
6706         * run-rtl-passes.c: Include debug.h.
6707         (run_rtl_passes): Call debug_hooks->assembly_start.
6708         * dwarf2out.c (dwarf2out_assembly_start): Return early if invoked
6709         multiple times.
6711         PR target/82981
6712         * internal-fn.c (expand_mul_overflow): Use OPTAB_WIDEN instead of
6713         OPTAB_DIRECT in calls to expand_simple_binop.
6715 2017-11-20  David Malcolm  <dmalcolm@redhat.com>
6717         PR c/81404
6718         * Makefile.in (C_COMMON_OBJS): Add c-family/known-headers.o.
6720 2017-11-20  David Malcolm  <dmalcolm@redhat.com>
6722         PR c++/72786
6723         * spellcheck.h (best_match::blithely_get_best_candidate): New
6724         accessor.
6726 2017-11-20  Jakub Jelinek  <jakub@redhat.com>
6728         * config/i386/i386.c (parse_mtune_ctrl_str): Start diagnostics
6729         with lower case letter.
6731 2017-11-20  Uros Bizjak  <ubizjak@gmail.com>
6733         * config/i386/i386.md (bswaphi2): New expander.
6734         (*bswaphi2_movbe): New insn pattern.
6735         (bswaphi -> rorhi pepehole2): New peephole pattern.
6737 2017-11-20  Jeff Law  <law@redhat.com>
6739         * Makefile.in (OBJS): Add gimple-ssa-evrp-analyze.o.
6740         * gimple-ssa-evrp-analyze.c: New file pulled from gimple-ssa-evrp.c.
6741         * gimple-ssa-evrp-analyze.h: New file pulled from gimple-ssa-evrp.c.
6742         * gimple-ssa-evrp.c: Remove bits moved into new files.  Include
6743         gimple-ssa-evrp-analyze.h.
6745         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Do not
6746         set BB_VISITED here.
6747         (evrp_range_analyzer::enter): Set BB_VISITED here instead.
6749 2017-11-20  Bin Cheng  <bin.cheng@arm.com>
6751         * tree-predcom.c: Add general comment on Store-Store chains.
6752         (split_data_refs_to_components): Postpone clearing eliminate_store_p
6753         flag in component.
6754         (get_chain_last_ref_at): Rename into...
6755         (get_chain_last_write_at): ...this.
6756         (get_chain_last_write_before_load): New function.
6757         (add_ref_to_chain): Promote type of chain from CT_STORE_LOAD to
6758         CT_STORE_STORE when write reference is added.
6759         (determine_roots_comp): Support load ref in CT_STORE_STORE chains.
6760         (is_inv_store_elimination_chain): Update get_chain_last_write_at call.
6761         (initialize_root_vars_store_elim_1): Ditto.
6762         (initialize_root_vars_store_elim_2): Ditto.  Replace rhs once default
6763         definition is created.
6764         (execute_pred_commoning_chain): Support load ref in CT_STORE_STORE
6765         chain by replacing it with dominant stored value.
6767 2017-11-20  Bin Cheng  <bin.cheng@arm.com>
6769         * tree-predcom.c (add_ref_to_chain): Remove check on distance.
6771 2017-11-20  Marc Glisse  <marc.glisse@inria.fr>
6773         * vr-values.c (extract_range_from_binary_expr): Use a full range
6774         for VR_VARYING.
6776 2017-11-20  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6778         * config/arm/arm.md (R4_REGNUM): Define constant.
6779         (nonsecure_call_internal): Remove r4 clobber.
6780         (nonsecure_call_value_internal): Likewise.
6781         * config/arm/thumb1.md (nonsecure_call_reg_thumb1_v5): Remove second
6782         clobber and resequence match_operands.
6783         (nonsecure_call_value_reg_thumb1_v5): Likewise.
6784         * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Likewise.
6785         (nonsecure_call_value_reg_thumb2): Likewise.
6787 2017-11-20  Jakub Jelinek  <jakub@redhat.com>
6789         PR tree-optimization/78821
6790         * gimple-ssa-store-merging.c (find_bswap_or_nop_load): Give up
6791         if base is TARGET_MEM_REF.  If base is not MEM_REF, set base_addr
6792         to the address of the base rather than the base itself.
6793         (find_bswap_or_nop_1): Just use pointer comparison for vuse check.
6794         (find_bswap_or_nop_finalize): New function.
6795         (find_bswap_or_nop): Use it.
6796         (bswap_replace): Return a tree rather than bool, change first
6797         argument from gimple * to gimple_stmt_iterator, allow inserting
6798         into an empty sequence, allow ins_stmt to be NULL - then emit
6799         all stmts into gsi.  Fix up MEM_REF address gimplification.
6800         (pass_optimize_bswap::execute): Adjust bswap_replace caller.
6801         (struct store_immediate_info): Add N and INS_STMT non-static
6802         data members.
6803         (store_immediate_info::store_immediate_info): Initialize them
6804         from newly added ctor args.
6805         (merged_store_group::apply_stores): Formatting fixes.  Sort by
6806         bitpos at the end.
6807         (stmts_may_clobber_ref_p): For stores call also
6808         refs_anti_dependent_p.
6809         (gather_bswap_load_refs): New function.
6810         (imm_store_chain_info::try_coalesce_bswap): New method.
6811         (imm_store_chain_info::coalesce_immediate_stores): Use it.
6812         (split_group): Handle LROTATE_EXPR and NOP_EXPR rhs_code specially.
6813         (imm_store_chain_info::output_merged_store): Fail if number of
6814         new estimated stmts is bigger or equal than old.  Handle LROTATE_EXPR
6815         and NOP_EXPR rhs_code.
6816         (pass_store_merging::process_store): Compute n and ins_stmt, if
6817         ins_stmt is non-NULL and the store rhs is otherwise invalid, use
6818         LROTATE_EXPR rhs_code.  Pass n and ins_stmt to store_immediate_info
6819         ctor.
6820         (pass_store_merging::execute): Calculate dominators.
6822         * tree-ssa-math-opts.c (nop_stats, bswap_stats, struct symbolic_number,
6823         BITS_PER_MARKER, MARKER_MASK, MARKER_BYTE_UNKNOWN, HEAD_MARKER, CMPNOP,
6824         CMPXCHG, do_shift_rotate, verify_symbolic_number_p,
6825         init_symbolic_number, find_bswap_or_nop_load, perform_symbolic_merge,
6826         find_bswap_or_nop_1, find_bswap_or_nop, pass_data_optimize_bswap,
6827         class pass_optimize_bswap, bswap_replace,
6828         pass_optimize_bswap::execute): Moved to ...
6829         * gimple-ssa-store-merging.c: ... this file.
6830         Include optabs-tree.h.
6831         (nop_stats, bswap_stats, do_shift_rotate, verify_symbolic_number_p,
6832         init_symbolic_number, find_bswap_or_nop_load, perform_symbolic_merge,
6833         find_bswap_or_nop_1, find_bswap_or_nop, bswap_replace): Put into
6834         anonymous namespace, remove static keywords.
6835         (pass_optimize_bswap::gate): Test BITS_PER_UNIT == 8 here...
6836         (pass_optimize_bswap::execute): ... rather than here.  Formatting fix.
6838 2017-11-20  Jan Hubicka  <hubicka@ucw.cz>
6840         PR bootstrap/83062
6841         * ipa-inline.c (can_inline_edge_p): Fix typo in previous patch.
6843 2017-11-20  Aldy Hernandez  <aldyh@redhat.com>
6845         * vec.h (debug_helper): New function.
6846         (DEFINE_DEBUG_VEC): New macro.
6847         * hash-set.h (debug_helper): New function.
6848         (DEFINE_DEBUG_HASH_SET): New macro.
6849         * cfg.c (debug_slim (edge)): New function.
6850         Call DEFINE_DEBUG_VEC for edges.
6851         Call DEFINE_DEBUG_HASH_SET for edges.
6852         * cfghooks.c (debug_slim (basic_block)): New function.
6853         Call DEFINE_DEBUG_VEC for basic blocks.
6854         Call DEFINE_DEBUG_HASH_SET for basic blocks.
6855         * print-tree.c (debug_slim): New function to handle trees.
6856         Call DEFINE_DEBUG_VEC for trees.
6857         Call DEFINE_DEBUG_HASH_SET for trees.
6858         (debug (vec<tree, va_gc>) &): Remove.
6859         (debug (<vec<tree, va_gc>) *): Remove.
6860         * print-rtl.c (debug_slim): New function to handle const_rtx.
6861         Call DEFINE_DEBUG_VEC for rtx_def.
6862         Call DEFINE_DEBUG_VEC for rtx_insn.
6863         Call DEFINE_DEBUG_HASH_SET for rtx_def.
6864         Call DEFINE_DEBUG_HASH_SET for rtx_insn.
6865         * sel-sched-dump.c (debug (vec<rtx_insn *> &): Remove.
6866         (debug (vec<rtx_insn *> *ptr): Remove.
6867         (debug_insn_vector): Remove.
6868         * stor-layout.c (debug_rli): Call debug() instead of debug_vec_tree.
6870 2017-11-20  Tom de Vries  <tom@codesourcery.com>
6872         PR rtl-optimization/82020
6873         * simplify-rtx.c (simplify_ternary_operation): Fix comparison mode of
6874         IF_THEN_ELSE condition.
6876 2017-11-19  Jeff Law  <law@redhat.com>
6878         * tree-ssa-dom.c (record_equivalences_from_phis): Fix handling
6879         of degenerates resulting from ignoring an edge.
6881 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6883         PR ipa/81360
6884         * ipa-inline.c (can_inline_edge_p): Also check that caller is optimized
6886 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6888         PR ipa/83001
6889         * profile-count.c (profile_count::to_sreal_scale): Fix return value
6890         for uninitialied counts.
6892 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6894         PR ipa/60243
6895         * tree-inline.c (estimate_num_insns): Set to 1 at least.
6897 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6899         PR target/82713
6900         * i386.c (ix86_builtin_vectorization_cost): Be ready for insane types.
6902 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6904         * config/arc/arc.h (FUNCTION_PROFILER): Remove semicolon after
6905         "do while (0)".
6907 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6909         * config/phoenix.h (TARGET_OS_CPP_BUILTINS): Remove semicolon after
6910         "do {} while (0)".
6912 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6914         * config/visium/visium.h (ASM_OUTPUT_CASE_END): Remove semicolon after
6915         macro body.
6917 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6919         * config/ft32/ft32.h (REGISTER_TARGET_PRAGMAS): Remove semicolon after
6920         "do {} while (0)".
6921         * config/spu/spu.h (REGISTER_TARGET_PRAGMAS): Same.
6923 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6925         * config/mcore/mcore-elf.h (MCORE_EXPORT_NAME): Remove semicolon after
6926         "do {} while (0)".
6927         * config/mcore/mcore.h (ASM_OUTPUT_ALIGNED_COMMON): After missing
6928         semicolon after MCORE_EXPORT_NAME call.
6930 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6932         PR target/82961
6933         * vmsdbgout.c (vmsdbgout_early_finish): New function.
6934         (vmsdbg_debug_hooks): Set early_finish field to vmsdbgout_early_finish.
6936 2017-11-18  Jan Hubicka  <hubicka@ucw.cz>
6938         * cgraphclones.c (cgraph_edge::clone): Rename gcov_count to prof_count.
6939         (cgraph_edge::clone): Cleanup updating of profile.
6940         * ipa-cp.c (update_profiling_info): Likewise.
6941         * ipa-inline-transform.c (inline_transform): Likewise.
6942         * ipa-inline.c (inline_small_functions): Add missing space to dump.
6943         * ipa-split.c (execute_split_functions): Do not split when function
6944         is cold.
6945         * predict.c (estimate_bb_frequencies): Cleanup updating of profile.
6946         * profile-count.c (profile_count::dump): Add global0.
6947         (profile_count::to_cgraph_frequency): Do not ICE when entry is
6948         undefined.
6949         (profile_count::to_sreal_scale): Likewise.
6950         (profile_count::adjust_for_ipa_scaling): Fix typo in comment.
6951         (profile_count::combine_with_ipa_count): New function.
6952         * profile-count.h (profile_guessed_global0adjusted): New.
6953         (profile_count::adjusted_zero): New.
6954         (profile_count::global0adjusted): New.
6955         (profile_count::combine_with_ipa_count): New.
6956         * tree-inline.c (copy_edges_for_bb): Add NUM/DEN arugment;
6957         correct profile of return block of split functions.
6958         (copy_cfg_body): Remove unused profile_count.
6959         (copy_body): Likewise.
6960         (expand_call_inline): Update.
6961         (tree_function_versioning): Update.
6963 2017-11-18  Aldy Hernandez  <aldyh@redhat.com>
6965         * hash-set.h (hash_set::empty): New.
6966         * tree-ssa-threadbackward.h: Delete.
6967         * tree-ssa-threadbackward.c (class thread_jumps): New.
6968         Move max_threaded_paths into class.
6969         (fsm_find_thread_path): Remove arguments that are now in class.
6970         (profitable_jump_thread_path): Rename to...
6971         (thread_jumps::profitable_jump_thread_path): ...this.
6972         (convert_and_register_jump_thread_path): Rename to...
6973         (thread_jumps::convert_and_register_current_path): ...this.
6974         (check_subpath_and_update_thread_path): Rename to...
6975         (thread_jumps::check_subpath_and_update_thread_path): ...this.
6976         (register_jump_thread_path_if_profitable): Rename to...
6977         (thread_jumps::register_jump_thread_path_if_profitable): ...this.
6978         (handle_phi): Rename to...
6979         (thread_jumps::handle_phi): ...this.
6980         (handle_assignment): Rename to...
6981         (thread_jumps::handle_assignment): ...this.
6982         (fsm_find_control_statement_thread_paths): Rename to...
6983         (thread_jumps::fsm_find_control_statement_thread_paths): ...this.
6984         (find_jump_threads_backwards): Rename to...
6985         (thread_jumps::find_jump_threads_backwards): ...this.
6986         Initialize path local data.
6987         (pass_thread_jumps::execute): Call find_jump_threads_backwards
6988         from within thread_jumps class.
6989         (pass_early_thread_jumps::execute): Same.
6991 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6993         * cgraph.c (cgraph_node::dump): Do IPA sanity checking on IPA counts.
6995 2017-11-17  Steve Ellcey  <sellcey@cavium.com>
6997         * config/aarch64/aarch64-simd.md (fnma<mode>4): Move neg operator
6998         to canonical location.
7000 2017-11-17  Steve Ellcey  <sellcey@cavium.com>
7002         PR target/81356
7003         * config/aarch64/aarch64.c (aarch64_use_by_pieces_infrastructure_p):
7004         Remove.
7005         (TARGET_USE_BY_PIECES_INFRASTRUCTURE_P): Remove define.
7007 2017-11-17  H.J. Lu  <hongjiu.lu@intel.com>
7009         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Use
7010         rcrt1.o%s/grcrt1.o%s for -static-pie.
7012 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
7014         * i386.c (ix86_multiplication_cost, ix86_division_cost,
7015         ix86_shift_rotate_cost): Break out from ...
7016         (ix86_rtx_costs): ... here.
7017         (ix86_add_stmt_cost): Use rtx cost machinery to compute cost of
7018         vector operations.
7020 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
7022         * predict.c (determine_unlikely_bbs): Set cgraph node count to 0
7023         when entry block was promoted unlikely.
7024         (estimate_bb_frequencies): Increase frequency scale.
7025         * profile-count.h (profile_count): Export precision info.
7027 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
7029         * tree-tailcall.c (eliminate_tail_call): Be more careful about not
7030         disturbin profile of entry block.
7032 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
7034         * ipa-fnsummary.c (estimate_node_size_and_time): Be more tolerant for
7035         roundoff errors.
7037 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
7039         * ipa-cp.c (update_profiling_info): Handle conversion to local profile.
7040         * tree-cfg.c (execute_fixup_cfg): Do fixup same way as inliner does.
7042 2017-11-17  Jeff Law  <law@redhat.com>
7044         * gimple-ssa-evrp.c (class evrp_range_analyzer): New class extracted
7045         from evrp_dom_walker class.  Various methods moved into new class.
7046         (evrp_range_analyzer::evrp_range_analyzer): Constructor for new class.
7047         (evrp_range_analyzer::enter): New method.
7048         (evrp_range_analyzer::leave): New method.
7049         (evrp_dom_walker): Remove delegators no longer needed by this class.
7050         Replace vr_values data member with evrp_range_analyzer
7052         * gimple-ssa-evrp.c (evrp_dom_walker::record_ranges_from_phis): New
7053         method extracted from evrp_dom_walker::before_dom_children.
7054         (evrp_dom_walker::record_ranges_from_stmt): Likewise.
7055         (evrp_dom_walker::record_ranges_from_incoming_edge): Likewise.
7057         * gimple-ssa-evrp.c (evrp_dom_walker): Add cleanup method.
7058         Add private copy constructor and move assignment operators.
7059         Privatize methods and class data where trivially possible.
7060         (evrp_dom_walker::cleanup): New function, extracted from
7061         execute_early_vrp.  Simplify access to class data.
7063         * vr-values.h (get_output_for_vrp): Prototype.
7064         * vr-values.c (get_output_for_vrp): New function extracted from
7065         vrp_visit_assignment_or_call and extract_range_from_stmt.
7066         (vrp_visit_assignment_or_call): Use get_output_for_vrp.  Simplify.
7068 2017-11-17  Luis Machado  <luis.machado@linaro.org>
7070         * config/aarch64/aarch64.c
7071         (qdf24xx_prefetch_tune) <default_opt_level>: Set to -1.
7072         (qdf24xx_tunings) <autoprefetcher_model>: Set to
7073         tune_params::AUTOPREFETCHER_WEAK.
7075 2017-11-17  Tamar Christina  <tamar.christina@arm.com>
7077         PR target/82641
7078         * config/arm/arm.c (arm_valid_target_attribute_rec):
7079         Parse "arch=" and "+<ext>".
7080         (arm_valid_target_attribute_tree): Re-init global options.
7081         (arm_option_override): Make non-static.
7082         (arm_options_perform_arch_sanity_checks): Make errors fatal.
7083         * gcc/config/arm/arm-c.c (__ARM_FEATURE_CMSE): Support undef.
7084         (__ARM_FEATURE_CRC32): Support undef.
7085         * config/arm/arm_acle.h (__ARM_FEATURE_CRC32): Replace with pragma.
7086         * doc/extend.texi (ARM Function Attributes): Add pragma and target.
7088 2017-11-17  David Malcolm  <dmalcolm@redhat.com>
7090         * gdbinit.in (break-on-diagnostic): New command.
7092 2017-11-17  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
7094         * config/i386/i386.c (ix86_expand_epilogue): Change simple
7095         return to indirect jump for EH return if control-flow
7096         protection is enabled. Change explicit 'false' argument in
7097         pro_epilogue_adjust_stack with a value of
7098         flag_cf_protection.
7099         * config/i386/i386.md (simple_return_indirect_internal):
7100         Remove SImode restriction to support 64-bit.
7102 2017-11-17  Segher Boessenkool  <segher@kernel.crashing.org>
7104         * combine.c (added_notes_insn): New.
7105         (try_combine): Handle added_notes_insn like added_links_insn.
7106         Rewrite return value code.
7107         (distribute_notes): Set added_notes_insn to the earliest insn we added
7108         a note to.
7110 2017-11-17  Segher Boessenkool  <segher@kernel.crashing.org>
7112         PR rtl-optimization/82621
7113         * combine.c (try_combine): Do not split PARALLELs of two SETs if the
7114         dest of one of those SETs is unused.
7116 2017-11-17  Richard Biener  <rguenther@suse.de>
7118         PR fortran/83017
7119         * tree-core.h (enum annot_expr_kind): Add annot_expr_parallel_kind.
7120         * tree-pretty-print.c (dump_generic_node): Handle
7121         annot_expr_parallel_kind.
7122         * tree-cfg.c (replace_loop_annotate_in_block): Likewise.
7123         * gimplify.c (gimple_boolify): Likewise.
7125 2017-11-17  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
7127         * config.gcc (extra_headers): Add cet.h for x86 targets.
7128         * config/i386/cet.h: New file.
7129         * doc/install.texi: Add --enable-cet/--disable-cet.
7131 2017-11-17  Richard Biener  <rguenther@suse.de>
7133         PR tree-optimization/83017
7134         * tree-parloops.c (MIN_PER_THREAD): Use --param parloops-min-per-thread.
7135         (gen_parallel_loop): Properly count iterations.
7136         (parallelize_loops): Handle loop->can_be_parallel independent
7137         of flag_loop_parallelize_all.  Make static profitability test match
7138         the runtime one.
7139         * params.def (PARAM_PARLOOPS_MIN_PER_THREAD): New.
7140         * invoke.texi (parloops-min-per-thread): Document.
7142 2017-11-17  Vineet Gupta  <vgupta@synopsys.com>
7144         * config/arc/linux.h: GLIBC_DYNAMIC_LINKER update per glibc
7145         upstreaming review comments.
7147 2017-11-17  Tamar Christina  <tamar.christina@arm.com>
7149         * expr.c (copy_blkmode_to_reg): Fix bitsize for targets
7150         with fast unaligned access.
7151         * doc/sourcebuild.texi (word_mode_no_slow_unalign): New.
7153 2017-11-17  Thomas Preud'homme  <thomas.preudhomme@arm.com>
7155         * config/arm/arm.c (cmse_nonsecure_entry_clear_before_return): Allocate
7156         to_clear_arg_regs_bitmap to the same size as to_clear_bitmap.
7158 2017-11-17  Richard Biener  <rguenther@suse.de>
7160         * tree-ssa-pre.c (phi_translate_1): Remove redundant constant
7161         folding of references.
7163 2017-11-17  Qing Zhao  <qing.zhao@oracle.com>
7165         PR middle-end/78809
7166         * gimple-fold.c (gimple_fold_builtin_string_compare): Add handling
7167         of replacing call to strncmp with corresponding call to strcmp when
7168         meeting conditions.
7170 2017-11-17  Sergey Shalnov  <Sergey.Shalnov@intel.com>
7172         * config/i386/x86-tune.def (X86_TUNE_AVX256_OPTIMAL): Add tuning
7173         option prefer-avx256 for skylake-avx512 configuration.
7174         * config/i386/i386.c (ix86_option_override_internal): Ditto.
7175         (get_builtin_code_for_version): Ditto.
7177 2017-11-17  Chung-Ju Wu  <jasonwucj@gmail.com>
7178             Monk Chiang  <sh.chiang04@gmail.com>
7180         * config/nds32/nds32.h (FIRST_PSEUDO_REGISTER): Modify.
7181         (FIXED_REGISTERS): Reserve more register numbers.
7182         (CALL_USED_REGISTERS): Likewise.
7183         (REG_ALLOC_ORDER): Likewise.
7184         (REG_CLASS_CONTENTS): Likewise.
7185         (REGISTER_NAMES): Likewise.
7187 2017-11-17  Chung-Ju Wu  <jasonwucj@gmail.com>
7188             Kito Cheng  <kito.cheng@gmail.com>
7190         * config/nds32/nds32-modes.def: Add vector mode V4QI V2HI V8QI V4HI
7191         V2SI.
7192         * config/nds32/iterators.md: Add vector mode iterators and attributes.
7194 2017-11-16  Steven Munroe  <munroesj@gcc.gnu.org>
7196         * config/rs6000/mmintrin.h (_mm_add_pi32[_ARCH_PWR]): Correct
7197         parameter list for vec_splats.
7199 2017-11-16  Joseph Myers  <joseph@codesourcery.com>
7201         * doc/invoke.texi (-std=c17): Refer to 2018 expected publication
7202         date of C17.
7203         (-std=c18, -std=iso9899:2018, -std=gnu18): Document option aliases.
7205 2017-11-16  Thomas Koenig  <tkoenig@gcc.gnu.org>
7207         PR bootstrap/82856
7208         * doc/install.texi: Document incompatibility of Perl >=5.6.26
7209         with the required version of automake 1.11.6.
7211 2017-11-16  Pat Haugen  <pthaugen@us.ibm.com>
7213         * rs6000/power9.md (power9fpdiv): New automaton and cpu_unit defined
7214         for it.
7215         (DU_C2_3_power9): Correct reservation combinations.
7216         (FP_DIV_power9, VEC_DIV_power9): New.
7217         (power9-alu): Split out rotate/shift...
7218         (power9-rot): ...to here, correct dispatch resource.
7219         (power9-cracked-alu, power9-mul, power9-mul-compare): Correct dispatch
7220         resource.
7221         (power9-fp): Correct latency.
7222         (power9-sdiv): Add div/sqrt resource.
7223         (power9-ddiv): Correct latency, add div/sqrt resource.
7224         (power9-sqrt, power9-dsqrt): Add div/sqrt resource.
7225         (power9-vecfdiv, power9-vecdiv): Correct latency, add div/sqrt
7226         resource.
7227         (power9-qpdiv, power9-qpmul): Adjust resource usage.
7229 2017-11-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
7231         * config/rs6000/rs6000.c (rs6000_expand_builtin): Do not do the
7232         switch statement mapping KF built-ins to TF built-ins if we don't
7233         have the proper ISA 3.0 assembler support.
7235 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7237         * tree-emutls.c (lower_emutls_data): Remove unused bb_freq.
7238         (lower_emutls_function_body): Do not compute it.
7240 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7242         * ipa-split.c (split_bb_info): Turn time to sreal.
7243         (split_point): Likewise.
7244         (dump_split_point): Likewise.
7245         (fine_split_points): Likewise.
7246         (execute_split_functions): Only zero split_bbs; turn time to sreals.
7248 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7250         * ipa-fnsummary.c (analyze_function_body): Accumulate time consistently
7251         in sreal.
7253 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7255         * predict.c (combine_predictions_for_bb): Preserve zero predicted
7256         edges.
7257         (expensive_function_p): Remove useless assert.
7258         (determine_unlikely_bbs): Propagate also forward; determine cold blocks
7260 2017-11-16  Martin Sebor  <msebor@redhat.com>
7262         PR tree-optimization/82588
7263         PR tree-optimization/82583
7264         * tree-vrp.c (check_array_ref): Handle flexible array members,
7265         string literals, and inner indices.
7266         (search_for_addr_array): Add detail to diagnostics.
7268 2017-11-16  Nathan Sidwell  <nathan@acm.org>
7270         PR c++/82836
7271         PR c++/82737
7272         * tree.h (COPY_DECL_RTL): Rename parms for clarity.
7273         (SET_DECL_ASSEMBLER_NAME): Forward to
7274         overwrite_decl_assembler_name.
7275         (COPY_DECL_ASSEMBLER_NAME): Rename parms for clarity.
7276         (overwrite_decl_assembler_name): Declare.
7277         * tree.c (overwrite_decl_assembler_name): New.
7278         * langhooks-def.h (lhd_overwrite_decl_assembler_name): Declare.
7279         (LANG_HOOKS_OVERWRITE_DECL_ASSEMBLER_NAME): Provide default.
7280         (LANG_HOOKS_INITIALIZER): Add it.
7281         * langhooks.h (struct lang_hooks): Add overwrite_decl_assembler_name.
7282         * langhooks.c (lhd_set_decl_assembler_name): Use
7283         SET_DECL_ASSEMBLER_NAME.
7284         (lhd_overwrite_decl_assembler_name): Default implementation.
7286 2017-11-16  Wilco Dijkstra  <wdijkstr@arm.com>
7287             Jackson Woodruff  <jackson.woodruff@arm.com>
7289         PR tree-optimization/71026
7290         * match.pd: Canonicalize constant multiplies in division.
7292 2017-11-16  Wilco Dijkstra  <wdijkstr@arm.com>
7294         * opts.c (default_options_table): Add OPT_fomit_frame_pointer entry.
7295         * common/config/alpha/alpha-common.c (TARGET_OPTION_OPTIMIZATION_TABLE):
7296         Remove OPT_fomit_frame_pointer entry.
7297         * common/config/arc/arc-common.c: Likewise.
7298         * common/config/arm/arm-common.c: Likewise.
7299         * common/config/avr/avr-common.c: Likewise.
7300         * common/config/c6x/c6x-common.c: Likewise.
7301         * common/config/cr16/cr16-common.c: Likewise.
7302         * common/config/cris/cris-common.c: Likewise.
7303         * common/config/epiphany/epiphany-common.c: Likewise.
7304         * common/config/fr30/fr30-common.c: Likewise.
7305         * common/config/frv/frv-common.c: Likewise.
7306         * common/config/ia64/ia64-common.c: Likewise.
7307         * common/config/iq2000/iq2000-common.c: Likewise.
7308         * common/config/lm32/lm32-common.c: Likewise.
7309         * common/config/m32r/m32r-common.c: Likewise.
7310         * common/config/mcore/mcore-common.c: Likewise.
7311         * common/config/microblaze/microblaze-common.c: Likewise.
7312         * common/config/mips/mips-common.c: Likewise.
7313         * common/config/mmix/mmix-common.c: Likewise.
7314         * common/config/mn10300/mn10300-common.c: Likewise.
7315         * common/config/nios2/nios2-common.c: Likewise.
7316         * common/config/pa/pa-common.c: Likewise.
7317         * common/config/pdp11/pdp11-common.c: Likewise.
7318         * common/config/powerpcspe/powerpcspe-common.c: Likewise.
7319         * common/config/riscv/riscv-common.c: Likewise.
7320         * common/config/rs6000/rs6000-common.c: Likewise.
7321         * common/config/rx/rx-common.c: Likewise.
7322         * common/config/s390/s390-common.c: Likewise.
7323         * common/config/sh/sh-common.c: Likewise.
7324         * common/config/sparc/sparc-common.c: Likewise.
7325         * common/config/tilegx/tilegx-common.c: Likewise.
7326         * common/config/tilepro/tilepro-common.c: Likewise.
7327         * common/config/v850/v850-common.c: Likewise.
7328         * common/config/visium/visium-common.c: Likewise.
7329         * common/config/xstormy16/xstormy16-common.c: Likewise.
7330         * common/config/xtensa/xtensa-common.c: Likewise.
7331         * invoke.texi (-fomit-frame-pointer): Update documentation.
7333 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7335         * tree-cfg.c (gimple_find_sub_bbs): Do not compute freq.
7337 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7339         * cfg.c (scale_bbs_frequencies_int,
7340         cale_bbs_frequencies_gcov_type): Remove.
7341         * cfg.h (scale_bbs_frequencies_int,
7342         cale_bbs_frequencies_gcov_type): Remove.
7344 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7346         * tree-ssa-loop-manip.c
7347         (scale_dominated_blocks_in_loop): Update to profile counts.
7348         (tree_transform_and_unroll_loop): Likewise.
7350 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7352         * tree-vect-loop-manip.c (vect_do_peeling): Do not use
7353         scale_bbs_frequencies_int.
7355 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7357         * final.c (compute_alignments): Use counts rather than frequencies.
7359 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7361         * cfgloopanal.c: Include sreal.h
7362         (average_num_loop_insns): Use counts and sreal for accounting.
7364 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7366         * cfgloopmanip.c (duplicate_loop_to_header_edge): Cleanup profile
7367         manipulation.
7369 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7370             Kito Cheng  <kito.cheng@gmail.com>
7372         * config/nds32/constraints.md: Provide more constraints.
7373         * config/nds32/nds32.h (enum reg_class, REG_CLASS_NAMES,
7374         REG_CLASS_CONTENTS): Define R5_REG, R8_REG, and FRAME_POINTER_REG to
7375         support constraints usage.
7377 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7379         * config/nds32/constants.md (UNSPEC_VOLATILE_FUNC_RETURN): Remove.
7381 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7382             Kito Cheng  <kito.cheng@gmail.com>
7384         * config/nds32/nds32.opt: Add mext-perf, mext-perf2, mext-string.
7385         * config/nds32/nds32.opt: Refine the layout.
7386         * config/nds32/nds32.c (TARGET_EXT_PERF, TARGET_EXT_PERF2,
7387         TARGET_EXT_STRING): Support new options.
7388         * config/nds32/nds32.h: Likewise.
7389         * config/nds32/nds32.md: Likewise.
7390         * config/nds32/nds32-predicates.c: Likewise.
7391         * config/nds32/constraints.md: Likewise.
7392         * common/config/nds32/nds32-common.c: Likewise.
7394 2017-11-16  Julia Koval  <julia.koval@intel.com>
7396         PR target/82983
7397         * config/i386/gfniintrin.h: Add sse check.
7398         * config/i386/i386.c (ix86_expand_builtin): Fix gfni check.
7400 2017-11-16  Julia Koval  <julia.koval@intel.com>
7402         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VBMI2_SET,
7403         OPTION_MASK_ISA_AVX512VBMI2_UNSET): New.
7404         (ix86_handle_option): Handle -mavx512vbmi2.
7405         * config/i386/cpuid.h: Add bit_AVX512VBMI2.
7406         * config/i386/driver-i386.c (host_detect_local_cpu): Handle new bit.
7407         * config/i386/i386-c.c (__AVX512VBMI2__): New.
7408         * config/i386/i386.c (ix86_target_string): Handle -mavx512vbmi2.
7409         (ix86_valid_target_attribute_inner_p): Ditto.
7410         * config/i386/i386.h (TARGET_AVX512VBMI2, TARGET_AVX512VBMI2_P): New.
7411         * config/i386/i386.opt (mavx512vbmi2): New option.
7412         * doc/invoke.texi: Add new option.
7414 2017-11-16  Julia Koval  <julia.koval@intel.com>
7416         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8, _mm256_gf2p8mul_epi8,
7417         _mm_mask_gf2p8mul_epi8, _mm_maskz_gf2p8mul_epi8,
7418         _mm256_mask_gf2p8mul_epi8, _mm256_maskz_gf2p8mul_epi8,
7419         _mm512_mask_gf2p8mul_epi8, _mm512_maskz_gf2p8mul_epi8,
7420         _mm512_gf2p8mul_epi8): New intrinsics.
7421         * config/i386/i386-builtin-types.def
7422         (V64QI_FTYPE_V64QI_V64QI): New type.
7423         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8mulb_v64qi,
7424         __builtin_ia32_vgf2p8mulb_v64qi_mask, __builtin_ia32_vgf2p8mulb_v32qi,
7425         __builtin_ia32_vgf2p8mulb_v32qi_mask, __builtin_ia32_vgf2p8mulb_v16qi,
7426         __builtin_ia32_vgf2p8mulb_v16qi_mask): New builtins.
7427         * config/i386/sse.md (vgf2p8mulb_*): New pattern.
7428         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
7430 2017-11-15  Uros Bizjak  <ubizjak@gmail.com>
7432         * config/i386/i386.c (x86_print_call_or_nop): Emit 5 byte nop
7433         explicitly as a stream of bytes.
7435 2017-11-15  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
7437         * config/rs6000/altivec.h (vec_xst_be): New #define.
7438         * config/rs6000/altivec.md (altivec_vperm_<mode>_direct): Rename
7439         and externalize from *altivec_vperm_<mode>_internal.
7440         * config/rs6000/rs6000-builtin.def (XL_BE_V16QI): Remove macro
7441         instantiation.
7442         (XL_BE_V8HI): Likewise.
7443         (XL_BE_V4SI): Likewise.
7444         (XL_BE_V4SI): Likewise.
7445         (XL_BE_V2DI): Likewise.
7446         (XL_BE_V4SF): Likewise.
7447         (XL_BE_V2DF): Likewise.
7448         (XST_BE): Add BU_VSX_OVERLOAD_X macro instantiation.
7449         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Correct
7450         all array entries with these keys: VSX_BUILTIN_VEC_XL,
7451         VSX_BUILTIN_VEC_XL_BE, VSX_BUILTIN_VEC_XST.  Add entries for key
7452         VSX_BUILTIN_VEC_XST_BE.
7453         * config/rs6000/rs6000.c (altivec_expand_xl_be_builtin): Remove.
7454         (altivec_expand_builtin): Remove handling for VSX_BUILTIN_XL_BE_*
7455         built-ins.
7456         (altivec_init_builtins): Replace conditional calls to def_builtin
7457         for __builtin_vsx_ld_elemrev_{v8hi,v16qi} and
7458         __builtin_vsx_st_elemrev_{v8hi,v16qi} based on TARGET_P9_VECTOR
7459         with unconditional calls.  Remove calls to def_builtin for
7460         __builtin_vsx_le_be_<mode>.  Add a call to def_builtin for
7461         __builtin_vec_xst_be.
7462         * config/rs6000/vsx.md (vsx_ld_elemrev_v8hi): Convert define_insn
7463         to define_expand, and add alternate RTL generation for P8.
7464         (*vsx_ld_elemrev_v8hi_internal): New define_insn based on
7465         vsx_ld_elemrev_v8hi.
7466         (vsx_ld_elemrev_v16qi): Convert define_insn to define_expand, and
7467         add alternate RTL generation for P8.
7468         (*vsx_ld_elemrev_v16qi_internal): New define_insn based on
7469         vsx_ld_elemrev_v16qi.
7470         (vsx_st_elemrev_v8hi): Convert define_insn
7471         to define_expand, and add alternate RTL generation for P8.
7472         (*vsx_st_elemrev_v8hi_internal): New define_insn based on
7473         vsx_st_elemrev_v8hi.
7474         (vsx_st_elemrev_v16qi): Convert define_insn to define_expand, and
7475         add alternate RTL generation for P8.
7476         (*vsx_st_elemrev_v16qi_internal): New define_insn based on
7477         vsx_st_elemrev_v16qi.
7479 2017-11-15  H.J. Lu  <hongjiu.lu@intel.com>
7481         PR target/82990
7482         * config/i386/i386.c (pass_insert_vzeroupper::gate): Remove
7483         TARGET_AVX512ER check.
7484         (ix86_option_override_internal): Set MASK_VZEROUPPER if
7485         neither -mzeroupper nor -mno-zeroupper is used and
7486         TARGET_EMIT_VZEROUPPER is set.
7487         * config/i386/i386.h (TARGET_EMIT_VZEROUPPER): New.
7488         * config/i386/x86-tune.def: Add X86_TUNE_EMIT_VZEROUPPER.
7490 2017-11-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
7492         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
7493         folding of vector compares.
7494         (fold_build_vec_cmp): New helper function.
7495         (fold_compare_helper): New helper function.
7496         (builtin_function_type): Add compare builtins to the list of functions
7497         having unsigned arguments.  Cosmetic updates to comment indentation.
7498         * config/rs6000/vsx.md (vcmpneb, vcmpneh, vcmpnew): Update to specify
7499         the not+eq combination.
7501 2017-11-15  Bin Cheng  <bin.cheng@arm.com>
7503         PR tree-optimization/82726
7504         PR tree-optimization/70754
7505         * tree-predcom.c (order_drefs_by_pos): New function.
7506         (combine_chains): Move code setting has_max_use_after to...
7507         (try_combine_chains): ...here.  New parameter.  Sort combined chains
7508         according to position information.
7509         (tree_predictive_commoning_loop): Update call to above function.
7510         (update_pos_for_combined_chains, pcom_stmt_dominates_stmt_p): New.
7512 2017-11-15  Bin Cheng  <bin.cheng@arm.com>
7514         PR tree-optimization/82726
7515         Revert
7516         2017-01-23  Bin Cheng  <bin.cheng@arm.com>
7518         PR tree-optimization/70754
7519         * tree-predcom.c (stmt_combining_refs): New parameter INSERT_BEFORE.
7520         (reassociate_to_the_same_stmt): New parameter INSERT_BEFORE.  Insert
7521         combined stmt before it if not NULL.
7522         (combine_chains): Process refs reversely and compute dominance point
7523         for root ref.
7525         Revert
7526         2017-02-23  Bin Cheng  <bin.cheng@arm.com>
7528         PR tree-optimization/79663
7529         * tree-predcom.c (combine_chains): Process refs in reverse order
7530         only for ZERO length chains, and add explaining comment.
7532 2017-11-15  Tamar Christina  <tamar.christina@arm.com>
7534         * config/arm/arm-cpus.in (armv8_3, ARMv8_3a, armv8.3-a): New
7535         * config/arm/arm-tables.opt (armv8.3-a): Regenerated.
7536         * doc/invoke.texi (ARM Options): Add armv8.3-a.
7538 2017-11-15  Tamar Christina  <tamar.christina@arm.com>
7540         * config/arm/arm.h (TARGET_DOTPROD): Add arm_arch8_2.
7542 2017-11-15  Martin Liska  <mliska@suse.cz>
7544         * tree-cfg.c (pass_warn_function_return::execute):
7545         Compare warn_return_type for greater than zero.
7547 2017-11-15  Sebastian Peryt  <sebastian.peryt@intel.com>
7549         PR target/82941
7550         PR target/82942
7551         * config/i386/i386.c (pass_insert_vzeroupper): Modify gate condition
7552         to return true on Xeon and not on Xeon Phi.
7553         (ix86_check_avx256_register): Changed to ...
7554         (ix86_check_avx_upper_register): ... this. Add extra check for
7555         VALID_AVX512F_REG_OR_XI_MODE.
7556         (ix86_avx_u128_mode_needed): Changed
7557         ix86_check_avx256_register to ix86_check_avx_upper_register.
7558         (ix86_check_avx256_stores): Changed to ...
7559         (ix86_check_avx_upper_stores): ... this. Changed
7560         ix86_check_avx256_register to ix86_check_avx_upper_register.
7561         (ix86_avx_u128_mode_after): Changed
7562         avx_reg256_found to avx_upper_reg_found. Changed
7563         ix86_check_avx256_stores to ix86_check_avx_upper_stores.
7564         (ix86_avx_u128_mode_entry): Changed
7565         ix86_check_avx256_register to ix86_check_avx_upper_register.
7566         (ix86_avx_u128_mode_exit): Ditto.
7567         * config/i386/i386.h: (host_detect_local_cpu): New define.
7569 2017-11-15  Dominik Infuehr  <dominik.infuehr@theobroma-systems.com>
7571         * config/arm/xgene1.md (xgene1): Split into automatons
7572         xgene1_main, xgene1_decoder, xgene1_div, xgene1_simd.
7573         (xgene1_f_load): Adjust reservations and/or types.
7574         (xgene1_f_store): Likewise.
7575         (xgene1_load_pair): Likewise.
7576         (xgene1_store_pair): Likewise.
7577         (xgene1_fp_load1): Likewise.
7578         (xgene1_load1): Likewise.
7579         (xgene1_store1): Likewise.
7580         (xgene1_move): Likewise.
7581         (xgene1_alu): Likewise.
7582         (xgene1_simd): Likewise.
7583         (xgene1_bfm): Likewise.
7584         (xgene1_neon_load1): Likewise.
7585         (xgene1_neon_store1): Likewise.
7586         (xgene1_neon_logic): Likewise.
7587         (xgene1_neon_st1): Likewise.
7588         (xgene1_neon_ld1r): Likewise.
7589         (xgene1_alu_cond): Added.
7590         (xgene1_shift_reg): Likwise.
7591         (xgene1_bfx): Likewise.
7592         (xgene1_mul): Split into xgene1_mul32, xgene1_mul64.
7594 2017-11-15  Jakub Jelinek  <jakub@redhat.com>
7596         PR target/82981
7597         * internal-fn.c: Include gimple-ssa.h, tree-phinodes.h and
7598         ssa-iterators.h.
7599         (can_widen_mult_without_libcall): New function.
7600         (expand_mul_overflow): If only checking unsigned mul overflow,
7601         not result, and can do efficiently MULT_HIGHPART_EXPR, emit that.
7602         Don't use WIDEN_MULT_EXPR if it would involve a libcall, unless
7603         no other way works.  Add MULT_HIGHPART_EXPR + MULT_EXPR support.
7604         (expand_DIVMOD): Formatting fix.
7605         * expmed.h (expand_mult): Add NO_LIBCALL argument.
7606         * expmed.c (expand_mult): Likewise.  Use OPTAB_WIDEN rather
7607         than OPTAB_LIB_WIDEN if NO_LIBCALL is true, and allow it to fail.
7609         PR tree-optimization/82977
7610         * tree-ssa-strlen.c (strlen_optimize_stmt): Pass a reference to a copy
7611         constructed temporary to strlen_to_stridx.put.
7613 2017-11-15  Martin Liska  <mliska@suse.cz>
7615         * configure.ac: Remove -fkeep-inline-functions from coverage_flags.
7616         * configure: Regenerate.
7618 2017-11-15  Martin Liska  <mliska@suse.cz>
7620         PR target/82927
7621         * config/sh/sh-mem.cc: Use proper probability for
7622         REG_BR_PROB_NOTE.
7624 2017-11-14  Jeff Law  <law@redhat.com>
7626         * explow.c (anti_adjust_stack_and_probe_stack_clash): Avoid probing
7627         the red zone for stack_clash_protection_final_dynamic_probe targets
7628         when the total dynamic stack size is zero bytes.
7630         * tree-ssa-threadupdate.c (thread_through_all_blocks): Thread
7631         blocks is post order.
7633 2017-11-15  Alexandre Oliva  <aoliva@redhat.com>
7635         * dumpfile.h (TDF_COMPARE_DEBUG): New.
7636         * final.c (rest_of_clean_state): Set it for the
7637         -fcompare-debug dump.
7638         * tree-pretty-print.c (dump_generic_node): Omit OBJ_TYPE_REF
7639         class when TDF_COMPARE_DEBUG is set.
7641         * dwarf2out.c (gen_producer_string): Discard
7642         OPT_fcompare_debug.
7644 2017-11-15  Joseph Myers  <joseph@codesourcery.com>
7646         PR c/81156
7647         * doc/extend.texi (Other Builtins): Document __builtin_tgmath.
7648         * ginclude/tgmath.h (__tg_cplx, __tg_ldbl, __tg_dbl, __tg_choose)
7649         (__tg_choose_2, __tg_choose_3, __TGMATH_REAL_1_2)
7650         (__TGMATH_REAL_2_3): Remove macros.
7651         (__TGMATH_CPLX, __TGMATH_CPLX_2, __TGMATH_REAL, __TGMATH_REAL_2)
7652         (__TGMATH_REAL_3, __TGMATH_CPLX_ONLY): Define using
7653         __builtin_tgmath.
7654         (frexp, ldexp, nexttoward, scalbn, scalbln): Define using
7655         __TGMATH_REAL_2.
7656         (remquo): Define using __TGMATH_REAL_3.
7658 2017-11-14  Jeff Law  <law@redhat.com>
7660         * vr-values.c: New file with contents extracted from tree-vrp.c.
7661         * Makefile.in (OBJS): Add vr-values.o
7662         * tree-vrp.h (set_value_range_to_nonnull): Prototype.
7663         (set_value_range, set_and_canonicalize_value_range): Likewise.
7664         (vrp_bitmap_equal_p, range_is_nonnull): Likewise.
7665         (value_range_constant_singleton, symbolic_range_p): Likewise.
7666         (compare_values, compare_values_warnv, vrp_val_is_min): Likewise.
7667         (vrp_val_is_max, copy_value_range, set_value_range_to_value): Likewise.
7668         (extract_range_from_binary_expr_1, vrp_val_min, vrp_val_max): Likewise.
7669         (set_value_range_to_null, range_int_cst_p, opreand_less_p): Likewise.
7670         (find_case_label_range, find_case_label_index): Likewise.
7671         (zero_nonzero_bits_from_vr, overflow_comparison_p): Likewise.
7672         (range_int_cst_singleton_p, value_inside_range): Likewise.
7673         (get_single_symbol): Likewise.
7674         (switch_update): Move structure definition here.
7675         (to_remove_edges, to_update_switch_stmts): Provide externs.
7676         * tree-vrp.c: Move all methods for vr-values class to vr-values.c
7677         (vrp_val_max, vrp_val_min, vrp_val_is_max): Make externally visible.
7678         (vrp_val_is_min, set_value_range): Likewise.
7679         (set_and_canonicalize_value_range, copy_value_range): Likewise.
7680         (set_value_range_to_value, set_value_range_to_nonnull): Likewise.
7681         (set_value_range_to_null, vrp_bitmap_equal_p): Likewise.
7682         (range_is_nonnull, range_int_cst_p): Likewwise.
7683         (range_int_cst_singleton_p, symbolic_range_p): Likewise.
7684         (get_single_symbol, operand_less_p): Likewise
7685         (compare_values_warnv, compare_values): Likewise.
7686         (value_inside_range, value_range_constant_singleton): Likewise.
7687         (zero_nonzero_bitgs_from_vr): Likewise.
7688         (extract_range_from_binary_expr_1): Likewise.
7689         (overflow_comparison_p): Likewise.
7690         (to_remove_edges, to_update_switch_stmts): Likewise.
7691         (find_case_label-index, find_case_label_range): Likewise.
7692         (switch_update, set_value_range_to_nonnegative): Remove.
7693         (set_value_range_to_truthvalue): Likewise.
7694         (symbolic_range_based_on_p, gimple_assign_nonzero_p): Likewise.
7695         (gimple_stmt_nonzero_p, compare_ranges): Likewise.
7696         (compare_range_with_value, vrp_valueize, vrp_valueize_1): Likewise.
7697         (find_case_label_ranges, test_for_singularity): Likewise.
7698         (range_fits_type_p, simplify_conversion_using_ranges): LIkewise.
7699         (x_vr_values): Move to its remaining use site.
7701 2017-11-10  Jeff Law  <law@redhat.com>
7703         * vr-values.h (VR_INITIALIZER): Move #define here.
7704         * gimple-ssa-evrp.c: New file with contents extracted from tree-vrp.c
7705         * Makefile.in (OBJS): Add tree-evrp.o
7706         * tree-vrp.h (assert_info): Move structure definition here.
7707         (set_value_range_to_varying): Prototype.
7708         (vrp_operand_equal_p, range_includes_zero_p): Likewise.
7709         (infer_value_range, register_edge_assert_for): Likewise.
7710         (stmt_interesting_for_vrp): Likewise.
7711         * tree-vrp.c: Move all methods for evrp class into gimple-ssa-evrp.c.
7712         (set_value_range_to_varying): No longer static.
7713         (vrp_operand_equal_p, range_includes_zero_p): Likewise.
7714         (infer_value_range, register_edge_assert_for): Likewise.
7716 2017-11-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
7718         * config/rs6000/rs6000.md (bswapdi2): On 32-bit ISA 3.0, don't
7719         generate the XXBRD instruction.
7721         * config/rs6000/rs6000-c.c (is_float128_p): New helper function.
7722         (rs6000_builtin_type_compatible): Treat _Float128 and long double
7723         as being compatible if -mabi=ieeelongdouble.
7724         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_HW_1): New macros
7725         to setup float128 built-ins with hardware support.
7726         (BU_FLOAT128_HW_2): Likewise.
7727         (BU_FLOAT128_HW_3): Likewise.
7728         (BU_FLOAT128_HW_VSX_1): Likewise.
7729         (BU_FLOAT128_HW_VSX_2): Likewise.
7730         (scalar_extract_expq): Change float128 built-in functions to
7731         accommodate having both KFmode and TFmode functions.  Use the
7732         KFmode variant as the default.
7733         (scalar_extract_sigq): Likewise.
7734         (scalar_test_neg_qp): Likewise.
7735         (scalar_insert_exp_q): Likewise.
7736         (scalar_insert_exp_qp): Likewise.
7737         (scalar_test_data_class_qp): Likewise.
7738         (sqrtf128_round_to_odd): Delete processing the round to odd
7739         built-in functions as special built-in functions, and define them
7740         as float128 built-ins.  Use the KFmode variant as the default.
7741         (truncf128_round_to_odd): Likewise.
7742         (addf128_round_to_odd): Likewise.
7743         (subf128_round_to_odd): Likewise.
7744         (mulf128_round_to_odd): Likewise.
7745         (divf128_round_to_odd): Likewise.
7746         (fmaf128_round_to_odd): Likewise.
7747         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Add
7748         support for KFmode and TFmode xststdcqp calls.
7749         (rs6000_expand_builtin): If long double is IEEE 128-bit floating
7750         point, switch the built-in handlers for the get/set float128
7751         exponent, get float128 mantissa, float128 test built-ins, and the
7752         float128 round to odd built-in functions.  Eliminate creating the
7753         float128 round to odd built-in functions as special built-ins.
7754         (rs6000_init_builtins): Eliminate special creation of the float128
7755         round to odd built-in functions.
7756         * config/rs6000/vsx.md (xsxexpqp_<mode>): Change float128 built-in
7757         function insns to support both TFmode and KFmode variants.
7758         (xsxsigqp_<mode>): Likewise.
7759         (xsiexpqpf_<mode>): Likewise.
7760         (xsiexpqp_<mode>): Likewise.
7761         (xststdcqp_<mode>): Likewise.
7762         (xststdcnegqp_<mode>): Likewise.
7763         (xststdcqp_<mode>): Likewise.
7765 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7767         * tree-ssa-threadupdate.c (compute_path_counts): Remove
7768         unused path_in_freq_ptr parameter.
7769         (ssa_fix_duplicate_block_edges): Do not pass around path_in_freq
7771 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7773         * ipa-inline.c (edge_badness): Dump sreal frequency.
7774         (compute_inlined_call_time): Match natural implementaiton ...
7775         * ipa-fnsummary.c (estimate_edge_size_and_time): ... here; remove
7776         forgotten division by CGRAPH_FREQ_BASE.
7778 2017-11-14  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
7780         * config.gcc (*-*-solaris2*): Enable default_use_cxa_atexit since
7781         Solaris 11.  Update comment.
7782         * configure.ac (gcc_cv_ld_pid): Adapt comment for Solaris 12
7783         renaming.
7784         * config/sol2.h (STARTFILE_SPEC): Likewise.
7785         * configure: Regenerate.
7787 2017-11-14  Carl Love  <cel@us.ibm.com>
7789         * config/rs6000/rs6000.c (swap_endian_selector_for_mode): Remove
7790         le_ and be_ prefixes to swap* variables.  Remove
7791         if (VECTOR_ELT_ORDER_BIG) statement. Remove E_V16QImode case
7792         statements.
7794 2017-11-14  Jason Merrill  <jason@redhat.com>
7796         Support GTY((cache)) on hash_map.
7797         * hash-traits.h (ggc_remove): Add ggc_maybe_mx member function.
7798         (ggc_cache_remove): Override it instead of ggc_mx.
7799         * hash-table.h (gt_ggc_mx): Call it instead of ggc_mx.
7800         (gt_cleare_cache): Call ggc_mx instead of gt_ggc_mx.
7801         * hash-map-traits.h (simple_hashmap_traits): Add maybe_mx member.
7802         (simple_cache_map_traits): Override maybe_mx.
7803         * hash-map.h (hash_entry): Add ggc_maybe_mx and keep_cache_entry.
7804         (hash_map): Friend gt_cleare_cache.
7805         (gt_cleare_cache): New.
7806         * tree.h (tree_cache_traits): New hash_map traits class.
7807         (tree_cache_map): New typedef.
7809 2017-11-14  Richard Biener  <rguenther@suse.de>
7811         * tree-cfgcleanup.c (cleanup_control_expr_graph): Remove first_p
7812         paramter and handling.
7813         (cleanup_control_flow_bb): Likewise.
7814         (cleanup_control_flow_pre): New helper performing a DFS walk
7815         to call cleanup_control_flow_bb in PRE order.
7816         (cleanup_tree_cfg_1): Do the first phase of cleanup_control_flow_bb
7817         via cleanup_control_flow_pre.
7819 2017-11-14  James Greenhalgh  <james.greenhalgh@arm.com>
7821         * config/aarch64/aarch64-simd.md
7822         (aarch64_simd_bsl<mode>_internal): Remove DImode.
7823         (*aarch64_simd_bsl<mode>_alt): Likewise.
7824         (aarch64_simd_bsldi_internal): New.
7825         (aarch64_simd_bsldi_alt): Likewise.
7827 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7829         * tracer.c (better_p): Do not compare frequencies.
7830         * reg-stack.c (better_edge): Likewise.
7831         * shrink-wrap.c (try_shrink_wrapping): Do not convert to gcov counts
7832         and back.
7834 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7836         * auto-profile.c (afdo_annotate_cfg): Use update_max_bb_count.
7837         * cgraphunit.c (cgraph_node::expand_thunk): Use update_max_bb_count.
7838         * ipa-utils.c (ipa_merge_profiles): Use update_max_bb_count.
7839         * lto-streamer-in.c (input_function): Use update_max_bb_count.
7840         * omp-expand.c (expand_omp_taskreg): Use update_max_bb_count.
7841         * predict.c (maybe_hot_frequency_p): Inline to ...
7842         (maybe_hot_count_p): ... here; rewrite to counts.
7843         (counts_to_freqs): Rename to ...
7844         (update_max_bb_count): ... this one.
7845         (expensive_function_p): Use counts.
7846         (estimate_bb_frequencies): Update.
7847         (rebuild_frequencies): Update.
7848         * predict.h (counts_to_freqs): Rename to ...
7849         (update_max_bb_count): ... this one.
7850         * profile.c (compute_branch_probabilities): Add debug info
7851         * tree-inline.c (expand_call_inline): Update debug info.
7852         (optimize_inline_calls): Use update_max_bb_count..
7853         (tree_function_versioning): Use update_max_bb_count..
7854         * value-prof.c (gimple_value_profile_transformations):
7855         Do not use update_max_bb_count.
7857 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7859         * ipa-inline.c (compute_uninlined_call_time, compute_inlined_call_time):
7860         always use frequencies.
7862 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7864         * bb-reorder.c: Remove frequencies from comments.
7865         (better_edge_p): Use profile counts.
7866         (find_traces): Dump profile counts.
7867         (rotate_loop): Use profile counts.
7868         (find_traces_1_round): Likewise.
7869         (connect_better_edge_p): Use counts instead of probabilities for
7870         reverse walk.
7871         (copy_bb_p): Drop early check for non-0 frequency.
7872         (sanitize_hot_paths): Update comments.
7874 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7876         * ipa-split.c (struct split_point): Add count.
7877         (consider_split): Do not compute incoming frequency; compute incoming
7878         count and store it to split_point.
7879         (split_function): Set count of the call to split part correctly.
7881 2017-11-13  Carl Love  <cel@us.ibm.com>
7883         * config/rs6000/altivec.md (altivec_vsumsws_be): Add define_expand.
7885 2017-11-13  Tom Tromey  <tom@tromey.com>
7887         * doc/cpp.texi (Variadic Macros): Document __VA_OPT__.
7889 2017-11-13  Carl Love  <cel@us.ibm.com>
7891         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
7892         Add support for builtins:
7893         unsigned int vec_first_{,miss}_match_{,or_eos}index,
7894         vector {un,}signed {char,int,short},
7895         vector {un,}signed {char,int,short}) arguments.
7896         * config/rs6000/rs6000-builtin.def (VFIRSTMATCHINDEX,
7897         VFIRSTMATCHOREOSINDEX, VFIRSTMISMATCHINDEX, VFIRSTMISMATCHOREOSINDEX):
7898         Add BU_P9V_AV_2 expansions for the builtins.
7899         * config/rs6000/altivec.h (vec_first_match_index,
7900         vec_first_mismatch_index, vec_first_match_or_eos_index,
7901         vec_first_mismatch_or_eos_index): Add #defines for the builtins.
7902         * config/rs6000/rs6000-protos.h (bytes_in_mode): Add
7903         new extern declaration.
7904         * config/rs6000/rs6000.c (bytes_in_mode): Add new function.
7905         * config/rs6000/vsx.md (first_match_index_<mode>,
7906         first_match_or_eos_index_<mode>, first_mismatch_index_<mode>,
7907         first_mismatch_or_eos_index_<mode>): Add define expand.
7908         (vctzlsbb_<mode>): Add mode field to define_insn for vctzlsbb.
7909         * doc/extend.texi: Update the built-in documenation file for the new
7910         built-in functions.
7912 2017-11-13  Michael Meissner  <meissner@linux.vnet.ibm.com>
7914         * match.pd: Convert fminf<N>, fminf<N>x, fmax<N>, and fmax<N>x
7915         into the min/max operations for _Float<N> and _Float<N>X types.
7917 2017-11-13  Eric Botcazou  <ebotcazou@adacore.com>
7919         PR lto/81351
7920         * dwarf2out.c (do_eh_frame): New static variable.
7921         (dwarf2out_begin_prologue): Set it.
7922         (dwarf2out_frame_finish): Test it instead of dwarf2out_do_eh_frame.
7924 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7926         * tree-ssa-coalesce.c (coalesce_cost): Fix formating.
7928         * tree-ssa-sink.c (select_best_block): Do not use frequencies.
7930 2017-11-13  Eric Botcazou  <ebotcazou@adacore.com>
7932         PR lto/81351
7933         * debug.h (dwarf2out_do_eh_frame): Declare.
7934         * dwarf2cfi.c (dwarf2out_do_eh_frame): New predicate.
7935         (dwarf2out_do_frame): Use it.
7936         (dwarf2out_do_cfi_asm): Likewise.
7937         * dwarf2out.c (dwarf2out_frame_finish): Likewise.
7938         (dwarf2out_assembly_start): Likewise.
7939         (dwarf2out_begin_prologue): Fix comment.
7940         * toplev.c (compile_file): Always call dwarf2out_frame_finish
7941         if the target needs either debug or unwind DWARF2 info.
7942         * lto-opts.c (lto_write_options): Do not save -fexceptions,
7943         -fnon-call-exceptions, -ffp-contract, -fmath-errno, -fsigned-zeros,
7944         -ftrapping-math, -ftrapv and -fwrapv.
7946 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7948         * cgraph.c (cgraph_edge::sreal_frequency): New function.
7949         * cgraph.h (cgraph_edge::sreal_frequency): Declare.
7950         * ipa-fnsummary.c (dump_ipa_call_summary): Use sreal_frequency.
7951         (estimate_edge_size_and_time): Likewise.
7952         (ipa_merge_fn_summary_after_inlining): Likewise.
7953         * ipa-inline.c (cgraph_freq_base_rec): Remove.
7954         (compute_uninlined_call_time): Use sreal_frequency.
7955         (compute_inlined_call_time): Likewise.
7956         (ipa_inline): Do not initialize cgraph_freq_base_rec.
7957         * profile-count.c: Include sreal.h.
7958         (profile_count::to_sreal_scale): New.
7959         * profile-count.h: Forward declare sreal.
7960         (profile_count::to_sreal_scale): Declare.
7962 2017-11-13  Nathan Sidwell  <nathan@acm.org>
7964         * diagnostic.c (maybe_line_and_column): New.
7965         (diagnostic_get_location_text): Use it.
7966         (diagnostic_report_current_module): Likewise.
7967         (test_diagnostic_get_location_text): Add tests.
7969 2017-11-13  Luis Machado  <luis.machado@linaro.org>
7971         * doc/md.texi (Specifying processor pipeline description): Fix
7972         incorrect latency for the div instruction example.
7974 2017-11-13  Jakub Jelinek  <jakub@redhat.com>
7976         PR tree-optimization/78821
7977         * gimple-ssa-store-merging.c (compatible_load_p): Don't require
7978         that bit_not_p is the same.
7979         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
7980         (split_group): Count precisely bit_not_p bits in each statement.
7981         (invert_op): New function.
7982         (imm_store_chain_info::output_merged_store): Use invert_op to
7983         emit BIT_XOR_EXPR with a xor_mask instead of BIT_NOT_EXPR if some
7984         but not all orig_stores have BIT_NOT_EXPR in the corresponding spots.
7986 2017-11-13  Martin Liska  <mliska@suse.cz>
7988         * gcov.c (struct coverage_info): Remove typedef of coverage_t.
7989         (struct source_info): Likewise.
7990         (add_branch_counts): Likewise.
7991         (add_line_counts): Likewise.
7992         (function_summary): Likewise.
7993         (output_intermediate_line): Likewise.
7994         (generate_results): Likewise.
7996 2017-11-13  Martin Liska  <mliska@suse.cz>
7998         * gcov.c (struct block_info): Remove typedef for block_t.
7999         (struct line_info): Likewise.
8000         (line_info::has_block): Likewise.
8001         (EXIT_BLOCK): Likewise.
8002         (unblock): Likewise.
8003         (circuit): Likewise.
8004         (get_cycles_count): Likewise.
8005         (process_file): Likewise.
8006         (read_graph_file): Likewise.
8007         (solve_flow_graph): Likewise.
8008         (find_exception_blocks): Likewise.
8009         (add_line_counts): Likewise.
8010         (accumulate_line_info): Likewise.
8011         (output_line_details): Likewise.
8013 2017-11-13  Martin Liska  <mliska@suse.cz>
8015         * gcov.c (struct arc_info): Remove typedef for arc_t.
8016         (struct line_info): Likewise.
8017         (add_branch_counts): Likewise.
8018         (output_branch_count): Likewise.
8019         (function_info::~function_info): Likewise.
8020         (circuit): Likewise.
8021         (output_intermediate_line): Likewise.
8022         (read_graph_file): Likewise.
8023         (solve_flow_graph): Likewise.
8024         (find_exception_blocks): Likewise.
8025         (add_line_counts): Likewise.
8026         (accumulate_line_info): Likewise.
8027         (output_line_details): Likewise.
8028         (output_function_details): Likewise.
8030 2017-11-13  Martin Liska  <mliska@suse.cz>
8032         * gcov.c (struct function_info): Remove typedef for function_t.
8033         (struct source_info): Likewise.
8034         (source_info::get_functions_at_location): Likewise.
8035         (solve_flow_graph): Likewise.
8036         (find_exception_blocks): Likewise.
8037         (add_line_counts): Likewise.
8038         (output_intermediate_file): Likewise.
8039         (process_file): Likewise.
8040         (generate_results): Likewise.
8041         (release_structures): Likewise.
8042         (read_graph_file): Likewise.
8043         (read_count_file): Likewise.
8044         (accumulate_line_counts): Likewise.
8045         (output_lines): Likewise.
8047 2017-11-13  Martin Liska  <mliska@suse.cz>
8049         * gcov.c (function_info::function_info): Remove num_counts
8050         and add vector<gcov_type>.
8051         (function_info::~function_info): Use the vector.
8052         (process_file): Likewise.
8053         (read_graph_file): Likewise.
8054         (read_count_file): Likewise.
8055         (solve_flow_graph): Likewise.
8057 2017-11-13  Martin Liska  <mliska@suse.cz>
8059         * gcov.c (function_info::is_artificial): New function.
8060         (process_file): Erase all artificial early.
8061         (generate_results): Skip as all artificial are already
8062         removed.
8064 2017-11-13  Martin Liska  <mliska@suse.cz>
8066         * gcov.c (read_graph_file): Store to global vector of functions.
8067         (read_count_file): Iterate the vector.
8068         (process_file): Likewise.
8069         (generate_results): Likewise.
8070         (release_structures): Likewise.
8072 2017-11-13  Jakub Jelinek  <jakub@redhat.com>
8074         PR tree-optimization/82954
8075         * gimple-ssa-store-merging.c
8076         (imm_store_chain_info::coalesce_immediate_stores): If
8077         !infof->ops[N].base_addr, split group if info->ops[N].base_addr.
8079 2017-11-13  Richard Sandiford  <richard.sandiford@linaro.org>
8081         * config/aarch64/aarch64-simd.md (aarch64_store_lane0<mode>):
8082         Upddate call to ENDIAN_LANE_N.
8083         (aarch64_<sur>dot_lane<vsi2qi>): Use aarch64_endian_lane_rtx.
8084         (aarch64_<sur>dot_laneq<vsi2qi>): Likewise.
8085         (*aarch64_simd_vec_copy_lane<mode>): Update calls to ENDIAN_LANE_N
8086         and use aarch64_endian_lane_rtx.
8087         (*aarch64_simd_vec_copy_lane_<vswap_width_name><mode>): Likewise.
8089 2017-11-12  Tom de Vries  <tom@codesourcery.com>
8091         * config/riscv/riscv.h (ASM_OUTPUT_LABELREF): Wrap in do {} while (0).
8093 2017-11-12  Tom de Vries  <tom@codesourcery.com>
8095         * config/elfos.h (ASM_OUTPUT_ASCII): Remove semicolon after macro body.
8097 2017-11-12  Tom de Vries  <tom@codesourcery.com>
8099         * config/cr16/cr16.h (ASM_OUTPUT_LABELREF): Remove semicolon after macro
8100         body.
8101         * config/powerpcspe/xcoff.h (ASM_OUTPUT_LABELREF): Same.
8102         * config/rs6000/xcoff.h (ASM_OUTPUT_LABELREF): Same.
8103         * defaults.h (ASM_OUTPUT_LABELREF): Same.
8105 2017-11-11  Martin Sebor  <msebor@redhat.com>
8107         PR c/81117
8108         * doc/extend.texi (attribute nonstring): Remove spurious argument.
8110         PR bootstrap/82948
8111         * prefic.c (translate_name): Replace strncpy with memcpy to
8112         avoid -Wstringop-truncation.
8114 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
8116         * tree-ssa-loop-im.c (execute_sm_if_changed): Do not compute freq_sum.
8118 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
8120         * predict.c (maybe_hot_frequency_p): Do not use cfun.
8122 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
8124         * tree-ssa-tail-merge.c (replace_block_by): Fix and re-enable profile
8125         merging.
8127 2017-11-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
8129         * config/rs6000/rs6000.md (bswaphi2_reg): On ISA 3.0 systems,
8130         enable generating XXBRH if the value is in a vector register.
8131         (bswapsi2_reg): On ISA 3.0 systems, enable generating XXBRW if the
8132         value is in a vector register.
8133         (bswapdi2_reg): On ISA 3.0 systems, always use XXBRD to do
8134         register to register bswap64's instead of doing the GPR sequence
8135         used on previous machines.
8136         (bswapdi2_xxbrd): New insn.
8137         (bswapdi2_reg): Disallow on ISA 3.0.
8138         (register to register bswap64 splitter): Do not split the insn on
8139         ISA 3.0 systems that use XXBRD.
8141 2017-11-10  Martin Sebor  <msebor@redhat.com>
8143         PR c/81117
8144         * config/darwin-c.c (framework_construct_pathname): Replace strncpy
8145         with memcpy.
8146         (find_subframework_file): Same.
8148 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
8150         * auto-profile.c (afdo_indirect_call): Drop frequency.
8151         * cgraph.c (symbol_table::create_edge): Drop frequency argument.
8152         (cgraph_node::create_edge): Drop frequency argument.
8153         (cgraph_node::create_indirect_edge): Drop frequency argument.
8154         (cgraph_edge::make_speculative): Drop frequency arguments.
8155         (cgraph_edge::resolve_speculation): Do not update frequencies
8156         (cgraph_edge::dump_edge_flags): Do not dump frequency.
8157         (cgraph_node::dump): Check consistency in IPA mode.
8158         (cgraph_edge::maybe_hot_p): Use IPA counter.
8159         (cgraph_edge::verify_count_and_frequency): Rename to ...
8160         (cgraph_edge::verify_count): ... this one; drop frequency checking.
8161         (cgraph_node::verify_node): Update.
8162         * cgraph.h (struct cgraph_edge): Drop frequency.
8163         (cgraph_edge::frequency): New function.
8164         * cgraphbuild.c (pass_build_cgraph_edges::execute): Donot pass
8165         frequencies.
8166         (cgraph_edge::rebuild_edges): Likewise.
8167         * cgraphclones.c (cgraph_edge::clone): Scale only counts.
8168         (duplicate_thunk_for_node): Do not pass frequency.
8169         (cgraph_node::create_clone): Scale only counts.
8170         (cgraph_node::create_virtual_clone): Do not pass frequency.
8171         (cgraph_node::create_edge_including_clones): Do not pass frequency.
8172         (cgraph_node::create_version_clone): Do not pass frequency.
8173         * cgraphunit.c (cgraph_node::analyze): Do not pass frequency.
8174         (cgraph_node::expand_thunk): Do not pass frequency.
8175         (cgraph_node::create_wrapper): Do not pass frequency.
8176         * gimple-iterator.c (update_call_edge_frequencies): Do not pass
8177         frequency.
8178         * gimple-streamer-in.c (input_bb): Scale only IPA counts.
8179         * ipa-chkp.c (chkp_produce_thunks): Do not pass frequency.
8180         * ipa-cp.c (ipcp_lattice::print): Use frequency function.
8181         (gather_caller_stats): Use frequency function.
8182         (ipcp_cloning_candidate_p): Use frequency function.
8183         (ipcp_propagate_stage): Use frequency function.
8184         (get_info_about_necessary_edges): Use frequency function.
8185         (update_profiling_info): Update only IPA profile.
8186         (update_specialized_profile): Use frequency functoin.
8187         (perhaps_add_new_callers): Update only IPA profile.
8188         * ipa-devirt.c (ipa_devirt): Use IPA profile.
8189         * ipa-fnsummary.c (redirect_to_unreachable): Do not set frequrency.
8190         (dump_ipa_call_summary): Use frequency function.
8191         (estimate_edge_size_and_time): Use frequency function.
8192         (ipa_merge_fn_summary_after_inlining): Use frequency function.
8193         * ipa-inline-analysis.c (do_estimate_edge_time): Use IPA profile.
8194         * ipa-inline-transform.c (update_noncloned_frequencies): Rename to ..
8195         (update_noncloned_counts): ... ths one; scale counts only.
8196         (clone_inlined_nodes): Do not scale frequency.
8197         (inline_call): Do not pass frequency.
8198         * ipa-inline.c (compute_uninlined_call_time): Use IPA profile.
8199         (compute_inlined_call_time): Use IPA profile.
8200         (want_inline_small_function_p): Use IPA profile.
8201         (want_inline_self_recursive_call_p): Use IPA profile.
8202         (edge_badness): Use IPA profile.
8203         (lookup_recursive_calls): Use IPA profile.
8204         (recursive_inlining): Do not pass frequency.
8205         (resolve_noninline_speculation): Do not update frequency.
8206         (inline_small_functions): Collect max of IPA profile.
8207         (dump_overall_stats): Dump IPA porfile.
8208         (dump_inline_stats): Dump IPA porfile.
8209         (ipa_inline): Collect IPA stats.
8210         * ipa-inline.h (clone_inlined_nodes): Update prototype.
8211         * ipa-profile.c (ipa_propagate_frequency_1): Use frequency function.
8212         (ipa_propagate_frequency): Use frequency function.
8213         (ipa_profile): Cleanup.
8214         * ipa-prop.c (ipa_make_edge_direct_to_target): Do not pass frequency
8215         * ipa-utils.c (ipa_merge_profiles): Merge all profiles.
8216         * lto-cgraph.c (lto_output_edge): Do not stream frequency.
8217         (input_node): Do not stream frequency.
8218         (input_edge): Do not stream frequency.
8219         (merge_profile_summaries): Scale only IPA profiles.
8220         * omp-simd-clone.c (simd_clone_adjust): Do not pass frequency.
8221         * predict.c (drop_profile): Do not recompute frequency.
8222         * trans-mem.c (ipa_tm_insert_irr_call): Do not pass frequency.
8223         (ipa_tm_insert_gettmclone_call): Do not pass frequency.
8224         * tree-cfg.c (execute_fixup_cfg): Drop profile to global0 if needed.
8225         * tree-chkp.c (chkp_copy_bounds_for_assign): Do not pass frequency.
8226         * tree-emutls.c (gen_emutls_addr): Do not pass frequency.
8227         * tree-inline.c (copy_bb): Do not scale frequency.
8228         (expand_call_inline): Do not scale frequency.
8229         (tree_function_versioning): Do not scale frequency.
8230         * ubsan.c (ubsan_create_edge): Do not pass frequency.
8232 2017-11-10  Julia Koval  <julia.koval@intel.com>
8234         * config/i386/gfniintrin.h (_mm_gf2p8affine_epi64_epi8)
8235         (_mm256_gf2p8affine_epi64_epi8, _mm_mask_gf2p8affine_epi64_epi8)
8236         (_mm_maskz_gf2p8affine_epi64_epi8, _mm256_mask_gf2p8affine_epi64_epi8)
8237         (_mm256_maskz_gf2p8affine_epi64_epi8)
8238         (_mm512_mask_gf2p8affine_epi64_epi8, _mm512_gf2p8affine_epi64_epi8)
8239         (_mm512_maskz_gf2p8affine_epi64_epi8): New intrinsics.
8240         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8affineqb_v64qi)
8241         (__builtin_ia32_vgf2p8affineqb_v32qi)
8242         (__builtin_ia32_vgf2p8affineqb_v16qi): New builtins.
8243         * config/i386/sse.md (vgf2p8affineqb_<mode><mask_name>): New pattern.
8245 2017-11-10  Tamar Christina  <tamar.christina@arm.com>
8247         PR target/82641
8248         * config/arm/arm.c
8249         (arm_option_override): Refactor.
8250         (arm_option_reconfigure_globals): New.
8251         (arm_options_perform_arch_sanity_checks): New.
8252         * config/arm/arm-protos.h (arm_option_reconfigure_globals):
8253         New prototype.
8254         (arm_options_perform_arch_sanity_checks): Likewise
8256 2017-11-10  Pat Haugen  <pthaugen@us.ibm.com>
8258         * rs6000/power9.md (power9-qpdiv): Correct DFU pipe usage.
8259         (power9-qpmul): New.
8260         * rs6000/rs6000.md ("type" attr): Add qmul.
8261         (mul<mode>3, fma<mode>4_hw, *fms<mode>4_hw, *nfma<mode>4_hw,
8262         *nfms<mode>4_hw, mul<mode>3_odd, fma<mode>4_odd, *fms<mode>4_odd,
8263         *nfma<mode>4_odd, *nfms<mode>4_odd): Change type to qmul.
8265 2017-11-10  Martin Sebor  <msebor@redhat.com>
8267         PR c/81117
8268         * builtins.c (compute_objsize): Handle arrays that
8269         compute_builtin_object_size likes to fail for.  Make extern.
8270         * builtins.h (compute_objsize): Declare.
8271         (check_strncpy_sizes): New function.
8272         (expand_builtin_strncpy): Call check_strncpy_sizes.
8273         * gimple-fold.c (gimple_fold_builtin_strncpy): Implement
8274         -Wstringop-truncation.
8275         (gimple_fold_builtin_strncat): Same.
8276         * gimple.c (gimple_build_call_from_tree): Set call location.
8277         * tree-ssa-strlen.c (strlen_to_stridx): New global variable.
8278         (maybe_diag_bound_equal_length, is_strlen_related_p): New functions.
8279         (handle_builtin_stxncpy, handle_builtin_strncat): Same.
8280         (handle_builtin_strlen): Use strlen_to_stridx.
8281         (strlen_optimize_stmt): Handle flavors of strncat, strncpy, and
8282         stpncpy.
8283         Use strlen_to_stridx.
8284         (pass_strlen::execute): Release strlen_to_stridx.
8285         * doc/invoke.texi (-Wsizeof-pointer-memaccess): Document enhancement.
8286         (-Wstringop-truncation): Document new option.
8288 2017-11-10  Martin Liska  <mliska@suse.cz>
8290         PR gcov-profile/82702
8291         * gcov.c (main): Handle intermediate files in a different
8292         way.
8293         (get_gcov_intermediate_filename): New function.
8294         (output_gcov_file): Remove support of intermediate files.
8295         (generate_results): Allocate intermediate file.
8296         (release_structures): Clean-up properly fn_end.
8297         (output_intermediate_file): Start iterating with line 1.
8299 2017-11-10  Jakub Jelinek  <jakub@redhat.com>
8301         PR tree-optimization/82929
8302         * gimple-ssa-store-merging.c (struct store_immediate_info): Add
8303         ops_swapped_p non-static data member.
8304         (store_immediate_info::store_immediate_info): Clear it.
8305         (imm_store_chain_info::coalesce_immediate_stores): If swapping
8306         ops set ops_swapped_p.
8307         (count_multiple_uses): Handle ops_swapped_p.
8309 2017-11-10  Martin Liska  <mliska@suse.cz>
8311         * coverage.c (coverage_init): Stream information about
8312         support of has_unexecuted_blocks.
8313         * doc/gcov.texi: Document that.
8314         * gcov-dump.c (dump_gcov_file): Support it in gcov_dump tool.
8315         * gcov.c (read_graph_file): Likewise.
8316         (output_line_beginning): Fix a small issue with
8317         color output.
8319 2017-11-10  Bin Cheng  <bin.cheng@arm.com>
8321         * tree-predcom.c (determine_roots_comp): Avoid memory leak by freeing
8322         reference of trivial component.
8324 2017-11-10  Jakub Jelinek  <jakub@redhat.com>
8326         PR bootstrap/82916
8327         * gimple-ssa-store-merging.c
8328         (pass_store_merging::terminate_all_aliasing_chains): For
8329         gimple_store_p stmts also call refs_output_dependent_p.
8331         PR rtl-optimization/82913
8332         * compare-elim.c (try_merge_compare): Punt if def_insn is not
8333         single set.
8335 2017-11-09  Jeff Law  <law@redhat.com>
8337         * vr-values.h: New file with vr_values class.
8338         * tree-vrp.c: Include vr-values.h
8339         (vrp_value_range_pool, vrp_equiv_obstack, num_vr_values): Move static
8340         data objects into the vr_values class.
8341         (vr_value, values_propagated, vr_phi_edge_counts): Likewise.
8342         (get_value_range): Make it a member function within vr_values class.
8343         (set_defs_to_varying, update_value_range, add_equivalence): Likewise.
8344         (vrp_stmt_computes_nonzero_p, op_with_boolean_value_range_p): Likewise.
8345         (op_with_constant_singleton_value_range): Likewise.
8346         (extract_range_for_var_from_comparison_expr): Likewise.
8347         (extract_range_from_assert, extract_range_from_ssa_name): Likewise.
8348         (extract_range_from_binary_expr): Likewise.
8349         (extract_range_from_unary_expr): Likewise.
8350         (extract_range_from_cond_expr, extrat_range_from_comparison): Likewise.
8351         (check_for_binary_op_overflow, extract_range_basic): Likewise.
8352         (extract_range_from_assignment, adjust_range_with_scev): Likewise.
8353         (dump_all_value_ranges, get_vr_for_comparison): Likewise.
8354         (compare_name_with_value, compare_names): Likewise.
8355         (vrp_evaluate_conditional_warnv_with_ops_using_ranges): Likewise.
8356         (vrp_evaluate_conditional_warnv_with_ops): Likewise.  Remove prototype.
8357         (vrp_evaluate_conditional, vrp_visit_cond_stmt): Likewise.
8358         (vrp_visit_switch_stmt, extract_range_from_stmt): Likewise.
8359         (extract_range_from_phi_node): Likewise.
8360         (simplify_truth_ops_using_ranges): Likewise.
8361         (simplify_div_or_mod_using_ranges): Likewise.
8362         (simplify_min_or_max_using_ranges, simplify_abs_using_ranges): Likewise.
8363         (simplify_bit_ops_using_ranges, simplify_cond_using_ranges_1): Likewise.
8364         (simplify_cond_using_ranges_2, simplify_switch_using_ranges): Likewise.
8365         (simplify_float_conversion_using_ranges): Likewise.
8366         (simplify_internal_call_using_ranges): Likewise.
8367         (two_valued_val_range_p, simplify_stmt_using_ranges): Likewise.
8368         (vrp_visit_assignment_or_call): Likewise.  Smuggle class instance
8369         poitner via x_vr_values for calls into gimple folder.
8370         (vrp_initialize_lattice): Make this the vr_values ctor.
8371         (vrp_free_lattice): Make this the vr_values dtor.
8372         (set_vr_value): New function.
8373         (class vrp_prop): Add vr_values data member.  Add various member
8374         functions as well as member functions that delegate to vr_values.
8375         (check_array_ref): Make a member function within vrp_prop class.
8376         (search_for_addr_array, vrp_initialize): Likewise.
8377         (vrp_finalize): Likewise.  Revamp to avoid direct access to
8378         vr_value, values_propagated, etc.
8379         (check_array_bounds): Extract vrp_prop class instance pointer from
8380         walk info structure.  Use it to call member functions.
8381         (check_all_array_refs): Make a member function within vrp_prop class.
8382         Smuggle class instance pointer via walk info structure.
8383         (x_vr_values): New local static.
8384         (vrp_valueize): Use x_vr_values to get class instance.
8385         (vr_valueize_1): Likewise.
8386         (class vrp_folder): Add vr_values data member.  Add various member
8387         functions as well as member functions that delegate to vr_values.
8388         (fold_predicate_in): Make a mber fucntion within vrp_folder class.
8389         (simplify_stmt_for_jump_threading): Extract smuggled vr_values
8390         class instance from vr_values.  Use it to call member functions.
8391         (vrp_dom_walker): Add vr_values data member.
8392         (vrp_dom_walker::after_dom_children): Smuggle vr_values class
8393         instance via x_vr_values.
8394         (identify_jump_threads): Accept vr_values as argument.  Store
8395         it into the walker structure.
8396         (evrp_dom_walker): Add vr_values class data member.  Add various
8397         delegators.
8398         (evrp_dom_walker::try_find_new_range): Use vr_values data
8399         member to access the memory allocator.
8400         (evrp_dom_walker::before_dom_children): Store vr_values class
8401         instance into the vrp_folder class.
8402         (evrp_dom_walker::push_value_range): Rework to avoid direct
8403         access to num_vr_values and vr_value.
8404         (evrp_dom_walker::pop_value_range): Likewise.
8405         (execute_early_vrp): Remove call to vrp_initialize_lattice.
8406         Use vr_values to get to dump_all_value_ranges member function.
8407         Remove call to vrp_free_lattice.  Call vrp_initialize, vrp_finalize,
8408         and simplify_cond_using_ranges_2 via vrp_prop class instance.
8409         Pass vr_values class instance down to identify_jump_threads.
8410         Remove call to vrp_free_lattice.
8411         (debug_all_value_ranges): Remove.
8413         * tree-vrp.c (vrp_prop): Move class to earlier point in the file.
8414         (vrp_folder): Likewise.
8416         * tree-vrp.c (set_value_range): Do not reference vrp_equiv_obstack.
8417         Get it from the existing bitmap instead.
8418         (vrp_intersect_ranges_1): Likewise.
8420 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8422         * gimple-ssa-store-merging.c (struct store_immediate_info): Add
8423         bit_not_p field.
8424         (store_immediate_info::store_immediate_info): Add bitnotp argument,
8425         set bit_not_p to it.
8426         (imm_store_chain_info::coalesce_immediate_stores): Break group
8427         if bit_not_p is different.
8428         (count_multiple_uses, split_group,
8429         imm_store_chain_info::output_merged_store): Handle info->bit_not_p.
8430         (handled_load): Avoid multiple chained BIT_NOT_EXPRs.
8431         (pass_store_merging::process_store): Handle BIT_{AND,IOR,XOR}_EXPR
8432         result inverted using BIT_NOT_EXPR, compute bit_not_p, pass it
8433         to store_immediate_info ctor.
8435 2017-11-09  Jim Wilson  <jimw@sifive.com>
8437         * collect2.c (OBJECT_FORMAT_COFF): Remove EXTENDED_COFF support.
8438         (scan_prog_file): Likewise.
8440 2017-11-09  Jan Hubicka  <hubicka@ucw.cz>
8442         * bb-reorder.c (max_entry_frequency): Remove.
8443         (find_traces, rotate_loop, mark_bb_visited, connect_better_edge_p,
8444         connect_traces, push_to_next_round_p): Remove prototypes.
8445         (find_traces_1_round): Use counts only.
8446         (push_to_next_round_p): Likewise.
8447         (find_traces): Likewise.
8448         (rotate_loop): Likewise.
8449         (find_traces_1_round): Likewise.
8450         (connect_traces): Likewise.
8451         (edge_order): Likewise.
8453 2017-11-09  Thomas Preud'homme  <thomas.preudhomme@arm.com>
8455         * config/arm/arm.c (output_return_instruction): Add comments to
8456         indicate requirement for cmse_nonsecure_entry return to account
8457         for the size of clearing instruction output here.
8458         (thumb_exit): Likewise.
8459         * config/arm/thumb2.md (thumb2_cmse_entry_return): Fix length for
8460         return in hardfloat mode.
8462 2017-11-09  Segher Boessenkool  <segher@kernel.crashing.org>
8464         * config/rs6000/rs6000.c (machine_function): Add a bool,
8465         "toc_is_wrapped_separately".
8466         (rs6000_option_override_internal): Enable OPTION_MASK_SAVE_TOC_INDIRECT
8467         if it wasn't explicitly set or unset, we are optimizing for speed, and
8468         doing separate shrink-wrapping.
8469         (rs6000_get_separate_components): Enable the TOC component if
8470         saving the TOC register in the prologue.
8471         (rs6000_components_for_bb): Handle the TOC component.
8472         (rs6000_emit_prologue_components): Store the TOC register where needed.
8473         (rs6000_set_handled_components): Mark TOC as handled, if handled.
8474         (rs6000_emit_prologue): Don't save the TOC if that is already done.
8476 2017-11-09  Martin Jambor  <mjambor@suse.cz>
8478         * ipa-param-manipulation.c: New file.
8479         * ipa-param-manipulation.h: Likewise.
8480         * Makefile.in (OBJS): Add ipa-param-manipulation.o.
8481         (PLUGIN_HEADERS): Addded ipa-param-manipulation.h
8482         * ipa-param.h (ipa_parm_op): Moved to ipa-param-manipulation.h.
8483         (ipa_parm_adjustment): Likewise.
8484         (ipa_parm_adjustment_vec): Likewise.
8485         (ipa_get_vector_of_formal_parms): Moved declaration to
8486         ipa-param-manipulation.h.
8487         (ipa_get_vector_of_formal_parm_types): Likewise.
8488         (ipa_modify_formal_parameters): Likewise.
8489         (ipa_modify_call_arguments): Likewise.
8490         (ipa_combine_adjustments): Likewise.
8491         (ipa_dump_param_adjustments): Likewise.
8492         (ipa_modify_expr): Likewise.
8493         (ipa_get_adjustment_candidate): Likewise.
8494         * ipa-prop.c (ipa_get_vector_of_formal_parms): Moved to
8495         ipa-param-manipulation.c.
8496         (ipa_get_vector_of_formal_parm_types): Likewise.
8497         (ipa_modify_formal_parameters): Likewise.
8498         (ipa_modify_call_arguments): Likewise.
8499         (ipa_modify_expr): Likewise.
8500         (get_ssa_base_param): Likewise.
8501         (ipa_get_adjustment_candidate): Likewise.
8502         (index_in_adjustments_multiple_times_p): Likewise.
8503         (ipa_combine_adjustments): Likewise.
8504         (ipa_dump_param_adjustments): Likewise.
8505         * tree-sra.c: Also include ipa-param-manipulation.h
8506         * omp-simd-clone.c: Include ipa-param-manipulation.h instead of
8507         ipa-param.h.
8509 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8510             Alan Hayward  <alan.hayward@arm.com>
8511             David Sherwood  <david.sherwood@arm.com>
8513         * doc/sourcebuild.texi (vect_masked_store): Document.
8515 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8516             Alan Hayward  <alan.hayward@arm.com>
8517             David Sherwood  <david.sherwood@arm.com>
8519         * doc/sourcebuild.texi (vect_align_stack_vars): Document.
8521 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8522             Alan Hayward  <alan.hayward@arm.com>
8523             David Sherwood  <david.sherwood@arm.com>
8525         * doc/sourcebuild.texi (vect_variable_length): Document.
8527 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8528             Alan Hayward  <alan.hayward@arm.com>
8529             David Sherwood  <david.sherwood@arm.com>
8531         * doc/sourcebuild.texi (vect_unaligned_possible): Document.
8533 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8534             Alan Hayward  <alan.hayward@arm.com>
8535             David Sherwood  <david.sherwood@arm.com>
8537         * doc/sourcebuild.texi (vect_element_align_preferred): Document.
8539 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8540             Alan Hayward  <alan.hayward@arm.com>
8541             David Sherwood  <david.sherwood@arm.com>
8543         * doc/sourcebuild.texi (vect_perm_short, vect_perm_byte): Document
8544         previously undocumented selectors.
8545         (vect_perm3_byte, vect_perm3_short, vect_perm3_int): Document.
8547 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8549         * doc/rtl.texi (const_vector): Say that elements can be
8550         const_wide_ints too.
8551         * emit-rtl.h (valid_for_const_vec_duplicate_p): Declare.
8552         * emit-rtl.c (valid_for_const_vec_duplicate_p): New function.
8553         (gen_vec_duplicate): Use it instead of CONSTANT_P.
8554         * optabs.c (expand_vector_broadcast): Likewise.
8556 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8557             Alan Hayward  <alan.hayward@arm.com>
8558             David Sherwood  <david.sherwood@arm.com>
8560         * tree-ssa-loop-ivopts.c (get_address_cost): Try using a
8561         scaled index even if the unscaled address was invalid.
8562         Don't increase the complexity of using a scale in that case.
8564 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8565             Alan Hayward  <alan.hayward@arm.com>
8566             David Sherwood  <david.sherwood@arm.com>
8568         * doc/rtl.texi: Rewrite the subreg rules so that they partition
8569         the inner register into REGMODE_NATURAL_SIZE bytes rather than
8570         UNITS_PER_WORD bytes.
8571         * emit-rtl.c (validate_subreg): Divide subregs into blocks
8572         based on REGMODE_NATURAL_SIZE of the inner mode.
8573         (gen_lowpart_common): Split the SCALAR_FLOAT_MODE_P and
8574         !SCALAR_FLOAT_MODE_P cases.  Use REGMODE_NATURAL_SIZE for the latter.
8575         * expmed.c (lowpart_bit_field_p): Divide the value up into
8576         chunks of REGMODE_NATURAL_SIZE rather than UNITS_PER_WORD.
8577         * expr.c (store_constructor): Use REGMODE_NATURAL_SIZE to test
8578         whether something is likely to occupy more than one register.
8580 2017-11-09  Jan Hubicka  <hubicka@ucw.cz>
8582         PR ipa/82879
8583         * ipa-inline-transform.c (update_noncloned_frequencies): Use
8584         profile_count::adjust_for_ipa_scaling.
8585         * tree-inline.c (copy_bb, copy_cfg_body): Likewise.
8586         * profile-count.c (profile_count::adjust_for_ipa_scaling): New member
8587         function.
8588         * profile-count.h (profile_count::adjust_for_ipa_scaling): Declare.
8590 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8592         * gimple-ssa-store-merging.c (count_multiple_uses): New function.
8593         (split_group): Add total_orig and total_new arguments, estimate the
8594         number of statements related to the store group without store merging
8595         and with store merging.
8596         (imm_store_chain_info::output_merged_store): Adjust split_group
8597         callers, punt if estimated number of statements with store merging
8598         is not smaller than estimated number of statements without it.
8599         Formatting fix.
8600         (handled_load): Remove has_single_use checks.
8601         (pass_store_merging::process_store): Likewise.
8603 2017-11-09  Richard Biener  <rguenther@suse.de>
8605         PR tree-optimization/82902
8606         * tree-ssa-phiprop.c (propagate_with_phi): Test proper type.
8608 2017-11-09  Martin Liska  <mliska@suse.cz>
8610         PR target/82863
8611         * emit-rtl.c (init_emit_regs): Initialize split_branch_probability to
8612         uninitialized.
8614 2017-11-09  Martin Liska  <mliska@suse.cz>
8616         PR tree-optimization/82669
8617         * sbitmap.h (bmp_iter_set_init): Remove non needed check.
8619 2017-11-09  Martin Liska  <mliska@suse.cz>
8621         PR gcov-profile/48463
8622         * coverage.c (coverage_begin_function): Output also end locus
8623         of a function and information whether the function is
8624         artificial.
8625         * gcov-dump.c (tag_function): Parse and print the information.
8626         * gcov.c (INCLUDE_MAP): Add include.
8627         (INCLUDE_SET): Likewise.
8628         (struct line_info): Move earlier in the source file because
8629         of vector<line_info> in function_info structure.
8630         (line_info::line_info): Likewise.
8631         (line_info::has_block): Likewise.
8632         (struct source_info): Add new member index.
8633         (source_info::get_functions_at_location): New function.
8634         (function_info::group_line_p): New function.
8635         (output_intermediate_line): New function.
8636         (output_intermediate_file): Use the mentioned function.
8637         (struct function_start): New.
8638         (struct function_start_pair_hash): Likewise.
8639         (process_file): Add code that identifies group functions.
8640         Assign lines either to global or function scope.
8641         (generate_results): Skip artificial functions.
8642         (find_source): Assign index for each source file.
8643         (read_graph_file): Read new flag artificial and end_line.
8644         (add_line_counts): Assign it either to global of function scope.
8645         (accumulate_line_counts): Isolate core of the function to
8646         accumulate_line_info and call it for both function and global
8647         scope lines.
8648         (accumulate_line_info): New function.
8649         (output_line_beginning): Fix GNU coding style.
8650         (print_source_line): New function.
8651         (output_line_details): Likewise.
8652         (output_function_details): Likewise.
8653         (output_lines): Iterate both source (global) scope and function
8654         scope.
8655         (struct function_line_start_cmp): New class.
8656         * doc/gcov.texi: Reflect changes in documentation.
8658 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8660         PR debug/82837
8661         * dwarf2out.c (const_ok_for_output_1): Reject NEG in addition to NOT.
8662         (mem_loc_descriptor): Handle (const (neg (...))) as (neg (const (...)))
8663         and similarly for not instead of neg.
8665 2017-11-08  Andi Kleen  <ak@linux.intel.com>
8667         * config/i386/i386.opt: Add -mforce-indirect-call.
8668         * config/i386/predicates.md: Check for flag_force_indirect_call.
8669         * doc/invoke.texi: Document -mforce-indirect-call
8671 2017-11-08  Kito Cheng  <kito.cheng@gmail.com>
8673         * config/riscv/riscv-protos.h (riscv_slow_unaligned_access_p):
8674         New extern.
8675         (MOVE_RATIO): Use riscv_slow_unaligned_access_p.
8676         config/riscv/riscv.c (predict.h): New include.
8677         (riscv_slow_unaligned_access_p): No longer static.
8678         (riscv_block_move_straight): Add require.
8679         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): Delete.
8681 2017-11-08  Jakub Jelinek  <jakub@redhat.com>
8683         PR target/82855
8684         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>,
8685         <avx512>_eq<mode>3<mask_scalar_merge_name>_1): Use
8686         nonimmediate_operand predicate for operand 1 instead of
8687         register_operand.
8689 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8691         * config/aarch64/aarch64-simd.md (store_pair_lanes<mode>):
8692         New pattern.
8693         * config/aarch64/constraints.md (Uml): New constraint.
8694         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): New
8695         predicate.
8697 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8699         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
8700         of two vec_duplicates into a vec_concat.
8702 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8704         * simplify-rtx.c (simplify_ternary_operation, VEC_MERGE):
8705         Simplify vec_merge of vec_duplicate and vec_concat.
8706         * config/aarch64/constraints.md (Utq): New constraint.
8707         * config/aarch64/aarch64-simd.md (load_pair_lanes<mode>): New
8708         define_insn.
8710 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8712         * simplify-rtx.c (simplify_ternary_operation, VEC_MERGE):
8713         Simplify vec_merge of vec_duplicate and const_vector.
8714         * config/aarch64/predicates.md (aarch64_simd_or_scalar_imm_zero):
8715         New predicate.
8716         * config/aarch64/aarch64-simd.md (*aarch64_combinez<mode>): Use VDC
8717         mode iterator.  Update predicate on operand 1 to
8718         handle non-const_vec constants.  Delete constraints.
8719         (*aarch64_combinez_be<mode>): Likewise for operand 2.
8721 2017-11-08  Jakub Jelinek  <jakub@redhat.com>
8723         PR tree-optimization/78821
8724         * gimple-ssa-store-merging.c (struct store_operand_info): Add bit_not_p
8725         data member.
8726         (store_operand_info::store_operand_info): Initialize it to false.
8727         (pass_store_merging::terminate_all_aliasing_chains): Rewritten to use
8728         ref_maybe_used_by_stmt_p and stmt_may_clobber_ref_p on lhs of each
8729         store in the group, and if chain_info is non-NULL, to ignore altogether
8730         that chain.
8731         (compatible_load_p): Fail if bit_not_p does not match.
8732         (imm_store_chain_info::output_merged_store): Handle bit_not_p loads.
8733         (handled_load): Fill in bit_not_p.  Handle BIT_NOT_EXPR.
8734         (pass_store_merging::process_store): Adjust
8735         terminate_all_aliasing_chains calls to pass NULL in all current spots,
8736         call terminate_all_aliasing_chains newly when adding a store into
8737         a chain with non-NULL chain_info.
8739 2017-11-08  Wilco Dijkstra  <wdijkstr@arm.com>
8741         * config/aarch64/aarch64.c (aarch64_can_eliminate): Simplify logic.
8743 2017-11-08  Wilco Dijkstra  <wdijkstr@arm.com>
8745         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
8746         Remove.
8747         (aarch64_layout_frame): Initialise emit_frame_chain.
8748         (aarch64_can_eliminate): Remove omit leaf frame pointer code.
8749         (TARGET_FRAME_POINTER_REQUIRED): Remove define.
8751 2017-11-08  Martin Liska  <mliska@suse.cz>
8753         * gimplify.c (expand_FALLTHROUGH_r): Simplify usage
8754         of gimple_call_internal_p.
8756 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8758         * config/mips/mips.h (ASM_OUTPUT_LABELREF): Wrap in "do {} while (0)".
8760 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8762         * config/mips/mips.h (ASM_OUTPUT_CASE_END): Remove semicolon after
8763         "do {} while (0)".
8765 2017-11-08  Martin Liska  <mliska@suse.cz>
8767         PR sanitizer/82792
8768         * gimplify.c (expand_FALLTHROUGH_r): Skip IFN_ASAN_MARK.
8770 2017-11-07  Eric Botcazou  <ebotcazou@adacore.com>
8772         * gimple-pretty-print.c (dump_profile): Return "" instead of NULL.
8774 2017-11-07  Jakub Jelinek  <jakub@redhat.com>
8776         PR target/82855
8777         * config/i386/i386.md (SWI1248_AVX512BWDQ2_64): New mode iterator.
8778         (*cmp<mode>_ccz_1): New insn with $k alternative.
8780         PR target/82855
8781         * config/i386/i386.c (ix86_swap_binary_operands_p): Treat
8782         RTX_COMM_COMPARE as commutative as well.
8783         (ix86_binary_operator_ok): Formatting fix.
8784         * config/i386/sse.md (*mul<mode>3<mask_name><round_name>,
8785         *<code><mode>3<mask_name><round_saeonly_name>,
8786         *<code><mode>3<mask_name>, *<code>tf3, *mul<mode>3<mask_name>,
8787         *<s>mul<mode>3_highpart<mask_name>,
8788         *vec_widen_umult_even_v16si<mask_name>,
8789         *vec_widen_umult_even_v8si<mask_name>,
8790         *vec_widen_umult_even_v4si<mask_name>,
8791         *vec_widen_smult_even_v16si<mask_name>,
8792         *vec_widen_smult_even_v8si<mask_name>, *sse4_1_mulv2siv2di3<mask_name>,
8793         *avx2_pmaddwd, *sse2_pmaddwd, *<sse4_1_avx2>_mul<mode>3<mask_name>,
8794         *avx2_<code><mode>3, *avx512f_<code><mode>3<mask_name>,
8795         *sse4_1_<code><mode>3<mask_name>, *<code>v8hi3,
8796         *sse4_1_<code><mode>3<mask_name>, *<code>v16qi3, *avx2_eq<mode>3,
8797         <avx512>_eq<mode>3<mask_scalar_merge_name>_1, *sse4_1_eqv2di3,
8798         *sse2_eq<mode>3, <mask_codefor><code><mode>3<mask_name>,
8799         *<code><mode>3, *<sse2_avx2>_uavg<mode>3<mask_name>,
8800         *<ssse3_avx2>_pmulhrsw<mode>3<mask_name>, *ssse3_pmulhrswv4hi3): Use
8801         !(MEM_P (operands[1]) && MEM_P (operands[2])) condition instead of
8802         ix86_binary_operator_ok.  Formatting fixes.
8803         (*<plusminus_insn><mode>3<mask_name><round_name>,
8804         *<plusminus_insn><mode>3, *<plusminus_insn><mode>3_m): Formatting
8805         fixes.
8807 2017-11-07  Segher Boessenkool  <segher@kernel.crashing.org>
8809         * config/rs6000/rs6000.md (GPR2): New mode_iterator.
8810         ("cstore<mode>4"): Don't always expand with rs6000_emit_int_cmove for
8811         eq and ne if TARGET_ISEL.
8812         (cmp): New code_iterator.
8813         (UNS, UNSU_, UNSIK): New code_attrs.
8814         (<code><GPR:mode><GPR2:mode>2_isel): New define_insn_and_split.
8815         ("eq<mode>3"): New define_expand, rename the define_insn_and_split
8816         to...
8817         ("eq<mode>3"): ... this.
8818         ("ne<mode>3"): New define_expand, rename the define_insn_and_split
8819         to...
8820         ("ne<mode>3"): ... this.
8822 2017-11-07  Julia Koval  <julia.koval@intel.com>
8824         PR target/82812
8825         * common/config/i386/i386-common.c
8826         (OPTION_MASK_ISA_GENERAL_REGS_ONLY_UNSET): Remove MPX from flag.
8827         (ix86_handle_option): Move MPX to isa_flags2 and GFNI to isa_flags.
8828         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto.
8829         * config/i386/i386.opt: Ditto.
8830         * config/i386/i386.c (ix86_target_string): Ditto.
8831         (ix86_option_override_internal): Ditto.
8832         (ix86_init_mpx_builtins): Move MPX to args2.
8833         (ix86_expand_builtin): Special handling for OPTION_MASK_ISA_GFNI.
8834         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8affineinvqb_v64qi,
8835         __builtin_ia32_vgf2p8affineinvqb_v64qi_mask,
8836         __builtin_ia32_vgf2p8affineinvqb_v32qi,
8837         __builtin_ia32_vgf2p8affineinvqb_v32qi_mask,
8838         __builtin_ia32_vgf2p8affineinvqb_v16qi,
8839         __builtin_ia32_vgf2p8affineinvqb_v16qi_mask): Move to ARGS array.
8841 2017-11-07  Uros Bizjak  <ubizjak@gmail.com>
8843         PR target/80425
8844         * config/i386.i386.md (*zero_extendsidi2): Change (?r,*Yj), (?*Yi,r)
8845         and (*x,m) to ($r,Yj), ($Yi,r) and ($x,m).
8846         (zero-extendsidi peephole2): Remove peephole.
8848 2017-11-07  Eric Botcazou  <ebotcazou@adacore.com>
8850         PR c/53037
8851         * stor-layout.c: Include attribs.h.
8852         (handle_warn_if_not_align): Replace test on TYPE_USER_ALIGN with
8853         explicit lookup of "aligned" attribute.
8855 2017-11-07  Andrew Waterman  <andrew@sifive.com>
8857         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): New prototype.
8858         (riscv_expand_block_move): Likewise.
8859         * config/riscv/riscv.h (MOVE_RATIO): Tune cost to movmemsi
8860         implementation.
8861         (RISCV_MAX_MOVE_BYTES_PER_LOOP_ITER): New define.
8862         (RISCV_MAX_MOVE_BYTES_STRAIGHT): New define.
8863         * config/riscv/riscv.c (riscv_block_move_straight): New function.
8864         (riscv_adjust_block_mem): Likewise.
8865         (riscv_block_move_loop): Likewise.
8866         (riscv_expand_block_move): Likewise.
8867         * config/riscv/riscv.md (movmemsi): New pattern.
8869 2017-11-07  Michael Clark  <michaeljclark@mac.com>
8871         * config/riscv/linux.h (MUSL_ABI_SUFFIX): New define.
8872         (MUSL_DYNAMIC_LINKER): Likewise.
8874 2017-11-07  Richard Sandiford  <richard.sandiford@linaro.org>
8876         * config/aarch64/aarch64.md (ashl<mode>3, ashr<mode>3, lshr<mode>3)
8877         (rotr<mode>3, rotl<mode>3): Use aarch64_reg_or_imm instead of
8878         nonmmory_operand.
8880 2017-11-07  Richard Biener  <rguenther@suse.de>
8882         * match.pd: Fix build.
8884 2017-11-07  Wilco Dijkstra  <wdijkstr@arm.com>
8885             Jackson Woodruff  <jackson.woodruff@arm.com>
8887         PR tree-optimization/71026
8888         * match.pd: Canonicalize negate in division.
8890 2017-11-07  Sudakshina Das  <sudi.das@arm.com>
8892         PR middle-end/80131
8893         * match.pd: Simplify 1 << (C - x) where C = precision (x) - 1.
8895 2017-11-07  Marc Glisse  <marc.glisse@inria.fr>
8897         * match.pd ((a&~b)|(a^b),(a&~b)^~a,(a|b)&~(a^b),a|~(a^b),
8898         (a|b)|(a&^b),(a&b)|~(a^b),~(~a&b),~X^Y): New transformations.
8900 2017-11-07  Marc Glisse  <marc.glisse@inria.fr>
8902         * fold-const.c (negate_expr_p) [PLUS_EXPR, MINUS_EXPR]: Handle
8903         non-scalar integral types.
8904         * match.pd (negate_expr_p): Handle MINUS_EXPR.
8905         (-(A-B), -(~A)): New transformations.
8907 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8909         * config/powerpcspe/aix43.h (SUBTARGET_OVERRIDE_OPTIONS): Remove
8910         semicolon after "do {} while (0)".
8911         * config/powerpcspe/aix51.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8912         * config/powerpcspe/aix52.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8913         * config/powerpcspe/aix53.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8914         * config/powerpcspe/aix61.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8915         * config/powerpcspe/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8917 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8919         * config/rs6000/aix43.h (SUBTARGET_OVERRIDE_OPTIONS): Remove semicolon
8920         after "do {} while (0)".
8921         * config/rs6000/aix51.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8922         * config/rs6000/aix52.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8923         * config/rs6000/aix53.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8924         * config/rs6000/aix61.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8925         * config/rs6000/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8927 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8929         PR other/82784
8930         * config/arm/arm.c (HANDLE_NARROW_SHIFT_ARITH): Remove semicolon after
8931         "while {} do (0)".
8932         (arm_rtx_costs_internal): Add missing semicolon after
8933         HANDLE_NARROW_SHIFT_ARITH call.
8935 2017-11-06  Segher Boessenkool  <segher@kernel.crashing.org>
8937         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
8938         disable isel if it was not set explicitly.
8940 2017-11-06  James Bowman  <james.bowman@ftdichip.com>
8942         * gcc/dwarf2out.c (modified_type_die): Retain ADDR_SPACE qualifiers.
8943         (add_type_attribute) likewise.
8945 2017-11-06  H.J. Lu  <hongjiu.lu@intel.com>
8947         * config/i386/i386.c (ix86_can_use_return_insn_p): Use reference
8948         of struct ix86_frame.
8949         (ix86_initial_elimination_offset): Likewise.
8950         (ix86_expand_split_stack_prologue): Likewise.
8952 2017-11-06  Marc Glisse  <marc.glisse@inria.fr>
8954         * tree-vrp.h (enum value_range_type): Update stale comment.
8956 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8957             Alan Hayward  <alan.hayward@arm.com>
8958             David Sherwood  <david.sherwood@arm.com>
8960         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm)
8961         (aarch64_expand_vec_perm_const): Take the number of units too.
8962         * config/aarch64/aarch64.c (aarch64_expand_vec_perm)
8963         (aarch64_expand_vec_perm_const): Likewise.
8964         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>)
8965         (vec_perm<mode>): Update accordingly.
8967 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8968             Alan Hayward  <alan.hayward@arm.com>
8969             David Sherwood  <david.sherwood@arm.com>
8971         * config/aarch64/aarch64-protos.h (aarch64_simd_vect_par_cnst_half):
8972         Take the number of units too.
8973         * config/aarch64/aarch64.c (aarch64_simd_vect_par_cnst_half): Likewise.
8974         (aarch64_simd_check_vect_par_cnst_half): Update call accordingly,
8975         but check for a vector mode before rather than after the call.
8976         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
8977         (move_hi_quad_<mode>, vec_unpack<su>_hi_<mode>)
8978         (vec_unpack<su>_lo_<mode, vec_widen_<su>mult_lo_<mode>)
8979         (vec_widen_<su>mult_hi_<mode>, vec_unpacks_lo_<mode>)
8980         (vec_unpacks_hi_<mode>, aarch64_saddl2<mode>, aarch64_uaddl2<mode>)
8981         (aarch64_ssubl2<mode>, aarch64_usubl2<mode>, widen_ssum<mode>3)
8982         (widen_usum<mode>3, aarch64_saddw2<mode>, aarch64_uaddw2<mode>)
8983         (aarch64_ssubw2<mode>, aarch64_usubw2<mode>, aarch64_sqdmlal2<mode>)
8984         (aarch64_sqdmlsl2<mode>, aarch64_sqdmlal2_lane<mode>)
8985         (aarch64_sqdmlal2_laneq<mode>, aarch64_sqdmlsl2_lane<mode>)
8986         (aarch64_sqdmlsl2_laneq<mode>, aarch64_sqdmlal2_n<mode>)
8987         (aarch64_sqdmlsl2_n<mode>, aarch64_sqdmull2<mode>)
8988         (aarch64_sqdmull2_lane<mode>, aarch64_sqdmull2_laneq<mode>)
8989         (aarch64_sqdmull2_n<mode>): Update accordingly.
8991 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8992             Alan Hayward  <alan.hayward@arm.com>
8993             David Sherwood  <david.sherwood@arm.com>
8995         * config/aarch64/aarch64-protos.h (aarch64_reverse_mask): Take
8996         the number of units too.
8997         * config/aarch64/aarch64.c (aarch64_reverse_mask): Likewise.
8998         * config/aarch64/aarch64-simd.md (vec_load_lanesoi<mode>)
8999         (vec_store_lanesoi<mode>, vec_load_lanesci<mode>)
9000         (vec_store_lanesci<mode>, vec_load_lanesxi<mode>)
9001         (vec_store_lanesxi<mode>): Update accordingly.
9003 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
9004             Alan Hayward  <alan.hayward@arm.com>
9005             David Sherwood  <david.sherwood@arm.com>
9007         * config/aarch64/aarch64-protos.h (aarch64_endian_lane_rtx): Declare.
9008         * config/aarch64/aarch64.c (aarch64_endian_lane_rtx): New function.
9009         * config/aarch64/aarch64.h (ENDIAN_LANE_N): Take the number
9010         of units rather than the mode.
9011         * config/aarch64/iterators.md (nunits): New mode attribute.
9012         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args):
9013         Use aarch64_endian_lane_rtx instead of GEN_INT (ENDIAN_LANE_N ...).
9014         * config/aarch64/aarch64-simd.md (aarch64_dup_lane<mode>)
9015         (aarch64_dup_lane_<vswap_width_name><mode>, *aarch64_mul3_elt<mode>)
9016         (*aarch64_mul3_elt_<vswap_width_name><mode>): Likewise.
9017         (*aarch64_mul3_elt_to_64v2df, *aarch64_mla_elt<mode>): Likewise.
9018         (*aarch64_mla_elt_<vswap_width_name><mode>, *aarch64_mls_elt<mode>)
9019         (*aarch64_mls_elt_<vswap_width_name><mode>, *aarch64_fma4_elt<mode>)
9020         (*aarch64_fma4_elt_<vswap_width_name><mode>):: Likewise.
9021         (*aarch64_fma4_elt_to_64v2df, *aarch64_fnma4_elt<mode>): Likewise.
9022         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
9023         (*aarch64_fnma4_elt_to_64v2df, reduc_plus_scal_<mode>): Likewise.
9024         (reduc_plus_scal_v4sf, reduc_<maxmin_uns>_scal_<mode>): Likewise.
9025         (reduc_<maxmin_uns>_scal_<mode>): Likewise.
9026         (*aarch64_get_lane_extend<GPI:mode><VDQQH:mode>): Likewise.
9027         (*aarch64_get_lane_zero_extendsi<mode>): Likewise.
9028         (aarch64_get_lane<mode>, *aarch64_mulx_elt_<vswap_width_name><mode>)
9029         (*aarch64_mulx_elt<mode>, *aarch64_vgetfmulx<mode>): Likewise.
9030         (aarch64_sq<r>dmulh_lane<mode>, aarch64_sq<r>dmulh_laneq<mode>)
9031         (aarch64_sqrdml<SQRDMLH_AS:rdma_as>h_lane<mode>): Likewise.
9032         (aarch64_sqrdml<SQRDMLH_AS:rdma_as>h_laneq<mode>): Likewise.
9033         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>): Likewise.
9034         (aarch64_sqdml<SBINQOPS:as>l_laneq<mode>): Likewise.
9035         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
9036         (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>_internal): Likewise.
9037         (aarch64_sqdmull_lane<mode>, aarch64_sqdmull_laneq<mode>): Likewise.
9038         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
9039         (aarch64_sqdmull2_laneq<mode>_internal): Likewise.
9040         (aarch64_vec_load_lanesoi_lane<mode>): Likewise.
9041         (aarch64_vec_store_lanesoi_lane<mode>): Likewise.
9042         (aarch64_vec_load_lanesci_lane<mode>): Likewise.
9043         (aarch64_vec_store_lanesci_lane<mode>): Likewise.
9044         (aarch64_vec_load_lanesxi_lane<mode>): Likewise.
9045         (aarch64_vec_store_lanesxi_lane<mode>): Likewise.
9046         (aarch64_simd_vec_set<mode>): Update use of ENDIAN_LANE_N.
9047         (aarch64_simd_vec_setv2di): Likewise.
9049 2017-11-06  Carl Love  <cel@us.ibm.com>
9051         * config/rs6000/rs6000-c.c (P8V_BUILTIN_VEC_REVB): Add power 8
9052         definitions.
9053         (P9V_BUILTIN_VEC_REVB): Remove the power 9 instance definitions.
9054         * config/rs6000/altivec.h (vec_revb): Change the #define from power 9
9055         to power 8.
9056         * config/rs6000/r6000-protos.h (swap_endian_selector_for_mode): Add new
9057         extern declaration.
9058         * config/rs6000/rs6000.c (swap_endian_selector_for_mode): Add function.
9059         * config/rs6000/rs6000-builtin.def (BU_P8V_VSX_1, BU_P8V_OVERLOAD_1):
9060         Add power 8 macro expansions.
9061         (BU_P9V_OVERLOAD_1): Remove power 9 overload expansion.
9062         * config/rs6000/vsx.md (revb_<mode>): Add define_expand to generate
9063         power 8 instructions.  (VSX_XXBR): Add iterator.
9065 2017-11-06  Wilco Dijkstra  <wdijkstr@arm.com>
9067         * config/arm/arm.md (predicable_short_it): Change default to "no",
9068         improve documentation, remove uses that are identical to the default.
9069         (enabled_for_depr_it): Rename to enabled_for_short_it.
9070         * gcc/config/arm/arm-fixed.md (predicable_short_it):
9071         Remove default uses.
9072         * gcc/config/arm/ldmstm.md (predicable_short_it): Likewise.
9073         * gcc/config/arm/sync.md (predicable_short_it): Likewise.
9074         * gcc/config/arm/thumb2.md (predicable_short_it): Likewise.
9075         * gcc/config/arm/vfp.md (predicable_short_it): Likewise.
9077 2017-11-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
9079         PR target/82748
9080         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_1): Delete
9081         float128 helper macros, which are no longer used after deleting
9082         the old 'q' built-in functions, and moving the round to odd
9083         built-in functions to being special built-in functions.
9084         (BU_FLOAT128_2): Likewise.
9085         (BU_FLOAT128_1_HW): Likewise.
9086         (BU_FLOAT128_2_HW): Likewise.
9087         (BU_FLOAT128_3_HW): Likewise.
9088         (FABSQ): Delete old 'q' built-in functions.
9089         (COPYSIGNQ): Likewise.
9090         (SQRTF128_ODD): Move round to odd built-in functions to be
9091         special built-in functions, so that we can handle
9092         -mabi=ieeelongdouble.
9093         (TRUNCF128_ODD): Likewise.
9094         (ADDF128_ODD): Likewise.
9095         (SUBF128_ODD): Likewise.
9096         (MULF128_ODD): Likewise.
9097         (DIVF128_ODD): Likewise.
9098         (FMAF128_ODD): Likewise.
9099         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Map old 'q'
9100         built-in names to 'f128'.
9101         * config/rs6000/rs6000.c (rs6000_fold_builtin): Remove folding the
9102         old 'q' built-in functions, as the machine independent code for
9103         'f128' built-in functions handles this.
9104         (rs6000_expand_builtin): Add expansion for float128 round to odd
9105         functions, keying off on -mabi=ieeelongdouble of whether to use
9106         the KFmode or TFmode variant.
9107         (rs6000_init_builtins): Initialize the _Float128 round to odd
9108         built-in functions.
9109         * doc/extend.texi (PowerPC Built-in Functions): Document the old
9110         _Float128 'q' built-in functions are now mapped into the new
9111         'f128' built-in functions.
9113 2017-11-06  David Edelsohn  <dje.gcc@gmail.com>
9115         * collect2.c (add_lto_object): Compile for OBJECT_COFF.
9116         (scan_prog_file): Don't skip PASS_LTOINFO. Scan for LTO objects.
9118 2017-11-06  David Malcolm  <dmalcolm@redhat.com>
9120         PR jit/82826
9121         * ipa-fnsummary.c (ipa_fnsummary_c_finalize): New function.
9122         * ipa-fnsummary.h (ipa_fnsummary_c_finalize): New decl.
9123         * toplev.c: Include "ipa-fnsummary.h".
9124         (toplev::finalize): Call ipa_fnsummary_c_finalize.
9126 2017-11-06  Jakub Jelinek  <jakub@redhat.com>
9128         PR tree-optimization/82838
9129         * gimple-ssa-store-merging.c
9130         (imm_store_chain_info::output_merged_store): Call force_gimple_operand_1
9131         on a separate gimple_seq which is then appended to seq.
9133 2017-11-06  Jeff Law  <law@redhat.com>
9135         PR target/82788
9136         * config/i386/i386.c (PROBE_INTERVAL): Remove.
9137         (get_probe_interval): New functions.
9138         (ix86_adjust_stack_and_probe_stack_clash): Use get_probe_interval.
9139         (ix86_adjust_stack_and_probe): Likewise.
9140         (output_adjust_stack_and_probe): Likewise.
9141         (ix86_emit_probe_stack_range): Likewise.
9142         (ix86_expand_prologue): Likewise.
9144 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
9146         PR tree-optimization/82816
9147         * tree-ssa-math-opts.c (convert_mult_to_widen): Return false
9148         if the modes of the two types are the same.
9149         (convert_plusminus_to_widen): Likewise.
9151 2017-11-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
9153         * config/rs6000/altivec.md (*p9_vadu<mode>3) Rename to
9154         p9_vadu<mode>3.
9155         (usadv16qi): New define_expand.
9156         (usadv8hi): New define_expand.
9158 2017-11-06  Jan Hubicka  <hubicka@ucw.cz>
9160         PR bootstrap/82832
9161         * ipa-inline-transform.c (update_noncloned_frequencies): Always
9162         scale.
9163         (inline_transform): Likewise.
9164         * predict.c (counts_to_freqs): Remove useless conditional.
9165         * profile-count.h (profile_count::apply_scale): Move sanity check.
9166         * tree-inline.c (copy_bb): Always scale.
9167         (copy_cfg_body): Likewise.
9169 2017-11-06  Christophe Lyon  <christophe.lyon@linaro.org>
9171         PR target/67591
9172         * config/arm/arm.md (*sub_shiftsi): Add predicable_short_it
9173         attribute.
9174         (*cmp_ite0): Add enabled_for_depr_it attribute.
9175         (*cmp_ite1): Likewise.
9177 2017-11-06  Segher Boessenkool  <segher@kernel.crashing.org>
9179         * config/rs6000/rs6000.c (rs6000_insn_cost): Handle TYPE_MFCR and
9180         TYPE_MFCRF.
9182 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
9184         * tree-vrp.c (vrp_int_const_binop): Return true on success and
9185         return the value by pointer.
9186         (extract_range_from_multiplicative_op_1): Update accordingly.
9187         Return as soon as an operation fails.
9189 2017-11-05  Tom de Vries  <tom@codesourcery.com>
9191         PR other/82784
9192         * asan.c (DEF_SANITIZER_BUILTIN_1): Factor out of ...
9193         (DEF_SANITIZER_BUILTIN): ... here.
9194         (initialize_sanitizer_builtins): Use DEF_SANITIZER_BUILTIN_1 instead of
9195         DEF_SANITIZER_BUILTIN in if stmt.  Add missing semicolon.
9197 2017-11-05  Tom de Vries  <tom@codesourcery.com>
9199         PR other/82784
9200         * config/elfos.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove semicolon after
9201         macro body.
9202         (ASM_OUTPUT_CASE_LABEL): Add semicolon after
9203         ASM_OUTPUT_BEFORE_CASE_LABEL call.
9204         * config/arc/arc.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove semicolon
9205         after macro body.
9206         * config/m68k/m68kelf.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
9207         * config/mips/mips.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
9208         * config/v850/v850.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
9210 2017-11-05  Tom de Vries  <tom@codesourcery.com>
9212         PR other/82784
9213         * graphite-scop-detection.c (DEBUG_PRINT): Remove semicolon after
9214         "do {} while (0)".
9216 2017-11-04  Michael Clark  <michaeljclark@mac.com>
9218         * config/riscv/riscv.c (riscv_print_operand): Add a 'i' format.
9219         config/riscv/riscv.md (addsi3): Use 'i' for immediates.
9220         (adddi3): Likewise.
9221         (*addsi3_extended): Likewise.
9222         (*addsi3_extended2): Likewise.
9223         (<optab>si3): Likewise.
9224         (<optab>di3): Likewise.
9225         (<optab><mode>3): Likewise.
9226         (<*optabe>si3_internal): Likewise.
9227         (zero_extendqi<SUPERQI:mode>2): Likewise.
9228         (*add<mode>hi3): Likewise.
9229         (*xor<mode>hi3): Likewise.
9230         (<optab>di3): Likewise.
9231         (*<optab>si3_extend): Likewise.
9232         (*sge<u>_<X:mode><GPR:mode>): Likewise.
9233         (*slt<u>_<X:mode><GPR:mode>): Likewise.
9234         (*sle<u>_<X:mode><GPR:mode>): Likewise.
9236 2017-11-04  Andrew Waterman  <andrew@sifive.com>
9238         * config/riscv/riscv.c (riscv_option_override): Conditionally set
9239         TARGET_STRICT_ALIGN based upon -mtune argument.
9241 2017-11-04  Andrew Waterman  <andrew@sifive.com>
9243         * config/riscv/riscv.h (SLOW_BYTE_ACCESS): Change to 1.
9245 2017-11-04  Daniel Santos  <daniel.santos@pobox.com>
9247         * config/i386/i386.c (choose_basereg): Use optional scratch
9248         register and add assertion.
9249         (x86_emit_outlined_ms2sysv_save): Use scratch register when
9250         needed, and don't allocate stack.
9251         (ix86_expand_prologue): Rearrange where SSE saves/stub call is
9252         emitted, correct wrong allocation with -mcall-ms2sysv-xlogues.
9253         (ix86_emit_outlined_ms2sysv_restore): Fix non-immediate offsets.
9255 2017-11-03  Jeff Law  <law@redhat.com>
9257         * config/i386/i386.c (ix86_emit_restore_reg_using_pop): Prototype.
9258         (ix86_adjust_stack_and_probe_stack_clash): Use a push/pop sequence
9259         to probe at the start of a noreturn function.
9261 2017-11-03  Jakub Jelinek  <jakub@redhat.com>
9263         PR tree-optimization/78821
9264         * gimple-ssa-store-merging.c: Update the file comment.
9265         (MAX_STORE_ALIAS_CHECKS): Define.
9266         (struct store_operand_info): New type.
9267         (store_operand_info::store_operand_info): New constructor.
9268         (struct store_immediate_info): Add rhs_code and ops data members.
9269         (store_immediate_info::store_immediate_info): Add rhscode, op0r
9270         and op1r arguments to the ctor, initialize corresponding data members.
9271         (struct merged_store_group): Add load_align_base and load_align
9272         data members.
9273         (merged_store_group::merged_store_group): Initialize them.
9274         (merged_store_group::do_merge): Update them.
9275         (merged_store_group::apply_stores): Pick the constant for
9276         encode_tree_to_bitpos from one of the two operands, or skip
9277         encode_tree_to_bitpos if neither operand is a constant.
9278         (class pass_store_merging): Add process_store method decl.  Remove
9279         bool argument from terminate_all_aliasing_chains method decl.
9280         (pass_store_merging::terminate_all_aliasing_chains): Remove
9281         var_offset_p argument and corresponding handling.
9282         (stmts_may_clobber_ref_p): New function.
9283         (compatible_load_p): New function.
9284         (imm_store_chain_info::coalesce_immediate_stores): Terminate group
9285         if there is overlap and rhs_code is not INTEGER_CST.  For
9286         non-overlapping stores terminate group if rhs is not mergeable.
9287         (get_alias_type_for_stmts): Change first argument from
9288         auto_vec<gimple *> & to vec<gimple *> &.  Add IS_LOAD, CLIQUEP and
9289         BASEP arguments.  If IS_LOAD is true, look at rhs1 of the stmts
9290         instead of lhs.  Compute *CLIQUEP and *BASEP in addition to the
9291         alias type.
9292         (get_location_for_stmts): Change first argument from
9293         auto_vec<gimple *> & to vec<gimple *> &.
9294         (struct split_store): Remove orig_stmts data member, add orig_stores.
9295         (split_store::split_store): Create orig_stores rather than orig_stmts.
9296         (find_constituent_stmts): Renamed to ...
9297         (find_constituent_stores): ... this.  Change second argument from
9298         vec<gimple *> * to vec<store_immediate_info *> *, push pointers
9299         to info structures rather than the statements.
9300         (split_group): Rename ALLOW_UNALIGNED argument to
9301         ALLOW_UNALIGNED_STORE, add ALLOW_UNALIGNED_LOAD argument and handle
9302         it.  Adjust find_constituent_stores caller.
9303         (imm_store_chain_info::output_merged_store): Handle rhs_code other
9304         than INTEGER_CST, adjust split_group, get_alias_type_for_stmts and
9305         get_location_for_stmts callers.  Set MR_DEPENDENCE_CLIQUE and
9306         MR_DEPENDENCE_BASE on the MEM_REFs if they are the same in all stores.
9307         (mem_valid_for_store_merging): New function.
9308         (handled_load): New function.
9309         (pass_store_merging::process_store): New method.
9310         (pass_store_merging::execute): Use process_store method.  Adjust
9311         terminate_all_aliasing_chains caller.
9313 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9315         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9316         Return true for more constants, symbols and label references.
9317         (aarch64_valid_floating_const): Remove unused function.
9319 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9321         PR target/82786
9322         * config/aarch64/aarch64.c (aarch64_layout_frame):
9323         Undo forcing of LR at bottom of frame.
9325 2017-11-03  Jeff Law  <law@redhat.com>
9327         PR target/82823
9328         * config/i386/i386.c (ix86_expand_prologue): Tighten assert
9329         for int_registers_saved.
9331         * cfganal.c (single_pred_edge_ignoring_loop_edges): New function
9332         extracted from tree-ssa-dom.c.
9333         * cfganal.h (single_pred_edge_ignoring_loop_edges): Prototype.
9334         * tree-ssa-dom.c (single_incoming_edge_ignoring_loop_edges): Remove.
9335         (record_equivalences_from_incoming_edge): Add additional argument
9336         to single_pred_edge_ignoring_loop_edges call.
9337         * tree-ssa-uncprop.c (single_incoming_edge_ignoring_loop_edges): Remove.
9338         (uncprop_dom_walker::before_dom_children): Add additional argument
9339         to single_pred_edge_ignoring_loop_edges call.
9340         * tree-ssa-sccvn.c (sccvn_dom_walker::before_dom_children): Use
9341         single_pred_edge_ignoring_loop_edges rather than open coding.
9342         * tree-vrp.c (evrp_dom_walker::before_dom_children): Similarly.
9344 2017-11-03  Marc Glisse  <marc.glisse@inria.fr>
9346         * match.pd (-(-A)): Rewrite.
9348 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9350         * config/rs6000/rs60000-protos.h (rs6000_emit_sISEL): Delete.
9351         (rs6000_emit_int_cmove): New declaration.
9352         * config/rs6000/rs6000.c (rs6000_emit_int_cmove): Delete declaration.
9353         (rs6000_emit_sISEL): Delete.
9354         (rs6000_emit_int_cmove): Make non-static.
9355         * config/rs6000/rs6000.md (cstore<mode>4): Use rs6000_emit_int_cmove
9356         instead of rs6000_emit_sISEL.
9358 2017-11-03  Jan Hubicka  <hubicka@ucw.cz>
9360         * asan.c (create_cond_insert_point): Maintain profile.
9361         * ipa-utils.c (ipa_merge_profiles): Be sure only IPA profiles are
9362         merged.
9363         * basic-block.h (struct basic_block_def): Remove frequency.
9364         (EDGE_FREQUENCY): Use to_frequency
9365         * bb-reorder.c (push_to_next_round_p): Use only IPA counts for global
9366         heuristics.
9367         (find_traces): Update to use to_frequency.
9368         (find_traces_1_round): Likewise; use only IPA counts.
9369         (bb_to_key): Likewise.
9370         (connect_traces): Use IPA counts only.
9371         (copy_bb_p): Update to use to_frequency.
9372         (fix_up_crossing_landing_pad): Likewise.
9373         (sanitize_hot_paths): Likewise.
9374         * bt-load.c (basic_block_freq): Likewise.
9375         * cfg.c (init_flow): Set count_max to uninitialized.
9376         (check_bb_profile): Remove frequencies; check counts.
9377         (dump_bb_info): Do not dump frequencies.
9378         (update_bb_profile_for_threading): Update counts only.
9379         (scale_bbs_frequencies_int): Likewise.
9380         (MAX_SAFE_MULTIPLIER): Remove.
9381         (scale_bbs_frequencies_gcov_type): Update counts only.
9382         (scale_bbs_frequencies_profile_count): Update counts only.
9383         (scale_bbs_frequencies): Update counts only.
9384         * cfg.h (struct control_flow_graph): Add count-max.
9385         (update_bb_profile_for_threading): Update prototype.
9386         * cfgbuild.c (find_bb_boundaries): Do not update frequencies.
9387         (find_many_sub_basic_blocks): Likewise.
9388         * cfgcleanup.c (try_forward_edges): Likewise.
9389         (try_crossjump_to_edge): Likewise.
9390         * cfgexpand.c (expand_gimple_cond): Likewise.
9391         (expand_gimple_tailcall): Likewise.
9392         (construct_init_block): Likewise.
9393         (construct_exit_block): Likewise.
9394         * cfghooks.c (verify_flow_info): Check consistency of counts.
9395         (dump_bb_for_graph): Do not dump frequencies.
9396         (split_block_1): Do not update frequencies.
9397         (split_edge): Do not update frequencies.
9398         (make_forwarder_block): Do not update frequencies.
9399         (duplicate_block): Do not update frequencies.
9400         (account_profile_record): Do not update frequencies.
9401         * cfgloop.c (find_subloop_latch_edge_by_profile): Use IPA counts
9402         for global heuristics.
9403         * cfgloopanal.c (average_num_loop_insns): Update to use to_frequency.
9404         (expected_loop_iterations_unbounded): Use counts only.
9405         * cfgloopmanip.c (scale_loop_profile): Simplify.
9406         (create_empty_loop_on_edge): Simplify
9407         (loopify): Simplify
9408         (duplicate_loop_to_header_edge): Simplify
9409         * cfgrtl.c (force_nonfallthru_and_redirect): Update profile.
9410         (update_br_prob_note): Take care of removing note when profile
9411         becomes undefined.
9412         (relink_block_chain): Do not dump frequency.
9413         (rtl_account_profile_record): Use to_frequency.
9414         * cgraph.c (symbol_table::create_edge): Convert count to ipa count.
9415         (cgraph_edge::redirect_call_stmt_to_calle): Conver tcount to ipa count.
9416         (cgraph_update_edges_for_call_stmt_node): Likewise.
9417         (cgraph_edge::verify_count_and_frequency): Update.
9418         (cgraph_node::verify_node): Temporarily disable frequency verification.
9419         * cgraphbuild.c (compute_call_stmt_bb_frequency): Use
9420         to_cgraph_frequency.
9421         (cgraph_edge::rebuild_edges): Convert to ipa counts.
9422         * cgraphunit.c (init_lowered_empty_function): Do not initialize
9423         frequencies.
9424         (cgraph_node::expand_thunk): Update profile.
9425         * except.c (dw2_build_landing_pads): Do not update frequency.
9426         * final.c (compute_alignments): Use to_frequency.
9427         (dump_basic_block_info): Do not dump frequency.
9428         * gimple-pretty-print.c (dump_profile): Do not dump frequency.
9429         (dump_gimple_bb_header): Do not dump frequency.
9430         * gimple-ssa-isolate-paths.c (isolate_path): Do not update frequency;
9431         do update count.
9432         * gimple-streamer-in.c (input_bb): Do not stream frequency.
9433         * gimple-streamer-out.c (output_bb): Do not stream frequency.
9434         * haifa-sched.c (sched_pressure_start_bb): Use to_freuqency.
9435         (init_before_recovery): Do not update frequency.
9436         (sched_create_recovery_edges): Do not update frequency.
9437         * hsa-gen.c (convert_switch_statements): Do not update frequency.
9438         * ipa-cp.c (ipcp_propagate_stage): Update search for max_count.
9439         (ipa_cp_c_finalize): Set max_count to uninitialized.
9440         * ipa-fnsummary.c (get_minimal_bb): Use counts.
9441         (param_change_prob): Use counts.
9442         * ipa-profile.c (ipa_profile_generate_summary): Do not summarize
9443         local profiles.
9444         * ipa-split.c (consider_split): Use to_frequency.
9445         (split_function): Use to_frequency.
9446         * ira-build.c (loop_compare_func): Likewise.
9447         (mark_loops_for_removal): Likewise.
9448         (mark_all_loops_for_removal): Likewise.
9449         * loop-doloop.c (doloop_modify): Do not update frequency.
9450         * loop-unroll.c (unroll_loop_runtime_iterations): Do not update
9451         frequency.
9452         * lto-streamer-in.c (input_function): Update count_max.
9453         * omp-expand.c (expand_omp_taskreg): Update count_max.
9454         * omp-simd-clone.c (simd_clone_adjust): Update profile.
9455         * predict.c (maybe_hot_frequency_p): Use to_frequency.
9456         (maybe_hot_count_p): Use ipa counts only.
9457         (maybe_hot_bb_p): Simplify.
9458         (maybe_hot_edge_p): Simplify.
9459         (probably_never_executed): Do not take frequency argument.
9460         (probably_never_executed_bb_p): Do not pass frequency.
9461         (probably_never_executed_edge_p): Likewise.
9462         (combine_predictions_for_bb): Check that profile is nonzero.
9463         (propagate_freq): Do not set frequency.
9464         (drop_profile): Simplify.
9465         (counts_to_freqs): Simplify.
9466         (expensive_function_p): Use to_frequency.
9467         (propagate_unlikely_bbs_forward): Simplify.
9468         (determine_unlikely_bbs): Simplify.
9469         (estimate_bb_frequencies): Add hack to silence graphite issues.
9470         (compute_function_frequency): Use ipa counts.
9471         (pass_profile::execute): Update.
9472         (rebuild_frequencies): Use counts only.
9473         (force_edge_cold): Use counts only.
9474         * profile-count.c (profile_count::dump): Dump new count types.
9475         (profile_count::differs_from_p): Check compatiblity.
9476         (profile_count::to_frequency): New function.
9477         (profile_count::to_cgraph_frequency): New function.
9478         * profile-count.h (struct function): Declare.
9479         (enum profile_quality): Add profile_guessed_local and
9480         profile_guessed_global0.
9481         (class profile_proability): Decrease number of bits to 29;
9482         update from_reg_br_prob_note and to_reg_br_prob_note.
9483         (class profile_count: Update comment; decrease number of bits
9484         to 61. Check compatibility.
9485         (profile_count::compatible_p): New private member function.
9486         (profile_count::ipa_p): New member function.
9487         (profile_count::operator<): Handle global zero correctly.
9488         (profile_count::operator>): Handle global zero correctly.
9489         (profile_count::operator<=): Handle global zero correctly.
9490         (profile_count::operator>=): Handle global zero correctly.
9491         (profile_count::nonzero_p): New member function.
9492         (profile_count::force_nonzero): New member function.
9493         (profile_count::max): New member function.
9494         (profile_count::apply_scale): Handle IPA scalling.
9495         (profile_count::guessed_local): New member function.
9496         (profile_count::global0): New member function.
9497         (profile_count::ipa): New member function.
9498         (profile_count::to_frequency): Declare.
9499         (profile_count::to_cgraph_frequency): Declare.
9500         * profile.c (OVERLAP_BASE): Delete.
9501         (compute_frequency_overlap): Delete.
9502         (compute_branch_probabilities): Do not use compute_frequency_overlap.
9503         * regs.h (REG_FREQ_FROM_BB): Use to_frequency.
9504         * sched-ebb.c (rank): Use counts only.
9505         * shrink-wrap.c (handle_simple_exit): Use counts only.
9506         (try_shrink_wrapping): Use counts only.
9507         (place_prologue_for_one_component): Use counts only.
9508         * tracer.c (find_best_predecessor): Use to_frequency.
9509         (find_trace): Use to_frequency.
9510         (tail_duplicate): Use to_frequency.
9511         * trans-mem.c (expand_transaction): Do not update frequency.
9512         * tree-call-cdce.c: Do not update frequency.
9513         * tree-cfg.c (gimple_find_sub_bbs): Likewise.
9514         (gimple_merge_blocks): Likewise.
9515         (gimple_split_edge): Likewise.
9516         (gimple_duplicate_sese_region): Likewise.
9517         (gimple_duplicate_sese_tail): Likewise.
9518         (move_sese_region_to_fn): Likewise.
9519         (gimple_account_profile_record): Likewise.
9520         (insert_cond_bb): Likewise.
9521         * tree-complex.c (expand_complex_div_wide): Likewise.
9522         * tree-eh.c (lower_resx): Update profile.
9523         * tree-inline.c (copy_bb): Simplify count scaling; do not scale
9524         frequencies.
9525         (initialize_cfun): Do not initialize frequencies
9526         (freqs_to_counts): Delete.
9527         (copy_cfg_body): Ignore count parameter.
9528         (copy_body): Update.
9529         (expand_call_inline): Update count_max.
9530         (optimize_inline_calls): Update count_max.
9531         (tree_function_versioning): Update count_max.
9532         * tree-ssa-coalesce.c (coalesce_cost_bb): Use to_frequency.
9533         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Do not update
9534         frequency.
9535         * tree-ssa-loop-im.c (execute_sm_if_changed): Use counts only.
9536         * tree-ssa-loop-ivcanon.c (unloop_loops): Do not update freuqency.
9537         (try_peel_loop): Likewise.
9538         * tree-ssa-loop-ivopts.c (get_scaled_computation_cost_at): Use
9539         to_frequency.
9540         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): Pass -1.
9541         (tree_transform_and_unroll_loop): Do not use frequencies
9542         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations):
9543         Use reliable prediction only.
9544         * tree-ssa-loop-unswitch.c (hoist_guard): Do not use frequencies.
9545         * tree-ssa-sink.c (select_best_block): Use to_frequency.
9546         * tree-ssa-tail-merge.c (replace_block_by): Temporarily disable
9547         probability scaling.
9548         * tree-ssa-threadupdate.c (create_block_for_threading): Do
9549         not update frequency
9550         (any_remaining_duplicated_blocks): Likewise.
9551         (update_profile): Likewise.
9552         (estimated_freqs_path): Delete.
9553         (freqs_to_counts_path): Delete.
9554         (clear_counts_path): Delete.
9555         (ssa_fix_duplicate_block_edges): Likewise.
9556         (duplicate_thread_path): Likewise.
9557         * tree-switch-conversion.c (gen_inbound_check): Use counts.
9558         * tree-tailcall.c (decrease_profile): Do not update frequency.
9559         (eliminate_tail_call): Likewise.
9560         * tree-vect-loop-manip.c (vect_do_peeling): Likewise.
9561         * tree-vect-loop.c (scale_profile_for_vect_loop): Likewise.
9562         (optimize_mask_stores): Likewise.
9563         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
9564         * ubsan.c (ubsan_expand_null_ifn): Update profile.
9565         (ubsan_expand_ptr_ifn): Update profile.
9566         * value-prof.c (gimple_ic): Simplify.
9567         * value-prof.h (gimple_ic): Update prototype.
9568         * ipa-inline-transform.c (inline_transform): Fix scaling conditoins.
9569         * ipa-inline.c (compute_uninlined_call_time): Be sure that
9570         counts are nonzero.
9571         (want_inline_self_recursive_call_p): Likewise.
9572         (resolve_noninline_speculation): Only cummulate defined counts.
9573         (inline_small_functions): Use nonzero_p.
9574         (ipa_inline): Do not access freed node.
9576 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9578         * config/aarch64/aarch64.c (aarch64_override_options_internal):
9579         Set PARAM_SCHED_PRESSURE_ALGORITHM to SCHED_PRESSURE_MODEL.
9581 2017-11-03  Kito Cheng  <kito.cheng@gmail.com>
9583         * config/riscv/riscv.c (riscv_legitimize_move): Handle
9584         non-legitimate address.
9586 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9588         * config/rs6000/rs6000.md (*lt0_disi): Delete.
9589         (*lt0_<mode>di, *lt0_<mode>si): New.
9591 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9593         * config/rs6000/rs6000.md (move_from_CR_ov_bit): Change condition to
9594         TARGET_PAIRED_FLOAT.
9596 2017-11-03  Siddhesh Poyarekar  <siddhesh.poyarekar@linaro.org>
9597             Jim Wilson  <jim.wilson@linaro.org>
9599         * config/aarch64/aarch64-cores.def (saphira): New CPU.
9600         * config/aarch64/aarch64-tune.md: Regenerated.
9601         * doc/invoke.texi (AArch64 Options/-mtune): Add "saphira".
9602         * gcc/config/aarch64/aarch64.c (saphira_tunings): New tuning table.
9604 2017-11-03  Cupertino Miranda  <cmiranda@synopsys.com>
9606         * config/arc/arc.c (arc_save_restore): Corrected CFA note.
9607         (arc_expand_prologue): Restore blink for millicode.
9608         * config/arc/linux.h (LINK_EH_SPEC): Defined.
9610 2017-11-03  Richard Sandiford  <richard.sandiford@linaro.org>
9612         PR target/82809
9613         * config/i386/i386.c (ix86_vector_duplicate_value): Use
9614         gen_vec_duplicate after forcing the scalar into a register.
9616 2017-11-02  Segher Boessenkool  <segher@kernel.crashing.org>
9618         * combine (try_combine): Print the insns input to try_combine to the
9619         dump file.
9621 2017-11-02  Steve Ellcey  <sellcey@cavium.com>
9623         PR target/79868
9624         * config/aarch64/aarch64-c.c (aarch64_pragma_target_parse):
9625         Remove second argument from aarch64_process_target_attr call.
9626         * config/aarch64/aarch64-protos.h (aarch64_process_target_attr):
9627         Ditto.
9628         * config/aarch64/aarch64.c (aarch64_attribute_info): Change
9629         field type.
9630         (aarch64_handle_attr_arch): Remove second argument.
9631         (aarch64_handle_attr_cpu): Ditto.
9632         (aarch64_handle_attr_tune): Ditto.
9633         (aarch64_handle_attr_isa_flags): Ditto.
9634         (aarch64_process_one_target_attr): Ditto.
9635         (aarch64_process_target_attr): Ditto.
9636         (aarch64_option_valid_attribute_p): Remove second argument.
9637         on aarch64_process_target_attr call.
9639 2017-11-02  David Malcolm  <dmalcolm@redhat.com>
9641         * diagnostic.c: Include "selftest-diagnostic.h".
9642         (selftest::assert_location_text): New function.
9643         (selftest::test_diagnostic_get_location_text): New function.
9644         (selftest::diagnostic_c_tests): Call it.
9646 2017-11-02  David Malcolm  <dmalcolm@redhat.com>
9648         * Makefile.in (OBJS-libcommon): Add selftest-diagnostic.o.
9649         * diagnostic-show-locus.c: Include "selftest-diagnostic.h".
9650         (class selftest::test_diagnostic_context): Move to...
9651         * selftest-diagnostic.c: New file.
9652         * selftest-diagnostic.h: New file.
9654 2017-11-02  James Bowman  <james.bowman@ftdichip.com>
9656         * config/ft32/ft32.c (ft32_addr_space_legitimate_address_p): increase
9657         offset range for FT32B.
9658         * config/ft32/ft32.h: option "mcompress" enables relaxation.
9659         * config/ft32/ft32.md: Add TARGET_NOPM.
9660         * config/ft32/ft32.opt: Add mft32b, mcompress, mnopm.
9661         * gcc/doc/invoke.texi: Add mft32b, mcompress, mnopm.
9663 2017-11-02  Wilco Dijkstra  <wdijkstr@arm.com>
9665         * config/aarch64/aarch64.h (MALLOC_ABI_ALIGNMENT): New define.
9667 2017-11-02  Jeff Law  <law@redhat.com>
9669         * gimple-ssa-sprintf.c (sprintf_dom_walker): Remove
9670         virtual keyword on FINAL OVERRIDE members.
9672         * tree-ssa-propagate.h (ssa_propagation_engine): Group
9673         virtuals together.  Add virtual destructor.
9674         (substitute_and_fold_engine): Similarly.
9676 2017-11-02  Jan Hubicka  <hubicka@ucw.cz>
9678         * x86-tune.def (X86_TUNE_USE_INCDEC): Enable for Haswell+.
9680 2017-11-02  Richard Biener  <rguenther@suse.de>
9682         PR tree-optimization/82795
9683         * tree-if-conv.c (predicate_mem_writes): Remove bogus assert.
9685 2017-11-02  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
9687         * acinclude.m4 (gcc_AC_INITFINI_ARRAY): Don't require
9688         gcc_SUN_LD_VERSION.
9689         (gcc_GAS_CHECK_FEATURE): Remove.
9690         * configure.ac (ld_vers) <*-*-solaris2*>: Move comments from
9691         gcc_AC_INITFINI_ARRAY here.  Update for Solaris 11.4 changes.
9692         * configure: Regenerate.
9694 2017-11-02  Claudiu Zissulescu <claziss@synopsys.com>
9696         * config/arc/arc.c (hwloop_optimize): Account for empty
9697         body loops.
9699 2017-11-02  Richard Biener  <rguenther@suse.de>
9701         PR middle-end/82765
9702         * varasm.c (decode_addr_const): Make offset HOST_WIDE_INT.
9703         Truncate ARRAY_REF index and element size.
9705 2017-11-01  Palmer Dabbelt  <palmer@dabbelt.com>
9707         * doc/invoke.texi (RISC-V Options): Use "@minus{}2 GB", not "-2 GB".
9709 2017-11-01  Jeff Law  <law@redhat.com>
9711         * tree-ssa-ccp.c (ccp_folder): New class derived from
9712         substitute_and_fold_engine.
9713         (ccp_folder::get_value): New member function.
9714         (ccp_folder::fold_stmt): Renamed from ccp_fold_stmt.
9715         (ccp_fold_stmt): Remove prototype.
9716         (ccp_finalize): Call substitute_and_fold from the ccp_class.
9717         * tree-ssa-copy.c (copy_folder): New class derived from
9718         substitute_and_fold_engine.
9719         (copy_folder::get_value): Renamed from get_value.
9720         (fini_copy_prop): Call substitute_and_fold from copy_folder class.
9721         * tree-vrp.c (vrp_folder): New class derived from
9722         substitute_and_fold_engine.
9723         (vrp_folder::fold_stmt): Renamed from vrp_fold_stmt.
9724         (vrp_folder::get_value): New member function.
9725         (vrp_finalize): Call substitute_and_fold from vrp_folder class.
9726         (evrp_dom_walker::before_dom_children): Similarly for replace_uses_in.
9727         * tree-ssa-propagate.h (substitute_and_fold_engine): New class to
9728         provide a class interface to folder/substitute routines.
9729         (ssa_prop_fold_stmt_fn): Remove typedef.
9730         (ssa_prop_get_value_fn): Likewise.
9731         (subsitute_and_fold): Remove prototype.
9732         (replace_uses_in): Likewise.
9733         * tree-ssa-propagate.c (substitute_and_fold_engine::replace_uses_in):
9734         Renamed from replace_uses_in.  Call the virtual member function
9735         (substitute_and_fold_engine::replace_phi_args_in): Similarly.
9736         (substitute_and_fold_dom_walker): Remove initialization of
9737         data member entries for calbacks.  Add substitute_and_fold_engine
9738         member and initialize it.
9739         (substitute_and_fold_dom_walker::before_dom_children0: Use the
9740         member functions for get_value, replace_phi_args_in c
9741         replace_uses_in, and fold_stmt calls.
9742         (substitute_and_fold_engine::substitute_and_fold): Renamed from
9743         substitute_and_fold.  Remove assert.   Update ctor call.
9745         * tree-ssa-propagate.h (ssa_prop_visit_stmt_fn): Remove typedef.
9746         (ssa_prop_visit_phi_fn): Likewise.
9747         (class ssa_propagation_engine): New class to provide an interface
9748         into ssa_propagate.
9749         * tree-ssa-propagate.c (ssa_prop_visit_stmt): Remove file scoped
9750         variable.
9751         (ssa_prop_visit_phi): Likewise.
9752         (ssa_propagation_engine::simulate_stmt): Moved into class.
9753         Call visit_phi/visit_stmt from the class rather than via
9754         file scoped static variables.
9755         (ssa_propagation_engine::simulate_block): Moved into class.
9756         (ssa_propagation_engine::process_ssa_edge_worklist): Similarly.
9757         (ssa_propagation_engine::ssa_propagate): Similarly.  No longer
9758         set file scoped statics for the visit_stmt/visit_phi callbacks.
9759         * tree-complex.c (complex_propagate): New class derived from
9760         ssa_propagation_engine.
9761         (complex_propagate::visit_stmt): Renamed from complex_visit_stmt.
9762         (complex_propagate::visit_phi): Renamed from complex_visit_phi.
9763         (tree_lower_complex): Call ssa_propagate via the complex_propagate
9764         class.
9765         * tree-ssa-ccp.c: (ccp_propagate): New class derived from
9766         ssa_propagation_engine.
9767         (ccp_propagate::visit_phi): Renamed from ccp_visit_phi_node.
9768         (ccp_propagate::visit_stmt): Renamed from ccp_visit_stmt.
9769         (do_ssa_ccp): Call ssa_propagate from the ccp_propagate class.
9770         * tree-ssa-copy.c (copy_prop): New class derived from
9771         ssa_propagation_engine.
9772         (copy_prop::visit_stmt): Renamed from copy_prop_visit_stmt.
9773         (copy_prop::visit_phi): Renamed from copy_prop_visit_phi_node.
9774         (execute_copy_prop): Call ssa_propagate from the copy_prop class.
9775         * tree-vrp.c (vrp_prop): New class derived from ssa_propagation_engine.
9776         (vrp_prop::visit_stmt): Renamed from vrp_visit_stmt.
9777         (vrp_prop::visit_phi): Renamed from vrp_visit_phi_node.
9778         (execute_vrp): Call ssa_propagate from the vrp_prop class.
9780 2017-11-01  Jakub Jelinek  <jakub@redhat.com>
9782         PR rtl-optimization/82778
9783         PR rtl-optimization/82597
9784         * compare-elim.c (struct comparison): Add in_a_setter field.
9785         (find_comparison_dom_walker::before_dom_children): Remove killed
9786         bitmap and df_simulate_find_defs call, instead walk the defs.
9787         Compute last_setter and initialize in_a_setter.  Merge definitions
9788         with first initialization for a few variables.
9789         (try_validate_parallel): Use insn_invalid_p instead of
9790         recog_memoized.  Return insn rather than just the pattern.
9791         (try_merge_compare): Fix up comment.  Don't uselessly test if
9792         in_a is a REG_P.  Use cmp->in_a_setter instead of walking UD
9793         chains.
9794         (execute_compare_elim_after_reload): Remove df_chain_add_problem
9795         call.
9797 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9798             Alan Hayward  <alan.hayward@arm.com>
9799             David Sherwood  <david.sherwood@arm.com>
9801         * config/aarch64/aarch64.c (aarch64_rtx_costs): Use
9802         aarch64_hard_regno_nregs to get the number of registers
9803         in a mode.
9805 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9806             Alan Hayward  <alan.hayward@arm.com>
9807             David Sherwood  <david.sherwood@arm.com>
9809         * config/aarch64/constraints.md (Upl): Rename to...
9810         (Uaa): ...this.
9811         * config/aarch64/aarch64.md
9812         (*zero_extend<SHORT:mode><GPI:mode>2_aarch64, *addsi3_aarch64_uxtw):
9813         Update accordingly.
9815 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9816             Alan Hayward  <alan.hayward@arm.com>
9817             David Sherwood  <david.sherwood@arm.com>
9819         * config/aarch64/aarch64.c (aarch64_add_constant_internal)
9820         (aarch64_add_constant, aarch64_add_sp, aarch64_sub_sp): Move
9821         earlier in file.
9823 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9824             Alan Hayward  <alan.hayward@arm.com>
9825             David Sherwood  <david.sherwood@arm.com>
9827         * config/aarch64/aarch64.c (aarch64_evpc_trn, aarch64_evpc_uzp)
9828         (aarch64_evpc_zip, aarch64_evpc_ext, aarch64_evpc_rev)
9829         (aarch64_evpc_dup): Generate rtl direcly, rather than using
9830         named expanders.
9831         (aarch64_expand_vec_perm_const_1): Explicitly check for permutes
9832         of a single element.
9833         * config/aarch64/iterators.md: Add a comment above the permute
9834         unspecs to say that they are generated directly by
9835         aarch64_expand_vec_perm_const.
9836         * config/aarch64/aarch64-simd.md: Likewise the permute instructions.
9838 2017-11-01  Nathan Sidwell  <nathan@acm.org>
9840         * tree-dump.c (dequeue_and_dump): Use HAS_DECL_ASSEMBLER_NAME_P.
9842 2017-11-01  Palmer Dabbelt  <palmer@dabbelt.com>
9844         * doc/invoke.texi (RISC-V Options): Explicitly name the medlow
9845         and medany code models, and describe what they do.
9847 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9849         Revert accidental duplicate:
9851         * combine.c (can_change_dest_mode): Reject changes in
9852         REGMODE_NATURAL_SIZE.
9854 2017-11-01  Segher Boessenkool  <segher@kernel.crashing.org>
9856         PR rtl-optimization/64682
9857         PR rtl-optimization/69567
9858         PR rtl-optimization/69737
9859         PR rtl-optimization/82683
9860         * combine.c (distribute_notes) <REG_DEAD>: If the new I2 sets the same
9861         register mentioned in the note, drop the note, unless it came from I3,
9862         in which case it should go to I3 again.
9864 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9866         * tree-ssa-dse.c (normalize_ref): Check whether the ranges overlap
9867         and return false if not.
9868         (clear_bytes_written_by, live_bytes_read): Update accordingly.
9870 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9872         * tree-ssa-alias.h (ranges_overlap_p): Return false if either
9873         range is known to be empty.
9875 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9876             Alan Hayward  <alan.hayward@arm.com>
9877             David Sherwood  <david.sherwood@arm.com>
9879         * simplify-rtx.c (simplify_const_unary_operation): Use GET_MODE_NUNITS
9880         and CONST_VECTOR_NUNITS instead of computing the number of units from
9881         the byte sizes of the vector and element.
9882         (simplify_binary_operation_1): Likewise.
9883         (simplify_const_binary_operation): Likewise.
9884         (simplify_ternary_operation): Likewise.
9886 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9887             Alan Hayward  <alan.hayward@arm.com>
9888             David Sherwood  <david.sherwood@arm.com>
9890         * var-tracking.c (INT_MEM_OFFSET): Replace with...
9891         (int_mem_offset): ...this new function.
9892         (var_mem_set, var_mem_delete_and_set, var_mem_delete)
9893         (find_mem_expr_in_1pdv, dataflow_set_preserve_mem_locs)
9894         (same_variable_part_p, use_type, add_stores, vt_get_decl_and_offset):
9895         Update accordingly.
9897 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9898             Alan Hayward  <alan.hayward@arm.com>
9899             David Sherwood  <david.sherwood@arm.com>
9901         * lower-subreg.c (interesting_mode_p): New function.
9902         (compute_costs, find_decomposable_subregs, decompose_register)
9903         (simplify_subreg_concatn, can_decompose_p, resolve_simple_move)
9904         (resolve_clobber, dump_choices): Use it.
9906 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9907             Alan Hayward  <alan.hayward@arm.com>
9908             David Sherwood  <david.sherwood@arm.com>
9910         * rtlhash.c (add_rtx): Use add_hwi for 'w' and add_int for 'i'.
9912 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9913             Alan Hayward  <alan.hayward@arm.com>
9914             David Sherwood  <david.sherwood@arm.com>
9916         * alias.c (find_base_value, find_base_term): Only process integer
9917         truncations.  Check the precision rather than the size.
9919 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9920             Alan Hayward  <alan.hayward@arm.com>
9921             David Sherwood  <david.sherwood@arm.com>
9923         * machmode.h (is_narrower_int_mode): New function
9924         * optabs.c (expand_float, expand_fix): Use it.
9925         * dwarf2out.c (rotate_loc_descriptor): Likewise.
9927 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9928             Alan Hayward  <alan.hayward@arm.com>
9929             David Sherwood  <david.sherwood@arm.com>
9931         * rtl.h (narrower_subreg_mode): New function.
9932         * ira-color.c (update_costs_from_allocno): Use it.
9934 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9935             Alan Hayward  <alan.hayward@arm.com>
9936             David Sherwood  <david.sherwood@arm.com>
9938         * optabs-query.h (convert_optab_p): New function, split out from...
9939         (convert_optab_handler): ...here.
9940         (widening_optab_handler): Delete.
9941         (find_widening_optab_handler): Remove permit_non_widening parameter.
9942         (find_widening_optab_handler_and_mode): Likewise.  Provide an
9943         override that operates on mode class wrappers.
9944         * optabs-query.c (widening_optab_handler): Delete.
9945         (find_widening_optab_handler_and_mode): Remove permit_non_widening
9946         parameter.  Assert that the two modes are the same class and that
9947         the "from" mode is narrower than the "to" mode.  Use
9948         convert_optab_handler instead of widening_optab_handler.
9949         * expmed.c (expmed_mult_highpart_optab): Use convert_optab_handler
9950         instead of widening_optab_handler.
9951         * expr.c (expand_expr_real_2): Update calls to
9952         find_widening_optab_handler.
9953         * optabs.c (expand_widen_pattern_expr): Likewise.
9954         (expand_binop_directly): Take the insn_code as a parameter.
9955         (expand_binop): Only call find_widening_optab_handler for
9956         conversion optabs; use optab_handler otherwise.  Update calls
9957         to find_widening_optab_handler and expand_binop_directly.
9958         Use convert_optab_handler instead of widening_optab_handler.
9959         * tree-ssa-math-opts.c (convert_mult_to_widen): Update calls to
9960         find_widening_optab_handler and use scalar_mode rather than
9961         machine_mode.
9962         (convert_plusminus_to_widen): Likewise.
9964 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9965             Alan Hayward  <alan.hayward@arm.com>
9966             David Sherwood  <david.sherwood@arm.com>
9968         * machmode.h (fixed_size_mode): New class.
9969         * rtl.h (get_pool_mode): Return fixed_size_mode.
9970         * gengtype.c (main): Add fixed_size_mode.
9971         * target.def (get_raw_result_mode): Return a fixed_size_mode.
9972         (get_raw_arg_mode): Likewise.
9973         * doc/tm.texi: Regenerate.
9974         * targhooks.h (default_get_reg_raw_mode): Return a fixed_size_mode.
9975         * targhooks.c (default_get_reg_raw_mode): Likewise.
9976         * config/ia64/ia64.c (ia64_get_reg_raw_mode): Likewise.
9977         * config/mips/mips.c (mips_get_reg_raw_mode): Likewise.
9978         * config/msp430/msp430.c (msp430_get_raw_arg_mode): Likewise.
9979         (msp430_get_raw_result_mode): Likewise.
9980         * config/avr/avr-protos.h (regmask): Use as_a <fixed_side_mode>
9981         * dbxout.c (dbxout_parms): Require fixed-size modes.
9982         * expr.c (copy_blkmode_from_reg, copy_blkmode_to_reg): Likewise.
9983         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Likewise.
9984         * omp-low.c (lower_oacc_reductions): Likewise.
9985         * simplify-rtx.c (simplify_immed_subreg): Take fixed_size_modes.
9986         (simplify_subreg): Update accordingly.
9987         * varasm.c (constant_descriptor_rtx::mode): Change to fixed_size_mode.
9988         (force_const_mem): Update accordingly.  Return NULL_RTX for modes
9989         that aren't fixed-size.
9990         (get_pool_mode): Return a fixed_size_mode.
9991         (output_constant_pool_2): Take a fixed_size_mode.
9993 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9994             Alan Hayward  <alan.hayward@arm.com>
9995             David Sherwood  <david.sherwood@arm.com>
9997         * doc/rtl.texi (vec_series): Document.
9998         (const): Say that the operand can be a vec_series.
9999         * rtl.def (VEC_SERIES): New rtx code.
10000         * rtl.h (const_vec_series_p_1): Declare.
10001         (const_vec_series_p): New function.
10002         * emit-rtl.h (gen_const_vec_series): Declare.
10003         (gen_vec_series): Likewise.
10004         * emit-rtl.c (const_vec_series_p_1, gen_const_vec_series)
10005         (gen_vec_series): Likewise.
10006         * optabs.c (expand_mult_highpart): Use gen_const_vec_series.
10007         * simplify-rtx.c (simplify_unary_operation): Handle negations
10008         of vector series.
10009         (simplify_binary_operation_series): New function.
10010         (simplify_binary_operation_1): Use it.  Handle VEC_SERIES.
10011         (test_vector_ops_series): New function.
10012         (test_vector_ops): Call it.
10013         * config/powerpcspe/altivec.md (altivec_lvsl): Use
10014         gen_const_vec_series.
10015         (altivec_lvsr): Likewise.
10016         * config/rs6000/altivec.md (altivec_lvsl, altivec_lvsr): Likewise.
10018 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
10019             Alan Hayward  <alan.hayward@arm.com>
10020             David Sherwood  <david.sherwood@arm.com>
10022         * doc/rtl.texi (const): Update description of address constants.
10023         Say that vector constants are allowed too.
10024         * common.md (E, F): Use CONSTANT_P instead of checking for
10025         CONST_VECTOR.
10026         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
10027         checking for CONST_VECTOR.
10028         * expmed.c (make_tree): Use build_vector_from_val for a CONST
10029         VEC_DUPLICATE.
10030         * expr.c (expand_expr_real_2): Check for vector modes instead
10031         of checking for CONST_VECTOR.
10032         * rtl.h (const_vec_p): New function.
10033         (const_vec_duplicate_p): Check for a CONST VEC_DUPLICATE.
10034         (unwrap_const_vec_duplicate): Handle them here too.
10036 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
10037             David Malcolm  <dmalcolm@redhat.com>
10038             Alan Hayward  <alan.hayward@arm.com>
10039             David Sherwood  <david.sherwood@arm.com>
10041         * rtl.h (vec_duplicate_p): New function.
10042         * selftest-rtl.c (assert_rtx_eq_at): New function.
10043         * selftest-rtl.h (ASSERT_RTX_EQ): New macro.
10044         (assert_rtx_eq_at): Declare.
10045         * selftest.h (selftest::simplify_rtx_c_tests): Declare.
10046         * selftest-run-tests.c (selftest::run_tests): Call it.
10047         * simplify-rtx.c: Include selftest.h and selftest-rtl.h.
10048         (simplify_unary_operation_1): Recursively handle vector duplicates.
10049         (simplify_binary_operation_1): Likewise.  Handle VEC_SELECTs of
10050         vector duplicates.
10051         (simplify_subreg): Handle subregs of vector duplicates.
10052         (make_test_reg, test_vector_ops_duplicate, test_vector_ops)
10053         (selftest::simplify_rtx_c_tests): New functions.
10055 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
10056             Alan Hayward  <alan.hayward@arm.com>
10057             David Sherwood  <david.sherwood@arm.com>
10059         * emit-rtl.h (gen_const_vec_duplicate): Declare.
10060         (gen_vec_duplicate): Likewise.
10061         * emit-rtl.c (gen_const_vec_duplicate_1): New function, split
10062         out from...
10063         (gen_const_vector): ...here.
10064         (gen_const_vec_duplicate, gen_vec_duplicate): New functions.
10065         (gen_rtx_CONST_VECTOR): Use gen_const_vec_duplicate for constants
10066         whose elements are all equal.
10067         * optabs.c (expand_vector_broadcast): Use gen_const_vec_duplicate.
10068         * simplify-rtx.c (simplify_const_unary_operation): Likewise.
10069         (simplify_relational_operation): Likewise.
10070         * config/aarch64/aarch64.c (aarch64_simd_gen_const_vector_dup):
10071         Likewise.
10072         (aarch64_simd_dup_constant): Use gen_vec_duplicate.
10073         (aarch64_expand_vector_init): Likewise.
10074         * config/arm/arm.c (neon_vdup_constant): Likewise.
10075         (neon_expand_vector_init): Likewise.
10076         (arm_expand_vec_perm): Use gen_const_vec_duplicate.
10077         (arm_block_set_unaligned_vect): Likewise.
10078         (arm_block_set_aligned_vect): Likewise.
10079         * config/arm/neon.md (neon_copysignf<mode>): Likewise.
10080         * config/i386/i386.c (ix86_expand_vec_perm): Likewise.
10081         (expand_vec_perm_even_odd_pack): Likewise.
10082         (ix86_vector_duplicate_value): Use gen_vec_duplicate.
10083         * config/i386/sse.md (one_cmpl<mode>2): Use CONSTM1_RTX.
10084         * config/ia64/ia64.c (ia64_expand_vecint_compare): Use
10085         gen_const_vec_duplicate.
10086         * config/ia64/vect.md (addv2sf3, subv2sf3): Use CONST1_RTX.
10087         * config/mips/mips.c (mips_gen_const_int_vector): Use
10088         gen_const_vec_duplicate.
10089         (mips_expand_vector_init): Use CONST0_RTX.
10090         * config/powerpcspe/altivec.md (abs<mode>2, nabs<mode>2): Likewise.
10091         (define_split): Use gen_const_vec_duplicate.
10092         * config/rs6000/altivec.md (abs<mode>2, nabs<mode>2): Use CONST0_RTX.
10093         (define_split): Use gen_const_vec_duplicate.
10094         * config/s390/vx-builtins.md (vec_genmask<mode>): Likewise.
10095         (vec_ctd_s64, vec_ctd_u64, vec_ctsl, vec_ctul): Likewise.
10096         * config/spu/spu.c (spu_const): Likewise.
10098 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
10099             Alan Hayward  <alan.hayward@arm.com>
10100             David Sherwood  <david.sherwood@arm.com>
10102         * combine.c (can_change_dest_mode): Reject changes in
10103         REGMODE_NATURAL_SIZE.
10105 2017-10-31  Sandra Loosemore  <sandra@codesourcery.com>
10107         * configure.ac (--enable-libssp): New.
10108         (gcc_cv_libc_provides_ssp): Check for explicit setting before
10109         trying to determine target-specific default.  Adjust indentation.
10110         * configure: Regenerated.
10111         * doc/install.texi (Configuration): Expand --disable-libssp
10112         documentation.
10114 2017-10-31  Daniel Santos  <daniel.santos@pobox.com>
10116         config/i386/i386.c (ix86_expand_epilogue): Correct stack
10117         calculation.
10119 2017-10-31  Martin Jambor  <mjambor@suse.cz>
10121         PR c++/81702
10122         * gimple-fold.c (gimple_get_virt_method_for_vtable): Remove assert.
10124 2017-10-31  David Malcolm  <dmalcolm@redhat.com>
10126         * auto-profile.c (autofdo_source_profile::read): Use
10127         UNKNOWN_LOCATION rather than 0.
10128         * diagnostic-core.h (warning_at_rich_loc): Rename to...
10129         (warning_at): ...this overload.
10130         (warning_at_rich_loc_n): Rename to...
10131         (warning_n): ...this overload.
10132         (error_at_rich_loc): Rename to...
10133         (error_at): ...this overload.
10134         (pedwarn_at_rich_loc): Rename to...
10135         (pedwarn): ...this overload.
10136         (permerror_at_rich_loc): Rename to...
10137         (permerror): ...this overload.
10138         (inform_at_rich_loc): Rename to...
10139         (inform): ...this overload.
10140         * diagnostic.c: (diagnostic_n_impl): Delete location_t-based decl.
10141         (diagnostic_n_impl_richloc): Rename to...
10142         (diagnostic_n_impl): ...this rich_location *-based decl.
10143         (inform_at_rich_loc): Rename to...
10144         (inform): ...this, and add an assertion.
10145         (inform_n): Update for removal of location_t-based diagnostic_n_impl.
10146         (warning_at_rich_loc): Rename to...
10147         (warning_at): ...this, and add an assertion.
10148         (warning_at_rich_loc_n): Rename to...
10149         (warning_n): ...this, and add an assertion.
10150         (warning_n): Update location_t-based implementation for removal of
10151         location_t-based diagnostic_n_impl.
10152         (pedwarn_at_rich_loc): Rename to...
10153         (pedwarn): ...this, and add an assertion.
10154         (permerror_at_rich_loc): Rename to...
10155         (permerror): ...this, and add an assertion.
10156         (error_n): Update for removal of location_t-based diagnostic_n_impl.
10157         (error_at_rich_loc): Rename to...
10158         (error_at): ...this, and add an assertion.
10159         * gcc.c (do_spec_1): Use UNKNOWN_LOCATION rather than 0.
10160         (driver::do_spec_on_infiles): Likewise.
10161         * substring-locations.c (format_warning_va): Update for renaming
10162         of inform_at_rich_loc.
10164 2017-10-31  Michael Meissner  <meissner@linux.vnet.ibm.com>
10166         * builtins.def (DEF_FLOATN_BUILTIN): Change most _Float<N> and
10167         _Float<N>X built-in functions so that the variant without the
10168         "__builtin_" prefix is only enabled for the GNU C and Objective C
10169         languages when they are in non-strict ANSI/ISO mode.
10170         (DEF_EXT_LIB_FLOATN_NX_BUILTINS): Likewise.
10171         * target.def (floatn_builtin_p): Add a target hook to control
10172         whether _Float<N> and _Float<N>X built-in functions without the
10173         "__builtin_" prefix are enabled, and return true for C and
10174         Objective C in the default hook.  Include langhooks.h in
10175         targhooks.c.
10176         * targhooks.h (default_floatn_builtin_p): Likewise.
10177         * targhooks.c (default_floatn_builtin_p): Likewise.
10178         * doc/tm.texi.in (TARGET_FLOATN_BUILTIN_P): Document the
10179         floatn_builtin_p target hook.
10180         * doc/tm.texi (TARGET_FLOATN_BUILTIN_P): Likewise.
10182 2017-10-31  Matthew Fortune  <matthew.fortune@imgtec.com>
10183             Eric Botcazou  <ebotcazou@adacore.com>
10185         PR rtl-optimization/81803
10186         * lra-constraints.c (curr_insn_transform): Also reload the whole
10187         register for a strict subreg no wider than a word if this is for
10188         a WORD_REGISTER_OPERATIONS target.
10190 2017-10-31  Jason Merrill  <jason@redhat.com>
10192         * gdbinit.in: Skip over inlines from timevar.h.
10194 2017-10-31  Martin Liska  <mliska@suse.cz>
10196         * doc/gcov.texi: Document new option.
10197         * gcov.c (print_usage): Likewise print it.
10198         (process_args): Support the argument.
10199         (format_count): New function.
10200         (format_gcov): Use the function.
10202 2017-10-31  Martin Liska  <mliska@suse.cz>
10204         * gcov.c (struct name_map): do not use typedef.
10205         Define operator== and operator<.
10206         (name_search): Remove.
10207         (name_sort): Remove.
10208         (main): Do not allocate names.
10209         (process_file): Add vertical space.
10210         (generate_results): Use std::find.
10211         (release_structures): Do not release memory.
10212         (find_source): Use std::find.
10214 2017-10-31  Martin Liska  <mliska@suse.cz>
10216         * gcov.c (struct line_info): Remove it's typedef.
10217         (line_info::line_info): Add proper ctor.
10218         (line_info::has_block): Do not use a typedef.
10219         (struct source_info): Do not use typedef.
10220         (circuit): Likewise.
10221         (get_cycles_count): Likewise.
10222         (output_intermediate_file): Iterate via vector iterator.
10223         (add_line_counts): Use std::vector methods.
10224         (accumulate_line_counts): Likewise.
10225         (output_lines): Likewise.
10227 2017-10-31  Martin Liska  <mliska@suse.cz>
10229         * gcov.c (struct source_info): Remove typedef.
10230         (source_info::source_info): Add proper ctor.
10231         (accumulate_line_counts): Use struct, not it's typedef.
10232         (output_gcov_file): Likewise.
10233         (output_lines): Likewise.
10234         (main): Do not allocate an array.
10235         (output_intermediate_file): Use size of vector container.
10236         (process_file): Resize the vector.
10237         (generate_results): Do not preallocate, use newly added vector
10238         lines.
10239         (release_structures): Do not release sources.
10240         (find_source): Use vector methods.
10241         (add_line_counts): Do not use typedef.
10243 2017-10-31  Martin Liska  <mliska@suse.cz>
10245         * doc/gcov.texi: Document that.
10246         * gcov.c (add_line_counts): Mark lines with a non-executed
10247         statement.
10248         (output_line_beginning): Handle such lines.
10249         (output_lines): Pass new argument.
10250         (output_intermediate_file): Print it in intermediate format.
10252 2017-10-31  Martin Liska  <mliska@suse.cz>
10254         * color-macros.h: New file.
10255         * diagnostic-color.c: Factor out color related to macros to
10256         color-macros.h.
10257         * doc/gcov.texi: Document -k option.
10258         * gcov.c (INCLUDE_STRING): Include string.h.
10259         (print_usage): Add -k option.
10260         (process_args): Parse it.
10261         (pad_count_string): New function.
10262         (output_line_beginning): Likewise.
10263         (DEFAULT_LINE_START): New macro.
10264         (output_lines): Support color output.
10266 2017-10-31  Martin Liska  <mliska@suse.cz>
10268         PR gcov-profile/82633
10269         * doc/gcov.texi: Document -fkeep-{static,inline}-functions and
10270         their interaction with GCOV infrastructure.
10271         * configure.ac: Add -fkeep-{inline,static}-functions to
10272         coverage_flags.
10273         * configure: Regenerate.
10275 2017-10-31  Uros Bizjak  <ubizjak@gmail.com>
10277         PR target/82772
10278         * config/alpha/sync.md (fetchop_constr) <and>: Change to "rINM".
10280 2017-10-31  Segher Boessenkool  <segher@kernel.crashing.org>
10282         PR target/82674
10283         * config/rs6000/rs6000.md (allocate_stack): Force update interval
10284         into a register if it does not fit into an immediate offset field.
10286 2017-10-31  Olivier Hainque  <hainque@adacore.com>
10288         * gcc/Makefile.in (FLAGS_TO_PASS): Pass libsubdir as well.
10290 2017-10-31  Julia Koval  <julia.koval@intel.com>
10292         * config.gcc: Add gfniintrin.h.
10293         * config/i386/gfniintrin.h: New.
10294         * config/i386/i386-builtin-types.def
10295         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
10296         __builtin_ia32_vgf2p8affineinvqb_v64qi_mask,
10297         __builtin_ia32_vgf2p8affineinvqb_v32qi,
10298         __builtin_ia32_vgf2p8affineinvqb_v32qi_mask,
10299         __builtin_ia32_vgf2p8affineinvqb_v16qi,
10300         __builtin_ia32_vgf2p8affineinvqb_v16qi_mask): New builtins.
10301         * config/i386/i386-builtin.def (V64QI_FTYPE_V64QI_V64QI_INT_V64QI_UDI,
10302         V32QI_FTYPE_V32QI_V32QI_INT_V32QI_USI,
10303         V16QI_FTYPE_V16QI_V16QI_INT_V16QI_UHI,
10304         V64QI_FTYPE_V64QI_V64QI_INT): New types.
10305         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
10306         * config/i386/immintrin.h: Include gfniintrin.h.
10307         * config/i386/sse.md (vgf2p8affineinvqb_*) New pattern.
10309 2017-10-30  Eric Botcazou  <ebotcazou@adacore.com>
10311         * gcc.c (HAVE_TARGET_EXECUTABLE_SUFFIX): Remove old kludge.
10313 2017-10-30  Wilco Dijkstra  <wdijkstr@arm.com>
10315         * config/arm/arm.md (ashldi3): Remove shift by 1 expansion.
10316         (arm_ashldi3_1bit): Remove pattern.
10317         (ashrdi3): Remove shift by 1 expansion.
10318         (arm_ashrdi3_1bit): Remove pattern.
10319         (lshrdi3): Remove shift by 1 expansion.
10320         (arm_lshrdi3_1bit): Remove pattern.
10321         * config/arm/arm.c (arm_rtx_costs_internal): Slightly increase
10322         cost of ashldi3 by 1.
10323         * config/arm/neon.md (ashldi3_neon): Remove shift by 1 expansion.
10324         (<shift>di3_neon): Likewise.
10326 2017-10-30  Dominik Infuehr  <dominik.infuehr@theobroma-systems.com>
10328         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov): Rename
10329         both identically named patterns to (*aarch64_simd_mov<VD:mode>)
10330         and (*aarch64_simd_mov<VQ:mode>).
10331         (*aarch64_simd_mov<VD:mode>): Change type attribute to match
10332         pattern alternative.
10333         (*aarch64_simd_mov<VQ:mode>): Re-order and change type
10334         attributes to match pattern alternative.
10336 2017-10-30  Steven Munroe  <munroesj@gcc.gnu.org>
10338         * config.gcc (powerpc*-*-*): Add emmintrin.h.
10339         * config/rs6000/emmintrin.h: New file.
10340         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include emmintrin.h.
10342 2017-10-30  Wilco Dijkstra  <wdijkstr@arm.com>
10344         * config/arm/vfp.md (movdi_vfp): Merge changes from movdi_vfp_cortexa8.
10345         * (movdi_vfp_cortexa8): Remove pattern.
10347 2017-10-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10349         * doc/install.texi (Specific, alpha*-*-*): Remove DEC OSF/1
10350         etc. reference.
10351         (Specific, alpha*-dec-osf5.1): Remove.
10352         (Specific, mips-sgi-irix5): Remove.
10353         (Specific, mips-sgi-irix6): Remove.
10355 2017-10-30  Jakub Jelinek  <jakub@redhat.com>
10357         PR middle-end/22141
10358         * gimple-ssa-store-merging.c (merged_store_group::apply_stores): Fix
10359         arguments to clear_bit_region_be.
10361 2017-10-30  Jim Wilson  <wilson@tuliptree.org>
10363         * gimplify.c: Include memmodel.h.
10365 2017-10-30  Martin Jambor  <mjambor@suse.cz>
10367         * omp-grid.c (grid_attempt_target_gridification): Also insert a
10368         condition whether loop should be executed at all.
10370 2017-10-30  Will Schmidt  <will_schmidt@vnet.ibm.com>
10372         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
10373         gimple folding of vec_madd() intrinsics.
10374         * config/rs6000/altivec.md (mulv8hi3): Rename altivec_vmladduhm to
10375         fmav8hi4.  (altivec_vmladduhm): Rename to fmav8hi4.
10376         * config/rs6000/rs6000-builtin.def: Rename vmladduhm to fmav8hi4
10378 2017-10-30  Richard Biener  <rguenther@suse.de>
10380         PR tree-optimization/82762
10381         Revert
10382         2017-10-23  Richard Biener  <rguenther@suse.de>
10384         PR tree-optimization/82129
10385         Revert
10386         2017-08-01  Richard Biener  <rguenther@suse.de>
10388         PR tree-optimization/81181
10389         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
10390         (compute_antic): ... end of iteration here.
10392 2017-10-30  Joseph Myers  <joseph@codesourcery.com>
10394         * doc/invoke.texi (C Dialect Options): Document -std=c17,
10395         -std=iso9899:2017 and -std=gnu17.
10396         * doc/standards.texi (C Language): Document C17 support.
10397         * doc/cpp.texi (Overview): Mention -std=c17.
10398         (Standard Predefined Macros): Document C11 and C17 values of
10399         __STDC_VERSION__.  Do not refer to C99 support as incomplete.
10400         * doc/extend.texi (Inline): Do not list individual options for
10401         standards newer than C99.
10402         * dwarf2out.c (highest_c_language, gen_compile_unit_die): Handle
10403         "GNU C17".
10404         * config/rl78/rl78.c (rl78_option_override): Handle "GNU C17"
10405         language name.
10407 2017-10-30  Maxim Ostapenko  <m.ostapenko@samsung.com>
10409         * asan.c (asan_finish_file): Align asan globals array by shadow
10410         granularity.
10412 2017-10-30  Jakub Jelinek  <jakub@redhat.com>
10414         PR middle-end/22141
10415         * gimple-ssa-store-merging.c: Include rtl.h and expr.h.
10416         (struct store_immediate_info): Add bitregion_start and bitregion_end
10417         fields.
10418         (store_immediate_info::store_immediate_info): Add brs and bre
10419         arguments and initialize bitregion_{start,end} from those.
10420         (struct merged_store_group): Add bitregion_start, bitregion_end,
10421         align_base and mask fields.  Drop unnecessary struct keyword from
10422         struct store_immediate_info.  Add do_merge method.
10423         (clear_bit_region_be): Use memset instead of loop storing zeros.
10424         (merged_store_group::do_merge): New method.
10425         (merged_store_group::merge_into): Use do_merge.  Allow gaps in between
10426         stores as long as the surrounding bitregions have no gaps.
10427         (merged_store_group::merge_overlapping): Use do_merge.
10428         (merged_store_group::apply_stores): Test that bitregion_{start,end}
10429         is byte aligned, rather than requiring that start and width are
10430         byte aligned.  Drop unnecessary struct keyword from
10431         struct store_immediate_info.  Allocate and populate also mask array.
10432         Make start of the arrays relative to bitregion_start rather than
10433         start and size them according to bitregion_{end,start} difference.
10434         (struct imm_store_chain_info): Drop unnecessary struct keyword from
10435         struct store_immediate_info.
10436         (pass_store_merging::gate): Punt if BITS_PER_UNIT or CHAR_BIT is not 8.
10437         (pass_store_merging::terminate_all_aliasing_chains): Drop unnecessary
10438         struct keyword from struct store_immediate_info.
10439         (imm_store_chain_info::coalesce_immediate_stores): Allow gaps in
10440         between stores as long as the surrounding bitregions have no gaps.
10441         Formatting fixes.
10442         (struct split_store): Add orig non-static data member.
10443         (split_store::split_store): Initialize orig to false.
10444         (find_constituent_stmts): Return store_immediate_info *, non-NULL
10445         if there is exactly a single original stmt.  Change stmts argument
10446         to pointer from reference, if NULL, don't push anything to it.  Add
10447         first argument, use it to optimize skipping over orig stmts that
10448         are known to be before bitpos already.  Simplify.
10449         (split_group): Return unsigned int count how many stores are or
10450         would be needed rather than a bool.  Add allow_unaligned argument.
10451         Change split_stores argument from reference to pointer, if NULL,
10452         only do a dry run computing how many stores would be produced.
10453         Rewritten algorithm to use both alignment and misalign if
10454         !allow_unaligned and handle bitfield stores with gaps.
10455         (imm_store_chain_info::output_merged_store): Set start_byte_pos
10456         from bitregion_start instead of start.  Compute allow_unaligned
10457         here, if true, do 2 split_group dry runs to compute which one
10458         produces fewer stores and prefer aligned if equal.  Punt if
10459         new count is bigger or equal than original before emitting any
10460         statements, rather than during that.  Remove no longer needed
10461         new_ssa_names tracking.  Replace num_stmts with
10462         split_stores.length ().  Use 32-bit stack allocated entries
10463         in split_stores auto_vec.  Try to reuse original store lhs/rhs1
10464         if possible.  Handle bitfields with gaps.
10465         (pass_store_merging::execute): Ignore bitsize == 0 stores.
10466         Compute bitregion_{start,end} for the stores and construct
10467         store_immediate_info with that.  Formatting fixes.
10469 2017-10-30  Uros Bizjak  <ubizjak@gmail.com>
10471         PR target/82725
10472         * config/i386/i386.c (legitimate_pic_address_disp_p): Allow
10473         UNSPEC_DTPOFF and UNSPEC_NTPOFF with SImode immediate offset.
10475 2017-10-29  Jim Wilson  <wilson@tuliptree.org>
10477         * gimplify.c: Include tm_p.h.
10479         * common.opt (gcoff): Re-add as ignored option.
10480         (gcoff1, gcoff2, gcoff3): Likewise.
10482         * Makefile.in (OBJS): Delete sdbout.o.
10483         (GTFILES): Delete $(srcdir)/sdbout.c.
10484         * debug.h: Delete sdb_debug_hooks.
10485         * final.c: Delete sdbout.h include.
10486         (final_scan_insn): Delete SDB_DEBUG check.
10487         (rest_of_clean_state): Likewise.
10488         * output.h: Delete sdb_begin_function_line.
10489         * sdbout.c: Delete.
10490         * sdbout.h: Delete.
10491         * toplev.c: Delete sdbout.h include.
10492         (process_options): Delete SDB_DEBUG check.
10493         * tree-core.h (tree_type_common): Delete pointer field of
10494         tree_type_symtab.
10495         * tree.c (copy_node): Clear TYPE_SYMTAB_DIE instead of
10496         TYPE_SYMTAB_POINTER.
10497         * tree.h (TYPE_SYMTAB_POINTER): Delete.
10498         (TYPE_SYMTAB_IS_POINTER): Delete.
10499         (TYPE_SYMTAB_IS_DIE): Renumber.
10500         * xcoffout.c: Refer to former sdbout.c file.
10501         (xcoffout_begin_prologue): Use past tense for sdbout.c reference.
10503         * doc/install.texi (--with-stabs): Delete COFF and ECOFF info.
10504         * doc/invoke.texi (SEEALSO): Delete adb and sdb references.
10505         (Debugging Options): Delete -gcoff.
10506         (-gstabs): Delete SDB reference.
10507         (-gcoff): Delete.
10508         (-gcoff@var{level}): Delete.
10509         * doc/passes.texi (Debugging information output): Delete SDB and
10510         sdbout.c references.
10511         * doc/tm.texi: Regenerate.
10512         * doc/tm.texi.in (DWARF_CIE_DATA_ALIGNMENT): Delete SDB from xref.
10513         (SDB and DWARF): Change node name to DWARF and delete SDB and COFF
10514         references.
10515         (DEBUGGER_AUTO_OFFSET): Delete COFF and SDB references.
10516         (PREFERRED_DEBUGGING_TYPE): Delete SDB_DEBUG and -gcoff references.
10517         (SDB_DEBUGGING_INFO): Delete.
10518         (PUT_SDB_@dots{}, SDB_DELIM, SDB_ALLOW_UNKNOWN_REFERENCES)
10519         SDB_ALLOW_FORWARD_REFERENCES, SDB_OUTPUT_SOURCE_LINE): Delete.
10520         * target.def (output_source_filename): Delete COFF reference.
10522         * common.opt (gcoff): Delete.
10523         (gxcoff+): Update Negative chain.
10524         * defaults.h: Delete all references to SDB_DEBUGGING_INFO and
10525         SDB_DEBUG.
10526         * dwarf2out.c (gen_array_type_die): Change SDB to debuggers.
10527         * flag-types.h (enum debug_info_type): Delete SDB_DEBUG.
10528         * function.c (number_blocks): Delete SDB_DEBUGGING_INFO, SDB_DEBUG,
10529         and SDB references.
10530         (expand_function_start): Change sdb reference to past tense.
10531         (expand_function_end): Change sdb reference to past tense.
10532         * gcc.c (cpp_unique_options): Delete gcoff3 reference.
10533         * opts.c (debug_type_names): Delete coff entry.
10534         (common_handle_option): Delete OPT_gcoff case.
10535         * system.h (SDB_DEBUG, SDB_DEBUGGING_INFO): Poison.
10537         * config/dbxcoff.h (PREFERRED_DEBUGGING_TYPE): Set to DBX_DEBUG.
10538         * config/cris/cris.h: Delete SDB reference in comment.
10539         * config/i386/cygming.h: Don't define SDB_DEBUGGING_INFO.
10540         (ASM_DECLARE_FUNCTION_NAME): Delete SDB reference from comment.
10541         * config/i386/gas.h: Don't define SDB_DEBUGGING_INFO.
10542         * config/i386/i386.c (svr4_dbx_register_map): Change SDB references
10543         to past tense.
10544         (ix86_expand_prologue): Likewise.
10545         * config/i386/winnt.c (i386_pe_start_function): Don't check SDB_DEBUG.
10546         * config/ia64/ia64.h: Likewise.
10547         * config/m68k/m68kelf.h (DBX_REGISTER_NUMBER): Delete SDB reference.
10548         * config/mips/mips.h (SUBTARGET_ASM_DEBUGGING_SPEC): Delete gcoff*
10549         support.
10550         * config/mmix/mmix.h: Likewise.
10551         * config/nds32/nds32.c: Likewise.
10552         * config/stormy/storym16.h: Likewise.
10553         * config/visium/visium.h: Likewise.
10554         * config/vx-common.h (SDB_DEBUGGING_INFO): Delete undef.
10556 2017-10-28  Sandra Loosemore  <sandra@codesourcery.com>
10558         * config/nios2/nios2.h (FRAME_GROWS_DOWNWARD): Define to 1.
10559         * config/nios2/nios2.c (nios2_initial_elimination_offset):  Make
10560         FRAME_POINTER_REGNUM point at high end of local var area.
10562 2017-10-27  Eric Botcazou  <ebotcazou@adacore.com>
10564         * bb-reorder.c (find_traces_1_round): Fix off-by-one index.
10565         Move comment around.  Do not reset best_edge for a copiable
10566         destination if the copy would cause a partition change.
10567         (better_edge_p): Remove redundant check.
10569 2017-10-27  Uros Bizjak  <ubizjak@gmail.com>
10571         * config/i386/i386-protos.h (ix86_fp_compare_mode): Remove prototype.
10573 2017-10-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
10575         * builtins.c (CASE_MATHFN_FLOATN): New helper macro to add cases
10576         for math functions that have _Float<N> and _Float<N>X variants.
10577         (mathfn_built_in_2): Add support for math functions that have
10578         _Float<N> and _Float<N>X variants.
10579         (DEF_INTERNAL_FLT_FLOATN_FN): New helper macro.
10580         (expand_builtin_mathfn_ternary): Add support for fma with
10581         _Float<N> and _Float<N>X variants.
10582         (expand_builtin): Likewise.
10583         (fold_builtin_3): Likewise.
10584         * builtins.def (DEF_EXT_LIB_FLOATN_NX_BUILTINS): New macro to
10585         create math function _Float<N> and _Float<N>X variants as external
10586         library builtins.
10587         (BUILT_IN_COPYSIGN _Float<N> and _Float<N>X variants) Use
10588         DEF_EXT_LIB_FLOATN_NX_BUILTINS to make built-in functions using
10589         the __builtin_ prefix and if not strict ansi, without the prefix.
10590         (BUILT_IN_FABS _Float<N> and _Float<N>X variants): Likewise.
10591         (BUILT_IN_FMA _Float<N> and _Float<N>X variants): Likewise.
10592         (BUILT_IN_FMAX _Float<N> and _Float<N>X variants): Likewise.
10593         (BUILT_IN_FMIN _Float<N> and _Float<N>X variants): Likewise.
10594         (BUILT_IN_NAN _Float<N> and _Float<N>X variants): Likewise.
10595         (BUILT_IN_SQRT _Float<N> and _Float<N>X variants): Likewise.
10596         * builtin-types.def (BT_FN_FLOAT16_FLOAT16_FLOAT16_FLOAT16): New
10597         function signatures for fma _Float<N> and _Float<N>X variants.
10598         (BT_FN_FLOAT32_FLOAT32_FLOAT32_FLOAT32): Likewise.
10599         (BT_FN_FLOAT64_FLOAT64_FLOAT64_FLOAT64): Likewise.
10600         (BT_FN_FLOAT128_FLOAT128_FLOAT128_FLOAT128): Likewise.
10601         (BT_FN_FLOAT32X_FLOAT32X_FLOAT32X_FLOAT32X): Likewise.
10602         (BT_FN_FLOAT64X_FLOAT64X_FLOAT64X_FLOAT64X): Likewise.
10603         (BT_FN_FLOAT128X_FLOAT128X_FLOAT128X_FLOAT128X): Likewise.
10604         * gencfn-macros.c (print_case_cfn): Add support for math functions
10605         that have _Float<N> and _Float<N>X variants.
10606         (print_define_operator_list): Likewise.
10607         (fltfn_suffixes): Likewise.
10608         (main): Likewise.
10609         * internal-fn.def (DEF_INTERNAL_FLT_FLOATN_FN): New helper macro
10610         for math functions that have _Float<N> and _Float<N>X variants.
10611         (SQRT): Add support for sqrt, copysign, fmin and fmax _Float<N>
10612         and _Float<N>X variants.
10613         (COPYSIGN): Likewise.
10614         (FMIN): Likewise.
10615         (FMAX): Likewise.
10616         * fold-const.c (tree_call_nonnegative_warnv_p): Add support for
10617         copysign, fma, fmax, fmin, and sqrt _Float<N> and _Float<N>X
10618         variants.
10619         (integer_valued_read_call_p): Likewise.
10620         * fold-const-call.c (fold_const_call_ss): Likewise.
10621         (fold_const_call_sss): Add support for copysign, fmin, and fmax
10622         _Float<N> and _Float<N>X variants.
10623         (fold_const_call_ssss): Add support for fma _Float<N> and
10624         _Float<N>X variants.
10625         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Add
10626         support for copysign and fma _Float<N> and _Float<N>X variants.
10627         (backprop::process_builtin_call_use): Likewise.
10628         * tree-call-cdce.c (can_test_argument_range); Add support for
10629         sqrt _Float<N> and _Float<N>X variants.
10630         (edom_only_function): Likewise.
10631         (get_no_error_domain): Likewise.
10632         * tree-ssa-math-opts.c (internal_fn_reciprocal): Likewise.
10633         * tree-ssa-reassoc.c (attempt_builtin_copysign): Add support for
10634         copysign _Float<N> and _Float<N>X variants.
10635         * config/rs6000/rs6000-builtin.def (SQRTF128): Delete, this is now
10636         handled by machine independent code.
10637         (FMAF128): Likewise.
10638         * doc/cpp.texi (Common Predefined Macros): Document defining
10639         __FP_FAST_FMAF<N> and __FP_FAST_FMAF<N>X if the backend supports
10640         fma _Float<N> and _Float<N>X variants.
10642 2017-10-27  Uros Bizjak  <ubizjak@gmail.com>
10644         PR target/82692
10645         * config/i386/i386-modes.def (CCFPU): Remove definition.
10646         * config/i386/i386.c (put_condition_mode): Remove CCFPU mode handling.
10647         (ix86_cc_modes_compatible): Ditto.
10648         (ix86_expand_carry_flag_compare): Ditto.
10649         (ix86_expand_int_movcc): Ditto.
10650         (ix86_expand_int_addcc): Ditto.
10651         (ix86_reverse_condition): Ditto.
10652         (ix86_unordered_fp_compare): Rename from ix86_fp_compare_mode.
10653         Return true/false for unordered/ordered fp comparisons.
10654         (ix86_cc_mode): Always return CCFPmode for float mode comparisons.
10655         (ix86_prepare_fp_compare_args): Update for rename.
10656         (ix86_expand_fp_compare): Update for rename.  Generate unordered
10657         compare RTXes wrapped with UNSPEC_NOTRAP unspec.
10658         (ix86_expand_sse_compare_and_jump): Ditto.
10659         * config/i386/predicates.md (fcmov_comparison_operator):
10660         Remove CCFPU mode handling.
10661         (ix86_comparison_operator): Ditto.
10662         (ix86_carry_flag_operator): Ditto.
10663         * config/i386/i386.md (UNSPEC_NOTRAP): New unspec.
10664         (*cmpu<mode>_i387): Wrap compare RTX with UNSPEC_NOTRAP unspec.
10665         (*cmpu<mode>_cc_i387): Ditto.
10666         (FPCMP): Remove mode iterator.
10667         (unord): Remove mode attribute.
10668         (unord_subst): New define_subst transformation
10669         (unord): New define_subst attribute.
10670         (unordered): Ditto.
10671         (*cmpi<unord><MODEF:mode>): Rewrite using unord_subst transformation.
10672         (*cmpi<unord>xf_i387): Ditto.
10673         * config/i386/sse.md (<sse>_<unord>comi<round_saeonly_name>): Merge
10674         from <sse>_comi<round_saeonly_name> and <sse>_ucomi<round_saeonly_name>
10675         using unord_subst transformation.
10676         * config/i386/subst.md (SUBST_A): Remove CCFP and CCFPU modes.
10677         (round_saeonly): Also handle CCFP mode.
10678         * reg-stack.c (subst_stack_regs_pat): Handle UNSPEC_NOTRAP unspec.
10679         Remove UNSPEC_SAHF unspec handling.
10681 2017-10-27  Jan Hubicka  <hubicka@ucw.cz>
10683         * x86-tune.def (X86_TUNE_INTER_UNIT_MOVES_TO_VEC): Disable for Zen.
10685 2017-10-27  Jeff Law  <law@redhat.com>
10687         * gimple-ssa-sprintf.c: Include domwalk.h.
10688         (class sprintf_dom_walker): New class, derived from dom_walker.
10689         (sprintf_dom_walker::before_dom_children): New function.
10690         (struct call_info): Moved into sprintf_dom_walker class
10691         (compute_formath_length, handle_gimple_call): Likewise.
10692         (sprintf_length::execute): Call the dominator walker rather
10693         than walking the statements.
10695         * tree-vrp.c (check_all_array_refs): Do not use wi->info to smuggle
10696         gimple statement locations.
10697         (check_array_bounds): Corresponding changes.  Get the statement's
10698         location directly from wi->stmt.
10700 2017-10-27  Palmer Dabbelt  <palmer@dabbelt.com>
10702         PR target/82717
10703         * doc/invoke.texi (RISC-V) <-mabi>: Correct and improve.
10705 2017-10-27  Jan Hubicka  <hubicka@ucw.cz>
10707         * config/i386/x86-tune.def (X86_TUNE_PARTIAL_REG_DEPENDENCY,
10708         X86_TUNE_MOVX): Disable for Haswell and newer CPUs.
10710 2017-10-27  Jakub Jelinek  <jakub@redhat.com>
10712         PR target/82703
10713         * config/i386/i386-protos.h (maybe_get_pool_constant): Removed.
10714         * config/i386/i386.c (maybe_get_pool_constant): Removed.
10715         (ix86_split_to_parts): Use avoid_constant_pool_reference instead of
10716         maybe_get_pool_constant.
10717         * config/i386/predicates.md (zero_extended_scalar_load_operand):
10718         Likewise.
10720 2017-10-27  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10722         * doc/install.texi (Specific, i?86-*-solaris2.10): Simplify gas
10723         2.26 caveat.  Update gas and gld versions.
10724         (Specific, *-*-solaris2*): Update binutils version.  Remove caveat
10725         reference.
10727 2017-10-27  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
10729         * cgraph.h (set_malloc_flag): Declare.
10730         * cgraph.c (set_malloc_flag_1): New function.
10731         (set_malloc_flag): Likewise.
10732         * ipa-fnsummary.h (ipa_call_summary): Add new field is_return_callee.
10733         * ipa-fnsummary.c (ipa_call_summary::reset): Set is_return_callee to
10734         false.
10735         (read_ipa_call_summary): Add support for reading is_return_callee.
10736         (write_ipa_call_summary): Stream is_return_callee.
10737         * ipa-inline.c (ipa_inline): Remove call to ipa_free_fn_summary.
10738         * ipa-pure-const.c: Add headers ssa.h, alloc-pool.h, symbol-summary.h,
10739         ipa-prop.h, ipa-fnsummary.h.
10740         (pure_const_names): Change to static.
10741         (malloc_state_e): Define.
10742         (malloc_state_names): Define.
10743         (funct_state_d): Add field malloc_state.
10744         (varying_state): Set malloc_state to STATE_MALLOC_BOTTOM.
10745         (check_retval_uses): New function.
10746         (malloc_candidate_p): Likewise.
10747         (analyze_function): Add support for malloc attribute.
10748         (pure_const_write_summary): Stream malloc_state.
10749         (pure_const_read_summary): Add support for reading malloc_state.
10750         (dump_malloc_lattice): New function.
10751         (propagate_malloc): New function.
10752         (warn_function_malloc): New function.
10753         (ipa_pure_const::execute): Call propagate_malloc and
10754         ipa_free_fn_summary.
10755         (pass_local_pure_const::execute): Add support for malloc attribute.
10756         * ssa-iterators.h (RETURN_FROM_IMM_USE_STMT): New macro.
10757         * doc/invoke.texi: Document Wsuggest-attribute=malloc.
10759 2017-10-27  Martin Liska  <mliska@suse.cz>
10761         PR gcov-profile/82457
10762         * doc/invoke.texi: Document that one needs a non-strict ISO mode
10763         for fork-like functions to be properly instrumented.
10765 2017-10-27  Richard Biener  <rguenther@suse.de>
10767         PR middle-end/81659
10768         * tree-eh.c (pass_lower_eh_dispatch::execute): Free dominator
10769         info when we redirected EH.
10771 2017-10-26  Michael Collison  <michael.collison@arm.com>
10773         * config/aarch64/aarch64.md(<optab>_trunc><vf><GPI:mode>2):
10774         New pattern.
10775         (<optab>_trunchf<GPI:mode>2: New pattern.
10776         (<optab>_trunc<vgp><GPI:mode>2: New pattern.
10777         * config/aarch64/iterators.md (wv): New mode attribute.
10778         (vf, VF): New mode attributes.
10779         (vgp, VGP): New mode attributes.
10780         (s): Update attribute with SImode and DImode prefixes.
10782 2017-10-26  Sandra Loosemore  <sandra@codesourcery.com>
10784         * config/nios2/constraints.md ("S"): Match r0rel_constant_p too.
10785         * config/nios2/nios2-protos.h (r0rel_constant_p): Declare.
10786         * config/nios2/nios2.c: (nios2_r0rel_sec_regex): New.
10787         (nios2_option_overide): Initialize it.  Don't allow R0-relative
10788         addressing with PIC.
10789         (nios2_rtx_costs): Handle r0rel_constant_p like gprel_constant_p.
10790         (nios2_symbolic_constant_p): Likewise.
10791         (nios2_legitimate_address_p): Likewise.
10792         (nios2_r0rel_section_name_p): New.
10793         (nios2_symbol_ref_in_r0rel_data_p): New.
10794         (nios2_emit_move_sequence): Handle r0rel_constant_p.
10795         (r0rel_constant_p): New.
10796         (nios2_print_operand_address): Handle r0rel_constant_p.
10797         (nios2_cdx_narrow_form_p): Likewise.
10798         * config/nios2/nios2.opt (mr0rel-sec=): New option.
10799         * doc/invoke.texi (Option Summary): Add -mr0rel-sec.
10800         (Nios II Options): Document -mr0rel-sec.
10802 2017-10-26  Sandra Loosemore  <sandra@codesourcery.com>
10804         * config/nios2/nios2.c: Include xregex.h.
10805         (nios2_gprel_sec_regex): New.
10806         (nios2_option_overide): Initialize it.  Don't allow GP-relative
10807         addressing with PIC.
10808         (nios2_small_section_name_p): Check for regex match.
10809         * config/nios2/nios2.opt (mgprel-sec=): New option.
10810         * doc/invoke.texi (Option Summary): Add -mgprel-sec.
10811         (Nios II Options): Document -mgprel-sec.
10813 2017-10-26  Jim Wilson  <wilson@tuliptree.org>
10815         * doc/invoke.texi (-fdebug-prefix-map): Expand documentation.
10817 2017-10-26  Tom de Vries  <tom@codesourcery.com>
10819         PR tree-optimization/82707
10820         * gimple.c (gimple_copy): Fix unsharing of
10821         GIMPLE_OMP_{SINGLE,TARGET,TEAMS}.
10823 2017-10-26  Olga Makhotina  <olga.makhotina@intel.com>
10825         * config/i386/avx512fintrin.h (_mm512_cmpeq_pd_mask,
10826         _mm512_cmple_pd_mask, _mm512_cmplt_pd_mask,
10827         _mm512_cmpneq_pd_mask, _mm512_cmpnle_pd_mask,
10828         _mm512_cmpnlt_pd_mask, _mm512_cmpord_pd_mask,
10829         _mm512_cmpunord_pd_mask, _mm512_mask_cmpeq_pd_mask,
10830         _mm512_mask_cmple_pd_mask, _mm512_mask_cmplt_pd_mask,
10831         _mm512_mask_cmpneq_pd_mask, _mm512_mask_cmpnle_pd_mask,
10832         _mm512_mask_cmpnlt_pd_mask, _mm512_mask_cmpord_pd_mask,
10833         _mm512_mask_cmpunord_pd_mask, _mm512_cmpeq_ps_mask,
10834         _mm512_cmple_ps_mask, _mm512_cmplt_ps_mask,
10835         _mm512_cmpneq_ps_mask, _mm512_cmpnle_ps_mask,
10836         _mm512_cmpnlt_ps_mask, _mm512_cmpord_ps_mask,
10837         _mm512_cmpunord_ps_mask, _mm512_mask_cmpeq_ps_mask,
10838         _mm512_mask_cmple_ps_mask, _mm512_mask_cmplt_ps_mask,
10839         _mm512_mask_cmpneq_ps_mask, _mm512_mask_cmpnle_ps_mask,
10840         _mm512_mask_cmpnlt_ps_mask, _mm512_mask_cmpord_ps_mask,
10841         _mm512_mask_cmpunord_ps_mask): New intrinsics.
10843 2017-10-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
10845         * config/rs6000/aix.h (TARGET_IEEEQUAD_DEFAULT): Set long double
10846         default to IBM.
10847         * config/rs6000/darwin.h (TARGET_IEEEQUAD_DEFAULT): Likewise.
10848         * config/rs6000/rs6000.opt (-mabi=ieeelongdouble): Move the
10849         warning to rs6000.c.  Remove the Undocumented flag, since it has
10850         been documented.
10851         (-mabi=ibmlongdouble): Likewise.
10852         * config/rs6000/rs6000.c (TARGET_IEEEQUAD_DEFAULT): If it is not
10853         already set, set the default format for long double.
10854         (rs6000_debug_reg_global): Print whether long double is IBM or
10855         IEEE.
10856         (rs6000_option_override_internal): Rework setting long double
10857         format.  Only warn if the user is changing the long double default
10858         and they did not use -Wno-psabi.
10859         * doc/invoke.texi (PowerPC options): Update the documentation for
10860         -mabi=ieeelongdouble and -mabi=ibmlongdouble.
10862 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10863             Alan Hayward  <alan.hayward@arm.com>
10864             David Sherwood  <david.sherwood@arm.com>
10866         * rtl.h (wider_subreg_mode): New function.
10867         * ira.h (ira_sort_regnos_for_alter_reg): Take a machine_mode *
10868         rather than an unsigned int *.
10869         * ira-color.c (regno_max_ref_width): Replace with...
10870         (regno_max_ref_mode): ...this new variable.
10871         (coalesced_pseudo_reg_slot_compare): Update accordingly.
10872         Use wider_subreg_mode.
10873         (ira_sort_regnos_for_alter_reg): Likewise.  Take a machine_mode *
10874         rather than an unsigned int *.
10875         * lra-constraints.c (uses_hard_regs_p): Use wider_subreg_mode.
10876         (process_alt_operands): Likewise.
10877         (invariant_p): Likewise.
10878         * lra-spills.c (assign_mem_slot): Likewise.
10879         (add_pseudo_to_slot): Likewise.
10880         * lra.c (collect_non_operand_hard_regs): Likewise.
10881         (add_regs_to_insn_regno_info): Likewise.
10882         * reload1.c (regno_max_ref_width): Replace with...
10883         (regno_max_ref_mode): ...this new variable.
10884         (reload): Update accordingly.  Update call to
10885         ira_sort_regnos_for_alter_reg.
10886         (alter_reg): Update to use regno_max_ref_mode.  Call wider_subreg_mode.
10887         (init_eliminable_invariants): Update to use regno_max_ref_mode.
10888         (scan_paradoxical_subregs): Likewise.
10890 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10892         * config/aarch64/aarch64.h (EXIT_IGNORE_STACK): Set if alloca is used.
10893         (aarch64_frame): Add emit_frame_chain boolean.
10894         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
10895         Move eh_return case to aarch64_layout_frame.
10896         (aarch64_layout_frame): Initialize emit_frame_chain.
10897         (aarch64_expand_prologue): Use emit_frame_chain.
10899 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10901         * config/aarch64/aarch64.c (aarch64_layout_frame):
10902         Ensure LR is always stored at the bottom of the callee-saves.
10903         Remove rarely used frame layout which saves callee-saves at top of
10904         frame, so the store of LR can be used as a valid probe in all cases.
10906 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10908         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10909         Improve unaligned TImode/TFmode base/offset split.
10911 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10912             Alan Hayward  <alan.hayward@arm.com>
10913             David Sherwood  <david.sherwood@arm.com>
10915         * caller-save.c (mark_referenced_regs):  Use read_modify_subreg_p.
10916         * combine.c (find_single_use_1): Likewise.
10917         (expand_field_assignment): Likewise.
10918         (move_deaths): Likewise.
10919         * lra-constraints.c (simplify_operand_subreg): Likewise.
10920         (curr_insn_transform): Likewise.
10921         * lra.c (collect_non_operand_hard_regs): Likewise.
10922         (add_regs_to_insn_regno_info): Likewise.
10923         * rtlanal.c (reg_referenced_p): Likewise.
10924         (covers_regno_no_parallel_p): Likewise.
10926 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10928         * wide-int-print.cc (print_hex): Loop based on extract_uhwi.
10929         Don't print any bits outside the precision of the value.
10930         * wide-int.cc (test_printing): Add some new tests.
10932 2017-10-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10934         * configure.ac (gcc_cv_as_ix86_xbrace_comment): Check if assembler
10935         supports -xbrace_comment option.
10936         * configure: Regenerate.
10937         * config.in: Regenerate.
10938         * config/i386/sol2.h (ASM_XBRACE_COMMENT_SPEC): Define.
10939         (ASM_CPU_SPEC): Use it.
10941 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10943         * target.def (static_rtx_alignment): New hook.
10944         * targhooks.h (default_static_rtx_alignment): Declare.
10945         * targhooks.c (default_static_rtx_alignment): New function.
10946         * doc/tm.texi.in (TARGET_STATIC_RTX_ALIGNMENT): New hook.
10947         * doc/tm.texi: Regenerate.
10948         * varasm.c (force_const_mem): Use targetm.static_rtx_alignment
10949         instead of targetm.constant_alignment.  Remove call to
10950         set_mem_attributes.
10951         * config/cris/cris.c (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10952         (cris_preferred_mininum_alignment): New function, split out from...
10953         (cris_constant_alignment): ...here.
10954         (cris_static_rtx_alignment): New function.
10955         * config/i386/i386.c (ix86_static_rtx_alignment): New function,
10956         split out from...
10957         (ix86_constant_alignment): ...here.
10958         (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10959         * config/mmix/mmix.c (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10960         (mmix_static_rtx_alignment): New function.
10961         * config/spu/spu.c (spu_static_rtx_alignment): New function.
10962         (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10964 2017-10-26  Tamar Christina  <tamar.christina@arm.com>
10966         PR target/81800
10967         * config/aarch64/aarch64.md (lrint<GPF:mode><GPI:mode>2):
10968         Add flag_trapping_math and flag_fp_int_builtin_inexact.
10970 2017-10-25  Palmer Dabbelt  <palmer@dabbelt.com>
10972         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Define.
10973         * config/riscv/pic.md (local_pic_load): Rename to local_pic_load_s,
10974         mark as a sign-extending load.
10975         (local_pic_load_u): Define.
10977 2017-10-25  Eric Botcazou  <ebotcazou@adacore.com>
10979         PR middle-end/82062
10980         * fold-const.c (operand_equal_for_comparison_p): Also return true
10981         if ARG0 is a simple variant of ARG1 with narrower precision.
10982         (fold_ternary_loc): Always pass unstripped operands to the predicate.
10984 2017-10-25  Jan Hubicka  <hubicka@ucw.cz>
10986         * i386.c (ix86_builtin_vectorization_cost): Compute scatter/gather
10987         cost correctly.
10988         * i386.h (processor_costs): Add gather_static, gather_per_elt,
10989         scatter_static, scatter_per_elt.
10990         * x86-tune-costs.h: Add new cost entries.
10992 2017-10-25  Richard Biener  <rguenther@suse.de>
10994         * tree-ssa-sccvn.h (vn_eliminate): Declare.
10995         * tree-ssa-pre.c (class eliminate_dom_walker, eliminate,
10996         class pass_fre): Move to ...
10997         * tree-ssa-sccvn.c (class eliminate_dom_walker, vn_eliminate,
10998         class pass_fre): ... here and adjust for statistics.
11000 2017-10-25  Jakub Jelinek  <jakub@redhat.com>
11002         PR libstdc++/81706
11003         * attribs.c (attribute_value_equal): Use omp_declare_simd_clauses_equal
11004         for comparison of OMP_CLAUSEs regardless of flag_openmp{,_simd}.
11005         (duplicate_one_attribute, copy_attributes_to_builtin): New functions.
11006         * attribs.h (duplicate_one_attribute, copy_attributes_to_builtin): New
11007         declarations.
11009 2017-10-25  Richard Biener  <rguenther@suse.de>
11011         * tree-ssa-pre.c (need_eh_cleanup, need_ab_cleanup, el_to_remove,
11012         el_to_fixup, el_todo, el_avail, el_avail_stack, eliminate_avail,
11013         eliminate_push_avail, eliminate_insert): Move inside...
11014         (class eliminate_dom_walker): ... this class in preparation
11015         of move.
11016         (fini_eliminate): Remove by merging with ...
11017         (eliminate): ... this function.  Adjust for class changes.
11018         (pass_pre::execute): Remove fini_eliminate call.
11019         (pass_fre::execute): Likewise.
11021 2017-10-24  Jakub Jelinek  <jakub@redhat.com>
11023         PR target/82460
11024         * config/i386/sse.md (UNSPEC_VPERMI2, UNSPEC_VPERMI2_MASK): Remove.
11025         (VPERMI2, VPERMI2I): New mode iterators.
11026         (<avx512>_vpermi2var<mode>3_maskz): Remove 3 define_expand patterns.
11027         (<avx512>_vpermi2var<mode>3<sd_maskz_name>): Remove 3 define_insn
11028         patterns.
11029         (<avx512>_vpermi2var<mode>3_mask): New define_expand using VPERMI2
11030         mode iterator.  Remove 3 old define_insn patterns.
11031         (*<avx512>_vpermi2var<mode>3_mask): 2 new define_insn patterns.
11032         (<avx512>_vpermt2var<mode>3_maskz): Adjust 1 define_expand to use
11033         VPERMI2 mode iterator, remove the other two expanders.
11034         (<avx512>_vpermt2var<mode>3<sd_maskz_name>): Adjust 1 define_insn
11035         to use VPERMI2 mode iterator, add another alternative for vpermi2*
11036         instructions, remove the other two patterns.
11037         (<avx512>_vpermt2var<mode>3_mask): Adjust 1 define_insn to use VPERMI2
11038         mode iterator, remove the other two patterns.
11039         * config/i386/i386.c (ix86_expand_vec_perm_vpermi2): Renamed to ...
11040         (ix86_expand_vec_perm_vpermt2): ... this.  Swap mask and op0
11041         arguments, use gen_*vpermt2* expanders instead of gen_*vpermi2*
11042         and adjust argument order accordingly.
11043         (ix86_expand_vec_perm): Adjust caller.
11044         (expand_vec_perm_1): Likewise.
11045         (expand_vec_perm_vpermi2_vpshub2): Rename to ...
11046         (expand_vec_perm_vpermt2_vpshub2): ... this.
11047         (ix86_expand_vec_perm_const_1): Adjust caller.
11048         (ix86_vectorize_vec_perm_const_ok): Adjust comments.
11050         PR target/82370
11051         * config/i386/sse.md (VIMAX_AVX2): Remove V4TImode.
11052         (VIMAX_AVX2_AVX512BW, VIMAX_AVX512VL): New mode iterators.
11053         (vec_shl_<mode>): Remove unused expander.
11054         (avx512bw_<shift_insn><mode>3): New define_insn.
11055         (<sse2_avx2>_ashl<mode>3, <sse2_avx2>_lshr<mode>3): Replaced by ...
11056         (<sse2_avx2>_<shift_insn><mode>3): ... this.  New define_insn.
11058 2017-10-24  Paolo Carlini  <paolo.carlini@oracle.com>
11060         PR c++/82466
11061         * doc/invoke.texi ([Wbuiltin-declaration-mismatch]): Extend
11062         description.
11064 2017-10-24  Wilco Dijkstra  <wdijkstr@arm.com>
11066         PR rtl-optimization/82396
11067         * gcc/haifa-sched.c (ready_sort_real): Remove qsort workaround.
11068         (autopref_multipass_init): Simplify initialization.
11069         (autopref_rank_data): Simplify sort order.
11070         * gcc/sched-int.h (autopref_multipass_data_): Remove
11071         multi_mem_insn_p, min_offset and max_offset.
11073 2017-10-24  Wilco Dijkstra  <wdijkstr@arm.com>
11075         PR middle-end/60580
11076         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
11077         Check special value of flag_omit_frame_pointer.
11078         (aarch64_can_eliminate): Likewise.
11079         (aarch64_override_options_after_change_1): Simplify handling of
11080         -fomit-frame-pointer and -fomit-leaf-frame-pointer.
11082 2017-10-24  Richard Biener  <rguenther@suse.de>
11084         PR tree-optimization/82697
11085         * tree-ssa-phiopt.c (cond_store_replacement): Use alias-set
11086         zero for conditional load and unconditional store.
11088 2017-10-24  H.J. Lu  <hongjiu.lu@intel.com>
11090         * doc/install.texi: Document bootstrap-cet.
11092 2017-10-24  H.J. Lu  <hongjiu.lu@intel.com>
11094         PR target/82659
11095         * config/i386/i386.c (rest_of_insert_endbranch): Don't insert
11096         ENDBR instruction at function entrance if function is only
11097         called directly.
11099 2017-10-24  Jakub Jelinek  <jakub@redhat.com>
11101         PR target/82628
11102         * config/i386/i386.md (addcarry<mode>, subborrow<mode>): Change
11103         patterns to better describe from which operation the CF is computed.
11104         (addcarry<mode>_0, subborrow<mode>_0): New patterns.
11105         * config/i386/i386.c (ix86_expand_builtin) <case handlecarry>: Pass
11106         one LTU with [DT]Imode and another one with [SD]Imode.  If arg0
11107         is 0, use _0 suffixed expanders instead of emitting a comparison
11108         before it.
11110 2017-10-06  Sergey Shalnov  <Sergey.Shalnov@intel.com>
11112         * config/i386/i386.md(*movsf_internal, *movdf_internal):
11113         Avoid 512-bit AVX modes for TARGET_PREFER_AVX256.
11115 2017-10-24  Eric Botcazou  <ebotcazou@adacore.com>
11117         PR middle-end/82569
11118         * tree-outof-ssa.h (always_initialized_rtx_for_ssa_name_p): Delete.
11119         * expr.c (expand_expr_real_1) <expand_decl_rtl>: Revert latest change.
11120         * loop-iv.c (iv_get_reaching_def): Likewise.
11121         * cfgexpand.c (expand_one_ssa_partition): Initialize the RTX if the
11122         variable is promoted and the partition contains undefined values.
11124 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
11126         * config/nios2/nios2.c (nios2_rtx_costs): Make costs better
11127         reflect reality.
11128         (nios2_address_cost): Define.
11129         (nios2_legitimize_address): Recognize (exp + constant) directly.
11130         (TARGET_ADDRESS_COST): Define.
11132 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
11134         * config/nios2/nios2-protos.h (nios2_large_constant_p): Declare.
11135         (nios2_symbolic_memory_operand_p): Declare.
11136         (nios2_split_large_constant): Declare.
11137         (nios2_split_symbolic_memory_operand): Declare.
11138         * config/nios2/nios2.c: Adjust includes.
11139         (nios2_symbolic_constant_allowed): New.
11140         (nios2_symbolic_constant_p): New.
11141         (nios2_plus_symbolic_constant_p): New.
11142         (nios2_valid_addr_expr_p): Recognize addresses involving
11143         symbolic constants.
11144         (nios2_legitimate_address_p): Likewise, also LO_SUM.
11145         (nios2_symbolic_memory_operand_p): New.
11146         (nios2_large_constant_p): New.
11147         (nios2_split_large_constant): New.
11148         (nios2_split_plus_large_constant): New.
11149         (nios2_split_symbolic_memory_operand): New.
11150         (nios2_legitimize_address): Code refactoring.  Handle addresses
11151         involving symbolic constants.
11152         (nios2_emit_move_sequence): Likewise.
11153         (nios2_print_operand): Improve error output.
11154         (nios2_print_operand_address): Handle LO_SUM.
11155         (nios2_cdx_narrow_form_p): Likewise.
11156         * config/nios2/nios2.md (movqi_internal): Add splitter for memory
11157         operands involving symbolic constants.
11158         (movhi_internal, movsi_internal): Likewise.
11159         (zero_extendhisi2, zero_extendqi<mode>2): Likewise.
11160         (extendhisi2, extendqi<mode>2): Likewise.
11162 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
11164         * tree-pass.h (PROP_rtl_split_insns): Define.
11165         * recog.c (pass_data_split_all_insns): Provide PROP_rtl_split_insns.
11167 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
11169         * config/nios2/nios2.c (TARGET_LRA_P): Don't override.
11171 2017-10-23  Jakub Jelinek  <jakub@redhat.com>
11173         PR debug/82630
11174         * target.def (const_not_ok_for_debug_p): Default to
11175         default_const_not_ok_for_debug_p instead of hook_bool_rtx_false.
11176         * targhooks.h (default_const_not_ok_for_debug_p): New declaration.
11177         * targhooks.c (default_const_not_ok_for_debug_p): New function.
11178         * dwarf2out.c (const_ok_for_output_1): Only reject UNSPECs for
11179         which targetm.const_not_ok_for_debug_p returned true.
11180         * config/arm/arm.c (arm_const_not_ok_for_debug_p): Return true
11181         for UNSPECs.
11182         * config/powerpcspe/powerpcspe.c (rs6000_const_not_ok_for_debug_p):
11183         Likewise.
11184         * config/rs6000/rs6000.c (rs6000_const_not_ok_for_debug_p): Likewise.
11185         * config/i386/i386.c (ix86_delegitimize_address_1): Don't delegitimize
11186         UNSPEC_GOTOFF with addend into addend - _GLOBAL_OFFSET_TABLE_ + symbol
11187         if !base_term_p.
11188         (ix86_const_not_ok_for_debug_p): New function.
11189         (i386_asm_output_addr_const_extra): Handle UNSPEC_GOTOFF.
11190         (TARGET_CONST_NOT_OK_FOR_DEBUG_P): Redefine.
11192 2017-10-23  David Malcolm  <dmalcolm@redhat.com>
11194         PR bootstrap/82610
11195         * system.h: Conditionally include "unique-ptr.h" if
11196         INCLUDE_UNIQUE_PTR is defined.
11197         * unique-ptr-tests.cc: Remove include of "unique-ptr.h" in favor
11198         of defining INCLUDE_UNIQUE_PTR before including "system.h".
11200 2017-10-23  Sebastian Perta  <sebastian.perta@renesas.com>
11202         * config/rl78/rl78.md: New define_expand "subdi3".
11204 2017-10-23  H.J. Lu  <hongjiu.lu@intel.com>
11206         PR target/82673
11207         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Skip
11208         DF_REF_INSN if DF_REF_INSN_INFO is false.
11210 2017-10-23  Jan Hubicka  <hubicka@ucw.cz>
11212         * i386.c (dimode_scalar_chain::compute_convert_gain): Use
11213         xmm_move instead of sse_move.
11214         (sse_store_index): New function.
11215         (ix86_register_move_cost): Be more sensible about mismatch stall;
11216         model AVX moves correctly; make difference between sse->integer and
11217         integer->sse.
11218         (ix86_builtin_vectorization_cost): Model correctly aligned and unaligned
11219         moves; make difference between SSE and AVX.
11220         * i386.h (processor_costs): Remove sse_move; add xmm_move, ymm_move
11221         and zmm_move. Increase size of sse load and store tables;
11222         add unaligned load and store tables; add ssemmx_to_integer.
11223         * x86-tune-costs.h: Update all entries according to real
11224         move latencies from Agner Fog's manual and chip documentation.
11226 2017-10-23  Jakub Jelinek  <jakub@redhat.com>
11228         PR target/82628
11229         * config/i386/predicates.md (x86_64_dwzext_immediate_operand): New.
11230         * config/i386/constraints.md (Wf): New constraint.
11231         * config/i386/i386.md (UNSPEC_SBB): New unspec.
11232         (cmp<dwi>_doubleword): Removed.
11233         (sub<mode>3_carry_ccc, *sub<mode>3_carry_ccc_1): New patterns.
11234         (sub<mode>3_carry_ccgz): Use unspec instead of compare.
11235         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>: Don't
11236         expand with cmp<dwi>_doubleword.  For LTU and GEU use
11237         sub<mode>3_carry_ccc instead of sub<mode>3_carry_ccgz and use CCCmode.
11239         * common.opt (gcolumn-info): Enable by default.
11240         * doc/invoke.texi (gcolumn-info): Document new default.
11242 2017-10-23  Richard Biener  <rguenther@suse.de>
11244         PR tree-optimization/82672
11245         * graphite-isl-ast-to-gimple.c (graphite_copy_stmts_from_block):
11246         Fold the stmt if we propagated into it.
11248 2017-10-23  Richard Biener  <rguenther@suse.de>
11250         * tree-ssa-pre.c (bitmap_remove_from_set): Rename to...
11251         (bitmap_remove_expr_from_set): ... this.  All callers call this
11252         for non-constant values.
11253         (bitmap_set_subtract): Rename to...
11254         (bitmap_set_subtract_expressions): ... this.  Adjust and
11255         optimize.
11256         (bitmap_set_contains_value): Remove superfluous check.
11257         (bitmap_set_replace_value): Inline into single caller ...
11258         (bitmap_value_replace_in_set): ... here and simplify.
11259         (dependent_clean): Merge into ...
11260         (clean): ... this using an overload.  Adjust.
11261         (prune_clobbered_mems): Adjust.
11262         (compute_antic_aux): Likewise.
11263         (compute_partial_antic_aux): Likewise.
11265 2017-10-23  Richard Biener  <rguenther@suse.de>
11267         PR tree-optimization/82129
11268         Revert
11269         2017-08-01  Richard Biener  <rguenther@suse.de>
11271         PR tree-optimization/81181
11272         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
11273         (compute_antic): ... end of iteration here.
11275 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11277         * target.def (starting_frame_offset): New hook.
11278         * doc/tm.texi (STARTING_FRAME_OFFSET): Remove in favor of...
11279         (TARGET_STARTING_FRAME_OFFSET): ...this new hook.
11280         * doc/tm.texi.in: Regenerate.
11281         * hooks.h (hook_hwi_void_0): Declare.
11282         * hooks.c (hook_hwi_void_0): New function.
11283         * doc/rtl.texi: Refer to TARGET_STARTING_FRAME_OFFSET instead of
11284         STARTING_FRAME_OFFSET.
11285         * builtins.c (expand_builtin_setjmp_receiver): Likewise.
11286         * reload1.c (reload): Likewise.
11287         * cfgexpand.c (expand_used_vars): Use targetm.starting_frame_offset
11288         instead of STARTING_FRAME_OFFSET.
11289         * function.c (try_fit_stack_local): Likewise.
11290         (assign_stack_local_1): Likewise
11291         (instantiate_virtual_regs): Likewise.
11292         * rtlanal.c (rtx_addr_can_trap_p_1): Likewise.
11293         * config/avr/avr.md (nonlocal_goto_receiver): Likewise.
11294         * config/aarch64/aarch64.h (STARTING_FRAME_OFFSET): Delete.
11295         * config/alpha/alpha.h (STARTING_FRAME_OFFSET): Likewise.
11296         * config/arc/arc.h (STARTING_FRAME_OFFSET): Likewise.
11297         * config/arm/arm.h (STARTING_FRAME_OFFSET): Likewise.
11298         * config/bfin/bfin.h (STARTING_FRAME_OFFSET): Likewise.
11299         * config/c6x/c6x.h (STARTING_FRAME_OFFSET): Likewise.
11300         * config/cr16/cr16.h (STARTING_FRAME_OFFSET): Likewise.
11301         * config/cris/cris.h (STARTING_FRAME_OFFSET): Likewise.
11302         * config/fr30/fr30.h (STARTING_FRAME_OFFSET): Likewise.
11303         * config/frv/frv.h (STARTING_FRAME_OFFSET): Likewise.
11304         * config/ft32/ft32.h (STARTING_FRAME_OFFSET): Likewise.
11305         * config/h8300/h8300.h (STARTING_FRAME_OFFSET): Likewise.
11306         * config/i386/i386.h (STARTING_FRAME_OFFSET): Likewise.
11307         * config/ia64/ia64.h (STARTING_FRAME_OFFSET): Likewise.
11308         * config/m32c/m32c.h (STARTING_FRAME_OFFSET): Likewise.
11309         * config/m68k/m68k.h (STARTING_FRAME_OFFSET): Likewise.
11310         * config/mcore/mcore.h (STARTING_FRAME_OFFSET): Likewise.
11311         * config/mn10300/mn10300.h (STARTING_FRAME_OFFSET): Likewise.
11312         * config/moxie/moxie.h (STARTING_FRAME_OFFSET): Likewise.
11313         * config/msp430/msp430.h (STARTING_FRAME_OFFSET): Likewise.
11314         * config/nds32/nds32.h (STARTING_FRAME_OFFSET): Likewise.
11315         * config/nios2/nios2.h (STARTING_FRAME_OFFSET): Likewise.
11316         * config/nvptx/nvptx.h (STARTING_FRAME_OFFSET): Likewise.
11317         * config/pdp11/pdp11.h (STARTING_FRAME_OFFSET): Likewise.
11318         * config/riscv/riscv.h (STARTING_FRAME_OFFSET): Likewise.
11319         * config/rl78/rl78.h (STARTING_FRAME_OFFSET): Likewise.
11320         * config/rx/rx.h (STARTING_FRAME_OFFSET): Likewise.
11321         * config/s390/s390.h (STARTING_FRAME_OFFSET): Likewise.
11322         * config/sh/sh.h (STARTING_FRAME_OFFSET): Likewise.
11323         * config/sparc/sparc.c (sparc_compute_frame_size): Likewise.
11324         * config/sparc/sparc.h (STARTING_FRAME_OFFSET): Likewise.
11325         * config/spu/spu.h (STARTING_FRAME_OFFSET): Likewise.
11326         * config/stormy16/stormy16.h (STARTING_FRAME_OFFSET): Likewise.
11327         * config/tilegx/tilegx.h (STARTING_FRAME_OFFSET): Likewise.
11328         * config/tilepro/tilepro.h (STARTING_FRAME_OFFSET): Likewise.
11329         * config/v850/v850.h (STARTING_FRAME_OFFSET): Likewise.
11330         * config/visium/visium.h (STARTING_FRAME_OFFSET): Likewise.
11331         * config/avr/avr.h (STARTING_FRAME_OFFSET): Likewise.
11332         * config/avr/avr-protos.h (avr_starting_frame_offset): Likewise.
11333         * config/avr/avr.c (avr_starting_frame_offset): Make static and
11334         return a HOST_WIDE_INT.
11335         (avr_builtin_setjmp_frame_value): Use it instead of
11336         STARTING_FRAME_OFFSET.
11337         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11338         * config/epiphany/epiphany.h (STARTING_FRAME_OFFSET): Delete.
11339         * config/epiphany/epiphany.c (epiphany_starting_frame_offset):
11340         New function.
11341         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11342         * config/iq2000/iq2000.h (STARTING_FRAME_OFFSET): Delete.
11343         * config/iq2000/iq2000.c (iq2000_starting_frame_offset): New function.
11344         (TARGET_CONSTANT_ALIGNMENT): Redefine.
11345         * config/lm32/lm32.h (STARTING_FRAME_OFFSET): Delete.
11346         * config/lm32/lm32.c (lm32_starting_frame_offset): New function.
11347         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11348         * config/m32r/m32r.h (STARTING_FRAME_OFFSET): Delete.
11349         * config/m32r/m32r.c (m32r_starting_frame_offset): New function.
11350         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11351         * config/microblaze/microblaze.h (STARTING_FRAME_OFFSET): Delete.
11352         * config/microblaze/microblaze.c (microblaze_starting_frame_offset):
11353         New function.
11354         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11355         * config/mips/mips.h (STARTING_FRAME_OFFSET): Delete.
11356         * config/mips/mips.c (mips_compute_frame_info): Refer to
11357         TARGET_STARTING_FRAME_OFFSET instead of STARTING_FRAME_OFFSET.
11358         (mips_starting_frame_offset): New function.
11359         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11360         * config/mmix/mmix.h (STARTING_FRAME_OFFSET): Delete.
11361         * config/mmix/mmix-protos.h (mmix_starting_frame_offset): Delete.
11362         * config/mmix/mmix.c (mmix_starting_frame_offset): Make static
11363         and return a HOST_WIDE_INT.
11364         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11365         (mmix_initial_elimination_offset): Refer to
11366         TARGET_STARTING_FRAME_OFFSET instead of STARTING_FRAME_OFFSET.
11367         * config/pa/pa.h (STARTING_FRAME_OFFSET): Delete.
11368         * config/pa/pa.c (pa_starting_frame_offset): New function.
11369         (pa_compute_frame_size): Use it instead of STARTING_FRAME_OFFSET.
11370         (pa_expand_prologue): Likewise.
11371         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11372         * config/powerpcspe/aix.h (STARTING_FRAME_OFFSET): Split out
11373         !FRAME_GROWS_DOWNWARD handling to...
11374         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11375         * config/powerpcspe/darwin.h (STARTING_FRAME_OFFSET): Split out
11376         !FRAME_GROWS_DOWNWARD handling to...
11377         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11378         * config/powerpcspe/powerpcspe.h (STARTING_FRAME_OFFSET): Split out
11379         !FRAME_GROWS_DOWNWARD handling to...
11380         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11381         * config/powerpcspe/powerpcspe.c (TARGET_STARTING_FRAME_OFFSET):
11382         Redefine.
11383         (rs6000_starting_frame_offset): New function.
11384         * config/rs6000/aix.h (STARTING_FRAME_OFFSET): Split out
11385         !FRAME_GROWS_DOWNWARD handling to...
11386         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11387         * config/rs6000/darwin.h (STARTING_FRAME_OFFSET): Split out
11388         !FRAME_GROWS_DOWNWARD handling to...
11389         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11390         * config/rs6000/rs6000.h (STARTING_FRAME_OFFSET): Split out
11391         !FRAME_GROWS_DOWNWARD handling to...
11392         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11393         * config/rs6000/rs6000.c (TARGET_STARTING_FRAME_OFFSET): Refine.
11394         (rs6000_starting_frame_offset): New function.
11395         * config/vax/elf.h (STARTING_FRAME_OFFSET): Delete.
11396         * config/vax/vax.h (STARTING_FRAME_OFFSET): Delete.
11397         * config/vax/vax.c (vax_starting_frame_offset): New function.
11398         (vax_expand_prologue): Use it instead of STARTING_FRAME_OFFSET.
11399         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11400         * config/xtensa/xtensa.h (STARTING_FRAME_OFFSET): Delete.
11401         * config/xtensa/xtensa.c (xtensa_starting_frame_offset): New function.
11402         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11403         * system.h (STARTING_FRAME_OFFSET): Poison.
11405 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11407         * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
11408         SCALAR_TYPE_MODE instead of TYPE_MODE.
11410 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11411             Alan Hayward  <alan.hayward@arm.com>
11412             David Sherwood  <david.sherwood@arm.com>
11414         * dwarf2out.c (loc_list_from_tree_1): Use SCALAR_INT_TYPE_MODE
11416 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11417             Alan Hayward  <alan.hayward@arm.com>
11418             David Sherwood  <david.sherwood@arm.com>
11420         * expmed.c (expand_shift_1): Use scalar_mode for scalar_mode.
11422 2017-10-23  Richard Biener  <rguenther@suse.de>
11424         PR tree-optimization/82129
11425         * tree-ssa-pre.c (bitmap_set_and): Remove.
11426         (compute_antic_aux): Compute ANTIC_OUT intersection in a way
11427         canonicalizing expressions in the set to those with lowest
11428         ID rather than taking that from the first edge.
11430 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11432         * combine.c (rtx_equal_for_field_assignment_p): Use
11433         byte_lowpart_offset.
11435 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11436             Alan Hayward  <alan.hayward@arm.com>
11437             David Sherwood  <david.sherwood@arm.com>
11439         * internal-fn.c (expand_direct_optab_fn): Don't assign directly
11440         to a SUBREG_PROMOTED_VAR.
11442 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11443             Alan Hayward  <alan.hayward@arm.com>
11444             David Sherwood  <david.sherwood@arm.com>
11446         * cfgexpand.c (expand_debug_expr): Use GET_MODE_UNIT_PRECISION.
11447         (expand_debug_source_expr): Likewise.
11448         * combine.c (combine_simplify_rtx): Likewise.
11449         * cse.c (fold_rtx): Likewise.
11450         * optabs.c (expand_float): Likewise.
11451         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11452         (simplify_binary_operation_1): Likewise.
11454 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11455             Alan Hayward  <alan.hayward@arm.com>
11456             David Sherwood  <david.sherwood@arm.com>
11458         * combine.c (simplify_comparison): Use HWI_COMPUTABLE_MODE_P.
11459         (record_promoted_value): Likewise.
11460         * expr.c (expand_expr_real_2): Likewise.
11461         * ree.c (update_reg_equal_equiv_notes): Likewise.
11462         (combine_set_extension): Likewise.
11463         * rtlanal.c (low_bitmask_len): Likewise.
11464         * simplify-rtx.c (neg_const_int): Likewise.
11465         (simplify_binary_operation_1): Likewise.
11467 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11468             Alan Hayward  <alan.hayward@arm.com>
11469             David Sherwood  <david.sherwood@arm.com>
11471         * lra-spills.c (assign_mem_slot): Use subreg_size_lowpart_offset.
11472         * regcprop.c (maybe_mode_change): Likewise.
11473         * reload1.c (alter_reg): Likewise.
11475 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11477         * inchash.h (inchash::hash::add_wide_int): New function.
11478         * lto-streamer-out.c (hash_tree): Use it.
11480 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11482         * inchash.h (inchash::hash::add_wide_int): Rename to...
11483         (inchash::hash::add_hwi): ...this.
11484         * ipa-devirt.c (hash_odr_vtable): Update accordingly.
11485         (polymorphic_call_target_hasher::hash): Likewise.
11486         * ipa-icf.c (sem_function::get_hash, sem_function::init): Likewise.
11487         (sem_item::add_expr, sem_item::add_type, sem_variable::get_hash)
11488         (sem_item_optimizer::update_hash_by_addr_refs): Likewise.
11489         * lto-streamer-out.c (hash_tree): Likewise.
11490         * optc-save-gen.awk: Likewise.
11491         * tree.c (add_expr): Likewise.
11493 2017-10-22  Uros Bizjak  <ubizjak@gmail.com>
11495         PR target/52451
11496         * config/i386/i386.c (ix86_fp_compare_mode): Return CCFPmode
11497         for ordered inequality comparisons even with TARGET_IEEE_FP.
11499 2017-10-22  Uros Bizjak  <ubizjak@gmail.com>
11501         PR target/82628
11502         * config/i386/i386.md (cmp<dwi>_doubleword): New pattern.
11503         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>:
11504         Expand with cmp<dwi>_doubleword.
11506 2017-10-21  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11508         * extend.texi: Add x86 specific to 'nocf_check' attribute.
11509         List CET intrinsics.
11510         * invoke.texi: Add -mcet, -mibt, -mshstk options.  Add x86
11511         specific to -fcf-protection option.
11513 2017-10-21  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11515         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET): New.
11516         (OPTION_MASK_ISA_SHSTK_SET): Likewise.
11517         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
11518         (OPTION_MASK_ISA_SHSTK_UNSET): Likewise.
11519         (ix86_handle_option): Add -mibt, -mshstk, -mcet handling.
11520         * config.gcc (extra_headers): Add cetintrin.h for x86 targets.
11521         (extra_objs): Add cet.o for Linux/x86 targets.
11522         (tmake_file): Add i386/t-cet for Linux/x86 targets.
11523         * config/i386/cet.c: New file.
11524         * config/i386/cetintrin.h: Likewise.
11525         * config/i386/t-cet: Likewise.
11526         * config/i386/cpuid.h (bit_SHSTK): New.
11527         (bit_IBT): Likewise.
11528         * config/i386/driver-i386.c (host_detect_local_cpu): Detect and
11529         pass IBT and SHSTK bits.
11530         * config/i386/i386-builtin-types.def
11531         (VOID_FTYPE_UNSIGNED_PVOID): New.
11532         (VOID_FTYPE_UINT64_PVOID): Likewise.
11533         * config/i386/i386-builtin.def: Add CET intrinsics.
11534         * config/i386/i386-c.c (ix86_target_macros_internal): Add
11535         OPTION_MASK_ISA_IBT, OPTION_MASK_ISA_SHSTK handling.
11536         * config/i386/i386-passes.def: Add pass_insert_endbranch pass.
11537         * config/i386/i386-protos.h (make_pass_insert_endbranch): New
11538         prototype.
11539         * config/i386/i386.c (rest_of_insert_endbranch): New.
11540         (pass_data_insert_endbranch): Likewise.
11541         (pass_insert_endbranch): Likewise.
11542         (make_pass_insert_endbranch): Likewise.
11543         (ix86_notrack_prefixed_insn_p): Likewise.
11544         (ix86_target_string): Add -mibt, -mshstk flags.
11545         (ix86_option_override_internal): Add flag_cf_protection
11546         processing.
11547         (ix86_valid_target_attribute_inner_p): Set OPT_mibt, OPT_mshstk.
11548         (ix86_print_operand): Add 'notrack' prefix output.
11549         (ix86_init_mmx_sse_builtins): Add CET intrinsics.
11550         (ix86_expand_builtin): Expand CET intrinsics.
11551         (x86_output_mi_thunk): Add 'endbranch' instruction.
11552         * config/i386/i386.h (TARGET_IBT): New.
11553         (TARGET_IBT_P): Likewise.
11554         (TARGET_SHSTK): Likewise.
11555         (TARGET_SHSTK_P): Likewise.
11556         * config/i386/i386.md (unspecv): Add UNSPECV_NOP_RDSSP,
11557         UNSPECV_INCSSP, UNSPECV_SAVEPREVSSP, UNSPECV_RSTORSSP,
11558         UNSPECV_WRSS, UNSPECV_WRUSS, UNSPECV_SETSSBSY, UNSPECV_CLRSSBSY.
11559         (builtin_setjmp_setup): New pattern.
11560         (builtin_longjmp): Likewise.
11561         (rdssp<mode>): Likewise.
11562         (incssp<mode>): Likewise.
11563         (saveprevssp): Likewise.
11564         (rstorssp): Likewise.
11565         (wrss<mode>): Likewise.
11566         (wruss<mode>): Likewise.
11567         (setssbsy): Likewise.
11568         (clrssbsy): Likewise.
11569         (nop_endbr): Likewise.
11570         * config/i386/i386.opt: Add -mcet, -mibt, -mshstk and -mcet-switch
11571         options.
11572         * config/i386/immintrin.h: Include <cetintrin.h>.
11573         * config/i386/linux-common.h
11574         (file_end_indicate_exec_stack_and_cet): New prototype.
11575         (TARGET_ASM_FILE_END): New.
11577 2017-10-20  Jan Hubicka  <hubicka@ucw.cz>
11579         * i386.c (ix86_builtin_vectorization_cost): Use existing rtx_cost
11580         latencies instead of having separate table; make difference between
11581         integer and float costs.
11582         * i386.h (processor_costs): Remove scalar_stmt_cost,
11583         scalar_load_cost, scalar_store_cost, vec_stmt_cost, vec_to_scalar_cost,
11584         scalar_to_vec_cost, vec_align_load_cost, vec_unalign_load_cost,
11585         vec_store_cost.
11586         * x86-tune-costs.h: Remove entries which has been removed in
11587         procesor_costs from all tables; make cond_taken_branch_cost
11588         and cond_not_taken_branch_cost COST_N_INSNS based.
11590 2017-10-20  Jan Hubicka  <hubicka@ucw.cz>
11592         * x86-tune-costs.h (intel_cost, generic_cost): Fix move costs.
11594 2017-10-20  Jakub Jelinek  <jakub@redhat.com>
11596         * config/i386/i386.md (isa): Remove fma_avx512f.
11597         * config/i386/sse.md (<avx512>_fmadd_<mode>_mask<round_name>,
11598         <avx512>_fmadd_<mode>_mask3<round_name>,
11599         <avx512>_fmsub_<mode>_mask<round_name>,
11600         <avx512>_fmsub_<mode>_mask3<round_name>,
11601         <avx512>_fnmadd_<mode>_mask<round_name>,
11602         <avx512>_fnmadd_<mode>_mask3<round_name>,
11603         <avx512>_fnmsub_<mode>_mask<round_name>,
11604         <avx512>_fnmsub_<mode>_mask3<round_name>,
11605         <avx512>_fmaddsub_<mode>_mask<round_name>,
11606         <avx512>_fmaddsub_<mode>_mask3<round_name>,
11607         <avx512>_fmsubadd_<mode>_mask<round_name>,
11608         <avx512>_fmsubadd_<mode>_mask3<round_name>): Remove isa attribute.
11609         (*vec_widen_umult_even_v16si<mask_name>,
11610         *vec_widen_smult_even_v16si<mask_name>): Likewise.
11611         (<mask_codefor>avx512bw_dbpsadbw<mode><mask_name>): Likewise.
11613 2017-10-20  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11615         * extend.texi: Add 'nocf_check' documentation.
11616         * gimple.texi: Add second parameter to
11617         gimple_build_call_from_tree.
11618         * invoke.texi: Add -fcf-protection documentation.
11619         * rtl.texi: Add REG_CALL_NOTRACK documenation.
11621 2017-10-20  Richard Biener  <rguenther@suse.de>
11623         PR tree-optimization/82473
11624         * tree-vect-loop.c (vectorizable_reduction): Properly get at
11625         the largest input type.
11627 2017-10-20  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11629         * c-attribs.c (handle_nocf_check_attribute): New function.
11630         (c_common_attribute_table): Add 'nocf_check' handling.
11631         * gimple-parser.c: Add second argument NULL to
11632         gimple_build_call_from_tree.
11633         * attrib.c (comp_type_attributes): Check nocf_check attribute.
11634         * cfgexpand.c (expand_call_stmt): Set REG_CALL_NOCF_CHECK for
11635         call insn.
11636         * combine.c (distribute_notes): Add REG_CALL_NOCF_CHECK handling.
11637         * common.opt: Add fcf-protection flag.
11638         * emit-rtl.c (try_split): Add REG_CALL_NOCF_CHECK handling.
11639         * flag-types.h: Add enum cf_protection_level.
11640         * gimple.c (gimple_build_call_from_tree): Add second parameter.
11641         Add 'nocf_check' attribute propagation to gimple call.
11642         * gimple.h (gf_mask): Add GF_CALL_NOCF_CHECK.
11643         (gimple_build_call_from_tree): Update prototype.
11644         (gimple_call_nocf_check_p): New function.
11645         (gimple_call_set_nocf_check): Likewise.
11646         * gimplify.c: Add second argument to gimple_build_call_from_tree.
11647         * ipa-icf.c: Add nocf_check attribute in statement hash.
11648         * recog.c (peep2_attempt): Add REG_CALL_NOCF_CHECK handling.
11649         * reg-notes.def: Add REG_NOTE (CALL_NOCF_CHECK).
11650         * toplev.c (process_options): Add flag_cf_protection handling.
11652 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11654         * x86-tune-costs.h (core_cost): Fix div, move and sqrt latencies.
11656 2017-10-20  Richard Biener  <rguenther@suse.de>
11658         PR tree-optimization/82603
11659         * tree-if-conv.c (predicate_mem_writes): Make sure to only
11660         remove false predicated stores.
11662 2017-10-20  Richard Biener  <rguenther@suse.de>
11664         * graphite-isl-ast-to-gimple.c
11665         (translate_isl_ast_to_gimple::graphite_copy_stmts_from_block):
11666         Remove return value and simplify, dump copied stmt after lhs
11667         adjustment.
11668         (translate_isl_ast_to_gimple::translate_isl_ast_node_user):
11669         Reduce dump verbosity.
11670         (gsi_insert_earliest): Likewise.
11671         (translate_isl_ast_to_gimple::copy_bb_and_scalar_dependences): Adjust.
11672         * graphite.c (print_global_statistics): Adjust dumping.
11673         (print_graphite_scop_statistics): Likewise.
11674         (print_graphite_statistics): Do not dump loops here.
11675         (graphite_transform_loops): But here.
11677 2017-10-20  Nicolas Roche  <roche@adacore.com>
11679         * configure.ac (ACX_PROG_GNAT): Append "libgnat" to include search dir.
11680         * configure: Regenerate.
11682 2017-10-20  Jakub Jelinek  <jakub@redhat.com>
11684         PR target/82158
11685         * tree-cfg.c (pass_warn_function_return::execute): In noreturn
11686         functions when optimizing replace GIMPLE_RETURN stmts with
11687         calls to __builtin_unreachable ().
11689         PR sanitizer/82595
11690         * config/gnu-user.h (LIBTSAN_EARLY_SPEC): Add libtsan_preinit.o
11691         for -fsanitize=thread link of executables.
11692         (LIBLSAN_EARLY_SPEC): Add liblsan_preinit.o for -fsanitize=leak
11693         link of executables.
11695         PR target/82370
11696         * config/i386/sse.md (VI248_AVX2, VI248_AVX512BW, VI248_AVX512BW_2):
11697         New mode iterators.
11698         (<shift_insn><mode>3<mask_name>): Change the last of the 3
11699         define_insns for logical vector shifts to use VI248_AVX512BW
11700         iterator instead of VI48_AVX512, remove <mask_mode512bit_condition>
11701         condition, useless isa and prefix attributes.  Change the first
11702         2 of these define_insns to ...
11703         (<mask_codefor><shift_insn><mode>3<mask_name>): ... this, new
11704         define_insn for avx512vl.
11705         (<shift_insn><mode>3): ... and this, new define_insn without
11706         masking for non-avx512vl.
11708         PR target/82370
11709         * config/i386/sse.md (*andnot<mode>3,
11710         <mask_codefor><code><mode>3<mask_name>, *<code><mode>3): Split
11711         (=v,v,vm) alternative into (=x,x,xm) and (=v,v,vm), for 128-bit
11712         and 256-bit vectors, the (=x,x,xm) alternative and when mask is
11713         not applied use empty suffix even for TARGET_AVX512VL.
11714         * config/i386/subst.md (mask_prefix3, mask_prefix4): When mask
11715         is applied, supply evex,evex or evex,evex,evex instead of just
11716         evex.
11718 2017-10-20  Julia Koval  <julia.koval@intel.com>
11720         * common/config/i386/i386-common.c (OPTION_MASK_ISA_GFNI_SET,
11721         (OPTION_MASK_ISA_GFNI_UNSET): New.
11722         (ix86_handle_option): Handle OPT_mgfni.
11723         * config/i386/cpuid.h (bit_GFNI): New.
11724         * config/i386/driver-i386.c (host_detect_local_cpu): Detect gfni.
11725         * config/i386/i386-c.c (ix86_target_macros_internal): Define __GFNI__.
11726         * config/i386/i386.c (ix86_target_string): Add -mgfni.
11727         (ix86_valid_target_attribute_inner_p): Add OPT_mgfni.
11728         * config/i386/i386.h (TARGET_GFNI, TARGET_GFNI_P): New.
11729         * config/i386/i386.opt: Add mgfni.
11731 2017-10-20  Orlando Arias  <oarias@knights.ucf.edu>
11733         * config/msp430/msp430.c (msp430_option_override): Disable
11734         -fdelete-null-pointer-checks.
11735         * doc/invoke.text (-fdelete-null-pointer-checks): Document that.
11737 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11739         * x86-tune-costs.h (generic_cost, core_cost): Correct costs
11740         of x87 and SSE instructions.
11742 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11744         * asan.c (create_cond_insert_point): Do not update edge count.
11745         * auto-profile.c (afdo_propagate_edge): Update for edge count removal.
11746         (afdo_propagate_circuit): Likewise.
11747         (afdo_calculate_branch_prob): Likewise.
11748         (afdo_annotate_cfg): Likewise.
11749         * basic-block.h (struct edge_def): Remove count.
11750         (edge_def::count): New accessor.
11751         * bb-reorder.c (rotate_loop): Update.
11752         (find_traces_1_round): Update.
11753         (connect_traces): Update.
11754         (sanitize_hot_paths): Update.
11755         * cfg.c (unchecked_make_edge): Update.
11756         (make_single_succ_edge): Update.
11757         (check_bb_profile): Update.
11758         (dump_edge_info): Update.
11759         (update_bb_profile_for_threading): Update.
11760         (scale_bbs_frequencies_int): Update.
11761         (scale_bbs_frequencies_gcov_type): Update.
11762         (scale_bbs_frequencies_profile_count): Update.
11763         (scale_bbs_frequencies): Update.
11764         * cfganal.c (connect_infinite_loops_to_exit): Update.
11765         * cfgbuild.c (compute_outgoing_frequencies): Update.
11766         (find_many_sub_basic_blocks): Update.
11767         * cfgcleanup.c (try_forward_edges): Update.
11768         (try_crossjump_to_edge): Update
11769         * cfgexpand.c (expand_gimple_cond): Update
11770         (expand_gimple_tailcall): Update
11771         (construct_exit_block): Update
11772         * cfghooks.c (verify_flow_info): Update
11773         (redirect_edge_succ_nodup): Update
11774         (split_edge): Update
11775         (make_forwarder_block): Update
11776         (duplicate_block): Update
11777         (account_profile_record): Update
11778         * cfgloop.c (find_subloop_latch_edge_by_profile): Update.
11779         * cfgloopanal.c (expected_loop_iterations_unbounded): Update.
11780         * cfgloopmanip.c (scale_loop_profile): Update.
11781         (loopify): Update.
11782         (lv_adjust_loop_entry_edge): Update.
11783         * cfgrtl.c (try_redirect_by_replacing_jump): Update.
11784         (force_nonfallthru_and_redirect): Update.
11785         (purge_dead_edges): Update.
11786         (rtl_flow_call_edges_add): Update.
11787         * cgraphunit.c (init_lowered_empty_function): Update.
11788         (cgraph_node::expand_thunk): Update.
11789         * gimple-pretty-print.c (dump_probability): Update.
11790         (dump_edge_probability): Update.
11791         * gimple-ssa-isolate-paths.c (isolate_path): Update.
11792         * haifa-sched.c (sched_create_recovery_edges): Update.
11793         * hsa-gen.c (convert_switch_statements): Update.
11794         * ifcvt.c (dead_or_predicable): Update.
11795         * ipa-inline-transform.c (inline_transform): Update.
11796         * ipa-split.c (split_function): Update.
11797         * ipa-utils.c (ipa_merge_profiles): Update.
11798         * loop-doloop.c (add_test): Update.
11799         * loop-unroll.c (unroll_loop_runtime_iterations): Update.
11800         * lto-streamer-in.c (input_cfg): Update.
11801         (input_function): Update.
11802         * lto-streamer-out.c (output_cfg): Update.
11803         * modulo-sched.c (sms_schedule): Update.
11804         * postreload-gcse.c (eliminate_partially_redundant_load): Update.
11805         * predict.c (maybe_hot_edge_p): Update.
11806         (unlikely_executed_edge_p): Update.
11807         (probably_never_executed_edge_p): Update.
11808         (dump_prediction): Update.
11809         (drop_profile): Update.
11810         (propagate_unlikely_bbs_forward): Update.
11811         (determine_unlikely_bbs): Update.
11812         (force_edge_cold): Update.
11813         * profile.c (compute_branch_probabilities): Update.
11814         * reg-stack.c (better_edge): Update.
11815         * shrink-wrap.c (handle_simple_exit): Update.
11816         * tracer.c (better_p): Update.
11817         * trans-mem.c (expand_transaction): Update.
11818         (split_bb_make_tm_edge): Update.
11819         * tree-call-cdce.c: Update.
11820         * tree-cfg.c (gimple_find_sub_bbs): Update.
11821         (gimple_split_edge): Update.
11822         (gimple_duplicate_sese_region): Update.
11823         (gimple_duplicate_sese_tail): Update.
11824         (gimple_flow_call_edges_add): Update.
11825         (insert_cond_bb): Update.
11826         (execute_fixup_cfg): Update.
11827         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update.
11828         * tree-complex.c (expand_complex_div_wide): Update.
11829         * tree-eh.c (lower_resx): Update.
11830         (unsplit_eh): Update.
11831         (cleanup_empty_eh_move_lp): Update.
11832         * tree-inline.c (copy_edges_for_bb): Update.
11833         (freqs_to_counts): Update.
11834         (copy_cfg_body): Update.
11835         * tree-ssa-dce.c (remove_dead_stmt): Update.
11836         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
11837         * tree-ssa-loop-im.c (execute_sm_if_changed): Update.
11838         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update.
11839         (unloop_loops): Update.
11840         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
11841         * tree-ssa-loop-split.c (connect_loops): Update.
11842         (split_loop): Update.
11843         * tree-ssa-loop-unswitch.c (hoist_guard): Update.
11844         * tree-ssa-phionlycprop.c (propagate_rhs_into_lhs): Update.
11845         * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Update.
11846         * tree-ssa-reassoc.c (branch_fixup): Update.
11847         * tree-ssa-tail-merge.c (replace_block_by): Update.
11848         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges): Update.
11849         (compute_path_counts): Update.
11850         (update_profile): Update.
11851         (recompute_probabilities): Update.
11852         (update_joiner_offpath_counts): Update.
11853         (estimated_freqs_path): Update.
11854         (freqs_to_counts_path): Update.
11855         (clear_counts_path): Update.
11856         (ssa_fix_duplicate_block_edges): Update.
11857         (duplicate_thread_path): Update.
11858         * tree-switch-conversion.c (hoist_edge_and_branch_if_true): Update.
11859         (case_bit_test_cmp): Update.
11860         (collect_switch_conv_info): Update.
11861         (gen_inbound_check): Update.
11862         (do_jump_if_equal): Update.
11863         (emit_cmp_and_jump_insns): Update.
11864         * tree-tailcall.c (decrease_profile): Update.
11865         (eliminate_tail_call): Update.
11866         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
11867         (vect_do_peeling): Update.
11868         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
11869         * ubsan.c (ubsan_expand_null_ifn): Update.
11870         (ubsan_expand_ptr_ifn): Update.
11871         * value-prof.c (gimple_divmod_fixed_value): Update.
11872         (gimple_mod_pow2): Update.
11873         (gimple_mod_subtract): Update.
11874         (gimple_ic): Update.
11875         (gimple_stringop_fixed_value): Update.
11877 2017-10-19  Uros Bizjak  <ubizjak@gmail.com>
11879         PR target/82618
11880         * config/i386/i386.md (sub to cmp): New peephole2 pattern.
11882 2017-10-19  Alexander Monakov  <amonakov@ispras.ru>
11884         PR rtl-optimization/82395
11885         * ira-color.c (allocno_priority_compare_func): Fix comparison step
11886         based on non_spilled_static_chain_regno_p.
11888 2017-10-19  Uros Bizjak  <ubizjak@gmail.com>
11890         * config/i386/i386.c (output_387_binary_op): Rewrite SSE part.
11891         (ix86_emit_mode_set): Rewrite insn mnemonic construction.
11892         (ix86_prepare_fp_compare_args): Redefine is_sse as bool.
11894 2017-10-19  Martin Sebor  <msebor@redhat.com>
11896         PR tree-optimization/82596
11897         * tree.c (array_at_struct_end_p): Handle STRING_CST.
11899 2017-10-19  Eric Botcazou  <ebotcazou@adacore.com>
11901         * asan.c (handle_builtin_alloca): Deal with all alloca variants.
11902         (get_mem_refs_of_builtin_call): Likewise.
11903         * builtins.c (expand_builtin_apply): Adjust call to
11904         allocate_dynamic_stack_space.
11905         (expand_builtin_alloca): For __builtin_alloca_with_align_and_max, pass
11906         the third argument to allocate_dynamic_stack_space, otherwise -1.
11907         (expand_builtin): Deal with all alloca variants.
11908         (is_inexpensive_builtin): Likewise.
11909         * builtins.def (BUILT_IN_ALLOCA_WITH_ALIGN_AND_MAX): New.
11910         * calls.c (special_function_p): Deal with all alloca variants.
11911         (initialize_argument_information): Adjust call to
11912         allocate_dynamic_stack_space.
11913         (expand_call): Likewise.
11914         * cfgexpand.c (expand_call_stmt): Deal with all alloca variants.
11915         * doc/extend.texi (Built-ins): Add __builtin_alloca_with_align_and_max
11916         * explow.c (allocate_dynamic_stack_space): Add MAX_SIZE parameter and
11917         use it for the stack usage computation.
11918         * explow.h (allocate_dynamic_stack_space): Adjust prototype.
11919         * function.c (gimplify_parameters): Call build_alloca_call_expr.
11920         * gimple-ssa-warn-alloca.c (alloca_call_type): Simplify control flow.
11921         Take into account 3rd argument of __builtin_alloca_with_align_and_max.
11922         (in_loop_p): Remove first argument and useless check.
11923         (pass_walloca::execute): Remove useless test and adjust call to above.
11924         * gimple.c (gimple_build_call_from_tree): Deal with all alloc variants
11925         * gimplify.c (gimplify_vla_decl): Call build_alloca_call_expr.
11926         (gimplify_call_expr): Deal with all alloca variants.
11927         * hsa-gen.c (gen_hsa_alloca): Likewise.
11928         (gen_hsa_insns_for_call): Likewise.
11929         * ipa-pure-const.c (special_builtin_state): Likewise.
11930         * tree-chkp.c (chkp_build_returned_bound): Likewise.
11931         * tree-object-size.c (alloc_object_size): Likewise.
11932         * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
11933         (call_may_clobber_ref_p_1): Likewise.
11934         * tree-ssa-ccp.c (evaluate_stmt): Likewise.
11935         (ccp_fold_stmt): Likewise.
11936         (optimize_stack_restore): Likewise.
11937         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
11938         (mark_all_reaching_defs_necessary_1): Likewise.
11939         (propagate_necessity): Likewise.
11940         (eliminate_unnecessary_stmts): Likewise.
11941         * tree.c (build_common_builtin_nodes): Build
11942         BUILT_IN_ALLOCA_WITH_ALIGN_AND_MAX.
11943         (build_alloca_call_expr): New function.
11944         * tree.h (ALLOCA_FUNCTION_CODE_P): New macro.
11945         (CASE_BUILT_IN_ALLOCA): Likewise.
11946         (build_alloca_call_expr): Declare.
11947         * varasm.c (incorporeal_function_p): Deal with all alloca variants.
11949 2017-10-19  Eric Botcazou  <ebotcazou@adacore.com>
11951         PR debug/82509
11952         * dwarf2out.c (new_die_raw): New static inline function.
11953         (new_die): Use it to create the DIE.
11954         (add_AT_external_die_ref): Likewise.
11955         (clone_die): Likewise.
11956         (clone_as_declaration): Likewise.
11957         (dwarf2out_vms_debug_main_pointer): Likewise.
11958         (base_type_die): Likewise.  Remove early return for corner cases.
11959         Do not call add_pubtype on the DIE here.
11960         (is_base_type): Remove ERROR_MARK and return 0 for VOID_TYPE.
11961         (modified_type_die): Adjust the lookup for reverse order DIEs.  Skip
11962         typedefs for base types with DW_AT_endianity.  Make sure a DIE with
11963         native order exists for base types, attach the DIE manually and call
11964         add_pubtype on it.  Do not equate a reverse order DIE to the type.
11966 2017-10-19  Richard Earnshaw  <rearnsha@arm.com>
11968         * config/arm/arm.c (align_ok_ldrd_strd): New function.
11969         (mem_ok_for_ldrd_strd): New parameter align.  Extract the alignment of
11970         the mem into it.
11971         (gen_operands_ldrd_strd): Validate the alignment of the accesses.
11973 2017-10-19  Jakub Jelinek  <jakub@redhat.com>
11975         * flag-types.h (enum sanitize_code): Add SANITIZE_BUILTIN.  Or
11976         SANITIZE_BUILTIN into SANITIZE_UNDEFINED.
11977         * sanitizer.def (BUILT_IN_UBSAN_HANDLE_INVALID_BUILTIN,
11978         BUILT_IN_UBSAN_HANDLE_INVALID_BUILTIN_ABORT): New builtins.
11979         * opts.c (sanitizer_opts): Add builtin.
11980         * ubsan.c (instrument_builtin): New function.
11981         (pass_ubsan::execute): Call it.
11982         (pass_ubsan::gate): Enable even for SANITIZE_BUILTIN.
11983         * doc/invoke.texi: Document -fsanitize=builtin.
11985         * ubsan.c (ubsan_expand_null_ifn): Use _v1 suffixed type mismatch
11986         builtins, store max (log2 (align), 0) into uchar field instead of
11987         align into uptr field.
11988         (ubsan_expand_objsize_ifn): Use _v1 suffixed type mismatch builtins,
11989         store uchar 0 field instead of uptr 0 field.
11990         (instrument_nonnull_return): Use _v1 suffixed nonnull return builtin,
11991         instead of passing one address of struct with 2 locations pass
11992         two addresses of structs with 1 location each.
11993         * sanitizer.def (BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH,
11994         BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_ABORT,
11995         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN,
11996         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_ABORT): Removed.
11997         (BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_V1,
11998         BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_V1_ABORT,
11999         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_V1,
12000         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_V1_ABORT): New builtins.
12002 2017-10-19  Martin Liska  <mliska@suse.cz>
12004         PR driver/81829
12005         * file-find.c (remove_prefix): Remove.
12006         * file-find.h (remove_prefix): Likewise.
12007         * gcc-ar.c: Remove smartness of lookup.
12009 2017-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
12011         * config/rs6000/rs6000.md (*call_indirect_aix<mode>,
12012         *call_value_indirect_aix<mode>, *call_indirect_elfv2<mode>,
12013         *call_value_indirect_elfv2<mode>): Add correct mode to the unspec.
12015 2017-10-19  Jakub Jelinek  <jakub@redhat.com>
12017         PR target/82580
12018         * config/i386/i386.md (setcc + movzbl to xor + setcc): New peephole2.
12019         (setcc + and to xor + setcc): New peephole2.
12021 2017-10-19  Tom de Vries  <tom@codesourcery.com>
12023         * doc/sourcebuild.texi (Test Directives, Variants of
12024         dg-require-support): Add dg-require-stack-size.
12026 2017-10-19  Martin Liska  <mliska@suse.cz>
12028         PR sanitizer/82517
12029         * gimplify.c (gimplify_decl_expr): Do not instrument variables
12030         that have a large alignment.
12031         (gimplify_target_expr): Likewise.
12033 2017-10-18  Segher Boessenkool  <segher@kernel.crashing.org>
12035         PR rtl-optimization/82602
12036         * ira.c (rtx_moveable_p): Return false for volatile asm.
12038 2017-10-18  Uros Bizjak  <ubizjak@gmail.com>
12040         PR target/82580
12041         * config/i386/i386-modes.def (CCGZ): New CC mode.
12042         * config/i386/i386.md (sub<mode>3_carry_ccgz): New insn pattern.
12043         * config/i386/predicates.md (ix86_comparison_operator):
12044         Handle CCGZmode.
12045         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>:
12046         Emulate LE, LEU, GT, GTU, LT, LTU, GE and GEU double-word comparisons
12047         with double-word subtraction.
12048         (put_condition_code): Handle CCGZmode.
12050 2017-10-18  Aldy Hernandez  <aldyh@redhat.com>
12052         * wide-int.cc (debug (const wide_int &)): New.
12053         (debug (const wide_int *)): New.
12054         (debug (const widest_int &)): New.
12055         (debug (const widest_int *)): New.
12057 2017-10-18  Vladimir Makarov  <vmakarov@redhat.com>
12059         PR middle-end/82556
12060         * lra-constraints.c (curr_insn_transform): Use non-input operand
12061         instead of output one for matched reload.
12063 2017-10-18  Bin Cheng  <bin.cheng@arm.com>
12065         * tree-loop-distribution.c (INCLUDE_ALGORITHM): New header file.
12066         (tree-ssa-loop-ivopts.h): New header file.
12067         (struct builtin_info): New fields.
12068         (classify_builtin_1): Compute and record base and offset parts for
12069         memset builtin partition by calling strip_offset.
12070         (offset_cmp, fuse_memset_builtins): New functions.
12071         (finalize_partitions): Fuse adjacent memset partitions by calling
12072         above function.
12073         * tree-ssa-loop-ivopts.c (strip_offset): Delete static declaration.
12074         Expose the interface.
12075         * tree-ssa-loop-ivopts.h (strip_offset): New declaration.
12077 2017-10-18  Bin Cheng  <bin.cheng@arm.com>
12079         PR tree-optimization/82574
12080         * tree-loop-distribution.c (find_single_drs): New parameter.  Check
12081         that data reference must be executed exactly once per iteration
12082         against the outermost loop in nest.
12083         (classify_partition): Update call to above function.
12085 2017-10-18  Richard Biener  <rguenther@suse.de>
12087         PR tree-optimization/82591
12088         * graphite.c (graphite_transform_loops): Move code gen message
12089         printing ...
12090         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
12091         Here.  Handle scop_to_isl_ast failing.
12092         (scop_to_isl_ast): Limit the number of ISL operations.
12094 2017-10-18  Richard Biener  <rguenther@suse.de>
12096         * graphite-isl-ast-to-gimple.c
12097         (translate_isl_ast_to_gimple::set_rename): Simplify.
12098         (translate_isl_ast_to_gimple::set_rename_for_each_def): Inline...
12099         (graphite_copy_stmts_from_block): ... here.
12100         (copy_bb_and_scalar_dependences): Simplify.
12101         (add_parameters_to_ivs_params): Canonicalize.
12102         (generate_entry_out_of_ssa_copies): Simplify.
12103         * graphite-sese-to-poly.c (extract_affine_name): Simplify
12104         by passing in ISL dimension.
12105         (parameter_index_in_region_1): Rename to ...
12106         (parameter_index_in_region): ... this.
12107         (extract_affine): Adjust assert, pass down parameter index.
12108         (add_param_constraints): Use range-info when available.
12109         (build_scop_context): Adjust.
12110         * sese.c (new_sese_info): Adjust.
12111         (free_sese_info): Likewise.
12112         * sese.h (bb_map_t, rename_map_t, phi_rename, init_back_edge_pair_t):
12113         Remove unused typedefs.
12114         (struct sese_info_t): Simplify rename_map, remove incomplete_phis.
12116 2017-10-18  Martin Liska  <mliska@suse.cz>
12118         * combine.c (simplify_compare_const): Add gcc_fallthrough.
12120 2017-10-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
12122         * config/s390/s390.c (s390_bb_fallthru_entry_likely): New function.
12123         (s390_sched_init): Do not reset s390_sched_state if we entered the
12124         current basic block via a fallthru edge and all others are unlikely.
12126 2017-10-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
12128         * config/s390/s390.c (NUM_SIDES): New variable.
12129         (LONGRUNNING_THRESHOLD): New variable.
12130         (LATENCY_FACTOR): New variable.
12131         (s390_sched_score): Decrease score for long-running instructions on
12132         wrong side.
12133         (s390_sched_variable_issue): Perform bookkeeping for long-running
12134         instructions.
12136 2017-10-18  Richard Biener  <rguenther@suse.de>
12138         * graphite-isl-ast-to-gimple.c (gcc_expression_from_isl_ast_expr_id):
12139         Simplify with removal of the parameter rename map.
12140         (set_rename): Likewise.
12141         (should_copy_to_new_region): Likewise.
12142         (graphite_copy_stmts_from_block): Likewise.
12143         (copy_bb_and_scalar_dependences): Remove initialization of
12144         unused copied_bb_map.
12145         (copy_def): Remove.
12146         (copy_internal_parameters): Likewise.
12147         (graphite_regenerate_ast_isl): Do not call copy_internal_parameters.
12148         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
12149         Use INTEGRAL_TYPE_P.
12150         (parameter_index_in_region_1): Rename to ...
12151         (assign_parameter_index_in_region): ... this.  Assert we have
12152         a parameter we handle.
12153         (scan_tree_for_params): Adjust.
12154         * sese.h (parameter_rename_map_t): Remove.
12155         (struct sese_info_t): Remove unused parameter_rename_map and
12156         copied_bb_map members.
12157         * sese.c (new_sese_info): Adjust.
12158         (free_sese_info): Likewise.
12160 2017-10-18  Martin Liska  <mliska@suse.cz>
12162         PR sanitizer/82545
12163         * asan.c (asan_expand_poison_ifn): Do not put gimple stmt
12164         on an abnormal edge.
12166 2017-10-18  Sebastian Huber  <sebastian.huber@embedded-brains.de>
12168         * doc/invoke.texi (ffunction-sections and fdata-sections):
12169         Update.
12171 2017-10-17  Eric Botcazou  <ebotcazou@adacore.com>
12173         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Bail out only if
12174         the use statement can throw internally.
12176 2017-10-17  Eric Botcazou  <ebotcazou@adacore.com>
12178         * config/visium/visium.c (visium_select_cc_mode): Return CCmode for
12179         any RTX present on the RHS of a SET.
12180         * compare-elim.c (try_eliminate_compare): Restore comment.
12182 2017-10-17  Jakub Jelinek  <jakub@redhat.com>
12184         * langhooks.h (struct lang_hooks): Document that tree_size langhook
12185         may be also called on tcc_type nodes.
12186         * langhooks.c (lhd_tree_size): Likewise.
12188 2017-10-17  David Malcolm  <dmalcolm@redhat.com>
12190         * gimple-ssa-sprintf.c (fmtwarn): Update for changed signature of
12191         format_warning_at_substring.
12192         (maybe_warn): Convert source_range * param to a location_t.  Pass
12193         UNKNOWN_LOCATION rather than NULL to fmtwarn.
12194         (format_directive): Remove code to extract source_ranges and
12195         source_range * in favor of just a location_t.
12196         (parse_directive): Pass UNKNOWN_LOCATION rather than NULL to
12197         fmtwarn.
12198         * substring-locations.c (format_warning_va): Convert
12199         source_range * param to a location_t.
12200         (format_warning_at_substring): Likewise.
12201         * substring-locations.h (format_warning_va): Likewise.
12202         (format_warning_at_substring): Likewise.
12204 2017-10-17  Jan Hubicka  <hubicka@ucw.cz>
12206         * target.h (enum vect_cost_for_stmt): Add vec_gather_load and
12207         vec_scatter_store
12208         * tree-vect-stmts.c (record_stmt_cost): Make difference between normal
12209         and scatter/gather ops.
12211         * aarch64/aarch64.c (aarch64_builtin_vectorization_cost): Add
12212         vec_gather_load and vec_scatter_store.
12213         * arm/arm.c (arm_builtin_vectorization_cost): Likewise.
12214         * powerpcspe/powerpcspe.c (rs6000_builtin_vectorization_cost): Likewise.
12215         * rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Likewise.
12216         * s390/s390.c (s390_builtin_vectorization_cost): Likewise.
12217         * spu/spu.c (spu_builtin_vectorization_cost): Likewise.
12218         * i386/i386.c (x86_builtin_vectorization_cost): Likewise.
12220 2017-10-17  Uros Bizjak  <ubizjak@gmail.com>
12222         * reg-stack.c (compare_for_stack_reg): Add bool argument.
12223         Detect FTST instruction and handle its register pops.  Only pop
12224         second operand if can_pop_second_op is true.
12225         (subst_stack_regs_pat) <case COMPARE>: Detect FCOMI instruction to
12226         set can_pop_second_op to false in the compare_for_stack_reg call.
12228         * config/i386/i386.md (*cmpi<FPCMP:unord><MODEF:mode>): Only call
12229         output_fp_compare for stack register operands.
12230         * config/i386/i386.c (output_fp_compare): Do not output SSE compare
12231         instructions here.  Do not emit stack register pops here.  Assert
12232         that FCOMPP pops next to top stack register.  Rewrite function.
12234 2017-10-17  Nathan Sidwell  <nathan@acm.org>
12236         PR middle-end/82577
12237         * alias.c (compare_base_decls): Check HAS_DECL_ASSEMBLER_NAME_P,
12238         use DECL_ASSEMBLER_NAME_RAW.
12240         PR middle-end/82546
12241         * tree.c (tree_code_size): Reformat.  Punt to lang hook for unknown
12242         TYPE nodes.
12244 2017-10-17  Qing Zhao <qing.zhao@oracle.com>
12245             Wilco Dijkstra <wilco.dijkstra@arm.com>
12247         * builtins.c (expand_builtin_update_setjmp_buf): Add a
12248         converstion to Pmode from the buf_addr.
12250 2017-10-17  Richard Biener  <rguenther@suse.de>
12252         * graphite-dependences.c (scop_get_reads_and_writes): Change
12253         output parameters to references.
12255 2017-10-17  Jackson Woodruff  <jackson.woodruff@arm.com>
12257         PR 71026/tree-optimization
12258         * fold-const.c (distribute_real_division): Removed.
12259         (fold_binary_loc): Remove calls to distribute_real_divison.
12261 2017-10-17  Richard Biener  <rguenther@suse.de>
12263         * graphite-scop-detection.c
12264         (scop_detection::stmt_has_simple_data_refs_p): Always use
12265         the full nest as region.
12266         (try_generate_gimple_bb): Likewise.
12267         * sese.c (scalar_evolution_in_region): Simplify now that
12268         SCEV can handle instantiation in regions.
12269         * tree-scalar-evolution.c (instantiate_scev_name): Also instantiate
12270         in the non-loop part of a function if requested.
12272 2017-10-17  Richard Biener  <rguenther@suse.de>
12274         PR tree-optimization/82563
12275         * graphite-isl-ast-to-gimple.c (generate_entry_out_of_ssa_copies):
12276         New function.
12277         (graphite_regenerate_ast_isl): Call it.
12278         * graphite-scop-detection.c (build_scops): Remove entry edge split.
12280 2017-10-17  Jakub Jelinek  <jakub@redhat.com>
12282         PR tree-optimization/82549
12283         * fold-const.c (optimize_bit_field_compare, fold_truth_andor_1):
12284         Formatting fixes.  Instead of calling make_bit_field_ref with negative
12285         bitpos return 0.
12287 2017-10-17  Olga Makhotina  <olga.makhotina@intel.com>
12289         * config/i386/avx512dqintrin.h (_mm_mask_reduce_sd,
12290         _mm_maskz_reduce_sd, _mm_mask_reduce_ss,=20
12291         _mm_maskz_reduce_ss): New.
12292         * config/i386/i386-builtin.def (__builtin_ia32_reducesd_mask,
12293         __builtin_ia32_reducess_mask): Ditto..
12294         (__builtin_ia32_reducesd, __builtin_ia32_reducess): Remove.
12295         * config/i386/sse.md (reduces<mode>): Renamed to ...
12296         (reduces<mode><mask_scalar_name>): ... this.
12297         (vreduce<ssescalarmodesuffix>\t{%3, %2, %1, %0|%0, %1, %2, %3}):
12298         Changed to ...
12299         (vreduce<ssescalarmodesuffix>\t{%3, %2, %1, %0<mask_scalar_operand4>|
12300         %0<mask_scalar_operand4>, %1, %2, %3}): ... this.
12302 2017-10-16  David Malcolm  <dmalcolm@redhat.com>
12304         * Makefile.in (OBJS): Add unique-ptr-tests.o.
12305         * selftest-run-tests.c (selftest::run_tests): Call
12306         selftest::unique_ptr_tests_cc_tests.
12307         * selftest.h (selftest::unique_ptr_tests_cc_tests): New decl.
12308         * unique-ptr-tests.cc: New file.
12310 2017-10-16  Vladimir Makarov  <vmakarov@redhat.com>
12312         PR sanitizer/82353
12313         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12314         locations.
12315         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12316         (make_hard_regno_born, make_hard_regno_dead): Update
12317         bb_killed_pseudos and bb_gen_pseudos for fixed regs.
12319 2017-10-16  Jeff Law  <law@redhat.com>
12321         * tree-ssa-dse.c (live_bytes_read): Fix thinko.
12323 2017-10-16  Jan Hubicka  <hubicka@ucw.cz>
12325         * x86-tune-costs.h (znver1_cost): Fix move cost tables.
12327 2017-10-16  Olivier Hainque  <hainque@adacore.com>
12329         * gcc/config.gcc (powerpc*-*-*spe*): Pick 8548 as the default
12330         with_cpu if we were configured for an e500v2 target cpu name.
12332 2017-10-16  Thomas Preud'homme  <thomas.preudhomme@arm.com>
12334         * config/arm/arm-cpus.in (cortex-m33): Add nodsp option.
12335         * doc/invoke.texi: Document +nodsp as a valid extension for
12336         -mcpu=cortex-m33.
12338 2017-10-16  Martin Liska  <mliska@suse.cz>
12340         * sbitmap.c (bitmap_bit_in_range_p_checking): New function.
12341         (test_set_range): Likewise.
12342         (test_range_functions): Rename to ...
12343         (test_bit_in_range): ... this.
12344         (sbitmap_c_tests): Add new test.
12346 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12348         * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
12349         New.
12350         (vdot_lane_u32, vdot_laneq_u32, vdotq_lane_u32, vdotq_laneq_u32): New.
12351         (vdot_lane_s32, vdot_laneq_s32, vdotq_lane_s32, vdotq_laneq_s32): New.
12353 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12355         * config/aarch64/aarch64-builtins.c
12356         (aarch64_types_quadopu_lane_qualifiers): New.
12357         (TYPES_QUADOPU_LANE): New.
12358         * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): New.
12359         (<sur>dot_prod<vsi2qi>, aarch64_<sur>dot_lane<vsi2qi>): New.
12360         (aarch64_<sur>dot_laneq<vsi2qi>): New.
12361         * config/aarch64/aarch64-simd-builtins.def (sdot, udot): New.
12362         (sdot_lane, udot_lane, sdot_laneq, udot_laneq): New.
12363         * config/aarch64/iterators.md (sur): Add UNSPEC_SDOT, UNSPEC_UDOT.
12364         (Vdottype, DOTPROD): New.
12365         (sur): Add SDOT and UDOT.
12367 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12369         * config/aarch64/aarch64.h (AARCH64_FL_DOTPROD): New.
12370         (AARCH64_ISA_DOTPROD, TARGET_DOTPROD): New.
12371         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12372         Add TARGET_DOTPROD.
12373         * config/aarch64/aarch64-option-extensions.def (dotprod): New.
12374         * config/aarch64/aarch64-cores.def (cortex-a55, cortex-a75):
12375         Enable TARGET_DOTPROD.
12376         (cortex-a75.cortex-a55): Likewise.
12377         * doc/invoke.texi (aarch64-feature-modifiers): Document dotprod.
12379 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12381         * config/arm/arm-builtins.c (arm_unsigned_uternop_qualifiers): New.
12382         (UTERNOP_QUALIFIERS, arm_umac_lane_qualifiers, UMAC_LANE_QUALIFIERS):
12383         New.
12384         * config/arm/arm_neon_builtins.def (sdot, udot, sdot_lane, udot_lane):
12385         New.
12386         * config/arm/iterators.md (DOTPROD, VSI2QI, vsi2qi): New.
12387         (UNSPEC_DOT_S, UNSPEC_DOT_U, opsuffix): New.
12388         * config/arm/neon.md (neon_<sup>dot<vsi2qi>): New.
12389         (neon_<sup>dot_lane<vsi2qi>, <sup>dot_prod<vsi2qi>): New.
12390         * config/arm/types.md (neon_dot, neon_dot_q): New.
12391         * config/arm/unspecs.md (sup): Add UNSPEC_DOT_S, UNSPEC_DOT_U.
12393 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12395         * config/arm/arm.h (TARGET_DOTPROD): New.
12396         * config/arm/arm.c (arm_arch_dotprod): New.
12397         (arm_option_reconfigure_globals): Add arm_arch_dotprod.
12398         * config/arm/arm-c.c (__ARM_FEATURE_DOTPROD): New.
12399         * config/arm/arm-cpus.in (armv8.2-a): Enabled +dotprod.
12400         (feature dotprod, group dotprod, ALL_SIMD_INTERNAL): New.
12401         (ALL_FPU_INTERNAL): Use ALL_SIMD_INTERNAL.
12402         * config/arm/t-multilib (v8_2_a_simd_variants): Add dotprod.
12403         * doc/invoke.texi (armv8.2-a): Document dotprod
12405 2017-10-14  Jan Hubicka  <hubicka@ucw.cz>
12407         * i386.c (ix86_vec_cost): New function.
12408         (ix86_rtx_costs): Handle vector operations better.
12409         * i386.h (struct processor_costs): Add sse_op, fmasd, fmass.
12410         * x86-tune-costs.h: Add new costs to all tables.
12412 2017-10-14  Jan Hubicka  <hubicka@ucw.cz>
12414         * i386.c (ix86_rtx_costs): Make difference between x87 and SSE
12415         operations.
12416         * i386.h (struct processor_costs): Add addss, mulss, mulsd, divss,
12417         divsd, sqrtss and sqrtsd
12418         * x86-tune-costs.h: Add new entries to all costs.
12419         (znver1_cost): Fix to match real instruction latencies.
12421 2017-10-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12422             Michael Collison <michael.collison@arm.com>
12424         * compare-elim.c: Include emit-rtl.h.
12425         (can_merge_compare_into_arith): New function.
12426         (try_validate_parallel): Likewise.
12427         (try_merge_compare): Likewise.
12428         (try_eliminate_compare): Call the above when no previous clobber
12429         is available.
12430         (execute_compare_elim_after_reload): Add DF_UD_CHAIN and DF_DU_CHAIN
12431         dataflow problems.
12433 2017-10-14  Jakub Jelinek  <jakub@redhat.com>
12435         PR middle-end/62263
12436         PR middle-end/82498
12437         * tree-ssa-phiopt.c (value_replacement): Comment fix.  Handle
12438         up to 2 preparation statements for ASSIGN in MIDDLE_BB.
12440         PR middle-end/62263
12441         PR middle-end/82498
12442         * tree-ssa-forwprop.c (simplify_rotate): Allow def_arg1[N]
12443         to be any operand_equal_p operands.  For & (B - 1) require
12444         B to be power of 2.  Recognize
12445         (X << (Y & (B - 1))) | (X >> ((-Y) & (B - 1))) and similar patterns.
12447 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12449         PR bootstrap/82553
12450         * optabs.c (expand_memory_blockage): Fix call of
12451         targetm.have_memory_blockage.
12453 2017-10-14  Jakub Jelinek  <jakub@redhat.com>
12455         PR bootstrap/82548
12456         * config.gcc (*-*-solaris2*, i[34567]86-*-cygwin*,
12457         x86_64-*-cygwin*, i[34567]86-*-mingw* | x86_64-*-mingw*): Append
12458         objects to extra_objs instead of overwriting it.
12460 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12462         * config/i386/sync.md (FILD_ATOMIC/FIST_ATOMIC FP load peephole2):
12463         Use any_fp_register_operand as operand[3] predicate.  Simplify
12464         equality test for operands[2] and operands[4] memory location.
12465         (LDX_ATOMIC/STX_ATOMIC FP load peephole2): Ditto.
12466         (FILD_ATOMIC/FIST_ATOMIC FP load peephole2 with mem blockage): New.
12467         (LDX_ATOMIC/LDX_ATOMIC FP load peephole2 with mem blockage): Ditto.
12468         (FILD_ATOMIC/FIST_ATOMIC FP store peephole2): Use
12469         any_fp_register_operand as operand[1] predicate.  Simplify
12470         equality test for operands[0] and operands[3] memory location.
12471         (LDX_ATOMIC/STX_ATOMIC FP store peephole2): Ditto.
12472         (FILD_ATOMIC/FIST_ATOMIC FP store peephole2 with mem blockage): New.
12473         (LDX_ATOMIC/LDX_ATOMIC FP storepeephole2 with mem blockage): Ditto.
12475 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12477         * target-insns.def: Add memory_blockage.
12478         * optabs.c (expand_memory_blockage): New function.
12479         (expand_asm_memory_barrier): Rename ...
12480         (expand_asm_memory_blockage): ... to this.
12481         (expand_mem_thread_fence): Call expand_memory_blockage
12482         instead of expand_asm_memory_barrier.
12483         (expand_mem_singnal_fence): Ditto.
12484         (expand_atomic_load): Ditto.
12485         (expand_atomic_store): Ditto.
12486         * doc/md.texi (Standard Pattern Names For Generation):
12487         Document memory_blockage instruction pattern.
12489 2017-10-13  Sebastian Perta  <sebastian.perta@renesas.com>
12491         * config/rl78/rl78.c (rl78_emit_libcall): New function.
12492         * config/rl78/rl78-protos.h (rl78_emit_libcall): New function.
12493         * config/rl78/rl78.md: New define_expand "adddi3".
12495 2017-10-13  Jan Hubicka  <hubicka@ucw.cz>
12497         * cfghooks.c (verify_flow_info): Disable check that all probabilities
12498         are set correctly.
12500 2017-10-13  Jeff Law  <law@redhat.com>
12502         * tree-ssa-reassoc.c (reassociate_bb): Clarify code slighly.
12504 2017-10-13  Jakub Jelinek  <jakub@redhat.com>
12506         PR target/82274
12507         * internal-fn.c (expand_mul_overflow): If both operands have
12508         the same highpart of -1 or 0 and the topmost bit of lowpart
12509         is different, overflow is if res <= 0 rather than res < 0.
12511 2017-10-13  Pat Haugen  <pthaugen@us.ibm.com>
12513         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Remove
12514         TARGET_P9_VECTOR code for unaligned_load case.
12516 2017-10-13  Jan Hubicka  <hubicka@ucw.cz>
12518         * cfghooks.c (verify_flow_info): Check that edge probabilities are set.
12520 2017-10-13  Nathan Sidwell  <nathan@acm.org>
12522         * tree-core.h (tree_contains_struct): Make bool.
12523         * tree.c (tree_contains_struct): Likewise.
12524         * tree.h (MARK_TS_BASE): Remove do ... while (0) idiom.
12525         (MARK_TS_TYPED, MARK_TS_COMMON, MARK_TS_TYPE_COMMON,
12526         MARK_TS_TYPE_WITH_LANG_SPECIFIC, MARK_TS_DECL_MINIMAL,
12527         MARK_TS_DECL_COMMON, MARK_TS_DECL_WRTL, MARK_TS_DECL_WITH_VIS,
12528         MARK_TS_DECL_NON_COMMON): Likewise, use comma operator.
12530 2017-10-13  Richard Biener  <rguenther@suse.de>
12532         * graphite-isl-ast-to-gimple.c
12533         (translate_isl_ast_to_gimple::get_rename_from_scev): Remove unused
12534         parameters and dominance check.
12535         (translate_isl_ast_to_gimple::graphite_copy_stmts_from_block): Adjust.
12536         (translate_isl_ast_to_gimple::copy_bb_and_scalar_dependences): Likewise.
12537         (translate_isl_ast_to_gimple::graphite_regenerate_ast_isl):
12538         Do not update SSA form here or do intermediate IL verification.
12539         * graphite.c: Include tree-ssa.h and tree-into-ssa.h.
12540         (graphite_initialize): Remove check on the number of loops in
12541         the function and inline into graphite_transform_loops.
12542         (graphite_finalize): Inline into graphite_transform_loops.
12543         (graphite_transform_loops): Perform SSA update and IL verification
12544         here.
12545         * params.def (PARAM_GRAPHITE_MIN_LOOPS_PER_FUNCTION): Remove.
12547 2017-10-13  Richard Biener  <rguenther@suse.de>
12549         * graphite-isl-ast-to-gimple.c (max_mode_int_precision,
12550         graphite_expression_type_precision): Avoid global constructor
12551         by moving ...
12552         (translate_isl_ast_to_gimple::translate_isl_ast_to_gimple): Here.
12553         (translate_isl_ast_to_gimple::graphite_expr_type): Add type member.
12554         (translate_isl_ast_to_gimple::translate_isl_ast_node_for): Use it.
12555         (translate_isl_ast_to_gimple::build_iv_mapping): Likewise.
12556         (translate_isl_ast_to_gimple::graphite_create_new_guard): Likewise.
12557         * graphite-sese-to-poly.c (build_original_schedule): Return nothing.
12559 2017-10-13  H.J. Lu  <hongjiu.lu@intel.com>
12561         PR target/82499
12562         * config/i386/i386.h (ix86_red_zone_size): New.
12563         * config/i386/i386.md (push peephole2s): Replace
12564         "!ix86_using_red_zone ()" with "ix86_red_zone_size == 0".
12566 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12567             Alan Hayward  <alan.hayward@arm.com>
12568             David Sherwood  <david.sherwood@arm.com>
12570         * combine.c (can_change_dest_mode): Reject changes in
12571         REGMODE_NATURAL_SIZE.
12573 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12574             Alan Hayward  <alan.hayward@arm.com>
12575             David Sherwood  <david.sherwood@arm.com>
12577         * cfgexpand.c (expand_debug_expr): Use GET_MODE_UNIT_BITSIZE.
12578         (expand_debug_source_expr): Likewise.
12579         * combine.c (combine_simplify_rtx): Likewise.
12580         * cse.c (fold_rtx): Likewise.
12581         * fwprop.c (canonicalize_address): Likewise.
12582         * targhooks.c (default_shift_truncation_mask): Likewise.
12584 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12585             Alan Hayward  <alan.hayward@arm.com>
12586             David Sherwood  <david.sherwood@arm.com>
12588         * optabs.c (add_equal_note): Use GET_MODE_UNIT_SIZE.
12589         (widened_mode): Likewise.
12590         (expand_unop): Likewise.
12591         * ree.c (transform_ifelse): Likewise.
12592         (merge_def_and_ext): Likewise.
12593         (combine_reaching_defs): Likewise.
12594         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12596 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12597             Alan Hayward  <alan.hayward@arm.com>
12598             David Sherwood  <david.sherwood@arm.com>
12600         * caller-save.c (replace_reg_with_saved_mem): Use byte_lowpart_offset.
12601         * combine.c (gen_lowpart_for_combine): Likewise.
12602         * dwarf2out.c (rtl_for_decl_location): Likewise.
12603         * final.c (alter_subreg): Likewise.
12604         * rtlhooks.c (gen_lowpart_general): Likewise.
12605         (gen_lowpart_if_possible): Likewise.
12607 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12608             Alan Hayward  <alan.hayward@arm.com>
12609             David Sherwood  <david.sherwood@arm.com>
12611         * calls.c (expand_call): Use subreg_lowpart_offset.
12612         * cse.c (cse_insn): Likewise.
12613         * regcprop.c (copy_value): Likewise.
12614         (copyprop_hardreg_forward_1): Likewise.
12616 2017-10-13  Jakub Jelinek  <jakub@redhat.com>
12618         PR target/82524
12619         * config/i386/i386.md (addqi_ext_1, andqi_ext_1,
12620         *andqi_ext_1_cc, *<code>qi_ext_1, *xorqi_ext_1_cc): Change
12621         =Q constraints to +Q and into insn condition add check
12622         that operands[0] and operands[1] are equal.
12623         (*addqi_ext_2, *andqi_ext_2, *<code>qi_ext_2): Change
12624         =Q constraints to +Q and into insn condition add check
12625         that operands[0] is equal to either operands[1] or operands[2].
12627         PR target/82498
12628         * fold-const.c (fold_binary_loc) <bit_rotate>: Code cleanups,
12629         instead of handling MINUS_EXPR twice (once for each argument),
12630         canonicalize operand order and handle just once, use rtype where
12631         possible.  Handle (A << B) | (A >> (-B & (Z - 1))).
12633         PR target/82498
12634         * config/i386/ia32intrin.h (__rold, __rord, __rolq, __rorq): Allow
12635         any values of __C while still being pattern recognizable as a simple
12636         rotate instruction.
12638 2017-10-13  Richard Biener  <rguenther@suse.de>
12640         PR tree-optimization/82451
12641         Revert
12642         2017-10-02  Richard Biener  <rguenther@suse.de>
12644         PR tree-optimization/82355
12645         * graphite-isl-ast-to-gimple.c (build_iv_mapping): Also build
12646         a mapping for the enclosing loop but avoid generating one for
12647         the loop tree root.
12648         (copy_bb_and_scalar_dependences): Remove premature codegen
12649         error on PHIs in blocks duplicated into multiple places.
12650         * graphite-scop-detection.c
12651         (scop_detection::stmt_has_simple_data_refs_p): For a loop not
12652         in the region use it as loop and nest to analyze the DR in.
12653         (try_generate_gimple_bb): Likewise.
12654         * graphite-sese-to-poly.c (extract_affine_chrec): Adjust.
12655         (add_loop_constraints): For blocks in a loop not in the region
12656         create a dimension with a single iteration.
12657         * sese.h (gbb_loop_at_index): Remove assert.
12659         * cfgloop.c (loop_preheader_edge): For the loop tree root
12660         return the single successor of the entry block.
12661         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
12662         Reset the SCEV hashtable and niters.
12663         * graphite-scop-detection.c
12664         (scop_detection::graphite_can_represent_scev): Add SCOP parameter,
12665         assert that we only have POLYNOMIAL_CHREC that vary in loops
12666         contained in the region.
12667         (scop_detection::graphite_can_represent_expr): Adjust.
12668         (scop_detection::stmt_has_simple_data_refs_p): For loops
12669         not in the region set loop to NULL.  The nest is now the
12670         entry edge to the region.
12671         (try_generate_gimple_bb): Likewise.
12672         * sese.c (scalar_evolution_in_region): Adjust for
12673         instantiate_scev change.
12674         * tree-data-ref.h (graphite_find_data_references_in_stmt):
12675         Make nest parameter the edge into the region.
12676         (create_data_ref): Likewise.
12677         * tree-data-ref.c (dr_analyze_indices): Make nest parameter an
12678         entry edge into a region and adjust instantiate_scev calls.
12679         (create_data_ref): Likewise.
12680         (graphite_find_data_references_in_stmt): Likewise.
12681         (find_data_references_in_stmt): Pass the loop preheader edge
12682         from the nest argument.
12683         * tree-scalar-evolution.h (instantiate_scev): Make instantiate_below
12684         parameter the edge into the region.
12685         (instantiate_parameters): Use the loop preheader edge as entry.
12686         * tree-scalar-evolution.c (analyze_scalar_evolution): Handle
12687         NULL loop.
12688         (get_instantiated_value_entry): Make instantiate_below parameter
12689         the edge into the region.
12690         (instantiate_scev_name): Likewise.  Adjust dominance checks,
12691         when we cannot use loop-based instantiation instantiate by
12692         walking use-def chains.
12693         (instantiate_scev_poly): Adjust.
12694         (instantiate_scev_binary): Likewise.
12695         (instantiate_scev_convert): Likewise.
12696         (instantiate_scev_not): Likewise.
12697         (instantiate_array_ref): Remove.
12698         (instantiate_scev_3): Likewise.
12699         (instantiate_scev_2): Likewise.
12700         (instantiate_scev_1): Likewise.
12701         (instantiate_scev_r): Do not blindly handle N-operand trees.
12702         Do not instantiate array-refs.  Handle all constants and invariants.
12703         (instantiate_scev): Make instantiate_below parameter
12704         the edge into the region.
12705         (resolve_mixers): Use the loop preheader edge for the region
12706         parameter to instantiate_scev_r.
12707         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Adjust.
12709 2017-10-13  Richard Biener  <rguenther@suse.de>
12711         PR tree-optimization/82525
12712         * graphite-isl-ast-to-gimple.c
12713         (translate_isl_ast_to_gimple::widest_int_from_isl_expr_int): Split
12714         out from ...
12715         (translate_isl_ast_to_gimple::gcc_expression_from_isl_expr_int): Here.
12716         Fail code generation when we cannot represent the isl integer.
12717         (binary_op_to_tree): Elide modulo operations that are no-ops
12718         in the type we code generate.  Remove now superfluous code
12719         generation errors.
12721 2017-10-13  Richard Biener  <rguenther@suse.de>
12723         * graphite-scop-detection.c (loop_ivs_can_be_represented): Remove.
12724         (scop_detection::harmful_loop_in_region): Remove premature
12725         IV type restriction.
12726         (scop_detection::graphite_can_represent_scev): We can handle
12727         pointer IVs just fine.
12729 2017-10-13  Alan Modra  <amodra@gmail.com>
12731         * doc/extend.texi (Extended Asm <Clobbers>): Rename to
12732         "Clobbers and Scratch Registers".  Add paragraph on
12733         alternative to clobbers for scratch registers and OpenBLAS
12734         example.
12736 2017-10-13  Alan Modra  <amodra@gmail.com>
12738         * doc/extend.texi (Clobbers): Correct vax example.  Delete old
12739         example of a memory input for a string of known length.  Move
12740         commentary out of table.  Add a number of new examples
12741         covering array memory inputs.
12743 2017-10-12  Martin Liska  <mliska@suse.cz>
12745         PR tree-optimization/82493
12746         * sbitmap.c (bitmap_bit_in_range_p): Fix the implementation.
12747         (test_range_functions): New function.
12748         (sbitmap_c_tests): Likewise.
12749         * selftest-run-tests.c (selftest::run_tests): Run new tests.
12750         * selftest.h (sbitmap_c_tests): New function.
12752         * tree-ssa-dse.c (live_bytes_read): Fix thinko.
12754 2017-10-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
12756         * config/rs6000/amo.h: Fix spacing issue.
12758 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12760         PR target/82498
12761         * config/i386/i386.md (*ashl<mode>3_mask_1,
12762         *<shift_insn><mode>3_mask_1, *<rotate_insn><mode>3_mask_1,
12763         *<btsc><mode>_mask_1, *btr<mode>_mask_1): New define_insn_and_split
12764         patterns.
12766 2017-10-12  Jan Hubicka  <hubicka@ucw.cz>
12768         * profile-count.h (safe_scale_64bit): Fix GCC4.x path.
12769         (profile_probability): Set max_probability
12770         to (uint32_t) 1 << (n_bits - 2) and update accessors to avoid overlfows
12771         in temporaries.
12772         * profile-count.c (profile_probability::differs_from_p): Do not
12773         rely on max_probaiblity == 10000
12775 2017-10-12  Jeff Law  <law@redhat.com>
12777         * tree-ssa-dse.c (valid_ao_ref_for_dse): Reject ao_refs with
12778         negative offsets.
12780 2017-10-12  Martin Sebor  <msebor@redhat.com>
12782         PR other/82301
12783         PR c/82435
12784         * cgraphunit.c (maybe_diag_incompatible_alias): New function.
12785         (handle_alias_pairs): Call it.
12786         * common.opt (-Wattribute-alias): New option.
12787         * doc/extend.texi (ifunc attribute): Discuss C++ specifics.
12788         * doc/invoke.texi (-Wattribute-alias): Document.
12790 2017-10-12  Vladimir Makarov  <vmakarov@redhat.com>
12792         Revert
12793         2017-10-11  Vladimir Makarov  <vmakarov@redhat.com>
12794         PR sanitizer/82353
12795         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12796         locations.
12797         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12798         (make_hard_regno_born, make_hard_regno_dead): Update
12799         bb_killed_pseudos and bb_gen_pseudos.
12801 2017-10-12  Jan Hubicka  <hubicka@ucw.cz>
12803         * config/i386/x86-tune-sched.c (ix86_adjust_cost): Fix Zen support.
12805 2017-10-12  Uros Bizjak  <ubizjak@gmail.com>
12807         * config/alpha/alpha.c (alpha_split_conditional_move):
12808         Use std::swap instead of manually swapping.
12809         (alpha_stdarg_optimize_hook): Ditto.
12810         (alpha_canonicalize_comparison): Ditto.
12812 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12814         * tree-loop-distribution.c (struct builtin_info): New struct.
12815         (struct partition): Refactor fields into struct builtin_info.
12816         (partition_free): Free struct builtin_info.
12817         (build_size_arg_loc, build_addr_arg_loc): Delete.
12818         (generate_memset_builtin, generate_memcpy_builtin): Get memory range
12819         information from struct builtin_info.
12820         (find_single_drs): New function refactored from classify_partition.
12821         Also moved builtin validity checks to this function.
12822         (compute_access_range, alloc_builtin): New functions.
12823         (classify_builtin_st, classify_builtin_ldst): New functions.
12824         (classify_partition): Refactor code into functions find_single_drs,
12825         classify_builtin_st and classify_builtin_ldst.
12826         (distribute_loop): Don't do runtime alias check when distributing
12827         loop nest.
12828         (find_seed_stmts_for_distribution): New function.
12829         (pass_loop_distribution::execute): Refactor code finding seed
12830         stmts into above function.  Support distribution for the innermost
12831         two-level loop nest.  Adjust dump information.
12833 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12835         * tree-loop-distribution.c: Adjust the general comment.
12836         (NUM_PARTITION_THRESHOLD): New macro.
12837         (ssa_name_has_uses_outside_loop_p): Support loop nest distribution.
12838         (classify_partition): Skip builtin pattern of loop nest's inner loop.
12839         (merge_dep_scc_partitions): New parameter ignore_alias_p and use it
12840         in call to build_partition_graph.
12841         (finalize_partitions): New parameter.  Make loop distribution more
12842         conservative by fusing more partitions.
12843         (distribute_loop): Don't do runtime alias check in case of loop nest
12844         distribution.
12845         (find_seed_stmts_for_distribution): New function.
12846         (prepare_perfect_loop_nest): New function.
12847         (pass_loop_distribution::execute): Refactor code finding seed stmts
12848         and loop nest into above functions.  Support loop nest distribution.
12849         Adjust dump information accordingly.
12851 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12853         * tree-loop-distribution.c (break_alias_scc_partitions): Add comment
12854         and set PTYPE_SEQUENTIAL for merged partition.
12856 2017-10-12  Richard Biener  <rguenther@suse.de>
12858         PR tree-optimization/69728
12859         Revert
12860         2017-09-19  Richard Biener  <rguenther@suse.de>
12862         PR tree-optimization/69728
12863         * graphite-sese-to-poly.c (schedule_error): New global.
12864         (add_loop_schedule): Handle empty domain by failing the
12865         schedule.
12866         (build_original_schedule): Handle schedule_error.
12868         * graphite-sese-to-poly.c (add_loop_schedule): Handle empty
12869         domain by returning an unchanged schedule.
12871 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12873         * genrecog.c (validate_pattern): For VEC_SELECT verify that
12874         CONST_INT selectors are 0 to GET_MODE_NUNITS (imode) - 1.
12876 2017-10-12  Aldy Hernandez  <aldyh@redhat.com>
12878         * Makefile.in (TAGS): Merge all the *.def files into one pattern.
12879         Handle params.def.
12881 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12883         PR c++/82159
12884         * expr.c (store_field): Don't optimize away bitsize == 0 store
12885         from CALL_EXPR with addressable return type.
12887 2017-10-11  Segher Boessenkool  <segher@kernel.crashing.org>
12889         * config/rs6000/rs6000.h (TARGET_ISEL64): Delete.
12890         * config/rs6000/rs6000.md (sel): Delete mode attribute.
12891         (mov<mode>cc, isel_signed_<mode>, isel_unsigned_<mode>,
12892         *isel_reversed_signed_<mode>, *isel_reversed_unsigned_<mode>): Use
12893         TARGET_ISEL instead of TARGET_ISEL<sel>.
12895 2017-10-11  David Edelsohn  <dje.gcc@gmail.com>
12897         * config/rs6000/rs6000.c
12898         (rs6000_xcoff_asm_output_aligned_decl_common): Test for NULL decl.
12900 2017-10-11  Segher Boessenkool  <segher@kernel.crashing.org>
12902         * config/rs6000/predicates.md (zero_constant, all_ones_constant):
12903         Move up in file.
12904         (reg_or_cint_operand): Fix comment.
12905         (reg_or_zero_operand): New predicate.
12906         * config/rs6000/rs6000-protos.h (output_isel): Delete.
12907         * config/rs6000/rs6000.c (output_isel): Delete.
12908         * config/rs6000/rs6000.md (isel_signed_<mode>): Use reg_or_zero_operand
12909         instead of reg_or_cint_operand.  Output instruction directly (not via
12910         output_isel).
12911         (isel_unsigned_<mode>): Ditto.
12912         (*isel_reversed_signed_<mode>): Use reg_or_zero_operand instead of
12913         gpc_reg_operand.  Add an instruction alternative for this.  Output
12914         instruction directly.
12915         (*isel_reversed_unsigned_<mode>): Ditto.
12917 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12919         * config/i386/i386.c (ix86_canonicalize_comparison): New function.
12920         (TARGET_CANONICALIZE_COMPARISON): Define.
12922 2017-10-11  Qing Zhao  <qing.zhao@oracle.com>
12924         PR target/81422
12925         * config/aarch64/aarch64.c (aarch64_load_symref_appropriately):
12926         Check whether the dest is REG before adding REG_EQUIV note.
12928 2017-10-11  Vladimir Makarov  <vmakarov@redhat.com>
12930         PR sanitizer/82353
12931         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12932         locations.
12933         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12934         (make_hard_regno_born, make_hard_regno_dead): Update
12935         bb_killed_pseudos and bb_gen_pseudos.
12937 2017-10-11  Nathan Sidwell  <nathan@acm.org>
12939         * incpath.h (enum incpath_kind): Name enum, prefix values.
12940         (add_path, add_cpp_dir_path, get_added_cpp_dirs): Use incpath_kind.
12941         * incpath.c (heads, tails): Use INC_MAX.
12942         (add_env_var_paths, add_standard_paths): Use incpath_kind.
12943         (merge_include_chains, split_quote_chain,
12944         register_include_chains): Update incpath_kind names.
12945         (add_cpp_dir_path, add_path, get_added_cpp_dirs): Use incpath_kind.
12946         * config/darwin-c.c (add_system_framework_path): Update incpath_kind
12947         names.
12948         (add_framework_path, darwin_register_objc_includes): Likewise.
12949         * config/vms/vms-c.c (vms_c_register_includes): Likewise.
12951 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12953         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
12954         Do not use float_operator operator predicate.
12955         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
12956         * config/i386/predicates.md (float_operator): Remove predicate.
12958 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12960         * config/i386/i386.md (*jcc<mode>_0_i387): Remove insn pattern.
12961         (*jccxf_i387): Ditto.
12962         (*jcc<mode>_i387): Ditto.
12963         (*jccu<mode>_i387): Ditto.
12964         (*jcc<X87MODEF:mode>_<SWI24:mode>_i387): Ditto.
12965         (*jcc_*_i387 splitters): Remove.
12966         * config/i386/i386-protos.h (ix86_split_fp_branch): Remove prototype.
12967         * config/i386/i386.c (ix86_split_fp_branch): Remove.
12968         * config/i386/predicates.md (ix86_swapped_fp_comparison_operator):
12969         Remove predicate.
12971 2017-10-11  Jan Hubicka  <hubicka@ucw.cz>
12973         * profile-count.h (slow_safe_scale_64bit): New function.
12974         (safe_scale_64bit): New inline.
12975         (profile_count::max_safe_multiplier): Remove; use safe_scale_64bit.
12976         * profile-count.c: Include wide-int.h
12977         (slow_safe_scale_64bit): New.
12979 2017-10-11  Nathan Sidwell  <nathan@acm.org>
12981         * tree.h (DECL_ASSEMBLER_NAME_SET_P): Don't check
12982         HAS_DECL_ASSEMBLER_NAME_P.
12983         * gimple-expr.c (gimple_decl_printable_name: Check
12984         HAS_DECL_ASSEMBLER_NAME_P too.
12985         * ipa-utils.h (type_in_anonymous_namespace_p): Check
12986         DECL_ASSEMBLER_NAME_SET_P of TYPE_NAME.
12987         (odr_type_p): No need to assert TYPE_NAME is a TYPE_DECL.
12988         * passes.c (rest_of_decl_compilation): Check
12989         HAS_DECL_ASSEMBLER_NAME_P too.
12990         * recog.c (verify_changes): Likewise.
12991         * tree-pretty-print.c (dump_decl_name): Likewise.
12992         * tree-ssa-structalias.c (alias_get_name): Likewise.  Reimplement.
12994         * tree.h (DECL_ASSEMBLER_NAME_RAW): New.
12995         (SET_DECL_ASSEMBLER_NAME): Use it.
12996         (DECL_ASSEMBLER_NAME_SET_P): Likewise.
12997         (COPY_DECL_ASSEMBLER_NAME): Likewise.
12998         * tree.c (decl_assembler_name): Use DECL_ASSEMBLER_NAME_RAW.
13000 2017-10-11  Jan Hubicka  <hubicka@ucw.cz>
13002         * config.gcc (i386, x86_64): Add extra objects.
13003         * config/i386/i386-protos.h (ix86_rip_relative_addr_p): Declare.
13004         (ix86_min_insn_size): Declare.
13005         (ix86_issue_rate): Declare.
13006         (ix86_adjust_cost): Declare.
13007         (ia32_multipass_dfa_lookahead): Declare.
13008         (ix86_macro_fusion_p): Declare.
13009         (ix86_macro_fusion_pair_p): Declare.
13010         (ix86_bd_has_dispatch): Declare.
13011         (ix86_bd_do_dispatch): Declare.
13012         (ix86_core2i7_init_hooks): Declare.
13013         (ix86_atom_sched_reorder): Declare.
13014         * config/i386/i386.c Move all CPU cost tables to x86-tune-costs.h.
13015         (COSTS_N_BYTES): Move to x86-tune-costs.h.
13016         (DUMMY_STRINGOP_ALGS):Move to x86-tune-costs.h.
13017         (rip_relative_addr_p): Rename to ...
13018         (ix86_rip_relative_addr_p): ... this one; export.
13019         (memory_address_length): Update.
13020         (ix86_issue_rate): Move to x86-tune-sched.c.
13021         (ix86_flags_dependent): Move to x86-tune-sched.c.
13022         (ix86_agi_dependent): Move to x86-tune-sched.c.
13023         (exact_dependency_1): Move to x86-tune-sched.c.
13024         (exact_store_load_dependency): Move to x86-tune-sched.c.
13025         (ix86_adjust_cost): Move to x86-tune-sched.c.
13026         (ia32_multipass_dfa_lookahead): Move to x86-tune-sched.c.
13027         (ix86_macro_fusion_p): Move to x86-tune-sched.c.
13028         (ix86_macro_fusion_pair_p): Move to x86-tune-sched.c.
13029         (do_reorder_for_imul): Move to x86-tune-sched-atom.c.
13030         (swap_top_of_ready_list): Move to x86-tune-sched-atom.c.
13031         (ix86_sched_reorder): Move to x86-tune-sched-atom.c.
13032         (core2i7_first_cycle_multipass_init): Move to x86-tune-sched-core.c.
13033         (core2i7_dfa_post_advance_cycle): Move to x86-tune-sched-core.c.
13034         (min_insn_size): Rename to ...
13035         (ix86_min_insn_size): ... this one; export.
13036         (core2i7_first_cycle_multipass_begin): Move to x86-tune-sched-core.c.
13037         (core2i7_first_cycle_multipass_issue): Move to x86-tune-sched-core.c.
13038         (core2i7_first_cycle_multipass_backtrack): Move to
13039         x86-tune-sched-core.c.
13040         (core2i7_first_cycle_multipass_end): Move to x86-tune-sched-core.c.
13041         (core2i7_first_cycle_multipass_fini): Move to x86-tune-sched-core.c.
13042         (ix86_sched_init_global): Break up logic to ix86_core2i7_init_hooks.
13043         (ix86_avoid_jump_mispredicts): Update.
13044         (TARGET_SCHED_DISPATCH): Move to ix86-tune-sched-bd.c.
13045         (TARGET_SCHED_DISPATCH_DO): Move to ix86-tune-sched-bd.c.
13046         (TARGET_SCHED_REORDER): Move to ix86-tune-sched-bd.c.
13047         (DISPATCH_WINDOW_SIZE): Move to ix86-tune-sched-bd.c.
13048         (MAX_DISPATCH_WINDOWS): Move to ix86-tune-sched-bd.c.
13049         (MAX_INSN): Move to ix86-tune-sched-bd.c.
13050         (MAX_IMM): Move to ix86-tune-sched-bd.c.
13051         (MAX_IMM_SIZE): Move to ix86-tune-sched-bd.c.
13052         (MAX_IMM_32): Move to ix86-tune-sched-bd.c.
13053         (MAX_IMM_64): Move to ix86-tune-sched-bd.c.
13054         (MAX_LOAD): Move to ix86-tune-sched-bd.c.
13055         (MAX_STORE): Move to ix86-tune-sched-bd.c.
13056         (BIG): Move to ix86-tune-sched-bd.c.
13057         (enum dispatch_group): Move to ix86-tune-sched-bd.c.
13058         (enum insn_path): Move to ix86-tune-sched-bd.c.
13059         (get_mem_group): Move to ix86-tune-sched-bd.c.
13060         (is_cmp): Move to ix86-tune-sched-bd.c.
13061         (dispatch_violation): Move to ix86-tune-sched-bd.c.
13062         (is_branch): Move to ix86-tune-sched-bd.c.
13063         (is_prefetch): Move to ix86-tune-sched-bd.c.
13064         (init_window): Move to ix86-tune-sched-bd.c.
13065         (allocate_window): Move to ix86-tune-sched-bd.c.
13066         (init_dispatch_sched): Move to ix86-tune-sched-bd.c.
13067         (is_end_basic_block): Move to ix86-tune-sched-bd.c.
13068         (process_end_window): Move to ix86-tune-sched-bd.c.
13069         (allocate_next_window): Move to ix86-tune-sched-bd.c.
13070         (find_constant): Move to ix86-tune-sched-bd.c.
13071         (get_num_immediates): Move to ix86-tune-sched-bd.c.
13072         (has_immediate): Move to ix86-tune-sched-bd.c.
13073         (get_insn_path): Move to ix86-tune-sched-bd.c.
13074         (get_insn_group): Move to ix86-tune-sched-bd.c.
13075         (count_num_restricted): Move to ix86-tune-sched-bd.c.
13076         (fits_dispatch_window): Move to ix86-tune-sched-bd.c.
13077         (add_insn_window): Move to ix86-tune-sched-bd.c.
13078         (add_to_dispatch_window): Move to ix86-tune-sched-bd.c.
13079         (debug_dispatch_window_file): Move to ix86-tune-sched-bd.c.
13080         (debug_dispatch_window): Move to ix86-tune-sched-bd.c.
13081         (debug_insn_dispatch_info_file): Move to ix86-tune-sched-bd.c.
13082         (debug_ready_dispatch): Move to ix86-tune-sched-bd.c.
13083         (do_dispatch): Move to ix86-tune-sched-bd.c.
13084         (has_dispatch): Move to ix86-tune-sched-bd.c.
13085         * config/i386/t-i386: Add new object files.
13086         * config/i386/x86-tune-costs.h: New file.
13087         * config/i386/x86-tune-sched-atom.c: New file.
13088         * config/i386/x86-tune-sched-bd.c: New file.
13089         * config/i386/x86-tune-sched-core.c: New file.
13090         * config/i386/x86-tune-sched.c: New file.
13092 2017-10-11  Liu Hao  <lh_mouse@126.com>
13094         * pretty-print.c [_WIN32] (colorize_init): Remove.  Use
13095         the generic version below instead.
13096         (should_colorize): Recognize Windows consoles as terminals
13097         for MinGW targets.
13098         * pretty-print.c [__MINGW32__] (write_all): New function.
13099         [__MINGW32__] (find_esc_head): Likewise.
13100         [__MINGW32__] (find_esc_terminator): Likewise.
13101         [__MINGW32__] (eat_esc_sequence): Likewise.
13102         [__MINGW32__] (mingw_ansi_fputs): New function that handles
13103         ANSI escape codes.
13104         (pp_write_text_to_stream): Use mingw_ansi_fputs instead of fputs
13105         for MinGW targets.
13107 2017-10-11  Richard Biener  <rguenther@suse.de>
13109         * tree-ssa-loop-niter.c (infer_loop_bounds_from_pointer_arith):
13110         Properly call analyze_scalar_evolution with the loop of the stmt.
13112 2017-10-11  Richard Biener  <rguenther@suse.de>
13114         * tree.def (POLYNOMIAL_CHREC): Remove CHREC_VARIABLE tree operand.
13115         * tree-core.h (tree_base): Add chrec_var union member.
13116         * tree.h (CHREC_VAR): Remove.
13117         (CHREC_LEFT, CHREC_RIGHT, CHREC_VARIABLE): Adjust.
13118         * tree-chrec.h (build_polynomial_chrec): Adjust.
13119         * tree-chrec.c (reset_evolution_in_loop): Use build_polynomial_chrec.
13120         * tree-pretty-print.c (dump_generic_node): Use CHREC_VARIABLE.
13122 2017-10-11  Marc Glisse  <marc.glisse@inria.fr>
13124         * fold-const.c (fold_binary_loc) [X +- Y CMP X]: Move ...
13125         * match.pd: ... here.
13126         ((T) X == (T) Y): Relax condition.
13128 2017-10-11  Bin Cheng  <bin.cheng@arm.com>
13130         PR tree-optimization/82472
13131         * tree-loop-distribution.c (sort_partitions_by_post_order): Refine
13132         comment.
13133         (break_alias_scc_partitions): Update postorder number.
13135 2017-10-11  Martin Liska  <mliska@suse.cz>
13137         PR sanitizer/82490
13138         * opts.c (parse_no_sanitize_attribute): Do not use error_value
13139         variable.
13140         * opts.h (parse_no_sanitize_attribute): Remove last argument.
13142 2017-10-11  Martin Liska  <mliska@suse.cz>
13144         * print-rtl.c (print_insn): Move declaration of idbuf
13145         to same scope as name.
13147 2017-10-11  Martin Liska  <mliska@suse.cz>
13149         Revert r253637:
13151         PR sanitizer/82484
13152         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
13153         volatile arguments.
13155 2017-10-11  Martin Liska  <mliska@suse.cz>
13157         PR sanitizer/82484
13158         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
13159         volatile arguments.
13161 2017-10-11  Adhemerval Zanella  <adhemerval.zanella@linaro.org>
13163         * config.gcc (default_gnu_indirect_function): Default to yes for
13164         arm*-*-linux* with glibc.
13166 2017-10-11  Richard Biener  <rguenther@suse.de>
13168         * tree-scalar-evolution.c (get_scalar_evolution): Handle
13169         default-defs and types we do not want to analyze.
13170         (interpret_loop_phi): Replace unreachable code with an assert.
13171         (compute_scalar_evolution_in_loop): Remove and inline ...
13172         (analyze_scalar_evolution_1): ... here, replacing condition with
13173         what makes the intent clearer.  Remove handling of cases
13174         get_scalar_evolution now handles.
13176 2017-10-10  Jim Wilson  <wilson@tuliptree.org>
13178         PR rtl-optimization/81434
13179         * haifa-sched.c (prune_ready_list): Init min_cost_group to 0.  Update
13180         comment for main loop.  In sched_group_found if, also add checks for
13181         pass and min_cost_group.
13183 2017-10-10  Segher Boessenkool  <segher@kernel.crashing.org>
13185         * config/rs6000/rs6000.c (TARGET_INSN_COST): New.
13186         (rs6000_insn_cost): New function.
13187         * config/rs6000/rs6000.md (cost): New attribute.
13189 2017-10-10  Jakub Jelinek  <jakub@redhat.com>
13190             H.J. Lu  <hongjiu.lu@intel.com>
13192         PR target/79565
13193         PR target/82483
13194         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Add
13195         OPTION_MASK_ISA_MMX for __builtin_ia32_maskmovq,
13196         __builtin_ia32_vec_ext_v4hi and __builtin_ia32_vec_set_v4hi.
13197         (ix86_expand_builtin): Treat OPTION_MASK_ISA_MMX similarly
13198         to OPTION_MASK_ISA_AVX512VL - builtins that have both
13199         OPTION_MASK_ISA_MMX and some other bit set require both
13200         mmx and the ISAs without the mmx bit.
13201         * config/i386/i386-builtin.def (__builtin_ia32_cvtps2pi,
13202         __builtin_ia32_cvttps2pi, __builtin_ia32_cvtpi2ps,
13203         __builtin_ia32_pavgb, __builtin_ia32_pavgw, __builtin_ia32_pmulhuw,
13204         __builtin_ia32_pmaxub, __builtin_ia32_pmaxsw, __builtin_ia32_pminub,
13205         __builtin_ia32_pminsw, __builtin_ia32_psadbw, __builtin_ia32_pmovmskb,
13206         __builtin_ia32_pshufw, __builtin_ia32_cvtpd2pi,
13207         __builtin_ia32_cvttpd2pi, __builtin_ia32_cvtpi2pd,
13208         __builtin_ia32_pmuludq, __builtin_ia32_pabsb, __builtin_ia32_pabsw,
13209         __builtin_ia32_pabsd, __builtin_ia32_phaddw, __builtin_ia32_phaddd,
13210         __builtin_ia32_phaddsw, __builtin_ia32_phsubw, __builtin_ia32_phsubd,
13211         __builtin_ia32_phsubsw, __builtin_ia32_pmaddubsw,
13212         __builtin_ia32_pmulhrsw, __builtin_ia32_pshufb, __builtin_ia32_psignb,
13213         __builtin_ia32_psignw, __builtin_ia32_psignd, __builtin_ia32_movntq,
13214         __builtin_ia32_paddq, __builtin_ia32_psubq, __builtin_ia32_palignr):
13215         Add OPTION_MASK_ISA_MMX.
13217 2017-10-10  Andreas Tobler  <andreast@gcc.gnu.org>
13219         * config.gcc (armv7*-*-freebsd*): New target.
13220         (armv6*-*-freebsd*): Remove obsolete TARGET_FREEBSD_ARMv6 define.
13222 2017-10-10  Jan Hubicka  <hubicka@ucw.cz>
13224         * x86-tune.def (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI,
13225         X86_TUNE_ADJUST_UNROLL, X86_TUNE_ONE_IF_CONV_INSN): Move to right
13226         spot in the file.
13228 2017-10-10  Richard Sandiford  <richard.sandiford@linaro.org>
13230         * wide-int.h (wide_int_ref_storage): Make host_dependent_precision
13231         a template parameter.
13232         (WIDE_INT_REF_FOR): Update accordingly.
13233         * tree.h (wi::int_traits <const_tree>): Delete.
13234         (wi::tree_to_widest_ref, wi::tree_to_offset_ref): New typedefs.
13235         (wi::to_widest, wi::to_offset): Use them.  Expand commentary.
13236         (wi::tree_to_wide_ref): New typedef.
13237         (wi::to_wide): New function.
13238         * calls.c (get_size_range): Use wi::to_wide when operating on
13239         trees as wide_ints.
13240         * cgraph.c (cgraph_node::create_thunk): Likewise.
13241         * config/i386/i386.c (ix86_data_alignment): Likewise.
13242         (ix86_local_alignment): Likewise.
13243         * dbxout.c (stabstr_O): Likewise.
13244         * dwarf2out.c (add_scalar_info, gen_enumeration_type_die): Likewise.
13245         * expr.c (const_vector_from_tree): Likewise.
13246         * fold-const-call.c (host_size_t_cst_p, fold_const_call_1): Likewise.
13247         * fold-const.c (may_negate_without_overflow_p, negate_expr_p)
13248         (fold_negate_expr_1, int_const_binop_1, const_binop)
13249         (fold_convert_const_int_from_real, optimize_bit_field_compare)
13250         (all_ones_mask_p, sign_bit_p, unextend, extract_muldiv_1)
13251         (fold_div_compare, fold_single_bit_test, fold_plusminus_mult_expr)
13252         (pointer_may_wrap_p, expr_not_equal_to, fold_binary_loc)
13253         (fold_ternary_loc, multiple_of_p, fold_negate_const, fold_abs_const)
13254         (fold_not_const, round_up_loc): Likewise.
13255         * gimple-fold.c (gimple_fold_indirect_ref): Likewise.
13256         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Likewise.
13257         (alloca_call_type): Likewise.
13258         * gimple.c (preprocess_case_label_vec_for_gimple): Likewise.
13259         * godump.c (go_output_typedef): Likewise.
13260         * graphite-sese-to-poly.c (tree_int_to_gmp): Likewise.
13261         * internal-fn.c (get_min_precision): Likewise.
13262         * ipa-cp.c (ipcp_store_vr_results): Likewise.
13263         * ipa-polymorphic-call.c
13264         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Likewise.
13265         * ipa-prop.c (ipa_print_node_jump_functions_for_edge): Likewise.
13266         (ipa_modify_call_arguments): Likewise.
13267         * match.pd: Likewise.
13268         * omp-low.c (scan_omp_1_op, lower_omp_ordered_clauses): Likewise.
13269         * print-tree.c (print_node_brief, print_node): Likewise.
13270         * stmt.c (expand_case): Likewise.
13271         * stor-layout.c (layout_type): Likewise.
13272         * tree-affine.c (tree_to_aff_combination): Likewise.
13273         * tree-cfg.c (group_case_labels_stmt): Likewise.
13274         * tree-data-ref.c (dr_analyze_indices): Likewise.
13275         (prune_runtime_alias_test_list): Likewise.
13276         * tree-dump.c (dequeue_and_dump): Likewise.
13277         * tree-inline.c (remap_gimple_op_r, copy_tree_body_r): Likewise.
13278         * tree-predcom.c (is_inv_store_elimination_chain): Likewise.
13279         * tree-pretty-print.c (dump_generic_node): Likewise.
13280         * tree-scalar-evolution.c (iv_can_overflow_p): Likewise.
13281         (simple_iv_with_niters): Likewise.
13282         * tree-ssa-address.c (addr_for_mem_ref): Likewise.
13283         * tree-ssa-ccp.c (ccp_finalize, evaluate_stmt): Likewise.
13284         * tree-ssa-loop-ivopts.c (constant_multiple_of): Likewise.
13285         * tree-ssa-loop-niter.c (split_to_var_and_offset)
13286         (refine_value_range_using_guard, number_of_iterations_ne_max)
13287         (number_of_iterations_lt_to_ne, number_of_iterations_lt)
13288         (get_cst_init_from_scev, record_nonwrapping_iv)
13289         (scev_var_range_cant_overflow): Likewise.
13290         * tree-ssa-phiopt.c (minmax_replacement): Likewise.
13291         * tree-ssa-pre.c (compute_avail): Likewise.
13292         * tree-ssa-sccvn.c (vn_reference_fold_indirect): Likewise.
13293         (vn_reference_maybe_forwprop_address, valueized_wider_op): Likewise.
13294         * tree-ssa-structalias.c (get_constraint_for_ptr_offset): Likewise.
13295         * tree-ssa-uninit.c (is_pred_expr_subset_of): Likewise.
13296         * tree-ssanames.c (set_nonzero_bits, get_nonzero_bits): Likewise.
13297         * tree-switch-conversion.c (collect_switch_conv_info, array_value_type)
13298         (dump_case_nodes, try_switch_expansion): Likewise.
13299         * tree-vect-loop-manip.c (vect_gen_vector_loop_niters): Likewise.
13300         (vect_do_peeling): Likewise.
13301         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13302         * tree-vect-stmts.c (vectorizable_load): Likewise.
13303         * tree-vrp.c (compare_values_warnv, vrp_int_const_binop): Likewise.
13304         (zero_nonzero_bits_from_vr, ranges_from_anti_range): Likewise.
13305         (extract_range_from_binary_expr_1, adjust_range_with_scev): Likewise.
13306         (overflow_comparison_p_1, register_edge_assert_for_2): Likewise.
13307         (is_masked_range_test, find_switch_asserts, maybe_set_nonzero_bits)
13308         (vrp_evaluate_conditional_warnv_with_ops, intersect_ranges): Likewise.
13309         (range_fits_type_p, two_valued_val_range_p, vrp_finalize): Likewise.
13310         (evrp_dom_walker::before_dom_children): Likewise.
13311         * tree.c (cache_integer_cst, real_value_from_int_cst, integer_zerop)
13312         (integer_all_onesp, integer_pow2p, integer_nonzerop, tree_log2)
13313         (tree_floor_log2, tree_ctz, mem_ref_offset, tree_int_cst_sign_bit)
13314         (tree_int_cst_sgn, get_unwidened, int_fits_type_p): Likewise.
13315         (get_type_static_bounds, num_ending_zeros, drop_tree_overflow)
13316         (get_range_pos_neg): Likewise.
13317         * ubsan.c (ubsan_expand_ptr_ifn): Likewise.
13318         * config/darwin.c (darwin_mergeable_constant_section): Likewise.
13319         * config/aarch64/aarch64.c (aapcs_vfp_sub_candidate): Likewise.
13320         * config/arm/arm.c (aapcs_vfp_sub_candidate): Likewise.
13321         * config/avr/avr.c (avr_fold_builtin): Likewise.
13322         * config/bfin/bfin.c (bfin_local_alignment): Likewise.
13323         * config/msp430/msp430.c (msp430_attr): Likewise.
13324         * config/nds32/nds32.c (nds32_insert_attributes): Likewise.
13325         * config/powerpcspe/powerpcspe-c.c
13326         (altivec_resolve_overloaded_builtin): Likewise.
13327         * config/powerpcspe/powerpcspe.c (rs6000_aggregate_candidate)
13328         (rs6000_expand_ternop_builtin): Likewise.
13329         * config/rs6000/rs6000-c.c
13330         (altivec_resolve_overloaded_builtin): Likewise.
13331         * config/rs6000/rs6000.c (rs6000_aggregate_candidate): Likewise.
13332         (rs6000_expand_ternop_builtin): Likewise.
13333         * config/s390/s390.c (s390_handle_hotpatch_attribute): Likewise.
13335 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13337         * tree-vect-loop-manip.c (rename_variables_in_bb): Rename PHI nodes
13338         when copying loop nest with only one inner loop.
13340 2017-10-10  Richard Biener  <rguenther@suse.de>
13342         * tree-cfgcleanup.c (cleanup_tree_cfg_noloop): Avoid compacting
13343         blocks if SCEV is active.
13344         * tree-scalar-evolution.c (analyze_scalar_evolution_1): Remove
13345         dead code.
13346         (analyze_scalar_evolution): Handle cached evolutions the obvious way.
13347         (scev_initialize): Assert we are not yet initialized.
13349 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13351         * tree-loop-distribution.c (generate_loops_for_partition): Remove
13352         inner loop's exit stmt by making it always exit the loop, otherwise
13353         we would generate an infinite empty loop.
13355 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13357         * tree-vect-loop-manip.c (slpeel_tree_duplicate_loop_to_edge_cfg): Skip
13358         renaming variables in new preheader if it's deleted.
13360 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13362         * tree-loop-distribution.c (struct partition): Remove unused field
13363         loops of the structure.
13364         (partition_alloc, partition_free): Ditto.
13365         (build_rdg_partition_for_vertex): Ditto.
13367 2017-10-09  Jeff Law  <law@redhat.com>
13369         * targhooks.c (default_stack_clash_protection_final_dynamic_probe): Fix
13370         return type to match prototype and documentation.
13372 2010-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13374         * config/rs6000/rs6000.c (processor_costs): Move to ...
13375         * config/rs6000/rs6000.h: ... here.
13376         (rs6000_cost): Declare.
13378 2017-10-09  Eric Botcazou  <ebotcazou@adacore.com>
13380         * except.c (setjmp_fn): New global variable.
13381         (init_eh): Initialize it if DONT_USE_BUILTIN_SETJMP is defined.
13382         (sjlj_emit_function_enter): Call it instead of BUILTIN_SETJMP
13383         if DONT_USE_BUILTIN_SETJMP is defined.
13385 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13387         * target.def (insn_cost): New hook.
13388         * doc/tm.texi.in (TARGET_INSN_COST): New hook.
13389         * doc/tm.texi: Regenerate.
13390         * rtlanal.c (insn_cost): Use the new hook.
13392 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13394         * combine.c (combine_validate_cost): Compute the new insn_cost,
13395         not just pattern_cost.
13396         (try_combine): Adjust comment.
13398 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13400         * cfgrtl.c (rtl_account_profile_record): Replace insn_rtx_cost with
13401         insn_cost.
13402         * combine.c (uid_insn_cost): Adjust comment.
13403         (combine_validate_cost): Adjust comment.  Use pattern_cost instead
13404         of insn_rtx_cost
13405         (combine_instructions): Use insn_cost instead of insn_rtx_cost.
13406         * dse.c (find_shift_sequence): Ditto.
13407         * ifcvt.c (cheap_bb_rtx_cost_p): Ditto.
13408         (bb_valid_for_noce_process_p): Use pattern_cost.
13409         * rtl.h (insn_rtx_cost): Delete.
13410         (pattern_cost): New prototype.
13411         (insn_cost): New prototype.
13412         * rtlanal.c (insn_rtx_cost): Rename to...
13413         (pattern_cost): ... this.
13414         (insn_cost): New.
13416 2017-10-09  Uros Bizjak  <ubizjak@gmail.com>
13418         * config/i386/i386.md (*jcc_2): Remove insn pattern.
13419         (*jcc<mode>_0_r_i387): Ditto.
13420         (*jccxf_r_i387): Ditto.
13421         (*jcc<mode>_r_i387): Ditto.
13422         (*jccu<mode>_r_i387): Ditto.
13423         (*jcc<X87MODEF:mode>_<SWI24:mode>_r_i387): Ditto.
13424         (*jcc): Rename from *jcc_1.
13426 2017-10-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
13428         * config/rs6000/rs6000-p8swap.c (rs6000_analyze_swaps): Process
13429         deferred rescans after the lvx/stvx recombination pre-pass.
13431 2017-10-09  Michael Meissner  <meissner@linux.vnet.ibm.com>
13433         * config/rs6000/amo.h: New include file to provide ISA 3.0 atomic
13434         memory operation instruction support.
13435         * config.gcc (powerpc*-*-*): Include amo.h as an extra header.
13436         (rs6000-ibm-aix[789]*): Likewise.
13437         * doc/extend.texi (PowerPC Atomic Memory Operation Functions):
13438         Document new functions.
13440 2017-10-09  Richard Biener  <rguenther@suse.de>
13442         PR tree-optimization/82397
13443         * tree-data-ref.c (data_ref_compare_tree): Make sure to return
13444         equality only for semantically equal trees.
13446 2017-10-09  Richard Biener  <rguenther@suse.de>
13448         PR tree-optimization/82449
13449         * sese.c (scev_analyzable_p): Check whether the SCEV is linear.
13450         * tree-chrec.h (evolution_function_is_constant_p): Adjust to
13451         allow constant addresses.
13452         * tree-chrec.c (scev_is_linear_expression): Constant evolutions
13453         are linear.
13455 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13457         * config/s390/s390-builtins.def (vec_nabs, vec_vfi): Fix builtin
13458         flags.
13460 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13462         PR target/82463
13463         * config/s390/vecintrin.h (vec_madd, vec_msub): Fix macro
13464         definitions.
13466 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13468         PR target/82465
13469         * config/s390/s390-builtins.def (vec_sqrt): Fix builtin flags.
13471 2017-10-09  Jakub Jelinek  <jakub@redhat.com>
13473         PR target/82464
13474         * config/s390/s390-builtins.def (s390_vec_xor_flt_a,
13475         s390_vec_xor_flt_b, s390_vec_xor_flt_c): New.
13477 2017-10-09  Richard Sandiford  <richard.sandiford@linaro.org>
13479         * wide-int.h (WI_BINARY_OPERATOR_RESULT): New macro.
13480         (WI_BINARY_PREDICATE_RESULT): Likewise.
13481         (wi::binary_traits::operator_result): New type.
13482         (wi::binary_traits::predicate_result): Likewise.
13483         (generic_wide_int::operator~, unary generic_wide_int::operator-)
13484         (generic_wide_int::operator==, generic_wide_int::operator!=)
13485         (generic_wide_int::operator&, generic_wide_int::and_not)
13486         (generic_wide_int::operator|, generic_wide_int::or_not)
13487         (generic_wide_int::operator^, generic_wide_int::operator+
13488         (binary generic_wide_int::operator-, generic_wide_int::operator*):
13489         Delete.
13490         (operator~, unary operator-, operator==, operator!=, operator&)
13491         (operator|, operator^, operator+, binary operator-, operator*): New
13492         functions.
13493         * expr.c (get_inner_reference): Use wi::bit_and_not.
13494         * fold-const.c (fold_binary_loc): Likewise.
13495         * ipa-prop.c (ipa_compute_jump_functions_for_edge): Likewise.
13496         * tree-ssa-ccp.c (get_value_from_alignment): Likewise.
13497         (bit_value_binop): Likewise.
13498         * tree-ssa-math-opts.c (find_bswap_or_nop_load): Likewise.
13499         * tree-vrp.c (zero_nonzero_bits_from_vr): Likewise.
13500         (extract_range_from_binary_expr_1): Likewise.
13501         (masked_increment): Likewise.
13502         (simplify_bit_ops_using_ranges): Likewise.
13504 2017-10-09  Martin Jambor  <mjambor@suse.cz>
13506         PR hsa/82416
13507         * hsa-common.h (hsa_op_with_type): New method extend_int_to_32bit.
13508         * hsa-gen.c (hsa_extend_inttype_to_32bit): New function.
13509         (hsa_type_for_scalar_tree_type): Use it.  Always force min32int for
13510         COMPLEX types.
13511         (hsa_fixup_mov_insn_type): New function.
13512         (hsa_op_with_type::get_in_type): Use it.
13513         (hsa_build_append_simple_mov): Likewise.  Allow sub-32bit
13514         immediates in an assert.
13515         (hsa_op_with_type::extend_int_to_32bit): New method.
13516         (gen_hsa_insns_for_bitfield): Fixup instruction and intermediary
13517         types.  Convert to dest type if necessary.
13518         (gen_hsa_insns_for_bitfield_load): Fixup load type if necessary.
13519         (reg_for_gimple_ssa): Pass false as min32int to
13520         hsa_type_for_scalar_tree_type.
13521         (gen_hsa_addr): Fixup type when creating addresable temporary.
13522         (gen_hsa_cmp_insn_from_gimple): Extend operands if necessary.
13523         (gen_hsa_unary_operation): Extend operands and convert to dest type if
13524         necessary.  Call hsa_fixup_mov_insn_type.
13525         (gen_hsa_binary_operation): Changed operand types to hsa_op_with_type,
13526         extend operands and convert to dest type if necessary.
13527         (gen_hsa_insns_for_operation_assignment): Extend operands and convert
13528         to dest type if necessary.
13529         (set_output_in_type): Call hsa_fixup_mov_insn_type.  Just ude dest
13530         if conversion nt necessary and size matches.
13531         (gen_hsa_insns_for_load): Call hsa_fixup_mov_insn_type, convert
13532         to dest type if necessary.
13533         (gen_hsa_insns_for_store): Call hsa_fixup_mov_insn_type.
13534         (gen_hsa_insns_for_switch_stmt): Likewise. Also extend operands if
13535         necessary.
13536         (gen_hsa_clrsb): Likewise.
13537         (gen_hsa_ffs): Likewise.
13538         (gen_hsa_divmod): Extend operands and convert to dest type if
13539         necessary.
13540         (gen_hsa_atomic_for_builtin): Change type of op to hsa_op_with_type.
13542 2017-10-08  Segher Boessenkool  <segher@kernel.crashing.org>
13544         * config/rs6000/rs6000.md (conditional branch): Clean up formatting.
13545         Remove empty default arguments.  Use a brace block as output
13546         statement.
13547         (conditional return): Ditto.
13548         (jump): Ditto.
13549         (indirect_jump): Ditto.  Use b%T0 instead of bctr/blr.
13550         (tablejump, tablejumpsi, tablejumpdi, *tablejump<mode>_internal1):
13551         Ditto.
13552         (group_ending_nop): Ditto.
13553         (doloop_end): Ditto.
13554         (ctr<mode>, ctr<mode>_internal1, ctr<mode>_internal2): Ditto.
13555         (splitters for those): Ditto.
13557 2017-10-08  Segher Boessenkool  <segher@kernel.crashing.org>
13559         * config/rs6000/rs6000-string.c (expand_strncmp_align_check): Invert
13560         a conditional jump (and the compare for it) so that pc_rtx is the
13561         last operand.
13562         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Adjust
13563         for the deleted and renamed ctr<mode>_internal[234] patterns.
13564         * config/rs6000/rs6000.md: Delete second conditional branch pattern.
13565         Delete second conditional return pattern.
13566         (ctr<mode>_internal2): Delete this second bdnz pattern.
13567         (ctr<mode>_internal3): Rename to ctr<mode>_internal2.
13568         (ctr<mode>_internal4): Delete this second bdz pattern.
13570 2017-10-08  Eric Botcazou  <ebotcazou@adacore.com>
13572         * tree-outof-ssa.h (ssaexpand): Add partitions_for_undefined_values.
13573         (always_initialized_rtx_for_ssa_name_p): New predicate.
13574         * tree-outof-ssa.c (remove_ssa_form): Initialize new field of SA.
13575         (finish_out_of_ssa): Free new field of SA.
13576         * tree-ssa-coalesce.h (get_undefined_value_partitions): Declare.
13577         * tree-ssa-coalesce.c: Include tree-ssa.h.
13578         (get_parm_default_def_partitions): Remove extern keyword.
13579         (get_undefined_value_partitions): New function.
13580         * expr.c (expand_expr_real_1) <expand_decl_rtl>: For a SSA_NAME, do
13581         not set SUBREG_PROMOTED_VAR_P on the sub-register if it may contain
13582         uninitialized bits.
13583         * loop-iv.c (iv_get_reaching_def): Disqualify all subregs.
13585 2017-10-08  Eric Botcazou  <ebotcazou@adacore.com>
13587         * builtins.def (BUILT_IN_SETJMP): Revert latest change.
13589 2017-10-08  Jan Hubicka  <hubicka@ucw.cz>
13591         * config/i386/i386.c (ix86_expand_set_or_movmem): Disable 512bit loops
13592         for targets that preffer 128bit.
13594 2017-10-08  Jan Hubicka  <hubicka@ucw.cz>
13596         * config/i386/i386.c (has_dispatch): Disable for Ryzen.
13598 2017-10-08  Olivier Hainque  <hainque@adacore.com>
13600         * config/arm/arm.c (arm_set_return_address): Use MEM_VOLATILE_P
13601         on the target mem instead of RTX_FRAME_RELATED_P on the insn to
13602         prevent DSE.
13603         (thumb_set_return_address): Likewise.
13605 2017-10-08  Olivier Hainque  <hainque@adacore.com>
13607         * common/config/arm/arm-common.c (arm_except_unwind_info):
13608         Handle DWARF2_UNWIND_INFO.
13610 2017-10-07  Michael Collison <michael.collison@arm.com>
13612         * config/aarch64/aarch64.md (*aarch64_reg_<optab>_minus<mode>3):
13613         New pattern.
13615 2017-10-07  Eric Botcazou  <ebotcazou@adacore.com>
13617         * builtins.def (BUILT_IN_SETJMP): Declare as library builtin instead
13618         of GCC builtin if DONT_USE_BUILTIN_SETJMP is defined.
13619         * except.c (sjlj_emit_function_enter): If DONT_USE_BUILTIN_SETJMP is
13620         defined, force the creation of a new block for a dispatch label.
13622 2017-10-07  Jan Hubicka  <hubicka@ucw.cz>
13624         * invoke.texi (Wsuggest-attribute=cold): Document.
13625         * common.opt (Wsuggest-attribute=cold): New
13626         * ipa-pure-const.c (warn_function_cold): New function.
13627         * predict.c (compute_function_frequency): Use it.
13628         * predict.h (warn_function_cold): Declare.
13630 2017-10-06  Jan Hubicka  <hubicka@ucw.cz>
13632         * tree-switch-conversion.c (do_jump_if_equal, emit_cmp_and_jump_insns):
13633         Update profile.
13635 2017-10-06  Martin Liska  <mliska@suse.cz>
13637         * sanopt.c (struct sanopt_tree_triplet_hash): Remove inline
13638         keyword for member functions.
13639         (struct sanopt_tree_couple): New struct.
13640         (struct sanopt_tree_couple_hash): New function.
13641         (struct sanopt_ctx): Add new hash_map.
13642         (has_dominating_ubsan_ptr_check): New function.
13643         (record_ubsan_ptr_check_stmt): Likewise.
13644         (maybe_optimize_ubsan_ptr_ifn): Likewise.
13645         (sanopt_optimize_walker): Handle IFN_UBSAN_PTR.
13646         (pass_sanopt::execute): Handle also SANITIZE_POINTER_OVERFLOW.
13648 2017-10-06  Sudakshina Das  <sudi.das@arm.com>
13650         PR target/82440
13651         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Only call
13652         aarch64_simd_valid_immediate on CONST_VECTORs.
13653         (aarch64_reg_or_bic_imm): Likewise.
13655 2017-10-06  Wilco Dijkstra  <wdijkstr@arm.com>
13657         PR rtl-optimization/82396
13658         * haifa-sched.c (ready_sort_real): Disable qsort checking.
13660 2017-10-06  Sebastian Pop  <sebpop@gmail.com>
13662         * graphite-dependences.c (scop_get_reads): Move code to...
13663         (scop_get_must_writes): Move code to...
13664         (scop_get_may_writes): Move code to...
13665         (scop_get_reads_and_writes): ... here.
13666         (scop_get_dependences): Call scop_get_reads_and_writes.
13668 2017-10-06  Jakub Jelinek  <jakub@redhat.com>
13670         PR tree-optimization/82434
13671         * fold-const.h (can_native_encode_type_p,
13672         can_native_encode_string_p): Remove.
13673         * fold-const.c (native_encode_int): Formatting fixes.  If ptr is NULL,
13674         don't encode anything, just return what would be otherwise returned.
13675         (native_encode_fixed, native_encode_complex, native_encode_vector):
13676         Likewise.
13677         (native_encode_string): Likewise.  Inline by hand
13678         can_native_encode_string_p.
13679         (can_native_encode_type_p): Remove.
13680         (can_native_encode_string_p): Remove.
13681         * tree-vect-stmts.c (vectorizable_store): Instead of testing just
13682         STRING_CSTs using can_native_encode_string_p, test all
13683         CONSTANT_CLASS_P values using native_encode_expr with NULL ptr.
13684         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Remove last
13685         argument from native_encode_expr.
13686         (rhs_valid_for_store_merging_p): Use native_encode_expr with NULL ptr.
13687         (pass_store_merging::execute): Don't unnecessarily look for 3 stmts,
13688         but just 2.
13690 2017-10-06  Richard Biener  <rguenther@suse.de>
13692         PR tree-optimization/82397
13693         * tree-vect-data-refs.c (dr_group_sort_cmp): Do not use
13694         operand_equal_p but rely on data_ref_compare_tree for detecting
13695         equalities.
13696         (vect_analyze_data_ref_accesses): Use data_ref_compare_tree
13697         to match up with dr_group_sort_cmp.
13699 2017-10-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13701         PR target/82322
13702         * config/s390/s390-builtins.def (s390_vfi): Define new overloaded
13703         builtin.
13704         * config/s390/s390-builtin-types.def: Regenerate.
13706 2017-10-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13708         PR target/82317
13709         * config/s390/s390-builtin-types.def: Regenerate.
13710         * config/s390/s390-builtins.def (s390_vfmaxdb_4, s390_vfmindb_4):
13711         Change flag from B_VXE to B_VX.
13712         (s390_vec_min_dbl): Remove B_VXE flag.
13714 2017-10-06  Richard Biener  <rguenther@suse.de>
13716         * graphite-isl-ast-to-gimple.c: Include ssa.h and tree-ssa.h.
13717         (translate_isl_ast_to_gimple::translate_pending_phi_nodes,
13718         translate_isl_ast_to_gimple::is_valid_rename,
13719         translate_isl_ast_to_gimple::get_rename,
13720         translate_isl_ast_to_gimple::get_def_bb_for_const,
13721         translate_isl_ast_to_gimple::get_new_name,
13722         translate_isl_ast_to_gimple::collect_all_ssa_names,
13723         translate_isl_ast_to_gimple::copy_loop_phi_args,
13724         translate_isl_ast_to_gimple::collect_all_ssa_names,
13725         translate_isl_ast_to_gimple::copy_loop_phi_args,
13726         translate_isl_ast_to_gimple::copy_loop_phi_nodes,
13727         translate_isl_ast_to_gimple::add_close_phis_to_merge_points,
13728         translate_isl_ast_to_gimple::add_close_phis_to_outer_loops,
13729         translate_isl_ast_to_gimple::copy_loop_close_phi_args,
13730         translate_isl_ast_to_gimple::copy_loop_close_phi_nodes,
13731         translate_isl_ast_to_gimple::copy_cond_phi_args,
13732         translate_isl_ast_to_gimple::copy_cond_phi_nodes,
13733         translate_isl_ast_to_gimple::edge_for_new_close_phis,
13734         translate_isl_ast_to_gimple::add_phi_arg_for_new_expr,
13735         translate_isl_ast_to_gimple::rename_uses,
13736         translate_isl_ast_to_gimple::rename_all_uses): Remove.
13737         (translate_isl_ast_to_gimple::get_rename_from_scev): Simplify.
13738         (set_rename_for_each_def): Likewise.
13739         (graphite_copy_stmts_from_block): Handle debug stmt resetting
13740         here.  Handle rewriting SCEV analyzable uses here.
13741         (copy_bb_and_scalar_dependences): Generate code for PHI
13742         copy-in/outs.
13743         (graphite_regenerate_ast_isl): Adjust.
13744         * graphite-scop-detection.c (trivially_empty_bb_p): Move to sese.[ch].
13745         (add_write, add_read): New functions.
13746         (build_cross_bb_scalars_def): Use it and simplify.
13747         (build_cross_bb_scalars_use): Likewise.
13748         (graphite_find_cross_bb_scalar_vars): Inline into...
13749         (try_generate_gimple_bb): ...here.  Add dependences for PHIs,
13750         simulating out-of-SSA.  Compute liveout and add dependencies.
13751         (build_scops): Force an empty entry block.
13752         * sese.h (sese_info_t::liveout, sese_info_t::debug_liveout): New
13753         members.
13754         (sese_build_liveouts): Declare.
13755         (sese_trivially_empty_bb_p): Likewise.
13756         * sese.c (sese_build_liveouts_bb): Properly handle PHIs,
13757         compute liveout and debug_liveout.
13758         (sese_bad_liveouts_use): Remove.
13759         (sese_reset_debug_liveouts_bb): Likewise.
13760         (sese_reset_debug_liveouts): Rewrite in terms of debug_liveout.
13761         (sese_build_liveouts): Build liveout and debug_liveout and store
13762         it in region.
13763         (new_sese_info): Adjust.
13764         (free_sese_info): Likewise.
13765         (sese_insert_phis_for_liveouts): Reset debug stmts from here,
13766         do not build liveout here.
13767         (move_sese_in_condition): Adjust region entry.
13768         (scev_analyzable_p): Match up with chrec_apply requirements.
13769         (sese_trivially_empty_bb_p): New.
13770         * tree-into-ssa.c (get_reaching_def): Properly support generating
13771         default-defs for incremental rewrite of anonymous names.
13773 2017-10-06  Richard Biener  <rguenther@suse.de>
13775         * graphite-sese-to-poly.c (extract_affine): For casts increasing
13776         precision do not perform modulo reduction.
13778 2017-10-06  Richard Biener  <rguenther@suse.de>
13780         PR tree-optimization/82436
13781         * tree-vect-slp.c (vect_supported_load_permutation_p): More
13782         conservatively choose the vectorization factor when checking
13783         whether we can perform the required load permutation.
13784         (vect_transform_slp_perm_load): Assert when we may not fail.
13786 2017-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
13788         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Correct error
13789         message for incompatible -msdata=* and -mcall-* options.
13791 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13793         * config/i386/i386.c (ia32_multipass_dfa_lookahead): Default to issue
13794         rate for post-reload scheduling.
13796 2017-10-05  Tamar Christina  <tamar.christina@arm.com>
13798         * doc/sourcebuild.texi (vect_sizes_16B_8B, vect_sizes_32B_16B): New.
13800 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13802         * config/i386/i386.c (znver1_cost): Set branch_cost to 3 (instead of 2)
13803         to improve monte carlo in scimark.
13805 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13807         * config/i386/i386.c (ix86_size_cost, i386_cost, i486_cost,
13808         pentium_cost, lakemont_cost, pentiumpro_cost, geode_cost, k6_cost,
13809         athlon_cost, k8_cost, amdfam10_cost, btver1_cost, btver2_cost,
13810         pentium4_cost, nocona_cost): Set reassociation width to 1.
13811         (bdver1_cost, bdver2_cost, bdver3_cost, bdver4_cost): Set reassociation
13812         width to 2 for fp operations and 1 otherwise.
13813         (znver1_cost): Set scalar reassoc width to 4 and vector to 3 and 6
13814         for int and fp.
13815         (atom_cost): Set reassociation width to 2.
13816         (slm_cost, generic_cost): Set fp reassociation width
13817         to 2 and 1 otherwise.
13818         (intel_cost): Set fp reassociation width to 4 and 1 otherwise.
13819         (core_cost): Set fp reassociation width to 4 and vector to 2.
13820         (ix86_reassociation_width): Rewrite using cost table; special case
13821         plus/minus on Zen; honor X86_TUNE_SSE_SPLIT_REGS
13822         and TARGET_AVX128_OPTIMAL.
13823         * config/i386/i386.h (processor_costs): Add
13824         reassoc_int, reassoc_fp, reassoc_vec_int, reassoc_vec_fp.
13825         (TARGET_VECTOR_PARALLEL_EXECUTION, TARGET_REASSOC_INT_TO_PARALLEL,
13826         TARGET_REASSOC_FP_TO_PARALLEL): Remove.
13827         * x86-tune.def (X86_TUNE_REASSOC_INT_TO_PARALLEL): Remove.
13828         (X86_TUNE_REASSOC_FP_TO_PARALLEL): Remove.
13829         (X86_TUNE_VECTOR_PARALLEL_EXECUTION):  Remove.
13831 2017-10-05  Nathan Sidwell  <nathan@acm.org>
13833         * doc/invoke.texi (Wparentheses): Document C++ MVP behaviour.
13835 2017-10-05  Tamar Christina  <tamar.christina@arm.com>
13837         * config/arm/arm.c (arm_test_fpu_data): New.
13838         (arm_run_selftests): Call arm_test_fpu_data.
13840 2017-10-04  Nathan Sidwell  <nathan@acm.org>
13842         * toplev.c (toplev::main): Remove excess parens on pretty_printer
13843         decl.
13844         * caller-save.c (insert_save): Remove excess parens on TO_SAVE parm.
13846 2017-10-04  Sudakshina Das  <sudi.das@arm.com>
13848         * config/aarch64/aarch64-protos.h (enum simd_immediate_check): New
13849         check type for aarch64_simd_valid_immediate.
13850         (aarch64_output_simd_mov_immediate): Update prototype.
13851         (aarch64_simd_valid_immediate): Update prototype.
13852         * config/aarch64/aarch64-simd.md (orr<mode>3): modified pattern to add
13853         support for ORR-immediate.
13854         (and<mode>3): modified pattern to add support for BIC-immediate.
13855         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Function
13856         now checks for valid immediate for BIC and ORR based on new enum
13857         argument.
13858         (aarch64_output_simd_mov_immediate): Function now used to output
13859         BIC/ORR imm as well based on new enum argument.
13860         * config/aarch64/constraints.md (Do): New vector immediate constraint.
13861         (Db) : Likewise.
13862         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): New predicate.
13863         (aarch64_reg_or_bic_imm): Likewise.
13865 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13867         * config/s390/vx-builtins.md ("vec_mergeh<mode>")
13868         ("vec_mergel<mode>"): Change mode iterator to V_128_NOSINGLE.
13870 2017-10-04  Wilco Dijkstra  <wdijkstr@arm.com>
13872         Revert r253399:
13874         PR rtl-optimization/82396
13875         * haifa-sched.c (autopref_multipass_init): Simplify
13876         initialization.
13877         (autopref_rank_data): Simplify sort order.
13878         * sched-int.h (autopref_multipass_data_): Remove
13879         multi_mem_insn_p, min_offset and max_offset.
13881 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13883         * doc/sourcebuild.texi: Document vect_peeling_profitable.
13885 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13887         * doc/sourcebuild.texi: Document vect_intdouble_cvt and
13888         vect_doubleint_cvt.
13890 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13892         * doc/sourcebuild.texi: Document vect_long_mult.
13894 2017-10-04  Richard Sandiford  <richard.sandiford@linaro.org>
13896         PR tree-optimization/82413
13897         * fold-const.c (build_range_check): Use widest_int when comparing
13898         the maximum ETYPE value with HIGH.
13900 2017-10-04  Wilco Dijkstra  <wdijkstr@arm.com>
13902         PR rtl-optimization/82396
13903         * haifa-sched.c (autopref_multipass_init): Simplify
13904         initialization.
13905         (autopref_rank_data): Simplify sort order.
13906         * sched-int.h (autopref_multipass_data_): Remove
13907         multi_mem_insn_p, min_offset and max_offset.
13909 2017-10-04  Jakub Jelinek  <jakub@redhat.com>
13911         PR tree-optimization/82381
13912         * tree-ssa-reassoc.c (sort_by_operand_rank): Check for different
13913         oeN->rank first.  Return 1 or -1 if one op is SSA_NAME and the other
13914         is not.
13916         PR tree-optimization/82374
13917         * omp-low.c (create_omp_child_function): Copy DECL_ATTRIBUTES,
13918         DECL_FUNCTION_SPECIFIC_OPTIMIZATION,
13919         DECL_FUNCTION_SPECIFIC_TARGET and DECL_FUNCTION_VERSIONED from
13920         current_function_decl to the new decl.
13922 2017-10-03  Michael Meissner  <meissner@linux.vnet.ibm.com>
13924         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_2_HW): Define new
13925         helper macro for IEEE float128 hardware built-in functions.
13926         (SQRTF128_ODD): Add built-in functions with the round-to-odd
13927         semantics.
13928         (TRUNCF128_ODD): Likewise.
13929         (ADDF128_ODD): Likewise.
13930         (SUBF128_ODD): Likewise.
13931         (MULF128_ODD): Likewise.
13932         (DIVF128_ODD): Likewise.
13933         (FMAF128_ODD): Likewise.
13934         * config/rs6000/rs6000.md (UNSPEC_ROUND_TO_ODD): Rename to
13935         UNSPEC_TRUNC_ROUND_TO_ODD.
13936         (UNSPEC_TRUNC_ROUND_TO_ODD): Likewise.
13937         (UNSPEC_ADD_ROUND_TO_ODD): New unspec codes for the IEEE 128-bit
13938         floating point round to odd instructions.
13939         (UNSPEC_SUB_ROUND_TO_ODD): Likewise.
13940         (UNSPEC_MUL_ROUND_TO_ODD): Likewise.
13941         (UNSPEC_DIV_ROUND_TO_ODD): Likewise.
13942         (UNSPEC_FMA_ROUND_TO_ODD): Likewise.
13943         (UNSPEC_SQRT_ROUND_TO_ODD): Likewise.
13944         (trunc<mode>sf2_hw): Change the truncate with round to odd
13945         expansion to use UNSPEC_TRUNC_ROUND_TO_ODD.
13946         (add<mode>3_odd): Add insns for IEEE 128-bit floating point round
13947         to odd hardware instructions.
13948         (sub<mode>3_odd): Likewise.
13949         (mul<mode>3_odd): Likewise.
13950         (div<mode>3_odd): Likewise.
13951         (sqrt<mode>2_odd): Likewise.
13952         (fma<mode>4_odd): Likewise.
13953         (fms<mode>4_odd): Likewise.
13954         (nfma<mode>4_odd): Likewise.
13955         (nfms<mode>4_odd): Likewise.
13956         (trunc<mode>df2_odd): Change the truncate with round to odd
13957         expansion to use UNSPEC_TRUNC_ROUND_TO_ODD.  Add a generator
13958         function.
13959         * doc/extend.texi (PowerPC built-in functions): Update documentation
13960         for existing IEEE float128-bit built-in functions.  Add built-in
13961         functions that generate the IEEE 128-bit floating point round to
13962         odd instructions.
13964 2017-10-03  Segher Boessenkool  <segher@kernel.crashing.org>
13966         PR rtl-optimization/77729
13967         * simplify-rtx.c (simplify_binary_operation_1): Delete the (X&C1)|C2
13968         to (X&(C1&~C2))|C2 transformations.
13970 2017-10-03  Martin Jambor  <mjambor@suse.cz>
13972         PR tree-optimization/82363
13973         * tree-sra.c (propagate_subaccesses_across_link): In unrecoverable
13974         mismatch, mark lacc written regardless of racc.
13976 2017-10-03  Jakub Jelinek  <jakub@redhat.com>
13978         PR tree-optimization/82381
13979         * tree-ssa-reassoc.c (sort_by_operand_rank): Don't check
13980         stmt_to_insert nor wheather SSA_NAMEs are default defs.
13981         Return 1 or -1 if one of bba and bbb is NULL. If bb_rank is equal,
13982         fallthrough into reassoc_stmt_dominates_stmt_p.
13984         PR target/82386
13985         * combine.c (combine_instructions): Don't combine in unreachable
13986         basic blocks.
13988 2017-10-02  Peter Bergner  <bergner@vnet.ibm.com>
13990         PR target/80210
13991         * config/rs6000/rs6000.c (rs6000_option_override_internal): Rewrite
13992         function to not use the have_cpu variable.  Do not set cpu_index,
13993         rs6000_cpu_index or rs6000_tune_index if we end up using TARGET_DEFAULT
13994         or the default cpu.
13995         (rs6000_valid_attribute_p): Remove duplicate initializations of
13996         old_optimize and func_optimize.
13997         (rs6000_pragma_target_parse): Call rs6000_activate_target_options ().
13998         (rs6000_activate_target_options): Make global.
13999         * config/rs6000/rs6000-protos.h (rs6000_activate_target_options): Add
14000         prototype.
14002 2017-10-02  Jakub Jelinek  <jakub@redhat.com>
14004         * tree-dfa.c (get_ref_base_and_extent): Set *pmax_size to -1
14005         if *poffset + *pmax_size overflows in HOST_WIDE_INT.
14006         Set *poffset to 0 and *psize and *pmax_size to -1 if
14007         *poffset + *psize overflows in HOST_WIDE_INT.
14009         PR tree-optimization/82387
14010         PR tree-optimization/82388
14011         PR tree-optimization/82389
14012         * tree-ssa-dse.c (dse_classify_store): Test byte_tracking_enabled
14013         instead of live_bytes non-NULL.
14015 2017-10-02  Georg-Johann Lay  <avr@gjlay.de>
14017         PR target/41076
14018         * confg/avr/avr.md (*iorhi3.ashift8-ext.zerox): Add "r,r,0"
14019         alternative.
14021 2017-10-02  Richard Biener  <rguenther@suse.de>
14023         * graphite-isl-ast-to-gimple.c (set_codegen_error): With
14024         -fchecking and --param graphite-allow-codegen-errors=0 ICE.
14025         * params.def (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): New param.
14027 2017-10-02  Richard Sandiford  <richard.sandiford@linaro.org>
14029         * tree.h (wi::int_traits <const_tree>::decompose): Assert that the
14030         requested precision matches the type's.
14031         * calls.c (alloc_max_size): Calculate the new candidate size as
14032         a widest_int and use wi::to_widest when comparing it with the
14033         current candidate size.
14034         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Compare with
14035         zero rather than integer_zero_node.
14036         * match.pd: Check for a no-op conversion before using wi::add
14037         rather than after.  Use tree_to_uhwi when summing small shift
14038         counts into an unsigned int.
14040 2017-10-02  Richard Sandiford  <richard.sandiford@linaro.org>
14041             Alan Hayward  <alan.hayward@arm.com>
14042             David Sherwood  <david.sherwood@arm.com>
14044         PR target/71307
14045         * config/aarch64/aarch64.h (POINTER_AND_FP_REGS): New reg class.
14046         (REG_CLASS_NAMES, REG_CLASS_CONTENTS): Update accordingly.
14047         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
14048         POINTER_AND_FP_REGS.
14050 2017-10-02  Richard Biener  <rguenther@suse.de>
14052         PR tree-optimization/82355
14053         * graphite-isl-ast-to-gimple.c (build_iv_mapping): Also build
14054         a mapping for the enclosing loop but avoid generating one for
14055         the loop tree root.
14056         (copy_bb_and_scalar_dependences): Remove premature codegen
14057         error on PHIs in blocks duplicated into multiple places.
14058         * graphite-scop-detection.c
14059         (scop_detection::stmt_has_simple_data_refs_p): For a loop not
14060         in the region use it as loop and nest to analyze the DR in.
14061         (try_generate_gimple_bb): Likewise.
14062         * graphite-sese-to-poly.c (extract_affine_chrec): Adjust.
14063         (add_loop_constraints): For blocks in a loop not in the region
14064         create a dimension with a single iteration.
14065         * sese.h (gbb_loop_at_index): Remove assert.
14067 2017-10-01  Kevin Buettner  <kevinb@redhat.com>
14069         * omp-expand.c (adjust_context_scope): New function.
14070         (expand_parallel_call): Call adjust_context_scope.
14072 2017-10-01  Jeff Law  <law@redhat.com>
14074         * tree-ssa-dom.c (optimize_stmt): Make this a method within the
14075         dom_opt_dom_walker class with direct access to private members.
14076         Add comments.  Call test_for_singularity.
14077         (dom_opt_dom_walker::before_dom_children): Corresponding changes.
14078         (dom_opt_dom_walker::after_dom_children): Do not lazily initialize
14079         m_dummy_cond anymore.
14080         (class dom_opt_dom_walker): Initialize m_dummy_cond member in the
14081         class ctor.
14082         (pass_dominator:execute): Build the dummy_cond here and pass it
14083         to the dom_opt_dom_walker ctor.
14084         (test_for_singularity): New function.
14086 2017-09-30  Krister Walfridsson  <krister.walfridsson@gmail.com>
14087             Maya Rashish  <coypu@sdf.org>
14089         * config.gcc (*-*-netbsd*): New variable nbsd_tm_file containing
14090         netbsd.h, netbsd-stdint.h, and netbsd-elf.h.
14091         (alpha*-*-netbsd*) Use nbsd_tm_file.
14092         (arm*-*-netbsdelf*) Likewise.
14093         (i[34567]86-*-netbsdelf*) Likewise.
14094         (x86_64-*-netbsd*) Likewise.
14095         (mips*-*-netbsd*) Likewise.
14096         (powerpc-*-netbsd*) Likewise.
14097         (sh*-*-netbsd*) Likewise.
14098         (sparc-*-netbsdelf*) Likewise.
14099         (sparc64-*-netbsd*) Likewise.
14100         (m68k*-*-netbsdelf*) Use nbsd_tm_file and add CHAR_FAST8/SHORT_FAST16
14101         to tm_defines.
14102         (vax-*-netbsdelf*) Likewise.
14103         * config/netbsd-stdint.h (INT_FAST8_TYPE): Check CHAR_FAST8.
14104         (UINT_FAST8_TYPE) Likewise.
14105         (INT_FAST16_TYPE) Check CHAR_FAST16.
14106         (UINT_FAST16_TYPE) Likewise.
14108 2017-09-30  Jakub Jelinek  <jakub@redhat.com>
14110         PR target/82361
14111         * config/i386/i386.md
14112         (TARGET_USE_8BIT_IDIV zext divmodsi4 splitter): New define_split.
14113         (divmodsi4_zext_1, divmodsi4_zext_2, *divmodsi4_zext_1,
14114         *divmodsi4_zext_2): New define_insn_and_split.
14115         (*divmodsi4_noext_zext_1, *divmodsi4_noext_zext_2): New define_insn.
14116         (TARGET_USE_8BIT_IDIV zext udivmodsi4 splitter): New define_split.
14117         (udivmodsi4_zext_1, udivmodsi4_zext_2, *udivmodsi4_zext_1,
14118         *udivmodsi4_zext_2, *udivmodsi4_pow2_zext_1, *udivmodsi4_pow2_zext_2):
14119         New define_insn_and_split.
14120         (*udivmodsi4_noext_zext_1, *udivmodsi4_noext_zext_2): New define_insn.
14121         * config/i386/i386.c (ix86_split_idivmod): Handle operands[0] or
14122         operands[1] having DImode when mode is SImode.
14124         * config/i386/i386.c (ix86_split_idivmod): Use mode instead of
14125         always SImode for DIV and MOD in REG_EQUAL notes.
14127 2017-09-29  Yury Gribov  <tetra2005@gmail.com>
14129         PR middle-end/82319
14130         * match.pd: Fix handling of NaNs in pattern.
14132 2017-09-29  Jeff Law  <law@redhat.com>
14134         * sbitmap.c (bitmap_bit_in_range_p): New function.
14135         * sbitmap.h (bitmap_bit_in_range_p): Prototype.
14136         * tree-ssa-dse.c (live_bytes_read): New function.
14137         (dse_classify_store): Ignore reads of dead bytes.
14139         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Fix
14140         typos and whitespace errors.
14141         * config/i386/predicates.md (address_no_seg_operand): Likewise.
14142         * config/s390/s390.c (s390_emit_prologue): Likewise.
14144 2017-09-29  Vladimir Makarov  <vmakarov@redhat.com>
14146         PR target/81481
14147         * ira-costs.c (scan_one_insn): Don't take into account PIC equiv
14148         with a symbol for LRA.
14150 2017-09-29  Vladimir Makarov  <vmakarov@redhat.com>
14152         PR rtl-optimization/82338
14153         * lra-constraints.c (inherit_in_ebb): Check usage_insns check.
14155 2017-09-29  Alexander Monakov  <amonakov@ispras.ru>
14157         * genmodes.c (calc_wider_mode): Suppress qsort macro.
14158         * system.h [CHECKING_P] (qsort): Redirect to qsort_chk.
14159         (qsort_chk): Declare.
14160         * vec.c [CHECKING_P] (qsort_chk_error): New static function.
14161         (qsort_chk): New function.
14163 2017-09-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
14165         PR tree-optimization/82337
14166         * gimple-ssa-strength-reduction.c (find_phi_def): Don't record a
14167         phi definition if the PHI result appears in an abnormal PHI.
14168         (find_basis_for_base_expr): Don't record a basis if the LHS of the
14169         basis appears in an abnormal PHI.
14171 2017-09-29  Richard Biener  <rguenther@suse.de>
14173         * graphite-isl-ast-to-gimple.c
14174         (translate_isl_ast_to_gimple::set_codegen_error): New function.
14175         (binary_op_to_tree): Use it.
14176         (get_rename_from_scev): Likewise.
14177         (copy_loop_phi_nodes): Likewise.
14178         (copy_bb_and_scalar_dependences): Likewise.
14179         (translate_pending_phi_nodes): Likewise.
14181 2017-09-29  Jakub Jelinek  <jakub@redhat.com>
14183         PR target/82339
14184         * config/i386/i386.md (*movdi_internal peephole2): New -Os peephole
14185         for movabsq $(i32 << shift), r64.
14187 2017-09-28  Uros Bizjak  <ubizjak@gmail.com>
14189         * config/i386/i386.c (ix86_print_operand_address_as): Do not check
14190         index when encoding %esp as %rsp to avoid 0x67 prefix.
14192 2017-09-28  Sergey Shalnov  <Sergey.Shalnov@intel.com>
14194         * config/i386/i386.md (*movsf_internal, *movdf_internal):
14195         Return 256-bit AVX modes for TARGET_PREFER_AVX256.
14197 2017-09-28  Thomas Preud'homme  <thomas.preudhomme@arm.com>
14199         * config/arm/arm.c (arm_option_override): Forbid ARMv8-M Security
14200         Extensions with more than 16 double VFP registers.
14201         (cmse_nonsecure_entry_clear_before_return): Remove second entry of
14202         to_clear_mask and all code related to it.  Replace the remaining
14203         entry by a sbitmap and adapt code accordingly.
14205 2017-09-28  Henry Linjamäki  <henry.linjamaki@parmance.com>
14207         * brig-builtins.def: Change pure attributes to const.
14209 2017-09-28  Joseph Myers  <joseph@codesourcery.com>
14211         * config.gcc (default_gnu_indirect_function): Default to yes for
14212         sparc*-*-linux* with glibc.
14214 2017-09-28  Joseph Myers  <joseph@codesourcery.com>
14216         * config/aarch64/aarch64.c (aarch64_elf_asm_constructor)
14217         (aarch64_elf_asm_destructor): Pass SECTION_NOTYPE to get_section
14218         when creating .init_array and .fini_array sections with priority
14219         specified.
14221 2017-09-27  Christophe Lyon  <christophe.lyon@linaro.org>
14223         PR target/71727
14224         * config/aarch64/aarch64.c
14225         (aarch64_builtin_support_vector_misalignment): Always return false
14226         when misalignment is unknown.
14228 2017-09-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14230         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Revise
14231         this function to return false if the definition used by the swap
14232         instruction is artificial, or if the memory address from which the
14233         constant value is loaded is not represented by a base address held
14234         in a register or if the base address register is a frame or stack
14235         pointer.  Additionally, return false if the base address of the
14236         loaded constant is a SYMBOL_REF but is not considered to be a
14237         constant.
14238         (replace_swapped_load_constant): New function.
14239         (rs6000_analyze_swaps): Add a new pass to replace a swap of a
14240         loaded constant vector with a load of a swapped constant vector.
14242 2017-09-27  Carl Love  <cel@us.ibm.com>
14244         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
14245         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
14246         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
14247         fctiw instruction.
14249 2017-09-27  Alexander Monakov  <amonakov@ispras.ru>
14251         * haifa-sched.c (autopref_rank_for_schedule): Order 'irrelevant' insns
14252         first, always call autopref_rank_data otherwise.
14254 2017-09-27  Richard Biener  <rguenther@suse.de>
14256         * graphite-scop-detection.c (find_scop_parameters): Move
14257         loop bound handling ...
14258         (gather_bbs::before_dom_children): ... here, avoiding the need
14259         to build scop_info->loop_nest.
14260         (record_loop_in_sese): Remove.
14261         * sese.h (sese_info_t::loop_nest): Remove.
14262         * sese.c (new_sese_info): Do not allocate loop_nest.
14263         (free_sese_info): Do not free loop_nest.
14265 2017-09-27  Jakub Jelinek  <jakub@redhat.com>
14267         PR c++/82159
14268         * gimplify.c (gimplify_modify_expr): Don't optimize away zero sized
14269         lhs from calls if the lhs has addressable type.
14271 2017-09-27  Richard Biener  <rguenther@suse.de>
14273         * graphite.h (scop::max_alias_set): New member.
14274         * graphite-scop-detection.c: Remove references to non-existing
14275         --param in comments.
14276         (build_alias_sets): Record the maximum alias set used for drs.
14277         (build_scops): Support zero as unlimited for
14278         --param graphite-max-arrays-per-scop.
14279         * graphite-sese-to-poly.c (add_scalar_version_numbers): Remove
14280         and inline into ...
14281         (build_poly_sr_1): ... here.  Compute alias set based on the
14282         maximum alias set used for drs rather than
14283         PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP
14285 2017-09-27  Richard Biener  <rguenther@suse.de>
14287         * graphite-optimize-isl.c (get_schedule_for_node_st): Allow
14288         --param loop-block-tile-size=0 to disable tiling.
14290 2017-09-27  Richard Biener  <rguenther@suse.de>
14292         * doc/invoke.texi (graphite-max-bbs-per-function): Remove.
14293         (graphite-max-nb-scop-params): Document special value zero.
14294         * domwalk.h (dom_walker::STOP): New symbolical constant.
14295         (dom_walker::dom_walker): Add optional parameter for bb to
14296         RPO mapping.
14297         (dom_walker::~dom_walker): Declare.
14298         (dom_walker::before_dom_children): Document STOP return value.
14299         (dom_walker::m_user_bb_to_rpo): New member.
14300         (dom_walker::m_bb_to_rpo): Likewise.
14301         * domwalk.c (dom_walker::dom_walker): Compute bb to RPO
14302         mapping here if not provided by the user.
14303         (dom_walker::~dom_walker): Free bb to RPO mapping if not
14304         provided by the user.
14305         (dom_walker::STOP): Define.
14306         (dom_walker::walk): Do not compute bb to RPO mapping here.
14307         Support STOP return value from before_dom_children to stop
14308         walking.
14309         * graphite-optimize-isl.c (optimize_isl): If the schedule
14310         is the same still generate code if -fgraphite-identity
14311         or -floop-parallelize-all are given.
14312         * graphite-scop-detection.c: Include cfganal.h.
14313         (gather_bbs::gather_bbs): Get and pass through bb to RPO
14314         mapping.
14315         (gather_bbs::before_dom_children): Return STOP for BBs
14316         not in the region.
14317         (build_scops): Compute bb to RPO mapping and pass it to
14318         the domwalk.  Treat --param graphite-max-nb-scop-params=0
14319         as not limiting the number of params.
14320         * graphite.c (graphite_initialize): Remove limit on the
14321         number of basic-blocks in a function.
14322         * params.def (PARAM_GRAPHITE_MAX_BBS_PER_FUNCTION): Remove.
14323         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Adjust to documented
14324         default value of 10.
14326 2017-09-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
14328         * config/rs6000/vsx.md (peephole for optimizing move SF to GPR):
14329         Adjust code to eliminate needing to do the shift right 32-bits
14330         operation after XSCVDPSPN.
14332 2017-09-26  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
14334         * match.pd ((X / Y) == 0 -> X < Y): New pattern.
14335         ((X / Y) != 0 -> X >= Y): Likewise.
14337 2017-09-26  Carl Love  <cel@us.ibm.com>
14339         * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_XL_LEN_R,
14340         P9V_BUILTIN_VEC_XST_LEN_R): Add support for builtins
14341         vector unsigned char vec_xl_len_r (unsigned char *, size_t);
14342         void vec_xst_len_r (vector unsigned char, unsigned char *, size_t);
14343         * config/rs6000/altivec.h (vec_xl_len_r, vec_xst_len_r): Add defines.
14344         * config/rs6000/rs6000-builtin.def (XL_LEN_R, XST_LEN_R): Add
14345         definitions and overloading.
14346         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case
14347         statement for P9V_BUILTIN_XST_LEN_R.
14348         (altivec_init_builtins): Add def_builtin for P9V_BUILTIN_STXVLL.
14349         * config/rs6000/vsx.md (lxvll, stxvll, xl_len_r, xst_len_r): Add
14350         define_expand and define_insn for the instructions and builtins.
14351         * doc/extend.texi: Update the built-in documentation file for the new
14352         built-in functions.
14353         * config/rs6000/altivec.md (altivec_lvsl_reg, altivec_lvsr_reg): Add
14354         define_insn for the instructions
14356 2017-09-26  Krister Walfridsson  <krister.walfridsson@gmail.com>
14358         PR target/39570
14359         * gcc/config/netbsd-protos.h: New file.
14360         * gcc/config/netbsd.c: New file.
14361         * gcc/config/netbsd.h (SUBTARGET_INIT_BUILTINS): Define.
14362         * gcc/config/t-netbsd: New file.
14363         * gcc/config.gcc (tm_p_file): Add netbsd-protos.h.
14364         (tmake_file) Add t-netbsd.
14365         (extra_objs) Add netbsd.o.
14367 2017-09-26  Janus Weil  <janus@gcc.gnu.org>
14369         PR fortran/82143
14370         PR fortran/82324
14371         * doc/sourcebuild.texi: Document fortran_real_10 and fortran_real_16.
14373 2017-09-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
14375         * config/rs6000/rs6000.md (extendsi<mode>2): Add a splitter to do
14376         sign extension from a vector register to a GPR by doing a 32-bit
14377         direct move and then an EXTSW.
14378         (extendsi<mode>2 splitter): Likewise.
14379         (movsi_from_sf): Adjust code to eliminate doing a 32-bit shift
14380         right or vector extract after doing XSCVDPSPN.  Use
14381         zero_extendsidi2 instead of p8_mfvsrd_4_disf to move the value to
14382         the GPRs.
14383         (movdi_from_sf_zero_ext): Likewise.
14384         (reload_gpr_from_vsxsf): Likewise.
14385         (p8_mfvsrd_4_disf): Delete, no longer used.
14386         (movsi_from_df): Optimize converting a DFmode to a SFmode, and
14387         then needing to move the SFmode to a GPR to use the XSCVDPSP
14388         instruction instead of FRSP and XSCVDPSPN.
14389         * config/rs6000/vsx.md (vsx_xscvspdp_scalar2): Move insn so that
14390         it is adjacent to the other XSCVSPDP insns.
14391         (vsx_xscvdpsp_scalar): Use "ww" constraint instead of "f" to allow
14392         SFmode to be in traditional Altivec registers.
14393         (vsx_xscvdpspn): Eliminate useless alternative constraint.
14394         (vsx_xscvspdpn): Likewise.
14395         (vsx_xscvspdpn_scalar): Likewise.
14397 2017-09-26  Martin Jambor  <mjambor@suse.cz>
14399         * tree-sra.c (compare_access_positions): Put integral types first,
14400         stabilize sorting of integral types, remove conditions putting
14401         non-full-precision integers last.
14402         (sort_and_splice_var_accesses): Disable scalarization if a
14403         non-integert would be represented by a non-full-precision integer.
14405 2017-09-26  Joseph Myers  <joseph@codesourcery.com>
14407         * config/microblaze/linux.h (TARGET_ASM_FILE_END): Likewise.
14408         * config/pa/pa.h (NEED_INDICATE_EXEC_STACK): Likewise.
14409         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Likewise.
14410         * config/pa/pa.c (pa_hpux_file_end): Rename to pa_file_end.
14411         Define unconditionally, with [ASM_OUTPUT_EXTERNAL_REAL]
14412         conditionals inside the function instead of around it.  Call
14413         file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK.
14414         (TARGET_ASM_FILE_END): Define unconditionally to pa_file_end.
14416 2017-09-26  Richard Biener  <rguenther@suse.de>
14418         * graphite-scop-detection.c (scop_detection::build_scop_depth): Rewrite,
14419         fold in ...
14420         (scop_detection::build_scop_breadth): ... this.  Removed.
14421         (scop_detection::loop_is_valid_in_scop): Fold into single caller.
14422         (scop_detection::harmful_stmt_in_bb): Likewise.
14423         (scop_detection::graphite_can_represent_stmt): Likewise.
14424         (scop_detection::loop_body_is_valid_scop): Likewise.  Remove recursion.
14425         (scop_detection::can_represent_loop): Remove recursion, fold in ...
14426         (scop_detection::can_represent_loop_1): ... this.  Removed.
14427         (scop_detection::harmful_loop_in_region): Simplify after inlining
14428         the above and remove more quadraticness.
14429         (build_scops): Adjust.
14430         * tree-data-ref.c (loop_nest_has_data_refs): Remove pointless
14431         quadraticness.
14433 2017-09-26  Jakub Jelinek  <jakub@redhat.com>
14435         PR target/82267
14436         * config/i386/i386.c (ix86_print_operand_address_as): Only test
14437         REGNO (base) == SP_REG if base is a REG.
14439         PR middle-end/35691
14440         * tree-ssa-reassoc.c (update_range_test): Dump r->exp each time
14441         if it is different SSA_NAME.
14442         (optimize_range_tests_cmp_bitwise): New function.
14443         (optimize_range_tests): Call it.
14445 2017-09-26  Richard Biener  <rguenther@suse.de>
14447         PR tree-optimization/82321
14448         * graphite.c (canonicalize_loop_closed_ssa): Properly check
14449         for the def being inside the loop.
14451 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14453         * config/s390/vx-builtins.md ("vmslg"): Add missing operand in
14454         assembler output.
14455         * config/s390/s390-builtins.def: Fix constraint on op4.
14457 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14459         * config/s390/s390.c (s390_expand_vec_compare): Use the new mode
14460         independent expanders.
14461         * config/s390/vector.md ("vec_cmpuneq", "vec_cmpltgt")
14462         ("vec_ordered", "vec_unordered"): New expanders.
14464 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14466         * config/s390/s390.c (s390_preferred_simd_mode): Return V4SFmode
14467         for SFmode.
14469 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14471         * config/s390/vector.md ("vec_unpacks_low_v16qi"): Rename to
14472         vec_unpacks_lo_v16qi.
14473         ("vec_unpacku_low_v16qi"): Rename to vec_unpacku_lo_v16qi.
14475 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14477         * config/s390/vector.md ("vec_unpacks_lo_v4sf")
14478         ("vec_unpacks_hi_v4sf", "vec_unpacks_lo_v2df")
14479         ("vec_unpacks_hi_v2df", "vec_pack_trunc_v2df"): New expanders.
14481 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14483         * config/s390/predicates.md ("const_shift_by_byte_operand"): New
14484         predicate.
14485         * config/s390/vector.md ("*vec_srb<mode>"): Change modes to V_128
14486         and V16QI.
14487         ("*vec_slb<mode>"): New insn pattern.
14488         ("vec_shr_<mode>"): New expander.
14489         * config/s390/vx-builtins.md ("vec_slb<mode>"): Turn into expander
14490         and force the shift count operand to V16QImode.
14491         ("vec_srb<mode>"): Set shift count mode to V16QI.
14493 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14495         * config/s390/vector.md ("vec_widen_umult_lo_<mode>")
14496         ("vec_widen_umult_hi_<mode>", "vec_widen_smult_lo_<mode>")
14497         ("vec_widen_smult_hi_<mode>"): New expander definitions.
14499 2017-09-26  Richard Earnshaw  <rearnsha@arm.com>
14501         PR target/82175
14502         * config/arm/arm.h (DRIVER_SELF_SPECS): Separate sub-rules with commas.
14504 2017-09-26  Richard Biener  <rguenther@suse.de>
14506         PR tree-optimization/82320
14507         * tree-ssa-sccvn.c (set_ssa_val_to): Changing undef to undef
14508         isn't a change.
14510 2017-09-25  Jeff Law  <law@redhat.com>
14512         * config/rs6000/rs6000-protos.h (output_probe_stack_range): Update
14513         prototype for new argument.
14514         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack_1): New function,
14515         mostly extracted from rs6000_emit_allocate_stack.
14516         (rs6000_emit_probe_stack_range_stack_clash): New function.
14517         (rs6000_emit_allocate_stack): Call
14518         rs6000_emit_probe_stack_range_stack_clash as needed.
14519         (rs6000_emit_probe_stack_range): Add additional argument
14520         to call to gen_probe_stack_range{si,di}.
14521         (output_probe_stack_range): New.
14522         (output_probe_stack_range_1): Renamed from output_probe_stack_range.
14523         (output_probe_stack_range_stack_clash): New.
14524         (rs6000_emit_prologue): Emit notes into dump file as requested.
14525         * rs6000.md (allocate_stack): Handle -fstack-clash-protection.
14526         (probe_stack_range<P:mode>): Operand 0 is now early-clobbered.
14527         Add additional operand and pass it to output_probe_stack_range.
14529 2017-09-25  Bin Cheng  <bin.cheng@arm.com>
14531         PR tree-optimization/82163
14532         * tree-ssa-loop-manip.h (verify_loop_closed_ssa): New parameter.
14533         (checking_verify_loop_closed_ssa): New parameter.
14534         * tree-ssa-loop-manip.c (check_loop_closed_ssa_use): Delete.
14535         (check_loop_closed_ssa_stmt): Delete.
14536         (check_loop_closed_ssa_def, check_loop_closed_ssa_bb): New functions.
14537         (verify_loop_closed_ssa): Check loop closed ssa form for LOOP.
14538         (tree_transform_and_unroll_loop): Check loop closed ssa form only for
14539         changed loops.
14541 2017-09-25  Pekka Jaaskelainen <pekka@parmance.com>
14543         * brig-builtins.def: Treat HSAIL barrier builtins as
14544         setjmp/longjump style functions.
14546 2017-09-25  Richard Sandiford  <richard.sandiford@linaro.org>
14548         * target.def (constant_alignment): New hook.
14549         * defaults.h (CONSTANT_ALIGNMENT): Delete.
14550         * doc/tm.texi.in (CONSTANT_ALIGNMENT): Replace with...
14551         (TARGET_CONSTANT_ALIGNMENT): ...this new hook.
14552         * doc/tm.texi: Regenerate.
14553         * targhooks.h (default_constant_alignment): Declare.
14554         (constant_alignment_word_strings): Likewise.
14555         * targhooks.c (default_constant_alignment): New function.
14556         (constant_alignment_word_strings): Likewise.
14557         * builtins.c (get_object_alignment_2): Use targetm.constant_alignment
14558         instead of CONSTANT_ALIGNMENT.
14559         * varasm.c (align_variable, get_variable_align, build_constant_desc)
14560         (force_const_mem): Likewise.
14561         * config/aarch64/aarch64.h (CONSTANT_ALIGNMENT): Delete.
14562         * config/aarch64/aarch64.c (aarch64_constant_alignment): New function.
14563         (aarch64_classify_address): Call it instead of CONSTANT_ALIGNMENT.
14564         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14565         * config/alpha/alpha.h (CONSTANT_ALIGNMENT): Delete commented-out
14566         definition.
14567         * config/arc/arc.h (CONSTANT_ALIGNMENT): Delete.
14568         * config/arc/arc.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14569         constant_alignment_word_strings.
14570         * config/arm/arm.h (CONSTANT_ALIGNMENT_FACTOR): Delete.
14571         (CONSTANT_ALIGNMENT): Likewise.
14572         * config/arm/arm.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14573         (arm_constant_alignment): New function.
14574         * config/bfin/bfin.h (CONSTANT_ALIGNMENT): Delete.
14575         * config/bfin/bfin.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14576         constant_alignment_word_strings.
14577         * config/cr16/cr16.h (CONSTANT_ALIGNMENT): Delete.
14578         * config/cr16/cr16.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14579         constant_alignment_word_strings.
14580         * config/cris/cris.h (CONSTANT_ALIGNMENT): Delete.
14581         * config/cris/cris.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14582         (cris_constant_alignment): New function.
14583         * config/epiphany/epiphany.h (CONSTANT_ALIGNMENT): Delete.
14584         * config/epiphany/epiphany.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14585         (epiphany_constant_alignment): New function.
14586         * config/fr30/fr30.h (CONSTANT_ALIGNMENT): Delete.
14587         * config/fr30/fr30.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14588         constant_alignment_word_strings.
14589         * config/frv/frv.h (CONSTANT_ALIGNMENT): Delete.
14590         * config/frv/frv.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14591         constant_alignment_word_strings.
14592         * config/ft32/ft32.h (CONSTANT_ALIGNMENT): Delete.
14593         * config/ft32/ft32.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14594         constant_alignment_word_strings.
14595         * config/i386/i386.h (CONSTANT_ALIGNMENT): Delete.
14596         * config/i386/i386-protos.h (ix86_constant_alignment): Delete.
14597         * config/i386/i386.c (ix86_constant_alignment): Make static.
14598         Use the same interface as the target hook.
14599         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14600         * config/ia64/ia64.h (CONSTANT_ALIGNMENT): Delete.
14601         * config/ia64/ia64.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14602         constant_alignment_word_strings.
14603         * config/iq2000/iq2000.h (CONSTANT_ALIGNMENT): Delete.
14604         * config/iq2000/iq2000.c (iq2000_constant_alignment): New function.
14605         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14606         * config/lm32/lm32.h (CONSTANT_ALIGNMENT): Delete.
14607         * config/lm32/lm32.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14608         constant_alignment_word_strings.
14609         * config/m32r/m32r.h (CONSTANT_ALIGNMENT): Delete.
14610         * config/m32r/m32r.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14611         constant_alignment_word_strings.
14612         * config/mcore/mcore.h (CONSTANT_ALIGNMENT): Delete.
14613         * config/mcore/mcore.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14614         constant_alignment_word_strings.
14615         * config/microblaze/microblaze.h (CONSTANT_ALIGNMENT): Delete.
14616         * config/microblaze/microblaze.c (microblaze_constant_alignment):
14617         New function.
14618         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14619         * config/mips/mips.h (CONSTANT_ALIGNMENT): Delete.
14620         * config/mips/mips.c (mips_constant_alignment): New function.
14621         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14622         * config/mmix/mmix.h (CONSTANT_ALIGNMENT): Delete.
14623         * config/mmix/mmix-protos.h (mmix_constant_alignment): Delete.
14624         * config/mmix/mmix.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14625         (mmix_constant_alignment): Make static.  Use the same interface
14626         as the target hook.
14627         * config/moxie/moxie.h (CONSTANT_ALIGNMENT): Delete.
14628         * config/moxie/moxie.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14629         constant_alignment_word_strings.
14630         * config/nios2/nios2.h (CONSTANT_ALIGNMENT): Delete.
14631         * config/nios2/nios2.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14632         constant_alignment_word_strings.
14633         * config/pa/pa.h (CONSTANT_ALIGNMENT): Delete.
14634         * config/pa/pa.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14635         constant_alignment_word_strings.
14636         * config/powerpcspe/powerpcspe.h (CONSTANT_ALIGNMENT): Delete.
14637         * config/powerpcspe/powerpcspe.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14638         (rs6000_constant_alignment): New function.
14639         * config/riscv/riscv.h (CONSTANT_ALIGNMENT): Delete.
14640         * config/riscv/riscv.c (riscv_constant_alignment): New function.
14641         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14642         * config/rs6000/rs6000.h (CONSTANT_ALIGNMENT): Delete.
14643         * config/rs6000/rs6000.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14644         (rs6000_constant_alignment): New function.
14645         * config/s390/s390.h (CONSTANT_ALIGNMENT): Delete.
14646         * config/s390/s390.c (s390_constant_alignment): New function.
14647         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14648         * config/sh/sh.h (CONSTANT_ALIGNMENT): Delete.
14649         * config/sh/sh.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14650         constant_alignment_word_strings.
14651         * config/sparc/sparc.h (CONSTANT_ALIGNMENT): Delete.
14652         * config/sparc/sparc.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14653         (sparc_constant_alignment): New function.
14654         * config/spu/spu.h (CONSTANT_ALIGNMENT): Delete.
14655         * config/spu/spu.c (spu_constant_alignment): New function.
14656         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14657         * config/stormy16/stormy16.h (CONSTANT_ALIGNMENT): Delete.
14658         * config/stormy16/stormy16.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14659         constant_alignment_word_strings.
14660         * config/tilegx/tilegx.h (CONSTANT_ALIGNMENT): Delete.
14661         * config/tilegx/tilegx.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14662         constant_alignment_word_strings.
14663         * config/tilepro/tilepro.h (CONSTANT_ALIGNMENT): Delete.
14664         * config/tilepro/tilepro.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14665         constant_alignment_word_strings.
14666         * config/visium/visium.h (CONSTANT_ALIGNMENT): Delete.
14667         * config/visium/visium.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14668         (visium_constant_alignment): New function.
14669         * config/xtensa/xtensa.h (CONSTANT_ALIGNMENT): Delete.
14670         * config/xtensa/xtensa.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14671         (xtensa_constant_alignment): New function.
14672         * system.h (CONSTANT_ALIGNMENT): Poison.
14674 2017-09-25  Will Schmidt  <will_schmidt@vnet.ibm.com>
14676         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
14677         for early folding of vector stores (ALTIVEC_BUILTIN_ST_*).
14678         (rs6000_builtin_valid_without_lhs): New helper function.
14679         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
14680         Remove obsoleted code for handling ALTIVEC_BUILTIN_VEC_ST.
14682 2017-09-25  Richard Sandiford  <richard.sandiford@linaro.org>
14684         * target.h (vec_perm_indices): Use unsigned short rather than
14685         unsigned char.
14686         (auto_vec_perm_indices): Likewise.
14687         * config/aarch64/aarch64.c (aarch64_vectorize_vec_perm_const_ok):
14688         Use unsigned int rather than unsigned char.
14689         * config/arm/arm.c (arm_vectorize_vec_perm_const_ok): Likewise.
14691 2017-09-25  Richard Biener  <rguenther@suse.de>
14693         * cfgloop.h (sort_sibling_loops): Declare.
14694         * cfgloop.c (sort_sibling_loops_cmp): New helper.
14695         (sort_sibling_loops): New function sorting the sibling loop list
14696         in RPO order.
14697         * graphite.c (graphite_transform_loops): Sort sibling loops.
14699 2017-09-25  Richard Sandiford  <richard.sandifird@linaro.org>
14701         * target.def (vec_perm_const_ok): Change sel parameter to
14702         vec_perm_indices.
14703         * optabs-query.c (can_vec_perm_p): Update accordingly.
14704         * doc/tm.texi: Regenerate.
14705         * config/aarch64/aarch64.c (expand_vec_perm_d): Change perm to
14706         auto_vec_perm_indices and remove separate nelt field.
14707         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
14708         (aarch64_evpc_ext, aarch64_evpc_rev, aarch64_evpc_dup)
14709         (aarch64_evpc_tbl, aarch64_expand_vec_perm_const_1)
14710         (aarch64_expand_vec_perm_const): Update accordingly.
14711         (aarch64_vectorize_vec_perm_const_ok): Likewise.  Change sel
14712         to vec_perm_indices.
14713         * config/arm/arm.c (expand_vec_perm_d): Change perm to
14714         auto_vec_perm_indices and remove separate nelt field.
14715         (arm_evpc_neon_vuzp, arm_evpc_neon_vzip, arm_evpc_neon_vrev)
14716         (arm_evpc_neon_vtrn, arm_evpc_neon_vext, arm_evpc_neon_vtbl)
14717         (arm_expand_vec_perm_const_1, arm_expand_vec_perm_const): Update
14718         accordingly.
14719         (arm_vectorize_vec_perm_const_ok): Likewise.  Change sel
14720         to vec_perm_indices.
14721         * config/i386/i386.c (ix86_vectorize_vec_perm_const_ok): Change
14722         sel to vec_perm_indices.
14723         * config/ia64/ia64.c (ia64_vectorize_vec_perm_const_ok): Likewise.
14724         * config/mips/mips.c (mips_vectorize_vec_perm_const_ok): Likewise.
14725         * config/powerpcspe/powerpcspe.c (rs6000_vectorize_vec_perm_const_ok):
14726         Likewise.
14727         * config/rs6000/rs6000.c (rs6000_vectorize_vec_perm_const_ok):
14728         Likewise.
14730 2017-09-25  Pierre-Marie de Rodat  <derodat@adacore.com>
14732         PR debug/82155
14733         * dwarf2out.c (dwarf2out_early_global_decl): Call dwarf2out_decl
14734         on the FUNCTION_DECL function context if it has a DIE that is a
14735         declaration.
14737 2017-09-25  Richard Biener  <rguenther@suse.de>
14739         PR tree-optimization/82285
14740         * tree-vect-patterns.c (vect_recog_bool_pattern): Also handle
14741         enumeral types.
14743 2017-09-25  Tom de Vries  <tom@codesourcery.com>
14745         PR target/80035
14746         PR target/81069
14747         * config/nvptx/nvptx.c (nvptx_output_call_insn): Add exit after call to
14748         noreturn function.
14750 2017-09-25  Richard Biener  <rguenther@suse.de>
14752         * graphite-optimize-isl.c (optimize_isl): Fail and dump if
14753         ISL errors other than isl_error_quota happen.  Dump if the
14754         schedule is the same.
14755         * graphite-sese-to-poly.c (build_poly_scop): Fail on ISL
14756         errors instead of aborting inside ISL.
14758 2017-09-25  Iain Sandoe  <iain@codesourcery.com>
14760         PR target/80556
14761         * config/i386/darwin.h (REAL_LIB_SPEC): New; put libSystem ahead
14762         of libgcc_eh for m64.
14763         * config/i386/darwin64.h: Likewise.
14765 2017-09-25  Richard Biener  <rguenther@suse.de>
14767         PR middle-end/82144
14768         * dwarf2out.c (gen_enumeration_type_die): Do not add alignment
14769         attribute for incomplete types nor twice for complete ones.
14771 2017-09-24  Uros Bizjak  <ubizjak@gmail.com>
14773         PR target/82267
14774         * config/i386/i386.c (ix86_print_operand_address_as): Encode
14775         %esp as %rsp to avoid 0x67 prefix if there is no index or base
14776         register.
14778 2017-09-23  Uros Bizjak  <ubizjak@gmail.com>
14780         PR bootstrap/82306
14781         * config/i386/i386.opt (mprefer-avx256): Use
14782         ix86_target_flags variable.
14783         * config/i386/i386.c (ix86_target_string): Move
14784         -mprefer-avx256 to flag2_opts.
14786 2017-09-22  Jakub Jelinek  <jakub@redhat.com>
14788         PR middle-end/35691
14789         * match.pd: Simplify x == -1 & y == -1 into (x & y) == -1
14790         and x != -1 | y != -1 into (x & y) != -1.
14792 2017-09-22  Steve Ellcey  <sellcey@cavium.com>
14794         * config.gcc: Add new case statement to set
14795         default_gnu_indirect_function.  Remove it from x86_64-*-linux*,
14796         i[34567]86-*, powerpc*-*-linux*spe*, powerpc*-*-linux*, s390-*-linux*,
14797         s390x-*-linux* case statements.   Added aarch64 to the list of
14798         supported architectures.
14800 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14802         PR tree-optimization/82289
14803         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs): Check
14804         STMT_VINFO_RELEVANT_P.
14806 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14807             Alan Hayward  <alan.hayward@arm.com>
14808             David Sherwood  <david.sherwood@arm.com>
14810         * tree-vrp.c (extract_range_from_multiplicative_op_1): Assert
14811         for VR_RANGE only; don't allow VR_ANTI_RANGE.
14812         (extract_range_from_binary_expr_1): Don't call
14813         extract_range_from_multiplicative_op_1 if !range_int_cst_p.
14815 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14816             Alan Hayward  <alan.hayward@arm.com>
14817             David Sherwood  <david.sherwood@arm.com>
14819         * target.def (preferred_vector_alignment): New hook.
14820         * doc/tm.texi.in (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): New
14821         hook.
14822         * doc/tm.texi: Regenerate.
14823         * targhooks.h (default_preferred_vector_alignment): Declare.
14824         * targhooks.c (default_preferred_vector_alignment): New function.
14825         * tree-vectorizer.h (dataref_aux): Add a target_alignment field.
14826         Expand commentary.
14827         (DR_TARGET_ALIGNMENT): New macro.
14828         (aligned_access_p): Update commentary.
14829         (vect_known_alignment_in_bytes): New function.
14830         * tree-vect-data-refs.c (vect_calculate_required_alignment): New
14831         function.
14832         (vect_compute_data_ref_alignment): Set DR_TARGET_ALIGNMENT.
14833         Calculate the misalignment based on the target alignment rather than
14834         the vector size.
14835         (vect_update_misalignment_for_peel): Use DR_TARGET_ALIGMENT
14836         rather than TYPE_ALIGN / BITS_PER_UNIT to update the misalignment.
14837         (vect_enhance_data_refs_alignment): Mask the byte misalignment with
14838         the target alignment, rather than masking the element misalignment
14839         with the number of elements in a vector.  Also use the target
14840         alignment when calculating the maximum number of peels.
14841         (vect_find_same_alignment_drs): Use vect_calculate_required_alignment
14842         instead of TYPE_ALIGN_UNIT.
14843         (vect_duplicate_ssa_name_ptr_info): Remove stmt_info parameter.
14844         Measure DR_MISALIGNMENT relative to DR_TARGET_ALIGNMENT.
14845         (vect_create_addr_base_for_vector_ref): Update call accordingly.
14846         (vect_create_data_ref_ptr): Likewise.
14847         (vect_setup_realignment): Realign by ANDing with
14848         -DR_TARGET_MISALIGNMENT.
14849         * tree-vect-loop-manip.c (vect_gen_prolog_loop_niters): Calculate
14850         the number of peels based on DR_TARGET_ALIGNMENT.
14851         * tree-vect-stmts.c (get_group_load_store_type): Compare the gap
14852         with the guaranteed alignment boundary when deciding whether
14853         overrun is OK.
14854         (vectorizable_mask_load_store): Interpret DR_MISALIGNMENT
14855         relative to DR_TARGET_ALIGNMENT instead of TYPE_ALIGN_UNIT.
14856         (ensure_base_align): Remove stmt_info parameter.  Get the
14857         target base alignment from DR_TARGET_ALIGNMENT.
14858         (vectorizable_store): Update call accordingly.   Interpret
14859         DR_MISALIGNMENT relative to DR_TARGET_ALIGNMENT instead of
14860         TYPE_ALIGN_UNIT.
14861         (vectorizable_load): Likewise.
14863 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14864             Alan Hayward  <alan.hayward@arm.com>
14865             David Sherwood  <david.sherwood@arm.com>
14867         * tree-vectorizer.h (vect_get_scalar_dr_size): New function.
14868         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Use it.
14869         (vect_enhance_data_refs_alignment): Likewise.
14871 2017-09-22  Richard Earnshaw  <richard.earnshaw@arm.com>
14873         * config/arm/parsecpu.awk (fatal): Note that we've encountered an
14874         error.  Only quit immediately if parsing is complete.
14875         (BEGIN): Initialize fatal_err and parse_done.
14876         (begin fpu, end fpu): Check number of arguments.
14877         (begin arch, end arch): Likewise.
14878         (begin cpu, end cpu): Likewise.
14879         (cname, tune for, tune flags, architecture, fpu, option): Likewise.
14880         (optalias): Likewise.
14882 2017-09-22  Richard Earnshaw  <richard.earnshaw@arm.com>
14884         * config.gcc (arm*-*-*): Don't add arm-isa.h to tm_p_file.
14885         * config/arm/arm-isa.h: Delete.  Move definitions to ...
14886         * arm-cpus.in: ... here.  Use new feature and fgroup values.
14887         * config/arm/arm.c (arm_option_override): Use lower case for feature
14888         bit names.
14889         * config/arm/arm.h (TARGET_HARD_FLOAT): Likewise.
14890         (TARGET_VFP3, TARGET_VFP5, TARGET_FMA): Likewise.
14891         * config/arm/parsecpu.awk (END): Add new command 'isa'.
14892         (isa_pfx): Delete.
14893         (print_isa_bits_for): New function.
14894         (gen_isa): New function.
14895         (gen_comm_data): Use print_isa_bits_for.
14896         (define feature): New keyword.
14897         (define fgroup): New keyword.
14898         * config/arm/t-arm (TM_H): Remove.
14899         (GTM_H): Add arm-isa.h.
14900         (arm-isa.h): Add rule to generate file.
14901         * common/config/arm/arm-common.c: (arm_canon_arch_option): Use lower
14902         case for feature bit names.
14904 2017-09-22  Richard Biener  <rguenther@suse.de>
14906         * graphite-isl-ast-to-gimple.c (graphite_verify): Inline into
14907         single caller.
14908         (graphite_regenerate_ast_isl): Do not reset SCEV.  Move debug
14909         print of no dependency loops ...
14910         * graphite.c (graphite_transform_loops): ... here.
14911         (canonicalize_loop_closed_ssa_form): Work from inner to outer
14912         loops.
14913         (same_close_phi_node, remove_duplicate_close_phi,
14914         make_close_phi_nodes_unique, defined_in_loop_p): Fold into ...
14915         (canonicalize_loop_closed_ssa): ... here and simplify.
14916         * graphite-optimize-isl.c: Include tree-vectorizer.h.
14917         (optimize_isl): Use dump_printf_loc to tell when we stopped
14918         optimizing because of an ISL timeout.
14920 2017-09-22  Richard Biener  <rguenther@suse.de>
14922         PR tree-optimization/82291
14923         * tree-if-conv.c (predicate_mem_writes): Make sure to
14924         remove writes in blocks predicated with false.
14926 2017-09-22  Richard Biener  <rguenther@suse.de>
14928         * sese.c: Include cfganal.h.
14929         (if_region_set_false_region): Remove.
14930         (create_if_region_on_edge): Likewise.
14931         (move_sese_in_condition): Re-implement without destroying
14932         dominators.
14934 2017-09-22  Richard Biener  <rguenther@suse.de>
14936         * graphite-isl-ast-to-gimple.c (translate_pending_phi_nodes):
14937         Verify both BBs contain loop PHI nodes before dispatching to
14938         copy_loop_phi_args.
14939         (graphite_regenerate_ast_isl): Do not recompute dominators,
14940         do not verify three times.  Restructure for clarity.
14941         * graphite-scop-detection.c (same_close_phi_node,
14942         remove_duplicate_close_phi, make_close_phi_nodes_unique,
14943         defined_in_loop_p, canonicalize_loop_closed_ssa,
14944         canonicalize_loop_closed_ssa_form): Simplify, remove excess
14945         checking and SSA rewrite, move to ...
14946         * graphite.c: ... here.  Include ssa.h and tree-ssa-loop-manip.h.
14947         (graphite_initialize): Do not pass in ctx, do not reset the
14948         SCEV cache, compute only dominators.
14949         (graphite_transform_loops): Allocate ISL ctx after
14950         graphite_initialize.  Call canonicalize_loop_closed_ssa_form.
14951         Maintain post-dominators only around build_scops.
14952         * sese.c (if_region_set_false_region): Make static.  Free
14953         and recompute dominators.
14954         (move_sese_in_condition): Assert we don't get called with
14955         post-dominators computed.
14956         * sese.h (if_region_set_false_region): Remove.
14958 2017-09-22  Sergey Shalnov  <sergey.shalnov@intel.com>
14960         * config/i386/sse.md ("mov<mode>_internal"): Use <sseinsnmode>
14961         mode attribute for TARGET_AVX512VL.
14963 2017-09-21  Sergey Shalnov  <sergey.shalnov@intel.com>
14965         * config/i386/i386.opt (mprefer-avx256): New option.
14966         * config/i386/i386.c (ix86_target_string): Add -mprefer-avx256
14967         to flag_opts.
14968         (ix86_preferred_simd_mode): Return 256-bit AVX modes
14969         for TARGET_PREFER_AVX256.
14970         * doc/invoke.texi (x86 Options): Document -mprefer-avx256.
14972 2017-09-21  Jeff Law  <law@redhat.com>
14974         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash):
14975         Fix dump output if the only stack space is for pushed registers.
14977 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14979         * config/spu/spu.c (spu_sched_adjust_cost): Update after renaming
14980         of insn_cost.
14982 2017-09-21  Martin Sebor  <msebor@redhat.com>
14984         PR c/81882
14985         * doc/extend.texi (attribute ifunc): Avoid relying on ill-formed
14986         code (in C++) or code that triggers warnings.
14988 2017-09-21  Eric Botcazou  <ebotcazou@adacore.com>
14990         * stor-layout.c (bit_from_pos): Do not distribute the conversion.
14992 2017-09-21  Segher Boessenkool  <segher@kernel.crashing.org>
14994         * haifa-sched.c: Rename insn_cost to insn_sched_cost.
14995         * sched-rgn.c: Ditto.
14996         * sel-sched-ir.c: Ditto.
14998 2017-09-21  Alexander Monakov  <amonakov@ispras.ru>
15000         * toplev.h (set_random_seed): Adjust return type.
15001         * toplev.c (init_local_tick): Move eager initialization of random_seed
15002         to get_random_seed.  Adjust comment.
15003         (init_random_seed): Inline to get_random_seed, delete.
15004         (get_random_seed): Initialize random_seed lazily.
15005         (set_random_seed): Do not return previous value.
15006         (print_switch_value): Do not call get_random_seed.
15008 2017-09-21  Evgeny Kudryashov  <kudryashov@ispras.ru>
15010         * cgraph.c (delete_function_version): New, broken out from...
15011         (cgraph_node::delete_function_version): ...here.  Rename to
15012         cgraph_node::delete_function_version_by_decl.  Update all uses.
15013         (cgraph_node::remove): Call delete_function_version.
15015 2017-09-21  Jakub Jelinek  <jakub@redhat.com>
15017         PR sanitizer/81715
15018         * tree-inline.c (expand_call_inline): Emit clobber stmts for
15019         VAR_DECLs to which addressable non-volatile parameters are mapped
15020         and for id->retvar after the return value assignment.  Clear
15021         id->retval and id->retbnd after inlining.
15023 2017-09-21  Richard Biener  <rguenther@suse.de>
15025         PR tree-optimization/82276
15026         PR tree-optimization/82244
15027         * tree-vrp.c (build_assert_expr_for): Set
15028         SSA_NAME_OCCURS_IN_ABNORMAL_PHI if the variable we assert on
15029         has it set.
15030         (remove_range_assertions): Revert earlier change.
15032 2017-09-21  Wilco Dijkstra  <wdijkstr@arm.com>
15034         PR target/71951
15035         * config/aarch64/aarch64.h (LIBGCC2_UNWIND_ATTRIBUTE): Define.
15037 2017-09-21  Richard Biener  <rguenther@suse.de>
15039         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
15040         Restore valid IL after code generation errors.
15041         * graphite.c (graphite_transform_loops): Diagnose code
15042         generation issues as MSG_MISSED_OPTIMIZATION and continue
15043         with processing SCOPs.
15045 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
15046             Alan Hayward  <alan.hayward@arm.com>
15047             David Sherwood  <david.sherwood@arm.com>
15049         * calls.c (compute_argument_addresses): Use simplify_gen_binary
15050         rather than choosing between plus_constant and gen_rtx_<CODE>.
15051         * expr.c (emit_push_insn): Likewise.
15052         (expand_expr_real_2): Likewise.
15054 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
15055             Alan Hayward  <alan.hayward@arm.com>
15056             David Sherwood  <david.sherwood@arm.com>
15058         * loop-unroll.c (split_iv): Call copy_rtx on the step.
15060 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
15061             Alan Hayward  <alan.hayward@arm.com>
15062             David Sherwood  <david.sherwood@arm.com>
15064         * tree.c (find_atomic_core_type): Check tree_fits_uhwi_p before
15065         calling tree_to_uhwi.
15067 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
15068             Alan Hayward  <alan.hayward@arm.com>
15069             David Sherwood  <david.sherwood@arm.com>
15071         * tree-ssa-ccp.c (get_value_for_expr): Use a positive test for
15072         INTEGER_CST rather than a negative test for ADDR_EXPR.
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-vrp.c (extract_range_from_binary_expr_1): Check
15079         int_cst_rangeN before calling value_range_constant_singleton (&vrN).
15081 2017-09-21  Richard Biener  <rguenther@suse.de>
15083         PR tree-optimization/71351
15084         * graphite-isl-ast-to-gimple.c (translate_isl_ast_to_gimple::
15085         graphite_create_new_loop_guard): Remove, fold remaining parts
15086         into caller ...
15087         (translate_isl_ast_node_for): ... here and simplify.
15089 2017-09-21  Jakub Jelinek  <jakub@redhat.com>
15091         PR target/82260
15092         * config/i386/i386.md (*movqi_internal): Replace (=q,q) alternative
15093         with (=Q,Q), (=R,R) and (=r,r) alternatives, only enable the
15094         latter two for 64-bit, renumber alternatives, for -Os imov (=q,n)
15095         alternative always use QI mode, for -Os imov (=R,R) alternative
15096         always use SI mode, for imov (=Q,Q) or (=r,r) alternatives
15097         ignore -Os.
15099 2017-09-20  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
15100             Jeff Law  <law@redhat.com>
15102         * config/s390/s390.c (MIN_UNROLL_PROBES): Define.
15103         (allocate_stack_space): New function, partially extracted from
15104         s390_emit_prologue.
15105         (s390_emit_prologue): Track offset to most recent stack probe.
15106         Code to allocate space moved into allocate_stack_space.
15107         Dump actions when no stack is allocated.
15108         (s390_prologue_plus_offset): New function.
15109         (s390_emit_stack_probe): Likewise.
15111 2017-09-20  Alexandre Oliva  <aoliva@redhat.com>
15113         * common.opt (Wa, Wl, Wp, g, gz=): Add
15114         RejectNegative.
15115         (gno-column-info): Remove.
15116         (gcolumn-info): Drop RejectNegative.
15117         (gno-): New prefix.
15118         (gno-record-gcc-switches): Remove.
15119         (grecord-gcc-switches): Drop RejectNegative.
15120         (gno-split-dwarf): Remove.
15121         (gsplit-dwarf): Drop RejectNegative.
15122         (gno-strict-dwarf): Remove.
15123         (gstrict-dwarf): Drop RejectNegative.
15124         * config/darwin.opt (gfull, gused): Add RejectNegative.
15125         * dwarf2out.c (gen_producer_string): Drop
15126         gno-record-gcc-switches handler.
15127         * optc-gen.awk: Add g to prefixes with negative forms.
15128         * opts-common.c (remapping_prefix_p): New.
15129         (find_opt): Check it.
15130         (generate_canonical_option): Test g prefix.
15131         (option_map): Add -gno- mapping.
15132         (add_misspelling_candidates): Check remapping_prefix_p.
15134 2017-09-20  Jeff Law  <law@redhat.com>
15136         * config/powerpcspe/powerpcspe.c (rs6000_expand_prologue): Fix
15137         thinko in stack clash protection support.
15139         * explow.c (compute_stack_clash_protection_loop_data): Use
15140         CONST_INT_P instead of explicit test.  Verify object is a
15141         CONST_INT_P before looking at INTVAL.
15142         (anti_adjust_stack_and_probe_stack_clash): Use CONST_INT_P
15143         instead of explicit test.
15145 2017-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
15147         PR target/77687
15148         * config/rs6000/rs6000.md (stack_restore_tie): Store to a scratch
15149         address instead of to r1 and r11.
15151 2017-09-20  Sebastian Peryt  <sebastian.peryt@intel.com>
15153         * config.gcc: Support "knm".
15154         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "knm".
15155         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15156         PROCESSOR_KNM.
15157         * config/i386/i386.c (m_KNM): Define.
15158         (processor_target_table): Add "knm".
15159         (PTA_KNM): Define.
15160         (ix86_option_override_internal): Add "knm".
15161         (ix86_issue_rate): Add PROCESSOR_KNM.
15162         (ix86_adjust_cost): Ditto.
15163         (ia32_multipass_dfa_lookahead): Ditto.
15164         (get_builtin_code_for_version): Handle PROCESSOR_KNM.
15165         (fold_builtin_cpu): Add M_INTEL_KNM.
15166         * config/i386/i386.h (processor_costs): Define TARGET_KNM.
15167         (processor_type): Add PROCESSOR_KNM.
15168         * config/i386/x86-tune.def: Add m_KNM.
15169         * doc/invoke.texi: Add knm as x86 -march=/-mtune= CPU type.
15171 2017-09-20  Richard Biener  <rguenther@suse.de>
15173         PR tree-optimization/80213
15174         * graphite-scop-detection.c (trivially_empty_bb_p): Labels
15175         are allowed in empty BBs as well.
15176         (canonicalize_loop_closed_ssa): Also look for other complex
15177         edges.
15178         (scop_detection::get_sese): Include the loop-closed PHI block
15179         in loop SESEs.
15180         (scop_detection::merge_sese): Remove code adding extra blocks.
15181         (scop_detection::region_has_one_loop): Adjust for get_sese changes.
15182         (build_scops): Assert the final returned scop is invalid.
15184 2017-09-20  Richard Biener  <rguenther@suse.de>
15186         PR tree-optimization/82264
15187         * tree-ssa-sccvn.c (vn_phi_eq): Use safe_dyn_cast to check
15188         for GIMPLE_CONDs.
15189         (vn_phi_lookup): Likewise.
15190         (vn_phi_insert): Likewise.
15192 2017-09-20  Jakub Jelinek  <jakub@redhat.com>
15194         * dwarf2out.c (tree_add_const_value_attribute): For INTEGER_CST
15195         that fits into uhwi or shwi, add DW_AT_const_value regardless
15196         of early_dwarf without going through RTL, using add_AT_unsigned
15197         or add_AT_int.
15199         * dwarf2out.c (DEBUG_LTO_DWO_INFO_SECTION): Reorder defines.
15200         (DEBUG_LTO_ABBREV_SECTION): Likewise.
15201         (DEBUG_LTO_MACINFO_SECTION): Likewise.
15202         (DEBUG_MACRO_SECTION): Likewise.
15203         (DEBUG_LTO_MACRO_SECTION): Likewise.
15204         (DEBUG_STR_DWO_SECTION): Likewise.
15205         (DEBUG_LTO_STR_DWO_SECTION): Likewise.
15206         (DEBUG_LTO_LINE_SECTION): Drop .dwo suffix from the name.
15207         (DEBUG_LTO_DWO_LINE_SECTION): Define.
15208         (DEBUG_LTO_LINE_STR_SECTION): Define.
15209         (init_sections_and_labels): Initialize debug_line_str_section
15210         variable.  Initialize debug_loc_section for -gdwarf-5 to
15211         DEBUG_LOCLISTS_SECTION.  Formatting fixes.
15213 2017-09-20  Richard Biener  <rguenther@suse.de>
15215         * graphite-sese-to-poly.c (extract_affine): Properly handle
15216         POINTER_PLUS_EXPR, BIT_NOT_EXPR and conversion to signed.
15218 2017-09-20  Richard Biener  <rguenther@suse.de>
15220         PR tree-optimization/81373
15221         * graphite-scop-detection.c (build_cross_bb_scalars_def):
15222         Force SESE live-out defs to be handled even if they are
15223         scev_analyzable_p.
15225 2017-09-19  Jeff Law  <law@redhat.com>
15227         * combine-stack-adj.c (combine_stack_adjustments_for_block): Do
15228         nothing for stack adjustments with REG_STACK_CHECK.
15229         * sched-deps.c (parse_add_or_inc): Reject insns with
15230         REG_STACK_CHECK from dependency breaking.
15231         * config/i386/i386.c (pro_epilogue_adjust_stack): Return insn.
15232         (ix86_adjust_satck_and_probe_stack_clash): Add REG_STACK_NOTEs.
15233         * reg-notes.def (STACK_CHECK): New note.
15235         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): New.
15236         (ix86_expand_prologue): Dump stack clash info as needed.
15237         Call ix86_adjust_stack_and_probe_stack_clash as needed.
15239         * function.c (dump_stack_clash_frame_info): New function.
15240         * function.h (dump_stack_clash_frame_info): Prototype.
15241         (enum stack_clash_probes): New enum.
15243         * config/alpha/alpha.c (alpha_expand_prologue): Also check
15244         flag_stack_clash_protection.
15245         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Likewise.
15246         (arm_expand_prologue, thumb1_expand_prologue): Likewise.
15247         (arm_frame_pointer_required): Likewise.
15248         * config/ia64/ia64.c (ia64_compute_frame_size): Likewise.
15249         (ia64_expand_prologue): Likewise.
15250         * config/mips/mips.c (mips_expand_prologue): Likewise.
15251         * config/powerpcspe/powerpcspe.c (rs6000_expand_prologue): Likewise.
15252         * config/sparc/sparc.c (sparc_expand_prologue): Likewise.
15253         (sparc_flat_expand_prologue): Likewise.
15254         * config/spu/spu.c (spu_expand_prologue): Likewise.
15256         * explow.c: Include "params.h".
15257         (anti_adjust_stack_and_probe_stack_clash): New function.
15258         (get_stack_check_protect): Likewise.
15259         (compute_stack_clash_protection_loop_data): Likewise.
15260         (emit_stack_clash_protection_loop_start): Likewise.
15261         (emit_stack_clash_protection_loop_end): Likewise.
15262         (allocate_dynamic_stack_space): Use get_stack_check_protect.
15263         Use anti_adjust_stack_and_probe_stack_clash.
15264         * explow.h (compute_stack_clash_protection_loop_data): Prototype.
15265         (emit_stack_clash_protection_loop_start): Likewise.
15266         (emit_stack_clash_protection_loop_end): Likewise.
15267         * rtl.h (get_stack_check_protect): Prototype.
15268         * target.def (stack_clash_protection_final_dynamic_probe): New hook.
15269         * targhooks.c (default_stack_clash_protection_final_dynamic_probe): New.
15270         * targhooks.h (default_stack_clash_protection_final_dynamic_probe):
15271         Prototype.
15272         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE):
15273         Add @hook.
15274         * doc/tm.texi: Rebuilt.
15275         * config/aarch64/aarch64.c (aarch64_expand_prologue): Use
15276         get_stack_check_protect.
15277         * config/alpha/alpha.c (alpha_expand_prologue): Likewise.
15278         * config/arm/arm.c (arm_expand_prologue): Likewise.
15279         (arm_frame_pointer_required): Likewise.
15280         * config/i386/i386.c (ix86_expand_prologue): Likewise.
15281         * config/ia64/ia64.c (ia64_expand_prologue): Likewise.
15282         * config/mips/mips.c (mips_expand_prologue): Likewise.
15283         * config/powerpcspe/powerpcspe.c (rs6000_emit_prologue): Likewise.
15284         * config/rs6000/rs6000.c (rs6000_emit_prologue): Likewise.
15285         * config/sparc/sparc.c (sparc_expand_prologue): Likewise.
15286         (sparc_flat_expand_prologue): Likewise.
15288         * common.opt (-fstack-clash-protection): New option.
15289         * flag-types.h (enum stack_check_type): Note difference between
15290         -fstack-check= and -fstack-clash-protection.
15291         * params.def (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE): New PARAM.
15292         (PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Likewise.
15293         * toplev.c (process_options): Issue warnings/errors for cases
15294         not handled with -fstack-clash-protection.
15295         * doc/invoke.texi (-fstack-clash-protection): Document new option.
15296         (-fstack-check): Note additional problem with -fstack-check=generic.
15297         Note that -fstack-check is primarily for Ada and refer users
15298         to -fstack-clash-protection for stack-clash-protection.
15299         Document new params for stack clash protection.
15301 2017-09-19  Uros Bizjak  <ubizjak@gmail.com>
15303         * config/i386/i386.c (ix86_split_long_move): Do not handle
15304         address used for LEA in a special way.
15306 2017-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
15308         * simplify-rtx.c (simplify_binary_operation_1): Fix typo in comment.
15310 2017-09-19  Martin Sebor  <msebor@redhat.com>
15312         PR c/81854
15313         * cgraphunit.c (handle_alias_pairs): Reject aliases between functions
15314         of incompatible types.
15316 2017-09-19  Will Schmidt  <will_schmidt@vnet.ibm.com>
15318         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
15319         for early folding of vector loads (ALTIVEC_BUILTIN_LVX_*).
15320         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
15321         Remove obsoleted code for handling ALTIVEC_BUILTIN_VEC_LD.
15323 2017-09-19  Richard Biener  <rguenther@suse.de>
15325         PR tree-optimization/82244
15326         * tree-vrp.c (remove_range_assertions): Do not propagate
15327         a constant to abnormals but replace the assert with a copy.
15329 2017-09-19  Alexander Monakov  <amonakov@ispras.ru>
15331         PR rtl-optimization/57878
15332         PR rtl-optimization/68988
15333         * lra-assigns.c (reload_pseudo_compare_func): Remove fragmentation
15334         avoidance test involving non_reload_pseudos.  Move frequency test
15335         below the general fragmentation avoidance test.
15337 2017-09-19  Richard Biener  <rguenther@suse.de>
15339         PR tree-optimization/69728
15340         * graphite-sese-to-poly.c (schedule_error): New global.
15341         (add_loop_schedule): Handle empty domain by failing the
15342         schedule.
15343         (build_original_schedule): Handle schedule_error.
15345 2017-09-19  Richard Biener  <rguenther@suse.de>
15347         * graphite-scop-detection.c (scop_detection::can_represent_loop):
15348         Do not iterate to sibling loops but only to siblings of inner
15349         loops.
15351 2017-09-18  Andreas Schwab  <schwab@linux-m68k.org>
15353         PR target/81613
15354         * config/m68k/m68k.md (moveq feeding equality comparison): Check
15355         that the registers are different.
15357 2017-09-18  Uros Bizjak  <ubizjak@gmail.com>
15359         * config/i386/i386.c (fold_builtin_cpu): Add M_AMDFAM17H
15360         to processor_model and "amdfam17h" to arch_names_table.
15361         * doc/extend.texi (__builtin_cpu_is): Document amdfam17h CPU name.
15363 2017-09-18  Jakub Jelinek  <jakub@redhat.com>
15365         PR c/82234
15366         * doc/extend.texi: Add @findex entry for __builtin_shuffle.
15368 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15369             Alan Hayward  <alan.hayward@arm.com>
15370             David Sherwood  <david.sherwood@arm.com>
15372         * tree-vectorizer.h (vect_slp_analyze_operations): Replace parameters
15373         with a vec_info *.
15374         * tree-vect-loop.c (vect_analyze_loop_operations): Update call
15375         accordingly.
15376         * tree-vect-slp.c (vect_slp_analyze_node_operations): Add vec_info *
15377         parameter.  Set SLP_TREE_NUMBER_OF_VEC_STMTS here rather than in
15378         vect_schedule_slp_instance.
15379         (vect_slp_analyze_operations): Replace parameters with a vec_info *.
15380         Update call to vect_slp_analyze_node_operations.  Simplify return
15381         value.
15382         (vect_slp_analyze_bb_1): Update call accordingly.
15383         (vect_schedule_slp_instance): Remove vectorization_factor parameter.
15384         Don't calculate SLP_TREE_NUMBER_OF_VEC_STMTS here.
15385         (vect_schedule_slp): Update call accordingly.
15387 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15388             Alan Hayward  <alan.hayward@arm.com>
15389             David Sherwood  <david.sherwood@arm.com>
15391         * tree-vect-data-refs.c (vect_get_smallest_scalar_type): Cope
15392         with types that aren't in fact scalar.
15394 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15396         * tree-vect-slp.c (vect_record_max_nunits): New function,
15397         split out from...
15398         (vect_build_slp_tree_1): ...here.
15399         (vect_build_slp_tree_2): Call it for phis too.
15401 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15403         * tree-vect-stmts.c (vectorizable_mask_load_store): Pass mask_vectype
15404         to vect_get_vec_def_for_operand when getting the mask operand.
15406 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15407             Alan Hayward  <alan.hayward@arm.com>
15408             David Sherwood  <david.sherwood@arm.com>
15410         * tree-vect-loop.c (vectorizable_live_operation): Fix type of
15411         bitstart.
15413 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15414             Alan Hayward  <alan.hayward@arm.com>
15415             David Sherwood  <david.sherwood@arm.com>
15417         * tree-vect-loop.c (vectorizable_live_operation): Fix element size
15418         calculation for vector booleans.
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-stmts.c (can_vectorize_live_stmts): New function,
15425         split out from...
15426         (vect_transform_stmt): ...here.
15427         (vect_analyze_stmt): Use it instead of calling
15428         vectorizable_live_operation directly.
15430 2017-09-18  Cesar Philippidis  <cesar@codesourcery.com>
15432         * omp-offload.c (oacc_xform_loop): Enable SIMD vectorization on
15433         non-SIMT targets in acc vector loops.
15435 2017-09-18  Claudiu Zissulescu  <claziss@synopsys.com>
15437         * configure.ac: Add arc and check if assembler supports gdwarf2.
15438         * configure: Regenerate.
15440 2017-09-18  Richard Biener  <rguenther@suse.de>
15442         PR tree-optimization/82220
15443         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Exclude
15444         epilogue niters from the min_profitable_iters compute.
15446 2017-09-18  Jakub Jelinek  <jakub@redhat.com>
15448         PR target/82145
15449         * config/i386/i386.c (ix86_init_large_pic_reg): Revert 2017-09-01
15450         changes.  Turn CODE_LABEL into NOTE_INSN_DELETED_LABEL immediately.
15451         (ix86_init_pic_reg): Revert 2017-09-01 changes.
15453 2017-09-18  Eric Botcazou  <ebotcazou@adacore.com>
15455         PR target/81361
15456         * dwarf2cfi.c (add_cfis_to_fde): Do not generate DW_CFA_set_loc after
15457         switching to a new text section.
15459 2017-09-18  Richard Biener  <rguenther@suse.de>
15461         * graphite-scop-detection.c (scop_detection::stmt_has_simple_data_ref):
15462         Simplify.
15463         (build_alias_set): Reject aliases with no access function.
15465 2017-09-18  Richard Biener  <rguenther@suse.de>
15467         PR tree-optimization/79622
15468         * graphite-scop-detection.c (build_cross_bb_scalars_def): Properly
15469         handle PHIs.
15470         (build_cross_bb_scalars_use): Likewise.
15472 2017-09-18  Pierre-Marie de Rodat  <derodat@adacore.com>
15474         * cgraph.h (cgraph_thunk_info): Fix a typo in a comment.
15476 2017-09-18  Alan Modra  <amodra@gmail.com>
15478         PR target/81996
15479         * gcc/config/rs6000/rs6000.c (rs6000_return_addr): Use
15480         stack_pointer_rtx for count 0.  Update comments.  Break up
15481         large rtl expression.
15483 2017-09-17  Daniel Santos  <daniel.santos@pobox.com>
15485         * config/i386/i386.c (xlogue_layout::STUB_NAME_MAX_LEN):
15486         Increase to 20 bytes.
15487         (xlogue_layout::s_stub_names): Add an additional size-2 diminsion.
15488         (xlogue_layout::get_stub_name): Modify to select the appropairate sse
15489         or avx version of the stub.
15491 2017-09-17  H.J. Lu  <hongjiu.lu@intel.com>
15493         PR target/82166
15494         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Properly
15495         compute the minimum stack alignment.  Also update preferred stack
15496         boundary for leaf functions.
15498 2017-09-16  Richard Sandiford  <richard.sandiford@linaro.org>
15500         PR tree-optimization/82228
15501         * tree-vect-loop.c (vectorizable_live_operation): Move initialization
15502         of ncopies.
15504 2017-09-16  Chung-Ju Wu  <jasonwucj@gmail.com>
15506         * common/config/nds32/nds32-common.c
15507         (nds32_option_optimization_table): Refine formatting.
15508         (nds32_option_optimization_table): Use -fsched-pressure and
15509         -fomit-frame-pointer for specific optimization level.
15511 2017-09-16  Chung-Ju Wu  <jasonwucj@gmail.com>
15513         * config/nds32/nds32.c: Refine formatting and comments.
15514         * config/nds32/nds32.h: Likewise.
15515         * config/nds32/nds32.md: Likewise.
15516         * config/nds32/nds32-cost.c: Likewise.
15517         * config/nds32/nds32-isr.c: Likewise.
15518         * config/nds32/nds32-md-auxiliary.c: Likewise.
15519         * config/nds32/nds32-multiple.md: Likewise.
15520         * config/nds32/nds32-predicates.c: Likewise.
15522 2017-09-15  Andrew Sutton  <andrew.n.sutton@gmail.com>
15523             Jakub Jelinek  <jakub@redhat.com>
15525         Add support for -std=c++2a.
15526         * doc/cpp.texi (__cplusplus): Document value for -std=c++2a
15527         or -std=gnu+2a.
15528         * doc/invoke.texi: Document -std=c++2a and -std=gnu++2a.
15530 2017-09-15  Steve Ellcey  <sellcey@cavium.com>
15532         PR target/82066
15533         * doc/extend.texi (Common Function Attributes): Add 
15534         references to ARM, AArch64, and S/390 specific attributes.
15535         (Function Specific Option Pragmas): Add AArch64 and S/390
15536         to list of back ends that support the target pragma.
15538 2017-09-15  Nathan Sidwell  <nathan@acm.org>
15540         * doc/standards.texi: Fix C++17 description.  Update URLs for
15541         C++11 & 14.
15543 2017-09-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
15545         * common.opt (Wcast-align=strict): New warning option.
15546         * doc/invoke.texi: Document -Wcast-align=strict. 
15548 2017-09-15  Pierre-Marie de Rodat  <derodat@adacore.com>
15550         * cgraph.h (cgraph_thunk_info): Add comments.
15551         * cgraph.c (cgraph_node::create_thunk): Adjust comment, make
15552         assert for VIRTUAL_* arguments stricter.
15554 2017-09-15  Jackson Woodruff  <jackson.woodruff@arm.com>
15556         PR tree-optimization/71026
15557         * match.pd: Move RDIV patterns from fold-const.c
15558         * fold-const.c (distribute_real_division): Removed.
15559         (fold_binary_loc): Remove calls to distribute_real_divison.
15561 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15563         * doc/invoke.texi: Document -std=c++17 and -std=gnu++17 and document
15564         c++1z and gnu++1z as deprecated.  Change other references to
15565         -std=c++1z to -std=c++17 and -std=gnu++1z to -std=gnu++17.
15566         Change -Wc++1z-compat to -Wc++17-compat.
15567         * doc/cpp.texi: Document -std=c++17 defines __cplusplus 201703L.
15568         * dwarf2out.c (highest_c_language): Handle C++17.
15569         (gen_compile_unit_die): Likewise.
15571 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15573         PR rtl-optimization/82192
15574         * combine.c (make_extraction): Don't look through non-paradoxical
15575         SUBREGs or TRUNCATE if pos + len is or might be bigger than
15576         inner's mode.
15578 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15579             Alan Hayward  <alan.hayward@arm.com>
15580             David Sherwood  <david.sherwood@arm.com>
15582         * target.def (function_arg_offset): New hook.
15583         * targhooks.h (default_function_arg_offset): Declare.
15584         * targhooks.c (default_function_arg_offset): New function.
15585         * function.c (locate_and_pad_parm): Use
15586         targetm.calls.function_arg_offset instead of FUNCTION_ARG_OFFSET.
15587         * doc/tm.texi.in (FUNCTION_ARG_OFFSET): Replace with...
15588         (TARGET_FUNCTION_ARG_OFFSET): ...this.
15589         * doc/tm.texi: Regenerate.
15590         * config/spu/spu.h (FUNCTION_ARG_OFFSET): Delete.
15591         * config/spu/spu.c (spu_function_arg_offset): New function.
15592         (TARGET_FUNCTION_ARG_OFFSET): Redefine.
15593         * system.h (FUNCTION_ARG_OFFSET): Poison.
15595 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15596             Alan Hayard  <alan.hayward@arm.com>
15597             David Sherwood  <david.sherwood@arm.com>
15599         * target.def (truly_noop_truncation): New hook.
15600         (mode_rep_extended): Refer to TARGET_TRULY_NOOP_TRUNCATION rather
15601         than TRULY_NOOP_TRUNCATION.
15602         * hooks.h (hook_bool_uint_uint_true): Declare.
15603         * hooks.c (hook_bool_uint_uint_true): New function.
15604         * doc/tm.texi.in (TRULY_NOOP_TRUNCATION): Replace with...
15605         (TARGET_TRULY_NOOP_TRUNCATION): ...this.
15606         * doc/tm.texi: Regenerate.
15607         * combine.c (make_extraction): Refer to TARGET_TRULY_NOOP_TRUNCATION
15608         rather than TRULY_NOOP_TRUNCATION in comments.
15609         (simplify_comparison): Likewise.
15610         (record_truncated_value): Likewise.
15611         * expmed.c (extract_bit_field_1): Likewise.
15612         (extract_split_bit_field): Likewise.
15613         * convert.c (convert_to_integer_1): Use targetm.truly_noop_truncation
15614         instead of TRULY_NOOP_TRUNCATION.
15615         * function.c (assign_parm_setup_block): Likewise.
15616         * machmode.h (TRULY_NOOP_TRUNCATION_MODES_P): Likewise.
15617         * rtlhooks.c: Include target.h.
15618         * config/aarch64/aarch64.h (TRULY_NOOP_TRUNCATION): Delete.
15619         * config/alpha/alpha.h (TRULY_NOOP_TRUNCATION): Delete.
15620         * config/arc/arc.h (TRULY_NOOP_TRUNCATION): Delete.
15621         * config/arm/arm.h (TRULY_NOOP_TRUNCATION): Delete.
15622         * config/avr/avr.h (TRULY_NOOP_TRUNCATION): Delete.
15623         * config/bfin/bfin.h (TRULY_NOOP_TRUNCATION): Delete.
15624         * config/c6x/c6x.h (TRULY_NOOP_TRUNCATION): Delete.
15625         * config/cr16/cr16.h (TRULY_NOOP_TRUNCATION): Delete.
15626         * config/cris/cris.h (TRULY_NOOP_TRUNCATION): Delete.
15627         * config/epiphany/epiphany.h (TRULY_NOOP_TRUNCATION): Delete.
15628         * config/fr30/fr30.h (TRULY_NOOP_TRUNCATION): Delete.
15629         * config/frv/frv.h (TRULY_NOOP_TRUNCATION): Delete.
15630         * config/ft32/ft32.h (TRULY_NOOP_TRUNCATION): Delete.
15631         * config/h8300/h8300.h (TRULY_NOOP_TRUNCATION): Delete.
15632         * config/i386/i386.h (TRULY_NOOP_TRUNCATION): Delete.
15633         * config/ia64/ia64.h (TRULY_NOOP_TRUNCATION): Delete.
15634         * config/iq2000/iq2000.h (TRULY_NOOP_TRUNCATION): Delete.
15635         * config/lm32/lm32.h (TRULY_NOOP_TRUNCATION): Delete.
15636         * config/m32c/m32c.h (TRULY_NOOP_TRUNCATION): Delete.
15637         * config/m32r/m32r.h (TRULY_NOOP_TRUNCATION): Delete.
15638         * config/m68k/m68k.h (TRULY_NOOP_TRUNCATION): Delete.
15639         * config/mcore/mcore.h (TRULY_NOOP_TRUNCATION): Delete.
15640         * config/microblaze/microblaze.h (TRULY_NOOP_TRUNCATION): Delete.
15641         * config/mips/mips.h (TRULY_NOOP_TRUNCATION): Delete.
15642         * config/mips/mips.c (mips_truly_noop_truncation): New function.
15643         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15644         * config/mips/mips.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15645         rather than TRULY_NOOP_TRUNCATION in comments.
15646         * config/mmix/mmix.h (TRULY_NOOP_TRUNCATION): Delete.
15647         * config/mn10300/mn10300.h (TRULY_NOOP_TRUNCATION): Delete.
15648         * config/moxie/moxie.h (TRULY_NOOP_TRUNCATION): Delete.
15649         * config/msp430/msp430.h (TRULY_NOOP_TRUNCATION): Delete.
15650         * config/nds32/nds32.h (TRULY_NOOP_TRUNCATION): Delete.
15651         * config/nios2/nios2.h (TRULY_NOOP_TRUNCATION): Delete.
15652         * config/nvptx/nvptx.h (TRULY_NOOP_TRUNCATION): Delete.
15653         * config/pa/pa.h (TRULY_NOOP_TRUNCATION): Delete.
15654         * config/pdp11/pdp11.h (TRULY_NOOP_TRUNCATION): Delete.
15655         * config/powerpcspe/powerpcspe.h (TRULY_NOOP_TRUNCATION): Delete.
15656         * config/riscv/riscv.h (TRULY_NOOP_TRUNCATION): Delete.
15657         * config/riscv/riscv.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15658         rather than TRULY_NOOP_TRUNCATION in comments.
15659         * config/rl78/rl78.h (TRULY_NOOP_TRUNCATION): Delete.
15660         * config/rs6000/rs6000.h (TRULY_NOOP_TRUNCATION): Delete.
15661         * config/rx/rx.h (TRULY_NOOP_TRUNCATION): Delete.
15662         * config/s390/s390.h (TRULY_NOOP_TRUNCATION): Delete.
15663         * config/sh/sh.h (MAYBE_BASE_REGISTER_RTX_P): Remove
15664         TRULY_NOOP_TRUNCATION condition.
15665         (MAYBE_INDEX_REGISTER_RTX_P): Likewise.
15666         (TRULY_NOOP_TRUNCATION): Delete.
15667         * config/sparc/sparc.h (TRULY_NOOP_TRUNCATION): Delete.
15668         * config/spu/spu.h (TRULY_NOOP_TRUNCATION): Delete.
15669         * config/spu/spu.c (spu_truly_noop_truncation): New function.
15670         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15671         * config/stormy16/stormy16.h (TRULY_NOOP_TRUNCATION): Delete.
15672         * config/tilegx/tilegx.h (TRULY_NOOP_TRUNCATION): Delete.
15673         * config/tilegx/tilegx.c (tilegx_truly_noop_truncation): New fuction.
15674         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15675         * config/tilegx/tilegx.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15676         rather than TRULY_NOOP_TRUNCATION in comments.
15677         * config/tilepro/tilepro.h (TRULY_NOOP_TRUNCATION): Delete.
15678         * config/v850/v850.h (TRULY_NOOP_TRUNCATION): Delete.
15679         * config/vax/vax.h (TRULY_NOOP_TRUNCATION): Delete.
15680         * config/visium/visium.h (TRULY_NOOP_TRUNCATION): Delete.
15681         * config/xtensa/xtensa.h (TRULY_NOOP_TRUNCATION): Delete.
15682         * system.h (TRULY_NOOP_TRUNCATION): Poison.
15684 2017-09-15  Christophe Lyon  <christophe.lyon@linaro.org>
15686         PR target/67591
15687         * config/arm/arm.md (*cmp_and): Add enabled_for_depr_it attribute.
15688         (*cmp_ior): Likewise.
15689         (*ior_scc_scc): Add alternative for enabled_for_depr_it attribute.
15690         (*ior_scc_scc_cmp): Likewise.
15691         (*and_scc_scc): Likewise.
15692         (*and_scc_scc_cmp): Likewise.
15694 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15695             Alan Hayard  <alan.hayward@arm.com>
15696             David Sherwood  <david.sherwood@arm.com>
15698         * target.def (can_change_mode_class): New hook.
15699         (mode_rep_extended): Refer to it instead of CANNOT_CHANGE_MODE_CLASS.
15700         (hard_regno_nregs): Likewise.
15701         * hooks.h (hook_bool_mode_mode_reg_class_t_true): Declare.
15702         * hooks.c (hook_bool_mode_mode_reg_class_t_true): New function.
15703         * doc/tm.texi.in (CANNOT_CHANGE_MODE_CLASS): Replace with...
15704         (TARGET_CAN_CHANGE_MODE_CLASS): ...this.
15705         (LOAD_EXTEND_OP): Update accordingly.
15706         * doc/tm.texi: Regenerate.
15707         * doc/rtl.texi: Refer to TARGET_CAN_CHANGE_MODE_CLASS instead of
15708         CANNOT_CHANGE_MODE_CLASS.
15709         * hard-reg-set.h (REG_CANNOT_CHANGE_MODE_P): Replace with...
15710         (REG_CAN_CHANGE_MODE_P): ...this new macro.
15711         * combine.c (simplify_set): Update accordingly.
15712         * emit-rtl.c (validate_subreg): Likewise.
15713         * recog.c (general_operand): Likewise.
15714         * regcprop.c (mode_change_ok): Likewise.
15715         * reload1.c (choose_reload_regs): Likewise.
15716         (inherit_piecemeal_p): Likewise.
15717         * rtlanal.c (simplify_subreg_regno): Likewise.
15718         * postreload.c (reload_cse_simplify_set): Use REG_CAN_CHANGE_MODE_P
15719         instead of CANNOT_CHANGE_MODE_CLASS.
15720         (reload_cse_simplify_operands): Likewise.
15721         * reload.c (push_reload): Use targetm.can_change_mode_class
15722         instead of CANNOT_CHANGE_MODE_CLASS.
15723         (push_reload): Likewise.  Also use REG_CAN_CHANGE_MODE_P instead of
15724         REG_CANNOT_CHANGE_MODE_P.
15725         * config/alpha/alpha.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15726         * config/alpha/alpha.c (alpha_can_change_mode_class): New function.
15727         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15728         * config/arm/arm.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15729         * config/arm/arm.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15730         (arm_can_change_mode_class): New function.
15731         * config/arm/neon.md: Refer to TARGET_CAN_CHANGE_MODE_CLASS rather
15732         than CANNOT_CHANGE_MODE_CLASS in comments.
15733         * config/i386/i386.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15734         * config/i386/i386-protos.h (ix86_cannot_change_mode_class): Delete.
15735         * config/i386/i386.c (ix86_cannot_change_mode_class): Replace with...
15736         (ix86_can_change_mode_class): ...this new function, inverting the
15737         sense of the return value.
15738         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15739         * config/ia64/ia64.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15740         * config/ia64/ia64.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15741         (ia64_can_change_mode_class): New function.
15742         * config/m32c/m32c.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15743         * config/m32c/m32c-protos.h (m32c_cannot_change_mode_class): Delete.
15744         * config/m32c/m32c.c (m32c_cannot_change_mode_class): Replace with...
15745         (m32c_can_change_mode_class): ...this new function, inverting the
15746         sense of the return value.
15747         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15748         * config/mips/mips.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15749         * config/mips/mips-protos.h (mips_cannot_change_mode_class): Delete.
15750         * config/mips/mips.c (mips_cannot_change_mode_class): Replace with...
15751         (mips_can_change_mode_class): ...this new function, inverting the
15752         sense of the return value.
15753         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15754         * config/msp430/msp430.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15755         * config/msp430/msp430.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15756         (msp430_can_change_mode_class): New function.
15757         * config/nvptx/nvptx.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15758         * config/nvptx/nvptx.c (nvptx_can_change_mode_class): New function.
15759         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15760         * config/pa/pa32-regs.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15761         * config/pa/pa64-regs.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15762         * config/pa/pa-protos.h (pa_cannot_change_mode_class): Delete.
15763         * config/pa/pa.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15764         (pa_cannot_change_mode_class): Replace with...
15765         (pa_can_change_mode_class): ...this new function, inverting the
15766         sense of the return value.
15767         (pa_modes_tieable_p): Refer to TARGET_CAN_CHANGE_MODE_CLASS rather
15768         than CANNOT_CHANGE_MODE_CLASS in comments.
15769         * config/pdp11/pdp11.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15770         * config/pdp11/pdp11-protos.h (pdp11_cannot_change_mode_class): Delete.
15771         * config/pdp11/pdp11.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15772         (pdp11_cannot_change_mode_class): Replace with...
15773         (pdp11_can_change_mode_class): ...this new function, inverting the
15774         sense of the return value.
15775         * config/powerpcspe/powerpcspe.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15776         * config/powerpcspe/powerpcspe-protos.h
15777         (rs6000_cannot_change_mode_class_ptr): Delete.
15778         * config/powerpcspe/powerpcspe.c
15779         (rs6000_cannot_change_mode_class_ptr): Delete.
15780         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15781         (rs6000_option_override_internal): Assign to
15782         targetm.can_change_mode_class instead of
15783         rs6000_cannot_change_mode_class_ptr.
15784         (rs6000_cannot_change_mode_class): Replace with...
15785         (rs6000_can_change_mode_class): ...this new function, inverting the
15786         sense of the return value.
15787         (rs6000_debug_cannot_change_mode_class): Replace with...
15788         (rs6000_debug_can_change_mode_class): ...this new function.
15789         * config/riscv/riscv.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15790         * config/riscv/riscv.c (riscv_can_change_mode_class): New function.
15791         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15792         * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15793         * config/rs6000/rs6000-protos.h (rs6000_cannot_change_mode_class_ptr):
15794         Delete.
15795         * config/rs6000/rs6000.c (rs6000_cannot_change_mode_class_ptr): Delete.
15796         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15797         (rs6000_option_override_internal): Assign to
15798         targetm.can_change_mode_class instead of
15799         rs6000_cannot_change_mode_class_ptr.
15800         (rs6000_cannot_change_mode_class): Replace with...
15801         (rs6000_can_change_mode_class): ...this new function, inverting the
15802         sense of the return value.
15803         (rs6000_debug_cannot_change_mode_class): Replace with...
15804         (rs6000_debug_can_change_mode_class): ...this new function.
15805         * config/s390/s390.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15806         * config/s390/s390-protos.h (s390_cannot_change_mode_class): Delete.
15807         * config/s390/s390.c (s390_cannot_change_mode_class): Replace with...
15808         (s390_can_change_mode_class): ...this new function, inverting the
15809         sense of the return value.
15810         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15811         * config/sh/sh.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15812         * config/sh/sh-protos.h (sh_cannot_change_mode_class): Delete.
15813         * config/sh/sh.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15814         (sh_cannot_change_mode_class): Replace with...
15815         (sh_can_change_mode_class): ...this new function, inverting the
15816         sense of the return value.
15817         * config/sparc/sparc.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15818         * config/sparc/sparc.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15819         (sparc_can_change_mode_class): New function.
15820         * config/spu/spu.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15821         * config/spu/spu.c (spu_can_change_mode_class): New function.
15822         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15823         * config/visium/visium.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15824         * config/visium/visium.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15825         (visium_can_change_mode_class): New function.
15826         * system.h (CANNOT_CHANGE_MODE_CLASS): Poison.
15828 2017-09-15  Richard Biener  <rguenther@suse.de>
15830         PR tree-optimization/82217
15831         * tree-ssa-sccvn.c (visit_phi): Properly handle all VN_TOP
15832         but not undefined case.
15834 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15836         PR target/82145
15837         * postreload.c (reload_cse_simplify_operands): Skip
15838         NOTE_INSN_DELETED_LABEL similarly to skipping CODE_LABEL.
15840 2017-09-15  Richard Biener  <rguenther@suse.de>
15842         PR tree-optimization/68823
15843         * graphite-scop-detection.c (build_alias_set): If we have a
15844         possible dependence check whether we can handle them by just
15845         looking at the DRs DR_ACCESS_FNs.
15846         (build_scops): If build_alias_set fails, fail the SCOP.
15848 2017-09-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
15850         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_1_HW): New macros
15851         to support float128 built-in functions that require the ISA 3.0
15852         hardware.
15853         (BU_FLOAT128_3_HW): Likewise.
15854         (SQRTF128): Add support for the IEEE 128-bit square root and fma
15855         built-in functions.
15856         (FMAF128): Likewise.
15857         (FMAQ): Likewise.
15858         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add
15859         support for built-in functions that need the ISA 3.0 IEEE 128-bit
15860         floating point instructions.
15861         (rs6000_invalid_builtin): Likewise.
15862         (rs6000_builtin_mask_names): Likewise.
15863         * config/rs6000/rs6000.h (MASK_FLOAT128_HW): Likewise.
15864         (RS6000_BTM_FLOAT128_HW): Likewise.
15865         (RS6000_BTM_COMMON): Likewise.
15866         * config/rs6000/rs6000.md (fma<mode>4_hw): Add a generator
15867         function.
15868         * doc/extend.texi (RS/6000 built-in functions): Document the
15869         IEEE 128-bit floating point square root and fused multiply-add
15870         built-in functions.
15872 2017-09-14  Pat Haugen  <pthaugen@us.ibm.com>
15874         * config/rs6000/rs6000.c (rs6000_set_up_by_prologue): Make sure the TOC
15875         reg (r2) isn't in the set of registers defined in the prologue.
15877 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15878             Alan Hayward  <alan.hayward@arm.com>
15879             David Sherwood  <david.sherwood@arm.com>
15881         * tree-vectorizer.h (_loop_vec_info): Add max_vectorization_factor.
15882         (LOOP_VINFO_MAX_VECT_FACTOR): New macro.
15883         (LOOP_VINFO_ORIG_VECT_FACTOR): Replace with...
15884         (LOOP_VINFO_ORIG_MAX_VECT_FACTOR): ...this new macro.
15885         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Update
15886         accordingly.
15887         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15888         max_vectorization_factor.
15889         (vect_analyze_loop_2): Set LOOP_VINFO_MAX_VECT_FACTOR.
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 (vect_min_worthwhile_factor): Delete.
15896         (vect_worthwhile_without_simd_p): Declare.
15897         * tree-vect-loop.c (vect_worthwhile_without_simd_p): New function.
15898         (vectorizable_reduction): Use it.
15899         * tree-vect-stmts.c (vectorizable_shift): Likewise.
15900         (vectorizable_operation): Likewise.
15902 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15903             Alan Hayward  <alan.hayward@arm.com>
15904             David Sherwood  <david.sherwood@arm.com>
15906         * tree-vectorizer.h (vect_get_num_copies): New function.
15907         * tree-vect-data-refs.c (vect_get_data_access_cost): Use it.
15908         * tree-vect-loop.c (vectorizable_reduction): Likewise.
15909         (vectorizable_induction): Likewise.
15910         (vectorizable_live_operation): Likewise.
15911         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
15912         (vectorizable_bswap): Likewise.
15913         (vectorizable_call): Likewise.
15914         (vectorizable_conversion): Likewise.
15915         (vectorizable_assignment): Likewise.
15916         (vectorizable_shift): Likewise.
15917         (vectorizable_operation): Likewise.
15918         (vectorizable_store): Likewise.
15919         (vectorizable_load): Likewise.
15920         (vectorizable_condition): Likewise.
15921         (vectorizable_comparison): Likewise.
15922         (vect_analyze_stmt): Pass the slp node to vectorizable_live_operation.
15924 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15925             Alan Hayward  <alan.hayward@arm.com>
15926             David Sherwood  <david.sherwood@arm.com>
15928         * tree-vect-loop.c (vectorizable_induction): Use gimple_build instead
15929         of vect_init_vector.
15931 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15932             Alan Hayward  <alan.hayward@arm.com>
15933             David Sherwood  <david.sherwood@arm.com>
15935         * gimple-fold.h (gimple_build_vector_from_val): Declare, and provide
15936         an inline wrapper that provides a location.
15937         (gimple_build_vector): Likewise.
15938         * gimple-fold.c (gimple_build_vector_from_val): New function.
15939         (gimple_build_vector): Likewise.
15940         * tree-vect-loop.c (get_initial_def_for_reduction): Use the new
15941         functions to build the initial value.  Always return a gimple value.
15942         (get_initial_defs_for_reduction): Likewise.  Only compute
15943         neutral_vec once.
15944         (vect_create_epilog_for_reduction): Don't call force_gimple_operand or
15945         vect_init_vector on the results from get_initial_def(s)_for_reduction.
15946         (vectorizable_induction): Use gimple_build_vector rather than
15947         vect_init_vector.
15949 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15950             Alan Hayward  <alan.hayward@arm.com>
15951             David Sherwood  <david.sherwood@arm.com>
15953         * target.h (vec_perm_indices): New typedef.
15954         (auto_vec_perm_indices): Likewise.
15955         * optabs-query.h: Include target.h
15956         (can_vec_perm_p): Take a vec_perm_indices *.
15957         * optabs-query.c (can_vec_perm_p): Likewise.
15958         (can_mult_highpart_p): Update accordingly.  Use auto_vec_perm_indices.
15959         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15960         * tree-vect-generic.c (lower_vec_perm): Likewise.
15961         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15962         (vect_grouped_load_supported): Likewise.
15963         (vect_shift_permute_load_chain): Likewise.
15964         (vect_permute_store_chain): Use auto_vec_perm_indices.
15965         (vect_permute_load_chain): Likewise.
15966         * fold-const.c (fold_vec_perm): Take vec_perm_indices.
15967         (fold_ternary_loc): Update accordingly.  Use auto_vec_perm_indices.
15968         Update uses of can_vec_perm_p.
15969         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Replace the
15970         mode with a number of elements.  Take a vec_perm_indices *.
15971         (vect_create_epilog_for_reduction): Update accordingly.
15972         Use auto_vec_perm_indices.
15973         (have_whole_vector_shift): Likewise.  Update call to can_vec_perm_p.
15974         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15975         (vect_transform_slp_perm_load): Likewise.
15976         (vect_schedule_slp_instance): Use auto_vec_perm_indices.
15977         * tree-vectorizer.h (vect_gen_perm_mask_any): Take a vec_perm_indices.
15978         (vect_gen_perm_mask_checked): Likewise.
15979         * tree-vect-stmts.c (vect_gen_perm_mask_any): Take a vec_perm_indices.
15980         (vect_gen_perm_mask_checked): Likewise.
15981         (vectorizable_mask_load_store): Use auto_vec_perm_indices.
15982         (vectorizable_store): Likewise.
15983         (vectorizable_load): Likewise.
15984         (perm_mask_for_reverse): Likewise.  Update call to can_vec_perm_p.
15985         (vectorizable_bswap): Likewise.
15987 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15988             Alan Hayward  <alan.hayward@arm.com>
15989             David Sherwood  <david.sherwood@arm.com>
15991         * tree.h (build_vector): Take a vec<tree> instead of a tree *.
15992         * tree.c (build_vector): Likewise.
15993         (build_vector_from_ctor): Update accordingly.
15994         (build_vector_from_val): Likewise.
15995         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
15996         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15997         * tree-vect-generic.c (add_rshift): Likewise.
15998         (expand_vector_divmod): Likewise.
15999         (optimize_vector_constructor): Likewise.
16000         * tree-vect-slp.c (vect_get_constant_vectors): Likewise.
16001         (vect_transform_slp_perm_load): Likewise.
16002         (vect_schedule_slp_instance): Likewise.
16003         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
16004         (vectorizable_call): Likewise.
16005         (vect_gen_perm_mask_any): Likewise.  Add elements in order.
16006         * expmed.c (make_tree): Likewise.
16007         * fold-const.c (fold_negate_expr_1): Use auto_vec<tree> when building
16008         a vector passed to build_vector.
16009         (fold_convert_const): Likewise.
16010         (exact_inverse): Likewise.
16011         (fold_ternary_loc): Likewise.
16012         (fold_relational_const): Likewise.
16013         (const_binop): Likewise.  Use VECTOR_CST_ELT directly when operating
16014         on VECTOR_CSTs, rather than going through vec_cst_ctor_to_array.
16015         (const_unop): Likewise.  Store the reduction accumulator in a
16016         variable rather than an array.
16017         (vec_cst_ctor_to_array): Take the number of elements as a parameter.
16018         (fold_vec_perm): Update calls accordingly.  Use auto_vec<tree> for
16019         the new vector, rather than constructing it after the input arrays.
16020         (native_interpret_vector): Use auto_vec<tree> when building
16021         a vector passed to build_vector.  Add elements in order.
16022         * tree-vect-loop.c (get_initial_defs_for_reduction): Use
16023         auto_vec<tree> when building a vector passed to build_vector.
16024         (vect_create_epilog_for_reduction): Likewise.
16025         (vectorizable_induction): Likewise.
16026         (get_initial_def_for_reduction): Likewise.  Fix indentation of
16027         case statements.
16028         * config/sparc/sparc.c (sparc_handle_vis_mul8x16): Change n_elts
16029         to a vec<tree> *.
16030         (sparc_fold_builtin): Use auto_vec<tree> when building a vector
16031         passed to build_vector.
16033 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
16034             Alan Hayward  <alan.hayward@arm.com>
16035             David Sherwood  <david.sherwood@arm.com>
16037         * tree-core.h (tree_base::u): Add an "nelts" field.
16038         (tree_vector): Use VECTOR_CST_NELTS as the length.
16039         * tree.c (tree_size): Likewise.
16040         (make_vector): Initialize VECTOR_CST_NELTS.
16041         * tree.h (VECTOR_CST_NELTS): Use the u.nelts field.
16042         * cfgexpand.c (expand_debug_expr): Use VECTOR_CST_NELTS instead of
16043         TYPE_VECTOR_SUBPARTS.
16044         * expr.c (const_vector_mask_from_tree): Consistently use "units"
16045         as the number of units, setting it from VECTOR_CST_NELTS.
16046         (const_vector_from_tree): Likewise.
16047         * fold-const.c (negate_expr_p): Use VECTOR_CST_NELTS instead of
16048         TYPE_VECTOR_SUBPARTS for the number of elements in a VECTOR_CST.
16049         (fold_negate_expr_1): Likewise.
16050         (fold_convert_const): Likewise.
16051         (const_binop): Likewise.  Differentiate the number of output and
16052         input elements.
16053         (const_unop): Likewise.
16054         (fold_ternary_loc): Use VECTOR_CST_NELTS for the number of elements
16055         in a VECTOR_CST, asserting that it is the same as TYPE_VECTOR_SUBPARTS
16056         in cases that did the opposite.
16058 2017-09-14  Richard Biener  <rguenther@suse.de>
16060         * tree-ssa-sccvn.c (visit_phi): Merge undefined values similar
16061         to VN_TOP.
16063 2017-09-14  Eric Botcazou  <ebotcazou@adacore.com>
16065         * dwarf2out.c (dwarf2out_source_line): Remove superfluous test.
16067 2017-09-14  Jakub Jelinek  <jakub@redhat.com>
16069         PR target/81325
16070         * cfgbuild.c (find_bb_boundaries): Ignore debug insns in decisions
16071         if and where to split a bb, except for splitting before debug insn
16072         sequences followed by non-label real insn.  Delete debug insns
16073         in between basic blocks.
16075         * combine.c (make_compound_operation_int): Formatting fixes.
16077         * config/alpha/elf.h (LINK_EH_SPEC): Add -static-pie support.
16078         * config/alpha/linux.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
16079         * config/netbsd.h (LINK_EH_SPEC): Likewise.
16080         * config/sol2.h (LINK_EH_SPEC): Likewise.
16081         * config/arm/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
16082         * config/s390/linux.h (LINK_SPEC): Likewise.
16083         * config/freebsd.h (LINK_EH_SPEC): Likewise.
16084         * config/openbsd.h (LINK_EH_SPEC): Likewise.
16085         * config/lm32/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
16086         * config/aarch64/aarch64-linux.h (LINUX_TARGET_LINK_SPEC): Likewise.
16087         * config/powerpcspe/sysv4.h (LINK_EH_SPEC): Likewise.
16088         * config/bfin/linux.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
16089         * config/i386/gnu-user64.h (GNU_USER_TARGET_LINK_SPEC): Fix a typo.
16090         * config/i386/gnu-user.h (GNU_USER_TARGET_LINK_SPEC): Formatting fix.
16092 2017-09-13  Jakub Jelinek  <jakub@redhat.com>
16094         * config/rs6000/sysv4.h (STARTFILE_LINUX_SPEC): Add -static-pie
16095         support.
16096         (ENDFILE_LINUX_SPEC): Likewise.
16097         (LINK_EH_SPEC): Likewise.
16098         * config/rs6000/linux64.h (LINK_SHLIB_SPEC): Likewise.
16099         (LINK_OS_LINUX_SPEC32): Likewise.
16100         (LINK_OS_LINUX_SPEC64): Likewise.
16101         * config/rs6000/linux.h (LINK_SHLIB_SPEC): Likewise.
16102         (LINK_OS_LINUX_SPEC): Likewise.
16104 2017-09-13  Martin Liska  <mliska@suse.cz>
16106         PR middle-end/82154
16107         * stmt.c (expand_sjlj_dispatch_table): Use CASE_LOW when
16108         CASE_HIGH is NULL_TREE.
16110 2017-09-13  Richard Sandiford  <richard.sandiford@linaro.org>
16111             Alan Hayward  <alan.hayward@arm.com>
16112             David Sherwood  <david.sherwood@arm.com>
16114         * target.def (secondary_memory_needed): New hook.
16115         (secondary_reload): Refer to TARGET_SECONDARY_MEMORY_NEEDED
16116         instead of SECONDARY_MEMORY_NEEDED.
16117         (secondary_memory_needed_mode): Likewise.
16118         * hooks.h (hook_bool_mode_reg_class_t_reg_class_t_false): Declare.
16119         * hooks.c (hook_bool_mode_reg_class_t_reg_class_t_false): New function.
16120         * doc/tm.texi.in (SECONDARY_MEMORY_NEEDED): Replace with...
16121         (TARGET_SECONDARY_MEMORY_NEEDED): ...this.
16122         (SECONDARY_MEMORY_NEEDED_RTX): Update reference accordingly.
16123         * doc/tm.texi: Regenerate.
16124         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED): Delete.
16125         * config/alpha/alpha.c (alpha_secondary_memory_needed): New function.
16126         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16127         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED): Delete.
16128         * config/i386/i386-protos.h (ix86_secondary_memory_needed): Delete.
16129         * config/i386/i386.c (inline_secondary_memory_needed): Put the
16130         mode argument first and change the reg_class arguments to reg_class_t.
16131         (ix86_secondary_memory_needed): Likewise.  Remove the strict parameter.
16132         Make static.  Update the call to inline_secondary_memory_needed.
16133         (ix86_register_move_cost): Update the call to
16134         inline_secondary_memory_needed.
16135         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16136         * config/ia64/ia64.h (SECONDARY_MEMORY_NEEDED): Delete commented-out
16137         definition.
16138         * config/ia64/ia64.c (spill_xfmode_rfmode_operand): Refer to
16139         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
16140         in comment.
16141         * config/mips/mips.h (SECONDARY_MEMORY_NEEDED): Delete.
16142         * config/mips/mips-protos.h (mips_secondary_memory_needed): Delete.
16143         * config/mips/mips.c (mips_secondary_memory_needed): Make static
16144         and match hook interface.  Add comment from mips.h.
16145         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16146         * config/mmix/mmix.md (truncdfsf2): Refer to
16147         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
16148         in comment.
16149         * config/pa/pa-64.h (SECONDARY_MEMORY_NEEDED): Rename to...
16150         (PA_SECONDARY_MEMORY_NEEDED): ...this, and put the mode argument first.
16151         * config/pa/pa.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16152         (pa_secondary_memory_needed): New function.
16153         * config/pdp11/pdp11.h (SECONDARY_MEMORY_NEEDED): Delete.
16154         * config/pdp11/pdp11-protos.h (pdp11_secondary_memory_needed): Delete.
16155         * config/pdp11/pdp11.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16156         (pdp11_secondary_memory_needed): Make static and match hook interface.
16157         * config/powerpcspe/powerpcspe.h (SECONDARY_MEMORY_NEEDED): Delete.
16158         * config/powerpcspe/powerpcspe-protos.h
16159         (rs6000_secondary_memory_needed_ptr): Delete.
16160         * config/powerpcspe/powerpcspe.c (rs6000_secondary_memory_needed_ptr):
16161         Delete.
16162         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16163         (rs6000_option_override_internal): Assign to
16164         targetm.secondary_memory_needed rather than
16165         rs6000_secondary_memory_needed_ptr.
16166         (rs6000_secondary_memory_needed): Match hook interface.
16167         (rs6000_debug_secondary_memory_needed): Likewise.
16168         * config/riscv/riscv.h (SECONDARY_MEMORY_NEEDED): Delete.
16169         * config/riscv/riscv.c (riscv_secondary_memory_needed): New function.
16170         (riscv_register_move_cost): Use it instead of SECONDARY_MEMORY_NEEDED.
16171         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16172         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED): Delete.
16173         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_ptr):
16174         Delete.
16175         * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_ptr): 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/s390/s390.h (SECONDARY_MEMORY_NEEDED): Delete.
16183         * config/s390/s390.c (s390_secondary_memory_needed): New function.
16184         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16185         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED): Delete.
16186         * config/sparc/sparc.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16187         (sparc_secondary_memory_needed): New function.
16188         * lra-constraints.c (check_and_process_move): Refer to
16189         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
16190         in comment.
16191         (curr_insn_transform): Likewise.
16192         (process_alt_operands): Use targetm.secondary_memory_needed
16193         instead of TARGET_SECONDARY_MEMORY_NEEDED.
16194         (check_secondary_memory_needed_p): Likewise.
16195         (choose_split_class): Likewise.
16196         * reload.c: Unconditionally include code that was previously
16197         conditional on SECONDARY_MEMORY_NEEDED.
16198         (push_secondary_reload): Use targetm.secondary_memory_needed
16199         instead of TARGET_SECONDARY_MEMORY_NEEDED.
16200         (push_reload): Likewise.
16201         * reload1.c: Unconditionally include code that was previously
16202         conditional on SECONDARY_MEMORY_NEEDED.
16203         (choose_reload_regs): Use targetm.secondary_memory_needed
16204         instead of TARGET_SECONDARY_MEMORY_NEEDED.
16205         (gen_reload): Likewise.
16206         * system.h (SECONDARY_MEMORY_NEEDED): Poison.
16208 2017-09-13  Richard Sandiford  <richard.sandiford@linaro.org>
16209             Alan Hayward  <alan.hayward@arm.com>
16210             David Sherwood  <david.sherwood@arm.com>
16212         * target.def (secondary_memory_needed_mode): New hook:
16213         * targhooks.c (default_secondary_memory_needed_mode): Declare.
16214         * targhooks.h (default_secondary_memory_needed_mode): New function.
16215         * doc/tm.texi.in (SECONDARY_MEMORY_NEEDED_MODE): Replace with...
16216         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): ...this.
16217         * doc/tm.texi: Regenerate.
16218         * lra-constraints.c (check_and_process_move): Use
16219         targetm.secondary_memory_needed_mode instead of
16220         TARGET_SECONDARY_MEMORY_NEEDED_MODE.
16221         (curr_insn_transform): Likewise.
16222         * reload.c (get_secondary_mem): Likewise.
16223         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16224         * config/alpha/alpha.c (alpha_secondary_memory_needed_mode): New
16225         function.
16226         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16227         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16228         * config/i386/i386.c (ix86_secondary_memory_needed_mode): New function.
16229         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16230         * config/powerpcspe/powerpcspe.h (SECONDARY_MEMORY_NEEDED_MODE):
16231         Delete.
16232         * config/powerpcspe/powerpcspe-protos.h
16233         (rs6000_secondary_memory_needed_mode): Delete.
16234         * config/powerpcspe/powerpcspe.c
16235         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16236         (rs6000_secondary_memory_needed_mode): Make static.
16237         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16238         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_mode):
16239         Delete.
16240         * config/rs6000/rs6000.c (TARGET_SECONDARY_MEMORY_NEEDED_MODE):
16241         Redefine.
16242         (rs6000_secondary_memory_needed_mode): Make static.
16243         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16244         * config/s390/s390.c (s390_secondary_memory_needed_mode): New function.
16245         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16246         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16247         * config/sparc/sparc.c (TARGET_SECONDARY_MEMORY_NEEDED_MODE):
16248         Redefine.
16249         (sparc_secondary_memory_needed_mode): New function.
16250         * system.h (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Poison.
16252 2017-09-13  Jackson Woodruff  <jackson.woodruff@arm.com>
16254         * config/aarch64/constraints.md (Umq): New constraint.
16255         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>):
16256         Change to use Umq.
16257         (mov<mode>): Update condition.
16259 2017-09-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16261         * gimple-ssa-store-merging.c (sort_by_bitpos): Compare store order
16262         when bitposition is the same.
16264 2017-09-13  Richard Biener  <rguenther@suse.de>
16266         * dwarf2out.c (output_die_symbol): Remove.
16267         (output_die): Do not output a DIEs symbol.
16269 2017-09-13  Richard Biener  <rguenther@suse.de>
16271         PR middle-end/82128
16272         * gimple-fold.c (gimple_fold_call): Update SSA name in-place to
16273         default-def to avoid breaking iterator update with the weird
16274         interaction with cgraph_update_edges_for_call_stmt_node.
16276 2017-09-13  Richard Biener  <rguenther@suse.de>
16278         * tree-cfg.c (verify_gimple_assign_binary): Add verification
16279         for WIDEN_SUM_EXPR, VEC_WIDEN_MULT_{HI,LO,EVEN,ODD}_EXPR,
16280         VEC_PACK_{TRUNC,SAT,FIX_TRUNC}_EXPR.
16281         (verify_gimple_assign_ternary): Add verification for DOT_PROD_EXPR.
16283 2017-09-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
16285         * config/aarch64/aarch64.c (aarch64_override_options_after_change_1):
16286         Disable pc relative literal load irrespective of
16287         TARGET_FIX_ERR_A53_84341 for default.
16289 2017-09-12  Eric Botcazou  <ebotcazou@adacore.com>
16291         * config/sparc/sparc.c (output_return): Output the source location of
16292         the insn in the delay slot, if any.
16293         (output_sibcall): Likewise.
16295 2017-09-12  H.J. Lu  <hongjiu.lu@intel.com>
16297         PR driver/81498
16298         * common.opt (-static-pie): New alias.
16299         (shared): Negate static-pie.
16300         (-no-pie): Update help text.
16301         (-pie): Likewise.
16302         (static-pie): New option.
16303         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Add
16304         -static-pie support.
16305         (GNU_USER_TARGET_ENDFILE_SPEC): Likewise.
16306         (LINK_EH_SPEC): Likewise.
16307         (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
16308         * config/i386/gnu-user.h (GNU_USER_TARGET_LINK_SPEC): Likewise.
16309         * config/i386/gnu-user64.h (GNU_USER_TARGET_LINK_SPEC): Likewise.
16310         * gcc.c (LINK_COMMAND_SPEC): Likewise.
16311         (init_gcc_specs): Likewise.
16312         (init_spec): Likewise.
16313         (display_help): Update help message for -pie.
16314         * doc/invoke.texi: Update -pie, -no-pie and -static.  Document
16315         -static-pie.
16317 2017-09-12  Wilco Dijkstra  <wdijkstr@arm.com>
16319         * config/aarch64/aarch64.md (movsi_aarch64): Remove all '*'.
16320         (movdi_aarch64): Likewise.
16321         (movti_aarch64): Likewise.
16323 2017-09-12 Simon Wright <simon@pushface.org>
16325         PR target/80204
16326         * config/darwin-driver.c (darwin_find_version_from_kernel): Eliminate
16327         calculation of the minor version, always output as 0.
16329 2017-09-12  Jakub Jelinek  <jakub@redhat.com>
16331         PR target/82112
16332         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): For
16333         ALTIVEC_BUILTIN_VEC_LD if arg1 has array type call default_conversion
16334         on it early, rather than manual conversion late.  For
16335         ALTIVEC_BUILTIN_VEC_ST if arg2 has array type call default_conversion
16336         instead of performing manual conversion.
16338 2017-09-12  Carl Love  <cel@us.ibm.com>
16340         * config/rs6000/altivec.md (vec_widen_umult_even_v4si,
16341         vec_widen_smult_even_v4si): Add define expands for vmuleuw, vmulesw,
16342         vmulouw, vmulosw.
16343         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
16344         VMULOSW): Add definitions.
16345         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
16346         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
16347         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
16348         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
16349         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
16351 2017-09-12  James Greenhalgh  <james.greenhalgh@arm.com>
16353         * config/aarch64/aarch64.md (movdi_aarch64): Set load/store
16354         types correctly.
16355         (movti_aarch64): Likewise.
16356         (movdf_aarch64): Likewise.
16357         (movtf_aarch64): Likewise.
16358         (load_pairdi): Likewise.
16359         (store_pairdi): Likewise.
16360         (load_pairdf): Likewise.
16361         (store_pairdf): Likewise.
16362         (loadwb_pair<GPI:mode>_<P:mode>): Likewise.
16363         (storewb_pair<GPI:mode>_<P:mode>): Likewise.
16364         (ldr_got_small_<mode>): Likewise.
16365         (ldr_got_small_28k_<mode>): Likewise.
16366         (ldr_got_tiny): Likewise.
16367         * config/aarch64/iterators.md (ldst_sz): New.
16368         (ldpstp_sz): Likewise.
16369         * config/aarch64/thunderx.md (thunderx_storepair): Split store_8
16370         to store_16.
16371         (thunderx_load): Split load_8 to load_16.
16372         * config/aarch64/thunderx2t99.md (thunderx2t99_loadpair): Split
16373         load_8 to load_16.
16374         (thunderx2t99_storepair_basic): Split store_8 to store_16.
16375         * config/arm/xgene1.md (xgene1_load_pair): Split load_8 to load_16.
16376         (xgene1_store_pair): Split store_8 to store_16.
16377         * config/aarch64/falkor.md (falkor_ld_3_ld): Split load_8 to load_16.
16378         (falkor_st_0_st_sd): Split store_8 to store_16.
16380 2017-09-12  James Greenhalgh  <james.greenhalgh@arm.com>
16382         * config/arm/types.md (type): Rename load1/2/3/4 to load_4/8/12/16
16383         and store1/2/3/4 to store_4/8/12/16.
16384         * config/aarch64/aarch64.md: Update for rename.
16385         * config/arm/arm.md: Likewise.: Likewise.
16386         * config/arm/arm.c: Likewise.
16387         * config/arm/thumb1.md: Likewise.
16388         * config/arm/thumb2.md: Likewise.
16389         * config/arm/vfp.md: Likewise.
16390         * config/arm/arm-generic.md: Likewise.
16391         * config/arm/arm1020e.md: Likewise.
16392         * config/arm/arm1026ejs.md: Likewise.
16393         * config/arm/arm1136jfs.md: Likewise.
16394         * config/arm/arm926ejs.md: Likewise.
16395         * config/arm/cortex-a15.md: Likewise.
16396         * config/arm/cortex-a17.md: Likewise.
16397         * config/arm/cortex-a5.md: Likewise.
16398         * config/arm/cortex-a53.md: Likewise.
16399         * config/arm/cortex-a57.md: Likewise.
16400         * config/arm/cortex-a7.md: Likewise.
16401         * config/arm/cortex-a8.md: Likewise.
16402         * config/arm/cortex-a9.md: Likewise.
16403         * config/arm/cortex-m4.md: Likewise.
16404         * config/arm/cortex-m7.md: Likewise.
16405         * config/arm/cortex-r4.md: Likewise.
16406         * config/arm/exynos-m1.md: Likewise.
16407         * config/arm/fa526.md: Likewise.
16408         * config/arm/fa606te.md: Likewise.
16409         * config/arm/fa626te.md: Likewise.
16410         * config/arm/fa726te.md: Likewise.
16411         * config/arm/fmp626.md: Likewise.
16412         * config/arm/iwmmxt.md: Likewise.
16413         * config/arm/ldmstm.md: Likewise.
16414         * config/arm/marvell-pj4.md: Likewise.
16415         * config/arm/xgene1.md: Likewise.
16416         * config/aarch64/thunderx.md: Likewise.
16417         * config/aarch64/thunderx2t99.md: Likewise.
16418         * config/aarch64/falkor.md: Likewise.
16420 2017-09-12  Martin Liska  <mliska@suse.cz>
16422         * attribs.c (private_lookup_attribute): New function.
16423         * attribs.h (private_lookup_attribute): Declared here.
16424         (lookup_attribute): Called from this place.
16426 2017-09-12  Richard Biener  <rguenther@suse.de>
16428         PR tree-optimization/82157
16429         * tree-ssa-pre.c (remove_dead_inserted_code): Do not remove
16430         stmts with side-effects.
16432 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16433             Alan Hayward  <alan.hayward@arm.com>
16434             David Sherwood <david.sherwood@arm.com>
16436         * target.def (hard_regno_nregs): New hook.
16437         (class_max_nregs): Refer to it instead of HARD_REGNO_NREGS.
16438         * targhooks.h (default_hard_regno_nregs): Declare.
16439         * targhooks.c (default_hard_regno_nregs): New function.
16440         * doc/tm.texi.in (HARD_REGNO_NREGS): Replace with...
16441         (TARGET_HARD_REGNO_NREGS): ...this hook.
16442         (HARD_REGNO_NREGS_HAS_PADDING): Update accordingly.
16443         (CLASS_MAX_NREGS): Likewise.
16444         * doc/tm.texi: Regenerate.
16445         * reginfo.c (init_reg_modes_target): Use targetm.hard_regno_nregs
16446         instead of HARD_REGNO_NREGS.
16447         * rtl.h (REG_NREGS): Refer to TARGET_HARD_REGNO_NREGS rather than
16448         HARD_REGNO_NREGS in the comment.
16449         * config/aarch64/aarch64.h (HARD_REGNO_NREGS): Delete.
16450         * config/aarch64/aarch64-protos.h (aarch64_hard_regno_nregs): Delete.
16451         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs): Make static.
16452         Return an unsigned int.
16453         (TARGET_HARD_REGNO_NREGS): Redefine.
16454         * config/alpha/alpha.h (HARD_REGNO_NREGS): Delete.
16455         * config/arc/arc.h (HARD_REGNO_NREGS): Delete.
16456         * config/arc/arc.c (TARGET_HARD_REGNO_NREGS): Redefine.
16457         (arc_hard_regno_nregs): New function.
16458         * config/arm/arm.h (HARD_REGNO_NREGS): Delete.
16459         * config/arm/arm.c (TARGET_HARD_REGNO_NREGS): Redefine.
16460         (arm_hard_regno_nregs): New function.
16461         * config/avr/avr.h (HARD_REGNO_NREGS): Delete.
16462         * config/bfin/bfin.h (HARD_REGNO_NREGS): Delete.
16463         * config/bfin/bfin.c (bfin_hard_regno_nregs): New function.
16464         (TARGET_HARD_REGNO_NREGS): Redefine.
16465         * config/c6x/c6x.h (HARD_REGNO_NREGS): Delete.
16466         * config/cr16/cr16.h (LONG_REG_P): Use targetm.hard_regno_nregs.
16467         (HARD_REGNO_NREGS): Delete.
16468         * config/cr16/cr16.c (TARGET_HARD_REGNO_NREGS): Redefine.
16469         (cr16_hard_regno_nregs): New function.
16470         (cr16_memory_move_cost): Use it instead of HARD_REGNO_NREGS.
16471         * config/cris/cris.h (HARD_REGNO_NREGS): Delete.
16472         * config/cris/cris.c (TARGET_HARD_REGNO_NREGS): Redefine.
16473         (cris_hard_regno_nregs): New function.
16474         * config/epiphany/epiphany.h (HARD_REGNO_NREGS): Delete.
16475         * config/fr30/fr30.h (HARD_REGNO_NREGS): Delete.
16476         (CLASS_MAX_NREGS): Use targetm.hard_regno_nregs.
16477         * config/frv/frv.h (HARD_REGNO_NREGS): Delete.
16478         (CLASS_MAX_NREGS): Remove outdated copy of documentation.
16479         * config/frv/frv-protos.h (frv_hard_regno_nregs): Delete.
16480         * config/frv/frv.c (TARGET_HARD_REGNO_NREGS): Redefine.
16481         (frv_hard_regno_nregs): Make static.  Take and return an
16482         unsigned int.
16483         (frv_class_max_nregs): Remove outdated copy of documentation.
16484         * config/ft32/ft32.h (HARD_REGNO_NREGS): Delete.
16485         * config/h8300/h8300.h (HARD_REGNO_NREGS): Delete.
16486         * config/h8300/h8300-protos.h (h8300_hard_regno_nregs): Delete.
16487         * config/h8300/h8300.c (h8300_hard_regno_nregs): Delete.
16488         * config/i386/i386.h (HARD_REGNO_NREGS): Delete.
16489         * config/i386/i386.c (ix86_hard_regno_nregs): New function.
16490         (TARGET_HARD_REGNO_NREGS): Redefine.
16491         * config/ia64/ia64.h (HARD_REGNO_NREGS): Delete.
16492         (CLASS_MAX_NREGS): Update comment.
16493         * config/ia64/ia64.c (TARGET_HARD_REGNO_NREGS): Redefine.
16494         (ia64_hard_regno_nregs): New function.
16495         * config/iq2000/iq2000.h (HARD_REGNO_NREGS): Delete.
16496         * config/lm32/lm32.h (HARD_REGNO_NREGS): Delete.
16497         * config/m32c/m32c.h (HARD_REGNO_NREGS): Delete.
16498         * config/m32c/m32c-protos.h (m32c_hard_regno_nregs): Delete.
16499         * config/m32c/m32c.c (m32c_hard_regno_nregs_1): Take and return
16500         an unsigned int.
16501         (m32c_hard_regno_nregs): Likewise.  Make static.
16502         (TARGET_HARD_REGNO_NREGS): Redefine.
16503         * config/m32r/m32r.h (HARD_REGNO_NREGS): Delete.
16504         * config/m68k/m68k.h (HARD_REGNO_NREGS): Delete.
16505         * config/m68k/m68k.c (TARGET_HARD_REGNO_NREGS): Redefine.
16506         (m68k_hard_regno_nregs): New function.
16507         * config/mcore/mcore.h (HARD_REGNO_NREGS): Delete.
16508         * config/microblaze/microblaze.h (HARD_REGNO_NREGS): Delete.
16509         * config/mips/mips.h (HARD_REGNO_NREGS): Delete.
16510         * config/mips/mips-protos.h (mips_hard_regno_nregs): Delete.
16511         * config/mips/mips.c (mips_hard_regno_nregs): Make static.
16512         Take and return an unsigned int.
16513         (TARGET_HARD_REGNO_NREGS): Redefine.
16514         * config/mmix/mmix.h (HARD_REGNO_NREGS): Delete.
16515         (CLASS_MAX_NREGS): Use targetm.hard_regno_nregs.
16516         * config/mn10300/mn10300.h (HARD_REGNO_NREGS): Delete.
16517         * config/moxie/moxie.h (HARD_REGNO_NREGS): Delete.
16518         * config/msp430/msp430.h (HARD_REGNO_NREGS): Delete.
16519         * config/msp430/msp430-protos.h (msp430_hard_regno_nregs): Delete.
16520         * config/msp430/msp430.c (TARGET_HARD_REGNO_NREGS): Redefine.
16521         (msp430_hard_regno_nregs): Make static.  Take and return an
16522         unsigned int.
16523         * config/nds32/nds32.h (HARD_REGNO_NREGS): Delete.
16524         * config/nds32/nds32-protos.h (nds32_hard_regno_nregs): Delete.
16525         * config/nds32/nds32.c (nds32_hard_regno_nregs): Delete.
16526         (nds32_hard_regno_mode_ok): Use targetm.hard_regno_nregs.
16527         * config/nios2/nios2.h (HARD_REGNO_NREGS): Delete.
16528         * config/nvptx/nvptx.h (HARD_REGNO_NREGS): Delete.
16529         * config/nvptx/nvptx.c (nvptx_hard_regno_nregs): New function.
16530         (TARGET_HARD_REGNO_NREGS): Redefine.
16531         * config/pa/pa32-regs.h (HARD_REGNO_NREGS): Rename to...
16532         (PA_HARD_REGNO_NREGS): ...this.
16533         * config/pa/pa64-regs.h (HARD_REGNO_NREGS): Rename to...
16534         (PA_HARD_REGNO_NREGS): ...this.
16535         * config/pa/pa.c (TARGET_HARD_REGNO_NREGS): Redefine.
16536         (pa_hard_regno_nregs): New function.
16537         * config/pdp11/pdp11.h (HARD_REGNO_NREGS): Delete.
16538         * config/pdp11/pdp11.c (TARGET_HARD_REGNO_NREGS): Redefine.
16539         (pdp11_hard_regno_nregs): New function.
16540         * config/powerpcspe/powerpcspe.h (HARD_REGNO_NREGS): Delete.
16541         * config/powerpcspe/powerpcspe.c (TARGET_HARD_REGNO_NREGS): Redefine.
16542         (rs6000_hard_regno_nregs_hook): New function.
16543         * config/riscv/riscv.h (HARD_REGNO_NREGS): Delete.
16544         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): Delete.
16545         * config/riscv/riscv.c (riscv_hard_regno_nregs): Make static.
16546         Take and return an unsigned int.  Move earlier in file.
16547         (TARGET_HARD_REGNO_NREGS): Redefine.
16548         * config/rl78/rl78.h (HARD_REGNO_NREGS): Delete.
16549         * config/rl78/rl78-protos.h (rl78_hard_regno_nregs): Delete.
16550         * config/rl78/rl78.c (TARGET_HARD_REGNO_NREGS): Reefine.
16551         (rl78_hard_regno_nregs): Make static.  Take and return an
16552         unsigned int.
16553         * config/rs6000/rs6000.h (HARD_REGNO_NREGS): Delete.
16554         * config/rs6000/rs6000.c (TARGET_HARD_REGNO_NREGS): Redefine.
16555         (rs6000_hard_regno_nregs_hook): New function.
16556         * config/rx/rx.h (HARD_REGNO_NREGS): Delete.
16557         * config/rx/rx.c (rx_hard_regno_nregs): New function.
16558         (TARGET_HARD_REGNO_NREGS): Redefine.
16559         * config/s390/s390.h (HARD_REGNO_NREGS): Delete.
16560         * config/s390/s390.c (REGNO_PAIR_OK): Use s390_hard_regno_nregs
16561         instead of HARD_REGNO_NREGS.
16562         (s390_hard_regno_nregs): New function.
16563         (s390_hard_regno_mode_ok): Add comment from s390.h.
16564         (TARGET_HARD_REGNO_NREGS): Redefine.
16565         * config/sh/sh.h (HARD_REGNO_NREGS): Delete.
16566         * config/sh/sh.c (TARGET_HARD_REGNO_NREGS): Redefine.
16567         (sh_hard_regno_nregs): New function.
16568         (sh_pass_in_reg_p): Use it.
16569         * config/sparc/sparc.h (HARD_REGNO_NREGS): Delete.
16570         * config/sparc/sparc.c (TARGET_HARD_REGNO_NREGS): Redefine.
16571         (sparc_hard_regno_nregs): New function.
16572         * config/spu/spu.h (HARD_REGNO_NREGS): Delete.
16573         * config/spu/spu.c (spu_hard_regno_nregs): New function.
16574         (spu_function_arg_advance): Use it, supplying a valid register number.
16575         (TARGET_HARD_REGNO_NREGS): Redefine.
16576         * config/stormy16/stormy16.h (HARD_REGNO_NREGS): Delete.
16577         * config/tilegx/tilegx.h (HARD_REGNO_NREGS): Delete.
16578         * config/tilepro/tilepro.h (HARD_REGNO_NREGS): Delete.
16579         * config/v850/v850.h (HARD_REGNO_NREGS): Delete.
16580         * config/vax/vax.h (HARD_REGNO_NREGS): Delete.
16581         * config/visium/visium.h (HARD_REGNO_NREGS): Delete.
16582         (CLASS_MAX_NREGS): Remove copy of old documentation.
16583         * config/visium/visium.c (TARGET_HARD_REGNO_NREGS): Redefine.
16584         (visium_hard_regno_nregs): New function.
16585         (visium_hard_regno_mode_ok): Use it instead of HARD_REGNO_NREGS.
16586         * config/xtensa/xtensa.h (HARD_REGNO_NREGS): Delete.
16587         * config/xtensa/xtensa.c (TARGET_HARD_REGNO_NREGS): Redefine.
16588         xtensa_hard_regno_nregs): New function.
16589         * system.h (HARD_REGNO_NREGS): Poison.
16591 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16593         * config/arm/arm.h (THUMB_SECONDARY_INPUT_RELOAD_CLASS): Use
16594         hard_regno_nregs instead of HARD_REGNO_NREGS.
16595         (THUMB_SECONDARY_OUTPUT_RELOAD_CLASS): Likewise.
16596         * config/c6x/c6x.c (c6x_expand_prologue): Likewise.
16597         (c6x_expand_epilogue): Likewise.
16598         * config/frv/frv.c (frv_alloc_temp_reg): Likewise.
16599         (frv_read_iacc_argument): Likewise.
16600         * config/sh/sh.c: Include regs.h.
16601         (sh_print_operand): Use hard_regno_nregs instead of HARD_REGNO_NREGS.
16602         (regs_used): Likewise.
16603         (output_stack_adjust): Likewise.
16604         * config/xtensa/xtensa.c (xtensa_copy_incoming_a7): Likewise.
16605         * expmed.c: Include regs.h.
16606         (store_bit_field_1): Use hard_regno_nregs instead of HARD_REGNO_NREGS.
16607         * ree.c: Include regs.h.
16608         (combine_reaching_defs): Use hard_regno_nregs instead of
16609         HARD_REGNO_NREGS.
16610         (add_removable_extension): Likewise.
16612 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16614         * regs.h (hard_regno_nregs): Turn into a function.
16615         (end_hard_regno): Update accordingly.
16616         * caller-save.c (setup_save_areas): Likewise.
16617         (save_call_clobbered_regs): Likewise.
16618         (replace_reg_with_saved_mem): Likewise.
16619         (insert_restore): Likewise.
16620         (insert_save): Likewise.
16621         * combine.c (can_change_dest_mode): Likewise.
16622         (move_deaths): Likewise.
16623         (distribute_notes): Likewise.
16624         * config/mips/mips.c (mips_hard_regno_call_part_clobbered): Likewise.
16625         * config/powerpcspe/powerpcspe.c (rs6000_cannot_change_mode_class)
16626         (rs6000_split_multireg_move): Likewise.
16627         (rs6000_register_move_cost): Likewise.
16628         (rs6000_memory_move_cost): Likewise.
16629         * config/rs6000/rs6000.c (rs6000_cannot_change_mode_class): Likewise.
16630         (rs6000_split_multireg_move): Likewise.
16631         (rs6000_register_move_cost): Likewise.
16632         (rs6000_memory_move_cost): Likewise.
16633         * cselib.c (cselib_reset_table): Likewise.
16634         (cselib_lookup_1): Likewise.
16635         * emit-rtl.c (set_mode_and_regno): Likewise.
16636         * function.c (aggregate_value_p): Likewise.
16637         * ira-color.c (setup_profitable_hard_regs): Likewise.
16638         (check_hard_reg_p): Likewise.
16639         (calculate_saved_nregs): Likewise.
16640         (assign_hard_reg): Likewise.
16641         (improve_allocation): Likewise.
16642         (calculate_spill_cost): Likewise.
16643         * ira-emit.c (modify_move_list): Likewise.
16644         * ira-int.h (ira_hard_reg_set_intersection_p): Likewise.
16645         (ira_hard_reg_in_set_p): Likewise.
16646         * ira.c (setup_reg_mode_hard_regset): Likewise.
16647         (clarify_prohibited_class_mode_regs): Likewise.
16648         (check_allocation): Likewise.
16649         * lra-assigns.c (find_hard_regno_for_1): Likewise.
16650         (lra_setup_reg_renumber): Likewise.
16651         (setup_try_hard_regno_pseudos): Likewise.
16652         (spill_for): Likewise.
16653         (assign_hard_regno): Likewise.
16654         (setup_live_pseudos_and_spill_after_risky_transforms): Likewise.
16655         * lra-constraints.c (in_class_p): Likewise.
16656         (lra_constraint_offset): Likewise.
16657         (simplify_operand_subreg): Likewise.
16658         (lra_constraints): Likewise.
16659         (split_reg): Likewise.
16660         (split_if_necessary): Likewise.
16661         (invariant_p): Likewise.
16662         (inherit_in_ebb): Likewise.
16663         * lra-lives.c (process_bb_lives): Likewise.
16664         * lra-remat.c (reg_overlap_for_remat_p): Likewise.
16665         (get_hard_regs): Likewise.
16666         (do_remat): Likewise.
16667         * lra-spills.c (assign_spill_hard_regs): Likewise.
16668         * mode-switching.c (create_pre_exit): Likewise.
16669         * postreload.c (reload_combine_recognize_pattern): Likewise.
16670         * recog.c (peep2_find_free_register): Likewise.
16671         * regcprop.c (kill_value_regno): Likewise.
16672         (set_value_regno): Likewise.
16673         (copy_value): Likewise.
16674         (maybe_mode_change): Likewise.
16675         (find_oldest_value_reg): Likewise.
16676         (copyprop_hardreg_forward_1): Likewise.
16677         * regrename.c (check_new_reg_p): Likewise.
16678         (regrename_do_replace): Likewise.
16679         * reload.c (push_reload): Likewise.
16680         (combine_reloads): Likewise.
16681         (find_dummy_reload): Likewise.
16682         (operands_match_p): Likewise.
16683         (find_reloads): Likewise.
16684         (find_equiv_reg): Likewise.
16685         (reload_adjust_reg_for_mode): Likewise.
16686         * reload1.c (count_pseudo): Likewise.
16687         (count_spilled_pseudo): Likewise.
16688         (find_reg): Likewise.
16689         (clear_reload_reg_in_use): Likewise.
16690         (free_for_value_p): Likewise.
16691         (allocate_reload_reg): Likewise.
16692         (choose_reload_regs): Likewise.
16693         (reload_adjust_reg_for_temp): Likewise.
16694         (emit_reload_insns): Likewise.
16695         (delete_output_reload): Likewise.
16696         * rtlanal.c (subreg_get_info): Likewise.
16697         * sched-deps.c (sched_analyze_reg): Likewise.
16698         * sel-sched.c (init_regs_for_mode): Likewise.
16699         (mark_unavailable_hard_regs): Likewise.
16700         (choose_best_reg_1): Likewise.
16701         (verify_target_availability): Likewise.
16702         * valtrack.c (dead_debug_insert_temp): Likewise.
16703         * var-tracking.c (track_loc_p): Likewise.
16704         (emit_note_insn_var_location): Likewise.
16705         * varasm.c (make_decl_rtl): Likewise.
16706         * reginfo.c (choose_hard_reg_mode): Likewise.
16707         (init_reg_modes_target): Refer directly to
16708         this_target_regs->x_hard_regno_nregs.
16710 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16712         * ira-costs.c (record_operand_costs): Use in_hard_reg_set_p
16713         instead of hard_regno_nregs.
16715 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16717         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Use
16718         end_hard_regno instead of hard_regno_nregs.
16719         * config/s390/s390.c (s390_reg_clobbered_rtx): Likewise.
16720         * config/sparc/sparc.h (ASM_DECLARE_REGISTER_GLOBAL): Likewise.
16721         * config/visium/visium.c (visium_hard_regno_mode_ok): Likewise.
16722         * ira-color.c (improve_allocation): Likewise.
16723         * lra-assigns.c (find_hard_regno_for_1): Likewise.
16724         * lra-lives.c (mark_regno_live): Likewise.
16725         (mark_regno_dead): Likewise.
16726         * lra-remat.c (operand_to_remat): Likewise.
16727         * lra.c (collect_non_operand_hard_regs): Likewise.
16728         * postreload.c (reload_combine_note_store): Likewise.
16729         (move2add_valid_value_p): Likewise.
16730         * reload.c (regno_clobbered_p): Likewise.
16732 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16734         * config/frv/frv.c (FOR_EACH_REGNO): Use END_REGNO instead of
16735         hard_regno_nregs.
16736         * config/v850/v850.c (v850_reorg): Likewise.
16737         * reload.c (refers_to_regno_for_reload_p): Likewise.
16738         (find_equiv_reg): Likewise.
16739         * reload1.c (reload_reg_reaches_end_p): Likewise.
16741 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16743         * caller-save.c (add_used_regs): Use REG_NREGS instead of
16744         hard_regno_nregs.
16745         * config/aarch64/aarch64.c (aarch64_split_combinev16qi): Likewise.
16746         * config/arm/arm.c (output_move_neon): Likewise.
16747         (arm_attr_length_move_neon): Likewise.
16748         (neon_split_vcombine): Likewise.
16749         * config/c6x/c6x.c (c6x_mark_reg_read): Likewise.
16750         (c6x_mark_reg_written): Likewise.
16751         (c6x_dwarf_register_span): Likewise.
16752         * config/i386/i386.c (ix86_save_reg): Likewise.
16753         * config/ia64/ia64.c (mark_reg_gr_used_mask): Likewise.
16754         (rws_access_reg): Likewise.
16755         * config/s390/s390.c (s390_call_saved_register_used): Likewise.
16756         * mode-switching.c (create_pre_exit): Likewise.
16757         * ree.c (combine_reaching_defs): Likewise.
16758         (add_removable_extension): Likewise.
16759         * regcprop.c (find_oldest_value_reg): Likewise.
16760         (copyprop_hardreg_forward_1): Likewise.
16761         * reload.c (reload_inner_reg_of_subreg): Likewise.
16762         (push_reload): Likewise.
16763         (combine_reloads): Likewise.
16764         (find_dummy_reload): Likewise.
16765         (reload_adjust_reg_for_mode): Likewise.
16766         * reload1.c (find_reload_regs): Likewise.
16767         (forget_old_reloads_1): Likewise.
16768         (reload_reg_free_for_value_p): Likewise.
16769         (reload_adjust_reg_for_temp): Likewise.
16770         (emit_reload_insns): Likewise.
16771         (delete_output_reload): Likewise.
16772         * sel-sched.c (choose_best_reg_1): Likewise.
16773         (choose_best_pseudo_reg): Likewise.
16775 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16776             Alan Hayward  <alan.hayward@arm.com>
16777             David Sherwood <david.sherwood@arm.com>
16779         * defaults.h (SLOW_UNALIGNED_ACCESS): Delete.
16780         * target.def (slow_unaligned_access): New hook.
16781         * targhooks.h (default_slow_unaligned_access): Declare.
16782         * targhooks.c (default_slow_unaligned_access): New function.
16783         * doc/tm.texi.in (SLOW_UNALIGNED_ACCESS): Replace with...
16784         (TARGET_SLOW_UNALIGNED_ACCESS): ...this.
16785         * doc/tm.texi: Regenerate.
16786         * config/alpha/alpha.h (SLOW_UNALIGNED_ACCESS): Delete.
16787         * config/arm/arm.h (SLOW_UNALIGNED_ACCESS): Delete.
16788         * config/i386/i386.h (SLOW_UNALIGNED_ACCESS): Delete commented-out
16789         definition.
16790         * config/powerpcspe/powerpcspe.h (SLOW_UNALIGNED_ACCESS): Delete.
16791         * config/powerpcspe/powerpcspe.c (TARGET_SLOW_UNALIGNED_ACCESS):
16792         Redefine.
16793         (rs6000_slow_unaligned_access): New function.
16794         (rs6000_emit_move): Use it instead of SLOW_UNALIGNED_ACCESS.
16795         (expand_block_compare): Likewise.
16796         (expand_strn_compare): Likewise.
16797         (rs6000_rtx_costs): Likewise.
16798         * config/riscv/riscv.h (SLOW_UNALIGNED_ACCESS): Delete.
16799         (riscv_slow_unaligned_access): Likewise.
16800         * config/riscv/riscv.c (riscv_slow_unaligned_access): Rename to...
16801         (riscv_slow_unaligned_access_p): ...this and make static.
16802         (riscv_option_override): Update accordingly.
16803         (riscv_slow_unaligned_access): New function.
16804         (TARGET_SLOW_UNALIGNED_ACCESS): Redefine.
16805         * config/rs6000/rs6000.h (SLOW_UNALIGNED_ACCESS): Delete.
16806         * config/rs6000/rs6000.c (TARGET_SLOW_UNALIGNED_ACCESS): Redefine.
16807         (rs6000_slow_unaligned_access): New function.
16808         (rs6000_emit_move): Use it instead of SLOW_UNALIGNED_ACCESS.
16809         (rs6000_rtx_costs): Likewise.
16810         * config/rs6000/rs6000-string.c (expand_block_compare)
16811         (expand_strn_compare): Use targetm.slow_unaligned_access instead
16812         of SLOW_UNALIGNED_ACCESS.
16813         * config/tilegx/tilegx.h (SLOW_UNALIGNED_ACCESS): Delete.
16814         * config/tilepro/tilepro.h (SLOW_UNALIGNED_ACCESS): Delete.
16815         * calls.c (expand_call): Use targetm.slow_unaligned_access instead
16816         of SLOW_UNALIGNED_ACCESS.
16817         * expmed.c (simple_mem_bitfield_p): Likewise.
16818         * expr.c (alignment_for_piecewise_move): Likewise.
16819         (emit_group_load_1): Likewise.
16820         (emit_group_store): Likewise.
16821         (copy_blkmode_from_reg): Likewise.
16822         (emit_push_insn): Likewise.
16823         (expand_assignment): Likewise.
16824         (store_field): Likewise.
16825         (expand_expr_real_1): Likewise.
16826         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
16827         * lra-constraints.c (simplify_operand_subreg): Likewise.
16828         * stor-layout.c (bit_field_mode_iterator::next_mode): Likewise.
16829         * gimple-ssa-store-merging.c: Likewise in block comment at start
16830         of file.
16831         * tree-ssa-strlen.c: Include target.h.
16832         (handle_builtin_memcmp): Use targetm.slow_unaligned_access instead
16833         of SLOW_UNALIGNED_ACCESS.
16834         * system.h (SLOW_UNALIGNED_ACCESS): Poison.
16836 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16838         PR rtl-optimization/82185
16839         * expmed.c (emit_store_flag_int): Only test tem if it has been
16840         initialized.
16842 2017-09-12  Richard Biener  <rguenther@suse.de>
16844         PR middle-end/82149
16845         * match.pd ((FTYPE) N CMP CST): Fix typo.
16847 2017-09-12  Simon Atanasyan  <simon.atanasyan@imgtec.com>
16849         * config/mips/mips.c (mips_attribute_table): Add 'short_call'
16850         attribute.
16851         (mips_near_type_p): Add 'short_call' attribute as a synonym
16852         for 'near'.
16853         * doc/extend.texi (short_call): Document new function attribute.
16855 2017-09-12  Jakub Jelinek  <jakub@redhat.com>
16857         PR target/82112
16858         * c-common.c (sync_resolve_size): Instead of c_dialect_cxx ()
16859         assertion check that in the condition.
16860         (get_atomic_generic_size): Likewise.  Before testing if parameter
16861         has pointer type, if it has array type, call for C++
16862         default_conversion to perform array-to-pointer conversion.
16864 2017-09-12  Richard Biener  <rguenther@suse.de>
16866         * tree-vect-generic.c (expand_vector_operations_1): Do nothing
16867         for operations we cannot scalarize.
16869 2017-09-12  Aldy Hernandez  <aldyh@redhat.com>
16871         * tree-ssa-threadbackward.c (fsm_find_thread_path): Make GC
16872         vectors heap vectors.  Clean up comments.
16873         Make visited_bbs a reference.
16874         (profitable_jump_thread_path): Make GC
16875         vectors heap vectors.  Clean up comments.
16876         Misc cleanups.
16877         (convert_and_register_jump_thread_path): Make GC vectors heap
16878         vectors.
16879         (check_subpath_and_update_thread_path): Same.  Clean up comments.
16880         Make visited_bbs a reference.
16881         (handle_phi): Abstract common code to to
16882         register_jump_thread_path_if_profitable.
16883         Rename VAR_BB to DEF_BB.
16884         Update comments.
16885         Make GC vectors heap vectors.
16886         Make visited_bbs a reference.
16887         (handle_assignment): Same.
16888         (register_jump_thread_path_if_profitable): New.
16889         (fsm_find_control_statement_thread_paths): Rename VAR_BB to
16890         DEF_BB.
16891         Make GC vectors heap vectors.  Clean up comments.
16892         Make visited_bbs a reference.
16893         (find_jump_threads_backwards): Make visited_bbs live in the stack.
16894         * tree-ssa-threadupdate.c (delete_jump_thread_path): Fix typo in
16895         comment.
16897 2017-09-11  Max Filippov  <jcmvbkbc@gmail.com>
16899         PR target/82181
16900         * config/xtensa/xtensa.c (xtensa_mem_offset): Check that both
16901         words of E_DImode object are reachable by xtensa_uimm8x4 access.
16903 2017-09-11  Vidya Praveen  <vidyapraveen@arm.com>
16905         Revert r251800 and r251799.
16907 2017-09-11  Martin Jambor  <mjambor@suse.cz>
16909         PR hsa/82119
16910         * hsa-gen.c (gen_hsa_phi_from_gimple_phi): Process ADDR_EXPRs in
16911         arguments in advance.
16912         * hsa-regalloc.c (naive_process_phi): New parameter predecessors,
16913         use it to find predecessor edges.
16914         (naive_outof_ssa): Collect vector of predecessors.
16916 2017-09-08  Jason Merrill  <jason@redhat.com>
16918         PR c++/70029 - ICE with ref-qualifier and -flto
16919         * langhooks.h (struct lang_hooks_for_types): Add
16920         copy_lang_qualifiers.
16921         * attribs.c (build_type_attribute_qual_variant): Use it.
16922         * langhooks-def.h (LANG_HOOKS_COPY_LANG_QUALIFIERS): Default to
16923         NULL.
16924         (LANG_HOOKS_FOR_TYPES_INITIALIZER): Use it.
16925         * tree.c (verify_type): Re-enable TYPE_CANONICAL main variant check.
16927 2017-09-08  Eric Botcazou  <ebotcazou@adacore.com>
16929         PR target/81988
16930         * config/sparc/sparc.md (mulsi3): Rename into *mulsi3_sp32.
16931         (*mulsi3_sp64): New instruction.
16932         (mulsi3): New expander.
16934 2017-09-08  Uros Bizjak  <ubizjak@gmail.com>
16936         * config/alpha/alpha.c (alpha_print_operand) <case 'S'>: Remove.
16938 2017-09-08  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
16940         * sancov.c: Include memmodel.h.
16942 2017-09-07  Eric Botcazou  <ebotcazou@adacore.com>
16944         PR target/80897
16945         * config/sparc/sparc.c (sparc_emit_set_symbolic_const64): Deal with too
16946         large offsets.
16948 2017-09-07  Carl Love  <cel@us.ibm.com>
16950         * config/rs6000/vsx.md (define_insn "*stxvl"): Add missing argument to
16951         the sldi instruction.
16953 2017-09-07  David Edelsohn  <dje.gcc@gmail.com>
16955         * sancov.c: Include tm_p.h.
16957 2017-09-07  Jakub Jelinek  <jakub@redhat.com>
16959         PR target/81979
16960         * output.h (switch_to_other_text_partition): New declaration.
16961         * varasm.c (switch_to_other_text_partition): New function.
16962         * config/rs6000/rs6000.c (uses_TOC): Return 2 if
16963         NOTE_INSN_SWITCH_TEXT_SECTIONS is seen before finding load_toc_* insn.
16964         (rs6000_elf_declare_function_name): If uses_TOC returned 2, switch
16965         to the other text partition before emitting LCL label and switch back
16966         after emitting the word after it.
16968 2017-09-07  Richard Biener  <rguenther@suse.de>
16970         * passes.def (pass_split_crit_edges): Remove instance before PRE.
16971         * tree-ssa-pre.c (pass_pre::execute): Instead manually split
16972         critical edges here, after loop init.
16973         (pass_data_pre): Remove PROP_no_crit_edges flags.
16974         * tree-ssa-sccvn.c (vn_reference_lookup_3): Use vn_valueize
16975         for valueization of call args to avoid leaking VN_TOP.
16976         (visit_use): Assert we do not visit default defs.
16977         (init_scc_vn): Use build_decl for VN_TOP to make name nicer.
16978         Use error_mark_node to more easily detect leaking VN_TOP.
16979         All default-defs are varying, not VN_TOP.  Mark them visited.
16980         (run_scc_vn): Make code match comment.
16982 2017-09-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
16984         * config/rs6000/rs6000-cpus.def (OTHER_VSX_VECTOR_MASKS): Delete
16985         OPTION_MASK_FLOAT128_KEYWORD.
16986         (POWERPC_MASKS): Likewise.
16987         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Delete
16988         support for the -mfloat128-type option, and make -mfloat128
16989         default on PowerPC Linux systems.  Define or undefine
16990         __FLOAT128__ and  __FLOAT128_HARDWARE__ for the current options.
16991         Define __float128 to be __ieee128 if IEEE 128-bit support is
16992         enabled, or undefine it.
16993         (rs6000_cpu_cpp_builtins): Delete defining __FLOAT128__ here.
16994         Delete defining __FLOAT128_TYPE__.
16995         * config/rs6000/rs6000.opt (x_TARGET_FLOAT128_TYPE): Delete the
16996         -mfloat128-type option and make -mfloat128 default on PowerPC
16997         Linux systems.
16998         (TARGET_FLOAT128_TYPE): Likewise.
16999         (-mfloat128-type): Likewise.
17000         * config/rs6000/rs6000.c (rs6000_option_override_internal):
17001         Delete the -mfloat128-type option and make -mfloat128 default on
17002         PowerPC Linux systems.  Always use __ieee128 to be the keyword for
17003         the IEEE 128-bit type, and map __float128 to __ieee128 if IEEE
17004         128-bit floating point is enabled.  Change tests from using
17005         -mfloat128-type to -mfloat128.
17006         (rs6000_mangle_type): Use the correct mangling for the __float128
17007         type even if normal long double is restricted to 64-bits.
17008         (floatn_mode): Enable the _Float128 type by default on VSX Linux
17009         systems.
17010         * config/rs6000/rs6000.h (MASK_FLOAT128_TYPE): Delete.
17011         (MASK_FLOAT128_KEYWORD): Define new shortcut macro.
17012         (RS6000BTM_FLOAT128): Define in terms of -mfloat128, not
17013         -mfloat128-type.
17014         * doc/invoke.texi (RS/6000 and PowerPC Options): Update
17015         documentation for -mfloat128.
17017 2017-09-06  Olivier Hainque  <hainque@adacore.com>
17019         * config.gcc (powerpc-wrs-vxworksspe): Now match as vxworks*spe.
17021 2017-09-06  Wish Wu  <wishwu007@gmail.com>
17022             Jakub Jelinek  <jakub@redhat.com>
17024         * asan.c (initialize_sanitizer_builtins): Add
17025         BT_FN_VOID_UINT8_UINT8, BT_FN_VOID_UINT16_UINT16,
17026         BT_FN_VOID_UINT32_UINT32, BT_FN_VOID_UINT64_UINT64,
17027         BT_FN_VOID_FLOAT_FLOAT, BT_FN_VOID_DOUBLE_DOUBLE and
17028         BT_FN_VOID_UINT64_PTR variables.
17029         * builtin-types.def (BT_FN_VOID_UINT8_UINT8): New fn type.
17030         (BT_FN_VOID_UINT16_UINT16): Likewise.
17031         (BT_FN_VOID_UINT32_UINT32): Likewise.
17032         (BT_FN_VOID_FLOAT_FLOAT): Likewise.
17033         (BT_FN_VOID_DOUBLE_DOUBLE): Likewise.
17034         (BT_FN_VOID_UINT64_PTR): Likewise.
17035         * common.opt (flag_sanitize_coverage): New variable.
17036         (fsanitize-coverage=trace-pc): Remove.
17037         (fsanitize-coverage=): Add.
17038         * flag-types.h (enum sanitize_coverage_code): New enum.
17039         * fold-const.c (fold_range_test): Disable non-short-circuit
17040         optimization if flag_sanitize_coverage.
17041         (fold_truth_andor): Likewise.
17042         * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
17043         * opts.c (COVERAGE_SANITIZER_OPT): Define.
17044         (coverage_sanitizer_opts): New array.
17045         (get_closest_sanitizer_option): Add OPTS argument, handle also
17046         OPT_fsanitize_coverage_.
17047         (parse_sanitizer_options): Adjusted to also handle
17048         OPT_fsanitize_coverage_.
17049         (common_handle_option): Add OPT_fsanitize_coverage_.
17050         * sancov.c (instrument_comparison, instrument_switch): New function.
17051         (sancov_pass): Add trace-cmp support.
17052         * sanitizer.def (BUILT_IN_SANITIZER_COV_TRACE_CMP1,
17053         BUILT_IN_SANITIZER_COV_TRACE_CMP2, BUILT_IN_SANITIZER_COV_TRACE_CMP4,
17054         BUILT_IN_SANITIZER_COV_TRACE_CMP8,
17055         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP1,
17056         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP2,
17057         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP4,
17058         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP8,
17059         BUILT_IN_SANITIZER_COV_TRACE_CMPF, BUILT_IN_SANITIZER_COV_TRACE_CMPD,
17060         BUILT_IN_SANITIZER_COV_TRACE_SWITCH): New builtins.
17061         * doc/invoke.texi: Document -fsanitize-coverage=trace-cmp.
17063 2017-09-06  Richard Earnshaw  <rearnsha@arm.com>
17065         * config/arm/parsecpu.awk (fatal): Note that we've encountered an
17066         error.  Only quit immediately if parsing is complete.
17067         (BEGIN): Initialize fatal_err and parse_done.
17068         (begin fpu, end fpu): Check number of arguments.
17069         (begin arch, end arch): Likewise.
17070         (begin cpu, end cpu): Likewise.
17071         (cname, tune for, tune flags, architecture, fpu, option): Likewise.
17072         (optalias): Likewise.
17074 2017-09-06  Richard Earnshaw  <rearnsha@arm.com>
17076         * config.gcc (arm*-*-*): Don't add arm-isa.h to tm_p_file.
17077         * config/arm/arm-isa.h: Delete.  Move definitions to ...
17078         * arm-cpus.in: ... here.  Use new feature and fgroup values.
17079         * config/arm/arm.c (arm_option_override): Use lower case for feature
17080         bit names.
17081         * config/arm/arm.h (TARGET_HARD_FLOAT): Likewise.
17082         (TARGET_VFP3, TARGET_VFP5, TARGET_FMA): Likewise.
17083         * config/arm/parsecpu.awk (END): Add new command 'isa'.
17084         (isa_pfx): Delete.
17085         (print_isa_bits_for): New function.
17086         (gen_isa): New function.
17087         (gen_comm_data): Use print_isa_bits_for.
17088         (define feature): New keyword.
17089         (define fgroup): New keyword.
17090         * config/arm/t-arm (OPTIONS_H_EXTRA): Add arm-isa.h
17091         (arm-isa.h): Add rule to generate file.
17092         * common/config/arm/arm-common.c: (arm_canon_arch_option): Use lower
17093         case for feature bit names.
17095 2017-09-06  Richard Biener  <rguenther@suse.de>
17097         * tree-ssa-pre.c (NECESSARY): Remove.
17098         (create_expression_by_pieces): Do not touch pass-local flags.
17099         (insert_into_preds_of_block): Likewise.
17100         (do_pre_regular_insertion): Likewise.
17101         (eliminate_insert): Likewise.
17102         (eliminate_dom_walker::before_dom_children): Likewise.
17103         (fini_eliminate): Do not look at inserted_exprs.
17104         (mark_operand_necessary): Remove.
17105         (remove_dead_inserted_code): Replace with simple work-list
17106         algorithm based on inserted_exprs and SSA uses.
17107         (pass_pre::execute): Re-order fini_eliminate and
17108         remove_dead_inserted_code.
17110 2017-09-06  Olivier Hainque  <hainque@adacore.com>
17112         * config/powerpcspe/vxworks.h (VXCPU_FOR_8548): Correct definition
17113         for VxWorks 7.  Adjust surrounding comments.
17115 2017-09-06  Richard Biener  <rguenther@suse.de>
17117         * gimple-ssa-strength-reduction.c
17118         (find_candidates_dom_walker::before_dom_children): Also allow
17119         pointer types.
17121 2017-09-06  Richard Biener  <rguenther@suse.de>
17123         PR tree-optimization/82108
17124         * tree-vect-stmts.c (vectorizable_load): Fix pointer adjustment
17125         for gap in the non-permutation SLP case.
17127 2017-09-06  Martin Jambor  <mjambor@suse.cz>
17129         PR tree-optimization/82078
17130         * tree-sra.c (sort_and_splice_var_accesses): Move call to
17131         add_access_to_work_queue...
17132         (build_accesses_from_assign): ...here.
17133         (propagate_all_subaccesses): Make sure racc is the group
17134         representative, if there is one.
17136 2017-09-06  Jakub Jelinek  <jakub@redhat.com>
17138         PR middle-end/82095
17139         * varasm.c (categorize_decl_for_section): Use SECCAT_TBSS for TLS vars with
17140         NULL DECL_INITIAL.
17142 2017-09-06  Richard Biener  <rguenther@suse.de>
17144         * gimple-ssa-strength-reduction.c
17145         (find_candidates_dom_walker::before_doom_children): Use a
17146         type and not a mode check.
17148 2017-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
17150         PR target/77308
17151         * config/arm/predicates.md (arm_general_adddi_operand): Create new
17152         non-vfp predicate.
17153         * config/arm/arm.md (*arm_adddi3, *arm_subdi3): Use new predicates.
17155 2017-09-05  Jeff Law  <law@redhat.com>
17157         PR tree-optimization/64910
17158         * tree-ssa-reassoc.c (reassociate_bb): Restrict last change to
17159         cases where we have 3 or more operands.
17161 2017-09-05  Jakub Jelinek  <jakub@redhat.com>
17163         PR middle-end/81768
17164         * omp-low.c (lower_omp_for): Recompute tree invariant if
17165         gimple_omp_for_initial/final is ADDR_EXPR.
17167         PR middle-end/81768
17168         * omp-expand.c (expand_omp_simd): Force second operands of COND_EXPR
17169         into gimple val before gimplification fo the COND_EXPR.
17171 2017-09-05  Aldy Hernandez  <aldyh@redhat.com>
17173         * tree-ssa-threadupdate.c (duplicate_thread_path): Remove unused
17174         REGION_COPY argument.
17175         (thread_through_all_blocks): Remove unused argument to
17176         duplicate_thread_path.
17178 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17179             Alan Hayward  <alan.hayward@arm.com>
17180             David Sherwood  <david.sherwood@arm.com>
17182         * config/aarch64/aarch64-protos.h (aarch64_gen_adjusted_ldpstp):
17183         Take a scalar_mode rather than a machine_mode.
17184         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
17185         * config/aarch64/aarch64.c (aarch64_simd_container_mode): Likewise.
17186         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
17187         (aarch64_gen_adjusted_ldpstp): Likewise.
17188         (aarch64_expand_vector_init): Use scalar_mode instead of machine_mode.
17190 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17191             Alan Hayward  <alan.hayward@arm.com>
17192             David Sherwood  <david.sherwood@arm.com>
17194         * config/aarch64/aarch64-protos.h (aarch64_is_extend_from_extract):
17195         Take a scalar_int_mode instead of a machine_mode.
17196         (aarch64_mask_and_shift_for_ubfiz_p): Likewise.
17197         (aarch64_output_scalar_simd_mov_immediate): Likewise.
17198         (aarch64_simd_scalar_immediate_valid_for_move): Likewise.
17199         (aarch64_simd_attr_length_rglist): Delete.
17200         * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Take
17201         a scalar_int_mode instead of a machine_mode.
17202         (aarch64_add_offset): Likewise.
17203         (aarch64_internal_mov_immediate): Likewise
17204         (aarch64_add_constant_internal): Likewise.
17205         (aarch64_add_constant): Likewise.
17206         (aarch64_movw_imm): Likewise.
17207         (aarch64_rtx_arith_op_extract_p): Likewise.
17208         (aarch64_mask_and_shift_for_ubfiz_p): Likewise.
17209         (aarch64_simd_scalar_immediate_valid_for_move): Likewise.
17210         Remove assert that the mode isn't a vector.
17211         (aarch64_output_scalar_simd_mov_immediate): Likewise.
17212         (aarch64_expand_mov_immediate): Update calls after above changes.
17213         (aarch64_output_casesi): Use as_a <scalar_int_mode>.
17214         (aarch64_and_bitmask_imm): Check for scalar integer modes.
17215         (aarch64_move_imm): Likewise.
17216         (aarch64_can_const_movi_rtx_p): Likewise.
17217         (aarch64_strip_extend): Likewise.
17218         (aarch64_extr_rtx_p): Likewise.
17219         (aarch64_rtx_costs): Likewise, using wode_mode as the mode of
17220         a CONST_INT when the mode parameter is VOIDmode.
17221         (aarch64_float_const_rtx_p): Use scalar_int_mode for a temporary.
17223 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17225         * machmode.h (bitwise_mode_for_mode): Return opt_mode.
17226         * stor-layout.c (bitwise_mode_for_mode): Likewise.
17227         (bitwise_type_for_mode): Update accordingly.
17229 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17231         * stor-layout.h (mode_for_size_tree): Return an opt_mode.
17232         * stor-layout.c (mode_for_size_tree): Likewise.
17233         (mode_for_array): Update accordingly.
17234         (layout_decl): Likewise.
17235         (compute_record_mode): Likewise.  Only set the mode once.
17237 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17239         * target.def (get_mask_mode): Change return type to opt_mode.
17240         Expand commentary.
17241         * doc/tm.texi: Regenerate.
17242         * targhooks.h (default_get_mask_mode): Return an opt_mode.
17243         * targhooks.c (default_get_mask_mode): Likewise.
17244         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
17245         * optabs-query.c (can_vec_mask_load_store_p): Update use of
17246         targetm.get_mask_mode.
17247         * tree.c (build_truth_vector_type): Likewise.
17249 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17251         * machmode.h (mode_for_vector): Return an opt_mode.
17252         * stor-layout.c (mode_for_vector): Likewise.
17253         (mode_for_int_vector): Update accordingly.
17254         (layout_type): Likewise.
17255         * config/i386/i386.c (emit_memmov): Likewise.
17256         (ix86_expand_set_or_movmem): Likewise.
17257         (ix86_expand_vector_init): Likewise.
17258         (ix86_get_mask_mode): Likewise.
17259         * config/powerpcspe/powerpcspe.c (rs6000_expand_vec_perm_const_1):
17260         Likewise.
17261         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Likewise.
17262         * expmed.c (extract_bit_field_1): Likewise.
17263         * expr.c (expand_expr_real_2): Likewise.
17264         * optabs-query.c (can_vec_perm_p): Likewise.
17265         (can_vec_mask_load_store_p): Likewise.
17266         * optabs.c (expand_vec_perm): Likewise.
17267         * targhooks.c (default_get_mask_mode): Likewise.
17268         * tree-vect-stmts.c (vectorizable_store): Likewise.
17269         (vectorizable_load): Likewise.
17270         (get_vectype_for_scalar_type_and_size): Likewise.
17272 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17274         * machmode.h (mode_for_int_vector): New function.
17275         * stor-layout.c (mode_for_int_vector): Likewise.
17276         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Use it.
17277         * config/powerpcspe/powerpcspe.c (rs6000_do_expand_vec_perm): Likewise.
17278         * config/rs6000/rs6000.c (rs6000_do_expand_vec_perm): Likewise.
17279         * config/s390/s390.c (s390_expand_vec_compare_cc): Likewise.
17280         (s390_expand_vcond): Likewise.
17282 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17284         * machmode.h (opt_machine_mode): New type.
17285         (opt_mode<T>): Allow construction from anything that can be
17286         converted to a T.
17287         (is_a, as_a, dyn_cast): Add overloads for opt_mode.
17288         (mode_for_size): Return an opt_machine_mode.
17289         * stor-layout.c (mode_for_size): Likewise.
17290         (mode_for_size_tree): Update call accordingly.
17291         (bitwise_mode_for_mode): Likewise.
17292         (make_fract_type): Likewise.
17293         (make_accum_type): Likewise.
17294         * caller-save.c (replace_reg_with_saved_mem): Update call
17295         accordingly.
17296         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17297         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17298         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17299         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17300         * expmed.c (extract_bit_field_1): Likewise.
17301         * reload.c (get_secondary_mem): Likewise.
17302         * varasm.c (assemble_integer): Likewise.
17303         * lower-subreg.c (simplify_subreg_concatn): Likewise.  Move
17304         early-out.
17306 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17308         * machmode.h (decimal_float_mode_for_size): New function.
17309         * real.h (REAL_VALUE_TO_TARGET_LONG_DOUBLE): Use float_mode_for_size.
17310         (REAL_VALUE_TO_TARGET_DOUBLE): Likewise.
17311         (REAL_VALUE_TO_TARGET_SINGLE): Likewise.
17312         (REAL_VALUE_TO_TARGET_DECIMAL128): Use decimal_float_mode_for_size.
17313         (REAL_VALUE_TO_TARGET_DECIMAL64): Likewise.
17314         (REAL_VALUE_TO_TARGET_DECIMAL32): Likewise.
17316 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17318         * builtins.c (expand_builtin_powi): Use int_mode_for_size.
17319         (get_builtin_sync_mode): Likewise.
17320         (expand_ifn_atomic_compare_exchange): Likewise.
17321         (expand_builtin_atomic_clear): Likewise.
17322         (expand_builtin_atomic_test_and_set): Likewise.
17323         (fold_builtin_atomic_always_lock_free): Likewise.
17324         * calls.c (compute_argument_addresses): Likewise.
17325         (emit_library_call_value_1): Likewise.
17326         (store_one_arg): Likewise.
17327         * combine.c (combine_instructions): Likewise.
17328         * config/aarch64/aarch64.c (aarch64_function_value): Likewise.
17329         * config/arm/arm.c (arm_function_value): Likewise.
17330         (aapcs_allocate_return_reg): Likewise.
17331         * config/c6x/c6x.c (c6x_expand_movmem): Likewise.
17332         * config/i386/i386.c (construct_container): Likewise.
17333         (ix86_gimplify_va_arg): Likewise.
17334         (ix86_expand_sse_cmp): Likewise.
17335         (emit_memmov): Likewise.
17336         (emit_memset): Likewise.
17337         (expand_small_movmem_or_setmem): Likewise.
17338         (ix86_expand_pextr): Likewise.
17339         (ix86_expand_pinsr): Likewise.
17340         * config/lm32/lm32.c (lm32_block_move_inline): Likewise.
17341         * config/microblaze/microblaze.c (microblaze_block_move_straight):
17342         Likewise.
17343         * config/mips/mips.c (mips_function_value_1) Likewise.
17344         (mips_block_move_straight): Likewise.
17345         (mips_expand_ins_as_unaligned_store): Likewise.
17346         * config/powerpcspe/powerpcspe.c
17347         (rs6000_darwin64_record_arg_advance_flush): Likewise.
17348         (rs6000_darwin64_record_arg_flush): Likewise.
17349         * config/rs6000/rs6000.c
17350         (rs6000_darwin64_record_arg_advance_flush): Likewise.
17351         (rs6000_darwin64_record_arg_flush): Likewise.
17352         * config/sparc/sparc.c (sparc_function_arg_1): Likewise.
17353         (sparc_function_value_1): Likewise.
17354         * config/spu/spu.c (adjust_operand): Likewise.
17355         (spu_emit_branch_or_set): Likewise.
17356         (arith_immediate_p): Likewise.
17357         * emit-rtl.c (gen_lowpart_common): Likewise.
17358         * expr.c (expand_expr_real_1): Likewise.
17359         * function.c (assign_parm_setup_block): Likewise.
17360         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Likewise.
17361         * reload1.c (alter_reg): Likewise.
17362         * stor-layout.c (mode_for_vector): Likewise.
17363         (layout_type): Likewise.
17365 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17367         * config/spu/spu.c (exp2_immediate_p): Use int_mode_for_mode.
17368         (spu_convert_move): Likewise.
17369         * lower-subreg.c (resolve_simple_move): Likewise.
17371 2017-09-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
17373         PR target/81833
17374         * config/rs6000/altivec.md (altivec_vsum2sws): Convert from a
17375         define_insn to a define_expand.
17376         (altivec_vsum2sws_direct): New define_insn.
17377         (altivec_vsumsws): Convert from a define_insn to a define_expand.
17379 2017-09-05  Wilco Dijkstra  <wdijkstr@arm.com>
17381         * config/arm/arm.c (arm_option_params_internal): Improve setting of
17382         max_insns_skipped.
17384 2017-09-05  H.J. Lu  <hongjiu.lu@intel.com>
17386         PR target/59501
17387         PR target/81624
17388         PR target/81769
17389         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Don't
17390         realign stack if stack alignment needed is less than incoming
17391         stack boundary.
17393 2017-09-05  Marek Polacek  <polacek@redhat.com>
17395         PR sanitizer/82072
17396         * convert.c (convert_to_integer_1) <case NEGATE_EXPR>: Move the ubsan
17397         check earlier.
17399 2017-09-05  Wilco Dijkstra  <wdijkstr@arm.com>
17401         * explow.c (get_dynamic_stack_size): Improve dynamic alignment.
17403 2017-09-05  Richard Biener  <rguenther@suse.de>
17405         PR tree-optimization/82084
17406         * fold-const.c (can_native_encode_string_p): Handle wide characters.
17408 2017-09-05  Richard Biener  <rguenther@suse.de>
17410         PR tree-optimization/82102
17411         * tree-ssa-pre.c (fini_eliminate): Check if lhs is NULL.
17413 2017-09-05  Martin Liska  <mliska@suse.cz>
17415         PR tree-optimization/82032
17416         * tree-cfg.c (generate_range_test): New function.
17417         * tree-cfg.h (generate_range_test): Declared here.
17418         * tree-cfgcleanup.c (convert_single_case_switch): New function.
17419         (cleanup_control_expr_graph): Use it.
17420         * tree-switch-conversion.c (try_switch_expansion): Remove
17421         assert.
17422         (emit_case_nodes): Use generate_range_test.
17424 2017-09-04  Uros Bizjak  <ubizjak@gmail.com>
17426         PR target/82098
17427         * config/i386/i386.md (*<btsc><mode>_mask): Add
17428         TARGET_USE_BT to insn constraint.
17429         (*btr<mode>_mask): Ditto.
17431 2017-09-04  Wilco Dijkstra  <wdijkstr@arm.com>
17433         * config/arm/arm.c (arm_legitimate_index_p): Add comment.
17434         (thumb2_legitimate_index_p): Use correct range for DI/DF mode.
17436 2017-09-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
17438         PR target/77308
17439         * config/arm/arm.md (*arm_adddi3, *arm_subdi3): Split early except for
17440         TARGET_NEON and TARGET_IWMMXT.
17441         (anddi3, iordi3, xordi3, one_cmpldi2): Split while expanding except for
17442         TARGET_NEON and TARGET_IWMMXT.
17443         (*one_cmpldi2_insn): Moved the body of one_cmpldi2 here.
17445 2017-09-04  Uros Bizjak  <ubizjak@gmail.com>
17447         * config/i386/i386-protos.h (ix86_tls_address_pattern_p) New prototype.
17448         (ix86_rewrite_tls_address): Ditto.
17449         * config/i386/i386.c (ix86_tls_address_pattern_p) New function.
17450         (ix86_rewrite_tls_address_1): Ditto.
17451         (ix86_rewrite_tls_address): Ditto.
17452         * config/i386/predicates.md (tls_address_pattern): New predicate.
17453         * config/i386/i386.md (TLS address splitter): New splitter.
17455 2017-09-04  Richard Biener  <rguenther@suse.de>
17457         PR tree-optimization/82084
17458         * fold-const.h (can_native_encode_string_p): Declare.
17459         * fold-const.c (can_native_encode_string_p): Factor out from ...
17460         (native_encode_string): ... here.
17461         * tree-vect-stmts.c (vectorizable_store): Call it to avoid
17462         vectorizing stores from constants we later cannot handle.
17464 2017-09-04  Marek Polacek  <polacek@redhat.com>
17466         PR c/81783
17467         * doc/invoke.texi: Update -Wtautological-compare documentation.
17469 2017-09-04  Jeff Law  <law@redhat.com>
17471         PR tree-optimization/64910
17472         * tree-ssa-reassoc.c (reassociate_bb): For bitwise binary ops,
17473         swap the first and last operand if the last is a constant.
17475 2017-09-04  Marek Polacek  <polacek@redhat.com>
17477         PR sanitizer/82072
17478         * convert.c (do_narrow): When sanitizing signed integer overflows,
17479         bail out for signed types.
17480         (convert_to_integer_1) <case NEGATE_EXPR>: Likewise.
17482 2017-09-04  Richard Biener  <rguenther@suse.de>
17484         PR tree-optimization/82060
17485         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
17486         Move devirtualization after stmt folding and before EH/AB/noreturn
17487         cleanup to get the stmt refs canonicalized.  Use a bool instead
17488         of gimple_modified_p since that doesn't work for NOPs.  Schedule
17489         NOPs generated by folding for removal.
17491 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17492             Alan Hayward  <alan.hayward@arm.com>
17493             David Sherwood  <david.sherwood@arm.com>
17495         * coretypes.h (pad_direction): New enum.
17496         * defaults.h (DEFAULT_FUNCTION_ARG_PADDING): Delete.
17497         (FUNCTION_ARG_PADDING): Likewise.
17498         * target.def (function_arg_padding): New hook.
17499         * targhooks.h (default_function_arg_padding): Declare.
17500         * targhooks.c (default_function_arg_padding): New function.
17501         * doc/tm.texi.in (FUNCTION_ARG_PADDING): Replace with...
17502         (TARGET_FUNCTION_ARG_PADDING): ...this.
17503         * doc/tm.texi: Regenerate.
17504         * calls.c (store_unaligned_arguments_into_pseudos): Use pad_direction
17505         instead of direction.
17506         (compute_argument_addresses): Likewise.
17507         (load_register_parameters): Likewise.
17508         (emit_library_call_value_1): Likewise.
17509         (store_one_arg): Use targetm.calls.function_arg_padding instead
17510         of FUNCTION_ARG_PADDING.
17511         (must_pass_in_stack_var_size_or_pad): Likewise.
17512         * expr.c (emit_group_load_1): Use pad_direction instead of direction.
17513         (emit_group_store): Likewise.
17514         (emit_single_push_insn_1): Use targetm.calls.function_arg_padding
17515         instead of FUNCTION_ARG_PADDING.
17516         (emit_push_insn): Likewise, and propagate enum change throughout
17517         function.
17518         * function.h (direction): Delete.
17519         (locate_and_pad_arg_data::where_pad): Use pad_direction instead
17520         of direction.
17521         * function.c (assign_parm_find_stack_rtl): Likewise.
17522         (assign_parm_setup_block_p): Likewise.
17523         (assign_parm_setup_block): Likewise.
17524         (gimplify_parameters): Likewise.
17525         (locate_and_pad_parm): Use targetm.calls.function_arg_padding
17526         instead of FUNCTION_ARG_PADDING, and propagate enum change throughout
17527         function.
17528         * config/aarch64/aarch64.h (FUNCTION_ARG_PADDING): Delete.
17529         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17530         * config/aarch64/aarch64-protos.h (aarch64_pad_arg_upward): Delete.
17531         * config/aarch64/aarch64.c (aarch64_pad_arg_upward): Replace with...
17532         (aarch64_function_arg_padding): ...this new function.
17533         (aarch64_gimplify_va_arg_expr): Use pad_direction instead of direction.
17534         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17535         * config/arm/arm.h (FUNCTION_ARG_PADDING): Delete.
17536         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17537         * config/arm/arm-protos.h (arm_pad_arg_upward): Delete.
17538         * config/arm/arm.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17539         (arm_pad_arg_upward): Replace with...
17540         (arm_function_arg_padding): ...this new function.
17541         * config/c6x/c6x.h (BLOCK_REG_PADDING): Use pad_direction instead
17542         of direction.
17543         * config/ia64/hpux.h (FUNCTION_ARG_PADDING): Delete.
17544         * config/ia64/ia64-protos.h (ia64_hpux_function_arg_padding): Delete.
17545         * config/ia64/ia64.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17546         (ia64_hpux_function_arg_padding): Replace with...
17547         (ia64_function_arg_padding): ...this new function.  Use pad_direction
17548         instead of direction.  Check for TARGET_HPUX.
17549         * config/iq2000/iq2000.h (FUNCTION_ARG_PADDING): Delete.
17550         * config/iq2000/iq2000.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17551         (iq2000_function_arg_padding): New function.
17552         * config/mips/mips-protos.h (mips_pad_arg_upward): Delete.
17553         * config/mips/mips.c (mips_pad_arg_upward): Replace with...
17554         (mips_function_arg_padding): ...this new function.
17555         (mips_pad_reg_upward): Update accordingly.
17556         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17557         * config/mips/mips.h (PAD_VARARGS_DOWN): Use
17558         targetm.calls.function_arg_padding.
17559         (FUNCTION_ARG_PADDING): Delete.
17560         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17561         * config/nios2/nios2.h (FUNCTION_ARG_PADDING): Delete.
17562         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17563         * config/nios2/nios2-protos.h (nios2_function_arg_padding): Delete.
17564         (nios2_block_reg_padding): Return pad_direction instead of direction.
17565         * config/nios2/nios2.c (nios2_block_reg_padding): Return pad_direction
17566         instead of direction.
17567         (nios2_function_arg_padding): Likewise.  Make static.
17568         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17569         * config/pa/pa.h (FUNCTION_ARG_PADDING): Delete.
17570         (BLOCK_REG_PADDING): Use targetm.calls.function_arg_padding.
17571         * config/pa/pa-protos.h (pa_function_arg_padding): Delete.
17572         * config/pa/pa.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17573         (pa_function_arg_padding): Make static.  Return pad_direction instead
17574         of direction.
17575         * config/powerpcspe/powerpcspe.h (FUNCTION_ARG_PADDING): Delete.
17576         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17577         * config/powerpcspe/aix.h (BLOCK_REG_PADDING): Use pad_direction
17578         instead of direction.  Use targetm.calls.function_arg_padding.
17579         * config/powerpcspe/darwin.h (BLOCK_REG_PADDING): Likewise.
17580         * config/powerpcspe/freebsd64.h (BLOCK_REG_PADDING): Likewise.
17581         * config/powerpcspe/linux64.h (BLOCK_REG_PADDING): Likewise.
17582         * config/powerpcspe/powerpcspe-protos.h (function_arg_padding): Delete.
17583         * config/powerpcspe/powerpcspe.c (TARGET_FUNCTION_ARG_PADDING):
17584         Redefine.
17585         (function_arg_padding): Rename to...
17586         (rs6000_function_arg_padding): ...this.  Make static.  Return
17587         pad_direction instead of direction.
17588         (rs6000_return_in_msb): Use rs6000_function_arg_padding.
17589         * config/rs6000/rs6000.h (FUNCTION_ARG_PADDING): Delete.
17590         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17591         * config/rs6000/aix.h (BLOCK_REG_PADDING): Use pad_direction
17592         instead of direction.  Use targetm.calls.function_arg_padding.
17593         * config/rs6000/darwin.h (BLOCK_REG_PADDING): Likewise.
17594         * config/rs6000/freebsd64.h (BLOCK_REG_PADDING): Likewise.
17595         * config/rs6000/linux64.h (BLOCK_REG_PADDING): Likewise.
17596         * config/rs6000/rs6000-protos.h (function_arg_padding): Delete.
17597         * config/rs6000/rs6000.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17598         (function_arg_padding): Rename to...
17599         (rs6000_function_arg_padding): ...this.  Make static.  Return
17600         pad_direction instead of direction.
17601         (rs6000_return_in_msb): Use rs6000_function_arg_padding.
17602         * config/s390/s390.h (FUNCTION_ARG_PADDING): Delete.
17603         * config/s390/s390.c (s390_function_arg_padding): New function.
17604         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17605         * config/sparc/sparc.h (FUNCTION_ARG_PADDING): Delete.
17606         * config/sparc/sparc-protos.h (function_arg_padding): Delete.
17607         * config/sparc/sparc.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17608         (function_arg_padding): Rename to...
17609         (sparc_function_arg_padding): ...this.  Make static.  Return
17610         pad_direction instead of direction.
17611         * config/spu/spu.h (FUNCTION_ARG_PADDING): Delete.
17612         * config/spu/spu.c (spu_function_arg_padding): New function.
17613         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17614         * system.h (FUNCTION_ARG_PADDING): Poison.
17616 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17617             Alan Hayward  <alan.hayward@arm.com>
17618             David Sherwood  <david.sherwood@arm.com>
17620         * target.def (modes_tieable_p): New hook.
17621         * doc/tm.texi (MODES_TIEABLE_P): Replace with...
17622         (TARGET_MODES_TIEABLE_P): ...this.
17623         * doc/tm.texi.in: Regenerate.
17624         * hooks.h (hook_bool_mode_mode_true): Declare.
17625         * hooks.c (hook_bool_mode_mode_true): New function.
17626         * combine.c (subst): Use targetm.modes_tieable_p instead of
17627         MODES_TIEABLE_P.
17628         * dse.c (find_shift_sequence): Likewise.
17629         * expmed.c (extract_low_bits): Likewise.
17630         * lower-subreg.c: Include target.h.
17631         (find_decomposable_subregs): Use targetm.modes_tieable_p instead of
17632         MODES_TIEABLE_P.
17633         * rtlanal.c (rtx_cost): Likewise.
17634         * config/aarch64/aarch64.h (MODES_TIEABLE_P): Delete.
17635         * config/aarch64/aarch64-protos.h (aarch64_modes_tieable_p): Delete.
17636         * config/aarch64/aarch64.c (aarch64_modes_tieable_p): Make static.
17637         (TARGET_MODES_TIEABLE_P): Redefine.
17638         * config/alpha/alpha.h (MODES_TIEABLE_P): Delete.
17639         * config/alpha/alpha.c (alpha_modes_tieable_p): New function.
17640         (TARGET_MODES_TIEABLE_P): Redefine.
17641         * config/arc/arc.h (MODES_TIEABLE_P): Delete.
17642         * config/arc/arc.c (TARGET_MODES_TIEABLE_P): Redefine.
17643         (arc_modes_tieable_p): New function.
17644         * config/arm/arm.h (MODES_TIEABLE_P): Delete.
17645         * config/arm/arm-protos.h (arm_modes_tieable_p): Delete.
17646         * config/arm/arm.c (TARGET_MODES_TIEABLE_P): Redefine.
17647         (arm_modes_tieable_p): Make static.
17648         * config/avr/avr.h (MODES_TIEABLE_P): Delete.
17649         * config/bfin/bfin.h (MODES_TIEABLE_P): Delete.
17650         * config/bfin/bfin.c (bfin_modes_tieable_p): New function.
17651         (TARGET_MODES_TIEABLE_P): Redefine.
17652         * config/c6x/c6x.h (MODES_TIEABLE_P): Delete.
17653         * config/c6x/c6x.c (c6x_modes_tieable_p): New function.
17654         (TARGET_MODES_TIEABLE_P): Redefine.
17655         * config/cr16/cr16.h (MODES_TIEABLE_P): Delete.
17656         * config/cr16/cr16.c (TARGET_MODES_TIEABLE_P): Redefine.
17657         (cr16_modes_tieable_p): New function.
17658         * config/cris/cris.h (MODES_TIEABLE_P): Delete.
17659         * config/epiphany/epiphany.h (MODES_TIEABLE_P): Delete.
17660         * config/fr30/fr30.h (MODES_TIEABLE_P): Delete.
17661         (TRULY_NOOP_TRUNCATION): Update comment.
17662         * config/frv/frv.h (MODES_TIEABLE_P): Delete.
17663         (TRULY_NOOP_TRUNCATION): Update comment.
17664         * config/frv/frv.c (TARGET_MODES_TIEABLE_P): Redefine.
17665         (frv_modes_tieable_p): New function.
17666         * config/ft32/ft32.h (MODES_TIEABLE_P): Delete.
17667         * config/h8300/h8300.h (MODES_TIEABLE_P): Delete.
17668         * config/h8300/h8300.c (h8300_modes_tieable_p): New function.
17669         (TARGET_MODES_TIEABLE_P): Redefine.
17670         * config/i386/i386.h (MODES_TIEABLE_P): Delete.
17671         * config/i386/i386-protos.h (ix86_modes_tieable_p): Delete.
17672         * config/i386/i386.c (ix86_modes_tieable_p): Make static.
17673         (TARGET_MODES_TIEABLE_P): Redefine.
17674         * config/ia64/ia64.h (MODES_TIEABLE_P): Delete.
17675         * config/ia64/ia64.c (TARGET_MODES_TIEABLE_P): Redefine.
17676         (ia64_modes_tieable_p): New function.
17677         * config/iq2000/iq2000.h (MODES_TIEABLE_P): Delete.
17678         * config/iq2000/iq2000.c (TARGET_MODES_TIEABLE_P): Redefine.
17679         (iq2000_modes_tieable_p): New function.
17680         * config/lm32/lm32.h (MODES_TIEABLE_P): Delete.
17681         * config/lm32/lm32.c (TARGET_MODES_TIEABLE_P): Redefine.
17682         (lm32_modes_tieable_p): New function.
17683         * config/m32c/m32c.h (MODES_TIEABLE_P): Delete.
17684         * config/m32c/m32c-protos.h (m32c_modes_tieable_p): Delete.
17685         * config/m32c/m32c.c (m32c_modes_tieable_p): Make static.
17686         (TARGET_MODES_TIEABLE_P): Redefine.
17687         * config/m32r/m32r.h (MODES_TIEABLE_P): Delete.
17688         * config/m32r/m32r.c (TARGET_MODES_TIEABLE_P): Redefine.
17689         (m32r_modes_tieable_p): New function.
17690         * config/m68k/m68k.h (MODES_TIEABLE_P): Delete.
17691         * config/m68k/m68k.c (TARGET_MODES_TIEABLE_P): Redefine.
17692         (m68k_modes_tieable_p): New function.
17693         * config/mcore/mcore.h (MODES_TIEABLE_P): Delete.
17694         * config/mcore/mcore.c (TARGET_MODES_TIEABLE_P): Redefine.
17695         (mcore_modes_tieable_p): New function.
17696         * config/microblaze/microblaze.h (MODES_TIEABLE_P): Delete.
17697         * config/microblaze/microblaze.c (microblaze_modes_tieable_p): New
17698         function.
17699         (TARGET_MODES_TIEABLE_P): Redefine.
17700         * config/mips/mips.h (MODES_TIEABLE_P): Delete.
17701         * config/mips/mips-protos.h (mips_modes_tieable_p): Delete.
17702         * config/mips/mips.c (mips_modes_tieable_p): Make static.
17703         (TARGET_MODES_TIEABLE_P): Redefine.
17704         * config/mmix/mmix.h (MODES_TIEABLE_P): Delete.
17705         * config/mn10300/mn10300.h (MODES_TIEABLE_P): Delete.
17706         * config/mn10300/mn10300-protos.h (mn10300_modes_tieable): Delete.
17707         * config/mn10300/mn10300.c (mn10300_modes_tieable): Rename to...
17708         (mn10300_modes_tieable_p): ...this and make static.
17709         (TARGET_MODES_TIEABLE_P): Redefine.
17710         * config/moxie/moxie.h (MODES_TIEABLE_P): Delete.
17711         * config/msp430/msp430.h (MODES_TIEABLE_P): Delete.
17712         * config/msp430/msp430-protos.h (msp430_modes_tieable_p): Delete.
17713         * config/msp430/msp430.c (TARGET_MODES_TIEABLE_P): Redefine.
17714         (msp430_modes_tieable_p): Make static.
17715         * config/nds32/nds32.h (MODES_TIEABLE_P): Delete.
17716         * config/nds32/nds32.c (nds32_modes_tieable_p): New function.
17717         (TARGET_MODES_TIEABLE_P): Redefine.
17718         * config/nios2/nios2.h (MODES_TIEABLE_P): Delete.
17719         * config/nvptx/nvptx.h (MODES_TIEABLE_P): Delete.
17720         * config/nvptx/nvptx.c (nvptx_modes_tieable_p): New function.
17721         (TARGET_MODES_TIEABLE_P): Redefine.
17722         * config/pa/pa.h (MODES_TIEABLE_P): Delete.
17723         * config/pa/pa-protos.h (pa_modes_tieable_p): Delete.
17724         * config/pa/pa.c (pa_modes_tieable_p): Make static.
17725         (TARGET_MODES_TIEABLE_P): Redefine.
17726         * config/pdp11/pdp11.h (MODES_TIEABLE_P): Delete.
17727         * config/pdp11/pdp11.c (TARGET_MODES_TIEABLE_P): Redefine.
17728         (pdp11_modes_tieable_p): New function.
17729         * config/powerpcspe/powerpcspe.h (MODES_TIEABLE_P): Delete.
17730         * config/powerpcspe/powerpcspe.c (TARGET_MODES_TIEABLE_P): Redefine.
17731         (rs6000_modes_tieable_p): New function.
17732         (rs6000_debug_reg_global): Use it instead of MODES_TIEABLE_P.
17733         * config/powerpcspe/powerpcspe.md: Update comment.
17734         * config/riscv/riscv.h (MODES_TIEABLE_P): Delete.
17735         * config/riscv/riscv.c (riscv_modes_tieable_p): New function.
17736         (TARGET_MODES_TIEABLE_P): Redefine.
17737         * config/rl78/rl78.h (MODES_TIEABLE_P): Delete.
17738         * config/rl78/rl78.c (TARGET_MODES_TIEABLE_P): Redefine.
17739         (rl78_modes_tieable_p): New function.
17740         * config/rs6000/rs6000.h (MODES_TIEABLE_P): Delete.
17741         * config/rs6000/rs6000.c (TARGET_MODES_TIEABLE_P): Redefine.
17742         (rs6000_modes_tieable_p): New function.
17743         (rs6000_debug_reg_global): Use it instead of MODES_TIEABLE_P.
17744         * config/rs6000/rs6000.md: Update comment.
17745         * config/rx/rx.h (MODES_TIEABLE_P): Delete.
17746         * config/rx/rx.c (rx_modes_tieable_p): New function.
17747         (TARGET_MODES_TIEABLE_P): Redefine.
17748         * config/s390/s390.h (MODES_TIEABLE_P): Delete.
17749         * config/s390/s390.c (s390_modes_tieable_p): New function.
17750         (TARGET_MODES_TIEABLE_P): Redefine.
17751         * config/sh/sh.h (MODES_TIEABLE_P): Delete.
17752         * config/sh/sh.c (TARGET_MODES_TIEABLE_P): Redefine.
17753         (sh_modes_tieable_p): New function.
17754         * config/sparc/sparc.h (MODES_TIEABLE_P): Delete.
17755         * config/sparc/sparc-protos.h (sparc_modes_tieable_p): Delete.
17756         * config/sparc/sparc.c (TARGET_MODES_TIEABLE_P): Redefine.
17757         (sparc_modes_tieable_p): Make static.
17758         * config/spu/spu.h (MODES_TIEABLE_P): Delete.
17759         * config/spu/spu.c (spu_modes_tieable_p): New function.
17760         (TARGET_MODES_TIEABLE_P): Redefine.
17761         * config/stormy16/stormy16.h (MODES_TIEABLE_P): Delete.
17762         * config/stormy16/stormy16.c (xstormy16_modes_tieable_p): New function.
17763         (TARGET_MODES_TIEABLE_P): Redefine.
17764         * config/tilegx/tilegx.h (MODES_TIEABLE_P): Delete.
17765         * config/tilepro/tilepro.h (MODES_TIEABLE_P): Delete.
17766         * config/v850/v850.h (MODES_TIEABLE_P): Delete.
17767         * config/v850/v850.c (v850_modes_tieable_p): New function.
17768         (TARGET_MODES_TIEABLE_P): Redefine.
17769         * config/vax/vax.h (MODES_TIEABLE_P): Delete.
17770         * config/visium/visium.h (MODES_TIEABLE_P): Delete.
17771         * config/visium/visium.c (TARGET_MODES_TIEABLE_P): Redefine.
17772         (visium_modes_tieable_p): New function.
17773         * config/xtensa/xtensa.h (MODES_TIEABLE_P): Delete.
17774         * config/xtensa/xtensa.c (TARGET_MODES_TIEABLE_P): Redefine.
17775         (xtensa_modes_tieable_p): New function.
17776         * system.h (MODES_TIEABLE_P): Poison.
17778 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17779             Alan Hayward  <alan.hayward@arm.com>
17780             David Sherwood  <david.sherwood@arm.com>
17782         * target.def (hard_regno_mode_ok): New hook.
17783         * doc/tm.texi (HARD_REGNO_MODE_OK): Replace with...
17784         (TARGET_HARD_REGNO_MODE_OK): ...this.
17785         * doc/tm.texi.in: Regenerate.
17786         * hooks.h (hook_bool_uint_mode_true): Declare.
17787         * hooks.c (hook_bool_uint_mode_true): New function.
17788         * doc/md.texi: Refer to targetm.hard_regno_mode_ok instead of
17789         HARD_REGNO_MODE_OK.
17790         * genpreds.c (write_insn_preds_c): Add an include of target.h.
17791         * alias.c (init_alias_target): Use targetm.hard_regno_mode_ok
17792         instead of HARD_REGNO_MODE_OK.
17793         * caller-save.c: Include target.h.
17794         (reg_save_code): Use targetm.hard_regno_mode_ok instead of
17795         HARD_REGNO_MODE_OK.
17796         * combine.c (can_combine_p): Likewise.
17797         (combinable_i3pat): Likewise.
17798         (can_change_dest_mode): Likewise.
17799         * expr.c (init_expr_target): Likewise.
17800         (convert_move): Likewise.
17801         (convert_modes): Likewise.
17802         * ira.c (setup_prohibited_class_mode_regs): Likewise.
17803         (setup_prohibited_mode_move_regs): Likewise.
17804         * ira.h (target_ira): Likewise.
17805         * lra-assigns.c (find_hard_regno_for_1): Likewise.
17806         * lra-constraints.c (process_alt_operands): Likewise.
17807         (split_reg): Likewise.
17808         * recog.c (peep2_find_free_register): Likewise.
17809         * ree.c (combine_reaching_defs): Likewise.
17810         * regcprop.c (maybe_mode_change): Likewise.
17811         * reginfo.c (init_reg_sets_1): Likewise.
17812         (choose_hard_reg_mode): Likewise.
17813         (simplifiable_subregs): Likewise.
17814         * regrename.c (check_new_reg_p): Likewise.
17815         * reload.c (find_valid_class): Likewise.
17816         (find_valid_class_1): Likewise.
17817         (reload_inner_reg_of_subreg): Likewise.
17818         (push_reload): Likewise.
17819         (combine_reloads): Likewise.
17820         (find_dummy_reload): Likewise.
17821         (find_reloads): Likewise.
17822         * reload1.c (find_reg): Likewise.
17823         (set_reload_reg): Likewise.
17824         (allocate_reload_reg): Likewise.
17825         (choose_reload_regs): Likewise.
17826         (reload_adjust_reg_for_temp): Likewise.
17827         * rtlanal.c (subreg_size_offset_from_lsb): Likewise.
17828         (simplify_subreg_regno): Likewise.
17829         * sel-sched.c (init_regs_for_mode): Likewise.
17830         * varasm.c (make_decl_rtl): Likewise.
17831         * config/aarch64/aarch64.h (HARD_REGNO_MODE_OK): Delete.
17832         (MODES_TIEABLE_P): Use targetm.hard_regno_mode_ok instead of
17833         HARD_REGNO_MODE_OK.
17834         * config/aarch64/aarch64-protos.h (aarch64_hard_regno_mode_ok): Delete.
17835         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Make static.
17836         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17837         * config/alpha/alpha.h (HARD_REGNO_MODE_OK): Delete.
17838         * config/alpha/alpha.c (alpha_hard_regno_mode_ok): New function.
17839         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17840         * config/arc/arc.h (arc_hard_regno_mode_ok): Delete.
17841         (arc_mode_class): Delete.
17842         (HARD_REGNO_MODE_OK): Delete.
17843         * config/arc/arc.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17844         (arc_hard_regno_mode_ok): Rename old array to...
17845         (arc_hard_regno_mode_ok_modes): ...this.
17846         (arc_conditional_register_usage): Update accordingly.
17847         (arc_mode_class): Make static.
17848         (arc_hard_regno_mode_ok): New function.
17849         * config/arm/arm.h (HARD_REGNO_MODE_OK): Delete.
17850         * config/arm/arm-protos.h (arm_hard_regno_mode_ok): Delete.
17851         * config/arm/arm.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17852         (arm_hard_regno_mode_ok): Make static.
17853         * config/arm/arm.md (movdi): Use targetm.hard_regno_mode_ok instead of
17854         HARD_REGNO_MODE_OK.
17855         * config/avr/avr-protos.h (avr_hard_regno_mode_ok): Delete.
17856         * config/avr/avr.h (HARD_REGNO_MODE_OK): Delete.
17857         * config/avr/avr.c (avr_hard_regno_mode_ok): Make static and
17858         return a bool.
17859         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17860         * config/bfin/bfin-protos.h (hard_regno_mode_ok): Delete.
17861         * config/bfin/bfin.h (HARD_REGNO_MODE_OK): Delete.
17862         * config/bfin/bfin.c (hard_regno_mode_ok): Rename to...
17863         (bfin_hard_regno_mode_ok): ...this.  Make static and return a bool.
17864         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17865         * config/bfin/predicates.md (valid_reg_operand): Use
17866         targetm.hard_regno_mode_ok instead of HARD_REGNO_MODE_OK.
17867         * config/c6x/c6x.h (HARD_REGNO_MODE_OK): Delete.
17868         * config/c6x/c6x.c (c6x_hard_regno_mode_ok): New function.
17869         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17870         * config/cr16/cr16.h (HARD_REGNO_MODE_OK): Delete.
17871         * config/cr16/cr16-protos.h (cr16_hard_regno_mode_ok): Delete.
17872         * config/cr16/cr16.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17873         (cr16_hard_regno_mode_ok): Make static and return a bool.
17874         * config/cris/cris.h (HARD_REGNO_MODE_OK): Delete.
17875         * config/cris/cris.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17876         (cris_hard_regno_mode_ok): New function.
17877         * config/epiphany/epiphany.h (epiphany_hard_regno_mode_ok): Delete.
17878         (epiphany_mode_class): Delete.
17879         (HARD_REGNO_MODE_OK): Delete.
17880         * config/epiphany/epiphany-protos.h (hard_regno_mode_ok): Delete.
17881         * config/epiphany/epiphany.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17882         (hard_regno_mode_ok): Rename to...
17883         (epiphany_hard_regno_mode_ok): ...this.  Make static and return a bool.
17884         * config/fr30/fr30.h (HARD_REGNO_MODE_OK): Delete.
17885         * config/fr30/fr30.md: Refer to targetm.hard_regno_mode_ok instead of
17886         HARD_REGNO_MODE_OK.
17887         * config/frv/frv.h (HARD_REGNO_MODE_OK): Delete.
17888         * config/frv/frv-protos.h (frv_hard_regno_mode_ok): Delete.
17889         * config/frv/frv.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17890         (frv_hard_regno_mode_ok): Make static and return a bool.
17891         * config/frv/frv.md: Refer to targetm.hard_regno_mode_ok instead of
17892         HARD_REGNO_MODE_OK.
17893         * config/ft32/ft32.h (HARD_REGNO_MODE_OK): Delete.
17894         * config/h8300/h8300.h (HARD_REGNO_MODE_OK): Delete.
17895         * config/h8300/h8300-protos.h (h8300_hard_regno_mode_ok): Delete.
17896         * config/h8300/h8300.c (h8300_hard_regno_mode_ok): Make static
17897         and return a bool.
17898         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17899         * config/i386/i386.h (HARD_REGNO_MODE_OK): Delete.
17900         * config/i386/i386-protos.h (ix86_hard_regno_mode_ok): Delete.
17901         * config/i386/i386.c (ix86_hard_regno_mode_ok): Make static and
17902         return a bool.
17903         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17904         * config/ia64/ia64.h (HARD_REGNO_MODE_OK): Delete.
17905         * config/ia64/ia64.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17906         (ia64_hard_regno_mode_ok): New function.
17907         * config/iq2000/iq2000.h (HARD_REGNO_MODE_OK): Delete.
17908         * config/iq2000/iq2000.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17909         (iq2000_hard_regno_mode_ok): New function.
17910         * config/lm32/lm32.h (HARD_REGNO_MODE_OK): Delete.
17911         * config/lm32/lm32.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17912         (lm32_hard_regno_mode_ok): New function.
17913         * config/m32c/m32c.h (HARD_REGNO_MODE_OK): Delete.
17914         * config/m32c/m32c-protos.h (m32c_hard_regno_ok): Delete.
17915         * config/m32c/m32c.c (class_can_hold_mode): Use m32c_hard_regno_mode_ok
17916         instead of HARD_REGNO_MODE_OK.
17917         (m32c_hard_regno_ok): Rename to...
17918         (m32c_hard_regno_mode_ok): ...this.  Make static and return a bool.
17919         (m32c_cannot_change_mode_class): Update accordingly.
17920         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17921         * config/m32r/m32r.h (m32r_hard_regno_mode_ok): Delete.
17922         (m32r_mode_class): Delete.
17923         (HARD_REGNO_MODE_OK): Delete.
17924         * config/m32r/m32r.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17925         (m32r_hard_regno_mode_ok): Rename to...
17926         (m32r_hard_regno_modes): ...this.
17927         (m32r_mode_class): Make static.
17928         (m32r_hard_regno_mode_ok): New function.
17929         * config/m68k/m68k.h (HARD_REGNO_MODE_OK): Delete.
17930         * config/m68k/m68k-protos.h (m68k_regno_mode_ok): Delete.
17931         * config/m68k/m68k.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17932         (m68k_hard_regno_mode_ok): Make static.
17933         * config/mcore/mcore.h (HARD_REGNO_MODE_OK): Delete.
17934         * config/mcore/mcore.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17935         (mcore_hard_regno_mode_ok): New function.
17936         * config/microblaze/microblaze.h (microblaze_hard_regno_mode_ok)
17937         (HARD_REGNO_MODE_OK): Delete.
17938         * config/microblaze/microblaze.c (microblaze_hard_regno_mode_ok):
17939         Rename to...
17940         (microblaze_hard_regno_mode_ok_p): ...this and make static.
17941         (microblaze_hard_regno_mode_ok): New function.
17942         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17943         * config/mips/mips.h (HARD_REGNO_MODE_OK): Delete.
17944         (mips_hard_regno_mode_ok): Delete.
17945         * config/mips/mips.c (mips_hard_regno_mode_ok): Rename to...
17946         (mips_hard_regno_mode_ok_p): ...this and make static.
17947         (mips_hard_regno_mode_ok_p): Rename to...
17948         (mips_hard_regno_mode_ok_uncached): ...this.
17949         (mips_hard_regno_mode_ok): New function.
17950         (mips_class_max_nregs): Use mips_hard_regno_mode_ok instead
17951         of HARD_REGNO_MODE_OK.
17952         (mips_option_override): Update after above name changes.
17953         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17954         * config/mmix/mmix.h (HARD_REGNO_MODE_OK): Delete.
17955         * config/mn10300/mn10300.h (HARD_REGNO_MODE_OK): Delete.
17956         * config/mn10300/mn10300-protos.h (mn10300_hard_regno_mode_ok): Delete.
17957         * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Make static.
17958         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17959         * config/moxie/moxie.h (HARD_REGNO_MODE_OK): Delete.
17960         * config/msp430/msp430.h (HARD_REGNO_MODE_OK): Delete.
17961         * config/msp430/msp430-protos.h (msp430_hard_regno_mode_ok): Delete.
17962         * config/msp430/msp430.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17963         (msp430_hard_regno_mode_ok): Make static and return a bool.
17964         * config/nds32/nds32.h (HARD_REGNO_MODE_OK): Delete.
17965         * config/nds32/nds32-protos.h (nds32_hard_regno_mode_ok): Delete.
17966         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Make static
17967         and return a bool.
17968         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17969         * config/nios2/nios2.h (HARD_REGNO_MODE_OK): Delete.
17970         * config/nvptx/nvptx.h (HARD_REGNO_MODE_OK): Delete.
17971         * config/pa/pa.h (MODES_TIEABLE_P): Update commentary.
17972         * config/pa/pa32-regs.h (HARD_REGNO_MODE_OK): Rename to...
17973         (PA_HARD_REGNO_MODE_OK): ...this
17974         * config/pa/pa64-regs.h (HARD_REGNO_MODE_OK): Rename to...
17975         (PA_HARD_REGNO_MODE_OK): ...this.
17976         * config/pa/pa.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17977         (pa_hard_regno_mode_ok): New function.
17978         * config/pdp11/pdp11.h (HARD_REGNO_MODE_OK): Delete.
17979         * config/pdp11/pdp11.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17980         (pdp11_hard_regno_mode_ok): New function.
17981         * config/powerpcspe/powerpcspe.h (HARD_REGNO_MODE_OK): Delete.
17982         * config/powerpcspe/powerpcspe-protos.h (rs6000_hard_regno_mode_ok_p):
17983         Delete.
17984         * config/powerpcspe/powerpcspe.c (rs6000_hard_regno_mode_ok_p):
17985         Make static.
17986         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17987         (rs6000_hard_regno_mode_ok): Rename to...
17988         (rs6000_hard_regno_mode_ok_uncached): ...this.
17989         (rs6000_init_hard_regno_mode_ok): Update accordingly.
17990         (rs6000_hard_regno_mode_ok): New function.
17991         * config/riscv/riscv.h (HARD_REGNO_MODE_OK): Delete.
17992         * config/riscv/riscv-protos.h (riscv_hard_regno_mode_ok_p): Delete.
17993         * config/riscv/riscv.c (riscv_hard_regno_mode_ok_p): Rename to...
17994         (riscv_hard_regno_mode_ok): ...this and make static.
17995         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17996         * config/rl78/rl78.h (HARD_REGNO_MODE_OK): Delete.
17997         * config/rl78/rl78-protos.h (rl78_hard_regno_mode_ok): Delete.
17998         * config/rl78/rl78.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17999         (rl78_hard_regno_mode_ok): Make static and return bool.
18000         * config/rs6000/rs6000.h (HARD_REGNO_MODE_OK): Delete.
18001         * config/rs6000/rs6000-protos.h (rs6000_hard_regno_mode_ok_p):
18002         Delete.
18003         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok_p): Make static.
18004         (TARGET_HARD_REGNO_MODE_OK): Redefine.
18005         (rs6000_hard_regno_mode_ok): Rename to...
18006         (rs6000_hard_regno_mode_ok_uncached): ...this.
18007         (rs6000_init_hard_regno_mode_ok): Update accordingly.
18008         (rs6000_hard_regno_mode_ok): New function.
18009         * config/rx/rx.h (HARD_REGNO_MODE_OK): Delete.
18010         * config/rx/rx.c (rx_hard_regno_mode_ok): New function.
18011         (TARGET_HARD_REGNO_MODE_OK): Redefine.
18012         * config/s390/s390.h (HARD_REGNO_MODE_OK): Delete.
18013         * config/s390/s390-protos.h (s390_hard_regno_mode_ok): Delete.
18014         * config/s390/s390.c (s390_hard_regno_mode_ok): Make static.
18015         (TARGET_HARD_REGNO_MODE_OK): Redefine.
18016         * config/sh/sh.h (HARD_REGNO_MODE_OK): Delete.
18017         * config/sh/sh-protos.h (sh_hard_regno_mode_ok): Delete.
18018         * config/sh/sh.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
18019         (sh_hard_regno_mode_ok): Make static.
18020         * config/sparc/constraints.md: Refer to targetm.hard_regno_mode_ok
18021         instead of HARD_REGNO_MODE_OK.
18022         * config/sparc/sparc.h (hard_regno_mode_classes): Delete.
18023         (sparc_mode_class): Delete.
18024         (HARD_REGNO_MODE_OK): Delete.
18025         * config/sparc/sparc.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
18026         (hard_regno_mode_classes): Make static.
18027         (sparc_mode_class): Likewise.
18028         (sparc_hard_regno_mode_ok): New function.
18029         * config/spu/spu.h (HARD_REGNO_MODE_OK): Delete.
18030         * config/stormy16/stormy16.h (HARD_REGNO_MODE_OK): Delete.
18031         * config/stormy16/stormy16.c (xstormy16_hard_regno_mode_ok): New
18032         function.
18033         (TARGET_HARD_REGNO_MODE_OK): Redefine.
18034         * config/tilegx/tilegx.h (HARD_REGNO_MODE_OK): Delete.
18035         * config/tilepro/tilepro.h (HARD_REGNO_MODE_OK): Delete.
18036         * config/v850/v850.h (HARD_REGNO_MODE_OK): Delete.
18037         * config/v850/v850.c (v850_hard_regno_mode_ok): New function.
18038         (TARGET_HARD_REGNO_MODE_OK): Redefine.
18039         * config/vax/vax.h (HARD_REGNO_MODE_OK): Delete.
18040         * config/visium/visium.h (HARD_REGNO_MODE_OK): Delete.
18041         * config/visium/visium.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
18042         (visium_hard_regno_mode_ok): New function.
18043         * config/visium/visium.md: Refer to targetm.hard_regno_mode_ok
18044         instead of HARD_REGNO_MODE_OK.
18045         * config/xtensa/xtensa.h (xtensa_hard_regno_mode_ok): Delete.
18046         (HARD_REGNO_MODE_OK): Delete.
18047         * config/xtensa/xtensa.c (xtensa_hard_regno_mode_ok): Rename to...
18048         (xtensa_hard_regno_mode_ok_p): ...this and make static.
18049         (xtensa_option_override): Update accordingly.
18050         (TARGET_HARD_REGNO_MODE_OK): Redefine.
18051         (xtensa_hard_regno_mode_ok): New function.
18052         * system.h (HARD_REGNO_MODE_OK): Poison.
18054 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
18055             Alan Hayward  <alan.hayward@arm.com>
18056             David Sherwood  <david.sherwood@arm.com>
18058         * target.def (hard_regno_call_part_clobbered): New hook.
18059         * doc/tm.texi.in (HARD_REGNO_CALL_PART_CLOBBERED): Replace with...
18060         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): ...this hook.
18061         * doc/tm.texi: Regenerate.
18062         * hooks.h (hook_bool_uint_mode_false): Declare.
18063         * hooks.c (hook_bool_uint_mode_false): New function.
18064         * regs.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18065         * cselib.c (cselib_process_insn): Use
18066         targetm.hard_regno_call_part_clobbered instead of
18067         HARD_REGNO_CALL_PART_CLOBBERED.
18068         * ira-conflicts.c (ira_build_conflicts): Likewise.
18069         * ira-costs.c (ira_tune_allocno_costs): Likewise.
18070         * lra-constraints.c (need_for_call_save_p): Likewise.
18071         * lra-lives.c: Include target.h.
18072         (check_pseudos_live_through_calls): Use
18073         targetm.hard_regno_call_part_clobbered instead of
18074         HARD_REGNO_CALL_PART_CLOBBERED.
18075         * regcprop.c: Include target.h.
18076         (copyprop_hardreg_forward_1): Use
18077         targetm.hard_regno_call_part_clobbered instead of
18078         HARD_REGNO_CALL_PART_CLOBBERED.
18079         * reginfo.c (choose_hard_reg_mode): Likewise.
18080         * regrename.c (check_new_reg_p): Likewise.
18081         * reload.c (find_equiv_reg): Likewise.
18082         * reload1.c (emit_reload_insns): Likewise.
18083         * sched-deps.c (deps_analyze_insn): Likewise.
18084         * sel-sched.c (init_regs_for_mode): Likewise.
18085         (mark_unavailable_hard_regs): Likewise.
18086         * targhooks.c (default_dwarf_frame_reg_mode): Likewise.
18087         * config/aarch64/aarch64.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18088         * config/aarch64/aarch64.c (aarch64_hard_regno_call_part_clobbered):
18089         New function.
18090         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18091         * config/avr/avr.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18092         * config/avr/avr-protos.h (avr_hard_regno_call_part_clobbered):
18093         Delete.
18094         * config/avr/avr.c (avr_hard_regno_call_part_clobbered): Make static
18095         and return a bool.
18096         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18097         * config/i386/i386.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18098         * config/i386/i386.c (ix86_hard_regno_call_part_clobbered): New
18099         function.
18100         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18101         * config/mips/mips.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18102         * config/mips/mips.c (mips_hard_regno_call_part_clobbered): New
18103         function.
18104         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18105         * config/powerpcspe/powerpcspe.h (HARD_REGNO_CALL_PART_CLOBBERED):
18106         Delete.
18107         * config/powerpcspe/powerpcspe.c
18108         (rs6000_hard_regno_call_part_clobbered): New function.
18109         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18110         * config/rs6000/rs6000.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18111         * config/rs6000/rs6000.c (rs6000_hard_regno_call_part_clobbered):
18112         New function.
18113         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18114         * config/s390/s390.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18115         * config/s390/s390.c (s390_hard_regno_call_part_clobbered): New
18116         function.
18117         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18118         * config/sh/sh.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18119         * system.h (HARD_REGNO_CALL_PART_CLOBBERED): Poison.
18121 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
18122             Alan Hayward  <alan.hayward@arm.com>
18123             David Sherwood  <david.sherwood@arm.com>
18125         * rtl.h (subreg_memory_offset): Declare.
18126         * emit-rtl.c (subreg_memory_offset): New function.
18127         * expmed.c (store_bit_field_1): Use it.
18128         * expr.c (undefined_operand_subword_p): Likewise.
18129         * simplify-rtx.c (simplify_subreg): Likewise.
18131 2017-09-04  Alexander Monakov  <amonakov@ispras.ru>
18133         PR rtl-optimization/57448
18134         PR target/67458
18135         PR target/81316
18136         * optabs.c (expand_atomic_load): Place compiler memory barriers if
18137         using atomic_load pattern.
18138         (expand_atomic_store): Likewise.
18140 2017-09-04  Jakub Jelinek  <jakub@redhat.com>
18142         PR sanitizer/81981
18143         * gimple-fold.c (gimple_fold_call): Optimize away useless UBSAN_PTR
18144         and UBSAN_BOUNDS internal calls.  Clean up IFN_UBSAN_OBJECT_SIZE
18145         handling.  Use replace_call_with_value with NULL instead of
18146         gsi_replace, unlink_stmt_vdef and release_defs.
18148         * gdbhooks.py (OptMachineModePrinter.to_string): Use 8 spaces
18149         instead of tab.
18151         * lra-remat.c (reg_overlap_for_remat_p): Fix a pasto.
18153 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
18155         PR bootstrap/82045
18156         * rtl.h (emit_library_call_value_1): Declare.
18157         (emit_library_call): Replace declaration with a series of overloads.
18158         Remove the parameter count argument.
18159         (emit_library_call_value): Likewise.
18160         * calls.c (emit_library_call_value_1): Make global.  Replace varargs
18161         with an "rtx_mode_t *".
18162         (emit_library_call_value): Delete.
18163         (emit_library_call): Likewise.
18164         * asan.c (asan_emit_stack_protection): Update calls accordingly.
18165         (asan_emit_allocas_unpoison): Likewise.
18166         * builtins.c (expand_builtin_powi): Likewise.
18167         (expand_asan_emit_allocas_unpoison): Likewise.
18168         * cfgexpand.c (expand_main_function): Likewise.
18169         * config/aarch64/aarch64.c (aarch64_trampoline_init): Likewise.
18170         * config/aarch64/aarch64.h (PROFILE_HOOK): Likewise.
18171         * config/alpha/alpha.c (alpha_trampoline_init): Likewise.
18172         * config/arm/arm.c (arm_trampoline_init): Likewise.
18173         (arm_call_tls_get_addr): Likewise.
18174         (arm_expand_divmod_libfunc): Likewise.
18175         * config/bfin/bfin.md (umulsi3_highpart): Likewise.
18176         (smulsi3_highpart): Likewise.
18177         * config/c6x/c6x.c (c6x_initialize_trampoline): Likewise.
18178         (c6x_expand_compare): Likewise.
18179         (c6x_expand_movmem): Likewise.
18180         * config/frv/frv.c (frv_trampoline_init): Likewise.
18181         * config/i386/i386.c (ix86_trampoline_init): Likewise.
18182         (ix86_expand_divmod_libfunc): Likewise.
18183         * config/ia64/ia64.c (ia64_expand_tls_address): Likewise.
18184         (ia64_expand_compare): Likewise.
18185         (ia64_profile_hook): Likewise.
18186         * config/ia64/ia64.md (save_stack_nonlocal): Likewise.
18187         (nonlocal_goto): Likewise.
18188         (restore_stack_nonlocal): Likewise.
18189         * config/m32r/m32r.c (block_move_call): Likewise.
18190         (m32r_trampoline_init): Likewise.
18191         * config/m68k/linux.h (FINALIZE_TRAMPOLINE): Likewise.
18192         * config/m68k/m68k.c (m68k_call_tls_get_addr): Likewise.
18193         (m68k_call_m68k_read_tp): Likewise.
18194         * config/microblaze/microblaze.c (microblaze_call_tls_get_addr)
18195         (microblaze_expand_divide): Likewise.
18196         * config/mips/mips.h (mips_args): Likewise.
18197         * config/mips/sdemtk.h (mips_sync_icache): Likewise.
18198         (MIPS_ICACHE_SYNC): Likewise.
18199         * config/nios2/nios2.c (nios2_emit_expensive_div): Likewise.
18200         (nios2_trampoline_init): Likewise.
18201         * config/pa/pa.c (hppa_tls_call): Likewise.
18202         (pa_trampoline_init): Likewise.
18203         * config/pa/pa.md (canonicalize_funcptr_for_compare): Likewise.
18204         * config/powerpcspe/powerpcspe.c (rs6000_legitimize_tls_address)
18205         (expand_strn_compare): Likewise.
18206         (rs6000_generate_compare): Likewise.
18207         (rs6000_expand_float128_convert): Likewise.
18208         (output_profile_hook): Likewise.
18209         (rs6000_trampoline_init): Likewise.
18210         * config/powerpcspe/powerpcspe.md (neg<mode>2): Likewise.
18211         * config/riscv/riscv.h (PROFILE_HOOK): Likewise.
18212         * config/rs6000/rs6000-string.c (expand_strn_compare): Likewise.
18213         * config/rs6000/rs6000.c (rs6000_legitimize_tls_address): Likewise.
18214         (rs6000_generate_compare): Likewise.
18215         (rs6000_expand_float128_convert): Likewise.
18216         (output_profile_hook): Likewise.
18217         (rs6000_trampoline_init): Likewise.
18218         * config/rs6000/rs6000.md (neg<mode>2): Likewise.
18219         * config/sh/sh.c (sh_trampoline_init): Likewise.
18220         * config/sparc/sparc.c (emit_soft_tfmode_libcall): Likewise.
18221         (sparc_emit_float_lib_cmp): Likewise.
18222         (sparc32_initialize_trampoline): Likewise.
18223         (sparc64_initialize_trampoline): Likewise.
18224         (sparc_profile_hook): Likewise.
18225         * config/spu/spu.c (ea_load_store): Likewise.
18226         * config/spu/spu.md (floatunssidf2): Likewise.
18227         * config/tilegx/tilegx.c (tilegx_trampoline_init): Likewise.
18228         * config/tilepro/tilepro.c (tilepro_trampoline_init): Likewise.
18229         * config/visium/visium.c (expand_block_move_4): Likewise.
18230         (expand_block_move_2): Likewise.
18231         (expand_block_move_1): Likewise.
18232         (expand_block_set_4): Likewise.
18233         (expand_block_set_2): Likewise.
18234         (expand_block_set_1): Likewise.
18235         (visium_trampoline_init): Likewise.
18236         (visium_profile_hook): Likewise.
18237         * config/xtensa/xtensa.c (xtensa_expand_nonlocal_goto): Likewise.
18238         (xtensa_setup_frame_addresses): Likewise.
18239         (xtensa_trampoline_init): Likewise.
18240         * except.c (sjlj_emit_function_enter): Likewise.
18241         (sjlj_emit_function_exit): Likewise.
18242         * explow.c (allocate_dynamic_stack_space): Likewise.
18243         (probe_stack_range): Likewise.
18244         * expr.c (convert_mode_scalar): Likewise.
18245         * optabs.c (expand_binop): Likewise.
18246         (expand_twoval_binop_libfunc): Likewise.
18247         (expand_unop): Likewise.
18248         (prepare_cmp_insn): Likewise.
18249         (prepare_float_lib_cmp): Likewise.
18250         (expand_float): Likewise.
18251         (expand_fix): Likewise.
18252         (expand_fixed_convert): Likewise.
18253         (maybe_emit_sync_lock_test_and_set): Likewise.
18254         (expand_atomic_compare_and_swap): Likewise.
18255         (expand_mem_thread_fence): Likewise.
18256         (expand_atomic_fetch_op): Likewise.
18258 2017-09-03  Gerald Pfeifer  <gerald@pfeifer.com>
18260         * doc/generic.texi (OpenACC): Adjust URL.
18261         * doc/invoke.texi (C Dialect Options): Ditto.
18263 2017-09-03  Uros Bizjak  <ubizjak@gmail.com>
18265         * config/i386/i386.md (*bt<mode>): Use nonimmediate_operand
18266         predicate for operand 1.  Add (m,<S>) constraint.
18267         (*jcc_bt<mode>): Use nonimmediate_operand predicate for operand 1.
18268         Prevent memory operand 1 with register operand 2.
18270 2017-09-01  Segher Boessenkool  <segher@kernel.crashing.org>
18272         PR rtl-optimization/82024
18273         * combine.c (try_combine): If the combination result is a PARALLEL,
18274         and we only need to retain the SET in there that would be placed
18275         at I2, check that we can place that at I3 instead, before doing so.
18277 2017-09-01  Jakub Jelinek  <jakub@redhat.com>
18279         PR target/81766
18280         * config/i386/i386.c (ix86_init_large_pic_reg): Return label
18281         instead of void.
18282         (ix86_init_pic_reg): Remember label from ix86_init_large_pic_reg,
18283         if non-NULL and preceded by NOTE_INSN_BASIC_BLOCK, swap the note
18284         and label.
18286 2017-09-01  Joerg Sonnenberger  <joerg@bec.de>
18287             Jeff Law  <law@redhat.com>
18289         * varasm.c (bss_initializer_p): Do not put constants into .bss
18290         (categorize_decl_for_section): Handle bss_initializer_p returning
18291         false when DECL_INITIAL is NULL.
18293 2017-09-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
18295         PR target/82012
18296         * config/s390/s390.c (s390_can_inline_p): New function.
18298 2017-09-01  Jeff Law  <law@redhat.com>
18300         PR tree-optimization/82052
18301         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
18302         Always initialize the returned slot after a hash table miss
18303         when INSERT is true.
18305 2017-09-01  Alexander Monakov  <amonakov@ispras.ru>
18307         * config/s390/s390.md (mem_signal_fence): Remove.
18308         * doc/md.texi (mem_signal_fence): Remove.
18309         * optabs.c (expand_mem_signal_fence): Remove uses of mem_signal_fence.
18310         Update comments.
18311         * target-insns.def (mem_signal_fence): Remove.
18313 2017-09-01  Jakub Jelinek  <jakub@redhat.com>
18315         PR sanitizer/81902
18316         * doc/invoke.texi: Document -fsanitize=pointer-overflow.
18318         PR sanitizer/81923
18319         * asan.c (create_odr_indicator): Strip name encoding from assembler
18320         name before appending it after __odr_asan_.
18322 2017-09-01  Martin Liska  <mliska@suse.cz>
18324         PR tree-optimization/82059
18325         * gimple-ssa-isolate-paths.c (isolate_path): Add profile and
18326         frequency only when an edge is redirected.
18328 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18330         * config/arc/arc-c.c (__ARC_LPC_WIDTH__): Add builtin define.
18331         * config/arc/arc.c (ARC_MAX_LOOP_LENGTH): Define.
18332         (arc_conditional_register_usage): Remove ARC600 lp_count
18333         exception.
18334         (arc_file_start): Emit Tag_ARC_CPU_variation.
18335         (arc_can_use_doloop_p): New conditions to use ZOLs.
18336         (hwloop_fail): New function.
18337         (hwloop_optimize): Likewise.
18338         (hwloop_pattern_reg): Likewise.
18339         (arc_doloop_hooks): New struct, to be used with reorg_loops.
18340         (arc_reorg_loops): New function, calls reorg_loops.
18341         (arc_reorg): Call arc_reorg_loops.  Remove old ZOL handling.
18342         (arc600_corereg_hazard): Remove ZOL checking, case handled by
18343         hwloop_optimize.
18344         (arc_loop_hazard): Remove function, functionality moved into
18345         hwloop_optimize.
18346         (arc_hazard): Remove arc_loop_hazard call.
18347         (arc_adjust_insn_length): Remove ZOL handling, functionality moved
18348         into hwloop_optimize.
18349         (arc_label_align): Remove ZOL handling.
18350         * config/arc/arc.h (LOOP_ALIGN): Changed to 0.
18351         * config/arc/arc.md (doloop_begin): Remove pattern.
18352         (doloop_begin_i): Likewise.
18353         (doloop_end_i): Likewise.
18354         (doloop_fallback): Likewise.
18355         (doloop_fallback_m): Likewise.
18356         (doloop_end): Reimplement expand.
18357         (arc_lp): New pattern for LP instruction.
18358         (loop_end): New pattern.
18359         (loop_fail): Likewise.
18360         (decrement_and_branch_until_zero): Likewise.
18361         * config/arc/arc.opt (mlpc-width): New option.
18362         * doc/invoke.texi (mlpc-width): Document option.
18364 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18366         * config/arc/arc.c (arc_ifcvt): Remove use of merge_blocks call.
18367         (arc_ccfsm_advance): Fix checking for delay slots.
18368         (arc_reorg): Add rtl dump after each call to arc_ifcvt.
18370 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18372         * config/arc/arc.md (movqi_insn): Add stores to save constant long
18373         immediates.
18374         (movhi_insn): Update store instruction constraint which are saving
18375         6-bit short immediates.
18376         (movsi_insn): Consider also short scaled load operations.
18377         (zero_extendhisi2_i): Use Usd constraint instead of T.
18378         (extendhisi2_i): Add q constraint.
18379         (arc_clzsi2): Add type and length attributes.
18380         (arc_ctzsi2): Likewise.
18381         * config/arc/constraints.md (Usc): Update constraint, the
18382         assembler can parse two relocations for a single instruction.
18384 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18386         * config/arc/arc.c (arc_use_anchors_for_symbol_p): New function.
18387         (TARGET_USE_ANCHORS_FOR_SYMBOL_P): Define.
18389 2017-08-31  Olivier Hainque  <hainque@adacore.com>
18391         * config.gcc (powerpc-wrs-vxworks|vxworksae|vxworksmils): Now
18392         match as powerpc-wrs-vxworks*.
18394 2017-08-31  James Greenhalgh  <james.greenhalgh@arm.com>
18396         * config/aarch64/aarch64-simd.md (aarch64_mla_elt_merge<mode>): Fix
18397         register constraint for by-element operand.
18398         (aarch64_mls_elt_merge<mode>): Likewise.
18400 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18402         * config/arc/arc.c (arc_can_follow_jump): Check for short
18403         branches.
18405 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18407         * config.gcc: Use g.opt for arc.
18408         * config/arc/arc.c (LEGITIMATE_SCALED_ADDRESS_P): Deleted,
18409         functionality moved to ...
18410         (legitimate_scaled_address_p): New function, ...here.
18411         (LEGITIMATE_SMALL_DATA_OFFSET_P): New define.
18412         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Use the above define.
18413         (legitimate_offset_address_p): Delete TARGET_NO_SDATA_SET
18414         condition.
18415         (arc_override_options): Handle G option.
18416         (arc_output_pic_addr_const): Correct function definition.
18417         (arc_legitimate_address_p): Use legitimate_scaled_address_p.
18418         (arc_decl_anon_ns_mem_p): Delete.
18419         (arc_in_small_data_p): Overhaul this function to take into
18420         consideration the value given via G option.
18421         (arc_rewrite_small_data_1): Renamed and corrected old
18422         arc_rewrite_small_data function.
18423         (arc_rewrite_small_data): New function.
18424         (small_data_pattern): Don't use pic_offset_table_rtx.
18425         * config/arc/arc.h (CC1_SPEC): Recognize G option.
18426         * config/arc/simdext.md (movmisalignv2hi): Use
18427         prepare_move_operands function.
18428         (mov*): Likewise.
18429         (movmisalign*): Likewise.
18430         * doc/invoke.texi (ARC options): Document -G option.
18432 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18434         * config/arc/arc-protos.h (compact_sda_memory_operand): Update
18435         prototype.
18436         * config/arc/arc.c (arc_print_operand): Output scalled address for
18437         sdata whenever is possible.
18438         (arc_in_small_data_p): Allow sdata for 64bit datum when double
18439         load/stores are available.
18440         (compact_sda_memory_operand): Check for the alignment required by
18441         code density instructions.
18442         * config/arc/arc.md (movsi_insn): Use newly introduced Us0
18443         constraint.
18444         * config/arc/constraints.md (Usd): Update constraint.
18445         (Us0): New constraint.
18446         (Usc): Update constraint.
18448 2017-08-31  Richard Biener  <rguenther@suse.de>
18450         PR middle-end/82054
18451         * dwarf2out.c (dwarf2out_early_global_decl): Process each
18452         function only once.
18454 2017-08-31  Tamar Christina  <tamar.christina@arm.com>
18456         * config/aarch64/aarch64-builtins.c (aarch64_init_simd_builtins):
18457         Resize type_signature.
18459 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18460             Alan Hayward  <alan.hayward@arm.com>
18461             David Sherwood  <david.sherwood@arm.com>
18463         * config/aarch64/aarch64.c (aarch64_base_register_rtx_p): Only allow
18464         subregs whose inner modes can be stored in GPRs.
18465         (aarch64_classify_index): Likewise.
18467 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18468             Alan Hayward  <alan.hayward@arm.com>
18469             David Sherwood  <david.sherwood@arm.com>
18471         * config/aarch64/iterators.md (V_cmp_result): Rename to...
18472         (V_INT_EQUIV): ...this.
18473         (v_cmp_result): Rename to...
18474         (v_int_equiv): ...this.
18475         * config/aarch64/aarch64.md (xorsign<mode>3): Update accordingly.
18476         * config/aarch64/aarch64-simd.md (xorsign<mode>3): Likewise.
18477         (copysign<mode>3): Likewise.
18478         (aarch64_simd_bsl<mode>_internal): Likewise.
18479         (aarch64_simd_bsl<mode>): Likewise.
18480         (vec_cmp<mode><mode>): Likewise.
18481         (vcond<mode><mode>): Likewise.
18482         (vcond<v_cmp_mixed><mode>): Likewise.
18483         (vcondu<mode><v_cmp_mixed>): Likewise.
18484         (aarch64_cm<optab><mode>): Likewise.
18485         (aarch64_cmtst<mode>): Likewise.
18486         (aarch64_fac<optab><mode>): Likewise.
18487         (vec_perm_const<mode>): Likewise.
18488         (vcond_mask_<mode><v_cmp_result>): Rename to...
18489         (vcond_mask_<mode><v_int_equiv>): ...this.
18490         (vec_cmp<mode><v_cmp_result>): Rename to...
18491         (vec_cmp<mode><v_int_equiv>): ...this.
18493 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18494             Alan Hayward  <alan.hayward@arm.com>
18495             David Sherwood  <david.sherwood@arm.com>
18497         * config/aarch64/aarch64-modes.def: Remove 32-, 48- and 64-byte
18498         vector modes.
18499         * config/aarch64/iterators.md (VRL2, VRL3, VRL4): Delete.
18500         * config/aarch64/aarch64.md (UNSPEC_LD2_DREG, UNSPEC_LD3_DREG)
18501         (UNSPEC_LD4_DREG): New unspecs.
18502         * config/aarch64/aarch64-simd.md (aarch64_ld2<mode>_dreg_le)
18503         (aarch64_ld2<mode>_dreg_be): Replace with...
18504         (aarch64_ld2<mode>_dreg): ...this pattern and use the new DREG
18505         unspec.
18506         (aarch64_ld3<mode>_dreg_le)
18507         (aarch64_ld3<mode>_dreg_be): Replace with...
18508         (aarch64_ld3<mode>_dreg): ...this pattern and use the new DREG
18509         unspec.
18510         (aarch64_ld4<mode>_dreg_le)
18511         (aarch64_ld4<mode>_dreg_be): Replace with...
18512         (aarch64_ld4<mode>_dreg): ...this pattern and use the new DREG
18513         unspec.
18515 2017-08-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
18517         PR tree-optimization/81987
18518         * gimple-ssa-strength-reduction.c (insert_initializers): Don't
18519         insert an initializer in a location not dominated by the stride
18520         definition.
18522 2017-08-30  Eric Botcazou  <ebotcazou@adacore.com>
18524         * tree-eh.c (lower_try_finally_switch): Set the location of the finally
18525         on the entire header of the finally block in the fallthru case.
18527 2017-08-30  Eric Botcazou  <ebotcazou@adacore.com>
18529         * varasm.c (decode_addr_const): Deal with INDIRECT_REF <INTEGER_CST>.
18531 2017-08-30  Pat Haugen  <pthaugen@us.ibm.com>
18533         * config/rs6000/rs6000.c (rs6000_emit_prologue_move_from_cr): Rename from
18534         rs6000_emit_move_from_cr and call renamed function.
18535         (rs6000_emit_prologue): Call renamed functions.
18536         * config/rs6000/rs6000.md (prologue_movesi_from_cr): Rename from
18537         movesi_from_cr, remove volatile CRs.
18539 2017-08-30  Jon Beniston  <jon@beniston.com>
18540             Richard Biener  <rguenther@suse.de>
18542         * tree-vect-patterns.c (vect_pattern_recog_1): Use VECTOR_TYPE_P instead
18543         of VECTOR_MODE_P check.
18544         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Allow single
18545         element vector types.
18547 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18549         * df.h (df_read_modify_subreg_p): Remove in favor of...
18550         * rtl.h (read_modify_subreg_p): ...this new function.  Take a
18551         const_rtx instead of an rtx.
18552         * cprop.c (local_cprop_find_used_regs): Update accordingly.
18553         * df-problems.c (df_word_lr_mark_ref): Likewise.
18554         * ira-lives.c (mark_pseudo_reg_live): Likewise.
18555         (mark_pseudo_reg_dead): Likewise.
18556         (mark_ref_dead): Likewise.
18557         * reginfo.c (init_subregs_of_mode): Likewise.
18558         * sched-deps.c (sched_analyze_1): Likewise.
18559         * df-scan.c (df_def_record_1): Likewise.
18560         (df_uses_record): Likewise.
18561         (df_read_modify_subreg_p): Remove in favor of...
18562         * rtlanal.c (read_modify_subreg_p): ...this new function.  Take a
18563         const_rtx instead of an rtx.
18565 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18566             Alan Hayward  <alan.hayward@arm.com>
18567             David Sherwood  <david.sherwood@arm.com>
18569         * rtl.h (partial_subreg_p): New function.
18570         * caller-save.c (save_call_clobbered_regs): Use it.
18571         * calls.c (expand_call): Likewise.
18572         * combine.c (combinable_i3pat): Likewise.
18573         (simplify_set): Likewise.
18574         (make_extraction): Likewise.
18575         (make_compound_operation_int): Likewise.
18576         (gen_lowpart_or_truncate): Likewise.
18577         (force_to_mode): Likewise.
18578         (make_field_assignment): Likewise.
18579         (reg_truncated_to_mode): Likewise.
18580         (record_truncated_value): Likewise.
18581         (move_deaths): Likewise.
18582         * cse.c (record_jump_cond): Likewise.
18583         (cse_insn): Likewise.
18584         * cselib.c (cselib_lookup_1): Likewise.
18585         * expmed.c (extract_bit_field_using_extv): Likewise.
18586         * function.c (assign_parm_setup_reg): Likewise.
18587         * ifcvt.c (noce_convert_multiple_sets): Likewise.
18588         * ira-build.c (create_insn_allocnos): Likewise.
18589         * lra-coalesce.c (merge_pseudos): Likewise.
18590         * lra-constraints.c (match_reload): Likewise.
18591         (simplify_operand_subreg): Likewise.
18592         (curr_insn_transform): Likewise.
18593         * lra-lives.c (process_bb_lives): Likewise.
18594         * lra.c (new_insn_reg): Likewise.
18595         (lra_substitute_pseudo): Likewise.
18596         * regcprop.c (mode_change_ok): Likewise.
18597         (maybe_mode_change): Likewise.
18598         (copyprop_hardreg_forward_1): Likewise.
18599         * reload.c (push_reload): Likewise.
18600         (find_reloads): Likewise.
18601         (find_reloads_subreg_address): Likewise.
18602         * reload1.c (alter_reg): Likewise.
18603         (eliminate_regs_1): Likewise.
18604         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
18606 2017-08-30  David Edelsohn  <dje.gcc@gmail.com>
18608         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Revert
18609         back to if statements, including unpack.
18611 2017-08-30  Martin Liska  <mliska@suse.cz>
18613         PR inline-asm/82001
18614         * ipa-icf-gimple.c (func_checker::compare_tree_list_operand):
18615         Rename to ...
18616         (func_checker::compare_asm_inputs_outputs): ... this function.
18617         (func_checker::compare_gimple_asm): Use the function to compare
18618         also ASM constrains.
18619         * ipa-icf-gimple.h: Rename the function.
18621 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18622             Alan Hayward  <alan.hayward@arm.com>
18623             David Sherwood  <david.sherwood@arm.com>
18625         * coretypes.h (complex_mode): New type.
18626         * gdbhooks.py (build_pretty_printer): Handle it.
18627         * machmode.h (complex_mode): New class.
18628         (complex_mode::includes_p): New function.
18629         (is_complex_int_mode): Likewise.
18630         (is_complex_float_mode): Likewise.
18631         * genmodes.c (get_mode_class): Handle complex mode classes.
18632         * function.c (expand_function_end): Use is_complex_int_mode.
18634 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18635             Alan Hayward  <alan.hayward@arm.com>
18636             David Sherwood  <david.sherwood@arm.com>
18638         * coretypes.h (scalar_mode_pod): New typedef.
18639         * gdbhooks.py (build_pretty_printer): Handle it.
18640         * machmode.h (gt_ggc_mx, gt_pch_nx): New functions.
18641         * fixed-value.h (fixed_value::mode): Change type to scalar_mode_pod.
18642         * fold-const.c (fold_convert_const_int_from_fixed): Use scalar_mode.
18643         * tree-streamer-in.c (unpack_ts_fixed_cst_value_fields): Use
18644         as_a <scalar_mode>.
18646 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18647             Alan Hayward  <alan.hayward@arm.com>
18648             David Sherwood  <david.sherwood@arm.com>
18650         * machmode.h (mode_for_vector): Take a scalar_mode instead
18651         of a machine_mode.
18652         * stor-layout.c (mode_for_vector): Likewise.
18653         * explow.c (promote_mode): Use as_a <scalar_mode>.
18654         * sdbout.c (sdbout_parms): Use is_a <scalar_mode>.
18656 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18657             Alan Hayward  <alan.hayward@arm.com>
18658             David Sherwood  <david.sherwood@arm.com>
18660         * target.def (preferred_simd_mode): Take a scalar_mode
18661         instead of a machine_mode.
18662         * targhooks.h (default_preferred_simd_mode): Likewise.
18663         * targhooks.c (default_preferred_simd_mode): Likewise.
18664         * config/arc/arc.c (arc_preferred_simd_mode): Likewise.
18665         * config/arm/arm.c (arm_preferred_simd_mode): Likewise.
18666         * config/c6x/c6x.c (c6x_preferred_simd_mode): Likewise.
18667         * config/epiphany/epiphany.c (epiphany_preferred_simd_mode): Likewise.
18668         * config/i386/i386.c (ix86_preferred_simd_mode): Likewise.
18669         * config/mips/mips.c (mips_preferred_simd_mode): Likewise.
18670         * config/nvptx/nvptx.c (nvptx_preferred_simd_mode): Likewise.
18671         * config/powerpcspe/powerpcspe.c (rs6000_preferred_simd_mode):
18672         Likewise.
18673         * config/rs6000/rs6000.c (rs6000_preferred_simd_mode): Likewise.
18674         * config/s390/s390.c (s390_preferred_simd_mode): Likewise.
18675         * config/sparc/sparc.c (sparc_preferred_simd_mode): Likewise.
18676         * config/aarch64/aarch64.c (aarch64_preferred_simd_mode): Likewise.
18677         (aarch64_simd_scalar_immediate_valid_for_move): Update accordingly.
18678         * doc/tm.texi: Regenerate.
18679         * optabs-query.c (can_vec_mask_load_store_p): Return false for
18680         non-scalar modes.
18682 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18683             Alan Hayward  <alan.hayward@arm.com>
18684             David Sherwood  <david.sherwood@arm.com>
18686         * target.def (scalar_mode_supported_p): Take a scalar_mode
18687         instead of a machine_mode.
18688         * targhooks.h (default_scalar_mode_supported_p): Likewise.
18689         * targhooks.c (default_scalar_mode_supported_p): Likewise.
18690         * config/aarch64/aarch64.c (aarch64_scalar_mode_supported_p): Likewise.
18691         * config/alpha/alpha.c (alpha_scalar_mode_supported_p): Likewise.
18692         * config/arm/arm.c (arm_scalar_mode_supported_p): Likewise.
18693         * config/avr/avr.c (avr_scalar_mode_supported_p): Likewise.
18694         * config/c6x/c6x.c (c6x_scalar_mode_supported_p): Likewise.
18695         * config/i386/i386.c (ix86_scalar_mode_supported_p): Likewise.
18696         * config/ia64/ia64.c (ia64_scalar_mode_supported_p): Likewise.
18697         * config/mips/mips.c (mips_scalar_mode_supported_p): Likewise.
18698         * config/msp430/msp430.c (msp430_scalar_mode_supported_p): Likewise.
18699         * config/pa/pa.c (pa_scalar_mode_supported_p): Likewise.
18700         * config/pdp11/pdp11.c (pdp11_scalar_mode_supported_p): Likewise.
18701         * config/powerpcspe/powerpcspe.c (rs6000_scalar_mode_supported_p):
18702         Likewise.
18703         * config/rs6000/rs6000.c (rs6000_scalar_mode_supported_p): Likewise.
18704         * config/s390/s390.c (s390_scalar_mode_supported_p): Likewise.
18705         * config/spu/spu.c (spu_scalar_mode_supported_p): Likewise.
18706         * config/tilegx/tilegx.c (tilegx_scalar_mode_supported_p): Likewise.
18707         * config/tilepro/tilepro.c (tilepro_scalar_mode_supported_p):
18708         Likewise.
18709         * doc/tm.texi: Regenerate.
18711 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18712             Alan Hayward  <alan.hayward@arm.com>
18713             David Sherwood  <david.sherwood@arm.com>
18715         * coretypes.h (opt_scalar_mode): New typedef.
18716         * gdbhooks.py (build_pretty_printers): Handle it.
18717         * machmode.h (mode_iterator::get_2xwider): Add overload for
18718         opt_mode<T>.
18719         * emit-rtl.c (init_emit_once): Use opt_scalar_mode when iterating
18720         over scalar modes.
18721         * expr.c (convert_mode_scalar): Likewise.
18722         * omp-low.c (omp_clause_aligned_alignment): Likewise.
18723         * optabs.c (expand_float): Likewise.
18724         (expand_fix): Likewise.
18725         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
18727 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18728             Alan Hayward  <alan.hayward@arm.com>
18729             David Sherwood  <david.sherwood@arm.com>
18731         * optabs.c (expand_float): Explicitly check for scalars before
18732         using a branching expansion.
18733         (expand_fix): Likewise.
18735 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18736             Alan Hayward  <alan.hayward@arm.com>
18737             David Sherwood  <david.sherwood@arm.com>
18739         * expr.c (convert_mode): Split scalar handling out into...
18740         (convert_mode_scalar): ...this new function.  Treat the modes
18741         as scalar_modes.
18743 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18744             Alan Hayward  <alan.hayward@arm.com>
18745             David Sherwood  <david.sherwood@arm.com>
18747         * omp-expand.c (expand_omp_atomic): Use is_int_mode, is_float_mode
18748         and scalar_mode.
18749         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Likewise.
18751 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18752             Alan Hayward  <alan.hayward@arm.com>
18753             David Sherwood  <david.sherwood@arm.com>
18755         * fixed-value.h (fixed_from_double_int): Take a scalar_mode
18756         rather than a machine_mode.
18757         (fixed_from_string): Likewise.
18758         (fixed_convert): Likewise.
18759         (fixed_convert_from_int): Likewise.
18760         (fixed_convert_from_real): Likewise.
18761         (real_convert_from_fixed): Likewise.
18762         * fixed-value.c (fixed_from_double_int): Likewise.
18763         (fixed_from_string): Likewise.
18764         (fixed_convert): Likewise.
18765         (fixed_convert_from_int): Likewise.
18766         (fixed_convert_from_real): Likewise.
18767         (real_convert_from_fixed): Likewise.
18768         * config/avr/avr.c (avr_out_round): Use as_a <scalar_mode>.
18770 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18771             Alan Hayward  <alan.hayward@arm.com>
18772             David Sherwood  <david.sherwood@arm.com>
18774         * emit-rtl.c (immed_double_const): Use is_a <scalar_mode> instead
18775         of separate mode class checks.  Do not allow vector modes here.
18776         (immed_wide_int_const): Use as_a <scalar_mode>.
18777         * explow.c (trunc_int_for_mode): Likewise.
18778         * rtl.h (wi::int_traits<rtx_mode_t>::get_precision): Likewise.
18779         (wi::shwi): Likewise.
18780         (wi::min_value): Likewise.
18781         (wi::max_value): Likewise.
18782         * dwarf2out.c (loc_descriptor): Likewise.
18783         * simplify-rtx.c (simplify_immed_subreg): Fix rtx_mode_t argument
18784         for CONST_WIDE_INT.
18786 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18787             Alan Hayward  <alan.hayward@arm.com>
18788             David Sherwood  <david.sherwood@arm.com>
18790         * tree.h (SCALAR_TYPE_MODE): New macro.
18791         * expr.c (expand_expr_addr_expr_1): Use it.
18792         (expand_expr_real_2): Likewise.
18793         * fold-const.c (fold_convert_const_fixed_from_fixed): Likeise.
18794         (fold_convert_const_fixed_from_int): Likewise.
18795         (fold_convert_const_fixed_from_real): Likewise.
18796         (native_encode_fixed): Likewise
18797         (native_encode_complex): Likewise
18798         (native_encode_vector): Likewise.
18799         (native_interpret_fixed): Likewise.
18800         (native_interpret_real): Likewise.
18801         (native_interpret_complex): Likewise.
18802         (native_interpret_vector): Likewise.
18803         * omp-simd-clone.c (simd_clone_adjust_return_type): Likewise.
18804         (simd_clone_adjust_argument_types): Likewise.
18805         (simd_clone_init_simd_arrays): Likewise.
18806         (simd_clone_adjust): Likewise.
18807         * stor-layout.c (layout_type): Likewise.
18808         * tree.c (build_minus_one_cst): Likewise.
18809         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
18810         * tree-inline.c (estimate_move_cost): Likewise.
18811         * tree-ssa-math-opts.c (convert_plusminus_to_widen): Likewise.
18812         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise.
18813         (vectorizable_reduction): Likewise.
18814         * tree-vect-patterns.c (vect_recog_widen_mult_pattern): Likewise.
18815         (vect_recog_mixed_size_cond_pattern): Likewise.
18816         (check_bool_pattern): Likewise.
18817         (adjust_bool_pattern): Likewise.
18818         (search_type_for_mask_1): Likewise.
18819         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
18820         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
18821         (vectorizable_load): Likewise.
18822         (vectorizable_store): Likewise.
18823         * ubsan.c (ubsan_encode_value): Likewise.
18824         * varasm.c (output_constant): Likewise.
18826 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18827             Alan Hayward  <alan.hayward@arm.com>
18828             David Sherwood  <david.sherwood@arm.com>
18830         * coretypes.h (scalar_mode): New class.
18831         * machmode.h (scalar_mode): Likewise.
18832         (scalar_mode::includes_p): New function.
18833         (mode_to_inner): Return a scalar_mode rather than a machine_mode.
18834         * gdbhooks.py (build_pretty_printers): Handle scalar_mode.
18835         * genmodes.c (get_mode_class): Handle remaining scalar modes.
18836         * cfgexpand.c (expand_debug_expr): Use scalar_mode.
18837         * expmed.c (store_bit_field_1): Likewise.
18838         (extract_bit_field_1): Likewise.
18839         * expr.c (write_complex_part): Likewise.
18840         (read_complex_part): Likewise.
18841         (emit_move_complex_push): Likewise.
18842         (expand_expr_real_2): Likewise.
18843         * function.c (assign_parm_setup_reg): Likewise.
18844         (assign_parms_unsplit_complex): Likewise.
18845         * optabs.c (expand_binop): Likewise.
18846         * rtlanal.c (subreg_get_info): Likewise.
18847         * simplify-rtx.c (simplify_immed_subreg): Likewise.
18848         * varasm.c (output_constant_pool_2): Likewise.
18850 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18851             Alan Hayward  <alan.hayward@arm.com>
18852             David Sherwood  <david.sherwood@arm.com>
18854         * expmed.c (extract_high_half): Use scalar_int_mode and remove
18855         assertion.
18856         (expmed_mult_highpart_optab): Likewise.
18857         (expmed_mult_highpart): Likewise.
18859 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18860             Alan Hayward  <alan.hayward@arm.com>
18861             David Sherwood  <david.sherwood@arm.com>
18863         * builtins.h (builtin_strncpy_read_str): Take a scalar_int_mode
18864         instead of a machine_mode.
18865         (builtin_memset_read_str): Likewise.
18866         * builtins.c (c_readstr): Likewise.
18867         (builtin_memcpy_read_str): Likewise.
18868         (builtin_strncpy_read_str): Likewise.
18869         (builtin_memset_read_str): Likewise.
18870         (builtin_memset_gen_str): Likewise.
18871         (expand_builtin_signbit): Use scalar_int_mode for local variables.
18872         * cfgexpand.c (convert_debug_memory_address): Take a scalar_int_mode
18873         instead of a machine_mode.
18874         * combine.c (simplify_if_then_else): Use scalar_int_mode for local
18875         variables.
18876         (make_extraction): Likewise.
18877         (try_widen_shift_mode): Take and return scalar_int_modes instead
18878         of machine_modes.
18879         * config/aarch64/aarch64.c (aarch64_libgcc_cmp_return_mode): Return
18880         a scalar_int_mode instead of a machine_mode.
18881         * config/avr/avr.c (avr_addr_space_address_mode): Likewise.
18882         (avr_addr_space_pointer_mode): Likewise.
18883         * config/cr16/cr16.c (cr16_unwind_word_mode): Likewise.
18884         * config/msp430/msp430.c (msp430_addr_space_pointer_mode): Likewise.
18885         (msp430_unwind_word_mode): Likewise.
18886         * config/spu/spu.c (spu_unwind_word_mode): Likewise.
18887         (spu_addr_space_pointer_mode): Likewise.
18888         (spu_addr_space_address_mode): Likewise.
18889         (spu_libgcc_cmp_return_mode): Likewise.
18890         (spu_libgcc_shift_count_mode): Likewise.
18891         * config/rl78/rl78.c (rl78_addr_space_address_mode): Likewise.
18892         (rl78_addr_space_pointer_mode): Likewise.
18893         (fl78_unwind_word_mode): Likewise.
18894         (rl78_valid_pointer_mode): Take a scalar_int_mode instead of a
18895         machine_mode.
18896         * config/alpha/alpha.c (vms_valid_pointer_mode): Likewise.
18897         * config/ia64/ia64.c (ia64_vms_valid_pointer_mode): Likewise.
18898         * config/mips/mips.c (mips_mode_rep_extended): Likewise.
18899         (mips_valid_pointer_mode): Likewise.
18900         * config/tilegx/tilegx.c (tilegx_mode_rep_extended): Likewise.
18901         * config/ft32/ft32.c (ft32_valid_pointer_mode): Likewise.
18902         (ft32_addr_space_pointer_mode): Return a scalar_int_mode instead
18903         of a machine_mode.
18904         (ft32_addr_space_address_mode): Likewise.
18905         * config/m32c/m32c.c (m32c_valid_pointer_mode): Take a
18906         scalar_int_mode instead of a machine_mode.
18907         (m32c_addr_space_pointer_mode): Return a scalar_int_mode instead
18908         of a machine_mode.
18909         (m32c_addr_space_address_mode): Likewise.
18910         * config/powerpcspe/powerpcspe.c (rs6000_abi_word_mode): Likewise.
18911         (rs6000_eh_return_filter_mode): Likewise.
18912         * config/rs6000/rs6000.c (rs6000_abi_word_mode): Likewise.
18913         (rs6000_eh_return_filter_mode): Likewise.
18914         * config/s390/s390.c (s390_libgcc_cmp_return_mode): Likewise.
18915         (s390_libgcc_shift_count_mode): Likewise.
18916         (s390_unwind_word_mode): Likewise.
18917         (s390_valid_pointer_mode): Take a scalar_int_mode rather than a
18918         machine_mode.
18919         * target.def (mode_rep_extended): Likewise.
18920         (valid_pointer_mode): Likewise.
18921         (addr_space.valid_pointer_mode): Likewise.
18922         (eh_return_filter_mode): Return a scalar_int_mode rather than
18923         a machine_mode.
18924         (libgcc_cmp_return_mode): Likewise.
18925         (libgcc_shift_count_mode): Likewise.
18926         (unwind_word_mode): Likewise.
18927         (addr_space.pointer_mode): Likewise.
18928         (addr_space.address_mode): Likewise.
18929         * doc/tm.texi: Regenerate.
18930         * dojump.c (prefer_and_bit_test): Take a scalar_int_mode rather than
18931         a machine_mode.
18932         (do_jump): Use scalar_int_mode for local variables.
18933         * dwarf2cfi.c (init_return_column_size): Take a scalar_int_mode
18934         rather than a machine_mode.
18935         * dwarf2out.c (convert_descriptor_to_mode): Likewise.
18936         (scompare_loc_descriptor_wide): Likewise.
18937         (scompare_loc_descriptor_narrow): Likewise.
18938         * emit-rtl.c (adjust_address_1): Use scalar_int_mode for local
18939         variables.
18940         * except.c (sjlj_emit_dispatch_table): Likewise.
18941         (expand_builtin_eh_copy_values): Likewise.
18942         * explow.c (convert_memory_address_addr_space_1): Likewise.
18943         Take a scalar_int_mode rather than a machine_mode.
18944         (convert_memory_address_addr_space): Take a scalar_int_mode rather
18945         than a machine_mode.
18946         (memory_address_addr_space): Use scalar_int_mode for local variables.
18947         * expmed.h (expand_mult_highpart_adjust): Take a scalar_int_mode
18948         rather than a machine_mode.
18949         * expmed.c (mask_rtx): Likewise.
18950         (init_expmed_one_conv): Likewise.
18951         (expand_mult_highpart_adjust): Likewise.
18952         (extract_high_half): Likewise.
18953         (expmed_mult_highpart_optab): Likewise.
18954         (expmed_mult_highpart): Likewise.
18955         (expand_smod_pow2): Likewise.
18956         (expand_sdiv_pow2): Likewise.
18957         (emit_store_flag_int): Likewise.
18958         (adjust_bit_field_mem_for_reg): Use scalar_int_mode for local
18959         variables.
18960         (extract_low_bits): Likewise.
18961         * expr.h (by_pieces_constfn): Take a scalar_int_mode rather than
18962         a machine_mode.
18963         * expr.c (pieces_addr::adjust):  Likewise.
18964         (can_store_by_pieces): Likewise.
18965         (store_by_pieces): Likewise.
18966         (clear_by_pieces_1): Likewise.
18967         (expand_expr_addr_expr_1): Likewise.
18968         (expand_expr_addr_expr): Use scalar_int_mode for local variables.
18969         (expand_expr_real_1): Likewise.
18970         (try_casesi): Likewise.
18971         * final.c (shorten_branches): Likewise.
18972         * fold-const.c (fold_convert_const_int_from_fixed): Change the
18973         type of "mode" to machine_mode.
18974         * internal-fn.c (expand_arith_overflow_result_store): Take a
18975         scalar_int_mode rather than a machine_mode.
18976         (expand_mul_overflow): Use scalar_int_mode for local variables.
18977         * loop-doloop.c (doloop_modify): Likewise.
18978         (doloop_optimize): Likewise.
18979         * optabs.c (expand_subword_shift): Take a scalar_int_mode rather
18980         than a machine_mode.
18981         (expand_doubleword_shift_condmove): Likewise.
18982         (expand_doubleword_shift): Likewise.
18983         (expand_doubleword_clz): Likewise.
18984         (expand_doubleword_popcount): Likewise.
18985         (expand_doubleword_parity): Likewise.
18986         (expand_absneg_bit): Use scalar_int_mode for local variables.
18987         (prepare_float_lib_cmp): Likewise.
18988         * rtl.h (convert_memory_address_addr_space_1): Take a scalar_int_mode
18989         rather than a machine_mode.
18990         (convert_memory_address_addr_space): Likewise.
18991         (get_mode_bounds): Likewise.
18992         (get_address_mode): Return a scalar_int_mode rather than a
18993         machine_mode.
18994         * rtlanal.c (get_address_mode): Likewise.
18995         * stor-layout.c (get_mode_bounds): Take a scalar_int_mode rather
18996         than a machine_mode.
18997         * targhooks.c (default_mode_rep_extended): Likewise.
18998         (default_valid_pointer_mode): Likewise.
18999         (default_addr_space_valid_pointer_mode): Likewise.
19000         (default_eh_return_filter_mode): Return a scalar_int_mode rather
19001         than a machine_mode.
19002         (default_libgcc_cmp_return_mode): Likewise.
19003         (default_libgcc_shift_count_mode): Likewise.
19004         (default_unwind_word_mode): Likewise.
19005         (default_addr_space_pointer_mode): Likewise.
19006         (default_addr_space_address_mode): Likewise.
19007         * targhooks.h (default_eh_return_filter_mode): Likewise.
19008         (default_libgcc_cmp_return_mode): Likewise.
19009         (default_libgcc_shift_count_mode): Likewise.
19010         (default_unwind_word_mode): Likewise.
19011         (default_addr_space_pointer_mode): Likewise.
19012         (default_addr_space_address_mode): Likewise.
19013         (default_mode_rep_extended): Take a scalar_int_mode rather than
19014         a machine_mode.
19015         (default_valid_pointer_mode): Likewise.
19016         (default_addr_space_valid_pointer_mode): Likewise.
19017         * tree-ssa-address.c (addr_for_mem_ref): Use scalar_int_mode for
19018         local variables.
19019         * tree-ssa-loop-ivopts.c (get_shiftadd_cost): Take a scalar_int_mode
19020         rather than a machine_mode.
19021         * tree-switch-conversion.c (array_value_type): Use scalar_int_mode
19022         for local variables.
19023         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19024         * var-tracking.c (use_narrower_mode): Take a scalar_int_mode rather
19025         than a machine_mode.
19027 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19028             Alan Hayward  <alan.hayward@arm.com>
19029             David Sherwood  <david.sherwood@arm.com>
19031         * dojump.c (do_jump_by_parts_greater_rtx): Change the type of
19032         the mode argument to scalar_int_mode.
19033         (do_jump_by_parts_zero_rtx): Likewise.
19034         (do_jump_by_parts_equality_rtx): Likewise.
19035         (do_jump_by_parts_greater): Take a mode argument.
19036         (do_jump_by_parts_equality): Likewise.
19037         (do_jump_1): Update calls accordingly.
19039 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19040             Alan Hayward  <alan.hayward@arm.com>
19041             David Sherwood  <david.sherwood@arm.com>
19043         * is-a.h (safe_dyn_cast): New function.
19044         * rtl.h (rtx_jump_table_data::get_data_mode): New function.
19045         (jump_table_for_label): Likewise.
19046         * final.c (final_addr_vec_align): Take an rtx_jump_table_data *
19047         instead of an rtx_insn *.
19048         (shorten_branches): Use dyn_cast instead of LABEL_P and
19049         JUMP_TABLE_DATA_P.  Use jump_table_for_label and
19050         rtx_jump_table_data::get_data_mode.
19051         (final_scan_insn): Likewise.
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         * combine.c (try_combine): Use is_a <scalar_int_mode> when
19058         trying to combine a full-register integer set with a subreg
19059         integer set.
19061 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19062             Alan Hayward  <alan.hayward@arm.com>
19063             David Sherwood  <david.sherwood@arm.com>
19065         * expr.c (expand_expr_addr_expr): Add a new_tmode local variable
19066         that is always either address_mode or pointer_mode.
19068 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19069             Alan Hayward  <alan.hayward@arm.com>
19070             David Sherwood  <david.sherwood@arm.com>
19072         * expr.c (expand_expr_real_2): Use word_mode instead of innermode
19073         when the two are known to be equal.
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         * simplify-rtx.c (simplify_const_unary_operation): Use
19080         is_a <scalar_int_mode> instead of checking for a nonzero
19081         precision.  Forcibly convert op_mode to a scalar_int_mode
19082         in that case.  More clearly differentiate the operand and
19083         result modes and use the former when deciding what the value
19084         of a count-bits operation should be.  Use is_int_mode instead
19085         of checking for a MODE_INT.  Remove redundant check for whether
19086         this mode has a zero precision.
19088 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19089             Alan Hayward  <alan.hayward@arm.com>
19090             David Sherwood  <david.sherwood@arm.com>
19092         * optabs.c (widen_leading): Change the type of the mode argument
19093         to scalar_int_mode.  Use opt_scalar_int_mode for the mode iterator.
19094         (widen_bswap): Likewise.
19095         (expand_parity): Likewise.
19096         (expand_ctz): Change the type of the mode argument to scalar_int_mode.
19097         (expand_ffs): Likewise.
19098         (epand_unop): Check for scalar integer modes before calling the
19099         above routines.
19101 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19102             Alan Hayward  <alan.hayward@arm.com>
19103             David Sherwood  <david.sherwood@arm.com>
19105         * expr.c (const_scalar_mask_from_tree): Add a mode argument.
19106         Expand commentary.
19107         (expand_expr_real_1): Update call accordingly.
19109 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19110             Alan Hayward  <alan.hayward@arm.com>
19111             David Sherwood  <david.sherwood@arm.com>
19113         * expmed.c (store_bit_field_using_insv): Add op0_mode and
19114         value_mode arguments.  Use scalar_int_mode internally.
19115         (store_bit_field_1): Rename the new integer mode from imode
19116         to op0_mode and use it instead of GET_MODE (op0).  Update calls
19117         to store_split_bit_field, store_bit_field_using_insv and
19118         store_fixed_bit_field.
19119         (store_fixed_bit_field): Add op0_mode and value_mode arguments.
19120         Use scalar_int_mode internally.  Use a bit count rather than a mode
19121         when calculating the largest bit size for get_best_mode.
19122         Update calls to store_split_bit_field and store_fixed_bit_field_1.
19123         (store_fixed_bit_field_1): Add mode and value_mode arguments.
19124         Remove assertion that OP0 has a scalar integer mode.
19125         (store_split_bit_field): Add op0_mode and value_mode arguments.
19126         Update calls to extract_fixed_bit_field.
19127         (extract_bit_field_using_extv): Add an op0_mode argument.
19128         Use scalar_int_mode internally.
19129         (extract_bit_field_1): Rename the new integer mode from imode to
19130         op0_mode and use it instead of GET_MODE (op0).  Update calls to
19131         extract_split_bit_field, extract_bit_field_using_extv and
19132         extract_fixed_bit_field.
19133         (extract_fixed_bit_field): Add an op0_mode argument.  Update calls
19134         to extract_split_bit_field and extract_fixed_bit_field_1.
19135         (extract_fixed_bit_field_1): Add a mode argument.  Remove assertion
19136         that OP0 has a scalar integer mode.  Use as_a <scalar_int_mode>
19137         on the target mode.
19138         (extract_split_bit_field): Add an op0_mode argument.  Update call
19139         to extract_fixed_bit_field.
19141 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19142             Alan Hayward  <alan.hayward@arm.com>
19143             David Sherwood  <david.sherwood@arm.com>
19145         * cse.c (cse_insn): Use opt_scalar_int_mode for the mode iterator.
19146         * explow.c (hard_function_value): Likewise.
19147         * expmed.c (extract_fixed_bit_field_1): Likewise.  Move the
19148         convert_to_mode call outside the loop.
19149         * expr.c (alignment_for_piecewise_move): Use opt_scalar_int_mode
19150         for the mode iterator.  Require the mode specified by max_pieces
19151         to exist.
19152         (emit_block_move_via_movmem): Use opt_scalar_int_mode for the
19153         mode iterator.
19154         (copy_blkmode_to_reg): Likewise.
19155         (set_storage_via_setmem): Likewise.
19156         * optabs.c (prepare_cmp_insn): Likewise.
19157         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
19158         * stor-layout.c (finish_bitfield_representative): Likewise.
19160 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19161             Alan Hayward  <alan.hayward@arm.com>
19162             David Sherwood  <david.sherwood@arm.com>
19164         * rtl.h (subreg_unpromoted_mode, subreg_promoted_mode): New functions.
19165         * expr.c (convert_move): Use them.
19166         (convert_modes): Likewise.
19167         (store_expr_with_bounds): Likewise.
19169 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19170             Alan Hayward  <alan.hayward@arm.com>
19171             David Sherwood  <david.sherwood@arm.com>
19173         * rtl.h (rtl_hooks::reg_nonzero_bits): Add a scalar_int_mode
19174         parameter for the mode of "x".  Remove the "known_x", "known_mode"
19175         and "known_ret" arguments.  Change the type of the mode argument
19176         to scalar_int_mode.
19177         (rtl_hooks:reg_num_sign_bit_copies): Likewise.
19178         * combine.c (reg_nonzero_bits_for_combine): Update accordingly.
19179         (reg_num_sign_bit_copies_for_combine): Likewise.
19180         * rtlanal.c (nonzero_bits1): Likewise.
19181         (num_sign_bit_copies1): Likewise.
19182         * rtlhooks-def.h (reg_nonzero_bits_general): Likewise.
19183         (reg_num_sign_bit_copies_general): Likewise.
19184         * rtlhooks.c (reg_num_sign_bit_copies_general): Likewise.
19185         (reg_nonzero_bits_general): Likewise.
19187 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19188             Alan Hayward  <alan.hayward@arm.com>
19189             David Sherwood  <david.sherwood@arm.com>
19191         * rtlanal.c (num_sign_bit_copies): Handle VOIDmode here rather
19192         than in subroutines.  Return 1 for non-integer modes.
19193         (cached_num_sign_bit_copies): Change the type of the mode parameter
19194         to scalar_int_mode.
19195         (num_sign_bit_copies1): Likewise.  Remove early exit for other mode
19196         classes.  Handle CONST_INT_P first and then check whether X also
19197         has a scalar integer mode.  Check the same thing for inner registers
19198         of a SUBREG and for values that are being extended or truncated.
19200 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19201             Alan Hayward  <alan.hayward@arm.com>
19202             David Sherwood  <david.sherwood@arm.com>
19204         * rtlanal.c (nonzero_bits): Handle VOIDmode here rather than
19205         in subroutines.  Return the mode mask for non-integer modes.
19206         (cached_nonzero_bits): Change the type of the mode parameter
19207         to scalar_int_mode.
19208         (nonzero_bits1): Likewise.  Remove early exit for other mode
19209         classes.  Handle CONST_INT_P first and then check whether X
19210         also has a scalar integer mode.
19212 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19213             Alan Hayward  <alan.hayward@arm.com>
19214             David Sherwood  <david.sherwood@arm.com>
19216         * expr.c (widest_int_mode_for_size): Make the comment match the code.
19217         Return a scalar_int_mode and assert that the size is greater than
19218         one byte.
19219         (by_pieces_ninsns): Update accordingly and remove VOIDmode handling.
19220         (op_by_pieces_d::op_by_pieces_d): Likewise.
19221         (op_by_pieces_d::run): Likewise.
19222         (can_store_by_pieces): Likewise.
19224 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19225             Alan Hayward  <alan.hayward@arm.com>
19226             David Sherwood  <david.sherwood@arm.com>
19228         * combine.c (extract_left_shift): Add a mode argument and update
19229         recursive calls.
19230         (make_compound_operation_int): Change the type of the mode parameter
19231         to scalar_int_mode and update the call to extract_left_shift.
19233 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19234             Alan Hayward  <alan.hayward@arm.com>
19235             David Sherwood  <david.sherwood@arm.com>
19237         * combine.c (simplify_and_const_int): Change the type of the mode
19238         parameter to scalar_int_mode.
19239         (simplify_and_const_int_1): Likewise.  Update recursive call.
19241 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19242             Alan Hayward  <alan.hayward@arm.com>
19243             David Sherwood  <david.sherwood@arm.com>
19245         * combine.c (simplify_compare_const): Check that the mode is a
19246         scalar_int_mode (rather than VOIDmode) before testing its
19247         precision.
19248         (simplify_comparison): Move COMPARISON_P handling out of the
19249         loop and restrict the latter part of the loop to scalar_int_modes.
19250         Check is_a <scalar_int_mode> before calling HWI_COMPUTABLE_MODE_P
19251         and when considering SUBREG_REGs.  Use is_int_mode instead of
19252         checking GET_MODE_CLASS against MODE_INT.
19254 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19255             Alan Hayward  <alan.hayward@arm.com>
19256             David Sherwood  <david.sherwood@arm.com>
19258         * combine.c (try_widen_shift_mode): Move check for equal modes to...
19259         (simplify_shift_const_1): ...here.  Use scalar_int_mode for
19260         shift_unit_mode and for modes involved in scalar shifts.
19262 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19263             Alan Hayward  <alan.hayward@arm.com>
19264             David Sherwood  <david.sherwood@arm.com>
19266         * combine.c (force_int_to_mode): New function, split out from...
19267         (force_to_mode): ...here.  Keep xmode up-to-date and use it
19268         instead of GET_MODE (x).
19270 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19271             Alan Hayward  <alan.hayward@arm.com>
19272             David Sherwood  <david.sherwood@arm.com>
19274         * optabs-query.h (extraction_insn::struct_mode): Change type to
19275         opt_scalar_int_mode and update comment.
19276         (extraction_insn::field_mode): Change type to scalar_int_mode.
19277         (extraction_insn::pos_mode): Likewise.
19278         * combine.c (make_extraction): Update accordingly.
19279         * optabs-query.c (get_traditional_extraction_insn): Likewise.
19280         (get_optab_extraction_insn): Likewise.
19281         * recog.c (simplify_while_replacing): Likewise.
19282         * expmed.c (narrow_bit_field_mem): Change the type of the mode
19283         parameter to opt_scalar_int_mode.
19285 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19286             Alan Hayward  <alan.hayward@arm.com>
19287             David Sherwood  <david.sherwood@arm.com>
19289         * machmode.h (bit_field_mode_iterator::next_mode): Take a pointer
19290         to a scalar_int_mode instead of a machine_mode.
19291         (bit_field_mode_iterator::m_mode): Change type to opt_scalar_int_mode.
19292         (get_best_mode): Return a boolean and use a pointer argument to store
19293         the selected mode.  Replace the limit mode parameter with a bit limit.
19294         * expmed.c (adjust_bit_field_mem_for_reg): Use scalar_int_mode
19295         for the values returned by bit_field_mode_iterator::next_mode.
19296         (store_bit_field): Update call to get_best_mode.
19297         (store_fixed_bit_field): Likewise.
19298         (extract_fixed_bit_field): Likewise.
19299         * expr.c (optimize_bitfield_assignment_op): Likewise.
19300         * fold-const.c (optimize_bit_field_compare): Likewise.
19301         (fold_truth_andor_1): Likewise.
19302         * stor-layout.c (bit_field_mode_iterator::next_mode): As above.
19303         Update for new type of m_mode.
19304         (get_best_mode): As above.
19306 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19307             Alan Hayward  <alan.hayward@arm.com>
19308             David Sherwood  <david.sherwood@arm.com>
19310         * expmed.c (strict_volatile_bitfield_p): Change the type of fieldmode
19311         to scalar_int_mode.  Remove check for SCALAR_INT_MODE_P.
19312         (store_bit_field): Check is_a <scalar_int_mode> before calling
19313         strict_volatile_bitfield_p.
19314         (extract_bit_field): Likewise.
19316 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19317             Alan Hayward  <alan.hayward@arm.com>
19318             David Sherwood  <david.sherwood@arm.com>
19320         * target.def (cstore_mode): Return a scalar_int_mode.
19321         * doc/tm.texi: Regenerate.
19322         * config/sparc/sparc.c (sparc_cstore_mode): Return a scalar_int_mode.
19323         * targhooks.h (default_cstore_mode): Likewise.
19324         * targhooks.c (default_cstore_mode): Likewise, using a forced
19325         conversion.
19326         * expmed.c (emit_cstore): Expect the target of the cstore to be
19327         a scalar_int_mode.
19329 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19330             Alan Hayward  <alan.hayward@arm.com>
19331             David Sherwood  <david.sherwood@arm.com>
19333         * cfgloop.h (rtx_iv): Change type of extend_mode and mode to
19334         scalar_int_mode.
19335         (niter_desc): Likewise mode.
19336         (iv_analyze): Add a mode parameter.
19337         (biv_p): Likewise.
19338         (iv_analyze_expr): Pass the mode paraeter before the rtx it describes
19339         and change its type to scalar_int_mode.
19340         * loop-iv.c: Update commentary at head of file.
19341         (iv_constant): Pass the mode paraeter before the rtx it describes
19342         and change its type to scalar_int_mode.  Remove VOIDmode handling.
19343         (iv_subreg): Change the type of the mode parameter to scalar_int_mode.
19344         (iv_extend): Likewise.
19345         (shorten_into_mode): Likewise.
19346         (iv_add): Use scalar_int_mode.
19347         (iv_mult): Likewise.
19348         (iv_shift): Likewise.
19349         (canonicalize_iv_subregs): Likewise.
19350         (get_biv_step_1): Pass the outer_mode parameter before the rtx
19351         it describes and change its mode to scalar_int_mode.   Also change
19352         the type of the returned inner_mode to scalar_int_mode.
19353         (get_biv_step): Likewise, turning outer_mode from a pointer
19354         into a direct parameter.  Update call to get_biv_step_1.
19355         (iv_analyze_biv): Add an outer_mode parameter.  Update calls to
19356         iv_constant and get_biv_step.
19357         (iv_analyze_expr): Pass the mode parameter before the rtx it describes
19358         and change its type to scalar_int_mode.  Don't initialise iv->mode
19359         to VOIDmode and remove later checks for its still being VOIDmode.
19360         Update calls to iv_analyze_op and iv_analyze_expr.  Check
19361         is_a <scalar_int_mode> when changing the mode under consideration.
19362         (iv_analyze_def): Ignore registers that don't have a scalar_int_mode.
19363         Update call to iv_analyze_expr.
19364         (iv_analyze_op): Add a mode parameter.  Reject subregs whose
19365         inner register is not also a scalar_int_mode.  Update call to
19366         iv_analyze_biv.
19367         (iv_analyze): Add a mode parameter.  Update call to iv_analyze_op.
19368         (biv_p): Add a mode parameter.  Update call to iv_analyze_biv.
19369         (iv_number_of_iterations): Use is_a <scalar_int_mode> instead of
19370         separate mode class checks.  Update calls to iv_analyze.  Remove
19371         fix-up of VOIDmodes after iv_analyze_biv.
19372         * loop-unroll.c (analyze_iv_to_split_insn): Reject registers that
19373         don't have a scalar_int_mode.  Update call to biv_p.
19375 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19376             Alan Hayward  <alan.hayward@arm.com>
19377             David Sherwood  <david.sherwood@arm.com>
19379         * cfgexpand.c (convert_debug_memory_address): Use
19380         as_a <scalar_int_mode>.
19381         * combine.c (expand_compound_operation): Likewise.
19382         (make_extraction): Likewise.
19383         (change_zero_ext): Likewise.
19384         (simplify_comparison): Likewise.
19385         * cse.c (cse_insn): Likewise.
19386         * dwarf2out.c (minmax_loc_descriptor): Likewise.
19387         (mem_loc_descriptor): Likewise.
19388         (loc_descriptor): Likewise.
19389         * expmed.c (init_expmed_one_mode): Likewise.
19390         (synth_mult): Likewise.
19391         (emit_store_flag_1): Likewise.
19392         (expand_divmod): Likewise.  Use HWI_COMPUTABLE_MODE_P instead
19393         of a comparison with size.
19394         * expr.c (expand_assignment): Use as_a <scalar_int_mode>.
19395         (reduce_to_bit_field_precision): Likewise.
19396         * function.c (expand_function_end): Likewise.
19397         * internal-fn.c (expand_arith_overflow_result_store): Likewise.
19398         * loop-doloop.c (doloop_modify): Likewise.
19399         * optabs.c (expand_binop): Likewise.
19400         (expand_unop): Likewise.
19401         (expand_copysign_absneg): Likewise.
19402         (prepare_cmp_insn): Likewise.
19403         (maybe_legitimize_operand): Likewise.
19404         * recog.c (const_scalar_int_operand): Likewise.
19405         * rtlanal.c (get_address_mode): Likewise.
19406         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19407         (simplify_cond_clz_ctz): Likewise.
19408         * tree-nested.c (get_nl_goto_field): Likewise.
19409         * tree.c (build_vector_type_for_mode): Likewise.
19410         * var-tracking.c (use_narrower_mode): Likewise.
19412 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19413             Alan Hayward  <alan.hayward@arm.com>
19414             David Sherwood  <david.sherwood@arm.com>
19416         * tree.h (SCALAR_INT_TYPE_MODE): New macro.
19417         * builtins.c (expand_builtin_signbit): Use it.
19418         * cfgexpand.c (expand_debug_expr): Likewise.
19419         * dojump.c (do_jump): Likewise.
19420         (do_compare_and_jump): Likewise.
19421         * dwarf2cfi.c (expand_builtin_init_dwarf_reg_sizes): Likewise.
19422         * expmed.c (make_tree): Likewise.
19423         * expr.c (expand_expr_real_2): Likewise.
19424         (expand_expr_real_1): Likewise.
19425         (try_casesi): Likewise.
19426         * fold-const-call.c (fold_const_call_ss): Likewise.
19427         * fold-const.c (unextend): Likewise.
19428         (extract_muldiv_1): Likewise.
19429         (fold_single_bit_test): Likewise.
19430         (native_encode_int): Likewise.
19431         (native_encode_string): Likewise.
19432         (native_interpret_int): Likewise.
19433         * gimple-fold.c (gimple_fold_builtin_memset): Likewise.
19434         * internal-fn.c (expand_addsub_overflow): Likewise.
19435         (expand_neg_overflow): Likewise.
19436         (expand_mul_overflow): Likewise.
19437         (expand_arith_overflow): Likewise.
19438         * match.pd: Likewise.
19439         * stor-layout.c (layout_type): Likewise.
19440         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
19441         * tree-ssa-math-opts.c (convert_mult_to_widen): Likewise.
19442         * tree-ssanames.c (get_range_info): Likewise.
19443         * tree-switch-conversion.c (array_value_type) Likewise.
19444         * tree-vect-patterns.c (vect_recog_rotate_pattern): Likewise.
19445         (vect_recog_divmod_pattern): Likewise.
19446         (vect_recog_mixed_size_cond_pattern): Likewise.
19447         * tree-vrp.c (extract_range_basic): Likewise.
19448         (simplify_float_conversion_using_ranges): Likewise.
19449         * tree.c (int_fits_type_p): Likewise.
19450         * ubsan.c (instrument_bool_enum_load): Likewise.
19451         * varasm.c (mergeable_string_section): Likewise.
19452         (narrowing_initializer_constant_valid_p): Likewise.
19453         (output_constant): Likewise.
19455 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19456             Alan Hayward  <alan.hayward@arm.com>
19457             David Sherwood  <david.sherwood@arm.com>
19459         * machmode.h (NARROWEST_INT_MODE): New macro.
19460         * expr.c (alignment_for_piecewise_move): Use it instead of
19461         GET_CLASS_NARROWEST_MODE (MODE_INT).
19462         (push_block): Likewise.
19463         * stor-layout.c (bit_field_mode_iterator::bit_field_mode_iterator):
19464         Likewise.
19465         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19467 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19468             Alan Hayward  <alan.hayward@arm.com>
19469             David Sherwood  <david.sherwood@arm.com>
19471         * postreload.c (move2add_valid_value_p): Change the type of the
19472         mode parameter to scalar_int_mode.
19473         (move2add_use_add2_insn): Add a mode parameter and use it instead
19474         of GET_MODE (reg).
19475         (move2add_use_add3_insn): Likewise.
19476         (reload_cse_move2add): Update accordingly.
19478 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19479             Alan Hayward  <alan.hayward@arm.com>
19480             David Sherwood  <david.sherwood@arm.com>
19482         * expr.c (expand_expr_real_2): Use scalar_int_mode for the
19483         double-word mode.
19484         * lower-subreg.c (resolve_shift_zext): Use is_a <scalar_int_mode>.
19485         * optabs.c (expand_unop): Likewise.
19487 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19488             Alan Hayward  <alan.hayward@arm.com>
19489             David Sherwood  <david.sherwood@arm.com>
19491         * dwarf2out.c (typed_binop): Change mode parameter to scalar_int_mode.
19492         (clz_loc_descriptor): Likewise.  Remove SCALAR_INT_MODE_P check.
19493         (popcount_loc_descriptor): Likewise.
19494         (bswap_loc_descriptor): Likewise.
19495         (rotate_loc_descriptor): Likewise.
19496         (mem_loc_descriptor): Add is_a <scalar_int_mode> checks before
19497         calling the functions above.
19499 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19500             Alan Hayward  <alan.hayward@arm.com>
19501             David Sherwood  <david.sherwood@arm.com>
19503         * combine.c (sign_extend_short_imm): Add is_a <scalar_int_mode>
19504         checks.
19505         (try_combine): Likewise.
19506         (simplify_if_then_else): Likewise.
19507         * cse.c (cse_insn): Likewise.
19508         * dwarf2out.c (mem_loc_descriptor): Likewise.
19509         * emit-rtl.c (gen_lowpart_common): Likewise.
19510         * simplify-rtx.c (simplify_truncation): Likewise.
19511         (simplify_binary_operation_1): Likewise.
19512         (simplify_const_relational_operation): Likewise.
19513         (simplify_ternary_operation): Likewise.
19514         * tree-ssa-loop-ivopts.c (force_expr_to_var_cost): Likewise.
19516 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19517             Alan Hayward  <alan.hayward@arm.com>
19518             David Sherwood  <david.sherwood@arm.com>
19520         * cse.c (cse_insn): Add is_a <scalar_int_mode> checks.
19521         * reload.c (push_reload): Likewise.
19522         (find_reloads): Likewise.
19524 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19525             Alan Hayward  <alan.hayward@arm.com>
19526             David Sherwood  <david.sherwood@arm.com>
19528         * combine.c (find_split_point): Add is_a <scalar_int_mode> checks.
19529         (make_compound_operation_int): Likewise.
19530         (change_zero_ext): Likewise.
19531         * expr.c (convert_move): Likewise.
19532         (convert_modes): Likewise.
19533         * fwprop.c (forward_propagate_subreg): Likewise.
19534         * loop-iv.c (get_biv_step_1): Likewise.
19535         * optabs.c (widen_operand): Likewise.
19536         * postreload.c (move2add_valid_value_p): Likewise.
19537         * recog.c (simplify_while_replacing): Likewise.
19538         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19539         (simplify_binary_operation_1): Likewise.  Remove redundant
19540         mode equality check.
19542 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19543             Alan Hayward  <alan.hayward@arm.com>
19544             David Sherwood  <david.sherwood@arm.com>
19546         * combine.c (combine_simplify_rtx): Add checks for
19547         is_a <scalar_int_mode>.
19548         (simplify_if_then_else): Likewise.
19549         (make_field_assignment): Likewise.
19550         (simplify_comparison): Likewise.
19551         * ifcvt.c (noce_try_bitop): Likewise.
19552         * loop-invariant.c (canonicalize_address_mult): Likewise.
19553         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19555 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19556             Alan Hayward  <alan.hayward@arm.com>
19557             David Sherwood  <david.sherwood@arm.com>
19559         * gimple-fold.c (gimple_fold_builtin_memory_op): Use
19560         is_a <scalar_int_mode> instead of != BLKmode.
19562 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19563             Alan Hayward  <alan.hayward@arm.com>
19564             David Sherwood  <david.sherwood@arm.com>
19566         * cfgexpand.c (expand_debug_expr): Use is_a <scalar_int_mode>
19567         instead of != VOIDmode.
19568         * combine.c (if_then_else_cond): Likewise.
19569         (change_zero_ext): Likewise.
19570         * dwarf2out.c (mem_loc_descriptor): Likewise.
19571         (loc_descriptor): Likewise.
19572         * rtlanal.c (canonicalize_condition): Likewise.
19573         * simplify-rtx.c (simplify_relational_operation_1): Likewise.
19575 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19576             Alan Hayward  <alan.hayward@arm.com>
19577             David Sherwood  <david.sherwood@arm.com>
19579         * simplify-rtx.c (simplify_binary_operation_1): Use
19580         is_a <scalar_int_mode> instead of !VECTOR_MODE_P.
19582 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19583             Alan Hayward  <alan.hayward@arm.com>
19584             David Sherwood  <david.sherwood@arm.com>
19586         * wide-int.h (int_traits<unsigned char>) New class.
19587         (int_traits<unsigned short>) Likewise.
19588         * cfgexpand.c (expand_debug_expr): Use is_a <scalar_int_mode>.
19589         Use GET_MODE_UNIT_PRECISION and remove redundant test for
19590         SCALAR_INT_MODE_P.
19591         * combine.c (set_nonzero_bits_and_sign_copies): Use
19592         is_a <scalar_int_mode>.
19593         (find_split_point): Likewise.
19594         (combine_simplify_rtx): Likewise.
19595         (simplify_logical): Likewise.
19596         (expand_compound_operation): Likewise.
19597         (expand_field_assignment): Likewise.
19598         (make_compound_operation): Likewise.
19599         (extended_count): Likewise.
19600         (change_zero_ext): Likewise.
19601         (simplify_comparison): Likewise.
19602         * dwarf2out.c (scompare_loc_descriptor): Likewise.
19603         (ucompare_loc_descriptor): Likewise.
19604         (minmax_loc_descriptor): Likewise.
19605         (mem_loc_descriptor): Likewise.
19606         (loc_descriptor): Likewise.
19607         * expmed.c (init_expmed_one_mode): Likewise.
19608         * lra-constraints.c (lra_constraint_offset): Likewise.
19609         * optabs.c (prepare_libcall_arg): Likewise.
19610         * postreload.c (move2add_note_store): Likewise.
19611         * reload.c (operands_match_p): Likewise.
19612         * rtl.h (load_extend_op): Likewise.
19613         * rtlhooks.c (gen_lowpart_general): Likewise.
19614         * simplify-rtx.c (simplify_truncation): Likewise.
19615         (simplify_unary_operation_1): Likewise.
19616         (simplify_binary_operation_1): Likewise.
19617         (simplify_const_binary_operation): Likewise.
19618         (simplify_const_relational_operation): Likewise.
19619         (simplify_subreg): Likewise.
19620         * stor-layout.c (bitwise_mode_for_mode): Likewise.
19621         * var-tracking.c (adjust_mems): Likewise.
19622         (prepare_call_arguments): Likewise.
19624 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19625             Alan Hayward  <alan.hayward@arm.com>
19626             David Sherwood  <david.sherwood@arm.com>
19628         * machmode.h (is_int_mode): New fuction.
19629         * combine.c (find_split_point): Use it.
19630         (combine_simplify_rtx): Likewise.
19631         (simplify_if_then_else): Likewise.
19632         (simplify_set): Likewise.
19633         (simplify_shift_const_1): Likewise.
19634         (simplify_comparison): Likewise.
19635         * config/aarch64/aarch64.c (aarch64_rtx_costs): Likewise.
19636         * cse.c (notreg_cost): Likewise.
19637         (cse_insn): Likewise.
19638         * cselib.c (cselib_lookup_1): Likewise.
19639         * dojump.c (do_jump_1): Likewise.
19640         (do_compare_rtx_and_jump): Likewise.
19641         * dse.c (get_call_args): Likewise.
19642         * dwarf2out.c (rtl_for_decl_init): Likewise.
19643         (native_encode_initializer): Likewise.
19644         * expmed.c (emit_store_flag_1): Likewise.
19645         (emit_store_flag): Likewise.
19646         * expr.c (convert_modes): Likewise.
19647         (store_field): Likewise.
19648         (expand_expr_real_1): Likewise.
19649         * fold-const.c (fold_read_from_constant_string): Likewise.
19650         * gimple-ssa-sprintf.c (get_format_string): Likewise.
19651         * optabs-libfuncs.c (gen_int_libfunc): Likewise.
19652         * optabs.c (expand_binop): Likewise.
19653         (expand_unop): Likewise.
19654         (expand_abs_nojump): Likewise.
19655         (expand_one_cmpl_abs_nojump): Likewise.
19656         * simplify-rtx.c (mode_signbit_p): Likewise.
19657         (val_signbit_p): Likewise.
19658         (val_signbit_known_set_p): Likewise.
19659         (val_signbit_known_clear_p): Likewise.
19660         (simplify_relational_operation_1): Likewise.
19661         * tree.c (vector_type_mode): Likewise.
19663 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19664             Alan Hayward  <alan.hayward@arm.com>
19665             David Sherwood  <david.sherwood@arm.com>
19667         * machmode.h (smallest_mode_for_size): Fix formatting.
19668         (smallest_int_mode_for_size): New function.
19669         * cfgexpand.c (expand_debug_expr): Use smallest_int_mode_for_size
19670         instead of smallest_mode_for_size.
19671         * combine.c (make_extraction): Likewise.
19672         * config/arc/arc.c (arc_expand_movmem): Likewise.
19673         * config/arm/arm.c (arm_expand_divmod_libfunc): Likewise.
19674         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
19675         * config/s390/s390.c (s390_expand_insv): Likewise.
19676         * config/sparc/sparc.c (assign_int_registers): Likewise.
19677         * config/spu/spu.c (spu_function_value): Likewise.
19678         (spu_function_arg): Likewise.
19679         * coverage.c (get_gcov_type): Likewise.
19680         (get_gcov_unsigned_t): Likewise.
19681         * dse.c (find_shift_sequence): Likewise.
19682         * expmed.c (store_bit_field_1): Likewise.
19683         * expr.c (convert_move): Likewise.
19684         (store_field): Likewise.
19685         * internal-fn.c (expand_arith_overflow): Likewise.
19686         * optabs-query.c (get_best_extraction_insn): Likewise.
19687         * optabs.c (expand_twoval_binop_libfunc): Likewise.
19688         * stor-layout.c (layout_type): Likewise.
19689         (initialize_sizetypes): Likewise.
19690         * targhooks.c (default_get_mask_mode): Likewise.
19691         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Likewise.
19693 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19694             Alan Hayward  <alan.hayward@arm.com>
19695             David Sherwood  <david.sherwood@arm.com>
19697         * machmode.h (opt_mode::else_blk): New function.
19698         (int_mode_for_mode): Declare.
19699         * stor-layout.c (int_mode_for_mode): Return an opt_scalar_int_mode.
19700         * builtins.c (expand_builtin_signbit): Adjust for new int_mode_for_mode
19701         return type.
19702         * cfgexpand.c (expand_debug_expr): Likewise.
19703         * combine.c (gen_lowpart_or_truncate): Likewise.
19704         (gen_lowpart_for_combine): Likewise.
19705         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Likewise.
19706         * config/avr/avr.c (avr_to_int_mode): Likewise.
19707         (avr_out_plus_1): Likewise.
19708         (avr_out_plus): Likewise.
19709         (avr_out_round): Likewise.
19710         * config/i386/i386.c (ix86_split_to_parts): Likewise.
19711         * config/s390/s390.c (s390_expand_vec_compare_cc): Likewise.
19712         (s390_expand_vcond): Likewise.
19713         * config/spu/spu.c (spu_split_immediate): Likewise.
19714         (spu_expand_mov): Likewise.
19715         * dse.c (get_stored_val): Likewise.
19716         * expmed.c (store_bit_field_1): Likewise.
19717         (convert_extracted_bit_field): Use int_mode_for_mode instead of
19718         int_mode_for_size.
19719         (extract_bit_field_1): Adjust for new int_mode_for_mode return type.
19720         (extract_low_bits): Likewise.
19721         * expr.c (emit_group_load_1): Likewise.  Separate out the BLKmode
19722         handling rather than repeating the check.
19723         (emit_group_store): Likewise.
19724         (emit_move_via_integer): Adjust for new int_mode_for_mode return type.
19725         * optabs.c (expand_absneg_bit): Likewise.
19726         (expand_copysign_absneg): Likewise.
19727         (expand_copysign_bit): Likewise.
19728         * tree-if-conv.c (ifcvt_can_use_mask_load_store): Likewise.
19729         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19730         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
19731         * var-tracking.c (prepare_call_arguments):  Likewise.
19732         * config/powerpcspe/powerpcspe.c (rs6000_do_expand_vec_perm): Use
19733         int_mode_for_mode instead of mode_for_size.
19734         * config/rs6000/rs6000.c (rs6000_do_expand_vec_perm): Likewise.
19736 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19737             Alan Hayward  <alan.hayward@arm.com>
19738             David Sherwood  <david.sherwood@arm.com>
19740         * machmode.h (int_mode_for_size): New function.
19741         * builtins.c (set_builtin_user_assembler_name): Use int_mode_for_size
19742         instead of mode_for_size.
19743         * calls.c (save_fixed_argument_area): Likewise.  Make use of BLKmode
19744         explicit.
19745         * combine.c (expand_field_assignment): Use int_mode_for_size
19746         instead of mode_for_size.
19747         (make_extraction): Likewise.
19748         (simplify_shift_const_1): Likewise.
19749         (simplify_comparison): Likewise.
19750         * dojump.c (do_jump): Likewise.
19751         * dwarf2out.c (mem_loc_descriptor): Likewise.
19752         * emit-rtl.c (init_derived_machine_modes): Likewise.
19753         * expmed.c (flip_storage_order): Likewise.
19754         (convert_extracted_bit_field): Likewise.
19755         * expr.c (copy_blkmode_from_reg): Likewise.
19756         * graphite-isl-ast-to-gimple.c (max_mode_int_precision): Likewise.
19757         * internal-fn.c (expand_mul_overflow): Likewise.
19758         * lower-subreg.c (simple_move): Likewise.
19759         * optabs-libfuncs.c (init_optabs): Likewise.
19760         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19761         * tree.c (vector_type_mode): Likewise.
19762         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
19763         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Likewise.
19764         * tree-vect-generic.c (expand_vector_parallel): Likewise.
19765         * tree-vect-stmts.c (vectorizable_load): Likewise.
19766         (vectorizable_store): Likewise.
19768 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19769             Alan Hayward  <alan.hayward@arm.com>
19770             David Sherwood  <david.sherwood@arm.com>
19772         * coretypes.h (pod_mode): New type.
19773         (scalar_int_mode_pod): New typedef.
19774         * machmode.h (pod_mode): New class.
19775         (int_n_data_t::m): Change type to scalar_int_mode_pod.
19776         * genmodes.c (emit_mode_int_n): Update accordingly.
19777         * lower-subreg.h (target_lower_subreg): Change type to
19778         scalar_int_mode_pod.
19779         * gdbhooks.py (build_pretty_printer): Handle pod_mode and
19780         scalar_int_mode_pod.
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         * config/powerpcspe/powerpcspe.h (rs6000_pmode): Change type from
19787         machine_mode to scalar_int_mode.
19788         * config/powerpcspe/powerpcspe.c (rs6000_pmode): Likewise.
19789         (rs6000_option_override_internal): Remove cast to int.
19790         * config/rs6000/rs6000.h (rs6000_pmode): Change type from
19791         machine_mode to scalar_int_mode.
19792         * config/rs6000/rs6000.c (rs6000_pmode): Likewise.
19793         (rs6000_option_override_internal): Remove cast to int.
19794         * config/s390/s390.h (Pmode): Remove cast to machine_mode.
19795         * config/epiphany/epiphany.h (RTX_OK_FOR_OFFSET_P): Add cast
19796         to machine_mode.
19797         * config/s390/s390.c (s390_expand_builtin): Likewise.
19798         * coretypes.h (scalar_int_mode): New type.
19799         (opt_scalar_int_mode): New typedef.
19800         * machmode.h (scalar_int_mode): New class.
19801         (scalar_int_mode::includes_p): New function.
19802         (byte_mode): Change type to scalar_int_mode.
19803         (word_mode): Likewise.
19804         (ptr_mode): Likewise.
19805         * emit-rtl.c (byte_mode): Likewise.
19806         (word_mode): Likewise.
19807         (ptr_mode): Likewise.
19808         (init_derived_machine_modes): Update accordingly.
19809         * genmodes.c (get_mode_class): Return scalar_int_mode for MODE_INT
19810         and MODE_PARTIAL_INT.
19811         * gdbhooks.py (build_pretty_printer): Handle scalar_int_mode and
19812         opt_scalar_int_mode.
19814 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19815             Alan Hayward  <alan.hayward@arm.com>
19816             David Sherwood  <david.sherwood@arm.com>
19818         * target.def (libgcc_floating_mode_supported_p): Take a
19819         scalar_float_mode.
19820         * doc/tm.texi: Regenerate.
19821         * targhooks.h (default_libgcc_floating_mode_supported_p): Take a
19822         scalar_float_mode.
19823         * targhooks.c (default_libgcc_floating_mode_supported_p): Likewise.
19824         * config/aarch64/aarch64.c (aarch64_libgcc_floating_mode_supported_p):
19825         Likewise.
19827 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19828             Alan Hayward  <alan.hayward@arm.com>
19829             David Sherwood  <david.sherwood@arm.com>
19831         * target.def (default_floatn_mode): Return an opt_scalar_float_mode.
19832         * doc/tm.texi: Regenerate.
19833         * config/arm/arm.c (arm_floatn_mode): Return an opt_scalar_float_mode.
19834         * config/powerpcspe/powerpcspe.c (rs6000_floatn_mode): Likewise.
19835         * config/rs6000/rs6000.c (rs6000_floatn_mode): Likewise.
19836         * targhooks.h (default_floatn_mode): Likewise.
19837         * targhooks.c (default_floatn_mode): Likewise.
19838         * tree.c (build_common_tree_nodes): Update accordingly.
19840 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19841             Alan Hayward  <alan.hayward@arm.com>
19842             David Sherwood  <david.sherwood@arm.com>
19844         * machmode.h (mode_iterator::start): Provide overload for opt_modes.
19845         (mode_iterator::iterate_p): Likewise.
19846         (mode_iterator::get_wider): Likewise.
19847         * expr.c (init_expr_target): Use opt_scalar_float_mode.
19849 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19850             Alan Hayward  <alan.hayward@arm.com>
19851             David Sherwood  <david.sherwood@arm.com>
19853         * coretypes.h (opt_scalar_float_mode): New typedef.
19854         * machmode.h (float_mode_for_size): New function.
19855         * emit-rtl.c (double_mode): Delete.
19856         (init_emit_once): Use float_mode_for_size.
19857         * stor-layout.c (layout_type): Likewise.
19858         * gdbhooks.py (build_pretty_printer): Handle opt_scalar_float_mode.
19860 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19861             Alan Hayward  <alan.hayward@arm.com>
19862             David Sherwood  <david.sherwood@arm.com>
19864         * output.h (assemble_real): Take a scalar_float_mode.
19865         * config/arm/arm.c (arm_assemble_integer): Update accordingly.
19866         * config/arm/arm.md (consttable_4): Likewise.
19867         (consttable_8): Likewise.
19868         (consttable_16): Likewise.
19869         * config/mips/mips.md (consttable_float): Likewise.
19870         * config/s390/s390.c (s390_output_pool_entry): Likewise.
19871         * varasm.c (assemble_real): Take a scalar_float_mode.
19872         (output_constant_pool_2): Update accordingly.
19873         (output_constant): Likewise.
19875 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19876             Alan Hayward  <alan.hayward@arm.com>
19877             David Sherwood  <david.sherwood@arm.com>
19879         * tree.h (SCALAR_FLOAT_TYPE_MODE): New macro.
19880         * builtins.c (expand_builtin_signbit): Use it instead of TYPE_MODE.
19881         * fold-const.c (fold_convert_const_real_from_fixed): Likewise.
19882         (native_encode_real): Likewise.
19883         (native_interpret_real): Likewise.
19884         * hsa-brig.c (emit_immediate_scalar_to_buffer): Likewise.
19885         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19887 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19888             Alan Hayward  <alan.hayward@arm.com>
19889             David Sherwood  <david.sherwood@arm.com>
19891         * optabs-libfuncs.c (gen_trunc_conv_libfunc): Use is_a
19892         <scalar_float_mode>.  Simplify.
19893         (gen_extend_conv_libfunc): Likewise.
19895 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19896             Alan Hayward  <alan.hayward@arm.com>
19897             David Sherwood  <david.sherwood@arm.com>
19899         * coretypes.h (scalar_float_mode): New type.
19900         * machmode.h (mode_traits::from_int): Use machine_mode if
19901         USE_ENUM_MODES is defined.
19902         (is_a): New function.
19903         (as_a): Likewise.
19904         (dyn_cast): Likewise.
19905         (scalar_float_mode): New class.
19906         (scalar_float_mode::includes_p): New function.
19907         (is_float_mode): Likewise.
19908         * gdbhooks.py (MachineModePrinter): New class.
19909         (build_pretty_printer): Use it for scalar_float_mode.
19910         * real.h (FLOAT_MODE_FORMAT): Use as_a <scalar_float_mode>.
19911         (format_helper::format_helper): Turn into a template.
19912         * genmodes.c (get_mode_class): New function.
19913         (emit_insn_modes_h): Give modes the class returned by get_mode_class,
19914         or machine_mode if none.
19915         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Use
19916         as_a <scalar_float_mode>.
19917         * dwarf2out.c (mem_loc_descriptor): Likewise.
19918         (insert_float): Likewise.
19919         (add_const_value_attribute): Likewise.
19920         * simplify-rtx.c (simplify_immed_subreg): Likewise.
19921         * optabs.c (expand_absneg_bit): Take a scalar_float_mode.
19922         (expand_unop): Update accordingly.
19923         (expand_abs_nojump): Likewise.
19924         (expand_copysign_absneg): Take a scalar_float_mode.
19925         (expand_copysign_bit): Likewise.
19926         (expand_copysign): Update accordingly.
19928 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19929             Alan Hayward  <alan.hayward@arm.com>
19930             David Sherwood  <david.sherwood@arm.com>
19932         * coretypes.h (opt_mode): New class.
19933         * machmode.h (opt_mode): Likewise.
19934         (opt_mode::else_void): New function.
19935         (opt_mode::require): Likewise.
19936         (opt_mode::exists): Likewise.
19937         (GET_MODE_WIDER_MODE): Turn into a function and return an opt_mode.
19938         (GET_MODE_2XWIDER_MODE): Likewise.
19939         (mode_iterator::get_wider): Update accordingly.
19940         (mode_iterator::get_2xwider): Likewise.
19941         (mode_iterator::get_known_wider): Likewise, turning into a template.
19942         * combine.c (make_extraction): Update use of GET_MODE_WIDER_MODE,
19943         forcing a wider mode to exist.
19944         * config/cr16/cr16.h (LONG_REG_P): Likewise.
19945         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
19946         * config/c6x/c6x.c (c6x_rtx_costs): Update use of
19947         GET_MODE_2XWIDER_MODE, forcing a wider mode to exist.
19948         * lower-subreg.c (init_lower_subreg): Likewise.
19949         * optabs-libfuncs.c (init_sync_libfuncs_1): Likewise, but not
19950         on the final iteration.
19951         * config/i386/i386.c (ix86_expand_set_or_movmem): Check whether
19952         a wider mode exists before asking for a move pattern.
19953         (get_mode_wider_vector): Update use of GET_MODE_WIDER_MODE,
19954         forcing a wider mode to exist.
19955         (expand_vselect_vconcat): Update use of GET_MODE_2XWIDER_MODE,
19956         returning false if no such mode exists.
19957         * config/ia64/ia64.c (expand_vselect_vconcat): Likewise.
19958         * config/mips/mips.c (mips_expand_vselect_vconcat): Likewise.
19959         * expmed.c (init_expmed_one_mode): Update use of GET_MODE_WIDER_MODE.
19960         Avoid checking for a MODE_INT if we already know the mode is not a
19961         SCALAR_INT_MODE_P.
19962         (extract_high_half): Update use of GET_MODE_WIDER_MODE,
19963         forcing a wider mode to exist.
19964         (expmed_mult_highpart_optab): Likewise.
19965         (expmed_mult_highpart): Likewise.
19966         * expr.c (expand_expr_real_2): Update use of GET_MODE_WIDER_MODE,
19967         using else_void.
19968         * lto-streamer-in.c (lto_input_mode_table): Likewise.
19969         * optabs-query.c (find_widening_optab_handler_and_mode): Likewise.
19970         * stor-layout.c (bit_field_mode_iterator::next_mode): Likewise.
19971         * internal-fn.c (expand_mul_overflow): Update use of
19972         GET_MODE_2XWIDER_MODE.
19973         * omp-low.c (omp_clause_aligned_alignment): Likewise.
19974         * tree-ssa-math-opts.c (convert_mult_to_widen): Update use of
19975         GET_MODE_WIDER_MODE.
19976         (convert_plusminus_to_widen): Likewise.
19977         * tree-switch-conversion.c (array_value_type): Likewise.
19978         * var-tracking.c (emit_note_insn_var_location): Likewise.
19979         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19980         Return false inside rather than outside the loop if no wider mode
19981         exists
19982         * optabs.c (expand_binop): Update use of GET_MODE_WIDER_MODE
19983         and GET_MODE_2XWIDER_MODE
19984         (can_compare_p): Use else_void.
19985         * gdbhooks.py (OptMachineModePrinter): New class.
19986         (build_pretty_printer): Use it for opt_mode.
19988 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19989             Alan Hayward  <alan.hayward@arm.com>
19990             David Sherwood  <david.sherwood@arm.com>
19992         * tree-switch-conversion.c (array_value_type): Only read TYPE_MODE
19993         once.  Use get_narrowest_mode instead of GET_CLASS_NARROWEST_MODE.
19995 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19996             Alan Hayward  <alan.hayward@arm.com>
19997             David Sherwood  <david.sherwood@arm.com>
19999         * machmode.h (mode_traits): New structure.
20000         (get_narrowest_mode): New function.
20001         (mode_iterator::start): Likewise.
20002         (mode_iterator::iterate_p): Likewise.
20003         (mode_iterator::get_wider): Likewise.
20004         (mode_iterator::get_known_wider): Likewise.
20005         (mode_iterator::get_2xwider): Likewise.
20006         (FOR_EACH_MODE_IN_CLASS): New mode iterator.
20007         (FOR_EACH_MODE): Likewise.
20008         (FOR_EACH_MODE_FROM): Likewise.
20009         (FOR_EACH_MODE_UNTIL): Likewise.
20010         (FOR_EACH_WIDER_MODE): Likewise.
20011         (FOR_EACH_2XWIDER_MODE): Likewise.
20012         * builtins.c (expand_builtin_strlen): Use new mode iterators.
20013         * combine.c (simplify_comparison): Likewise
20014         * config/i386/i386.c (type_natural_mode): Likewise.
20015         * cse.c (cse_insn): Likewise.
20016         * dse.c (find_shift_sequence): Likewise.
20017         * emit-rtl.c (init_derived_machine_modes): Likewise.
20018         (init_emit_once): Likewise.
20019         * explow.c (hard_function_value): Likewise.
20020         * expmed.c (extract_fixed_bit_field_1): Likewise.
20021         (extract_bit_field_1): Likewise.
20022         (expand_divmod): Likewise.
20023         (emit_store_flag_1): Likewise.
20024         * expr.c (init_expr_target): Likewise.
20025         (convert_move): Likewise.
20026         (alignment_for_piecewise_move): Likewise.
20027         (widest_int_mode_for_size): Likewise.
20028         (emit_block_move_via_movmem): Likewise.
20029         (copy_blkmode_to_reg): Likewise.
20030         (set_storage_via_setmem): Likewise.
20031         (compress_float_constant): Likewise.
20032         * omp-low.c (omp_clause_aligned_alignment): Likewise.
20033         * optabs-query.c (get_best_extraction_insn): Likewise.
20034         * optabs.c (expand_binop): Likewise.
20035         (expand_twoval_unop): Likewise.
20036         (expand_twoval_binop): Likewise.
20037         (widen_leading): Likewise.
20038         (widen_bswap): Likewise.
20039         (expand_parity): Likewise.
20040         (expand_unop): Likewise.
20041         (prepare_cmp_insn): Likewise.
20042         (prepare_float_lib_cmp): Likewise.
20043         (expand_float): Likewise.
20044         (expand_fix): Likewise.
20045         (expand_sfix_optab): Likewise.
20046         * postreload.c (move2add_use_add2_insn): Likewise.
20047         * reg-stack.c (reg_to_stack): Likewise.
20048         * reginfo.c (choose_hard_reg_mode): Likewise.
20049         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
20050         * stor-layout.c (mode_for_size): Likewise.
20051         (smallest_mode_for_size): Likewise.
20052         (mode_for_vector): Likewise.
20053         (finish_bitfield_representative): Likewise.
20054         * tree-ssa-math-opts.c (target_supports_divmod_p): Likewise.
20055         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
20056         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
20057         * var-tracking.c (prepare_call_arguments): Likewise.
20059 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20060             Alan Hayward  <alan.hayward@arm.com>
20061             David Sherwood  <david.sherwood@arm.com>
20063         * genconditions.c (write_header): Add a "#define USE_ENUM_MODES".
20064         * genmodes.c (emit_insn_modes_h): Define FOOmode to E_FOOmode if
20065         USE_ENUM_MODES is defined and to ((void) 0, E_FOOmode) otherwise.
20066         * machmode.h (mode_size): Move earlier in file.
20067         (mode_precision): Likewise.
20068         (mode_inner): Likewise.
20069         (mode_nunits): Likewise.
20070         (mode_unit_size): Likewise.
20071         (unit_unit_precision): Likewise.
20072         (mode_wider): Likewise.
20073         (mode_2xwider): Likewise.
20074         (machine_mode): New class.
20075         (mode_to_bytes): New function.
20076         (mode_to_bits): Likewise.
20077         (mode_to_precision): Likewise.
20078         (mode_to_inner): Likewise.
20079         (mode_to_unit_size): Likewise.
20080         (mode_to_unit_precision): Likewise.
20081         (mode_to_nunits): Likewise.
20082         (GET_MODE_SIZE): Use mode_to_bytes.
20083         (GET_MODE_BITSIZE): Use mode_to_bits.
20084         (GET_MODE_PRECISION): Use mode_to_precision.
20085         (GET_MODE_INNER): Use mode_to_inner.
20086         (GET_MODE_UNIT_SIZE): Use mode_to_unit_size.
20087         (GET_MODE_UNIT_PRECISION): Use mode_to_unit_precision.
20088         (GET_MODE_NUNITS): Use mode_to_nunits.
20089         * system.h (ALWAYS_INLINE): New macro.
20090         * config/powerpcspe/powerpcspe-c.c
20091         (altivec_resolve_overloaded_builtin): Use machine_mode instead of
20092         int for arg1_mode and arg2_mode.
20094 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20095             Alan Hayward  <alan.hayward@arm.com>
20096             David Sherwood  <david.sherwood@arm.com>
20098         * config/aarch64/aarch64-builtins.c (aarch64_simd_builtin_std_type):
20099         Prefix mode names with E_ in case statements.
20100         * config/aarch64/aarch64-elf.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20101         * config/aarch64/aarch64.c (aarch64_split_simd_combine): Likewise.
20102         (aarch64_split_simd_move): Likewise.
20103         (aarch64_gen_storewb_pair): Likewise.
20104         (aarch64_gen_loadwb_pair): Likewise.
20105         (aarch64_gen_store_pair): Likewise.
20106         (aarch64_gen_load_pair): Likewise.
20107         (aarch64_get_condition_code_1): Likewise.
20108         (aarch64_constant_pool_reload_icode): Likewise.
20109         (get_rsqrte_type): Likewise.
20110         (get_rsqrts_type): Likewise.
20111         (get_recpe_type): Likewise.
20112         (get_recps_type): Likewise.
20113         (aarch64_gimplify_va_arg_expr): Likewise.
20114         (aarch64_simd_container_mode): Likewise.
20115         (aarch64_emit_load_exclusive): Likewise.
20116         (aarch64_emit_store_exclusive): Likewise.
20117         (aarch64_expand_compare_and_swap): Likewise.
20118         (aarch64_gen_atomic_cas): Likewise.
20119         (aarch64_emit_bic): Likewise.
20120         (aarch64_emit_atomic_swap): Likewise.
20121         (aarch64_emit_atomic_load_op): Likewise.
20122         (aarch64_evpc_trn): Likewise.
20123         (aarch64_evpc_uzp): Likewise.
20124         (aarch64_evpc_zip): Likewise.
20125         (aarch64_evpc_ext): Likewise.
20126         (aarch64_evpc_rev): Likewise.
20127         (aarch64_evpc_dup): Likewise.
20128         (aarch64_gen_ccmp_first): Likewise.
20129         (aarch64_gen_ccmp_next): Likewise.
20130         * config/alpha/alpha.c (alpha_scalar_mode_supported_p): Likewise.
20131         (alpha_emit_xfloating_libcall): Likewise.
20132         (emit_insxl): Likewise.
20133         (alpha_arg_type): Likewise.
20134         * config/arc/arc.c (arc_vector_mode_supported_p): Likewise.
20135         (arc_preferred_simd_mode): Likewise.
20136         (arc_secondary_reload): Likewise.
20137         (get_arc_condition_code): Likewise.
20138         (arc_print_operand): Likewise.
20139         (arc_legitimate_constant_p): Likewise.
20140         * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20141         * config/arc/arc.md (casesi_load): Likewise.
20142         (casesi_compact_jump): Likewise.
20143         * config/arc/predicates.md (proper_comparison_operator): Likewise.
20144         (cc_use_register): Likewise.
20145         * config/arm/aout.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20146         * config/arm/arm-builtins.c (arm_simd_builtin_std_type): Likewise.
20147         (arm_init_iwmmxt_builtins): Likewise.
20148         * config/arm/arm.c (thumb1_size_rtx_costs): Likewise.
20149         (neon_expand_vector_init): Likewise.
20150         (arm_attr_length_move_neon): Likewise.
20151         (maybe_get_arm_condition_code): Likewise.
20152         (arm_emit_vector_const): Likewise.
20153         (arm_preferred_simd_mode): Likewise.
20154         (arm_output_iwmmxt_tinsr): Likewise.
20155         (thumb1_output_casesi): Likewise.
20156         (thumb2_output_casesi): Likewise.
20157         (arm_emit_load_exclusive): Likewise.
20158         (arm_emit_store_exclusive): Likewise.
20159         (arm_expand_compare_and_swap): Likewise.
20160         (arm_evpc_neon_vuzp): Likewise.
20161         (arm_evpc_neon_vzip): Likewise.
20162         (arm_evpc_neon_vrev): Likewise.
20163         (arm_evpc_neon_vtrn): Likewise.
20164         (arm_evpc_neon_vext): Likewise.
20165         (arm_validize_comparison): Likewise.
20166         * config/arm/neon.md (neon_vc<cmp_op><mode>): Likewise.
20167         * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Likewise.
20168         * config/avr/avr.c (avr_rtx_costs_1): Likewise.
20169         * config/c6x/c6x.c (c6x_vector_mode_supported_p): Likewise.
20170         (c6x_preferred_simd_mode): Likewise.
20171         * config/epiphany/epiphany.c (get_epiphany_condition_code): Likewise.
20172         (epiphany_rtx_costs): Likewise.
20173         * config/epiphany/predicates.md (proper_comparison_operator):
20174         Likewise.
20175         * config/frv/frv.c (condexec_memory_operand): Likewise.
20176         (frv_emit_move): Likewise.
20177         (output_move_single): Likewise.
20178         (output_condmove_single): Likewise.
20179         (frv_hard_regno_mode_ok): Likewise.
20180         (frv_matching_accg_mode): Likewise.
20181         * config/h8300/h8300.c (split_adds_subs): Likewise.
20182         (h8300_rtx_costs): Likewise.
20183         (h8300_print_operand): Likewise.
20184         (compute_mov_length): Likewise.
20185         (output_logical_op): Likewise.
20186         (compute_logical_op_length): Likewise.
20187         (compute_logical_op_cc): Likewise.
20188         (h8300_shift_needs_scratch_p): Likewise.
20189         (output_a_shift): Likewise.
20190         (compute_a_shift_length): Likewise.
20191         (compute_a_shift_cc): Likewise.
20192         (expand_a_rotate): Likewise.
20193         (output_a_rotate): Likewise.
20194         * config/i386/i386.c (classify_argument): Likewise.
20195         (function_arg_advance_32): Likewise.
20196         (function_arg_32): Likewise.
20197         (function_arg_64): Likewise.
20198         (function_value_64): Likewise.
20199         (ix86_gimplify_va_arg): Likewise.
20200         (ix86_legitimate_constant_p): Likewise.
20201         (put_condition_code): Likewise.
20202         (split_double_mode): Likewise.
20203         (ix86_avx256_split_vector_move_misalign): Likewise.
20204         (ix86_expand_vector_logical_operator): Likewise.
20205         (ix86_split_idivmod): Likewise.
20206         (ix86_expand_adjust_ufix_to_sfix_si): Likewise.
20207         (ix86_build_const_vector): Likewise.
20208         (ix86_build_signbit_mask): Likewise.
20209         (ix86_match_ccmode): Likewise.
20210         (ix86_cc_modes_compatible): Likewise.
20211         (ix86_expand_branch): Likewise.
20212         (ix86_expand_sse_cmp): Likewise.
20213         (ix86_expand_sse_movcc): Likewise.
20214         (ix86_expand_int_sse_cmp): Likewise.
20215         (ix86_expand_vec_perm_vpermi2): Likewise.
20216         (ix86_expand_vec_perm): Likewise.
20217         (ix86_expand_sse_unpack): Likewise.
20218         (ix86_expand_int_addcc): Likewise.
20219         (ix86_split_to_parts): Likewise.
20220         (ix86_vectorize_builtin_gather): Likewise.
20221         (ix86_vectorize_builtin_scatter): Likewise.
20222         (avx_vpermilp_parallel): Likewise.
20223         (inline_memory_move_cost): Likewise.
20224         (ix86_tieable_integer_mode_p): Likewise.
20225         (x86_maybe_negate_const_int): Likewise.
20226         (ix86_expand_vector_init_duplicate): Likewise.
20227         (ix86_expand_vector_init_one_nonzero): Likewise.
20228         (ix86_expand_vector_init_one_var): Likewise.
20229         (ix86_expand_vector_init_concat): Likewise.
20230         (ix86_expand_vector_init_interleave): Likewise.
20231         (ix86_expand_vector_init_general): Likewise.
20232         (ix86_expand_vector_set): Likewise.
20233         (ix86_expand_vector_extract): Likewise.
20234         (emit_reduc_half): Likewise.
20235         (ix86_emit_i387_round): Likewise.
20236         (ix86_mangle_type): Likewise.
20237         (ix86_expand_round_sse4): Likewise.
20238         (expand_vec_perm_blend): Likewise.
20239         (canonicalize_vector_int_perm): Likewise.
20240         (ix86_expand_vec_one_operand_perm_avx512): Likewise.
20241         (expand_vec_perm_1): Likewise.
20242         (expand_vec_perm_interleave3): Likewise.
20243         (expand_vec_perm_even_odd_pack): Likewise.
20244         (expand_vec_perm_even_odd_1): Likewise.
20245         (expand_vec_perm_broadcast_1): Likewise.
20246         (ix86_vectorize_vec_perm_const_ok): Likewise.
20247         (ix86_expand_vecop_qihi): Likewise.
20248         (ix86_expand_mul_widen_hilo): Likewise.
20249         (ix86_expand_sse2_abs): Likewise.
20250         (ix86_expand_pextr): Likewise.
20251         (ix86_expand_pinsr): Likewise.
20252         (ix86_preferred_simd_mode): Likewise.
20253         (ix86_simd_clone_compute_vecsize_and_simdlen): Likewise.
20254         * config/i386/sse.md (*andnot<mode>3): Likewise.
20255         (<mask_codefor><code><mode>3<mask_name>): Likewise.
20256         (*<code><mode>3): Likewise.
20257         * config/ia64/ia64.c (ia64_expand_vecint_compare): Likewise.
20258         (ia64_expand_atomic_op): Likewise.
20259         (ia64_arg_type): Likewise.
20260         (ia64_mode_to_int): Likewise.
20261         (ia64_scalar_mode_supported_p): Likewise.
20262         (ia64_vector_mode_supported_p): Likewise.
20263         (expand_vec_perm_broadcast): Likewise.
20264         * config/iq2000/iq2000.c (iq2000_move_1word): Likewise.
20265         (iq2000_function_arg_advance): Likewise.
20266         (iq2000_function_arg): Likewise.
20267         * config/m32c/m32c.c (m32c_preferred_reload_class): Likewise.
20268         * config/m68k/m68k.c (output_dbcc_and_branch): Likewise.
20269         (m68k_libcall_value): Likewise.
20270         (m68k_function_value): Likewise.
20271         (sched_attr_op_type): Likewise.
20272         * config/mcore/mcore.c (mcore_output_move): Likewise.
20273         * config/microblaze/microblaze.c (microblaze_function_arg_advance):
20274         Likewise.
20275         (microblaze_function_arg): Likewise.
20276         * config/mips/mips.c (mips16_build_call_stub): Likewise.
20277         (mips_print_operand): Likewise.
20278         (mips_mode_ok_for_mov_fmt_p): Likewise.
20279         (mips_vector_mode_supported_p): Likewise.
20280         (mips_preferred_simd_mode): Likewise.
20281         (mips_expand_vpc_loongson_even_odd): Likewise.
20282         (mips_expand_vec_unpack): Likewise.
20283         (mips_expand_vi_broadcast): Likewise.
20284         (mips_expand_vector_init): Likewise.
20285         (mips_expand_vec_reduc): Likewise.
20286         (mips_expand_msa_cmp): Likewise.
20287         * config/mips/mips.md (casesi_internal_mips16_<mode>): Likewise.
20288         * config/mn10300/mn10300.c (mn10300_print_operand): Likewise.
20289         (cc_flags_for_mode): Likewise.
20290         * config/msp430/msp430.c (msp430_print_operand): Likewise.
20291         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Likewise.
20292         (nds32_output_casesi_pc_relative): Likewise.
20293         * config/nds32/nds32.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20294         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Likewise.
20295         (nvptx_gen_unpack): Likewise.
20296         (nvptx_gen_pack): Likewise.
20297         (nvptx_gen_shuffle): Likewise.
20298         (nvptx_gen_wcast): Likewise.
20299         (nvptx_preferred_simd_mode): Likewise.
20300         * config/pa/pa.c (pa_secondary_reload): Likewise.
20301         * config/pa/predicates.md (base14_operand): Likewise.
20302         * config/powerpcspe/powerpcspe-c.c
20303         (altivec_resolve_overloaded_builtin): Likewise.
20304         * config/powerpcspe/powerpcspe.c (rs6000_setup_reg_addr_masks):
20305         Likewise.
20306         (rs6000_preferred_simd_mode): Likewise.
20307         (output_vec_const_move): Likewise.
20308         (rs6000_expand_vector_extract): Likewise.
20309         (rs6000_split_vec_extract_var): Likewise.
20310         (reg_offset_addressing_ok_p): Likewise.
20311         (rs6000_legitimate_offset_address_p): Likewise.
20312         (rs6000_legitimize_address): Likewise.
20313         (rs6000_emit_set_const): Likewise.
20314         (rs6000_const_vec): Likewise.
20315         (rs6000_emit_move): Likewise.
20316         (spe_build_register_parallel): Likewise.
20317         (rs6000_darwin64_record_arg_recurse): Likewise.
20318         (swap_selector_for_mode): Likewise.
20319         (spe_init_builtins): Likewise.
20320         (paired_init_builtins): Likewise.
20321         (altivec_init_builtins): Likewise.
20322         (do_load_for_compare): Likewise.
20323         (rs6000_generate_compare): Likewise.
20324         (rs6000_expand_float128_convert): Likewise.
20325         (emit_load_locked): Likewise.
20326         (emit_store_conditional): Likewise.
20327         (rs6000_output_function_epilogue): Likewise.
20328         (rs6000_handle_altivec_attribute): Likewise.
20329         (rs6000_function_value): Likewise.
20330         (emit_fusion_gpr_load): Likewise.
20331         (emit_fusion_p9_load): Likewise.
20332         (emit_fusion_p9_store): Likewise.
20333         * config/powerpcspe/predicates.md (easy_fp_constant): Likewise.
20334         (fusion_gpr_mem_load): Likewise.
20335         (fusion_addis_mem_combo_load): Likewise.
20336         (fusion_addis_mem_combo_store): Likewise.
20337         * config/rs6000/predicates.md (easy_fp_constant): Likewise.
20338         (fusion_gpr_mem_load): Likewise.
20339         (fusion_addis_mem_combo_load): Likewise.
20340         (fusion_addis_mem_combo_store): Likewise.
20341         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
20342         Likewise.
20343         * config/rs6000/rs6000-string.c (do_load_for_compare): Likewise.
20344         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Likewise.
20345         (rs6000_preferred_simd_mode): Likewise.
20346         (output_vec_const_move): Likewise.
20347         (rs6000_expand_vector_extract): Likewise.
20348         (rs6000_split_vec_extract_var): Likewise.
20349         (reg_offset_addressing_ok_p): Likewise.
20350         (rs6000_legitimate_offset_address_p): Likewise.
20351         (rs6000_legitimize_address): Likewise.
20352         (rs6000_emit_set_const): Likewise.
20353         (rs6000_const_vec): Likewise.
20354         (rs6000_emit_move): Likewise.
20355         (rs6000_darwin64_record_arg_recurse): Likewise.
20356         (swap_selector_for_mode): Likewise.
20357         (paired_init_builtins): Likewise.
20358         (altivec_init_builtins): Likewise.
20359         (rs6000_expand_float128_convert): Likewise.
20360         (emit_load_locked): Likewise.
20361         (emit_store_conditional): Likewise.
20362         (rs6000_output_function_epilogue): Likewise.
20363         (rs6000_handle_altivec_attribute): Likewise.
20364         (rs6000_function_value): Likewise.
20365         (emit_fusion_gpr_load): Likewise.
20366         (emit_fusion_p9_load): Likewise.
20367         (emit_fusion_p9_store): Likewise.
20368         * config/rx/rx.c (rx_gen_move_template): Likewise.
20369         (flags_from_mode): Likewise.
20370         * config/s390/predicates.md (s390_alc_comparison): Likewise.
20371         (s390_slb_comparison): Likewise.
20372         * config/s390/s390.c (s390_handle_vectorbool_attribute): Likewise.
20373         (s390_vector_mode_supported_p): Likewise.
20374         (s390_cc_modes_compatible): Likewise.
20375         (s390_match_ccmode_set): Likewise.
20376         (s390_canonicalize_comparison): Likewise.
20377         (s390_emit_compare_and_swap): Likewise.
20378         (s390_branch_condition_mask): Likewise.
20379         (s390_rtx_costs): Likewise.
20380         (s390_secondary_reload): Likewise.
20381         (__SECONDARY_RELOAD_CASE): Likewise.
20382         (s390_expand_cs): Likewise.
20383         (s390_preferred_simd_mode): Likewise.
20384         * config/s390/vx-builtins.md (vec_packsu_u<mode>): Likewise.
20385         * config/sh/sh.c (sh_print_operand): Likewise.
20386         (dump_table): Likewise.
20387         (sh_secondary_reload): Likewise.
20388         * config/sh/sh.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20389         * config/sh/sh.md (casesi_worker_1): Likewise.
20390         (casesi_worker_2): Likewise.
20391         * config/sparc/predicates.md (icc_comparison_operator): Likewise.
20392         (fcc_comparison_operator): Likewise.
20393         * config/sparc/sparc.c (sparc_expand_move): Likewise.
20394         (emit_soft_tfmode_cvt): Likewise.
20395         (sparc_preferred_simd_mode): Likewise.
20396         (output_cbranch): Likewise.
20397         (sparc_print_operand): Likewise.
20398         (sparc_expand_vec_perm_bmask): Likewise.
20399         (vector_init_bshuffle): Likewise.
20400         * config/spu/spu.c (spu_scalar_mode_supported_p): Likewise.
20401         (spu_vector_mode_supported_p): Likewise.
20402         (spu_expand_insv): Likewise.
20403         (spu_emit_branch_or_set): Likewise.
20404         (spu_handle_vector_attribute): Likewise.
20405         (spu_builtin_splats): Likewise.
20406         (spu_builtin_extract): Likewise.
20407         (spu_builtin_promote): Likewise.
20408         (spu_expand_sign_extend): Likewise.
20409         * config/tilegx/tilegx.c (tilegx_scalar_mode_supported_p): Likewise.
20410         (tilegx_simd_int): Likewise.
20411         * config/tilepro/tilepro.c (tilepro_scalar_mode_supported_p): Likewise.
20412         (tilepro_simd_int): Likewise.
20413         * config/v850/v850.c (const_double_split): Likewise.
20414         (v850_print_operand): Likewise.
20415         (ep_memory_offset): Likewise.
20416         * config/vax/vax.c (vax_rtx_costs): Likewise.
20417         (vax_output_int_move): Likewise.
20418         (vax_output_int_add): Likewise.
20419         (vax_output_int_subtract): Likewise.
20420         * config/visium/predicates.md (visium_branch_operator): Likewise.
20421         * config/visium/visium.c (rtx_ok_for_offset_p): Likewise.
20422         (visium_print_operand_address): Likewise.
20423         * config/visium/visium.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20424         * config/xtensa/xtensa.c (xtensa_mem_offset): Likewise.
20425         (xtensa_expand_conditional_branch): Likewise.
20426         (xtensa_copy_incoming_a7): Likewise.
20427         (xtensa_output_literal): Likewise.
20428         * dfp.c (decimal_real_maxval): Likewise.
20429         * targhooks.c (default_libgcc_floating_mode_supported_p): Likewise.
20431 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20432             Alan Hayward  <alan.hayward@arm.com>
20433             David Sherwood  <david.sherwood@arm.com>
20435         * genmodes.c (mode_size_inline): Add an E_ prefix to mode names.
20436         (mode_nunits_inline): Likewise.
20437         (mode_inner_inline): Likewise.
20438         (mode_unit_size_inline): Likewise.
20439         (mode_unit_precision_inline): Likewise.
20440         (emit_insn_modes_h): Likewise.  Also emit a #define of the
20441         unprefixed name.
20442         (emit_mode_wider): Add an E_ prefix to mode names.
20443         (emit_mode_complex): Likewise.
20444         (emit_mode_inner): Likewise.
20445         (emit_mode_adjustments): Likewise.
20446         (emit_mode_int_n): Likewise.
20447         * config/aarch64/aarch64-builtins.c (v8qi_UP, v4hi_UP, v4hf_UP)
20448         (v2si_UP, v2sf_UP, v1df_UP, di_UP, df_UP, v16qi_UP, v8hi_UP, v8hf_UP)
20449         (v4si_UP, v4sf_UP, v2di_UP, v2df_UP, ti_UP, oi_UP, ci_UP, xi_UP)
20450         (si_UP, sf_UP, hi_UP, hf_UP, qi_UP): Likewise.
20451         (CRC32_BUILTIN, ENTRY): Likewise.
20452         * config/aarch64/aarch64.c (aarch64_push_regs): Likewise.
20453         (aarch64_pop_regs): Likewise.
20454         (aarch64_process_components): Likewise.
20455         * config/alpha/alpha.c (alpha_emit_conditional_move): Likewise.
20456         * config/arm/arm-builtins.c (v8qi_UP, v4hi_UP, v4hf_UP, v2si_UP)
20457         (v2sf_UP, di_UP, v16qi_UP, v8hi_UP, v8hf_UP, v4si_UP, v4sf_UP)
20458         (v2di_UP, ti_UP, ei_UP, oi_UP, hf_UP, si_UP, void_UP): Likewise.
20459         * config/arm/arm.c (arm_init_libfuncs): Likewise.
20460         * config/i386/i386-builtin-types.awk (ix86_builtin_type_vect_mode):
20461         Likewise.
20462         * config/i386/i386-builtin.def (pcmpestr): Likewise.
20463         (pcmpistr): Likewise.
20464         * config/microblaze/microblaze.c (double_memory_operand): Likewise.
20465         * config/mmix/mmix.c (mmix_output_condition): Likewise.
20466         * config/powerpcspe/powerpcspe.c (rs6000_init_hard_regno_mode_ok):
20467         Likewise.
20468         * config/rl78/rl78.c (mduc_regs): Likewise.
20469         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
20470         (htm_expand_builtin): Likewise.
20471         * config/sh/sh.h (REGISTER_NATURAL_MODE): Likewise.
20472         * config/sparc/sparc.c (emit_save_or_restore_regs): Likewise.
20473         * config/xtensa/xtensa.c (print_operand): Likewise.
20474         * expmed.h (NUM_MODE_PARTIAL_INT): Likewise.
20475         (NUM_MODE_VECTOR_INT): Likewise.
20476         * genoutput.c (null_operand): Likewise.
20477         (output_operand_data): Likewise.
20478         * genrecog.c (print_parameter_value): Likewise.
20479         * lra.c (debug_operand_data): Likewise.
20481 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20483         * dwarf2out.c (scompare_loc_descriptor_wide)
20484         (scompare_loc_descriptor_narrow): New functions, split out from...
20485         (scompare_loc_descriptor): ...here.
20486         * expmed.c (emit_store_flag_int): New function, split out from...
20487         (emit_store_flag): ...here.
20489 2017-08-30  Richard Biener  <rguenther@suse.de>
20491         * dwarf2out.c (dwarf2out_finish): Remove setting AT_pubnames.
20492         (dwarf2out_early_finish): Move setting of AT_pubnames from
20493         early debug output to early finish.
20495 2017-08-30  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
20497         * gcc/config/msp430/msp430.h: Pass -mcode/data-region to the linker
20498         and -mdata-region to the assembler.
20500 2017-08-30  Richard Biener  <rguenther@suse.de>
20502         * dwarf2out.c (add_dwarf_attr): Check we don't add duplicate
20503         attributes.
20504         (gen_subprogram_die): Add DW_AT_object_pointer only early.
20505         (dwarf2out_early_global_decl): Only generate a DIE for the
20506         abstract origin if it doesn't already exist or is a declaration DIE.
20507         (resolve_addr): Do not add the linkage name twice when
20508         generating a stub DIE for the DW_TAG_GNU_call_site target.
20510 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20512         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
20513         Use machine_mode rather than int for arg1_mode.
20515 2017-08-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
20517         PR target/82015
20518         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Insure
20519         that the second argument of the built-in functions to unpack
20520         128-bit scalar types to 64-bit values is 0 or 1.  Change to use a
20521         switch statement instead a lot of if statements.
20522         * config/rs6000/rs6000.md (unpack<mode>, FMOVE128_VSX iterator):
20523         Allow 64-bit values to be in Altivec registers as well as
20524         traditional floating point registers.
20525         (pack<mode>, FMOVE128_VSX iterator): Likewise.
20527 2017-08-29  Alexander Monakov  <amonakov@ispras.ru>
20529         * ira-costs.c (record_address_regs): Handle both operands of PLUS for
20530         MAX_REGS_PER_ADDRESS == 1.
20532 2017-08-29  Uros Bizjak  <ubizjak@gmail.com>
20534         * config/i386/i386.opt (flag_fentry): Do not init to -1.
20535         * config/i386/i386.c (ix86_option_override_internal): Simplify
20536         setting of opts->x_flag_entry.
20538 2017-08-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20539             Jakub Jelinek  <jakub@redhat.com>
20540             Richard Biener  <rguenther@suse.de>
20542         PR tree-optimization/81503
20543         * gimple-ssa-strength-reduction.c (replace_mult_candidate): Ensure
20544         folded constant fits in the target type; reorder tests for clarity.
20546 2017-08-29  Martin Liska  <mliska@suse.cz>
20548         * passes.def: Include pass_lower_switch.
20549         * stmt.c (dump_case_nodes): Remove and move to
20550         tree-switch-conversion.
20551         (case_values_threshold): Likewise.
20552         (expand_switch_as_decision_tree_p): Likewise.
20553         (emit_case_decision_tree): Likewise.
20554         (expand_case): Likewise.
20555         (balance_case_nodes): Likewise.
20556         (node_has_low_bound): Likewise.
20557         (node_has_high_bound): Likewise.
20558         (node_is_bounded): Likewise.
20559         (emit_case_nodes): Likewise.
20560         (struct simple_case_node): New struct.
20561         (add_case_node): Remove.
20562         (emit_case_dispatch_table): Use vector instead of case_list.
20563         (reset_out_edges_aux): Remove.
20564         (compute_cases_per_edge): Likewise.
20565         (expand_case): Build list of simple_case_node.
20566         (expand_sjlj_dispatch_table): Use it.
20567         * tree-switch-conversion.c (struct case_node): Moved from
20568         stmt.c and adjusted.
20569         (emit_case_nodes): Likewise.
20570         (node_has_low_bound): Likewise.
20571         (node_has_high_bound): Likewise.
20572         (node_is_bounded): Likewise.
20573         (case_values_threshold): Likewise.
20574         (reset_out_edges_aux): Likewise.
20575         (compute_cases_per_edge): Likewise.
20576         (add_case_node): Likewise.
20577         (dump_case_nodes): Likewise.
20578         (balance_case_nodes): Likewise.
20579         (expand_switch_as_decision_tree_p): Likewise.
20580         (emit_jump): Likewise.
20581         (emit_case_decision_tree): Likewise.
20582         (try_switch_expansion): Likewise.
20583         (do_jump_if_equal): Likewise.
20584         (emit_cmp_and_jump_insns): Likewise.
20585         (fix_phi_operands_for_edge): New function.
20586         (record_phi_operand_mapping): Likewise.
20587         (class pass_lower_switch): New pass.
20588         (pass_lower_switch::execute): New function.
20589         (make_pass_lower_switch): Likewise.
20590         (conditional_probability):
20591         * timevar.def: Add TV_TREE_SWITCH_LOWERING.
20592         * tree-pass.h: Add make_pass_lower_switch.
20594 2017-08-29  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
20596         PR target/80993
20597         * gcc/config/msp430/msp430.c (msp430_attr): Mark interrupt
20598         handlers as used.
20600 2017-08-29  Richard Biener  <rguenther@suse.de>
20602         * dwarf2out.c (add_dwarf_attr): When checking is enabled verify
20603         we do not add a DW_AT_inline attribute twice.
20604         (gen_subprogram_die): Remove code setting DW_AT_inline on
20605         DECL_ABSTRACT_P nodes.
20607 2017-08-29  Richard Sandiford  <richard.sandiford@linaro.org>
20609         * gimplify.c (gimplify_call_expr): Copy the nothrow flag to
20610         calls to internal functions.
20611         (gimplify_modify_expr): Likewise.
20612         * tree-call-cdce.c (use_internal_fn): Likewise.
20613         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
20614         (convert_to_divmod): Set the nothrow flag.
20615         * tree-if-conv.c (predicate_mem_writes):  Likewise.
20616         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
20617         (vectorizable_call): Likewise.
20618         (vectorizable_store): Likewise.
20619         (vectorizable_load): Likewise.
20620         * tree-vect-patterns.c (vect_recog_pow_pattern): Likewise.
20621         (vect_recog_mask_conversion_pattern): Likewise.
20623 2017-08-29  Martin Liska  <mliska@suse.cz>
20625         PR other/39851
20626         * gcc.c (driver_handle_option): Add new argument.
20627         * opts-common.c (handle_option): Pass
20628         target_option_override_hook.
20629         * opts-global.c (lang_handle_option): Add new option.
20630         (set_default_handlers):  Add new argument.
20631         (decode_options): Likewise.
20632         * opts.c (target_handle_option): Likewise.
20633         (common_handle_option): Call target_option_override_hook.
20634         * opts.h (struct cl_option_handler_func): Add hook for
20635         target option override.
20636         (struct cl_option_handlers): Likewise.
20637         (set_default_handlers): Add new argument.
20638         (decode_options): Likewise.
20639         (common_handle_option): Likewise.
20640         (target_handle_option): Likewise.
20641         * toplev.c (toplev::main): Pass targetm.target_option.override
20642         hook.
20644 2017-08-29  Richard Biener  <rguenther@suse.de>
20645         Dominik Infuehr <dominik.infuehr@theobroma-systems.com>
20647         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Properly confine
20648         life to the active subtree.
20650 2017-08-28  Jeff Law  <law@redhat.com>
20652         * tree-ssa-dom.c (edge_info::record_simple_equiv): Call
20653         derive_equivalences.
20654         (derive_equivalences_from_bit_ior, record_temporary_equivalences):
20655         Code moved into....
20656         (edge_info::derive_equivalences): New private member function
20658         * tree-ssa-dom.c (class edge_info): Changed from a struct
20659         to a class.  Add ctor/dtor, methods and data members.
20660         (edge_info::edge_info): Renamed from allocate_edge_info.
20661         Initialize additional members.
20662         (edge_info::~edge_info): New.
20663         (free_dom_edge_info): Delete the edge info.
20664         (record_edge_info): Use new class & associated member functions.
20665         Tighten forms for testing for edge equivalences.
20666         (record_temporary_equivalences): Iterate over the simple
20667         equivalences rather than assuming there's only one per edge.
20668         (cprop_into_successor_phis): Iterate over the simple
20669         equivalences rather than assuming there's only one per edge.
20670         (optimize_stmt): Use operand_equal_p rather than pointer
20671         equality for mini-DSE code.
20673 2017-08-28  Nathan Sidwell  <nathan@acm.org>
20675         * gcc.c (execute): Fold SIGPIPE handling into switch
20676         statement.  Adjust internal error message.
20678 2017-08-28  Richard Biener  <rguenther@suse.de>
20680         PR debug/81993
20681         * dwarf2out.c (gen_remaining_tmpl_value_param_die_attributes):
20682         Do nothing for removed DIEs.
20684 2017-08-28  Richard Biener  <rguenther@suse.de>
20686         PR tree-optimization/81977
20687         * tree-ssa-sccvn.c (vn_reference_lookup_3): Fix look through
20688         memcpy.
20690 2017-08-28  Alexander Monakov  <amonakov@ispras.ru>
20692         PR target/80640
20693         * doc/md.texi (mem_thread_fence): Remove mention of mode.  Rewrite.
20694         * optabs.c (expand_mem_thread_fence): Emit a compiler barrier when
20695         using targetm.gen_mem_thread_fence.
20697 2017-08-27  Uros Bizjak  <ubizjak@gmail.com>
20699         PR target/81995
20700         * config/i386/i386.md (*<btsc><mode>): Change operand 2
20701         predicate to register_operand.  Reorder operands.
20702         (*btr<mode>): Ditto.
20703         (*<btsc><mode>_mask): Change operand 3 predicate to register_operand.
20704         (*btr<mode>_mask): Ditto.
20706 2017-08-25  Steven Munroe  <munroesj@gcc.gnu.org>
20708         * config.gcc (powerpc*-*-*): Add xmmintrin.h and mm_malloc.h.
20709         * config/rs6000/xmmintrin.h: New file.
20710         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include xmmintrin.h.
20712 2017-08-25  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20714         PR target/81504
20715         * config/rs6000/rs6000-p8swap.c (find_alignment_op): Add reference
20716         parameter and_insn and return it.
20717         (recombine_lvx_pattern): Insert a copy to ensure availability of
20718         the base register of the copied masking operation at the point of
20719         the instruction replacement.
20720         (recombine_stvx_pattern): Likewise.
20722 2017-08-25  Michael Meissner  <meissner@linux.vnet.ibm.com>
20724         * config/rs6000/rs6000.opt (-mpower9-dform-scalar): Delete
20725         undocumented switches.
20726         (-mpower9-dform-vector): Likewise.
20727         (-mpower9-dform): Likewise.
20728         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
20729         comments to delete references to -mpower9-dform* switches.
20730         * config/rs6000/predicates.md (vsx_quad_dform_memory_operand):
20731         Delete reference to -mpower9-dform* switches, test for
20732         -mpower9-vector instead.
20733         * config/rs6000/rs6000-cpus.def (ISA_3_0_MASKS_SERVER): Likewise.
20734         (OTHER_P9_VECTOR_MASKS): Likewise.
20735         (POWERPC_MASKS): Likewise.
20736         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Change
20737         tests against -mpower9-dform* to -mpower9-vector.  Delete code
20738         that checked for -mpower9-dform* consistancy with other options.
20739         Add test for -mpower9-misc to enable other power9 switches.
20740         (rs6000_init_hard_regno_mode_ok): Likewise.
20741         (rs6000_option_override_internal): Likewise.
20742         (rs6000_emit_prologue): Likewise.
20743         (rs6000_emit_epilogue): Likewise.
20744         (rs6000_opt_masks): Delete -mpower9-dform-{scalar,vector}.
20745         (rs6000_disable_incompatiable_switches): Delete -mpower9-dform.
20746         (emit_fusion_p9_load): Change tests for -mpower9-dform-scalar
20747         -mpower9-vector.
20748         (emit_fusion_p9_store): Likewise.
20749         * config/rs6000/rs6000.h (TARGET_P9_DFORM_SCALAR): Delete
20750         resetting these macros if the assembler does not support ISA 3.0
20751         instructions.
20752         (TARGET_P9_DFORM_VECTOR): Likewise.
20753         * config/rs6000/rs6000.md (peepholes to optimize altivec memory):
20754         Change to use -mpower9-vector instead of -mpower9-dform-scalar.
20756 2017-08-25  Alan Modra  <amodra@gmail.com>
20758         PR rtl-optimization/81747
20759         * cse.c (cse_extended_basic_block): Don't attempt to record
20760         equivalences for degenerate conditional branches that branch
20761         to their fall-through.
20763 2017-08-24  Martin Sebor  <msebor@redhat.com>
20765         PR middle-end/81908
20766         * gimple-fold.c (size_must_be_zero_p): New function.
20767         (gimple_fold_builtin_memory_op): Call it.
20769 2017-08-24  Steven Munroe  <munroesj@gcc.gnu.org>
20771         * config/rs6000/mm_malloc.h: New file.
20773 2017-08-24  Bin Cheng  <bin.cheng@arm.com>
20775         PR tree-optimization/81913
20776         * tree-ssa-loop-niter.c (number_of_iterations_cond): Skip niter
20777         analysis when either IVs in condition can wrap.
20779 2017-08-24  Uros Bizjak  <ubizjak@gmail.com>
20781         * dwarf2out.c (MAX_ARTIFICIAL_LABEL_BYTES): Increase to 40.
20782         * dwarf2cfi.c (MAX_ARTIFICIAL_LABEL_BYTES): Remove.
20784 2017-08-24  Richard Biener  <rguenther@suse.de>
20786         PR target/81921
20787         * targhooks.c (default_target_can_inline_p): Properly
20788         use target_option_default_node when no DECL_FUNCTION_SPECIFIC_TARGET
20789         is present and always compare.
20790         * config/i386/i386.c (ix86_valid_target_attribute_tree): Do not
20791         infer -mfpmath=sse from TARGET_SSE_P.
20792         (ix86_can_inline_p): Properly use target_option_default_node when
20793         no DECL_FUNCTION_SPECIFIC_TARGET is present and always compare.
20795 2017-08-24  Richard Biener  <rguenther@suse.de>
20797         PR debug/81936
20798         * dwarf2out.c (output_die): Handle flag_generate_offload like
20799         flag_generate_lto.
20800         (output_comp_unit): Likewise.
20801         (gen_array_type_die): Likewise.
20802         (dwarf2out_early_finish): Likewise.
20803         (note_variable_value_in_expr): Likewise.
20804         (dwarf2out_finish): Likewise.  Adjust assert.
20805         * cgraphunit.c (symbol_table::compile): Move setting of
20806         flag_generate_offload earlier ...
20807         (symbol_table::finalize_compilation_unit): ... here, before
20808         early debug finalization.
20810 2017-08-24  Richard Biener  <rguenther@suse.de>
20812         * config/i386/i386.c: Include symbol-summary.h, ipa-prop.h
20813         and ipa-fnsummary.h.
20814         (ix86_can_inline_p): When ix86_fpmath flags do not match
20815         check whether the callee uses FP math at all.
20817 2017-08-24  Aldy Hernandez  <aldyh@redhat.com>
20819         PR middle-end/81931
20820         * tree-ssanames.c (get_nonzero_bits): Use element_precision
20821         instead of TYPE_PRECISION.
20823 2017-08-24  Richard Sandiford  <richard.sandiford@linaro.org>
20824             Alan Hayward  <alan.hayward@arm.com>
20825             David Sherwood  <david.sherwood@arm.com>
20827         * combine.c (make_extraction): Use subreg_offset_from_lsb.
20829 2017-08-23  Daniel Santos  <daniel.santos@pobox.com>
20831         * config/i386/i386.h (ix86_frame::stack_realign_allocate_offset):
20832         Remove field.
20833         (ix86_frame::stack_realign_allocate): New field.
20834         (struct machine_frame_state): Modify comments.
20835         (machine_frame_state::sp_realigned_fp_end): New field.
20836         * config/i386/i386.c (ix86_compute_frame_layout): Rework stack frame
20837         layout calculation.
20838         (sp_valid_at): Add assertion to assure no attempt to access invalid
20839         offset of a realigned stack.
20840         (fp_valid_at): Likewise.
20841         (choose_baseaddr): Modify comments.
20842         (ix86_emit_outlined_ms2sysv_save): Adjust to changes in
20843         ix86_expand_prologue.
20844         (ix86_expand_prologue): Modify stack realignment and allocation.
20845         (ix86_expand_epilogue): Modify comments.
20846         * doc/sourcebuild.texi: Add documentation for target selectors avx2,
20847         avx2_runtime, avx512f, and avx512f_runtime.
20849 2017-08-23  Uros Bizjak  <ubizjak@gmail.com>
20851         * config/i386/i386.opt: Remove unneeded Init(0) initializations.
20852         (mstackrealign): Do not init to -1.
20853         * config/i386/i386.c (ix86_option_override_internal):
20854         Check opts_set, not opts when setting default value of
20855         opts->x_ix86_force_align_arg_pointer.
20857 2017-08-23  Richard Biener  <rguenther@suse.de>
20859         * function.c (fndecl_name): Use verbosity 1 (no arguments) for
20860         lang_hooks.decl_printable_name.
20861         * print-rtl-function.c (print_rtx_function): Likewise.
20862         * tree-pretty-print.c (dump_function_header): Likewise.
20864 2017-08-23  Richard Biener  <rguenther@suse.de>
20866         PR lto/81940
20867         * dwarf2out.c (dwarf2out_abstract_function): Handle LTO with
20868         -g0 at compile-time.
20870 2017-08-23  Tamar Christina  <tamar.christina@arm.com>
20872         PR middle-end/19706
20873         * doc/sourcebuild.texi (Other hardware attributes):
20874         Document xorsign.
20876 2017-08-23  Tamar Christina  <tamar.christina@arm.com>
20878         PR middle-end/19706
20879         * tree-ssa-math-opts.c (convert_expand_mult_copysign):
20880         Fix single-use check.
20882 2017-08-23  Thomas Preud'homme  <thomas.preudhomme@arm.com>
20884         * gcc.c (execute): Only test for SIGKILL and SIGQUIT if available.
20886 2017-08-22  Daniel Santos  <daniel.santos@pobox.com>
20888         * doc/install.texi: Modify to add more details on running selected
20889         tests.
20891 2017-08-22  Daniel Santos  <daniel.santos@pobox.com>
20893         * config/i386/i386.c (ix86_option_override_internal): Error when -mx32
20894         is combined with -mabi=ms.
20895         (ix86_function_type_abi): Limit errors for mixing -mx32 with attribute
20896         ms_abi.
20898 2017-08-22  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20900         PR tree-optimization/81488
20901         * gimple-ssa-strength-reduction (struct slsr_cand_d): Add visited
20902         and cached_basis fields.
20903         (MAX_SPREAD): New constant.
20904         (alloc_cand_and_find_basis): Initialize new fields.
20905         (clear_visited): New function.
20906         (create_phi_basis_1): Rename from create_phi_basis, set visited
20907         and cached_basis fields.
20908         (create_phi_basis): New wrapper function.
20909         (phi_add_costs_1): Rename from phi_add_costs, add spread
20910         parameter, set visited field, short-circuit when limits reached.
20911         (phi_add_costs): New wrapper function.
20912         (record_phi_increments_1): Rename from record_phi_increments, set
20913         visited field.
20914         (record_phi_increments): New wrapper function.
20915         (phi_incr_cost_1): Rename from phi_incr_cost, set visited field.
20916         (phi_incr_cost): New wrapper function.
20917         (all_phi_incrs_profitable_1): Rename from
20918         all_phi_incrs_profitable, set visited field.
20919         (all_phi_incrs_profitable): New wrapper function.
20921 2017-08-22  Richard Sandiford  <richard.sandiford@linaro.org>
20922             Alan Hayward  <alan.hayward@arm.com>
20923             David Sherwood  <david.sherwood@arm.com>
20925         * rtl.h (paradoxical_subreg_p): Define inline, and add a version
20926         that takes the outer and inner modes.
20927         * doc/rtl.texi: Use paradoxical_subreg_p instead of a GET_MODE_SIZE
20928         comparison as the canonical test for a paradoxical subreg.
20929         * combine.c (simplify_set): Use paradoxical_subreg_p.
20930         (make_extraction): Likewise.
20931         (force_to_mode): Likewise.
20932         (rtx_equal_for_field_assignment_p): Likewise.
20933         (gen_lowpart_for_combine): Likewise.
20934         (simplify_comparison): Likewise.
20935         * cse.c (equiv_constant): Likewise.
20936         * expmed.c (store_bit_field_1): Likewise.
20937         * final.c (alter_subreg): Likewise.
20938         * fwprop.c (propagate_rtx): Likewise.
20939         (forward_propagate_subreg): Likewise.
20940         * ira-conflicts.c (ira_build_conflicts): Likewise.
20941         * lower-subreg.c (simplify_gen_subreg_concatn): Likewise.
20942         * lra-constraints.c (curr_insn_transform): Likewise.
20943         (split_reg): Likewise.
20944         * lra-eliminations.c (move_plus_up): Likewise.
20945         (lra_eliminate_regs_1): Likewise.
20946         * recog.c (general_operand): Likewise.
20947         * ree.c (combine_reaching_defs): Likewise.
20948         * reload.c (push_reload): Likewise.
20949         (find_reloads): Likewise.
20950         * reload1.c (elimination_effects): Likewise.
20951         (compute_reload_subreg_offset): Likewise.
20952         (choose_reload_regs): Likewise.
20953         * rtlanal.c (subreg_lsb_1): Likewise.
20954         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
20955         (simplify_subreg): Likewise.
20956         * var-tracking.c (track_loc_p): Likewise.
20957         * emit-rtl.c (byte_lowpart_offset): Likewise.
20958         (paradoxical_subreg_p): Delete out-of-line definition.
20960 2017-08-22  Jeff Law  <law@redhat.com>
20962         PR tree-optimization/81741
20963         PR tree-optimization/71947
20964         * tree-ssa-dom.c: Include tree-inline.h.
20965         (record_temporary_equivalences): Only record SSA_NAME = SSA_NAME
20966         equivalences if one is more expensive to compute than the other.
20967         * tree-ssa-scopedtables.h (class const_or_copies): Make
20968         record_const_or_copy_raw method private.
20969         (class avail_exprs_stack): New method simplify_binary_operation.
20970         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr): Call
20971         avail_exprs_stack::simplify_binary_operation as needed.
20972         (avail_exprs_stack::simplify_binary_operation): New function.
20974 2017-08-22  Sebastian Huber  <sebastian.huber@embedded-brains.de>
20976         * config.gcc (powerpc-*-rtems*): Add rs6000/linux64.opt.
20977         * config/rs6000/rtems.h (ASM_PREFERRED_EH_DATA_FORMAT): New define.
20978         (DOT_SYMBOLS): Likewise.
20979         (MINIMAL_TOC_SECTION_ASM_OP): Likewise.
20980         (RELOCATABLE_NEEDS_FIXUP): Likewise.
20981         (RS6000_ABI_NAME): Likewise.
20982         (TARGET_CMODEL): Likewise.
20983         (TOC_SECTION_ASM_OP): Likewise.
20984         (SET_CMODEL): New macro.
20985         (SUBSUBTARGET_OVERRIDE_OPTIONS): Evaluate cmodel options.
20987 2017-08-22  Richard Biener  <rguenther@suse.de>
20989         * tree-inline.c (remap_type_1): Change asserts on TYPE_SIZE[_UNIT]
20990         to allow for free-lang-data replacements similar to verify_type_variant.
20992 2017-08-22  Richard Sandiford  <richard.sandiford@linaro.org>
20993             Alan Hayward  <alan.hayward@arm.com>
20994             David Sherwood  <david.sherwood@arm.com>
20996         * config/aarch64/aarch64.md (casesi): Use DImode rather than
20997         VOIDmode for the LABEL_REF.
20999 2017-08-22  Richard Biener  <rguenther@suse.de>
21001         * tree-cfg.c (gimple_split_edge): Avoid reallocating target PHI nodes.
21003 2017-08-22  Richard Biener  <rguenther@suse.de>
21005         * common.opt (feliminate-dwarf2-dups): Ignore.
21006         * doc/invoke.texi (feliminate-dwarf2-dups): Remove documentation.
21007         * dwarf2out.c (push_new_compile_unit, pop_compile_unit,
21008         same_die_p_wrap, compute_section_prefix,
21009         is_symbol_die, assign_symbol_names, break_out_includes): Remove.
21010         (comdat_symbol_id, comdat_symbol_number): Likewise.
21011         (cu_hash_table_entry, cu_hash_table_entry_hasher, cu_hash_type):
21012         Likewise.
21013         (check_duplicate_cu, record_comdat_symbol_number): Likewise.
21014         (output_die): Mark unreachable path unreachable.
21015         (dwarf2out_start_source_file): Do not create DW_TAG_GNU_BINCL.
21016         (dwarf2out_end_source_file): Do not create DW_TAG_GNU_EINCL.
21017         (dwarf2out_init): Remove code handling flag_eliminate_dwarf2_dups.
21018         (dwarf2out_early_finish): Likewise.
21020 2017-08-22  Aldy Hernandez  <aldyh@redhat.com>
21022         * wide-int.h (hwi_with_prec::hwi_with_prec): Sign extend.
21024 2017-08-22  Georg-Johann Lay  <avr@gjlay.de>
21026         PR target/81910
21027         * config/avr/avr.c (avr_handle_addr_attribute): Early return if
21028         not VAR_P. Filter attribute warnings with OPT_Wattributes.
21029         (avr_attribute_table) <io, io_low, address>: Initialize
21030         .decl_required with true.
21032 2017-08-21  Michael Meissner  <meissner@linux.vnet.ibm.com>
21034         * config/rs6000/rs6000-cpus.def (-mvsx-scalar-float): Delete
21035         undocumented debugging options.
21036         (-mvsx-scalar-double): Likewise.
21037         (-mallow-df-permute): Likewise.
21038         (-mvectorize-builtins): Likewise.
21039         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
21040         (rs6000_builtin_vectorized_function): Likewise.
21041         (rs6000_builtin_md_vectorized_function): Likewise.
21042         (rs6000_opt_vars): Likewise.
21044 2017-08-21  Uros Bizjak  <ubizjak@gmail.com>
21046         PR target/46091
21047         * config/i386/i386.md (*btsq_imm): Rename from *btsq.
21048         (*btrq_imm): Rename from *btrq.
21049         (*btcq_imm): Rename from *btcq.
21050         (btsc): New code attribute.
21051         (*<btsc><mode>): New insn pattern.
21052         (*btr<mode>): Ditto.
21053         (*<btsc><mode>_mask): New insn_and_split pattern.
21054         (*btr<mode>_mask): Ditto.
21056 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21057             Alan Hayward  <alan.hayward@arm.com>
21058             David Sherwood  <david.sherwood@arm.com>
21060         * function.c (pad_below): Simplify padding calculation.
21062 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21063             Alan Hayward  <alan.hayward@arm.com>
21064             David Sherwood  <david.sherwood@arm.com>
21066         * target.def (function_prologue): Remove frame size argument.
21067         (function_epilogue): Likewise.
21068         * doc/tm.texi: Regenerate.
21069         * final.c (final_start_function): Update call to function_prologue.
21070         (final_end_function): Update call to function_epilogue.
21071         (default_function_pro_epilogue): Remove frame size argument.
21072         * output.h (default_function_pro_epilogue): Likewise.
21073         * config/arm/arm.c (arm_output_function_epilogue): Likewise.
21074         (arm_output_function_prologue): Likewise.
21075         * config/frv/frv.c (frv_function_prologue): Likewise.
21076         (frv_function_epilogue): Likewise.
21077         * config/i386/i386.c (ix86_output_function_epilogue): Likewise.
21078         * config/ia64/ia64.c (ia64_output_function_prologue): Likewise.
21079         (ia64_output_function_epilogue): Likewise.
21080         * config/m32r/m32r.c (m32r_output_function_prologue): Likewise.
21081         (m32r_output_function_epilogue): Likewise.
21082         * config/microblaze/microblaze.c (microblaze_function_prologue)
21083         (microblaze_function_epilogue): Likewise.
21084         * config/mips/mips.c (mips_output_function_prologue): Likewise.
21085         (mips_output_function_epilogue): Likewise.
21086         * config/mmix/mmix.c (mmix_target_asm_function_prologue): Likewise.
21087         (mmix_target_asm_function_epilogue): Likewise.
21088         * config/msp430/msp430.c (msp430_start_function): Likewise.
21089         * config/nds32/nds32.c (nds32_asm_function_prologue): Likewise.
21090         (nds32_asm_function_epilogue): Likewise.
21091         * config/nios2/nios2.c (nios2_asm_function_prologue): Likewise.
21092         * config/pa/pa.c (pa_output_function_prologue): Likewise.
21093         (pa_output_function_epilogue): Likewise.
21094         * config/powerpcspe/powerpcspe.c (rs6000_output_function_prologue)
21095         (rs6000_output_function_epilogue): Likewise.
21096         * config/rl78/rl78.c (rl78_start_function): Likewise.
21097         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Likewise.
21098         (rs6000_output_function_epilogue): Likewise.
21099         * config/rx/rx.c (rx_output_function_prologue): Likewise.
21100         * config/sh/sh.c (sh_output_function_epilogue): Likewise.
21101         * config/sparc/sparc.c (sparc_asm_function_prologue): Likewise.
21102         (sparc_asm_function_epilogue): Likewise.
21104 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21106         * tree.h (type_has_mode_precision_p): New function.
21107         * convert.c (convert_to_integer_1): Use it.
21108         * expr.c (expand_expr_real_2): Likewise.
21109         (expand_expr_real_1): Likewise.
21110         * fold-const.c (fold_single_bit_test_into_sign_test): Likewise.
21111         * match.pd: Likewise.
21112         * tree-ssa-forwprop.c (simplify_rotate): Likewise.
21113         * tree-ssa-math-opts.c (convert_mult_to_fma): Likewise.
21114         * tree-tailcall.c (process_assignment): Likewise.
21115         * tree-vect-loop.c (vectorizable_reduction): Likewise.
21116         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern)
21117         (vect_recog_mult_pattern, vect_recog_divmod_pattern): Likewise.
21118         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
21119         (vectorizable_assignment): Likewise.
21120         (vectorizable_shift): Likewise.
21121         (vectorizable_operation): Likewise.
21122         * tree-vrp.c (register_edge_assert_for_2): Likewise.
21124 2017-08-21  Wilco Dijkstra  <wdijkstr@arm.com>
21126         * match.pd: Add pow (C, x) simplification.
21128 2017-08-21  Richard Biener  <rguenther@suse.de>
21130         PR tree-optimization/81900
21131         * tree-ssa-pre.c (compute_antic_aux): Properly compute changed
21132         for blocks with abnormal predecessors.
21133         (compute_antic): Do not set visited flag prematurely.
21135 2017-08-21  Georg-Johann Lay  <avr@gjlay.de>
21137         PR target/79883
21138         * config/avr/avr.c (avr_set_current_function): Typo in diagnostic.
21140 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21142         * stor-layout.h (vector_type_mode): Move to...
21143         * tree.h (vector_type_mode): ...here.
21144         * stor-layout.c (vector_type_mode): Move to...
21145         * tree.c (vector_type_mode): ...here.  Include rtl.h and regs.h.
21147 2017-08-21  Richard Biener  <rguenther@suse.de>
21149         * debug.h (struct gcc_debug_hooks): Add die_ref_for_decl and
21150         register_external_die hooks.
21151         (debug_false_tree_charstarstar_uhwistar): Declare.
21152         (debug_nothing_tree_charstar_uhwi): Likewise.
21153         * debug.c (do_nothing_debug_hooks): Adjust.
21154         (debug_false_tree_charstarstar_uhwistar): New do nothing.
21155         (debug_nothing_tree_charstar_uhwi): Likewise.
21156         * dbxout.c (dbx_debug_hooks): Adjust.
21157         (xcoff_debug_hooks): Likewise.
21158         * sdbout.c (sdb_debug_hooks): Likewise.
21159         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
21160         * dwarf2out.c (macinfo_label_base): New global.
21161         (dwarf2out_register_external_die): New function for the
21162         register_external_die hook.
21163         (dwarf2out_die_ref_for_decl): Likewise for die_ref_for_decl.
21164         (dwarf2_debug_hooks): Use them.
21165         (dwarf2_lineno_debug_hooks): Adjust.
21166         (struct die_struct): Add with_offset flag.
21167         (DEBUG_LTO_DWO_INFO_SECTION, DEBUG_LTO_INFO_SECTION,
21168         DEBUG_LTO_DWO_ABBREV_SECTION, DEBUG_LTO_ABBREV_SECTION,
21169         DEBUG_LTO_DWO_MACINFO_SECTION, DEBUG_LTO_MACINFO_SECTION,
21170         DEBUG_LTO_DWO_MACRO_SECTION, DEBUG_LTO_MACRO_SECTION,
21171         DEBUG_LTO_LINE_SECTION, DEBUG_LTO_DWO_STR_OFFSETS_SECTION,
21172         DEBUG_LTO_STR_DWO_SECTION, DEBUG_STR_LTO_SECTION): New macros
21173         defining section names for the early LTO debug variants.
21174         (reset_indirect_string): New helper.
21175         (add_AT_external_die_ref): Helper for dwarf2out_register_external_die.
21176         (print_dw_val): Add support for offsetted symbol references.
21177         (get_ultimate_context): Split out from is_cxx.
21178         (is_cxx): Use get_ultimate_context.
21179         (is_fortran): Add decl overload.
21180         (compute_comp_unit_symbol): Split out worker from
21181         compute_section_prefix.
21182         (compute_section_prefix): Call compute_comp_unit_symbol and
21183         set comdat_type_p here.
21184         (output_die): Skip DIE symbol output for the LTO added one.
21185         Handle DIE symbol references with offset.
21186         (output_comp_unit): Guard section name mangling properly.
21187         For LTO debug sections emit a symbol at the section beginning
21188         which we use to refer to its DIEs.
21189         (add_abstract_origin_attribute): For DIEs registered via
21190         dwarf2out_register_external_die directly refer to the early
21191         DIE rather than indirectly through the shadow one we created.
21192         Remove obsolete call to dwarf2out_abstract_function for
21193         non-function/block origins.
21194         (gen_array_type_die): When generating early LTO debug do
21195         not emit DW_AT_string_length.
21196         (gen_formal_parameter_die): Do not re-create DIEs for PARM_DECLs
21197         late when in LTO.  As suggested place a gcc_unreachable for
21198         the DECL_ABSTRACT_P case.
21199         (gen_subprogram_die): Avoid another specification DIE
21200         for early built declarations/definitions for the late LTO case.
21201         (gen_variable_die): Add type references for late duplicated VLA dies
21202         when in late LTO.
21203         (gen_inlined_subroutine_die): Do not call dwarf2out_abstract_function,
21204         we have the abstract instance already.
21205         (process_scope_var): Adjust decl DIE contexts in LTO which
21206         first puts them in limbo.
21207         (gen_decl_die): Do not generate type DIEs late apart from
21208         types for VLAs or for decls we do not yet have a DIE.  Do not
21209         call dwarf2out_abstract_function late.
21210         (dwarf2out_early_global_decl): Make sure to create DIEs
21211         for abstract instances of a decl first.
21212         (dwarf2out_late_global_decl): Adjust comment.
21213         (output_macinfo_op): With multiple macro sections use
21214         macinfo_label_base to distinguish labels.
21215         (output_macinfo): Likewise.  Update macinfo_label_base.
21216         Pass in the line info label.
21217         (note_variable_value_in_expr): When generating LTO resolve
21218         all variable values here by generating DIEs as needed.
21219         (init_sections_and_labels): Add early LTO debug flag parameter
21220         and generate different sections and names if set.  Add generation
21221         counter for the labels so we can have multiple of them.
21222         (reset_dies): Helper to allow DIEs to be output multiple times.
21223         (dwarf2out_finish): When outputting DIEs to the fat part of an
21224         LTO object first reset DIEs.
21225         (dwarf2out_early_finish): Output early DIEs when generating LTO.
21226         (modified_type_die): Check for decl_ultimate_origin being self
21227         before recursing.
21228         (gen_type_die_with_usage): Likewise.
21229         (gen_typedef_die): Allow decl_ultimate_origin being self.
21230         (set_decl_abstract_flags): Remove.
21231         (set_block_abstract_flags): Likewise.
21232         (dwarf2out_abstract_function): Treat the early generated DIEs
21233         as the abstract copy and only add DW_AT_inline and
21234         DW_AT_artificial here and call set_decl_origin_self.
21235         If the DIE has an abstract origin don't do anything.
21236         * tree.c (free_lang_data): Build a dummy TRANSLATION_UNIT_DECL
21237         if we have none yet (Go fails to build one, PR78628).
21238         (variably_modified_type_p): Prevent endless recursion for Ada
21239         cyclic pointer types.
21240         * lto-streamer-in.c: Include debug.h.
21241         (dref_queue): New global.
21242         (lto_read_tree_1): Stream in DIE references.
21243         (lto_input_tree): Register DIE references.
21244         (input_function): Stream DECL_DEBUG_ARGS.
21245         * lto-streamer-out.c: Include debug.h.
21246         (lto_write_tree_1): Output DIE references.
21247         (DFS::DFS_write_tree_body): Follow DECL_ABSTRACT_ORIGIN.
21248         Force a TRANSLATION_UNIT_DECL DECL_CONTEXT for file-scope decls.
21249         (output_function): Stream DECL_DEBUG_ARGS.
21250         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
21251         Stream DECL_ABSTRACT_ORIGIN.
21252         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
21253         (write_ts_decl_minimal_tree_pointers): Force a TRANSLATION_UNIT_DECL
21254         DECL_CONTEXT for file-scope decls.
21255         * lto-streamer.h (struct dref_entry): Declare.
21256         (dref_queue): Likewise.
21257         * cfgexpand.c (pass_expand::execute): Do not call the
21258         outlining_inline_function hook here.
21259         * lto-wrapper.c (debug_obj): New global.
21260         (tool_cleanup): Unlink it if required.
21261         (debug_objcopy): New function.
21262         (run_gcc): Handle early debug sections in the IL files by
21263         extracting them to separate files, partially linkin them and
21264         feeding the result back as result to the linker.
21265         * config/darwin.h (DEBUG_LTO_INFO_SECTION, DEBUG_LTO_ABBREV_SECTION,
21266         DEBUG_LTO_MACINFO_SECTION, DEBUG_LTO_LINE_SECTION,
21267         DEBUG_STR_LTO_SECTION, DEBUG_LTO_MACRO_SECTION): Put early debug
21268         sections into a separate segment.
21269         * config/darwin.c (darwin_asm_named_section): Handle __GNU_DWARF_LTO
21270         segments.
21271         (darwin_asm_dwarf_section): Likewise.
21272         (darwin_asm_output_dwarf_offset): Likewise.
21273         * config/i386/i386.c (make_resolver_func): Set DECL_IGNORED_P.
21275 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21276             Alan Hayward  <alan.hayward@arm.com>
21277             David Sherwood  <david.sherwood@arm.com>
21279         * read-md.h (md_reader::record_potential_iterator_use): Replace
21280         pointer argument with an rtx and an index.
21281         * read-rtl.c (iterator_group::apply_iterator): Likewise.
21282         (apply_mode_iterator): Likewise.
21283         (apply_code_iterator): Likewise.
21284         (apply_int_iterator): Likewise.
21285         (apply_subst_iterator): Likewise.
21286         (record_iterator_use): Likewise.
21287         (record_attribute_use): Likewise.
21288         (md_reader::record_potential_iterator_use): Likewise.  Update calls
21289         to record_iterator_use and apply_iterator.
21290         (iterator_use): Replace ptr with x and index.
21291         (attribute_use): Likewise.
21292         (apply_attribute_uses): Update calls to apply_iterator.
21293         (apply_iterators): Likewise.  Update initialization of iterator_use.
21294         (rtx_reader::read_rtx_code): Update calls to record_iterator_use
21295         and record_potential_iterator_use.
21296         (rtx_reader::read_rtx_operand): Likewise.
21298 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21299             Alan Hayward  <alan.hayward@arm.com>
21300             David Sherwood  <david.sherwood@arm.com>
21302         * varasm.c (const_rtx_hash_1): Don't hash in the mode of a
21303         CONST_WIDE_INT.
21305 2017-08-21  Richard Biener  <rguenther@suse.de>
21307         PR middle-end/81884
21308         * tree-ssa-alias.c (stmt_kills_ref_p): Handle array accesses
21309         at struct end conservatively when comparing common bases.
21311 2017-08-21  Richard Biener  <rguenther@suse.de>
21313         * tree-ssa-loop-im.c (struct lim_aux_data): Add ref index member.
21314         (mem_ref_in_stmt): Remove.
21315         (determine_max_movement): Use ref index to get at the reference.
21316         (invariantness_dom_walker::before_dom_children): Deal with
21317         lim data already initialized.
21318         (gather_mem_refs_stmt): Initialize lim data and record ref index.
21320 2017-08-19  Uros Bizjak  <ubizjak@gmail.com>
21322         * config/i386/i386.h (OPTION_MASK_ISA_ROUND): Remove.
21323         (TARGET_ISA_ROUND): Ditto.
21324         (TARGET_ROUND): Ditto.
21325         * config/i386/i386.c: Substitute TARGET_ROUND with TARGET_SSE4_1.
21326         * config/i386/i386.md: Ditto.
21327         * config/i386/sse.md: Ditto.
21328         * config/i386/i386-builtin.def: Substitute OPTION_MASK_ISA_ROUND
21329         with OPTION_MASK_ISA_SSE4_1.
21331 2017-08-19  Uros Bizjak  <ubizjak@gmail.com>
21333         PR target/81894
21334         * doc/extend.texi (x86 Built-in Functions): Correct the name of
21335         __builtin_ia32_lzcnt_u16.
21337 2017-08-18  Peter Bergner  <bergner@vnet.ibm.com>
21339         PR target/80210
21340         * config/rs6000/rs6000.c (rs6000_activate_target_options): New function.
21341         (rs6000_set_current_function): Rewrite function to use it.
21343 2017-08-18  H.J. Lu  <hongjiu.lu@intel.com>
21345         PR c/53037
21346         * print-tree.c (print_node): Support DECL_WARN_IF_NOT_ALIGN
21347         and TYPE_WARN_IF_NOT_ALIGN.
21348         * stor-layout.c (do_type_align): Merge DECL_WARN_IF_NOT_ALIGN.
21349         (handle_warn_if_not_align): New.
21350         (place_union_field): Call handle_warn_if_not_align.
21351         (place_field): Call handle_warn_if_not_align.
21352         Copy TYPE_WARN_IF_NOT_ALIGN.
21353         (finish_builtin_struct): Copy TYPE_WARN_IF_NOT_ALIGN.
21354         (layout_type): Likewise.
21355         * tree-core.h (tree_type_common): Add warn_if_not_align.  Set
21356         spare to 18.
21357         (tree_decl_common): Add warn_if_not_align.
21358         * tree.c (build_range_type_1): Copy TYPE_WARN_IF_NOT_ALIGN.
21359         * tree.h (TYPE_WARN_IF_NOT_ALIGN): New.
21360         (SET_TYPE_WARN_IF_NOT_ALIGN): Likewise.
21361         (DECL_WARN_IF_NOT_ALIGN): Likewise.
21362         (SET_DECL_WARN_IF_NOT_ALIGN): Likewise.
21363         * doc/extend.texi: Document warn_if_not_aligned attribute.
21364         * doc/invoke.texi: Document -Wif-not-aligned and -Wpacked-not-aligned.
21366 2017-08-17  Martin Liska  <mliska@suse.cz>
21368         PR bootstrap/81864
21369         * tree-loop-distribution.c (ddrs_table): Change type to pointer type.
21370         (get_data_dependence): Use it as pointer type.
21371         (distribute_loop): Likewise.
21373 2017-08-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21375         * config/rs6000/altivec.md (UNSPEC_VMRGOW_DIRECT): New constant.
21376         (p8_vmrgew_v4sf_direct): Generalize to p8_vmrgew_<mode>_direct.
21377         (p8_vmrgow_<mode>_direct): New define_insn.
21378         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Properly
21379         handle endianness for vmrgew and vmrgow permute patterns.
21381 2017-08-17  Peter Bergner  <bergner@vnet.ibm.com>
21383         * config/rs6000/altivec.md (VParity): Remove TARGET_VSX_TIMODE.
21384         * config/rs6000/rs6000-cpus.def: Remove comment.
21385         (ISA_2_7_MASKS_SERVER): Delete OPTION_MASK_VSX_TIMODE;
21386         (POWERPC_MASKS): Likewise.
21387         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Remove unneeded
21388         use of TARGET_VSX_TIMODE.
21389         (rs6000_setup_reg_addr_masks): Change TARGET_VSX_TIMODE to TARGET_VSX.
21390         (rs6000_init_hard_regno_mode_ok): Remove unneeded uses of
21391         TARGET_VSX_TIMODE.  Change use of TARGET_VSX_TIMODE to TARGET_VSX.
21392         (rs6000_option_override_internal): Remove dead code.
21393         (rs6000_legitimize_address): Change TARGET_VSX_TIMODE to TARGET_VSX.
21394         (rs6000_legitimize_reload_address): Likewise.
21395         (rs6000_legitimate_address_p): Likewise.
21396         (rs6000_opt_masks): Delete "vsx-timode".
21397         (rs6000_disable_incompatible_switches): Remove mention of -mvsx-timode
21398         from function comment.
21399         * config/rs6000/rs6000.h (MASK_VSX_TIMODE): Delete.
21400         * config/rs6000/rs6000.md (FMOVE128_GPR): Remove TARGET_VSX_TIMODE.
21401         (V16QI, V8HI, V4SI, V4SF, V2DI, V2DF, V1TI): Remove useless empty
21402         condition.
21403         * config/rs6000/rs6000.opt (mvsx-timode): Replace with stub.
21404         * config/rs6000/vector.md (VEC_IP): Remove TARGET_VSX_TIMODE.
21405         * config/rs6000/vsx.md (VSX_LE_128): Likewise.
21406         (VSX_TI): Likewise.
21407         (VSX_M): Likewise.
21408         (define_peephole2): Likewise.
21410 2017-08-17  Martin Sebor  <msebor@redhat.com>
21412         PR c/81859
21413         * pretty-print.c (pp_format): Use strnlen in %.*s to avoid reading
21414         past the end of an array.
21415         (test_pp_format): Add test cases.
21417 2017-08-17  Richard Sandiford  <richard.sandiford@linaro.org>
21419         * internal-fn.def (CLRSB, CLZ, CTZ, FFS, PARITY, POPCOUNT): Add
21420         missing ECF_NOTHROW flags.
21422 2017-08-17  Peter Bergner  <bergner@vnet.ibm.com>
21424         PR target/72804
21425         * config/rs6000/vsx.md (*vsx_le_permute_<mode>): Add support for
21426         operands residing in integer registers.
21427         (*vsx_le_perm_load_<mode>): Likewise.
21428         (*vsx_le_perm_store_<mode>): Likewise.
21429         (define_peephole2): Add peepholes to optimize the above.
21431 2017-08-17  Marek Polacek  <polacek@redhat.com>
21433         PR middle-end/81814
21434         * fold-const.c (operand_equal_for_comparison_p): Remove code that used
21435         to mimic what shorten_compare did.  Change the return type to bool.
21436         (fold_cond_expr_with_comparison): Update call to
21437         operand_equal_for_comparison_p.
21438         (fold_ternary_loc): Likewise.
21440 2017-08-17  Jackson Woodruff  <jackson.woodruff@arm.com>
21442         * aarch64-simd.md (mov<mode>): No longer force zero immediate into
21443         register.
21444         (*aarch64_simd_mov<mode>): Add new case for stp using zero immediate.
21446 2017-08-17  Richard Biener  <rguenther@suse.de>
21448         * tree-ssa-structalias.c (solve_graph): When propagating
21449         to successors update the graphs succ edges and avoid duplicate work.
21451 2017-08-17  Maxim Ostapenko  <m.ostapenko@samsung.com>
21453         PR target/81861
21454         * config/i386/i386.c (ix86_option_override_internal): Save target
21455         specific options after ix86_stack_protector_guard_reg was changed.
21457 2017-08-17  Richard Biener  <rguenther@suse.de>
21459         PR tree-optimization/81827
21460         * tree-ssa-structalias.c (struct variable_info): Add is_reg_var flag.
21461         (new_var_info): Initialize it conservatively.
21462         (get_call_vi): Mark register vars.
21463         (new_scalar_tmp_constraint_exp): Likewise.
21464         (handle_rhs_call): Likewise.
21465         (handle_const_call): Likewise.
21466         (create_function_info_for): Likewise.
21467         (solve_constraints): Sort varinfos to separate register from
21468         non-register vars to pack points-to solution bitmaps during
21469         iteration.
21471 2017-08-17  Marek Polacek  <polacek@redhat.com>
21473         * gimplify.c (gimplify_adjust_omp_clauses): Compare with 0 instead of 1.
21475 2017-08-17  Richard Biener  <rguenther@suse.de>
21477         * tree-vrp.c (vrp_int_const_binop): Do not set *overflow_p
21478         to true when overflow is undefined and we saturated the result.
21480 2017-08-17  Alan Modra  <amodra@gmail.com>
21482         PR target/80938
21483         * config/rs6000/rs6000.c (rs6000_savres_strategy): Revert 2017-08-09.
21484         Don't use store multiple if only one reg needs saving.
21485         (interesting_frame_related_regno): New function.
21486         (rs6000_frame_related): Don't emit frame info for regs that
21487         don't need saving.
21488         (rs6000_emit_epilogue): Likewise.
21490 2017-08-16  Nathan Sidwell  <nathan@acm.org>
21492         * tree-core.h (tree_type_non_common): Rename binfo to lang_1.
21493         * tree.h (TYPE_BINFO): Use type_non_common.maxval.
21494         (TYPE_LANG_SLOT_1): Use type_non_common.lang_1, for any type.
21495         * tree.c (free_lang_data_in_type): Use else-if chain.  Always
21496         clear TYPE_LANG_1.  Remove obsolete member-function stripping.
21497         (find_decls_types_r): Comment about TYPE_MAX_VALUES_RAW.
21498         (verify_type): Adjust for TYPE_BINFO move.
21499         * lto-streamer-out.c (DFS::DFS_write_tree_body): No need to
21500         process TYPE_BINFO directly.
21501         (hash_tree): Likewise.
21502         * tree-streamer-in.c (lto_input_ts_type_non_common_tree_pointers):
21503         Likewise.
21504         * tree-streamer-out.c (write_ts_type_non_common_tree_pointers):
21505         Likewise.
21507 2017-08-16  David Malcolm  <dmalcolm@redhat.com>
21509         * diagnostic-show-locus.c (colorizer::m_caret): Remove unused field.
21511 2017-08-16  Uros Bizjak  <ubizjak@gmail.com>
21513         PR target/46091
21514         * config/i386/i386.md (*anddi_1_btr): Change predicates of
21515         operand 0 and operand 1 to nomimmediate_operand. Add "m" constraint.
21516         Add ix86_binary_operator_ok to insn constraint.
21517         (*iordi_1_bts): Ditto.
21518         (*xordi_1_btc): Ditto.
21519         (*btsq): Change predicate of operand 0 to nonimmediate_operand.
21520         Update corresponding peephole2 pattern.
21521         (*btrq): Ditto.
21522         (*btcq): Ditto.
21524 2017-08-16  Bin Cheng  <bin.cheng@arm.com>
21526         PR tree-optimization/81832
21527         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Don't
21528         copy loop header which has IFN_LOOP_DIST_ALIAS call.
21530 2017-08-16  Marek Polacek  <polacek@redhat.com>
21532         PR middle/81695
21533         * fold-const.c (fold_indirect_ref_1): Restore original behavior
21534         regarding size_zero_node.
21536 2017-08-16  Martin Liska  <mliska@suse.cz>
21538         PR target/81753
21539         * config.gcc: Respect previously set extra_objs in case
21540         of darwin target.
21542 2017-08-16  Richard Sandiford  <richard.sandiford@linaro.org>
21544         PR tree-optimization/81835
21545         * tree-vect-loop.c (vect_is_simple_reduction): Simply checks for
21546         the phi SSA_NAME.  Check that the condition in a COND_EXPR does
21547         not depend on the phi.
21549 2017-08-16  Alan Modra  <amodra@gmail.com>
21551         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Delete
21552         dead code.
21554 2017-08-16  Alan Modra  <amodra@gmail.com>
21556         * config/rs6000/rs6000.c (rs6000_reg_live_or_pic_offset_p): Merge..
21557         (save_reg_p): ..into this.  Update all callers.
21558         (first_reg_to_save): Simplify.
21560 2017-08-16  Alan Modra  <amodra@gmail.com>
21562         * config/rs6000/rs6000.c (rs6000_savres_strategy): Don't restore
21563         fixed regs.
21565 2017-08-15  Joseph Myers  <joseph@codesourcery.com>
21567         PR target/78460
21568         PR target/67712
21569         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Only unroll for
21570         constant count if that count is less than 32.
21572 2017-08-15  Nathan Sidwell  <nathan@acm.org>
21574         * gcc.c (execute): Emit friendlier message if inferior is killed
21575         by an external cause.
21577 2017-08-15  Richard Biener  <rguenther@suse.de>
21579         PR tree-optimization/81790
21580         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Handle both
21581         CONSTRUCTORs from simplifying and VN.
21583 2017-08-14  Martin Sebor  <msebor@redhat.com>
21585         * builtin-attrs.def: Add comments.
21587 2017-08-14  Martin Sebor  <msebor@redhat.com>
21589         PR c/81117
21590         * doc/extend.texi (attribute nonstring): Document new attribute.
21592 2017-08-14  Martin Sebor  <msebor@redhat.com>
21594         PR c/81117
21595         * tree-diagnostic.c (default_tree_printer): Handle %G.
21596         * gimple-pretty-print.h (percent_G_format): Declare new function.
21597         * gimple-pretty-print.c (percent_G_format): Define.
21598         * tree-pretty-print.c (percent_K_format): Add argument.
21600 2017-08-14  Martin Sebor  <msebor@redhat.com>
21602         PR translation/79998
21603         * gimple-ssa-sprintf.c (pass_sprintf_length::handle_gimple_call):
21604         Remove a stray space.
21606 2017-08-14  Uros Bizjak  <ubizjak@gmail.com>
21608         PR target/46091
21609         * config/i386/i386.md (*anddi_1_btr): New insn_and_split pattern.
21610         (*iordi_1_bts): Ditto.
21611         (*xordi_1_btc): Ditto.
21613 2017-08-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21615         PR target/79845
21616         * config/rs6000/linux64.h (INVALID_64BIT): Use quoted strings.
21617         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
21618         Likewise.
21619         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Use
21620         quoted strings, and make more translator-friendly.
21621         (darwin_rs6000_override_options): Likewise.
21622         (rs6000_option_override_internal): Likewise.
21623         (rs6000_return_in_memory): Fix overlong line.
21624         (init_cmulative_args): Use quoted strings, and make more
21625         translator-friendly.
21626         (rs6000_pass_by_reference): Fix overlong line.
21627         (def_builtin): Use quoted strings.
21628         (altivec_expand_predicate_builtin): Use quoted strings, and make
21629         more translator-friendly.
21630         (htm_expand_builtin): Use quoted strings.
21631         (cpu_expand_builtin): Use quoted strings, and make more
21632         translator-friendly.
21633         (altivec_expand_builtin): Likewise.
21634         (paired_expand_predicate_builtin): Likewise.
21635         (rs6000_invalid_builtin): Likewise.
21636         (builtin_function_type): Use quoted strings.
21637         (rs6000_expand_split_stack_prologue): Use quoted strings, and make
21638         more translator-friendly.
21639         (rs6000_trampoline_init): Likewise.
21640         (rs6000_handle_altivec_attribute): Likewise.
21641         (rs6000_inner_target_options): Use quoted strings.
21642         (rs6000_disable_incompatible_switches): Likewise.
21643         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Use quoted
21644         strings, and make more translator-friendly.
21645         (SUBSUBTARGET_OVERRIDE_OPTIONS): Use quoted strings.
21647 2017-08-14  Bin Cheng  <bin.cheng@arm.com>
21649         PR tree-optimization/81799
21650         * tree-loop-distribution.c (version_loop_by_alias_check): Force
21651         cond_expr to simple gimple operand.
21653 2017-08-14  Wilco Dijkstra  <wdijkstr@arm.com>
21655         PR middle-end/46932
21656         * doc/sourcebuild.texi (autoincdec): Add autoincdec description.
21658 2017-08-14  Georg-Johann Lay  <avr@gjlay.de>
21660         PR target/81754
21661         PR target/81268
21662         * config/avr/avr.opt (mgas-isr-prologues): New Var avr_gasisr_prologues.
21663         * config/avr/avr.md (gasisr, *gasisr): Use it instead of
21664         TARGET_GASISR_PROLOGUES.
21665         * config/avr/avr.c (avr_option_override): Same.
21666         (avr_pass_pre_proep::execute): Same.
21668 2017-08-13  H.J. Lu  <hongjiu.lu@intel.com>
21670         PR target/81820
21671         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Replace
21672         frame pointer with stack pointer - UNITS_PER_WORD in debug insns.
21674 2017-08-13  Uros Bizjak  <ubizjak@gmail.com>
21676         * config/i386/i386.md (*load_tp_<mode>): Redefine as
21677         define_insn_and_split.  Split to a memory load from 0 in
21678         DEFAULT_TLS_SEG_REG address space.  Merge with *load_tp_x32
21679         using PTR mode iterator.
21680         (*load_tp_x32_zext"): Redefine as define_insn_and_split.
21681         Split to a memory load from 0 in DEFAULT_TLS_SEG_REG address space.
21682         (*add_tp_<mode>): Redefine as define_insn_and_split.
21683         Split to an add with a memory load from 0 in DEFAULT_TLS_SEG_REG
21684         address space.  Merge with *add_tp_x32 using PTR mode iterator.
21685         (*add_tp_x32_zext"): Redefine as define_insn_and_split.
21686         Split to an add with a  memory load from 0 in
21687         DEFAULT_TLS_SEG_REG address space.
21689 2017-08-12  Andrew Pinski  <apinski@cavium.com>
21691         * config/aarch64/aarch64-option-extensions.def (rdma):
21692         Fix feature string to what Linux prints out in /proc/cpuinfo.
21694 2017-08-12  Pierre-Marie de Rodat  <derodat@adacore.com>
21696         PR ada/79542
21697         * dwarf2out.c (modified_type_die): For C typedef types that have
21698         an ultimate origin, process the ultimate origin instead of the
21699         input type.
21700         (gen_typedef_die): Assert that input DECLs have no ultimate
21701         origin.
21702         (gen_type_die_with_usage): For typedef variants that have an
21703         ultimate origin, just call gen_decl_die on the original DECL.
21704         (process_scope_var): Avoid creating DIEs for local typedefs and
21705         concrete static variables.
21707 2017-08-12  Alan Modra  <amodra@gmail.com>
21709         PR target/81170
21710         PR target/81295
21711         * config/rs6000/sysv4.h (STARTFILE_LINUX_SPEC): Upgrade to
21712         match gnu-user.h startfile.
21713         (ENDFILE_LINUX_SPEC): Similarly.
21715 2017-08-11  Thomas Schwinge  <thomas@codesourcery.com>
21717         PR lto/81430
21718         * config/nvptx/nvptx.c (nvptx_override_options_after_change):
21719         Remove function.
21720         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Remove definition.
21722 2017-08-11  Tamar Christina  <tamar.christina@arm.com>
21723         * config/aarch64/aarch64.md (mov<mode>): Change.
21724         (*movhf_aarch64, *movsf_aarch64, *movdf_aarch64):
21725         aarch64_reg_or_fp_float into aarch64_reg_or_fp_zero.
21726         * config/aarch64/predicates.md (aarch64_reg_or_fp_float): Removed.
21728 2017-08-11  Eric Botcazou  <ebotcazou@adacore.com>
21730         * tree-sra.c (build_access_from_expr_1): Use more precise diagnostics
21731         for storage order barriers.
21733 2017-08-11  Martin Liska  <mliska@suse.cz>
21735         PR tree-opt/79987
21736         * tree-chkp.c (chkp_get_bounds_for_decl_addr): Do not instrument
21737         variables of void type.
21739 2017-08-11  Martin Liska  <mliska@suse.cz>
21741         * asan.c (asan_protect_global): Replace ASM_OUTPUT_DEF with
21742         TARGET_SUPPORTS_ALIASES.
21743         * cgraph.c (cgraph_node::create_same_body_alias): Likewise.
21744         * ipa-visibility.c (can_replace_by_local_alias): Likewise.
21745         (optimize_weakref): Likewise.
21746         * symtab.c (symtab_node::noninterposable_alias): Likewise.
21747         * varpool.c (varpool_node::create_extra_name_alias): Likewise.
21748         * defaults.h: Introduce TARGET_SUPPORTS_ALIASES.
21750 2017-08-11  Martin Liska  <mliska@suse.cz>
21752         PR ipa/81213
21753         * config/i386/i386.c (make_resolver_func): Do complete
21754         refactoring of the function.
21756 2017-08-10  Uros Bizjak  <ubizjak@gmail.com>
21758         PR target/81708
21759         * config/i386/i386.opt (mstack-protector-guard-symbol=): New option
21760         * config/i386/i386.c (ix86_stack_protect_guard): Use
21761         ix86_stack_protect_guard_symbol_str to generate varible declaration.
21762         * doc/invoke.texi (x86 Options): Document
21763         -mstack-protector-guard-symbol= option.
21765 2017-08-10  Uros Bizjak  <ubizjak@gmail.com>
21767         * config/i386/i386-protos.h (ix86_split_stack_guard): New prototype.
21768         * config/i386/i386.c (ix86_split_stack_guard): New function.
21769         (ix86_xpand_split_stack_prologue): Call ix86_split_stack_guard.
21770         (ix86_legitimate_address_p) <case UNSPEC_STACK_CHECK>: Remove.
21771         (i386_asm_output_addr_const_extra) <case UNSPEC_STACK_CHECK>: Ditto.
21772         (optput_pic_addr_const): Remove UNSPEC_STACK_CHECK handling.
21773         * config/i386/i386.md (unspec): Remove UNSPEC_STACK_CHECK.
21774         (split_stack_space_check): Call ix86_split_stack_guard.
21776 2017-08-10  Martin Sebor  <msebor@redhat.com>
21778         * print-tree.c (print_node): Print location using the established
21779         format %s:%i%i.
21780         Replace spaces with colons.
21781         (debug_raw, debug): Ditto.
21783 2017-08-10  Martin Sebor  <msebor@redhat.com>
21785         PR c++/81586
21786         * pretty-print.c (pp_format): Correct the handling of %s precision.
21788 2017-08-10  H.J. Lu  <hongjiu.lu@intel.com>
21790         PR target/81736
21791         * config/i386/i386.c (ix86_finalize_stack_realign_flags): Renamed
21792         to ...
21793         (ix86_finalize_stack_frame_flags): This.  Also clear
21794         frame_pointer_needed if -fno-omit-frame-pointer is used without
21795         stack access.
21796         (ix86_expand_prologue): Replace ix86_finalize_stack_realign_flags
21797         with ix86_finalize_stack_frame_flags.
21798         (ix86_expand_epilogue): Likewise.
21799         (ix86_expand_split_stack_prologue): Likewise.
21800         * doc/invoke.texi: Add a note for -fno-omit-frame-pointer.
21802 2017-08-10  Martin Liska  <mliska@suse.cz>
21804         PR c++/81355
21805         * c-attribs.c (handle_target_attribute):
21806         Report warning for an empty string argument of target attribute.
21808 2017-08-09  Jakub Jelinek  <jakub@redhat.com>
21810         PR c/81687
21811         * omp-low.c (omp_copy_decl): Don't remap FORCED_LABEL or DECL_NONLOCAL
21812         LABEL_DECLs.
21813         * tree-cfg.c (move_stmt_op): Don't adjust DECL_CONTEXT of FORCED_LABEL
21814         or DECL_NONLOCAL labels.
21815         (move_stmt_r) <case GIMPLE_LABEL>: Adjust DECL_CONTEXT of FORCED_LABEL
21816         or DECL_NONLOCAL labels here.
21818 2017-08-09  Will Schmidt  <will_schmidt@vnet.ibm.com>
21820         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add blurb
21821         to indicate when early gimple folding has been disabled.
21822         (rs6000_gimple_fold_builtin): Add debug content.
21823         (rs6000_invalid_builtin): Fix whitespace.
21824         (rs6000_expand_builtin): Fix whitespace.
21825         * config/rs6000/rs6000.opt: Add option for -mfold-gimple.
21827 2017-08-09  Segher Boessenkool  <segher@kernel.crashing.org>
21829         PR target/80938
21830         * config/rs6000/rs6000.c (rs6000_savres_strategy): Don't use
21831         SAVE_MULTIPLE if not all the registers that saves, should be saved.
21833 2017-08-09  Jim Wilson  <jim.wilson@linaro.org>
21835         * config/aarch64/aarch64-cores.def (falkor): Use falkor pipeline.
21836         (qdf24xx): Likewise.
21837         * config/aarch64/aarch64.md: Include falkor.md.
21838         * config/aarch64/falkor.md: New.
21840 2017-08-09  Marek Polacek  <polacek@redhat.com>
21842         PR c/81233
21843         * diagnostic-core.h (emit_diagnostic_valist): Add declaration.
21844         * diagnostic.c (emit_diagnostic): Add a comment.
21845         (emit_diagnostic_valist): New function.
21847 2017-08-09  Marek Polacek  <polacek@redhat.com>
21849         PR c/81417
21850         * input.c (make_location): New overload.
21851         * input.h (make_location): Declare.
21853 2017-08-08  Alan Modra  <amodra@gmail.com>
21854             H.J. Lu  <hongjiu.lu@intel.com>
21856         PR driver/81523
21857         * gcc.c (NO_PIE_SPEC): Delete.
21858         (PIE_SPEC): Define as !no-pie/pie.  Move static|shared|r
21859         exclusion..
21860         (LINK_PIE_SPEC): ..to here.
21861         (LINK_COMMAND_SPEC): Support -no-pie.
21862         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Correct
21863         chain of crtbegin*.o selection, update for PIE_SPEC changes and
21864         format.
21865         (GNU_USER_TARGET_ENDFILE_SPEC): Similarly.
21866         * config/sol2.h (STARTFILE_CRTBEGIN_SPEC): Similarly.
21867         (ENDFILE_CRTEND_SPEC): Similarly.
21869 2017-08-08  Uros Bizjak  <ubizjak@gmail.com>
21871         PR target/81708
21872         * config/i386/i386.opt (mstack-protector-guard-reg=): New option
21873         (mstack-protector-guard-offset=): Ditto.
21874         * config/i386/i386.c (ix86_option_override): Handle
21875         -mstack-protector-guard-reg= and -mstack-protector-guard-offset=
21876         options.
21877         (ix86_stack_protect_guard): Use ix86_stack_protect_guard_reg and
21878         ix86_stack_protect_guard_offset variables.
21879         (TARGET_STACK_PROTECT_GUARD): Always define.
21880         * doc/invoke.texi (x86 Options): Document -mstack-protector-guard-reg=
21881         and -mstack-protector-guard-offset= options.
21883 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21885         * tree-ssa-loop-ivopts.c (relate_compare_use_with_all_cands): Handle
21886         boundary case for the last candidate.
21888 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21890         * doc/invoke.texi: Document -ftree-loop-distribution for O3.
21891         * opts.c (default_options_table): Add OPT_ftree_loop_distribution.
21893 2017-08-08  Tamar Christina  <tamar.christina@arm.com>
21895         PR middle-end/19706
21896         * config/aarch64/aarch64.md (xorsign<mode>3): New optabs.
21897         * config/aarch64/aarch64-builtins.c
21898         (aarch64_builtin_vectorized_function): Added CASE_CFN_XORSIGN.
21899         * config/aarch64/aarch64-simd-builtins.def: Added xorsign BINOP.
21900         * config/aarch64/aarch64-simd.md: Added xorsign<mode>3
21902 2017-08-08  Tamar Christina  <tamar.christina@arm.com>
21903             Andrew Pinski <pinskia@gmail.com>
21905         PR middle-end/19706
21906         * internal-fn.def (XORSIGN): New.
21907         * optabs.def (xorsign_optab): New.
21908         * tree-ssa-math-opts.c (is_copysign_call_with_1): New.
21909         (convert_expand_mult_copysign): New.
21910         (pass_optimize_widening_mul::execute): Call
21911         convert_expand_mult_copysign.
21913 2017-08-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21915         PR tree-optimization/81354
21916         * gimple-ssa-strength-reduction.c (create_add_on_incoming_edge):
21917         Insert on edges rather than explicitly creating landing pads.
21918         (analyze_candidates_and_replace): Commit edge inserts.
21920 2017-08-08  Richard Biener  <rguenther@suse.de>
21922         PR middle-end/81719
21923         * tree-ssa-loop-niter.c: Include tree-dfa.h.
21924         (expand_simple_operations): Also look through ADDR_EXPRs with
21925         MEM_REF bases treating them as POINTER_PLUS_EXPR.
21927 2017-08-08  Richard Biener  <rguenther@suse.de>
21929         PR tree-optimization/81723
21930         * tree-vect-slp.c (struct bst_traits): New hash traits.
21931         (bst_fail): New global.
21932         (vect_build_slp_tree_2): New worker, split out from ...
21933         (vect_build_slp_tree): ... this now wrapping it with using
21934         bst_fail set to cache SLP tree build fails.  Properly handle
21935         max_tree_size.
21936         (vect_analyze_slp_instance): Allocate and free bst_fail.
21938 2017-08-08  Martin Liska  <mliska@suse.cz>
21940         PR tree-opt/81696
21941         * ipa-icf-gimple.c (func_checker::compare_cst_or_decl): Consider
21942         LABEL_DECLs that can be from a different function.
21944 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21946         PR tree-optimization/81744
21947         * tree-predcom.c (prepare_finalizers_chain): Deep copy expr of
21948         loop's number of iterations.
21950 2017-08-08  Martin Liska  <mliska@suse.cz>
21952         * asan.c: Include header files.
21953         * attribs.c (build_decl_attribute_variant): New function moved
21954         from tree.[ch].
21955         (build_type_attribute_qual_variant): Likewise.
21956         (cmp_attrib_identifiers): Likewise.
21957         (simple_cst_list_equal): Likewise.
21958         (omp_declare_simd_clauses_equal): Likewise.
21959         (attribute_value_equal): Likewise.
21960         (comp_type_attributes): Likewise.
21961         (build_type_attribute_variant): Likewise.
21962         (lookup_ident_attribute): Likewise.
21963         (remove_attribute): Likewise.
21964         (merge_attributes): Likewise.
21965         (merge_type_attributes): Likewise.
21966         (merge_decl_attributes): Likewise.
21967         (merge_dllimport_decl_attributes): Likewise.
21968         (handle_dll_attribute): Likewise.
21969         (attribute_list_equal): Likewise.
21970         (attribute_list_contained): Likewise.
21971         * attribs.h (lookup_attribute): New function moved from tree.[ch].
21972         (lookup_attribute_by_prefix): Likewise.
21973         * bb-reorder.c: Include header files.
21974         * builtins.c: Likewise.
21975         * calls.c: Likewise.
21976         * cfgexpand.c: Likewise.
21977         * cgraph.c: Likewise.
21978         * cgraphunit.c: Likewise.
21979         * convert.c: Likewise.
21980         * dwarf2out.c: Likewise.
21981         * final.c: Likewise.
21982         * fold-const.c: Likewise.
21983         * function.c: Likewise.
21984         * gimple-expr.c: Likewise.
21985         * gimple-fold.c: Likewise.
21986         * gimple-pretty-print.c: Likewise.
21987         * gimple.c: Likewise.
21988         * gimplify.c: Likewise.
21989         * hsa-common.c: Likewise.
21990         * hsa-gen.c: Likewise.
21991         * internal-fn.c: Likewise.
21992         * ipa-chkp.c: Likewise.
21993         * ipa-cp.c: Likewise.
21994         * ipa-devirt.c: Likewise.
21995         * ipa-fnsummary.c: Likewise.
21996         * ipa-inline.c: Likewise.
21997         * ipa-visibility.c: Likewise.
21998         * ipa.c: Likewise.
21999         * lto-cgraph.c: Likewise.
22000         * omp-expand.c: Likewise.
22001         * omp-general.c: Likewise.
22002         * omp-low.c: Likewise.
22003         * omp-offload.c: Likewise.
22004         * omp-simd-clone.c: Likewise.
22005         * opts-global.c: Likewise.
22006         * passes.c: Likewise.
22007         * predict.c: Likewise.
22008         * sancov.c: Likewise.
22009         * sanopt.c: Likewise.
22010         * symtab.c: Likewise.
22011         * toplev.c: Likewise.
22012         * trans-mem.c: Likewise.
22013         * tree-chkp.c: Likewise.
22014         * tree-eh.c: Likewise.
22015         * tree-into-ssa.c: Likewise.
22016         * tree-object-size.c: Likewise.
22017         * tree-parloops.c: Likewise.
22018         * tree-profile.c: Likewise.
22019         * tree-ssa-ccp.c: Likewise.
22020         * tree-ssa-live.c: Likewise.
22021         * tree-ssa-loop.c: Likewise.
22022         * tree-ssa-sccvn.c: Likewise.
22023         * tree-ssa-structalias.c: Likewise.
22024         * tree-ssa.c: Likewise.
22025         * tree-streamer-in.c: Likewise.
22026         * tree-vectorizer.c: Likewise.
22027         * tree-vrp.c: Likewise.
22028         * tsan.c: Likewise.
22029         * ubsan.c: Likewise.
22030         * varasm.c: Likewise.
22031         * varpool.c: Likewise.
22032         * tree.c: Remove functions moved to attribs.[ch].
22033         * tree.h: Likewise.
22034         * config/aarch64/aarch64.c: Add attrs.h header file.
22035         * config/alpha/alpha.c: Likewise.
22036         * config/arc/arc.c: Likewise.
22037         * config/arm/arm.c: Likewise.
22038         * config/avr/avr.c: Likewise.
22039         * config/bfin/bfin.c: Likewise.
22040         * config/c6x/c6x.c: Likewise.
22041         * config/cr16/cr16.c: Likewise.
22042         * config/cris/cris.c: Likewise.
22043         * config/darwin.c: Likewise.
22044         * config/epiphany/epiphany.c: Likewise.
22045         * config/fr30/fr30.c: Likewise.
22046         * config/frv/frv.c: Likewise.
22047         * config/ft32/ft32.c: Likewise.
22048         * config/h8300/h8300.c: Likewise.
22049         * config/i386/winnt.c: Likewise.
22050         * config/ia64/ia64.c: Likewise.
22051         * config/iq2000/iq2000.c: Likewise.
22052         * config/lm32/lm32.c: Likewise.
22053         * config/m32c/m32c.c: Likewise.
22054         * config/m32r/m32r.c: Likewise.
22055         * config/m68k/m68k.c: Likewise.
22056         * config/mcore/mcore.c: Likewise.
22057         * config/microblaze/microblaze.c: Likewise.
22058         * config/mips/mips.c: Likewise.
22059         * config/mmix/mmix.c: Likewise.
22060         * config/mn10300/mn10300.c: Likewise.
22061         * config/moxie/moxie.c: Likewise.
22062         * config/msp430/msp430.c: Likewise.
22063         * config/nds32/nds32-isr.c: Likewise.
22064         * config/nds32/nds32.c: Likewise.
22065         * config/nios2/nios2.c: Likewise.
22066         * config/nvptx/nvptx.c: Likewise.
22067         * config/pa/pa.c: Likewise.
22068         * config/pdp11/pdp11.c: Likewise.
22069         * config/powerpcspe/powerpcspe.c: Likewise.
22070         * config/riscv/riscv.c: Likewise.
22071         * config/rl78/rl78.c: Likewise.
22072         * config/rx/rx.c: Likewise.
22073         * config/s390/s390.c: Likewise.
22074         * config/sh/sh.c: Likewise.
22075         * config/sol2.c: Likewise.
22076         * config/sparc/sparc.c: Likewise.
22077         * config/spu/spu.c: Likewise.
22078         * config/stormy16/stormy16.c: Likewise.
22079         * config/tilegx/tilegx.c: Likewise.
22080         * config/tilepro/tilepro.c: Likewise.
22081         * config/v850/v850.c: Likewise.
22082         * config/vax/vax.c: Likewise.
22083         * config/visium/visium.c: Likewise.
22084         * config/xtensa/xtensa.c: Likewise.
22086 2017-08-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
22088         PR target/81593
22089         * config/rs6000/vsx.md (vsx_concat_<mode>, VSX_D): Cleanup
22090         constraints since the -mupper-regs-* switches have been
22091         eliminated.
22092         (vsx_concat_<mode>_1): New combiner insns to recognize inserting
22093         into a vector from a double word element that was extracted from
22094         another vector, and eliminate extra XXPERMDI instructions.
22095         (vsx_concat_<mode>_2): Likewise.
22096         (vsx_concat_<mode>_3): Likewise.
22097         (vsx_set_<mode>, VSX_D): Rewrite vector set in terms of vector
22098         concat to allow optimizing inserts from previous extracts.
22100 2017-08-07  Uros Bizjak  <ubizjak@gmail.com>
22102         * config/i386/i386.c (ix86_stack_protect_guard): Generate
22103         memory reference to a SSP offset in TLS address space.
22104         (ix86_print_operand) <case '@'>: Remove.
22105         (ix86_print_operand_punct_valid_p): Remove '@' code.
22106         * config/i386/i386.md (unspec): Remove UNSPEC_SP_TLS_SET and
22107         UNSPEC_SP_TLS_TEST.
22108         (stack_tls_protect_set_<mode>): Remove.
22109         (stack_protect_set): Do not call gen_stack_tls_protect_set_<mode>.
22110         (stack_tls_protect_test_<mode>): Remove.
22111         (stack_protect_test): Do not call gen_stack_tls_protect_test_<mode>.
22113 2017-08-07  Olivier Hainque  <hainque@adacore.com>
22115         PR target/81755
22116         * config/vxworksae.h (VXWORKS_HAVE_TLS): Define.
22118 2017-08-07  Douglas Rupp  <rupp@adacore.com>
22120         * Makefile.in (install-mkheaders): Fix typo, where the multi_dir
22121         variable was referenced as multidir in command.
22123 2017-08-07  Jakub Jelinek  <jakub@redhat.com>
22125         PR c/69389
22126         * gimplify.c (goa_stabilize_expr): Handle BIT_INSERT_EXPR and
22127         BIT_FIELD_REF.
22129 2017-08-07  Martin Liska  <mliska@suse.cz>
22131         * config/m32c/m32c.c: Add include of stringpool.h and attribs.h.
22132         * config/rl78/rl78.c: Add include of attribs.h.
22133         * config/sh/sh.c: Likewise.
22134         * config/v850/v850.c: Likewise.
22136 2017-08-07  Tom de Vries  <tom@codesourcery.com>
22138         PR middle-end/78266
22139         * omp-expand.c (expand_oacc_for): Ensure diff_type is large enough.
22141 2017-08-07  Martin Liska  <mliska@suse.cz>
22143         * config/mips/mips.c: Include attribs.h.
22145 2017-08-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
22147         PR fortran/68829
22148         * doc/invoke.texi: Document change in behvaior for -Ofast for
22149         Fortran.
22151 2017-08-07  Wilco Dijkstra  <wdijkstr@arm.com>
22153         * config/aarch64/aarch64.c (aarch64_pushwb_single_reg):
22154         Use gen_frame_mem.
22155         (aarch64_pop_regs): Likewise.
22156         (aarch64_gen_load_pair): Likewise.
22157         (aarch64_save_callee_saves): Likewise.
22158         (aarch64_restore_callee_saves): Likewise.
22160 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
22162         * config/i386/i386.c: Revert the last change.
22164 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
22166         PR target/81736
22167         * config/i386/i386.c (ix86_finalize_stack_realign_flags): Renamed
22168         to ...
22169         (ix86_finalize_stack_frame_flags): This.  Also clear
22170         frame_pointer_needed if -fno-omit-frame-pointer is used without
22171         stack access.
22172         (ix86_expand_prologue): Replace ix86_finalize_stack_realign_flags
22173         with ix86_finalize_stack_frame_flags.
22174         (ix86_expand_epilogue): Likewise.
22175         (ix86_expand_split_stack_prologue): Likewise.
22177 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
22179         PR target/81743
22180         * config/i386/i386.c (get_builtin_code_for_version): Set priority
22181         to P_AES for Westmere.
22183 2017-08-07  Jonathan Yong  <10walls@gmail.com>
22185         * config/i386/mingw.opt (fset-stack-executable): Removed.
22186         * config/i386/cygming.opt (fset-stack-executable): Moved
22187         from mingw.opt.
22188         * config/i386/cygwin.h: Define CHECK_EXECUTE_STACK_ENABLED.
22190 2017-08-07  Segher Boessenkool  <segher@kernel.crashing.org>
22192         * print-rtl.c (print_exp): Print NOT as "~" instead of as "!".
22194 2017-08-07  Marek Polacek  <polacek@redhat.com>
22196         PR middle-end/81737
22197         * fold-const.c (fold_indirect_ref_1): Check type_domain.
22199 2017-08-07  Martin Liska  <mliska@suse.cz>
22201         * attribs.h (canonicalize_attr_name): New function.
22202         (cmp_attribs): Move from c-format.c and adjusted.
22203         (is_attribute_p): Moved from tree.h.
22204         * tree-inline.c: Add new includes.
22205         * tree.c (cmp_attrib_identifiers): Use cmp_attribs.
22206         (private_is_attribute_p): Remove.
22207         (private_lookup_attribute): Likewise.
22208         (private_lookup_attribute_by_prefix): Simplify.
22209         (remove_attribute): Use is_attribute_p.
22210         * tree.h: Remove removed declarations.
22212 2017-08-07  Jakub Jelinek  <jakub@redhat.com>
22214         PR middle-end/81698
22215         * stmt.c (emit_case_dispatch_table): Add DEFAULT_EDGE argument,
22216         instead of computing it in the function.  Formatting fix.
22217         (expand_case): Don't rely on default_edge being the first edge,
22218         clear it if removing it, pass default_edge to
22219         emit_case_dispatch_table.
22220         (expand_sjlj_dispatch_table): Pass NULL as DEFAULT_EDGE, formatting
22221         fix.
22223 2017-08-06  Uros Bizjak  <ubizjak@gmail.com>
22225         * config/alpha/alpha.c (alpha_reorg): If trap is the last active
22226         insn in the function, emit NOP after the insn.
22228 2017-08-06  Tom de Vries  <tom@codesourcery.com>
22230         * omp-expand.c (expand_oacc_for): Add missing edge probability for tile
22231         and element loops.
22233 2017-08-06  Tom de Vries  <tom@codesourcery.com>
22235         * omp-expand.c (expand_oacc_for): Add missing edge probability for chunk
22236         loop.
22238 2017-08-04  Yury Gribov  <tetra2005@gmail.com>
22240         PR tree-optimization/57371
22241         * match.pd: New pattern.
22243 2017-08-04  Marek Polacek  <polacek@redhat.com>
22245         PR middle-end/81695
22246         * fold-const.c (fold_indirect_ref_1): For ((int *)&a + 4 -> a[1],
22247         perform the computation in offset_int.
22249 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22251         PR tree-optimization/81136
22252         * tree-vectorizer.h: Include tree-hash-traits.h.
22253         (vec_base_alignments): New typedef.
22254         (vec_info): Add a base_alignments field.
22255         (vect_record_base_alignments): Declare.
22256         * tree-data-ref.h (data_reference): Add an is_conditional_in_stmt
22257         field.
22258         (DR_IS_CONDITIONAL_IN_STMT): New macro.
22259         (create_data_ref): Add an is_conditional_in_stmt argument.
22260         * tree-data-ref.c (create_data_ref): Likewise.  Use it to initialize
22261         the is_conditional_in_stmt field.
22262         (data_ref_loc): Add an is_conditional_in_stmt field.
22263         (get_references_in_stmt): Set the is_conditional_in_stmt field.
22264         (find_data_references_in_stmt): Update call to create_data_ref.
22265         (graphite_find_data_references_in_stmt): Likewise.
22266         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Likewise.
22267         * tree-vect-data-refs.c (vect_analyze_data_refs): Likewise.
22268         (vect_record_base_alignment): New function.
22269         (vect_record_base_alignments): Likewise.
22270         (vect_compute_data_ref_alignment): Adjust base_addr and aligned_to
22271         for nested statements even if we fail to compute a misalignment.
22272         Use pooled base alignments for unconditional references.
22273         (vect_find_same_alignment_drs): Compare base addresses instead
22274         of base objects.
22275         (vect_analyze_data_refs_alignment): Call vect_record_base_alignments.
22276         * tree-vect-slp.c (vect_slp_analyze_bb_1): Likewise.
22278 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22280         * tree-vectorizer.h (vec_info): Add a constructor and destructor.
22281         Add an explicit name for the enum.  Use auto_vec for slp_instances
22282         and grouped_stores.
22283         (_loop_vec_info): Add a constructor and destructor.  Use auto_vec
22284         for all vectors.
22285         (_bb_vec_info): Add a constructor and destructor.
22286         (vinfo_for_stmt): Return NULL for uids of -1 as well.
22287         (destroy_loop_vec_info): Delete.
22288         (vect_destroy_datarefs): Likewise.
22289         * tree-vectorizer.c (vect_destroy_datarefs): Delete.
22290         (vec_info::vec_info): New function.
22291         (vec_info::~vec_info): Likewise.
22292         (vectorize_loops): Use delete instead of destroy_loop_vec_info.
22293         * tree-parloops.c (gather_scalar_reductions): Use delete instead of
22294         destroy_loop_vec_info.
22295         * tree-vect-loop.c (new_loop_vec_info): Replace with...
22296         (_loop_vec_info::_loop_vec_info): ...this.
22297         (destroy_loop_vec_info): Replace with...
22298         (_loop_vec_info::~_loop_vec_info): ...this.  Unconditionally delete
22299         the stmt_vec_infos.  Leave handling of vec_info information to its
22300         destructor.  Remove explicit vector releases.
22301         (vect_analyze_loop_form): Use new instead of new_loop_vec_info.
22302         (vect_analyze_loop): Use delete instead of destroy_loop_vec_info.
22303         * tree-vect-slp.c (new_bb_vec_info): Replace with...
22304         (_bb_vec_info::_bb_vec_info): ...this.  Don't reserve space in
22305         BB_VINFO_GROUPED_STORES or BB_VINFO_SLP_INSTANCES.
22306         (destroy_bb_vec_info): Replace with...
22307         (_bb_vec_info::~_bb_vec_info): ...this.  Leave handling of vec_info
22308         information to its destructor.
22309         (vect_slp_analyze_bb_1): Use new and delete instead of
22310         new_bb_vec_info and destroy_bb_vec_info.
22311         (vect_slp_bb): Replace 2 calls to destroy_bb_vec_info with a
22312         single delete.
22314 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22316         * tree-data-ref.h (subscript): Add access_fn field.
22317         (data_dependence_relation): Add could_be_independent_p.
22318         (SUB_ACCESS_FN, DDR_COULD_BE_INDEPENDENT_P): New macros.
22319         (same_access_functions): Move to tree-data-ref.c.
22320         * tree-data-ref.c (ref_contains_union_access_p): New function.
22321         (access_fn_component_p): Likewise.
22322         (access_fn_components_comparable_p): Likewise.
22323         (dr_analyze_indices): Add a reference to access_fn_component_p.
22324         (dump_data_dependence_relation): Use SUB_ACCESS_FN instead of
22325         DR_ACCESS_FN.
22326         (constant_access_functions): Likewise.
22327         (add_other_self_distances): Likewise.
22328         (same_access_functions): Likewise.  (Moved from tree-data-ref.h.)
22329         (initialize_data_dependence_relation): Use XCNEW and remove
22330         explicit zeroing of DDR_REVERSED_P.  Look for a subsequence
22331         of access functions that have the same type.  Allow the
22332         subsequence to end with different bases in some circumstances.
22333         Record the chosen access functions in SUB_ACCESS_FN.
22334         (build_classic_dist_vector_1): Replace ddr_a and ddr_b with
22335         a_index and b_index.  Use SUB_ACCESS_FN instead of DR_ACCESS_FN.
22336         (subscript_dependence_tester_1): Likewise dra and drb.
22337         (build_classic_dist_vector): Update calls accordingly.
22338         (subscript_dependence_tester): Likewise.
22339         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Check
22340         DDR_COULD_BE_INDEPENDENT_P.
22341         * tree-vectorizer.h (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Test
22342         comp_alias_ddrs instead of may_alias_ddrs.
22343         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
22344         New function.
22345         (vect_analyze_data_ref_dependence): Use it if
22346         DDR_COULD_BE_INDEPENDENT_P, but fall back to using the recorded
22347         distance vectors if that fails.
22348         (dependence_distance_ge_vf): New function.
22349         (vect_prune_runtime_alias_test_list): Use it.  Don't clear
22350         LOOP_VINFO_MAY_ALIAS_DDRS.
22352 2017-08-04  Richard Biener  <rguenther@suse.de>
22354         PR middle-end/81705
22355         * fold-const.c (fold_binary_loc): Properly restrict
22356         minus_var0 && minus_var1 case when associating undefined overflow
22357         entities.
22359 2017-08-04  Tom de Vries  <tom@codesourcery.com>
22361         * omp-simd-clone.c (simd_clone_adjust): Add missing edge probability.
22363 2017-08-03  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
22365         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
22366         Don't start diagnostic messages with a capital letter.
22367         * config/rs6000/rs6000.c (rs6000_option_override_internal):
22368         Likewise.
22369         (rs6000_invalid_builtin): Likewise.
22370         (rs6000_trampoline_init): Likewise.
22372 2017-08-03  Jakub Jelinek  <jakub@redhat.com>
22374         PR target/81621
22375         * bb-reorder.c (pass_partition_blocks::execute): Return TODO_df_finish
22376         after setting changeable df flags.
22378 2017-08-03  Richard Biener  <rguenther@suse.de>
22380         * tree-ssa-reassoc.c (should_break_up_subtract): Also break
22381         up if the use is in USE - X.
22383 2017-08-03  Alexander Monakov  <amonakov@ispras.ru>
22385         * toplev.c (dumpfile.h): New include.
22386         (internal_error_reentered): New static function.  Use it...
22387         (internal_error_function): ...here to handle reentered internal_error.
22389 2017-08-03  Richard Biener  <rguenther@suse.de>
22391         PR middle-end/81148
22392         * fold-const.c (split_tree): Add minus_var and minus_con
22393         arguments, remove unused loc arg.  Never generate NEGATE_EXPRs
22394         here but always use minus_*.
22395         (associate_trees): Assert we never associate with MINUS_EXPR
22396         and NULL first operand.  Do not recurse for PLUS_EXPR operands
22397         when associating as MINUS_EXPR either.
22398         (fold_binary_loc): Track minus_var and minus_con.
22400 2017-08-03  Tom de Vries  <tom@codesourcery.com>
22402         PR lto/81430
22403         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers): If
22404         ACCEL_COMPILER, apply finish_options on
22405         DECL_FUNCTION_SPECIFIC_OPTIMIZATION.
22407 2017-08-03  Tom de Vries  <tom@codesourcery.com>
22409         PR target/81662
22410         * config/nvptx/nvptx.c (nvptx_option_override): Emit sorry if
22411         function_entry_patch_area_size > 0.
22413 2017-08-03  Jakub Jelinek  <jakub@redhat.com>
22415         PR driver/81650
22416         * calls.c (alloc_max_size): Use HOST_WIDE_INT_UC (10??)
22417         instead of 10??LU, perform unit multiplication in wide_int,
22418         don't change alloc_object_size_limit if the limit is larger
22419         than SSIZE_MAX.
22421         PR tree-optimization/81655
22422         PR tree-optimization/81588
22423         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Handle also
22424         the case when ranges[i].low and high are 1 for unsigned type with
22425         precision 1.
22427         PR middle-end/81052
22428         * omp-low.c (diagnose_sb_0): Handle flag_openmp_simd like flag_openmp.
22429         (pass_diagnose_omp_blocks::gate): Enable also for flag_openmp_simd.
22431 2017-08-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
22433         * tree-vrp.h: Add include guard.
22435 2017-08-02  Uros Bizjak  <ubizjak@gmail.com>
22437         PR target/81644
22438         * config/i386/i386.md (unspecv): Add UNSPECV_UD2.
22439         (ud2): New insn pattern.
22440         * config/i386/i386.c (ix86_expand_epilogue):
22441         For naked functions, generate ud2 instead of trap insn.
22443 2017-08-02  Marek Polacek  <polacek@redhat.com>
22445         PR other/81667
22446         * alloc-pool.h (base_pool_allocator): Initialize m_elt_size.
22448 2017-08-02  Tom de Vries  <tom@codesourcery.com>
22449             Cesar Philippidis  <cesar@codesourcery.com>
22451         * config/nvptx/nvptx.c (nvptx_lockless_update, nvptx_lockfull_update):
22452         Add missing edge probabilities.
22454 2017-08-02  Tamar Christina  <tamar.christina@arm.com>
22456         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
22457         Correct endianness.
22459 2017-08-02  Jakub Jelinek  <jakub@redhat.com>
22461         PR middle-end/79499
22462         * function.c (thread_prologue_and_epilogue_insns): Determine blocks
22463         for find_many_sub_basic_blocks bitmap by looking up BLOCK_FOR_INSN
22464         of first NONDEBUG_INSN_P in each of the split_prologue_seq and
22465         prologue_seq sequences - if any.
22467 2017-08-02  Richard Biener  <rguenther@suse.de>
22469         * tree-vect-stmts.c (vectorizable_store): Perform vector extracts
22470         via vectors if supported, integer extracts via punning if supported
22471         or otherwise vector extracts.
22473 2017-08-02  Richard Biener  <rguenther@suse.de>
22475         * tree-ssa-pre.c (bitmap_insert_into_set_1): Remove and inline
22476         into ...
22477         (bitmap_insert_into_set): ... this.
22479 2017-08-02  Richard Biener  <rguenther@suse.de>
22481         PR tree-optimization/81633
22482         Revert
22483         2015-08-17  Alan Hayward  <alan.hayward@arm.com>
22485         PR tree-optimization/71752
22486         * tree-vect-slp.c (vect_get_slp_defs): Handle null operands.
22488 2017-08-01  Daniel Santos  <daniel.santos@pobox.com>
22490         * config/i386/i386.h (ix86_frame::outlined_save_offset): Remove field.
22491         (machine_function::call_ms2sysv_pad_out): Remove field.
22492         * config/i386/i386.c (xlogue_layout::get_stack_space_used): Modify.
22493         (ix86_compute_frame_layout): Likewise.
22495 2017-08-01  H.J. Lu  <hongjiu.lu@intel.com>
22497         PR target/81654
22498         * config/i386/i386.c (ix86_set_func_type): Disallow naked
22499         attribute with interrupt attribute.
22501 2017-08-01  Andrew Pinski  <apinski@cavium.com>
22503         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Check
22504         BIT_INSERT_EXPR's operand 1
22505         to see if the types precision matches.
22507 2017-08-01  Martin Liska  <mliska@suse.cz>
22509         PR middle-end/70140
22510         * builtins.c (expand_builtin_memcpy_args): Remove.
22511         (expand_builtin_memcpy): Call newly added function
22512         expand_builtin_memory_copy_args.
22513         (expand_builtin_memcpy_with_bounds): Likewise.
22514         (expand_builtin_mempcpy): Remove last argument.
22515         (expand_builtin_mempcpy_with_bounds): Likewise.
22516         (expand_builtin_memory_copy_args): New function created from
22517         expand_builtin_mempcpy_args with small modifications.
22518         (expand_builtin_mempcpy_args): Remove.
22519         (expand_builtin_stpcpy): Remove unused argument.
22520         (expand_builtin): Likewise.
22521         (expand_builtin_with_bounds): Likewise.
22523 2017-08-01  Martin Liska  <mliska@suse.cz>
22525         Revert r250771
22526         Make mempcpy more optimal (PR middle-end/70140).
22528 2017-08-01  Jakub Jelinek  <jakub@redhat.com>
22530         PR target/81622
22531         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): For
22532         __builtin_vec_cmpne verify both arguments are compatible vectors
22533         before looking at TYPE_MODE on the element type.  For __builtin_vec_ld
22534         verify arg1_type is a pointer or array type.  For __builtin_vec_st,
22535         move computation of aligned to after checking the argument types.
22536         Formatting fixes.
22538         PR target/80846
22539         * config/rs6000/vsx.md (vextract_fp_from_shorth,
22540         vextract_fp_from_shortl): Add element mode after mode in gen_vec_init*
22541         calls.
22543 2017-08-01  Jerome Lambourg  <lambourg@adacore.com>
22544             Doug Rupp  <rupp@adacore.com>
22545             Olivier Hainque  <hainque@adacore.com>
22547         * config.gcc (arm-wrs-vxworks*): Rework to handle arm-wrs-vxworks7 as
22548         well as arm-wrs-vxworks. Update target_cpu_name from arm6 (arch v3) to
22549         arm8 (arch v4).
22550         * config/arm/vxworks.h (MAYBE_TARGET_BPABI_CPP_BUILTINS): New, helper
22551         for TARGET_OS_CPP_BUILTIN.
22552         (TARGET_OS_CPP_BUILTIN): Invoke MAYBE_TARGET_BPABI_CPP_BUILTINS(),
22553         refine CPU definitions for arm_arch5 and add those for arm_arch6 and
22554         arm_arch7.
22555         (MAYBE_ASM_ABI_SPEC): New, helper for SUBTARGET_EXTRA_ASM_SPEC,
22556         passing required abi options to the assembler for EABI configurations.
22557         (EXTRA_CC1_SPEC): New macro, to help prevent the implicit production
22558         of .text.hot and .text.unlikely sections for kernel modules when
22559         using ARM style exceptions.
22560         (CC1_SPEC): Remove obsolete attempt at mimicking Diab toolchain
22561         options. Add EXTRA_CC1_SPEC.
22562         (VXWORKS_ENDIAN_SPEC): Adjust comment and remove handling of Diab
22563         toolchain options.
22564         (DWARF2_UNWIND_INFO): Redefine to handle the pre/post VxWorks 7
22565         transition.
22566         (ARM_TARGET2_DWARF_FORMAT): Define.
22567         * config/arm/t-vxworks: Adjust multilib control to removal of the
22568         Diab command line options.
22570 2017-08-01  Martin Liska  <mliska@suse.cz>
22572         PR gcov-profile/81561
22573         * gcov.c (unblock): Make unblocking safe as we need to preserve
22574         index correspondence of blocks and block_lists.
22576 2017-08-01  Richard Biener  <rguenther@suse.de>
22578         PR tree-optimization/81181
22579         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
22580         (compute_antic): ... end of iteration here.
22582 2017-08-01  James Greenhalgh  <james.greenhalgh@arm.com>
22584         * common.opt (ftree-vectorize): No longer set flag_tree_vectorize.
22585         (ftree-loop-vectorize): Set as EnabledBy ftree-vectorize.
22586         (ftree-slp-vectorize): Likewise.
22587         * omp-expand (expand_omp_simd): Remove flag_tree_vectorize, as it
22588         can no longer be set independent of flag_tree_loop_vectorize.
22589         * omp-general.c (emp_max_vf): Likewise.
22590         * opts.c (enable_fdo_optimizations): Remove references to
22591         flag_tree_vectorize, these are now implicit.
22592         (common_handle_option): Remove handling for OPT_ftree_vectorize,
22593         and leave it for the options machinery.
22595 2017-08-01  Martin Liska  <mliska@suse.cz>
22597         PR middle-end/70140
22598         * builtins.c (expand_builtin_memcpy_args): Remove.
22599         (expand_builtin_memcpy): Call newly added function
22600         expand_builtin_memory_copy_args.
22601         (expand_builtin_memcpy_with_bounds): Likewise.
22602         (expand_builtin_mempcpy): Remove last argument.
22603         (expand_builtin_mempcpy_with_bounds): Likewise.
22604         (expand_builtin_memory_copy_args): New function created from
22605         expand_builtin_mempcpy_args with small modifications.
22606         (expand_builtin_mempcpy_args): Remove.
22607         (expand_builtin_stpcpy): Remove unused argument.
22608         (expand_builtin): Likewise.
22609         (expand_builtin_with_bounds): Likewise.
22611 2017-08-01  Uros Bizjak  <ubizjak@gmail.com>
22613         PR target/81641
22614         * config/i386/i386.c (ix86_print_operand_address_as): For -masm=intel
22615         print "ds:" only for immediates in generic address space.
22617 2017-08-01  Uros Bizjak  <ubizjak@gmail.com>
22619         PR target/81639
22620         * config/i386/i386.c (ix86_funciton_naked): New prototype.
22621         (ix86_function_ok_for_sibcall): Return false for naked functions.
22623 2017-08-01  Richard Biener  <rguenther@suse.de>
22625         * tree-ssa-pre.c (print_pre_expr): Handle NULL expr.
22626         (compute_antic): Seed worklist with exit block predecessors.
22627         * cfganal.c (dfs_find_deadend): For a cycle return the source
22628         of the edge closing it.
22630 2017-08-01  Tamar Christina  <tamar.christina@arm.com>
22632         * config/aarch64/aarch64.c
22633         (aarch64_can_const_movi_rtx_p): Move 0 check.
22635 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22637         * tree.h (POINTER_TYPE_OVERFLOW_UNDEFINED): Delete.
22638         * fold-const.c (fold_comparison, fold_binary_loc): Delete use of
22639         above macro.
22640         * match.pd: Ditto in address comparison pattern.
22642 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22644         PR tree-optimization/81627
22645         * tree-predcom.c (prepare_finalizers): Always rewrite into loop
22646         closed ssa form for store-store chain.
22648 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22650         PR tree-optimization/81620
22651         * tree-predcom.c (add_ref_to_chain): Don't set has_max_use_after
22652         for store-store chain.
22654 2017-08-01  Jakub Jelinek  <jakub@redhat.com>
22656         PR tree-optimization/81588
22657         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If
22658         ranges[i].in_p, invert comparison code ccode.  For >/>=,
22659         swap rhs1 and rhs2 and comparison code unconditionally,
22660         for </<= don't do that.  Don't swap rhs1/rhs2 again if
22661         ranges[i].in_p, instead invert comparison code ccode if
22662         opcode or oe->rank is BIT_IOR_EXPR.
22664         PR target/80846
22665         * optabs.def (vec_extract_optab, vec_init_optab): Change from
22666         a direct optab to conversion optab.
22667         * optabs.c (expand_vector_broadcast): Use convert_optab_handler
22668         with GET_MODE_INNER as last argument instead of optab_handler.
22669         * expmed.c (extract_bit_field_1): Likewise.  Use vector from
22670         vector extraction if possible and optab is available.
22671         * expr.c (store_constructor): Use convert_optab_handler instead
22672         of optab_handler.  Use vector initialization from smaller
22673         vectors if possible and optab is available.
22674         * tree-vect-stmts.c (vectorizable_load): Likewise.
22675         * doc/md.texi (vec_extract, vec_init): Document that the optabs
22676         now have two modes.
22677         * config/i386/i386.c (ix86_expand_vector_init): Handle expansion
22678         of vec_init from half-sized vectors with the same element mode.
22679         * config/i386/sse.md (ssehalfvecmode): Add V4TI case.
22680         (ssehalfvecmodelower, ssescalarmodelower): New mode attributes.
22681         (reduc_plus_scal_v8df, reduc_plus_scal_v4df, reduc_plus_scal_v2df,
22682         reduc_plus_scal_v16sf, reduc_plus_scal_v8sf, reduc_plus_scal_v4sf,
22683         reduc_<code>_scal_<mode>, reduc_umin_scal_v8hi): Add element mode
22684         after mode in gen_vec_extract* calls.
22685         (vec_extract<mode>): Renamed to ...
22686         (vec_extract<mode><ssescalarmodelower>): ... this.
22687         (vec_extract<mode><ssehalfvecmodelower>): New expander.
22688         (rotl<mode>3, rotr<mode>3, <shift_insn><mode>3, ashrv2di3): Add
22689         element mode after mode in gen_vec_init* calls.
22690         (VEC_INIT_HALF_MODE): New mode iterator.
22691         (vec_init<mode>): Renamed to ...
22692         (vec_init<mode><ssescalarmodelower>): ... this.
22693         (vec_init<mode><ssehalfvecmodelower>): New expander.
22694         * config/i386/mmx.md (vec_extractv2sf): Renamed to ...
22695         (vec_extractv2sfsf): ... this.
22696         (vec_initv2sf): Renamed to ...
22697         (vec_initv2sfsf): ... this.
22698         (vec_extractv2si): Renamed to ...
22699         (vec_extractv2sisi): ... this.
22700         (vec_initv2si): Renamed to ...
22701         (vec_initv2sisi): ... this.
22702         (vec_extractv4hi): Renamed to ...
22703         (vec_extractv4hihi): ... this.
22704         (vec_initv4hi): Renamed to ...
22705         (vec_initv4hihi): ... this.
22706         (vec_extractv8qi): Renamed to ...
22707         (vec_extractv8qiqi): ... this.
22708         (vec_initv8qi): Renamed to ...
22709         (vec_initv8qiqi): ... this.
22710         * config/rs6000/vector.md (VEC_base_l): New mode attribute.
22711         (vec_init<mode>): Renamed to ...
22712         (vec_init<mode><VEC_base_l>): ... this.
22713         (vec_extract<mode>): Renamed to ...
22714         (vec_extract<mode><VEC_base_l>): ... this.
22715         * config/rs6000/paired.md (vec_initv2sf): Renamed to ...
22716         (vec_initv2sfsf): ... this.
22717         * config/rs6000/altivec.md (splitter, altivec_copysign_v4sf3,
22718         vec_unpacku_hi_v16qi, vec_unpacku_hi_v8hi, vec_unpacku_lo_v16qi,
22719         vec_unpacku_lo_v8hi, mulv16qi3, altivec_vreve<mode>2): Add
22720         element mode after mode in gen_vec_init* calls.
22721         * config/aarch64/aarch64-simd.md (vec_init<mode>): Renamed to ...
22722         (vec_init<mode><Vel>): ... this.
22723         (vec_extract<mode>): Renamed to ...
22724         (vec_extract<mode><Vel>): ... this.
22725         * config/aarch64/iterators.md (Vel): New mode attribute.
22726         * config/s390/s390.c (s390_expand_vec_strlen, s390_expand_vec_movstr):
22727         Add element mode after mode in gen_vec_extract* calls.
22728         * config/s390/vector.md (non_vec_l): New mode attribute.
22729         (vec_extract<mode>): Renamed to ...
22730         (vec_extract<mode><non_vec_l>): ... this.
22731         (vec_init<mode>): Renamed to ...
22732         (vec_init<mode><non_vec_l>): ... this.
22733         * config/s390/s390-builtins.def (s390_vlgvb, s390_vlgvh, s390_vlgvf,
22734         s390_vlgvf_flt, s390_vlgvg, s390_vlgvg_dbl): Add element mode after
22735         vec_extract mode.
22736         * config/arm/iterators.md (V_elem_l): New mode attribute.
22737         * config/arm/neon.md (vec_extract<mode>): Renamed to ...
22738         (vec_extract<mode><V_elem_l>): ... this.
22739         (vec_extractv2di): Renamed to ...
22740         (vec_extractv2didi): ... this.
22741         (vec_init<mode>): Renamed to ...
22742         (vec_init<mode><V_elem_l>): ... this.
22743         (reduc_plus_scal_<mode>, reduc_plus_scal_v2di, reduc_smin_scal_<mode>,
22744         reduc_smax_scal_<mode>, reduc_umin_scal_<mode>,
22745         reduc_umax_scal_<mode>, neon_vget_lane<mode>, neon_vget_laneu<mode>):
22746         Add element mode after gen_vec_extract* calls.
22747         * config/mips/mips-msa.md (vec_init<mode>): Renamed to ...
22748         (vec_init<mode><unitmode>): ... this.
22749         (vec_extract<mode>): Renamed to ...
22750         (vec_extract<mode><unitmode>): ... this.
22751         * config/mips/loongson.md (vec_init<mode>): Renamed to ...
22752         (vec_init<mode><unitmode>): ... this.
22753         * config/mips/mips-ps-3d.md (vec_initv2sf): Renamed to ...
22754         (vec_initv2sfsf): ... this.
22755         (vec_extractv2sf): Renamed to ...
22756         (vec_extractv2sfsf): ... this.
22757         (reduc_plus_scal_v2sf, reduc_smin_scal_v2sf, reduc_smax_scal_v2sf):
22758         Add element mode after gen_vec_extract* calls.
22759         * config/mips/mips.md (unitmode): New mode iterator.
22760         * config/spu/spu.c (spu_expand_prologue, spu_allocate_stack,
22761         spu_builtin_extract): Add element mode after gen_vec_extract* calls.
22762         * config/spu/spu.md (inner_l): New mode attribute.
22763         (vec_init<mode>): Renamed to ...
22764         (vec_init<mode><inner_l>): ... this.
22765         (vec_extract<mode>): Renamed to ...
22766         (vec_extract<mode><inner_l>): ... this.
22767         * config/sparc/sparc.md (veltmode): New mode iterator.
22768         (vec_init<VMALL:mode>): Renamed to ...
22769         (vec_init<VMALL:mode><VMALL:veltmode>): ... this.
22770         * config/ia64/vect.md (vec_initv2si): Renamed to ...
22771         (vec_initv2sisi): ... this.
22772         (vec_initv2sf): Renamed to ...
22773         (vec_initv2sfsf): ... this.
22774         (vec_extractv2sf): Renamed to ...
22775         (vec_extractv2sfsf): ... this.
22776         * config/powerpcspe/vector.md (VEC_base_l): New mode attribute.
22777         (vec_init<mode>): Renamed to ...
22778         (vec_init<mode><VEC_base_l>): ... this.
22779         (vec_extract<mode>): Renamed to ...
22780         (vec_extract<mode><VEC_base_l>): ... this.
22781         * config/powerpcspe/paired.md (vec_initv2sf): Renamed to ...
22782         (vec_initv2sfsf): ... this.
22783         * config/powerpcspe/altivec.md (splitter, altivec_copysign_v4sf3,
22784         vec_unpacku_hi_v16qi, vec_unpacku_hi_v8hi, vec_unpacku_lo_v16qi,
22785         vec_unpacku_lo_v8hi, mulv16qi3): Add element mode after mode in
22786         gen_vec_init* calls.
22788 2017-08-01  Richard Biener  <rguenther@suse.de>
22790         PR tree-optimization/81297
22791         * tree-vrp.c (get_single_symbol): Remove assert, instead drop
22792         TREE_OVERFLOW from INTEGER_CSTs.
22794 2017-07-31  Segher Boessenkool  <segher@kernel.crashing.org>
22796         * config/rs6000/rs6000.c (enum rs6000_reg_type): Delete trailing comma.
22798 2017-07-31  Carl Love  <cel@us.ibm.com>
22800         * config/rs6000/rs6000-c: Add support for built-in functions
22801         vector signed char vec_xl_be (signed long long, signed char *);
22802         vector unsigned char vec_xl_be (signed long long, unsigned char *);
22803         vector signed int vec_xl_be (signed long long, signed int *);
22804         vector unsigned int vec_xl_be (signed long long, unsigned int *);
22805         vector signed long long vec_xl_be (signed long long, signed long long *);
22806         vector unsigned long long vec_xl_be (signed long long, unsigned long long *);
22807         vector signed short vec_xl_be (signed long long, signed short *);
22808         vector unsigned short vec_xl_be (signed long long, unsigned short *);
22809         vector double vec_xl_be (signed long long, double *);
22810         vector float vec_xl_be (signed long long, float *);
22811         * config/rs6000/altivec.h (vec_xl_be): Add #define.
22812         * config/rs6000/rs6000-builtin.def (XL_BE_V16QI, XL_BE_V8HI, XL_BE_V4SI)
22813         XL_BE_V2DI, XL_BE_V4SF, XL_BE_V2DF, XL_BE): Add definitions
22814         for the builtins.
22815         * config/rs6000/rs6000.c (altivec_expand_xl_be_builtin): Add function.
22816         (altivec_expand_builtin): Add switch statement to call
22817         altivec_expand_xl_be for each builtin.
22818         (altivec_init_builtins): Add def_builtin for _builtin_vsx_le_be_v8hi,
22819         __builtin_vsx_le_be_v4si, __builtin_vsx_le_be_v2di,
22820         __builtin_vsx_le_be_v4sf, __builtin_vsx_le_be_v2df,
22821         __builtin_vsx_le_be_v16qi.
22822         * doc/extend.texi: Update the built-in documentation file for the
22823         new built-in functions.
22825 2017-07-31  Uros Bizjak  <ubizjak@gmail.com>
22827         PR target/25967
22828         * config/i386/i386.c (ix86_allocate_stack_slots_for_args):
22829         New function.
22830         (TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS): Define.
22832 2017-07-31  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
22834         * config.gcc: Add z14.
22835         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
22836         CPU model numbers for z13s and z14.
22837         * config/s390/s390-c.c (s390_resolve_overloaded_builtin): Replace
22838         arch12 with z14.
22839         * config/s390/s390-opts.h (enum processor_type): Rename
22840         PROCESSOR_ARCH12 to PROCESSOR_3906_Z14.
22841         * config/s390/s390.c (processor_table): Add field for CPU name to
22842         be passed to Binutils.
22843         (s390_asm_output_machine_for_arch): Use the new field in
22844         processor_table for Binutils.
22845         (s390_expand_builtin): Replace arch12 with z14.
22846         (s390_issue_rate): Rename PROCESSOR_ARCH12 to PROCESSOR_3906_Z14.
22847         (s390_get_sched_attrmask): Likewise.
22848         (s390_get_unit_mask): Likewise.
22849         * config/s390/s390.opt: Add z14 to processor_type enum.
22851 2017-07-31  Martin Jambor  <mjambor@suse.cz>
22853         PR hsa/81477
22854         * ipa-fnsummary.c (ipa_fn_summary_generate): Set versionable
22855         regardless of optimization level.
22857 2017-07-31  Jan Hubicka <hubicka@ucw.cz>
22858             Martin Liska  <mliska@suse.cz>
22860         * predict.def: Remove old comment and adjust probability.
22861         * gimplify.c (should_warn_for_implicit_fallthrough): Ignore
22862         PREDICT statements.
22864 2017-07-31  Uros Bizjak  <ubizjak@gmail.com>
22866         PR target/25967
22867         * config/i386/i386.c (ix86_function_naked): New function.
22868         (ix86_can_use_return_insn_p): Return false for naked functions.
22869         (ix86_expand_prologue): Skip prologue for naked functions.
22870         (ix86_expand_epilogue): Skip epilogue for naked functions
22871         and emit trap instruction.
22872         (ix86_warn_func_return): New function.
22873         (ix86_attribute_table): Add "naked" attribute specification.
22874         (TARGET_WARN_FUNC_RETURN): Define.
22875         * doc/extend.texi (x86 Function Attributes) <naked>: Document it.
22877 2017-07-31  Martin Liska  <mliska@suse.cz>
22879         * gimple-pretty-print.c (dump_gimple_label): Never dump BB info.
22880         (dump_gimple_bb_header): Always dump BB info.
22881         (pp_cfg_jump): Do not append info about BB when dumping a jump.
22883 2017-07-31  Martin Liska  <mliska@suse.cz>
22885         PR sanitize/81530
22886         * convert.c (convert_to_integer_1): Guard condition with flag_sanitize_p
22887         also with current_function_decl non-null equality.
22889 2017-07-31  Jakub Jelinek  <jakub@redhat.com>
22891         PR sanitizer/81604
22892         * ubsan.c (ubsan_type_descriptor): For UBSAN_PRINT_ARRAY don't
22893         change type to the element type, instead add eltype variable and
22894         use it where we are interested in the element type.
22896         PR tree-optimization/81603
22897         * ipa-polymorphic-call.c
22898         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Perform
22899         offset arithmetic in offset_int, bail out if the resulting bit offset
22900         doesn't fit into shwi.
22902 2017-07-31  Martin Liska  <mliska@suse.cz>
22904         * gimplify.c (mostly_copy_tree_r): Remove Java specific hunk.
22905         (gimplify_save_expr): Fix comment.
22907 2017-07-30  H.J. Lu  <hongjiu.lu@intel.com>
22909         PR target/79793
22910         * config/i386/i386.c (ix86_function_arg): Update arguments for
22911         exception handler.
22912         (ix86_compute_frame_layout): Set the initial stack offset to
22913         INCOMING_FRAME_SP_OFFSET.  Update red-zone offset with
22914         INCOMING_FRAME_SP_OFFSET.
22915         (ix86_expand_epilogue): Don't pop the 'ERROR_CODE' off the
22916         stack before exception handler returns.
22917         * config/i386/i386.h (INCOMING_FRAME_SP_OFFSET): Add the
22918         the 'ERROR_CODE' for exception handler.
22920 2017-07-30  Uros Bizjak  <ubizjak@gmail.com>
22922         * config/i386/i386.h (ASM_PRINTF_EXTENSIONS): New macro.
22923         (ASM_OUTPUT_REG_PUSH): Rewrite with new operand modifiers.
22924         (ASM_OUTPUT_REG_POP): Ditto.
22925         * config/i386/i386.c (ix86_asm_output_function_label): Use fputs
22926         instead of asm_fprintf to output pure string.
22928 2017-07-29  Jakub Jelinek  <jakub@redhat.com>
22930         * debug.h (struct gcc_debug_hooks): Add IMPLICIT argument
22931         to imported_module_or_decl hook.
22932         (debug_nothing_tree_tree_tree_bool): Remove.
22933         (debug_nothing_tree_tree_tree_bool_bool): New declaration.
22934         * debug.c (do_nothing_debug_hooks): Use
22935         debug_nothing_tree_tree_tree_bool_bool instead of
22936         debug_nothing_tree_tree_tree_bool.
22937         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
22938         * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Likewise.
22939         * sdbout.c (sdb_debug_hooks): Likewise.
22940         * dwarf2out.c (dwarf2_lineno_debug_hooks): Likewise.
22941         (gen_namespace_die): Add DW_AT_export_symbols attribute if
22942         langhook wants it.
22943         (dwarf2out_imported_module_or_decl): Add IMPLICIT argument,
22944         if true, -gdwarf-5 and decl will have DW_AT_export_symbols
22945         attribute, don't add anything.
22947 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22949         * fold-const.c (fold_build1_stat_loc): Remove _stat from name.
22950         (fold_build2_stat_loc): Likewise.
22951         (fold_build3_stat_loc): Likewise.
22952         * fold-const.h (fold_build1, fold_build2, fold_build3): Adjust.
22953         (fold_build1_loc): Remove macro.
22954         (fold_build2_loc): Likewise.
22955         (fold_build3_loc): Likewise.
22957 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22959         * gimple.c (gimple_build_debug_bind_stat): Remove _stat from name.
22960         (gimple_build_debug_bind_source_stat): Likewise.
22961         * gimple.h (gimple_build_debug_bind): Remove macro.
22962         (gimple_build_debug_bind_source): Likewise.
22964 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22966         * bitmap.c (bitmap_alloc): Adjust.
22967         (bitmap_gc_alloc): Likewise.
22968         * bitmap.h (bitmap_initialize_stat): Remove _stat from name.
22970 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22972         * bitmap.c (bitmap_obstack_alloc_stat): Rename to bitmap_alloc.
22973         (bitmap_gc_alloc_stat): Rename to bitmap_gc_alloc.
22974         * bitmap.h (bitmap_obstack_alloc_stat): Adjust prototype.
22975         (bitmap_gc_alloc_stat): Likewise.
22976         (BITMAP_ALLOC, BITMAP_GGC_ALLOC): Adjust.
22978 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22980         * rtl.c (shallow_copy_rtx_stat): Remove _stat from name.
22981         * rtl.h (shallow_copy_rtx): Remove macro.
22983 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22985         * emit-rtl.c (gen_raw_REG): Adjust.
22986         * gengenrtl.c (gendef): Likewise.
22987         * rtl.c (rtx_alloc_stat): Remove _stat from name.
22988         * rtl.h (rtx_alloc): Remove macro.
22990 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22992         * tree.c (build_tree_list_vec_stat): Remove _stat from name.
22993         (build_tree_list_stat): Likewise.
22994         * tree.h (build_tree_list): Remove macro.
22995         (build_tree_list_vec): Likewise.
22997 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22999         * tree.c (make_vector_stat): Remove _stat from name.
23000         (build_vector_stat): Likewise.
23001         * tree.h (make_vector_stat): Remove macro.
23002         (build_vector_stat): Likewise.
23004 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
23006         * tree.h (build_var_debug_value): Remove prototype.
23008 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
23010         * tree.c (tree_cons_stat): Remove _stat from name.
23011         * tree.h (tree_cons): Remove macro.
23013 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
23015         * tree.c (build_vl_exp_stat): Remove _stat from name.
23016         * tree.h (build_vl_exp): Remove macro.
23018 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
23020         * tree.c (build_decl_stat): Remove _stat from name.
23021         * tree.h (build_decl): Remove macro.
23023 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
23025         * gimple.c (gimple_build_with_ops_stat): Adjust.
23026         (gimple_alloc_stat): Remove _stat from name.
23027         * gimple.h (gimple_alloc): Remove macro.
23029 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
23031         * tree.c (make_tree_vec_stat): Remove _stat from name.
23032         (grow_tree_vec_stat): Likewise.
23033         * tree.h (make_tree_vec_stat): Adjust prototype.
23034         (grow_tree_vec_stat): Likewise.
23035         (make_tree_vec): Remove macro.
23036         (grow_tree_vec): Likewise.
23038 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
23040         * fold-const.c (fold_build1_stat_loc): Adjust.
23041         (fold_build2_stat_loc): Likewise.
23042         (fold_build3_stat_loc): Likewise.
23043         * tree.c (build0_stat): Remove _stat from name.
23044         (build1_stat): Likewise.
23045         (build2_stat): Likewise.
23046         (build3_stat): Likewise.
23047         (build4_stat): Likewise.
23048         (build5_stat): Likewise.
23049         * tree.h (build1_loc): Remove macro, and rename _stat function
23050         to this.
23051         (build2_loc): Likewise.
23052         (build3_loc): Likewise.
23053         (build4_loc): Likewise.
23054         (build5_loc): Likewise.
23056 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
23058         * tree.c (make_int_cst_stat): Remove _stat from name.
23059         * tree.h (make_int_cst_stat): Adjust prototype.
23060         (make_int_cst): Remove macro.
23062 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
23064         * tree.c (make_tre_binfo_stat): Remove _stat from name.
23065         * tree.h (make_tree_binfo_stat): Adjust prototype.
23066         (make_tree_binfo): Remove.
23068 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
23070         * tree.c (copy_node_stat): Rename to copy_node.
23071         (build_distinct_type_copy): Adjust.
23072         * tree.h (copy_node_stat): Adjust prototype.
23073         (copy_node): Remove macro.
23075 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
23077         * tree.c (make_node_stat): rename to make_node.
23078         (build_tree_list_stat): Adjust.
23079         (build0_stat): Likewise.
23080         (build2_stat): Likewise.
23081         (build3_stat): Likewise.
23082         (build4_stat): Likewise.
23083         (build5_stat): Likewise.
23084         (build_decl_stat): Likewise.
23085         * tree.h (make_node_stat): Adjust prototype.
23086         (make_node): remove macro.
23088 2017-07-28  Peter Bergner  <bergner@vnet.ibm.com>
23090         * config/rs6000/ppc-auxv.h (PPC_FEATURE2_DARN): New define.
23091         (PPC_FEATURE2_SCV): Likewise.
23092         * config/rs6000/rs6000.c (cpu_supports_info): Use them.
23094 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
23096         * config/aarch64/aarch64.c
23097         (aarch64_internal_mov_immediate): Add new special pattern.
23098         * config/aarch64/aarch64.md (*movdi_aarch64):
23099         Add reg/32bit const mov case.
23101 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
23102             Richard Sandiford <richard.sandiford@linaro.org>
23104         * config/aarch64/aarch64.md (mov<mode>): Generalize.
23105         (*movhf_aarch64, *movsf_aarch64, *movdf_aarch64):
23106         Add integer and movi cases.
23107         (movi-split-hf-df-sf split, fp16): New.
23108         (enabled): Added TARGET_FP_F16INST.
23109         * config/aarch64/iterators.md (GPF_HF): New.
23110         * config/aarch64/predicates.md (aarch64_reg_or_fp_float): New.
23112 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
23114         * config/aarch64/aarch64.c
23115         (aarch64_simd_container_mode): Add prototype.
23116         (aarch64_expand_mov_immediate): Add HI support.
23117         (aarch64_reinterpret_float_as_int, aarch64_float_const_rtx_p: New.
23118         (aarch64_can_const_movi_rtx_p): New.
23119         (aarch64_preferred_reload_class):
23120         Remove restrictions of using FP registers for certain SIMD operations.
23121         (aarch64_rtx_costs): Added new cost for CONST_DOUBLE moves.
23122         (aarch64_valid_floating_const): Add integer move validation.
23123         (aarch64_simd_imm_scalar_p): Remove.
23124         (aarch64_output_scalar_simd_mov_immediate): Generalize function.
23125         (aarch64_legitimate_constant_p): Expand list of supported cases.
23126         * config/aarch64/aarch64-protos.h
23127         (aarch64_float_const_rtx_p, aarch64_can_const_movi_rtx_p): New.
23128         (aarch64_reinterpret_float_as_int): New.
23129         (aarch64_simd_imm_scalar_p): Remove.
23130         * config/aarch64/constraints.md (Uvi): New.
23131         (Dd): Split into Ds and new Dd.
23132         * config/aarch64/aarch64.md (*movsi_aarch64):
23133         Add SIMD mov case.
23134         (*movdi_aarch64): Add SIMD mov case.
23136 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23138         * tree-predcom.c: (struct chain): Handle store-store chain in which
23139         stores for elimination only store loop invariant values.
23140         (execute_pred_commoning_chain): Ditto.
23141         (prepare_initializers_chain_store_elim): Ditto.
23142         (prepare_finalizers): Ditto.
23143         (is_inv_store_elimination_chain): New function.
23144         (initialize_root_vars_store_elim_1): New function.
23146 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23148         * tree-predcom.c: Revise general description of the pass.
23149         (enum chain_type): New enum type for store elimination.
23150         (struct chain): New field supporting store elimination.
23151         (struct component): Ditto.
23152         (dump_chain): Dump store-stores chain.
23153         (release_chain): Release resources.
23154         (split_data_refs_to_components): Compute and create component
23155         contains only stores for elimination.
23156         (get_chain_last_ref_at): New function.
23157         (make_invariant_chain): Initialization.
23158         (make_rooted_chain): Specify chain type in parameter and record it.
23159         (add_looparound_copies): Skip for store-stores chain.
23160         (determine_roots_comp): Compute type of chain and pass it to
23161         make_rooted_chain.
23162         (initialize_root_vars_store_elim_2): New function.
23163         (finalize_eliminated_stores): New function.
23164         (remove_stmt): Handle store for elimination.
23165         (execute_pred_commoning_chain): Execute predictive commoning on
23166         store-store chains.
23167         (determine_unroll_factor): Skip unroll for store-stores chain.
23168         (prepare_initializers_chain_store_elim): New function.
23169         (prepare_initializers_chain): Hanlde store-store chain.
23170         (prepare_finalizers_chain, prepare_finalizers): New function.
23171         (tree_predictive_commoning_loop): Return integer value indicating
23172         if loop is unrolled or lcssa form is corrupted.
23173         (tree_predictive_commoning): Rewrite for lcssa form if necessary.
23175 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23177         * tree-predcom.c (initialize_root): Delete.
23178         (execute_pred_commoning_chain): Initialize root vars and replace
23179         reference of non-combined chain directly, rather than call above
23180         function.
23182 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23184         * tree-predcom.c (ref_at_iteration): Add parameter NITERS.  Compute
23185         memory reference to DR at (NITERS + ITERS)-th iteration of loop.
23187 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23189         * tree-predcom.c (struct chain): New field init_seq.
23190         (release_chain): Release init_seq.
23191         (prepare_initializers_chain): Record intialization stmts in above
23192         field.
23193         (insert_init_seqs): New function.
23194         (tree_predictive_commoning_loop): Call insert_init_seqs.
23196 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23198         * tree-predcom.c (determine_roots_comp): Skip trivial components.
23200 2017-07-28  Richard Biener  <rguenther@suse.de>
23202         * match.pd: Remove superfluous :c.
23203         * genmatch.c (simplify::id): Add member.
23204         (lower_commutative, lower_opt_convert, lower_cond, lower_for):
23205         Copy id.
23206         (current_id): New global.
23207         (dt_node::parent): Move from ...
23208         (dt_operand::parent): ... here.  Add for_id member.
23209         (is_a_helper <dt_operand *>::test): DT_TRUE is also a dt_operand.
23210         (decision_tree::find_node): Relax order requirement when
23211         merging DT_TRUE nodes to ones inbetween the current simplify
23212         and the one we try to merge with.  Add diagnostic whenever
23213         we need to enforce pattern order by not merging.
23214         (decision_tree::insert): Set current_id.
23215         (decision_tree::print_node): Dump parent node and for_id.
23216         (parser::last_id): Add member.
23217         (parser::push_simplify): Assign unique id.
23218         (parser::parser): Initialize last_id.
23220 2017-07-28  Martin Liska  <mliska@suse.cz>
23222         PR sanitizer/81340
23223         * sanopt.c (sanitize_rewrite_addressable_params): Set VALUE_EXPR after
23224         gimple_build_debug_bind.
23226 2017-07-28  Richard Biener  <rguenther@suse.de>
23228         PR tree-optimization/81502
23229         * match.pd: Add pattern combining BIT_INSERT_EXPR with
23230         BIT_FIELD_REF.
23231         * tree-cfg.c (verify_expr): Verify types of BIT_FIELD_REF
23232         size/pos operands.
23233         (verify_gimple_assign_ternary): Likewise for BIT_INSERT_EXPR pos.
23234         * gimple-fold.c (maybe_canonicalize_mem_ref_addr): Use bitsizetype
23235         for BIT_FIELD_REF args.
23236         * fold-const.c (make_bit_field_ref): Likewise.
23237         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
23239 2017-07-28  Jakub Jelinek  <jakub@redhat.com>
23241         PR sanitizer/80998
23242         * sanopt.c (pass_sanopt::execute): Handle IFN_UBSAN_PTR.
23243         * tree-ssa-alias.c (call_may_clobber_ref_p_1): Likewise.
23244         * flag-types.h (enum sanitize_code): Add SANITIZER_POINTER_OVERFLOW.
23245         Or it into SANITIZER_UNDEFINED.
23246         * ubsan.c: Include gimple-fold.h and varasm.h.
23247         (ubsan_expand_ptr_ifn): New function.
23248         (instrument_pointer_overflow): New function.
23249         (maybe_instrument_pointer_overflow): New function.
23250         (instrument_object_size): Formatting fix.
23251         (pass_ubsan::execute): Call instrument_pointer_overflow
23252         and maybe_instrument_pointer_overflow.
23253         * internal-fn.c (expand_UBSAN_PTR): New function.
23254         * ubsan.h (ubsan_expand_ptr_ifn): Declare.
23255         * sanitizer.def (__ubsan_handle_pointer_overflow,
23256         __ubsan_handle_pointer_overflow_abort): New builtins.
23257         * tree-ssa-tail-merge.c (merge_stmts_p): Handle IFN_UBSAN_PTR.
23258         * internal-fn.def (UBSAN_PTR): New internal function.
23259         * opts.c (sanitizer_opts): Add pointer-overflow.
23260         * lto-streamer-in.c (input_function): Handle IFN_UBSAN_PTR.
23261         * fold-const.c (build_range_check): Compute pointer range check in
23262         integral type if pointer arithmetics would be needed.  Formatting
23263         fixes.
23265 2017-07-28  Martin Liska  <mliska@suse.cz>
23267         PR sanitizer/81460
23268         * sanopt.c (sanitize_rewrite_addressable_params): Do not rewrite
23269         parameters that are of a variable-length.
23271 2017-07-28  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23273         * config.gcc (powerpc-*-rtems*): Remove rs6000/eabi.h.  Add
23274         rs6000/biarch64.h.
23275         * config/rs6000/rtems.h (ASM_DECLARE_FUNCTION_SIZE): New macro.
23276         (ASM_OUTPUT_SPECIAL_POOL_ENTRY_P): Likewise.
23277         (CRT_CALL_STATIC_FUNCTION): Likewise.
23278         (ASM_DEFAULT_SPEC): New define.
23279         (ASM_SPEC32): Likewise.
23280         (ASM_SPEC64): Likewise.
23281         (ASM_SPEC_COMMON): Likewise.
23282         (ASM_SPEC): Likewise.
23283         (INVALID_64BIT): Likewise.
23284         (LINK_OS_DEFAULT_SPEC): Likewise.
23285         (LINK_OS_SPEC32): Likewise.
23286         (LINK_OS_SPEC64): Likewise.
23287         (POWERPC_LINUX): Likewise.
23288         (PTRDIFF_TYPE): Likewise.
23289         (RESTORE_FP_PREFIX): Likewise.
23290         (RESTORE_FP_SUFFIX): Likewise.
23291         (SAVE_FP_PREFIX): Likewise.
23292         (SAVE_FP_SUFFIX): Likewise.
23293         (SIZE_TYPE): Likewise.
23294         (SUBSUBTARGET_OVERRIDE_OPTIONS): Likewise.
23295         (TARGET_64BIT): Likewise.
23296         (TARGET_64BIT): Likewise.
23297         (TARGET_AIX): Likewise.
23298         (WCHAR_TYPE_SIZE): Likewise.
23299         (WCHAR_TYPE): Undefine.
23300         (TARGET_OS_CPP_BUILTINS): Add 64-bit PowerPC defines.
23301         (CPP_OS_DEFAULT_SPEC): Use previous CPP_OS_RTEMS_SPEC.
23302         (CPP_OS_RTEMS_SPEC): Delete.
23303         (SUBSUBTARGET_EXTRA_SPECS): Remove cpp_os_rtems.  Add
23304         asm_spec_common, asm_spec32, asm_spec64, link_os_spec32, and
23305         link_os_spec64.
23306         * config/rs6000/t-rtems: Add mcpu=e6500/m64 multilibs.
23308 2017-07-28  Jakub Jelinek  <jakub@redhat.com>
23310         PR tree-optimization/81578
23311         * tree-parloops.c (build_new_reduction): Bail out if
23312         reduction_code isn't one of the standard OpenMP reductions.
23313         Move the details printing after that decision.
23315 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23317         * config/rs6000/predicates.md (volatile_mem_operand): Remove code
23318         related to reload_in_progress.
23319         (splat_input_operand): Likewise.
23320         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_rtx):
23321         Delete prototype.
23322         * config/rs6000/rs6000.c (machine_function): Remove sdmode_stack_slot
23323         field.
23324         (TARGET_EXPAND_TO_RTL_HOOK): Delete.
23325         (TARGET_INSTANTIATE_DECLS): Likewise.
23326         (legitimate_indexed_address_p): Delete reload_in_progress code.
23327         (rs6000_debug_legitimate_address_p): Likewise.
23328         (rs6000_eliminate_indexed_memrefs): Likewise.
23329         (rs6000_emit_le_vsx_store): Likewise.
23330         (rs6000_emit_move_si_sf_subreg): Likewise.
23331         (rs6000_emit_move): Likewise.
23332         (register_to_reg_type): Likewise.
23333         (rs6000_pre_atomic_barrier): Likewise.
23334         (rs6000_machopic_legitimize_pic_address): Likewise.
23335         (rs6000_allocate_stack_temp): Likewise.
23336         (rs6000_address_for_fpconvert): Likewise.
23337         (rs6000_address_for_altivec): Likewise.
23338         (rs6000_secondary_memory_needed_rtx): Delete function.
23339         (rs6000_check_sdmode): Likewise.
23340         (rs6000_alloc_sdmode_stack_slot): Likewise.
23341         (rs6000_instantiate_decls): Likewise.
23342         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED_RTX): Delete.
23343         * config/rs6000/rs6000.md (splitter for *movsi_got_internal):
23344         Delete reload_in_progress.
23345         (*vec_reload_and_plus_<mptrsize>): Likewise.
23346         * config/rs6000/vsx.md (vsx_mul_v2di): Likewise.
23347         (vsx_div_v2di): Likewise.
23348         (vsx_udiv_v2di): Likewise.
23350 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23352         * config/rs6000/rs6000.opt (mlra): Replace with stub.
23353         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete OPTION_MASK_LRA.
23354         * config/rs6000/rs6000.c (TARGET_LRA_P): Delete.
23355         (rs6000_debug_reg_global): Delete print of LRA status.
23356         (rs6000_option_override_internal): Delete dead LRA related code.
23357         (rs6000_lra_p): Delete function.
23358         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mlra.
23360 2017-07-27  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23362         * config.gcc (riscv*-*-elf*): Add (riscv*-*-rtems*).
23363         * config/riscv/rtems.h: New file.
23365 2017-07-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23366             Sudakshina Das  <sudi.das@arm.com>
23368         * config/aarch64/aarch64.md
23369         (define_split for and<mode>3nr_compare): Move
23370         non aarch64_logical_operand to a register.
23371         (define_split for and_<SHIFT:optab><mode>3nr_compare0): Move non
23372         register immediate operand to a register.
23373         * config/aarch64/predicates.md (aarch64_mov_imm_operand): New.
23375 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23377         PR middle-end/81564
23378         * tree-cfg.c (group_case_labels_stmt): Handle already deleted blocks.
23380 2017-07-27  Richard Biener  <rguenther@suse.de>
23382         PR tree-optimization/81573
23383         PR tree-optimization/81494
23384         * tree-vect-loop.c (vect_create_epilog_for_reduction): Handle
23385         multi defuse cycle case.
23387 2017-07-27  Richard Biener  <rguenther@suse.de>
23389         PR tree-optimization/81571
23390         * tree-vect-slp.c (vect_build_slp_tree): Properly verify reduction
23391         PHIs.
23393 2017-07-27  Eric Botcazou  <ebotcazou@adacore.com>
23395         * config/sparc/sparc.c (sparc_option_override): Set MASK_FSMULD flag
23396         earlier and only if MASK_FPU is set.  Adjust formatting.
23398 2017-07-27  Martin Liska  <mliska@suse.cz>
23400         * opt-functions.awk: Add validation of value of Init.
23401         * optc-gen.awk: Pass new argument.
23403 2017-07-27  Martin Liska  <mliska@suse.cz>
23405         * auto-profile.c (autofdo_source_profile::update_inlined_ind_target):
23406         Fix wrong condition.
23408 2017-07-27  Martin Liska  <mliska@suse.cz>
23410         * auto-profile.c (afdo_annotate_cfg): Assign zero counts to
23411         BBs and edges seen by autoFDO.
23413 2017-07-27  Richard Biener  <rguenther@suse.de>
23415         PR tree-optimization/81502
23416         * tree-ssa.c (non_rewritable_lvalue_p): Handle BIT_INSERT_EXPR
23417         with incompatible but same sized type.
23418         (execute_update_addresses_taken): Likewise.
23420 2017-07-27  James Greenhalgh  <james.greenhalgh@arm.com>
23422         * tree-ssa-loop-ch.c (pass_ch::process_loop_p): Guard on
23423         flag_tree_loop_vectorize rather than flag_tree_vectorize.
23425 2017-07-27  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
23427         PR target/81534
23428         * config/s390/s390.md ("*atomic_compare_and_swap<mode>_1")
23429         ("*atomic_compare_and_swapdi_2", "*atomic_compare_and_swapsi_3"):
23430         Change s_operand to memory_operand.
23432 2017-07-27  Richard Sandiford  <richard.sandiford@linaro.org>
23434         * config/rs6000/rs6000-protos.h (rs6000_emit_le_vsx_permute): Declare.
23435         * config/rs6000/rs6000.c (rs6000_gen_le_vsx_permute): Replace with...
23436         (rs6000_emit_le_vsx_permute): ...this.  Take the destination as input.
23437         Emit instructions rather than returning an expression.  Handle TFmode
23438         and KFmode by casting to TImode.
23439         (rs6000_emit_le_vsx_load): Update to use rs6000_emit_le_vsx_permute.
23440         (rs6000_emit_le_vsx_store): Likewise.
23441         * config/rs6000/vsx.md (VSX_TI): New iterator.
23442         (*vsx_le_permute_<mode>): Use it instead of VSX_LE_128.
23443         (*vsx_le_undo_permute_<mode>): Likewise.
23444         (*vsx_le_perm_load_<mode>): Use rs6000_emit_le_vsx_permute to
23445         emit the split sequence.
23446         (*vsx_le_perm_store_<mode>): Likewise.
23448 2017-07-27  Jakub Jelinek  <jakub@redhat.com>
23450         PR tree-optimization/81555
23451         PR tree-optimization/81556
23452         * tree-ssa-reassoc.c (rewrite_expr_tree): Add NEXT_CHANGED argument,
23453         if true, force CHANGED for the recursive invocation.
23454         (reassociate_bb): Remember original length of ops array, pass
23455         len != orig_len as NEXT_CHANGED in rewrite_expr_tree call.
23457         * attribs.c (decl_attributes): Imply noinline, noclone and no_icf
23458         attributes for noipa attribute.  For naked attribute use
23459         lookup_attribute first before lookup_attribute_spec.
23460         * final.c (rest_of_handle_final): Disable IPA RA for functions with
23461         noipa attribute.
23462         * ipa-visibility.c (non_local_p): Fix comment typos.  Return true
23463         for functions with noipa attribute.
23464         (cgraph_externally_visible_p): Return true for functions with noipa
23465         attribute.
23466         * cgraph.c (cgraph_node::get_availability): Return AVAIL_INTERPOSABLE
23467         for functions with noipa attribute.
23468         * doc/extend.texi: Document noipa function attribute.
23469         * tree-ssa-structalias.c (refered_from_nonlocal_fn): Set *nonlocal_p
23470         also for functions with noipa attribute.
23471         (ipa_pta_execute): Set nonlocal_p also for nodes with noipa attribute.
23473 2017-07-26  Andrew Pinski  <apinski@cavium.com>
23475         * config/aarch64/aarch64.c (thunderx_vector_cost): Decrease cost of
23476         vec_unalign_load_cost and vec_unalign_store_cost.
23478 2017-07-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
23480         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Delete
23481         -mvsx-small-integer option.
23482         (ISA_3_0_MASKS_IEEE): Likewise.
23483         (OTHER_VSX_VECTOR_MASKS): Likewise.
23484         (POWERPC_MASKS): Likewise.
23485         * config/rs6000/rs6000.opt (-mvsx-small-integer): Likewise.
23486         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Simplify
23487         code, only testing for DImode being allowed in non-VSX floating
23488         point registers.
23489         (rs6000_init_hard_regno_mode_ok): Change TARGET_VSX_SMALL_INTEGER
23490         to TARGET_P8_VECTOR test.  Remove redundant VSX test inside of
23491         another VSX test.
23492         (rs6000_option_override_internal): Delete -mvsx-small-integer.
23493         (rs6000_expand_vector_set): Change TARGET_VSX_SMALL_INTEGER to
23494         TARGET_P8_VECTOR test.
23495         (rs6000_secondary_reload_simple_move): Likewise.
23496         (rs6000_preferred_reload_class): Delete TARGET_VSX_SMALL_INTEGER,
23497         since TARGET_P9_VECTOR was already tested.
23498         (rs6000_opt_masks): Remove -mvsx-small-integer.
23499         * config/rs6000/vsx.md (vsx_extract_<mode>): Delete
23500         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23501         used.
23502         (vsx_extract_<mode>_p9): Delete TARGET_VSX_SMALL_INTEGER, since a
23503         test for TARGET_VEXTRACTUB was used, and that uses
23504         TARGET_P9_VECTOR.
23505         (p9 extract splitter): Likewise.
23506         (vsx_extract_<mode>_di_p9): Likewise.
23507         (vsx_extract_<mode>_store_p9): Likewise.
23508         (vsx_extract_si): Delete TARGET_VSX_SMALL_INTEGER, since a test
23509         for TARGET_P9_VECTOR was used.  Delete code that is now dead with
23510         the elimination of TARGET_VSX_SMALL_INTEGER.
23511         (vsx_extract_<mode>_p8): Likewise.
23512         (vsx_ext_<VSX_EXTRACT_I:VS_scalar>_fl_<FL_CONV:mode>): Likewise.
23513         (vsx_ext_<VSX_EXTRACT_I:VS_scalar>_ufl_<FL_CONV:mode>): Likewise.
23514         (vsx_set_<mode>_p9): Likewise.
23515         (vsx_set_v4sf_p9): Likewise.
23516         (vsx_set_v4sf_p9_zero): Likewise.
23517         (vsx_insert_extract_v4sf_p9): Likewise.
23518         (vsx_insert_extract_v4sf_p9_2): Likewise.
23519         * config/rs6000/rs6000.md (sign extend splitter): Change
23520         TARGET_VSX_SMALL_INTEGER to TARGET_P8_VECTOR test.
23521         (floatsi<mode>2_lfiwax_mem): Likewise.
23522         (floatunssi<mode>2_lfiwzx_mem): Likewise.
23523         (float<QHI:mode><FP_ISA3:mode>2): Delete TARGET_VSX_SMALL_INTEGER,
23524         since a test for TARGET_P9_VECTOR was used.
23525         (float<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23526         (floatuns<QHI:mode><FP_ISA3:mode>2): Likewise.
23527         (floatuns<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23528         (fix_trunc<mode>si2): Change TARGET_VSX_SMALL_INTEGER to
23529         TARGET_P8_VECTOR test.
23530         (fix_trunc<mode>si2_stfiwx): Likewise.
23531         (fix_trunc<mode>si2_internal): Likewise.
23532         (fix_trunc<SFDF:mode><QHI:mode>2): Delete
23533         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23534         used.
23535         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
23536         (fixuns_trunc<mode>si2): Change TARGET_VSX_SMALL_INTEGER to
23537         TARGET_P8_VECTOR test.
23538         (fixuns_trunc<mode>si2_stfiwx): Likewise.
23539         (fixuns_trunc<SFDF:mode><QHI:mode>2): Delete
23540         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23541         used.
23542         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
23543         (fctiw<u>z_<mode>_smallint): Delete TARGET_VSX_SMALL_INTEGER,
23544         since a test for TARGET_P9_VECTOR was used.
23545         (splitter for loading small constants): Likewise.
23547 2017-07-26  Andrew Pinski  <apinski@cavium.com>
23549         * config/aarch64/aarch64.c (thunderx_vector_cost): Fix
23550         vec_fp_stmt_cost.
23552 2017-07-26  H.J. Lu  <hongjiu.lu@intel.com>
23554         PR target/81563
23555         * config/i386/i386.c (sp_valid_at): Properly check CFA offset.
23556         (fp_valid_at): Likewise.
23558 2017-07-26  James Greenhalgh  <james.greenhalgh@arm.com>
23560         * config/aarch64/aarch64.c (cortexa57_addrcost_table): Remove.
23561         (qdf24xx_addrcost_table): Likewise.
23562         (cortexa57_tunings): Update to use generic_branch_cost.
23563         (cortexa72_tunings): Likewise.
23564         (cortexa73_tunings): Likewise.
23565         (qdf24xx_tunings): Likewise.
23567 2017-07-26  James Greenhalgh  <james.greenhalgh@arm.com>
23569         * config/aarch64/aarch64.c (cortexa57_branch_cost): Remove.
23570         (thunderx2t99_branch_cost): Likewise.
23571         (cortexa35_tunings): Update to use generic_branch_cost.
23572         (cortexa53_tunings): Likewise.
23573         (cortexa57_tunings): Likewise.
23574         (cortexa72_tunings): Likewise.
23575         (cortexa73_tunings): Likewise.
23576         (thunderx2t99_tunings): Likewise.
23578 2017-07-26  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23580         * config/sparc/sparc.c (dump_target_flag_bits): Dump MASK_FSMULD.
23581         (sparc_option_override): Honour MASK_FSMULD.
23582         * config/sparc/sparc.h (MASK_FEATURES): Add MASK_FSMULD.
23583         * config/sparc/sparc.md (muldf3_extend): Use TARGET_FSMULD.
23584         * config/sparc/sparc.opt (mfsmuld): New option.
23585         * doc/invoke.texi (mfsmuld): Document option.
23587 2017-07-26  Marek Polacek  <polacek@redhat.com>
23589         PR middle-end/70992
23590         * tree.c (build2_stat): Don't set TREE_CONSTANT on divisions by zero.
23592 2017-07-26  Richard Biener  <rguenther@suse.de>
23594         * gimple-match-head.c (do_valueize): Return OP if valueize
23595         returns NULL_TREE.
23596         (get_def): New helper to get at the def stmt of a SSA name
23597         if valueize allows.
23598         * genmatch.c (dt_node::gen_kids_1): Use get_def instead of
23599         do_valueize to get at the def stmt.
23600         (dt_operand::gen_gimple_expr): Simplify do_valueize calls.
23602 2017-07-26  Wilco Dijkstra  <wdijkstr@arm.com>
23604         PR middle-end/46932
23605         * auto-inc-dec.c (parse_add_or_inc): Block autoinc on sfp.
23607 2017-07-26  Martin Liska  <mliska@suse.cz>
23609         PR sanitize/81186
23610         * function.c (expand_function_start): Make expansion of
23611         nonlocal_goto_save_area after parm_birth_insn.
23613 2017-07-26  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23615         * config/sparc/sparc.c (sparc_option_override): Remove MASK_FPU
23616         from all CPU target flags enable members.
23618 2017-07-26  Richard Biener  <rguenther@suse.de>
23620         * genmatch.c (dt_simplify::gen): Make iterator vars const.
23621         (decision_tree::gen): Make 'type' const.
23622         (write_predicate): Likewise.
23624 2017-07-24  Michael Meissner  <meissner@linux.vnet.ibm.com>
23626         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok):
23627         Eliminate TARGET_UPPER_REGS_{DF,DI,SF} usage.
23628         (rs6000_option_override_internal): Likewise.
23629         (rs6000_expand_vector_set): Likewise.
23630         * config/rs6000/rs6000.h (TARGET_UPPER_REGS_DF): Delete.
23631         (TARGET_UPPER_REGS_SF): Likewise.
23632         (TARGET_UPPER_REGS_DI): Likewise.
23633         (TARGET_VEXTRACTUB): Eliminate TARGET_UPPER_REGS_{DF,DI,SF}.
23634         (TARGET_DIRECT_MOVE_64BIT): Likewise.
23635         * config/rs6000/rs6000.md (ALTIVEC_DFORM): Likewise.
23636         (float<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23637         (Splitters for DI constants in Altivec registers): Likewise.
23638         * config/rs6000/vsx.md (vsx_set_<mode>_p9): Likewise.
23639         (vsx_set_v4sf_p9): Likewise.
23640         (vsx_set_v4sf_p9_zero): Likewise.
23641         (vsx_insert_extract_v4sf_p9): Likewise.
23642         (vsx_insert_extract_v4sf_p9_2): Likewise.
23644 2017-07-25  Carl Love  <cel@us.ibm.com>
23646         * doc/extend.texi: Update the built-in documentation file for the
23647         existing built-in functions
23648         vector signed char vec_cnttz (vector signed char);
23649         vector unsigned char vec_cnttz (vector unsigned char);
23650         vector signed short vec_cnttz (vector signed short);
23651         vector unsigned short vec_cnttz (vector unsigned short);
23652         vector signed int vec_cnttz (vector signed int);
23653         vector unsigned int vec_cnttz (vector unsigned int);
23654         vector signed long long vec_cnttz (vector signed long long);
23655         vector unsigned long long vec_cnttz (vector unsigned long long);
23657 2017-07-25  Andrew Pinski  <apinski@cavium.com>
23659         * tree-ssa-uninit.c (warn_uninitialized_vars): Don't warn about memory
23660         accesses where the use is for the first operand of a BIT_INSERT.
23662 2017-07-25  Jim Wilson  <jim.wilson@linaro.org>
23664         PR bootstrap/81521
23665         * config/i386/winnt-cxx.c (i386_pe_adjust_class_at_definition): Look
23666         for FUNCTION_DECLs in TYPE_FIELDS rather than TYPE_METHODS.
23668 2017-07-25  Jim Wilson  <jim.wilson@linaro.org>
23670         * config/i386/gstabs.h: Delete.
23671         * config/i386/openbsd.h, config/i386/t-openbsd: Likewise.
23673 2017-07-25  Uros Bizjak  <ubizjak@gmail.com>
23675         * config/i386/i386.c (ix86_decompose_address): Do not check for
23676         register RTX when looking at index_reg or base_reg.
23677         * config/i386/i386.h (INCOMING_RETURN_ADDR_RTX): Use stack_pointer_rtx.
23679 2017-07-25  Eric Botcazou  <ebotcazou@adacore.com>
23681         * gimple.c (gimple_assign_set_rhs_with_ops): Do not ask gsi_replace
23682         to update EH info here.
23684 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23686         * match.pd ((X * CST1) * CST2): Simplify to X * (CST1 * CST2).
23688 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23690         * match.pd ((X * CST) * Y): Reassociate to (X * Y) * CST.
23692 2017-07-25  Torsten Duwe  <duwe@suse.de>
23694         * common.opt: Introduce -fpatchable-function-entry
23695         command line option, and its variables function_entry_patch_area_size
23696         and function_entry_patch_area_start.
23697         * opts.c (common_handle_option): Add -fpatchable_function_entry_ case,
23698         including a two-value parser.
23699         * target.def (print_patchable_function_entry): New target hook.
23700         * targhooks.h (default_print_patchable_function_entry): New function.
23701         * targhooks.c (default_print_patchable_function_entry): Likewise.
23702         * toplev.c (process_options): Switch off IPA-RA if
23703         patchable function entries are being generated.
23704         * varasm.c (assemble_start_function): Look at the
23705         patchable-function-entry command line switch and current
23706         function attributes and maybe generate NOP instructions by
23707         calling the print_patchable_function_entry hook.
23708         * doc/extend.texi: Document patchable_function_entry attribute.
23709         * doc/invoke.texi: Document -fpatchable_function_entry
23710         command line option.
23711         * doc/tm.texi.in (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY):
23712         New target hook.
23713         * doc/tm.texi: Re-generate.
23715 2017-07-25  Jakub Jelinek  <jakub@redhat.com>
23717         PR target/81532
23718         * config/i386/constraints.md (Yd, Ye): Use ALL_SSE_REGS for
23719         TARGET_AVX512DQ rather than TARGET_AVX512BW.
23721 2017-07-25  Tamar Christina  <tamar.christina@arm.com>
23723         * config/arm/parsecpu.awk (all_cores): Remove duplicates.
23725 2017-07-25  Richard Biener  <rguenther@suse.de>
23727         PR tree-optimization/81455
23728         * tree-ssa-loop-unswitch.c (find_loop_guard): Make sure to
23729         not walk in cycles when looking for guards.
23731 2017-07-25  Richard Biener  <rguenther@suse.de>
23733         PR tree-optimization/81529
23734         * tree-vect-stmts.c (process_use): Disregard live induction PHIs
23735         when optimizing backedge uses.
23737 2017-07-25  David Edelsohn  <dje.gcc@gmail.com>
23739         * dwarf2asm.c (dw2_asm_output_nstring): Encode double quote
23740         character for AIX.
23741         * dwarf2out.c (output_macinfo): Copy debug_line_section_label
23742         to dl_section_ref.  On AIX, append an expression to subtract
23743         the size of the section length to dl_section_ref.
23745 2017-07-25  Segher Boessenkool  <segher@kernel.crashing.org>
23747         * configure.ac: If any of the config.* scripts fail, exit 1.
23748         * configure: Regenerate.
23750 2017-07-25  Richard Biener  <rguenther@suse.de>
23752         PR middle-end/81546
23753         * tree-ssa-operands.c (verify_imm_links): Remove cap on number
23754         of immediate uses, be more verbose on errors.
23756 2017-07-25  Richard Biener  <rguenther@suse.de>
23758         PR tree-optimization/81510
23759         * tree-vect-loop.c (vect_is_simple_reduction): When the
23760         reduction stmt is not inside the loop bail out.
23762 2017-07-25  Richard Biener  <rguenther@suse.de>
23764         PR tree-optimization/81303
23765         * tree-vect-loop-manip.c (vect_loop_versioning): Build
23766         profitability check against LOOP_VINFO_NITERSM1.
23768 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23770         * domwalk.c (cmp_bb_postorder): Simplify.
23771         (sort_bbs_postorder): New function.  Use it...
23772         (dom_walker::walk): ...here to optimize common cases.
23774 2017-07-25  Martin Liska  <mliska@suse.cz>
23776         PR ipa/81520
23777         * ipa-visibility.c (function_and_variable_visibility): Make the
23778         redirection just on target that supports aliasing.
23779         Fix GNU coding style.
23781 2017-07-25  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23783         PR libgcc/61152
23784         * config/aarch64/rtems.h: Add GCC Runtime Library Exception.
23785         Format changes.
23786         * config/arm/rtems.h: Likewise.
23787         * config/bfin/rtems.h: Likewise.
23788         * config/i386/rtemself.h: Likewise.
23789         * config/lm32/rtems.h: Likewise.
23790         * config/m32c/rtems.h: Likewise.
23791         * config/m68k/rtemself.h: Likewise.
23792         * config/microblaze/rtems.h: Likewise.
23793         * config/mips/rtems.h: Likewise.
23794         * config/moxie/rtems.h: Likewise.
23795         * config/nios2/rtems.h: Likewise.
23796         * config/powerpcspe/rtems.h: Likewise.
23797         * config/rs6000/rtems.h: Likewise.
23798         * config/rtems.h: Likewise.
23799         * config/sh/rtems.h: Likewise.
23800         * config/sh/rtemself.h: Likewise.
23801         * config/sparc/rtemself.h: Likewise.
23803 2017-07-25  Georg-Johann Lay  <avr@gjlay.de>
23805         PR 81487
23806         * hsa-brig.c (brig_init): Use xasprintf instead of asprintf.
23807         * gimple-pretty-print.c (dump_profile, dump_probability): Same.
23808         * tree-ssa-structalias.c (alias_get_name): Same.
23810 2017-07-25  Bin Cheng  <bin.cheng@arm.com>
23812         PR target/81414
23813         * config/aarch64/cortex-a57-fma-steering.c (analyze): Skip fmul/fmac
23814         instructions if no du chain is found.
23816 2017-07-25  Georg-Johann Lay  <avr@gjlay.de>
23818         * config/avr/avr-log.c (avr_log_vadump) ['T']: Print NULL-TREE.
23820 2017-07-25  Richard Biener  <rguenther@suse.de>
23822         PR middle-end/81505
23823         * fold-const.c (fold_negate_const): TREE_OVERFLOW should be
23824         sticky.
23826 2017-07-24  Michael Meissner  <meissner@linux.vnet.ibm.com>
23828         * config/rs6000/rs6000-cpus.def (ISA_2_6_MASKS_SERVER): Delete
23829         upper-regs options.
23830         (ISA_2_7_MASKS_SERVER): Likewise.
23831         (ISA_3_0_MASKS_IEEE): Likewise.
23832         (OTHER_P8_VECTOR_MASKS): Likewise.
23833         (OTHER_VSX_VECTOR_MASKS): Likewise.
23834         (POWERPC_MASKS): Likewise.
23835         (power7 cpu): Use ISA_2_6_MASKS_SERVER instead of using a
23836         duplicate list of options.
23837         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove
23838         explicit -mupper-regs options.
23839         * config/rs6000/rs6000.opt (-mvsx-scalar-memory): Delete
23840         -mupper-regs* options.  Delete -mvsx-scalar-memory, which was an
23841         alias for -mupper-regs-df.
23842         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Likewise.
23843         (rs6000_init_hard_regno_mode_ok): Likewise.
23844         (rs6000_option_override_internal): Likewise.
23845         (rs6000_opt_masks): Likewise.
23846         * config/rs6000/rs6000.h (TARGET_UPPER_REGS_DF): Define upper regs
23847         options in terms of whether -mvsx or -mpower8-vector was used.
23848         (TARGET_UPPER_REGS_DI): Likewise.
23849         (TARGET_UPPER_REGS_SF): Likewise.
23850         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete the
23851         -mupper-regs-* options.
23853 2017-07-24  Segher Boessenkool  <segher@kernel.crashing.org>
23855         * passes.c (emergency_dump_function): Print some empty lines and a
23856         header before the RTL dump.
23858 2017-07-24  Segher Boessenkool  <segher@kernel.crashing.org>
23860         * cfgrtl.c (rtl_dump_bb): Don't call NEXT_INSN on NULL.
23862 2017-07-24  Wilco Dijkstra  <wdijkstr@arm.com>
23864         PR target/79041
23865         * config/aarch64/aarch64.c (aarch64_classify_symbol):
23866         Avoid SYMBOL_SMALL_ABSOLUTE for literals with pc-relative literals.
23868 2017-07-24  Carl Love  <cel@us.ibm.com>
23870         * config/rs6000/rs6000-c.c: Add support for built-in functions
23871         vector float vec_extract_fp32_from_shorth (vector unsigned short);
23872         vector float vec_extract_fp32_from_shortl (vector unsigned short);
23873         * config/rs6000/altivec.h (vec_extract_fp_from_shorth,
23874         vec_extract_fp_from_shortl): Add defines for the two builtins.
23875         * config/rs6000/rs6000-builtin.def (VEXTRACT_FP_FROM_SHORTH,
23876         VEXTRACT_FP_FROM_SHORTL): Add BU_P9V_OVERLOAD_1 and BU_P9V_VSX_1
23877         new builtins.
23878         * config/rs6000/vsx.md vsx_xvcvhpsp): Add define_insn.
23879         (vextract_fp_from_shorth, vextract_fp_from_shortl): Add define_expands.
23880         * doc/extend.texi: Update the built-in documentation file for the
23881         new built-in function.
23883 2017-07-24  Jakub Jelinek  <jakub@redhat.com>
23885         PR bootstrap/81521
23886         * tree.def: Remove TYPE_METHODS documentation, adjust TYPE_FIELDS
23887         documentation.
23888         * doc/generic.texi: Likewise.
23889         * config/i386/winnt-cxx.c (i386_pe_adjust_class_at_definition): Look
23890         for FUNCTION_DECLs in TYPE_FIELDS rather than TYPE_METHODS.
23892 2017-07-24  Jackson Woodruff  <jackson.woodruff@arm.com>
23894         * config/aarch64/aarch64-simd.md (aarch64_mla_elt_merge<mode>): New.
23895         (aarch64_mls_elt_merge<mode>): Likewise.
23897 2017-07-23  Krister Walfridsson  <krister.walfridsson@gmail.com>
23899         * config.gcc (*-*-netbsd*): Remove check for NetBSD versions not
23900         having __cxa_atexit.
23902 2017-07-23  Michael Collison  <michael.collison@arm.com>
23904         * config/arm/arm.c (arm_option_override): Deprecate
23905         use of -mstructure-size-boundary.
23906         * config/arm/arm.opt: Deprecate -mstructure-size-boundary.
23907         * doc/invoke.texi: Deprecate -mstructure-size-boundary.
23909 2017-07-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23911         PR target/80695
23912         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost):
23913         Reduce cost estimate for direct moves.
23915 2017-07-23  Uros Bizjak  <ubizjak@gmail.com>
23917         PR target/80569
23918         * config/i386/i386.c (ix86_option_override_internal): Disable
23919         BMI, BMI2 and TBM instructions for -m16.
23921 2017-07-21  Carl Love  <cel@us.ibm.com>
23923         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
23924         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
23925         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
23926         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
23927         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
23928         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
23929         VMULOSW): New enum "unspec" values.
23930         (altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
23931         altivec_vmulosw): New patterns.
23932         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
23933         VMULOSW): Add definitions.
23935 2017-07-21  Jim Wilson  <jim.wilson@linaro.org>
23937         * config/aarch64/aarch64-cores.def (falkor): Add AARCH64_FL_RDMA.
23938         (qdf24xx): Likewise.
23939         * config/aarch64/aarch64-options-extensions.def (rdma); New.
23940         * config/aarch64/aarch64.h (AARCH64_FL_RDMA): New.
23941         (AARCH64_FL_V8_1): Renumber.
23942         (AARCH64_FL_FOR_ARCH8_1): Add AARCH64_FL_RDMA.
23943         (AARCH64_ISA_RDMA): Use AARCH64_FL_RDMA.
23944         * config/aarch64/arm_neon.h: Use +rdma instead of arch=armv8.1-a.
23945         * doc/invoke.texi (AArch64 Options): Mention +rmda in -march docs.  Add
23946         rdma to feature modifiers list.
23948 2017-07-21  Yury Gribov  <tetra2005@gmail.com>
23950         PR middle-end/56727
23951         * ipa-visibility (function_and_variable_visibility): Convert
23952         recursive PLT call to direct call if appropriate.
23954 2017-07-21  Andrew Pinski  <apinski@cavium.com>
23956         * tree-ssa-sccvn.c (vn_nary_op_eq): Check BIT_INSERT_EXPR's
23957         operand 1 to see if the types precision matches.
23958         * fold-const.c (operand_equal_p): Likewise.
23960 2017-07-21  Richard Biener  <rguenther@suse.de>
23962         PR tree-optimization/81303
23963         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs): Pass
23964         in datarefs vector.  Allow NULL dr0 for no peeling cost estimate.
23965         (vect_peeling_hash_get_lowest_cost): Adjust.
23966         (vect_enhance_data_refs_alignment): Likewise.  Use
23967         vect_get_peeling_costs_all_drs to compute the penalty for no
23968         peeling to match up costs.
23970 2017-07-21  Richard Biener  <rguenther@suse.de>
23972         PR tree-optimization/81500
23973         * tree-vect-loop.c (vect_is_simple_reduction): Properly fail if
23974         we didn't identify a reduction path.
23976 2017-07-21  Tom de Vries  <tom@codesourcery.com>
23977             Cesar Philippidis  <cesar@codesourcery.com>
23979         PR gcov-profile/81442
23980         * config/nvptx/nvptx.c (nvptx_goacc_reduction_init): Add missing edge
23981         probabilities.
23983 2017-07-21  Tom de Vries  <tom@codesourcery.com>
23985         PR lto/81430
23986         * config/nvptx/nvptx.c (nvptx_override_options_after_change): New
23987         function.
23988         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define to
23989         nvptx_override_options_after_change.
23991 2017-07-21  Ulrich Drepper  <drepper@redhat.com>
23993         * dwarf2out.c (output_file_names): Avoid double testing for
23994         dwarf_version >= 5.
23996 2017-07-21  Georg-Johann Lay  <avr@gjlay.de>
23998         * doc/invoke.texi (AVR Built-in Functions): Re-layout section.
24000 2016-07-21  Jan Hubicka  <hubicka@ucw.cz>
24002         * cfgcleanup.c (flow_find_cross_jump): Do not crossjump across
24003         hot/cold regions.
24004         (try_crossjump_to_edge): Do not punt on partitioned functions.
24006 2016-07-21  Jan Hubicka  <hubicka@ucw.cz>
24008         * bb-reorder.c (find_rarely_executed_basic_blocks_and_crossing_edges):
24009         Put all BBs reachable only via paths crossing cold region to cold
24010         region.
24011         * cfgrtl.c (find_bbs_reachable_by_hot_paths): New function.
24013 2016-07-21  Richard Biener  <rguenther@suse.de>
24015         PR tree-optimization/81303
24016         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Take
24017         into account prologue and epilogue iterations when raising
24018         min_profitable_iters to sth at least covering one vector iteration.
24020 2017-07-21  Tamar Christina  <tamar.christina@arm.com>
24022         * config/arm/arm.c (arm_test_cpu_arch_dat):
24023         Check for overlap.
24025 2017-07-20  Nathan Sidwell  <nathan@acm.org>
24027         Remove TYPE_METHODS.
24028         * tree.h (TYPE_METHODS): Delete.
24029         * dwarf2out.c (gen_member_die): Member fns are on TYPE_FIELDS.
24030         * dbxout.c (dbxout_type_fields): Ignore FUNCTION_DECLs.
24031         (dbxout_type_methods): Scan TYPE_FIELDS.
24032         (dbxout_type): Don't check TYPE_METHODS here.
24033         * function.c (use_register_for_decl): Always ignore register for
24034         class types when not optimizing.
24035         * ipa-devirt.c (odr_types_equivalent_p): Delete TYPE_METHODS scan.
24036         * tree.c (free_lang_data_in_type): Stitch out member functions and
24037         templates from TYPE_FIELDS.
24038         (build_distinct_type_copy, verify_type_variant,
24039         verify_type): Member fns are on TYPE_FIELDS.
24040         * tree-dump.c (dequeue_and_dump): No TYPE_METHODS.
24041         * tree-pretty-print.c (dump_generic_node): Likewise.
24043 2017-07-20  Jakub Jelinek  <jakub@redhat.com>
24045         PR target/80846
24046         * config/i386/i386.c (ix86_expand_vector_init_general): Handle
24047         V2TImode and V4TImode.
24048         (ix86_expand_vector_extract): Likewise.
24049         * config/i386/sse.md (VMOVE): Enable V4TImode even for just
24050         TARGET_AVX512F, instead of only for TARGET_AVX512BW.
24051         (ssescalarmode): Handle V4TImode and V2TImode.
24052         (VEC_EXTRACT_MODE): Add V4TImode and V2TImode.
24053         (*vec_extractv2ti, *vec_extractv4ti): New insns.
24054         (VEXTRACTI128_MODE): New mode iterator.
24055         (splitter for *vec_extractv?ti first element): New.
24056         (VEC_INIT_MODE): New mode iterator.
24057         (vec_init<mode>): Consolidate 3 expanders into one using
24058         VEC_INIT_MODE mode iterator.
24060 2017-07-20  Alexander Monakov  <amonakov@ispras.ru>
24062         * lra-assigns.c (pseudo_compare_func): Fix comparison step based on
24063         non_spilled_static_chain_regno_p.
24065 2017-07-20  Alexander Monakov  <amonakov@ispras.ru>
24067         * gimple-ssa-store-merging.c (sort_by_bitpos): Return 0 on equal bitpos.
24069 2017-07-20  Jan Hubicka  <hubicka@ucw.cz>
24071         * bb-reorder.c (connect_traces): Allow copying of blocks within
24072         single partition.
24074 2017-07-20  Richard Biener  <rguenther@suse.de>
24076         * gimple.h (gimple_phi_result): Add gphi * overload.
24077         (gimple_phi_result_ptr): Likewise.
24078         (gimple_phi_arg): Likewise.  Adjust index assert to only
24079         allow actual argument accesses rather than all slots available
24080         by capacity.
24081         (gimple_phi_arg_def): Add gphi * overload.
24082         * tree-phinodes.c (make_phi_node): Initialize only actual
24083         arguments.
24084         (resize_phi_node): Clear memory not covered by old node,
24085         do not initialize excess argument slots.
24086         (reserve_phi_args_for_new_edge): Initialize new argument slot
24087         completely.
24089 2017-07-20  Bin Cheng  <bin.cheng@arm.com>
24091         PR tree-optimization/81388
24092         Revert r238585:
24093         2016-07-21  Bin Cheng  <bin.cheng@arm.com>
24095         * tree-ssa-loop-niter.c (number_of_iterations_lt_to_ne): Clean up
24096         by removing computation of may_be_zero.
24098 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
24099             Tom de Vries  <tom@codesourcery.com>
24101         PR middle-end/81030
24102         * cfgbuild.c (find_many_sub_basic_blocks): Update REG_BR_PROB note
24103         when gimple level profile disagrees with what RTL expander did.
24105 2017-07-20  Richard Biener  <rguenther@suse.de>
24107         PR tree-optimization/61171
24108         * tree-vectorizer.h (slp_instance): Add reduc_phis member.
24109         (vect_analyze_stmt): Add slp instance parameter.
24110         (vectorizable_reduction): Likewise.
24111         * tree-vect-loop.c (vect_analyze_loop_operations): Adjust.
24112         (vect_is_simple_reduction): Deal with chains not detected
24113         as SLP reduction chain, specifically not properly associated
24114         chains containing a mix of plus/minus.
24115         (get_reduction_op): Remove.
24116         (get_initial_defs_for_reduction): Simplify, pass in whether
24117         this is a reduction chain, pass in the SLP node for the PHIs.
24118         (vect_create_epilog_for_reduction): Get the SLP instance as
24119         arg and adjust.
24120         (vectorizable_reduction): Get the SLP instance as arg.
24121         During analysis remember the SLP node with the PHIs in the
24122         instance.  Simplify getting at the vectorized reduction PHIs.
24123         * tree-vect-slp.c (vect_slp_analyze_node_operations): Pass
24124         through SLP instance.
24125         (vect_slp_analyze_operations): Likewise.
24126         * tree-vect-stms.c (vect_analyze_stmt): Likewise.
24127         (vect_transform_stmt): Likewise.
24129 2017-07-20  Tom de Vries  <tom@codesourcery.com>
24131         PR tree-optimization/81489
24132         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior): Move
24133         read of phi arg location to before loop that modifies phi.
24135 2017-07-20  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
24137         * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0):
24138         New pattern.
24140 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
24142         PR middle-end/81331
24143         * except.c (execute): Fix ordering issue.
24145 2018-07-19  Segher Boessenkool  <segher@kernel.crashing.org>
24147         PR rtl-optimization/81423
24148         * combine.c (make_compound_operation_int): Don't try to optimize
24149         the AND of a SUBREG of an LSHIFTRT if that SUBREG is paradoxical.
24151 2017-07-19  Segher Boessenkool  <segher@kernel.crashing.org>
24153         PR rtl-optimization/81423
24154         * simplify-rtx.c (simplify_truncation): Handle truncating an IOR
24155         with a constant that is -1 in the truncated to mode.
24157 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
24159         * predict.c (propagate_unlikely_bbs_forward): Break out from ...
24160         (determine_unlikely_bbs): ... here.
24161         * predict.h (propagate_unlikely_bbs_forward): Declare.
24162         * cfgexpand.c (pass_expand::execute): Use it.
24163         * bb-reorder.c (sanitize_hot_paths): Do not consider known to be
24164         unlikely edges.
24165         (find_rarely_executed_basic_blocks_and_crossing_edges): Use
24166         propagate_unlikely_bbs_forward.
24168 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
24170         PR middle-end/81331
24171         * except.c (maybe_add_nop_after_section_switch): New function.
24172         (execute): Use it.
24174 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24176         * gimple.h (gimple_phi_set_arg): Make assert more strict.
24178 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24180         * gimple.h (gimple_phi_arg): Make assert more strict.
24182 2017-07-19  Steven Munroe  <munroesj@gcc.gnu.org>
24184         * config.gcc (powerpc*-*-*): Add mmintrin.h.
24185         * config/rs6000/mmintrin.h: New file.
24186         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include mmintrin.h.
24188 2017-07-19  Jakub Jelinek  <jakub@redhat.com>
24190         PR tree-optimization/81346
24191         * match.pd: Optimize (X - 1U) <= INT_MAX-1U into (int) X > 0.
24193 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24195         * config/nvptx/nvptx.md (VECIM): Add V2DI.
24197 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24199         * config/nvptx/nvptx-modes.def: Add V2DImode.
24200         * config/nvptx/nvptx-protos.h (nvptx_data_alignment): Declare.
24201         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Handle V2DImode.
24202         (nvptx_output_mov_insn): Handle lack of mov.b128.
24203         (nvptx_print_operand): Handle 'H' and 'L' codes.
24204         (nvptx_vector_mode_supported): Allow V2DImode.
24205         (nvptx_preferred_simd_mode): New function.
24206         (nvptx_data_alignment): New function.
24207         (TARGET_VECTORIZE_PREFERRED_SIMD_MODE): Redefine to
24208         nvptx_preferred_simd_mode.
24209         * config/nvptx/nvptx.h (STACK_BOUNDARY, BIGGEST_ALIGNMENT): Change from
24210         64 to 128 bits.
24211         (DATA_ALIGNMENT): Define.  Set to nvptx_data_alignment.
24213 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24215         * config/nvptx/nvptx-modes.def: New file.  Add V2SImode.
24216         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Handle V2SImode.
24217         (nvptx_vector_mode_supported): New function.  Allow V2SImode.
24218         (TARGET_VECTOR_MODE_SUPPORTED_P): Redefine to nvptx_vector_mode_supported.
24219         * config/nvptx/nvptx.md (VECIM): New mode iterator. Add V2SI.
24220         (mov<VECIM>_insn): New define_insn.
24221         (define_expand "mov<VECIM>): New define_expand.
24223 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24225         * config/nvptx/nvptx.c (nvptx_print_operand): Handle v2 vector mode.
24227 2017-07-19  Jakub Jelinek  <jakub@redhat.com>
24229         PR tree-optimization/81346
24230         * fold-const.h (fold_div_compare, range_check_type): Declare.
24231         * fold-const.c (range_check_type): New function.
24232         (build_range_check): Use range_check_type.
24233         (fold_div_compare): No longer static, rewritten into
24234         a match.pd helper function.
24235         (fold_comparison): Don't call fold_div_compare here.
24236         * match.pd (X / C1 op C2): New optimization using fold_div_compare
24237         as helper function.
24239 2017-07-19  Nathan Sidwell  <nathan@acm.org>
24241         * tree.h (TYPE_MINVAL, TYPE_MAXVAL): Rename to ...
24242         (TYPE_MIN_VALUE_RAW, TYPE_MAX_VALUE_RAW): ... these.
24243         * tree.c (find_decls_types_r, verify_type): Use
24244         TYPE_{MIN,MAX}_VALUE_RAW.
24245         * lto-streamer-out.c (DFS::DFS_write_tree_body): Likewise.
24246         (hash_tree): Likewise.
24247         * tree-streamer-in.c (lto_input_ts_type_non_common_tree_pointers):
24248         Likewise.
24249         * tree-streamer-out.c (write_ts_type_non_common_tree_pointers):
24250         Likewise.
24252 2017-07-18  Tom de Vries  <tom@codesourcery.com>
24254         PR middle-end/81464
24255         * omp-expand.c (expand_omp_for_static_chunk): Handle
24256         equal-argument loop exit phi.
24258 2017-07-18  Uros Bizjak  <ubizjak@gmail.com>
24260         PR target/81471
24261         * config/i386/i386.md (rorx_immediate_operand): New mode attribute.
24262         (*bmi2_rorx<mode>3_1): Use rorx_immediate_operand as
24263         operand 2 predicate.
24264         (*bmi2_rorxsi3_1_zext): Use const_0_to_31_operand as
24265         operand 2 predicate.
24266         (ror,rol -> rorx splitters): Use const_int_operand as
24267         operand 2 predicate.
24269 2017-06-18  Richard Biener  <rguenther@suse.de>
24271         PR tree-optimization/81410
24272         * tree-vect-stmts.c (vectorizable_load): Properly adjust for
24273         the gap in the ! slp_perm SLP case after each group.
24275 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
24277         PR middle-end/81463
24278         * cfgloopmanip.c (scale_loop_profile): Watch out for zero frequency
24279         again.
24281 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
24283         PR middle-end/81462
24284         * predict.c (set_even_probabilities): Cleanup; do not affect
24285         probabilities that are already known.
24286         (combine_predictions_for_bb): Call even when count is set.
24288 2017-07-18  Nathan Sidwell  <nathan@acm.org>
24290         * tree-parloops.c (try_transform_to_exit_first_loop_alt): Use
24291         TYPE_MAX_VALUE.
24293 2017-07-18  Bin Cheng  <bin.cheng@arm.com>
24295         PR target/81408
24296         * tree-ssa-loop-niter.c (number_of_iterations_exit): Dump missed
24297         optimization for loop niter analysis.
24299 2017-07-18  Georg-Johann Lay  <avr@gjlay.de>
24301         PR target/81473
24302         * config/avr/avr.c (avr_optimize_casesi): Don't use
24303         INT8_MIN, INT8_MAX, UINT8_MAX, INT16_MIN, INT16_MAX, UINT16_MAX.
24305 2017-07-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
24307         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Remove
24308         body_cost_vec from _vect_peel_extended_info.
24309         (vect_peeling_hash_get_lowest_cost): Do not set body_cost_vec.
24310         (vect_peeling_hash_choose_best_peeling): Remove body_cost_vec and
24311         npeel.
24313 2017-07-18  Bin Cheng  <bin.cheng@arm.com>
24315         * config/arm/arm.c (emit_unlikely_jump): Remove unused var.
24317 2017-07-18  Richard Biener  <rguenther@suse.de>
24319         PR tree-optimization/80620
24320         PR tree-optimization/81403
24321         * tree-ssa-pre.c (phi_translate_1): Clear range and points-to
24322         info when re-using a VN table entry.
24324 2017-07-18  Richard Biener  <rguenther@suse.de>
24326         PR tree-optimization/81418
24327         * tree-vect-loop.c (vectorizable_reduction): Properly compute
24328         vectype_in.  Verify that with lane-reducing reduction operations
24329         we have a single def-use cycle.
24331 2017-07-17  Carl Love  <cel@us.ibm.com>
24333         Revert commit r249424   2017-06-20  Carl Love  <cel@us.ibm.com>
24335         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
24336         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
24337         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
24338         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
24339         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
24340         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
24341         VMULOSW): New enum "unspec" values.
24342         (vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
24343         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si,
24344         altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
24345         altivec_vmulosw): New patterns.
24346         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
24347         VMULOSW): Add definitions.
24349 2017-07-17  Uros Bizjak  <ubizjak@gmail.com>
24351         * config/alpha/alpha.c: Include predict.h.
24353 2017-07-17  Yury Gribov  <tetra2005@gmail.com>
24355         * tree-vrp.c (compare_assert_loc): Fix comparison function
24356         to return predictable results.
24358 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24360         * config/arc/arc.md (adddi3): Remove support for mexpand-adddi
24361         option.
24362         (subdi3): Likewise.
24363         * config/arc/arc.opt (mexpand-adddi): Deprecate it.
24364         * doc/invoke.texi (mexpand-adddi): Update text.
24366 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24368         * config/arc/arc.md (clzsi2): Expand to an arc_clzsi2 instruction
24369         that also clobbers the CC register. The old expand code is moved
24370         to ...
24371         (*arc_clzsi2): ... here.
24372         (ctzsi2): Expand to an arc_ctzsi2 instruction that also clobbers
24373         the CC register. The old expand code is moved to ...
24374         (arc_ctzsi2): ... here.
24376 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24378         * config/arc/arc.opt (mindexed-loads): Use initial value
24379         TARGET_INDEXED_LOADS_DEFAULT.
24380         (mauto-modify-reg): Use initial value
24381         TARGET_AUTO_MODIFY_REG_DEFAULT.
24382         * config/arc/elf.h (TARGET_INDEXED_LOADS_DEFAULT): Define.
24383         (TARGET_AUTO_MODIFY_REG_DEFAULT): Likewise.
24384         * config/arc/linux.h (TARGET_INDEXED_LOADS_DEFAULT): Define.
24385         (TARGET_AUTO_MODIFY_REG_DEFAULT): Likewise.
24387 2017-07-17  Martin Liska  <mliska@suse.cz>
24389         PR sanitizer/81302
24390         * opts.c (finish_options): Do not allow -fgnu-tm
24391         w/ -fsanitize={kernel-,}address.  Say sorry.
24393 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24395         PR target/81369
24396         * tree-loop-distribution.c (classify_partition): Only assert on
24397         numer of iterations.
24398         (merge_dep_scc_partitions): Delete prameter.  Update function call.
24399         (distribute_loop): Remove code handling loop with unknown niters.
24400         (pass_loop_distribution::execute): Skip loop with unknown niters.
24402 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24404         PR target/81369
24405         * tree-loop-distribution.c (merge_dep_scc_partitions): Sink call to
24406         function sort_partitions_by_post_order.
24408 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24410         PR tree-optimization/81374
24411         * tree-loop-distribution.c (pass_loop_distribution::execute): Record
24412         the max index of basic blocks, rather than number of basic blocks.
24414 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24416         * config/arc/arc-protos.h (arc_legitimate_pc_offset_p): Remove
24417         proto.
24418         (arc_legitimate_pic_operand_p): Likewise.
24419         * config/arc/arc.c (arc_legitimate_pic_operand_p): Remove
24420         function.
24421         (arc_needs_pcl_p): Likewise.
24422         (arc_legitimate_pc_offset_p): Likewise.
24423         (arc_legitimate_pic_addr_p): Remove LABEL_REF case, as this
24424         function is also used in constrains.md.
24425         (arc_legitimate_constant_p): Use arc_legitimate_pic_addr_p to
24426         validate pic constants. Handle CONST_INT, CONST_DOUBLE, MINUS and
24427         PLUS.  Only return true/false in known cases, otherwise assert.
24428         (arc_legitimate_address_p): Remove arc_legitimate_pic_addr_p as it
24429         is already called in arc_legitimate_constant_p.
24430         * config/arc/arc.h (CONSTANT_ADDRESS_P): Consider also LABEL for
24431         pic addresses.
24432         (LEGITIMATE_PIC_OPERAND_P): Use
24433         arc_raw_symbolic_reference_mentioned_p function.
24434         * config/arc/constraints.md (Cpc): Use arc_legitimate_pic_addr_p
24435         function.
24436         (Cal): Likewise.
24437         (C32): Likewise.
24439 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24440         Andrew Burgess  <andrew.burgess@embecosm.com>
24442         * config/arc/arc-protos.h (arc_compute_function_type): Change prototype.
24443         (arc_return_address_register): New function.
24444         * config/arc/arc.c (arc_handle_fndecl_attribute): New function.
24445         (arc_handle_fndecl_attribute): Add naked attribute.
24446         (TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS): Define.
24447         (TARGET_WARN_FUNC_RETURN): Likewise.
24448         (arc_allocate_stack_slots_for_args): New function.
24449         (arc_warn_func_return): Likewise.
24450         (machine_function): Change type fn_type.
24451         (arc_compute_function_type): Consider new naked function type,
24452         change function return type.
24453         (arc_must_save_register): Adapt to handle new
24454         arc_compute_function_type's return type.
24455         (arc_expand_prologue): Likewise.
24456         (arc_expand_epilogue): Likewise.
24457         (arc_return_address_regs): Delete.
24458         (arc_return_address_register): New function.
24459         (arc_epilogue_uses): Use above function.
24460         * config/arc/arc.h (arc_return_address_regs): Delete prototype.
24461         (arc_function_type): Change encoding, add naked type.
24462         (ARC_INTERRUPT_P): Change to handle the new encoding.
24463         (ARC_FAST_INTERRUPT_P): Likewise.
24464         (ARC_NORMAL_P): Define.
24465         (ARC_NAKED_P): Likewise.
24466         (arc_compute_function_type): Delete prototype.
24467         * config/arc/arc.md (in_ret_delay_slot): Use
24468         arc_return_address_register function.
24469         (simple_return): Likewise.
24470         (p_return_i): Likewise.
24472 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24474         PR tree-optimization/81428
24475         * match.pd (X / X -> one): Don't optimize _Fract divisions, as 1
24476         can't be built for those types.
24478 2017-07-17  Georg-Johann Lay  <avr@gjlay.de>
24480         Remove stuff dead since r239246.
24482         * config/avr/avr-arch.h (avr_inform_devices): Remove dead proto.
24483         * config/avr/avr-devices.c (mcu_name, comparator, avr_mcus_str)
24484         (avr_inform_devices): Remove dead stuff.
24486 2017-07-17  Tamar Christina  <tamar.christina@arm.com>
24488         * config/arm/arm_neon.h: Fix softp typo.
24490 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24492         PR tree-optimization/81365
24493         * tree-ssa-phiprop.c (propagate_with_phi): When considering hoisting
24494         aggregate moves onto bb predecessor edges, make sure there are no
24495         loads that could alias the lhs in between the start of bb and the
24496         loads from *phi.
24498 2017-07-17  Georg-Johann Lay  <avr@gjlay.de>
24500         PR 80929
24501         * config/avr/avr.c (avr_mul_highpart_cost): New static function.
24502         (avr_rtx_costs_1) [TRUNCATE]: Use it to compute mul_highpart cost.
24503         [LSHIFTRT, outer_code = TRUNCATE]: Same.
24505 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24507         PR tree-optimization/81396
24508         * tree-ssa-math-opts.c (struct symbolic_number): Add n_ops field.
24509         (init_symbolic_number): Initialize it to 1.
24510         (perform_symbolic_merge): Add n_ops from both operands into the new
24511         n_ops.
24512         (find_bswap_or_nop): Don't consider n->n == cmpnop computations
24513         without base_addr as useless if they need more than one operation.
24514         (bswap_replace): Handle !bswap case for NULL base_addr.
24516 2017-07-17  Tom de Vries  <tom@codesourcery.com>
24518         PR target/81069
24519         * config/nvptx/nvptx.c (nvptx_single): Insert diverging branch as late
24520         as possible.
24522 2017-07-17  Sebastian Huber  <sebastian.huber@embedded-brains.de>
24524         * config/sparc/rtemself.h (TARGET_OS_CPP_BUILTINS): Add
24525         conditional builtin define __FIX_LEON3FT_B2BST.
24527 2017-07-17  Daniel Cederman  <cederman@gaisler.com>
24529         * config/sparc/t-rtems: Add mfix-gr712rc multilibs. Replace
24530         MULTILIB_EXCEPTIONS with MULTILIB_REQUIRED. Match -mfix-gr712rc
24531         with -mfix-ut700.
24533 2017-07-16  Eric Botcazou  <ebotcazou@adacore.com>
24535         PR rtl-optimization/81424
24536         * optabs.c (prepare_cmp_insn): Use copy_to_reg instead of force_reg
24537         to remove potential trapping from operands if -fnon-call-exceptions.
24539 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24541         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Use
24542         profile_proability for scalling.
24543         * scale_profile_for_vect_loop.c (scale_profile_for_vect_loop): Likewise.
24545 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24547         * cgraph.c (cgraph_edge::redirect_call_stmt_to_caller): Cleanup.
24549 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24551         * cfgloopmanip.c (scale_loop_profile): Avoid use of REG_BR_PROB_BASE
24552         fixpoint arithmetics.
24554 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24556         * tree-ssa-loop-unswitch.c (hoist_guard): Avoid use of REG_BR_PROB_BASE
24557         fixpoint arithmetics.
24559 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24561         * asan.c (create_cond_insert_point): Avoid use of REG_BR_PROB_BASE
24562         fixpoint arithmetics.
24564 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24566         * profile-count.h (profile_probability::from_reg_br_prob_note,
24567         profile_probability::to_reg_br_prob_note): New functions.
24568         * doc/rtl.texi (REG_BR_PROB_NOTE): Update documentation.
24569         * reg-notes.h (REG_BR_PROB, REG_BR_PRED): Update docs.
24570         * predict.c (probability_reliable_p): Update.
24571         (edge_probability_reliable_p): Update.
24572         (br_prob_note_reliable_p): Update.
24573         (invert_br_probabilities): Update.
24574         (add_reg_br_prob_note): New function.
24575         (combine_predictions_for_insn): Update.
24576         * asan.c (asan_clear_shadow): Update.
24577         * cfgbuild.c (compute_outgoing_frequencies): Update.
24578         * cfgrtl.c (force_nonfallthru_and_redirect): Update.
24579         (update_br_prob_note): Update.
24580         (rtl_verify_edges): Update.
24581         (purge_dead_edges): Update.
24582         (fixup_reorder_chain): Update.
24583         * emit-rtl.c (try_split): Update.
24584         * ifcvt.c (cond_exec_process_insns): Update.
24585         (cond_exec_process_if_block): Update.
24586         (dead_or_predicable): Update.
24587         * internal-fn.c (expand_addsub_overflow): Update.
24588         (expand_neg_overflow): Update.
24589         (expand_mul_overflow): Update.
24590         * loop-doloop.c (doloop_modify): Update.
24591         * loop-unroll.c (compare_and_jump_seq): Update.
24592         * optabs.c (emit_cmp_and_jump_insn_1): Update.
24593         * predict.h: Update.
24594         * reorg.c (mostly_true_jump): Update.
24595         * rtl.h: Update.
24596         * config/aarch64/aarch64.c (aarch64_emit_unlikely_jump): Update.
24597         * config/alpha/alpha.c (emit_unlikely_jump): Update.
24598         * config/arc/arc.c: (emit_unlikely_jump): Update.
24599         * config/arm/arm.c: (emit_unlikely_jump): Update.
24600         * config/bfin/bfin.c (cbranch_predicted_taken_p): Update.
24601         * config/frv/frv.c (frv_print_operand_jump_hint): Update.
24602         * config/i386/i386.c (ix86_expand_split_stack_prologue): Update.
24603         (ix86_print_operand): Update.
24604         (ix86_split_fp_branch): Update.
24605         (predict_jump): Update.
24606         * config/ia64/ia64.c (ia64_print_operand): Update.
24607         * config/mmix/mmix.c (mmix_print_operand): Update.
24608         * config/powerpcspe/powerpcspe.c (output_cbranch): Update.
24609         (rs6000_expand_split_stack_prologue): Update.
24610         * config/rs6000/rs6000.c: Update.
24611         * config/s390/s390.c (s390_expand_vec_strlen): Update.
24612         (s390_expand_vec_movstr): Update.
24613         (s390_expand_cs_tdsi): Update.
24614         (s390_expand_split_stack_prologue): Update.
24615         * config/sh/sh.c (sh_print_operand): Update.
24616         (expand_cbranchsi4): Update.
24617         (expand_cbranchdi4): Update.
24618         * config/sparc/sparc.c (output_v9branch): Update.
24619         * config/spu/spu.c (get_branch_target): Update.
24620         (ea_load_store_inline): Update.
24621         * config/tilegx/tilegx.c (cbranch_predicted_p): Update.
24622         * config/tilepro/tilepro.c: Update.
24624 2017-07-16  Eric Botcazou  <ebotcazou@adacore.com>
24626         * gimplify.c (mostly_copy_tree_r): Revert latest change.
24627         (gimplify_save_expr): Likewise.
24629 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24631         * ipa-visibility.c (function_and_variable_visibility): Fix pasto.
24633 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24635         * ipa-fnsummary.c (pass_data_ipa_fn_summary): Use
24636         TV_IPA_FNSUMMARY.
24637         * timevar.def (TV_IPA_FNSUMMARY): Define.
24639 2017-07-16  Daniel Cederman  <cederman@gaisler.com>
24641         * config/sparc/sparc.md (divdf3_fix): Add NOP to prevent back
24642         to back store errata sensitive sequence from being generated.
24643         (sqrtdf2_fix): Likewise.
24645 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24647         * tree-ssa-threadupdate.c (compute_path_counts,
24648         update_joiner_offpath_counts): Use profile_probability.
24650 2017-07-15  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24652         Revert:
24653         2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24655         * config/arm/arm-c.c (arm_cpu_builtins): Define
24656         __ARM_FEATURE_NUMERIC_MAXMIN solely based on TARGET_VFP5.
24658 2017-07-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
24660         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
24661         array entries to represent __ieee128 versions of the
24662         scalar_test_data_class, scalar_test_neg, scalar_extract_exp,
24663         scalar_extract_sig, and scalar_insert_exp built-in functions.
24664         (altivec_resolve_overloaded_builtin): Add special case handling
24665         for the __builtin_scalar_insert_exp function, as represented by
24666         the P9V_BUILTIN_VEC_VSIEDP constant.
24667         * config/rs6000/rs6000-builtin.def (VSEEQP): Add scalar extract
24668         exponent support for __ieee128 argument.
24669         (VSESQP): Add scalar extract signature support for __ieee128
24670         argument.
24671         (VSTDCNQP): Add scalar test negative support for __ieee128
24672         argument.
24673         (VSIEQP): Add scalar insert exponent support for __int128 argument
24674         with __ieee128 result.
24675         (VSIEQPF): Add scalar insert exponent support for __ieee128
24676         argument with __ieee128 result.
24677         (VSTDCQP): Add scalar test data class support for __ieee128
24678         argument.
24679         (VSTDCNQP): Add overload support for scalar test negative with
24680         __ieee128 argument.
24681         (VSTDCQP): Add overload support for scalar test data class
24682         __ieee128 argument.
24683         * config/rs6000/vsx.md (UNSPEC_VSX_SXSIG) Replace
24684         UNSPEC_VSX_SXSIGDP.
24685         (UNSPEC_VSX_SIEXPQP): New constant.
24686         (xsxexpqp): New insn for VSX scalar extract exponent quad
24687         precision.
24688         (xsxsigqp): New insn for VSX scalar extract significand quad
24689         precision.
24690         (xsiexpqpf): New insn for VSX scalar insert exponent quad
24691         precision with floating point argument.
24692         (xststdcqp): New expand for VSX scalar test data class quad
24693         precision.
24694         (xststdcnegqp): New expand for VSX scalar test negative quad
24695         precision.
24696         (xststdcqp): New insn to match expansions for VSX scalar test data
24697         class quad precision and VSX scalar test negative quad precision.
24698         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Add
24699         special case operand checking to enforce that second operand of
24700         VSX scalar test data class with quad precision argument is a 7-bit
24701         unsigned literal.
24702         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
24703         prototypes and descriptions of __ieee128 versions of
24704         scalar_extract_exp, scalar_extract_sig, scalar_insert_exp,
24705         scalar_test_data_class, and scalar_test_neg built-in functions.
24707 2016-07-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24709         PR tree-optimization/81162
24710         * gimple-ssa-strength-reduction.c (replace_mult_candidate): Don't
24711         replace a negate with an add.
24713 2017-07-14  James Greenhalgh  <james.greenhalgh@arm.com>
24715         * doc/invoke.texi (arm/-mcpu): Document +crypto.
24717 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24719         * config/arm/arm-c.c (arm_cpu_builtins): Define
24720         __ARM_FEATURE_NUMERIC_MAXMIN solely based on TARGET_VFP5.
24722 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24724         * config/arm/arm-cpus.in (cortex-r52): Add new entry.
24725         (armv8-r): Set ARM Cortex-R52 as default CPU.
24726         * config/arm/arm-tables.opt: Regenerate.
24727         * config/arm/arm-tune.md: Regenerate.
24728         * config/arm/driver-arm.c (arm_cpu_table): Add entry for ARM
24729         Cortex-R52.
24730         * doc/invoke.texi: Mention -mtune=cortex-r52 and availability of fp.dp
24731         extension for -mcpu=cortex-r52.
24733 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24735         * config/arm/arm-isa.h (isa_bit_FP_ARMv8): Delete enumerator.
24736         (ISA_FP_ARMv8): Define as ISA_FPv5 and ISA_FP_D32.
24737         * config/arm/arm-cpus.in (armv8-r): Define fp.sp as enabling FPv5.
24738         (fp-armv8): Define it as FP_ARMv8 only.
24739         config/arm/arm.h (TARGET_FPU_ARMV8): Delete.
24740         (TARGET_VFP_FP16INST): Define using TARGET_VFP5 rather than
24741         TARGET_FPU_ARMV8.
24742         config/arm/arm.c (arm_rtx_costs_internal): Replace checks against
24743         TARGET_FPU_ARMV8 by checks against TARGET_VFP5.
24744         * config/arm/arm-builtins.c (arm_builtin_vectorized_function): Define
24745         first ARM_CHECK_BUILTIN_MODE definition using TARGET_VFP5 rather
24746         than TARGET_FPU_ARMV8.
24747         * config/arm/arm-c.c (arm_cpu_builtins): Likewise for
24748         __ARM_FEATURE_NUMERIC_MAXMIN macro definition.
24749         * config/arm/arm.md (cmov<mode>): Condition on TARGET_VFP5 rather than
24750         TARGET_FPU_ARMV8.
24751         * config/arm/neon.md (neon_vrint): Likewise.
24752         (neon_vcvt): Likewise.
24753         (neon_<fmaxmin_op><mode>): Likewise.
24754         (<fmaxmin><mode>3): Likewise.
24755         * config/arm/vfp.md (l<vrint_pattern><su_optab><mode>si2): Likewise.
24756         * config/arm/predicates.md (arm_cond_move_operator): Check against
24757         TARGET_VFP5 rather than TARGET_FPU_ARMV8 and fix spacing.
24759 2017-07-14  Jackson Woodruff  <jackson.woodruff@arm.com>
24761         * config/aarch64/aarch64.c (aarch64_print_operand): Move comments
24762         to top of function.
24764 2017-07-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24766         * gimple-ssa-store-merging.c (clear_bit_region): Replace reference to
24767         loop in comment with memset.
24769 2017-07-14  Martin Liska  <mliska@suse.cz>
24771         * cfgexpand.c (expand_gimple_basic_block): Remove dead comment.
24772         * dwarf2out.c (is_java): Remove the function.
24773         (output_pubname): Remove usage of the function.
24774         (lower_bound_default): Remove usage of DW_LANG_Java.
24775         (gen_compile_unit_die): Likewise.
24776         * gcc.c: Remove compiler defaults for .java and .zip files.
24777         * gimple-expr.c (remove_suffix): Change as there's no longer
24778         extension than 4-letter one.
24779         * gimplify.c (mostly_copy_tree_r): Remove Java-special part.
24780         (gimplify_save_expr): Likewise.
24781         * ipa-utils.h (polymorphic_type_binfo_p): Remove the comment
24782         as it's possible even for other languages than Java.
24783         * langhooks.h (struct lang_hooks): Remove Java from a comment.
24784         * lto-opts.c (lto_write_options): Remove reference to Java.
24785         * opts.c (strip_off_ending): Update file extension handling.
24786         * tree-cfg.c (verify_gimple_call): Remove comment with Java.
24787         * tree-eh.c (lower_resx): Likewise.
24788         * tree.c (free_lang_data_in_type): Remove dead code.
24789         (find_decls_types_r): Likewise.
24790         (build_common_builtin_nodes): Remove Java from a comment.
24791         (verify_type): Remove dead code.
24792         * varasm.c (assemble_external): Remove Java from a comment.
24794 2017-07-14  Martin Liska  <mliska@suse.cz>
24796         * opts.c (finish_options): Add quotes.
24797         (common_handle_option): Likewise.
24799 2017-07-14  Martin Liska  <mliska@suse.cz>
24801         * dbxout.c (get_lang_number): Do not handle GNU Pascal.
24802         * dbxout.h (extern void dbxout_stab_value_internal_label_diff):
24803         Remove N_SO_PASCAL.
24804         * dwarf2out.c (lower_bound_default): Do not handle
24805         DW_LANG_Pascal83.
24806         (gen_compile_unit_die): Likewise.
24807         * gcc.c: Remove default extension binding for GNU Pascal.
24808         * stmt.c: Remove Pascal language from a comment.
24809         * xcoffout.c: Likewise.
24811 2017-07-13  David Malcolm  <dmalcolm@redhat.com>
24813         PR c/81405
24814         * diagnostic-show-locus.c (fixit_cmp): New function.
24815         (layout::layout): Sort m_fixit_hints.
24816         (column_range::column_range): Assert that the values are valid.
24817         (struct char_span): New struct.
24818         (correction::overwrite): New method.
24819         (struct source_line): New struct.
24820         (line_corrections::add_hint): Add assertions.  Reimplement memcpy
24821         calls in terms of classes source_line and char_span, and
24822         correction::overwrite.
24823         (selftest::test_overlapped_fixit_printing_2): New function.
24824         (selftest::diagnostic_show_locus_c_tests): Call it.
24826 2017-07-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
24828         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Return
24829         early if there is no lhs.
24831 2017-07-13  Martin Liska  <mliska@suse.cz>
24833         * dwarf2out.c (gen_pointer_type_die): Remove dead code.
24834         (gen_reference_type_die): Likewise.
24835         * stor-layout.c: Remove Pascal-related comment.
24837 2017-07-13  Martin Liska  <mliska@suse.cz>
24839         * opts.c (finish_options): Add quotes to error messages.
24840         (parse_sanitizer_options): Likewise.
24842 2017-07-13  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24844         * doc/invoke.texi (armv8-r): Document +fp.sp ARMv8-R extension.
24846 2017-07-13  Richard Earnshaw  <rearnsha@arm.com>
24848         * config/arm/vxworks.h (TARGET_ENDIAN_DEFAULT): Define.
24850 2017-07-13  Maxim Ostapenko  <m.ostapenko@samsung.com>
24852         * asan.c (asan_emit_allocas_unpoison): Use ptr_mode for arguments
24853         during expansion.
24854         * builtins.c (expand_asan_emit_allocas_unpoison): Likewise.
24856 2017-07-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
24858         PR target/81193
24859         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If GLIBC
24860         provides the hardware capability bits, define the macro
24861         __BUILTIN_CPU_SUPPORTS__.
24862         * config/rs6000/rs6000.c (cpu_expand_builtin): Generate a warning
24863         if GLIBC does not provide the hardware capability bits.  Add a
24864         gcc_unreachable call if the built-in cpu function is neither
24865         __builtin_cpu_is nor __builtin_cpu_supports.
24866         (rs6000_get_function_versions_dispatcher): Change the warning
24867         that an old GLIBC is used which does not export the capability
24868         bits to be an error.
24869         * doc/extend.texi (target_clones attribute): Document the
24870         restriction that GLIBC 2.23 or newer is needed on the PowerPC.
24871         (PowerPC built-in functions): Document that GLIBC 2.23 or newer is
24872         needed by __builtin_cpu_is and __builtin_cpu_supports.  Document
24873         the macros defined by GCC if the newer GLIBC is available.
24875 2017-07-12  Jeff Law  <law@redhat.com>
24877         * config/riscv/riscv.c: Remove unnecessary includes.  Reorder
24878         remaining includes slightly.
24879         * config/riscv/riscv-builtins.c: Include profile-count.h.
24881 2017-07-12  Georg-Johann Lay  <avr@gjlay.de>
24883         PR target/79883
24884         * config/avr/avr.c (avr_set_current_function): In diagnostic
24885         messages: Quote keywords and (parts of) identifiers.
24886         [WITH_AVRLIBC]: Warn for functions named "ISR", "SIGNAL" or
24887         "INTERUPT".
24889 2017-07-12  Carl Love  <cel@us.ibm.com>
24891         * config/rs6000/rs6000-c.c: Add support for built-in functions
24892         vector bool char vec_revb (vector bool char);
24893         vector bool short vec_revb (vector short char);
24894         vector bool int vec_revb (vector bool int);
24895         vector bool long long vec_revb (vector bool long long);
24896         * doc/extend.texi: Update the built-in documentation file for the
24897         new built-in functions.
24899 2017-07-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24901         * config/s390/s390.md: Remove movcc splitter.
24903 2017-07-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24905         * config/s390/s390.c (s390_rtx_costs): Return proper costs for
24906         load/store on condition.
24908 2017-07-12  Georg-Johann Lay  <avr@gjlay.de>
24910         PR target/81407
24911         * config/avr/avr.c (avr_encode_section_info)
24912         [progmem && !TREE_READONLY]: Error if progmem object needs
24913         constructing.
24915 2017-07-11  Michael Collison  <michael.collison@arm.com>
24917         * config/aarch64/aarch64-simd.md (aarch64_sub<mode>_compare0):
24918         New pattern.
24920 2017-07-11  Carl Love  <cel@us.ibm.com>
24922         * config/rs6000/rs6000-c.c: Add support for builtins
24923         vector unsigned int vec_parity_lsbb (vector signed int);
24924         vector unsigned int vec_parity_lsbb (vector unsigned int);
24925         vector unsigned __int128 vec_parity_lsbb (vector signed __int128);
24926         vector unsigned __int128 vec_parity_lsbb (vector unsigned __int128);
24927         vector unsigned long long vec_parity_lsbb (vector signed long long);
24928         vector unsigned long long vec_parity_lsbb (vector unsigned long long);
24929         * config/rs6000/rs6000-builtin.def (VPARITY_LSBB): Add BU_P9V_OVERLOAD1.
24930         * config/rs6000/altivec.h (vec_parity_lsbb): Add define.
24931         * doc/extend.texi: Update the built-in documentation file for the
24932         new built-in functions.
24934 2017-07-11  David Malcolm  <dmalcolm@redhat.com>
24936         * diagnostic-show-locus.c: Include "gcc-rich-location.h".
24937         (layout::m_primary_loc): New field.
24938         (layout::layout): Initialize new field.  Move location filtering
24939         logic from here to...
24940         (layout::maybe_add_location_range): ...this new method.  Add
24941         support for filtering to just the lines already specified by other
24942         locations.
24943         (layout::will_show_line_p): New method.
24944         (gcc_rich_location::add_location_if_nearby): New method.
24945         (selftest::test_add_location_if_nearby): New test function.
24946         (selftest::diagnostic_show_locus_c_tests): Call it.
24947         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
24948         New method.
24950 2017-07-11  Tom de Vries  <tom@codesourcery.com>
24952         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG): New macro.
24953         (bb_first_real_insn): New function.
24954         (nvptx_single): Add extra initialization of broadcasted condition
24955         variables.
24957 2017-07-11  Nathan Sidwell  <nathan@acm.org>
24959         * dwarf2out.c (gen_member_die): Remove useless check for anon ctors.
24961 2017-07-11  Georg-Johann Lay  <avr@gjlay.de>
24963         * doc/extend.texi (AVR Function Attributes): Remove weblink to
24964         Binutils doc as TEXI will mess them up.
24965         * doc/invoke.texi (AVR Options): Same here.
24967 2017-07-11  Daniel Cederman  <cederman@gaisler.com>
24969         * config/sparc/sparc.opt (mfix-ut700): New option.
24970         (mfix-gr712rc): Likewise.
24971         (sparc_fix_b2bst): New variable.
24972         * doc/invoke.texi (SPARC options): Document them.
24973         (ARM options): Fix warnings.
24974         * config/sparc/sparc.c (sparc_do_work_around_errata): Insert NOP
24975         instructions to prevent sequences that can trigger the store-store
24976         errata for certain LEON3FT processors.
24977         (pass_work_around_errata::gate): Also test sparc_fix_b2bst.
24978         (sparc_option_override): Set sparc_fix_b2bst appropriately.
24979         * config/sparc/sparc.md (fix_b2bst): New attribute.
24980         (in_branch_delay): Prevent stores in delay slot if fix_b2bst.
24982 2017-07-10  Uros Bizjak  <ubizjak@gmail.com>
24984         PR target/81375
24985         * config/i386/i386.md (divsf3): Add TARGET_SSE to TARGET_SSE_MATH.
24986         (rcpps): Ditto.
24987         (*rsqrtsf2_sse): Ditto.
24988         (rsqrtsf2): Ditto.
24989         (div<mode>3): Macroize insn from divdf3 and divsf3
24990         using MODEF mode iterator.
24992 2017-07-10  Martin Sebor  <msebor@redhat.com>
24994         PR tree-optimization/80397
24995         * gimple-ssa-sprintf.c (format_integer): Use INTEGRAL_TYPE_P()
24996         instead of testing for equality to INTEGER_TYPE.
24998 2017-07-10  Vineet Gupta <vgupta@synopsys.com>
25000         * config.gcc: Remove uclibc from arc target spec.
25002 2017-07-10  Claudiu Zissulescu  <claziss@synopsys.com>
25004         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Define.
25006 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
25008         PR lto/80838
25009         * lto-wrapper.c (remove_option): New function.
25010         (merge_and_complain): Merge PIC/PIE options more realistically.
25012 2017-07-10  Georg-Johann Lay  <avr@gjlay.de>
25014         Better ISR prologues by supporting GASes __gcc_isr pseudo insn.
25016         PR target/20296
25017         PR target/81268
25018         * configure.ac [target=avr]: Add GAS check for -mgcc-isr.
25019         (HAVE_AS_AVR_MGCCISR_OPTION):  If so, AC_DEFINE it.
25020         * config.in: Regenerate.
25021         * configure: Regenerate.
25022         * doc/extend.texi (AVR Function Attributes) <no_gccisr>: Document it.
25023         * doc/invoke.texi (AVR Options) <-mgas-isr-prologues>: Document it.
25024         * config/avr/avr.opt (-mgas-isr-prologues): New option and...
25025         (TARGET_GASISR_PROLOGUES): ...target mask.
25026         * common/config/avr/avr-common.c
25027         (avr_option_optimization_table) [OPT_LEVELS_1_PLUS_NOT_DEBUG]:
25028         Set -mgas-isr-prologues.
25029         * config/avr/avr-passes.def (avr_pass_pre_proep): Add
25030         INSERT_PASS_BEFORE for it.
25031         * config/avr/avr-protos.h (make_avr_pass_pre_proep): New proto.
25032         * config/avr/avr.c (avr_option_override)
25033         [!HAVE_AS_AVR_MGCCISR_OPTION]: Unset TARGET_GASISR_PROLOGUES.
25034         (avr_no_gccisr_function_p, avr_hregs_split_reg): New static functions.
25035         (avr_attribute_table) <no_gccisr>: Add new function attribute.
25036         (avr_set_current_function) <is_no_gccisr>: Init machine field.
25037         (avr_pass_data_pre_proep, avr_pass_pre_proep): New pass data
25038         and rtl_opt_pass.
25039         (make_avr_pass_pre_proep): New function.
25040         (emit_push_sfr) <treg>: Add argument to function and use it
25041         instead of TMP_REG.
25042         (avr_expand_prologue) [machine->gasisr.maybe]: Emit gasisr insn
25043         and set machine->gasisr.yes.
25044         (avr_expand_epilogue) [machine->gasisr.yes]: Similar.
25045         (avr_asm_function_end_prologue) [machine->gasisr.yes]: Add
25046         __gcc_isr.n_pushed to .L__stack_usage.
25047         (TARGET_ASM_FINAL_POSTSCAN_INSN): Define to...
25048         (avr_asm_final_postscan_insn): ...this new static function.
25049         * config/avr/avr.h (machine_function)
25050         <is_no_gccisr, use_L__stack_usage>: New fields.
25051         <gasisr, gasisr.yes, gasisr.maybe, gasisr.regno>: New fields.
25052         * config/avr/avr.md (UNSPECV_GASISR): Add unspecv enum.
25053         (GASISR_Prologue, GASISR_Epilogue, GASISR_Done): New define_constants.
25054         (gasisr, *gasisr): New expander and insn.
25055         * config/avr/gen-avr-mmcu-specs.c (print_mcu)
25056         [HAVE_AS_AVR_MGCCISR_OPTION]: Print asm_gccisr spec.
25057         * config/avr/specs.h (ASM_SPEC) <asm_gccisr>: Add sub spec.
25059 2017-07-10  Richard Earnshaw  <rearnsha@arm.com>
25061         * config/arm/parsecpu.awk (gen_comm_data): Do not escape single quotes
25062         in quoted strings.
25064 2017-07-10  Georg-Johann Lay  <avr@gjlay.de>
25066         Move jump-tables out of .text again.
25068         PR target/81075
25069         * config/avr/avr.c (ASM_OUTPUT_ADDR_VEC_ELT): Remove function.
25070         (ASM_OUTPUT_ADDR_VEC): New function.
25071         (avr_adjust_insn_length) [JUMP_TABLE_DATA_P]: Return 0.
25072         (avr_final_prescan_insn) [avr_log.insn_addresses]: Dump
25073         INSN_ADDRESSes as asm comment.
25074         * config/avr/avr.h (JUMP_TABLES_IN_TEXT_SECTION): Adjust comment.
25075         (ASM_OUTPUT_ADDR_VEC_ELT): Remove define.
25076         (ASM_OUTPUT_ADDR_VEC): Define to avr_output_addr_vec.
25077         * config/avr/avr.md (*tablejump): Adjust comment.
25078         * config/avr/elf.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove.
25079         * config/avr/avr-log.c (avr_log_set_avr_log) <insn_addresses>:
25080         New detail.
25081         * config/avr/avr-protos.h (avr_output_addr_vec_elt): Remove proto.
25082         (avr_output_addr_vec): New proto.
25083         (avr_log_t) <insn_addresses>: New field.
25085 2017-07-09  H.J. Lu  <hongjiu.lu@intel.com>
25087         PR target/81313
25088         * config/i386/i386.c (ix86_function_arg_advance): Set
25089         outgoing_args_on_stack to true if there are outgoing arguments
25090         on stack.
25091         (ix86_function_arg): Likewise.
25092         (ix86_get_drap_rtx): Use DRAP only if there are outgoing
25093         arguments on stack and ACCUMULATE_OUTGOING_ARGS is false.
25094         * config/i386/i386.h (machine_function): Add
25095         outgoing_args_on_stack.
25097 2017-07-09  Krister Walfridsson  <krister.walfridsson@gmail.com>
25099         * config.gcc (*-*-netbsd*): Remove check for NetBSD versions not
25100         supporting pthreds.
25101         * config/netbsd.h (NETBSD_LIBGCC_SPEC): Always enable pthreads.
25103 2017-07-08  Richard Sandiford  <richard.sandiford@linaro.org>
25105         * Makefile.in (HOOKS_H, RTL_BASE_H, FUNCTION_H, EXPR_H, REGS_H)
25106         (REAL_H): Remove $(MACHMODE_H).
25107         (FIXED_VALUE_H, TREE_CORE_H, CFGLOOP_H): Remove $(MACHMODE_H) and
25108         double-int.h.
25109         (CORETYPES_H): Add signop.h, wide-int.h, wide-int-print.h,
25110         $(MACHMODE_H) and double-int.h.
25111         (build/min-insn-modes.o): Depend on $(CORETYPES_H) rather than
25112         $(MACHMODE_H).
25113         (gengtype-state.o, gengtype.o, build/gengtype.o): Don't depend on
25114         double-int.h.
25116 2017-07-07  Andrew Pinski  <apinski@cavium.com>
25118         * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Check
25119         prev_set and curr_set for AARCH64_FUSE_ALU_BRANCH.
25121 2017-07-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
25123         * config/rs6000/rs6000.c (rs6000_get_function_versions_dispatcher):
25124         Add warning if GCC was not configured to link against a GLIBC that
25125         exports the hardware capability bits.
25126         (make_resolver_func): Make resolver function private and not a
25127         COMDAT function.  Create the name with clone_function_name instead
25128         of make_unique_name.
25130         PR target/81348
25131         * config/rs6000/rs6000.md (HI sign_extend splitter): Use the
25132         correct operand in doing the split.
25134 2017-07-07 Carl Love  <cel@us.ibm.com>
25136         * config/rs6000/rs6000-c: Add support for built-in function
25137         vector unsigned short vec_pack_to_short_fp32 (vector float,
25138                                                       vector float).
25139         * config/rs6000/rs6000-builtin.def (CONVERT_4F32_8I16): Add
25140         BU_P9V_AV_2 and BU_P9V_OVERLOAD_2 definitions.
25141         * config/rs6000/altivec.h (vec_pack_to_short_fp32): Add define.
25142         * config/rs6000/altivec.md(UNSPEC_CONVERT_4F32_8I16): Add UNSPEC.
25143         (convert_4f32_8i16): Add define_expand.
25144         * doc/extend.texi: Update the built-in documentation file for the
25145         new built-in function.
25147 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25149         * config/sparc/m8.md: New file.
25150         * config/sparc/sparc.md: Include m8.md.
25152 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25154         * config/sparc/sparc.opt: New option -mvis4b.
25155         * config/sparc/sparc.c (dump_target_flag_bits): Handle MASK_VIS4B.
25156         (sparc_option_override): Handle VIS4B.
25157         (enum sparc_builtins): Define
25158         SPARC_BUILTIN_DICTUNPACK{8,16,32},
25159         SPARC_BUILTIN_FPCMP{LE,GT,EQ,NE}{8,16,32}SHL,
25160         SPARC_BUILTIN_FPCMPU{LE,GT}{8,16,32}SHL,
25161         SPARC_BUILTIN_FPCMPDE{8,16,32}SHL and
25162         SPARC_BUILTIN_FPCMPUR{8,16,32}SHL.
25163         (check_constant_argument): New function.
25164         (sparc_vis_init_builtins): Define builtins
25165         __builtin_vis_dictunpack{8,16,32},
25166         __builtin_vis_fpcmp{le,gt,eq,ne}{8,16,32}shl,
25167         __builtin_vis_fpcmpu{le,gt}{8,16,32}shl,
25168         __builtin_vis_fpcmpde{8,16,32}shl and
25169         __builtin_vis_fpcmpur{8,16,32}shl.
25170         (sparc_expand_builtin): Check that the constant operands to
25171         __builtin_vis_fpcmp*shl and _builtin_vis_dictunpack* are indeed
25172         constant and in range.
25173         * config/sparc/sparc-c.c (sparc_target_macros): Handle
25174         TARGET_VIS4B.
25175         * config/sparc/sparc.h (SPARC_IMM2_P): Define.
25176         (SPARC_IMM5_P): Likewise.
25177         * config/sparc/sparc.md (cpu_feature): Add new feagure "vis4b".
25178         (enabled): Handle vis4b.
25179         (UNSPEC_DICTUNPACK): New unspec.
25180         (UNSPEC_FPCMPSHL): Likewise.
25181         (UNSPEC_FPUCMPSHL): Likewise.
25182         (UNSPEC_FPCMPDESHL): Likewise.
25183         (UNSPEC_FPCMPURSHL): Likewise.
25184         (cpu_feature): New CPU feature `vis4b'.
25185         (dictunpack{8,16,32}): New insns.
25186         (FPCSMODE): New mode iterator.
25187         (fpcscond): New code iterator.
25188         (fpcsucond): Likewise.
25189         (fpcmp{le,gt,eq,ne}{8,16,32}{si,di}shl): New insns.
25190         (fpcmpu{le,gt}{8,16,32}{si,di}shl): Likewise.
25191         (fpcmpde{8,16,32}{si,di}shl): Likewise.
25192         (fpcmpur{8,16,32}{si,di}shl): Likewise.
25193         * config/sparc/constraints.md: Define constraints `q' for unsigned
25194         2-bit integer constants and `t' for unsigned 5-bit integer
25195         constants.
25196         * config/sparc/predicates.md (imm5_operand_dictunpack8): New
25197         predicate.
25198         (imm5_operand_dictunpack16): Likewise.
25199         (imm5_operand_dictunpack32): Likewise.
25200         (imm2_operand): Likewise.
25201         * doc/invoke.texi (SPARC Options): Document -mvis4b.
25202         * doc/extend.texi (SPARC VIS Built-in Functions): Document the
25203         ditunpack* and fpcmp*shl builtins.
25205 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25207         * config.gcc: Handle m8 in --with-{cpu,tune} options.
25208         * config.in: Add HAVE_AS_SPARC6 define.
25209         * config/sparc/driver-sparc.c (cpu_names): Add entry for the SPARC
25210         M8.
25211         * config/sparc/sol2.h (CPP_CPU64_DEFAULT_SPEC): Define for
25212         TARGET_CPU_m8.
25213         (ASM_CPU32_DEFAUILT_SPEC): Likewise.
25214         (CPP_CPU_SPEC): Handle m8.
25215         (ASM_CPU_SPEC): Likewise.
25216         * config/sparc/sparc-opts.h (enum processor_type): Add
25217         PROCESSOR_M8.
25218         * config/sparc/sparc.c (m8_costs): New struct.
25219         (sparc_option_override): Handle TARGET_CPU_m8.
25220         (sparc32_initialize_trampoline): Likewise.
25221         (sparc64_initialize_trampoline): Likewise.
25222         (sparc_issue_rate): Likewise.
25223         (sparc_register_move_cost): Likewise.
25224         * config/sparc/sparc.h (TARGET_CPU_m8): Define.
25225         (CPP_CPU64_DEFAULT_SPEC): Define for M8.
25226         (ASM_CPU64_DEFAULT_SPEC): Likewise.
25227         (CPP_CPU_SPEC): Handle M8.
25228         (ASM_CPU_SPEC): Likewise.
25229         (AS_M8_FLAG): Define.
25230         * config/sparc/sparc.md: Add m8 to the cpu attribute.
25231         * config/sparc/sparc.opt: New option -mcpu=m8 for sparc targets.
25232         * configure.ac (HAVE_AS_SPARC6): Check for assembler support for
25233         M8 instructions.
25234         * configure: Regenerate.
25235         * doc/invoke.texi (SPARC Options): Document -mcpu=m8 and
25236         -mtune=m8.
25238 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25240         * config/sparc/niagara7.md: Rework the DFA scheduler to use insn
25241         subtypes.
25242         * config/sparc/sparc.md: Remove the `v3pipe' insn attribute.
25243         ("*movdi_insn_sp32"): Do not set v3pipe.
25244         ("*movsi_insn"): Likewise.
25245         ("*movdi_insn_sp64"): Likewise.
25246         ("*movsf_insn"): Likewise.
25247         ("*movdf_insn_sp32"): Likewise.
25248         ("*movdf_insn_sp64"): Likewise.
25249         ("*zero_extendsidi2_insn_sp64"): Likewise.
25250         ("*sign_extendsidi2_insn"): Likewise.
25251         ("*mov<VM32:mode>_insn"): Likewise.
25252         ("*mov<VM64:mode>_insn_sp64"): Likewise.
25253         ("*mov<VM64:mode>_insn_sp32"): Likewise.
25254         ("<plusminus_insn><VADDSUB:mode>3"): Likewise.
25255         ("<vlop:code><VL:mode>3"): Likewise.
25256         ("*not_<vlop:code><VL:mode>3"): Likewise.
25257         ("*nand<VL:mode>_vis"): Likewise.
25258         ("*<vlnotop:code>_not1<VL:mode>_vis"): Likewise.
25259         ("*<vlnotop:code>_not2<VL:mode>_vis"): Likewise.
25260         ("one_cmpl<VL:mode>2"): Likewise.
25261         ("faligndata<VM64:mode>_vis"): Likewise.
25262         ("alignaddrsi_vis"): Likewise.
25263         ("alignaddrdi_vis"): Likweise.
25264         ("alignaddrlsi_vis"): Likewise.
25265         ("alignaddrldi_vis"): Likewise.
25266         ("fcmp<gcond:code><GCM:gcm_name><P:mode>_vis"): Likewise.
25267         ("bmaskdi_vis"): Likewise.
25268         ("bmasksi_vis"): Likewise.
25269         ("bshuffle<VM64:mode>_vis"): Likewise.
25270         ("cmask8<P:mode>_vis"): Likewise.
25271         ("cmask16<P:mode>_vis"): Likewise.
25272         ("cmask32<P:mode>_vis"): Likewise.
25273         ("pdistn<P:mode>_vis"): Likewise.
25274         ("<vis3_addsub_ss_patname><VASS:mode>3"): Likewise.
25276 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25278         * config/sparc/sparc.md ("subtype"): New insn attribute.
25279         ("*wrgsr_sp64"): Set insn subtype.
25280         ("*rdgsr_sp64"): Likewise.
25281         ("alignaddrsi_vis"): Likewise.
25282         ("alignaddrdi_vis"): Likewise.
25283         ("alignaddrlsi_vis"): Likewise.
25284         ("alignaddrldi_vis"): Likewise.
25285         ("<plusminus_insn><VADDSUB:mode>3"): Likewise.
25286         ("fexpand_vis"): Likewise.
25287         ("fpmerge_vis"): Likewise.
25288         ("faligndata<VM64:mode>_vis"): Likewise.
25289         ("bshuffle<VM64:mode>_vis"): Likewise.
25290         ("cmask8<P:mode>_vis"): Likewise.
25291         ("cmask16<P:mode>_vis"): Likewise.
25292         ("cmask32<P:mode>_vis"): Likewise.
25293         ("fchksm16_vis"): Likewise.
25294         ("v<vis3_shift_patname><GCM:mode>3"): Likewise.
25295         ("fmean16_vis"): Likewise.
25296         ("fp<plusminus_insn>64_vis"): Likewise.
25297         ("<plusminus_insn>v8qi3"): Likewise.
25298         ("<vis3_addsub_ss_patname><VASS:mode>3"): Likewise.
25299         ("<vis4_minmax_patname><VMMAX:mode>3"): Likewise.
25300         ("<vis4_uminmax_patname><VMMAX:mode>3"): Likewise.
25301         ("<vis3_addsub_ss_patname>v8qi3"): Likewise.
25302         ("<vis4_addsub_us_patname><VAUS:mode>3"): Likewise.
25303         ("*movqi_insn"): Likewise.
25304         ("*movhi_insn"): Likewise.
25305         ("*movsi_insn"): Likewise.
25306         ("movsi_pic_gotdata_op"): Likewise.
25307         ("*movdi_insn_sp32"): Likewise.
25308         ("*movdi_insn_sp64"): Likewise.
25309         ("movdi_pic_gotdata_op"): Likewise.
25310         ("*movsf_insn"): Likewise.
25311         ("*movdf_insn_sp32"): Likewise.
25312         ("*movdf_insn_sp64"): Likewise.
25313         ("*zero_extendhisi2_insn"): Likewise.
25314         ("*zero_extendqihi2_insn"): Likewise.
25315         ("*zero_extendqisi2_insn"): Likewise.
25316         ("*zero_extendqidi2_insn"): Likewise.
25317         ("*zero_extendhidi2_insn"): Likewise.
25318         ("*zero_extendsidi2_insn_sp64"): Likewise.
25319         ("ldfsr"): Likewise.
25320         ("prefetch_64"): Likewise.
25321         ("prefetch_32"): Likewise.
25322         ("tie_ld32"): Likewise.
25323         ("tie_ld64"): Likewise.
25324         ("*tldo_ldub_sp32"): Likewise.
25325         ("*tldo_ldub1_sp32"): Likewise.
25326         ("*tldo_ldub2_sp32"): Likewise.
25327         ("*tldo_ldub_sp64"): Likewise.
25328         ("*tldo_ldub1_sp64"): Likewise.
25329         ("*tldo_ldub2_sp64"): Likewise.
25330         ("*tldo_ldub3_sp64"): Likewise.
25331         ("*tldo_lduh_sp32"): Likewise.
25332         ("*tldo_lduh1_sp32"): Likewise.
25333         ("*tldo_lduh_sp64"): Likewise.
25334         ("*tldo_lduh1_sp64"): Likewise.
25335         ("*tldo_lduh2_sp64"): Likewise.
25336         ("*tldo_lduw_sp32"): Likewise.
25337         ("*tldo_lduw_sp64"): Likewise.
25338         ("*tldo_lduw1_sp64"): Likewise.
25339         ("*tldo_ldx_sp64"): Likewise.
25340         ("*mov<VM32:mode>_insn"): Likewise.
25341         ("*mov<VM64:mode>_insn_sp64"): Likewise.
25342         ("*mov<VM64:mode>_insn_sp32"): Likewise.
25344 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25346         * config/sparc/sparc.md ("type"): New insn type viscmp.
25347         ("fcmp<gcond:code><GCM:gcm_name><P:mode>_vis"): Set insn type to
25348         viscmp.
25349         ("fpcmp<gcond:code>8<P:mode>_vis"): Likewise.
25350         ("fucmp<gcond:code>8<P:mode>_vis"): Likewise.
25351         ("fpcmpu<gcond:code><GCM:gcm_name><P:mode>_vis"): Likewise.
25352         * config/sparc/niagara7.md ("n7_vis_logical_v3pipe"): Handle
25353         viscmp.
25354         ("n7_vis_logical_11cycle"): Likewise.
25355         * config/sparc/niagara4.md ("n4_vis_logical"): Likewise.
25356         * config/sparc/niagara2.md ("niag3_vis": Likewise.
25357         * config/sparc/niagara.md ("niag_vis"): Likewise.
25358         * config/sparc/ultra3.md ("us3_fga"): Likewise.
25359         * config/sparc/ultra1_2.md ("us1_fga_double"): Likewise.
25361 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25363         * config/sparc/sparc.md: New instruction type `bmask'.
25364         (bmaskdi_vis): Use the `bmask' type.
25365         (bmasksi_vis): Likewise.
25366         * config/sparc/ultra3.md (us3_array): Likewise.
25367         * config/sparc/niagara7.md (n7_array): Likewise.
25368         * config/sparc/niagara4.md (n4_array): Likewise.
25369         * config/sparc/niagara2.md (niag2_vis): Likewise.
25370         (niag3_vis): Likewise.
25371         * config/sparc/niagara.md (niag_vis): Likewise.
25373 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25375         * ipa-comdats.c: Remove optimize check from gate.
25376         * ipa-fnsummary.c (ipa_fn_summary_generate): do not generate summary
25377         for functions not optimized.
25378         (ipa_fn_summary_read): Skip optimize check.
25379         (ipa_fn_summary_write): Likewise.
25380         * ipa-inline-analysis.c (do_estimate_growth_1): Check that caller
25381         is optimized.
25382         * ipa-inline.c (can_inline_edge_p): Not optimized functions are
25383         uninlinable.
25384         (can_inline_edge_p): Check flag_pcc_struct_return for match.
25385         (check_callers): Give up on caller which is not optimized.
25386         (inline_small_functions): Likewise.
25387         (ipa_inline): Do not give up when not optimizing.
25388         * ipa-visbility.c (function_and_variable_visibility): Do not optimize
25389         away unoptimizes cdtors.
25390         (whole_program_function_and_variable_visibility): Do
25391         ipa_discover_readonly_nonaddressable_vars in LTO mode.
25392         * ipa.c (process_references): Do not check optimize.
25393         (symbol_table::remove_unreachable_nodes): Update optimize check.
25394         (set_writeonly_bit): Update optimize check.
25395         (pass_ipa_cdtor_merge::gate): Do not check optimize.
25396         (pass_ipa_single_use::gate): Remove.
25398 2017-07-06  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
25400         * config/rs6000/rs6000.c (union_defs, union_uses, insn_is_load_p,
25401         insn_is_store_p, insn_is_swap_p, const_load_sequence_p, v2df_reduction_p,
25402         rtx_is_swappable_p, insn_is_swappable_p, chain_contains_only_swaps,
25403         mark_swaps_for_removal, swap_const_vector_halves, adjust_subreg_index,
25404         permute_load, permute_store, adjust_extract, adjust_splat,
25405         adjust_xxpermdi, adjust_concat, adjust_vperm, handle_special_swappables,
25406         replace_swap_with_copy, dump_swap_insn_table,
25407         alignment_with_canonical_addr, alignment_mask, find_alignment_op,
25408         recombine_lvx_pattern, recombine_stvx_pattern,
25409         recombine_lvx_stvx_patterns, rs6000_analyze_swaps,
25410         make_pass_analyze_swaps): Move all code related to p8 swap optimizations
25411         to file rs6000-p8swap.c.
25412         * config/rs6000/rs6000-p8swap.c: New file.
25413         * config/rs6000/t-rs6000: Add rule to build rs6000-p8swap.o.
25414         * config.gcc: Add rs6000-p8swap.o to extra_objs for powerpc*-*-*
25415         and rs6000*-*-* targets.
25417 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25419         * Makefile.in (selftest): Remove dependency on s-selftest-c++.
25421 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25423         * lto-wrapper.c (merge_and_complain): Do not merge
25424         fexceptions, fnon_call_exceptions, ftrapv, ffp_contract_, fmath_errno,
25425         fsigned_zeros, ftrapping_math, fwrapv.
25426         (append_compiler_options): Do not track these options.
25427         (append_linker_options): Likewie
25429 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25431         * cgraphunit.c (cgraph_node::finalize_function): When
25432         !flag_toplevel_reorde set no_reorder flag.
25433         (varpool_node::finalize_decl): Likewise.
25434         (symbol_table::compile): Drop no toplevel reorder path.
25436 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25438         * bb-reorder.c (better_edge_p): Do not build traces across abnormal/eh
25439         edges; zero probability is not better than uninitialized.
25441 2017-07-06  Maxim Ostapenko  <m.ostapenko@samsung.com>
25443         * asan.h (asan_sanitize_allocas_p): Declare.
25444         * asan.c (asan_sanitize_allocas_p): New function.
25445         (handle_builtin_stack_restore): Bail out if !asan_sanitize_allocas_p.
25446         (handle_builtin_alloca): Likewise.
25447         * cfgexpand.c (expand_used_vars): Do not add allocas unpoisoning stuff
25448         if !asan_sanitize_allocas_p.
25449         * params.def (asan-instrument-allocas): Add new option.
25450         * params.h (ASAN_PROTECT_ALLOCAS): Define.
25451         * opts.c (common_handle_option): Disable allocas sanitization for
25452         KASan by default.
25454 2017-07-06  Maxim Ostapenko  <m.ostapenko@samsung.com>
25456         * asan.c: Include gimple-fold.h.
25457         (get_last_alloca_addr): New function.
25458         (handle_builtin_stackrestore): Likewise.
25459         (handle_builtin_alloca): Likewise.
25460         (asan_emit_allocas_unpoison): Likewise.
25461         (get_mem_refs_of_builtin_call): Add new parameter, remove const
25462         quallifier from first paramerer. Handle BUILT_IN_ALLOCA,
25463         BUILT_IN_ALLOCA_WITH_ALIGN and BUILT_IN_STACK_RESTORE builtins.
25464         (instrument_builtin_call): Pass gimple iterator to
25465         get_mem_refs_of_builtin_call.
25466         (last_alloca_addr): New global.
25467         * asan.h (asan_emit_allocas_unpoison): Declare.
25468         * builtins.c (expand_asan_emit_allocas_unpoison): New function.
25469         (expand_builtin): Handle BUILT_IN_ASAN_ALLOCAS_UNPOISON.
25470         * cfgexpand.c (expand_used_vars): Call asan_emit_allocas_unpoison
25471         if function calls alloca.
25472         * gimple-fold.c (replace_call_with_value): Remove static keyword.
25473         * gimple-fold.h (replace_call_with_value): Declare.
25474         * internal-fn.c: Include asan.h.
25475         * sanitizer.def (BUILT_IN_ASAN_ALLOCA_POISON,
25476         BUILT_IN_ASAN_ALLOCAS_UNPOISON): New builtins.
25478 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25480         * Makefile.in (SELFTEST_FLAGS): Drop "-x c", moving it to...
25481         (C_SELFTEST_FLAGS): New.
25482         (CPP_SELFTEST_FLAGS): New.
25483         (SELFTEST_DEPS): New, from deps of s-selftest.
25484         (C_SELFTEST_DEPS): New, from deps of s-selftest.
25485         (CPP_SELFTEST_DEPS): New.
25486         (selftest): Add dependency on s-selftest-c++.
25487         (s-selftest): Rename to...
25488         (s-selftest-c): ...this, moving deps to SELFTEST_DEPS
25489         and C_SELFTEST_DEPS, and using C_SELFTEST_FLAGS rather
25490         than SELFTEST_FLAGS.
25491         (selftest-gdb): Rename to...
25492         (selftest-c-gdb): ...this, using C_SELFTEST_DEPS and
25493         C_SELFTEST_FLAGS.
25494         (selftest-gdb): Reintroduce as an alias for selftest-c-gdb.
25495         (selftest-valgrind): Rename to...
25496         (selftest-c-valgrind): ...this, using C_SELFTEST_DEPS and
25497         C_SELFTEST_FLAGS.
25498         (selftest-valgrind): Reintroduce as an alias for
25499         selftest-c-valgrind.
25500         (s-selftest-c++): New.
25501         (selftest-c++-gdb): New.
25502         (selftest-c++-valgrind): New.
25504 2017-07-06  Olivier Hainque  <hainque@adacore.com>
25506         * gcc.c (process_command): When deciding if undefined variables
25507         should be ignored when processing specs, accept "gcc -v" as well.
25509 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25511         * auto-profile.c (afdo_set_bb_count, afdo_propagate_edge,
25512         afdo_annotate_cfg): Set counts/probabilities as determined by afdo.
25514 2017-07-06  Thomas Preud'homme  <thomas.preudhomme@arm.com>
25516         * config/arm/arm-cpus.in (armv8-r): Add new entry.
25517         * config/arm/arm-isa.h (ISA_ARMv8r): Define macro.
25518         * config/arm/arm-tables.opt: Regenerate.
25519         * config/arm/arm.h (enum base_architecture): Add BASE_ARCH_8R
25520         enumerator.
25521         * doc/invoke.texi: Mention -march=armv8-r and its extensions.
25523 2017-07-06  Carl Love  <cel@us.ibm.com>
25525         * ChangeLog: Clean up from mid air collision
25527 2017-07-06  Carl Love  <cel@us.ibm.com>
25529         * config/rs6000/rs6000-c.c: Add support for built-in functions
25530         vector signed int vec_subc (vector signed int, vector signed int);
25531         vector signed __int128 vec_subc (vector signed __int128,
25532                                          vector signed __int128);
25533         vector unsigned __int128 vec_subc (vector unsigned __int128,
25534                                            vector unsigned __int128);
25535         vector signed int vec_sube (vector signed int, vector signed int,
25536                                     vector signed int);
25537         vector unsigned int vec_sube (vector unsigned int,
25538                                       vector unsigned int,
25539                                       vector unsigned int);
25540         vector signed __int128 vec_sube (vector signed __int128,
25541                                          vector signed __int128,
25542                                          vector signed__int128);
25543         vector unsigned __int128 vec_sube (vector unsigned __int128,
25544                                            vector unsigned __int128,
25545                                            vector unsigned __int128);
25546         vector signed int vec_subec (vector signed int, vector signed int,
25547                                      vector signed int);
25548         vector unsigned int vec_subec (vector unsigned int,
25549                                        vector unsigned int,
25550                                        vector unsigned int);
25551         vector signed __int128 vec_subec (vector signed __int128,
25552                                           vector signed __int128,
25553                                           vector signed__int128);
25554         vector unsigned __int128 vec_subec (vector unsigned __int128,
25555                                             vector unsigned __int128,
25556                                             vector unsigned __int128);
25557         * config/rs6000/rs6000.c (ALTIVEC_BUILTIN_VEC_SUBE,
25558         ALTIVEC_BUILTIN_VEC_SUBEC): Add ef_builtins.
25559         * config/rs6000/rs6000-builtin.def (SUBE, SUBEC): Add
25560         BU_ALTIVEC_OVERLOAD_X definitions.
25561         * config/rs6000/altivec.h (vec_sube, vec_subec): Add builtin defines.
25562         * doc/extend.texi: Update the built-in documentation file for the new
25563         built-in functions.
25565 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25567         PR c++/79300
25568         * diagnostic-show-locus.c (layout::layout): Use start and finish
25569         spelling location for the start and finish of each range.
25570         * genmatch.c (linemap_client_expand_location_to_spelling_point):
25571         Add unused aspect param.
25572         * input.c (expand_location_1): Add "aspect" param, and use it
25573         to access the correct part of the location.
25574         (expand_location): Pass LOCATION_ASPECT_CARET to new param of
25575         expand_location_1.
25576         (expand_location_to_spelling_point): Likewise.
25577         (linemap_client_expand_location_to_spelling_point): Add "aspect"
25578         param, and pass it to expand_location_1.
25580 2017-07-06  Sebastian Peryt  <sebastian.peryt@intel.com>
25582         * config/i386/avx512fintrin.h (_mm_mask_getexp_round_ss,
25583         _mm_maskz_getexp_round_ss, _mm_mask_getexp_round_sd,
25584         _mm_maskz_getexp_round_sd, _mm_mask_getmant_round_sd,
25585         _mm_maskz_getmant_round_sd, _mm_mask_getmant_round_ss,
25586         _mm_maskz_getmant_round_ss, _mm_mask_getexp_ss, _mm_maskz_getexp_ss,
25587         _mm_mask_getexp_sd, _mm_maskz_getexp_sd, _mm_mask_getmant_sd,
25588         _mm_maskz_getmant_sd, _mm_mask_getmant_ss,
25589         _mm_maskz_getmant_ss): New intrinsics.
25590         (__builtin_ia32_getexpss128_mask): Changed to ...
25591         __builtin_ia32_getexpss128_round ... this.
25592         (__builtin_ia32_getexpsd128_mask): Changed to ...
25593         __builtin_ia32_getexpsd128_round ... this.
25594         * config/i386/i386-builtin-types.def
25595         ((V2DF, V2DF, V2DF, INT, V2DF, UQI, INT),
25596         (V4SF, V4SF, V4SF, INT, V4SF, UQI, INT)): New function type aliases.
25597         * config/i386/i386-builtin.def (__builtin_ia32_getexpsd_mask_round,
25598         __builtin_ia32_getexpss_mask_round,     __builtin_ia32_getmantsd_mask_round,
25599         __builtin_ia32_getmantss_mask_round): New builtins.
25600         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_INT_V2DF_UQI_INT,
25601         V4SF_FTYPE_V4SF_V4SF_INT_V4SF_UQI_INT): Handle new types.
25602         (CODE_FOR_avx512f_vgetmantv2df_mask_round,
25603         CODE_FOR_avx512f_vgetmantv4sf_mask_round): New cases.
25604         * config/i386/sse.md
25605         (avx512f_sgetexp<mode><round_saeonly_name>): Changed to ...
25606         avx512f_sgetexp<mode><mask_scalar_name>
25607         <round_saeonly_scalar_name> ... this.
25608         (vgetexp<ssescalarmodesuffix>\t{<round_saeonly_op3>%2, %1, %0|
25609         %0, %1, %2<round_saeonly_op3>}): Changed to ...
25610         vgetexp<ssescalarmodesuffix>
25611         \t{<round_saeonly_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25612         %0<mask_scalar_operand3>, %1, %2<round_saeonly_scalar_mask_op3>} ... this.
25613         (avx512f_vgetmant<mode><round_saeonly_name>): Changed to ...
25614         avx512f_vgetmant<mode><mask_scalar_name>
25615         <round_saeonly_scalar_name> ... this.
25616         (vgetmant<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
25617         %0, %1, %2<round_saeonly_op4>, %3}): Changed to ...
25618         vgetmant<ssescalarmodesuffix>
25619         \t{%3, <round_saeonly_scalar_mask_op4>%2, %1, %0<mask_scalar_operand4>|
25620         %0<mask_scalar_operand4>, %1, %2
25621         <round_saeonly_scalar_mask_op4>, %3} ... this.
25622         * config/i386/subst.md (mask_scalar_operand4,
25623         round_saeonly_scalar_mask_operand4,     round_saeonly_scalar_mask_op4,
25624         round_saeonly_scalar_nimm_predicate): New subst attributes.
25626 2017-07-06  Julia Koval  <julia.koval@intel.com>
25628         * config/i386/i386.c (ix86_erase_embedded_rounding):
25629         Remove code for old rounding pattern.
25631 2017-07-06  Richard Earnshaw  <rearnsha@arm.com>
25633         * config/arm/t-arm (GTM_H): Add arm-cpu.h.
25635 2017-07-06  Christophe Lyon  <christophe.lyon@linaro.org>
25637         * doc/sourcebuild.texi (Test Directives, Variants of
25638         dg-require-support): Add documentation for dg-require-stack-check.
25640 2017-07-05  Sebastian Peryt  <sebastian.peryt@intel.com>
25642         * config/i386/subst.md (mask_scalar, round_scalar,
25643         round_saeonly_scalar): New meta-templates.
25644         (mask_scalar_name, mask_scalar_operand3, round_scalar_name,
25645         round_scalar_mask_operand3, round_scalar_mask_op3,
25646         round_scalar_constraint, round_scalar_prefix, round_saeonly_scalar_name,
25647         round_saeonly_scalar_mask_operand3, round_saeonly_scalar_mask_op3,
25648         round_saeonly_scalar_constraint,
25649         round_saeonly_scalar_prefix): New subst attribute.
25650         * config/i386/sse.md
25651         (<sse>_vm<plusminus_insn><mode>3<mask_name><round_name>): Renamed to ...
25652         <sse>_vm<plusminus_insn><mode>3<mask_scalar_name>
25653         <round_scalar_name> ... this.
25654         (<sse>_vm<multdiv_mnemonic><mode>3<mask_name><round_name>): Renamed to ...
25655         <sse>_vm<multdiv_mnemonic><mode>3<mask_scalar_name>
25656         <round_scalar_name> ... this.
25657         (<sse>_vm<code><mode>3<mask_name><round_saeonly_name>): Renamed to ...
25658         <sse>_vm<code><mode>3<mask_scalar_name>
25659         <round_saeonly_scalar_name> ... this.
25660         (v<plusminus_mnemonic><ssescalarmodesuffix>
25661         \t{<round_mask_op3>%2, %1, %0<mask_operand3>|
25662         %0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}): Changed to ...
25663         v<plusminus_mnemonic><ssescalarmodesuffix>
25664         \t{<round_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25665         %0<mask_scalar_operand3>, %1, %<iptr>2<round_scalar_mask_op3>} ... this.
25666         (v<multdiv_mnemonic><ssescalarmodesuffix>
25667         \t{<round_mask_op3>%2, %1, %0<mask_operand3>|
25668         %0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}): Changed to ...
25669         v<multdiv_mnemonic><ssescalarmodesuffix>
25670         \t{<round_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25671         %0<mask_scalar_operand3>, %1, %<iptr>2<round_scalar_mask_op3>} ... this.
25672         (v<maxmin_float><ssescalarmodesuffix>
25673         \t{<round_saeonly_mask_op3>%2, %1, %0<mask_operand3>|
25674         %0<mask_operand3>, %1, %<iptr>2<round_saeonly_mask_op3>}): Changed to ...
25675         v<maxmin_float><ssescalarmodesuffix>
25676         \t{<round_saeonly_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25677         %0<mask_scalar_operand3>, %1, %<iptr>2
25678         <round_saeonly_scalar_mask_op3>} ... this.
25680 2017-07-05  Richard Earnshaw  <rearnsha@arm.com>
25682         * config/arm/arm.c (arm_fixed_condition_code_regs): New function.
25683         (TARGET_FIXED_CONDITION_CODE_REGS): Redefine.
25685 2017-07-05  Richard Sandiford  <richard.sandiford@linaro.org>
25686             Alan Hayward  <alan.hayward@arm.com>
25687             David Sherwood  <david.sherwood@arm.com>
25689         * combine.c (simplify_if_then_else): Remove "enum" before
25690         "machine_mode".
25691         * compare-elim.c (can_eliminate_compare): Likewise.
25692         * config/aarch64/aarch64-builtins.c (aarch64_simd_builtin_std_type):
25693         Likewise.
25694         (aarch64_lookup_simd_builtin_type): Likewise.
25695         (aarch64_simd_builtin_type): Likewise.
25696         (aarch64_init_simd_builtin_types): Likewise.
25697         (aarch64_simd_expand_args): Likewise.
25698         * config/aarch64/aarch64-protos.h (aarch64_simd_attr_length_rglist):
25699         Likewise.
25700         (aarch64_reverse_mask): Likewise.
25701         (aarch64_simd_emit_reg_reg_move): Likewise.
25702         (aarch64_gen_adjusted_ldpstp): Likewise.
25703         (aarch64_ccmp_mode_to_code): Likewise.
25704         (aarch64_operands_ok_for_ldpstp): Likewise.
25705         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
25706         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
25707         Likewise.
25708         (aarch64_min_divisions_for_recip_mul): Likewise.
25709         (aarch64_reassociation_width): Likewise.
25710         (aarch64_get_condition_code_1): Likewise.
25711         (aarch64_simd_emit_reg_reg_move): Likewise.
25712         (aarch64_simd_attr_length_rglist): Likewise.
25713         (aarch64_reverse_mask): Likewise.
25714         (aarch64_operands_ok_for_ldpstp): Likewise.
25715         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
25716         (aarch64_gen_adjusted_ldpstp): Likewise.
25717         * config/aarch64/cortex-a57-fma-steering.c (fma_node::rename):
25718         Likewise.
25719         * config/arc/arc.c (legitimate_offset_address_p): Likewise.
25720         * config/arm/arm-builtins.c (arm_simd_builtin_std_type): Likewise.
25721         (arm_lookup_simd_builtin_type): Likewise.
25722         (arm_simd_builtin_type): Likewise.
25723         (arm_init_simd_builtin_types): Likewise.
25724         (arm_expand_builtin_args): Likewise.
25725         * config/arm/arm-protos.h (arm_expand_builtin): Likewise.
25726         * config/ft32/ft32.c (ft32_libcall_value): Likewise.
25727         (ft32_setup_incoming_varargs): Likewise.
25728         (ft32_function_arg): Likewise.
25729         (ft32_function_arg_advance): Likewise.
25730         (ft32_pass_by_reference): Likewise.
25731         (ft32_arg_partial_bytes): Likewise.
25732         (ft32_valid_pointer_mode): Likewise.
25733         (ft32_addr_space_pointer_mode): Likewise.
25734         (ft32_addr_space_legitimate_address_p): Likewise.
25735         * config/i386/i386-protos.h (ix86_operands_ok_for_move_multiple):
25736         Likewise.
25737         * config/i386/i386.c (ix86_setup_incoming_vararg_bounds): Likewise.
25738         (ix86_emit_outlined_ms2sysv_restore): Likewise.
25739         (iamcu_alignment): Likewise.
25740         (canonicalize_vector_int_perm): Likewise.
25741         (ix86_noce_conversion_profitable_p): Likewise.
25742         (ix86_mpx_bound_mode): Likewise.
25743         (ix86_operands_ok_for_move_multiple): Likewise.
25744         * config/microblaze/microblaze-protos.h
25745         (microblaze_expand_conditional_branch_reg): Likewise.
25746         * config/microblaze/microblaze.c
25747         (microblaze_expand_conditional_branch_reg): Likewise.
25748         * config/powerpcspe/powerpcspe.c (rs6000_init_hard_regno_mode_ok):
25749         Likewise.
25750         (rs6000_reassociation_width): Likewise.
25751         (rs6000_invalid_binary_op): Likewise.
25752         (fusion_p9_p): Likewise.
25753         (emit_fusion_p9_load): Likewise.
25754         (emit_fusion_p9_store): Likewise.
25755         * config/riscv/riscv-protos.h (riscv_regno_mode_ok_for_base_p):
25756         Likewise.
25757         (riscv_hard_regno_mode_ok_p): Likewise.
25758         (riscv_address_insns): Likewise.
25759         (riscv_split_symbol): Likewise.
25760         (riscv_legitimize_move): Likewise.
25761         (riscv_function_value): Likewise.
25762         (riscv_hard_regno_nregs): Likewise.
25763         (riscv_expand_builtin): Likewise.
25764         * config/riscv/riscv.c (riscv_build_integer_1): Likewise.
25765         (riscv_build_integer): Likewise.
25766         (riscv_split_integer): Likewise.
25767         (riscv_legitimate_constant_p): Likewise.
25768         (riscv_cannot_force_const_mem): Likewise.
25769         (riscv_regno_mode_ok_for_base_p): Likewise.
25770         (riscv_valid_base_register_p): Likewise.
25771         (riscv_valid_offset_p): Likewise.
25772         (riscv_valid_lo_sum_p): Likewise.
25773         (riscv_classify_address): Likewise.
25774         (riscv_legitimate_address_p): Likewise.
25775         (riscv_address_insns): Likewise.
25776         (riscv_load_store_insns): Likewise.
25777         (riscv_force_binary): Likewise.
25778         (riscv_split_symbol): Likewise.
25779         (riscv_force_address): Likewise.
25780         (riscv_legitimize_address): Likewise.
25781         (riscv_move_integer): Likewise.
25782         (riscv_legitimize_const_move): Likewise.
25783         (riscv_legitimize_move): Likewise.
25784         (riscv_address_cost): Likewise.
25785         (riscv_subword): Likewise.
25786         (riscv_output_move): Likewise.
25787         (riscv_canonicalize_int_order_test): Likewise.
25788         (riscv_emit_int_order_test): Likewise.
25789         (riscv_function_arg_boundary): Likewise.
25790         (riscv_pass_mode_in_fpr_p): Likewise.
25791         (riscv_pass_fpr_single): Likewise.
25792         (riscv_pass_fpr_pair): Likewise.
25793         (riscv_get_arg_info): Likewise.
25794         (riscv_function_arg): Likewise.
25795         (riscv_function_arg_advance): Likewise.
25796         (riscv_arg_partial_bytes): Likewise.
25797         (riscv_function_value): Likewise.
25798         (riscv_pass_by_reference): Likewise.
25799         (riscv_setup_incoming_varargs): Likewise.
25800         (riscv_print_operand): Likewise.
25801         (riscv_elf_select_rtx_section): Likewise.
25802         (riscv_save_restore_reg): Likewise.
25803         (riscv_for_each_saved_reg): Likewise.
25804         (riscv_register_move_cost): Likewise.
25805         (riscv_hard_regno_mode_ok_p): Likewise.
25806         (riscv_hard_regno_nregs): Likewise.
25807         (riscv_class_max_nregs): Likewise.
25808         (riscv_memory_move_cost): Likewise.
25809         * config/rl78/rl78-protos.h (rl78_split_movsi): Likewise.
25810         * config/rl78/rl78.c (rl78_split_movsi): Likewise.
25811         (rl78_addr_space_address_mode): Likewise.
25812         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
25813         Likewise.
25814         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
25815         (rs6000_reassociation_width): Likewise.
25816         (rs6000_invalid_binary_op): Likewise.
25817         (fusion_p9_p): Likewise.
25818         (emit_fusion_p9_load): Likewise.
25819         (emit_fusion_p9_store): Likewise.
25820         * config/visium/visium-protos.h (prepare_move_operands): Likewise.
25821         (ok_for_simple_move_operands): Likewise.
25822         (ok_for_simple_move_strict_operands): Likewise.
25823         (ok_for_simple_arith_logic_operands): Likewise.
25824         (visium_legitimize_reload_address): Likewise.
25825         (visium_select_cc_mode): Likewise.
25826         (output_cbranch): Likewise.
25827         (visium_split_double_move): Likewise.
25828         (visium_expand_copysign): Likewise.
25829         (visium_expand_int_cstore): Likewise.
25830         (visium_expand_fp_cstore): Likewise.
25831         * config/visium/visium.c (visium_pass_by_reference): Likewise.
25832         (visium_function_arg): Likewise.
25833         (visium_function_arg_advance): Likewise.
25834         (visium_libcall_value): Likewise.
25835         (visium_setup_incoming_varargs): Likewise.
25836         (visium_legitimate_constant_p): Likewise.
25837         (visium_legitimate_address_p): Likewise.
25838         (visium_legitimize_address): Likewise.
25839         (visium_secondary_reload): Likewise.
25840         (visium_register_move_cost): Likewise.
25841         (visium_memory_move_cost): Likewise.
25842         (prepare_move_operands): Likewise.
25843         (ok_for_simple_move_operands): Likewise.
25844         (ok_for_simple_move_strict_operands): Likewise.
25845         (ok_for_simple_arith_logic_operands): Likewise.
25846         (visium_function_value_1): Likewise.
25847         (rtx_ok_for_offset_p): Likewise.
25848         (visium_legitimize_reload_address): Likewise.
25849         (visium_split_double_move): Likewise.
25850         (visium_expand_copysign): Likewise.
25851         (visium_expand_int_cstore): Likewise.
25852         (visium_expand_fp_cstore): Likewise.
25853         (visium_split_cstore): Likewise.
25854         (visium_select_cc_mode): Likewise.
25855         (visium_split_cbranch): Likewise.
25856         (output_cbranch): Likewise.
25857         (visium_print_operand_address): Likewise.
25858         * expmed.c (flip_storage_order): Likewise.
25859         * expmed.h (emit_cstore): Likewise.
25860         (flip_storage_order): Likewise.
25861         * genrecog.c (validate_pattern): Likewise.
25862         * hsa-gen.c (gen_hsa_addr): Likewise.
25863         * internal-fn.c (expand_arith_overflow): Likewise.
25864         * ira-color.c (allocno_copy_cost_saving): Likewise.
25865         * lra-assigns.c (find_hard_regno_for_1): Likewise.
25866         * lra-constraints.c (prohibited_class_reg_set_mode_p): Likewise.
25867         (process_invariant_for_inheritance): Likewise.
25868         * lra-eliminations.c (move_plus_up): Likewise.
25869         * omp-low.c (lower_oacc_reductions): Likewise.
25870         * simplify-rtx.c (simplify_subreg): Likewise.
25871         * target.def (TARGET_SETUP_INCOMING_VARARG_BOUNDS): Likewise.
25872         (TARGET_CHKP_BOUND_MODE): Likewise..
25873         * targhooks.c (default_chkp_bound_mode): Likewise.
25874         (default_setup_incoming_vararg_bounds): Likewise.
25875         * targhooks.h (default_chkp_bound_mode): Likewise.
25876         (default_setup_incoming_vararg_bounds): Likewise.
25877         * tree-ssa-math-opts.c (divmod_candidate_p): Likewise.
25878         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
25879         (have_whole_vector_shift): Likewise.
25880         * tree-vect-stmts.c (vectorizable_load): Likewise.
25881         * doc/tm.texi: Regenerate.
25883 2017-07-05  Georg-Johann Lay  <avr@gjlay.de>
25885         Graceful degrade if Binutils PR21472 is not available.
25887         PR target/81072
25888         * configure.ac [target=avr]: WARN instead of ERROR if avrxmega3
25889         .rodata in flash test fails.
25890         (HAVE_LD_AVR_AVRXMEGA3_RODATA_IN_FLASH): Define it if test passes.
25891         * confgure: Regenerate.
25892         * config.in: Regenerate.
25893         * config/avr/avr.c (avr_asm_named_section)
25894         [HAVE_LD_AVR_AVRXMEGA3_RODATA_IN_FLASH]: Only trigger
25895         __do_copy_data for stuff in .rodata if flash_pm_offset = 0.
25896         (avr_asm_init_sections): Same.
25898 2017-07-05  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
25900         * config/arm/neon.md (fma<VCVTF:mode>4): Remove %?.
25901         (fma<VH:mode>4_intrinsic): Likewise.
25902         (*fmsub<VCVTF:mode>4): Likewise.
25903         (*fmsub<VH:mode>4_intrinsic): Likewise.
25905 2017-07-05  Georg-Johann Lay  <avr@gjlay.de>
25907         PR target/81305
25908         * config/avr/avr.c (avr_out_movhi_mr_r_xmega) [CONSTANT_ADDRESS_P]:
25909         Don't depend on "optimize > 0".
25910         (out_movhi_r_mr, out_movqi_mr_r): Same.
25911         (out_movhi_mr_r, out_movqi_r_mr): Same.
25912         (avr_address_cost) [CONSTANT_ADDRESS_P]: Don't depend cost for
25913         io_address_operand on "optimize > 0".
25915 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25917         * tree-loop-distribution.c: Add general explanantion on the pass.
25918         (generate_loops_for_partition): Mark distributed loop.
25919         (pg_add_dependence_edges): New parameter.  Handle alias data
25920         dependence specially and record it in the parameter if asked.
25921         (struct pg_vdata, pg_edata, pg_edge_callback_data): New structs.
25922         (init_partition_graph_vertices, add_partition_graph_edge): New.
25923         (pg_skip_alias_edge, free_partition_graph_edata_cb): New.
25924         (free_partition_graph_vdata, build_partition_graph): New.
25925         (sort_partitions_by_post_order, merge_dep_scc_partitions): New.
25926         (pg_collect_alias_ddrs, break_alias_scc_partitions): New.
25927         (data_ref_segment_size, latch_dominated_by_data_ref): New.
25928         (compute_alias_check_pairs, version_loop_by_alias_check): New.
25929         (version_for_distribution_p, finalize_partitions): New.
25930         (distribute_loop): Handle alias data dependence specially.  Factor
25931         out loop fusion code as functions and call these functions.
25933 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25935         * tree-loop-distribution.c (classify_partition): New parameter and
25936         better handle reduction statement.
25937         (rdg_build_partitions): Revise comment.
25938         (distribute_loop): Compute statements in all partitions and pass it
25939         to classify_partition.
25941 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25943         * tree-loop-distribution.c (enum partition_type): New.
25944         (struct partition): New field type.
25945         (partition_merge_into): Add parameter.  Update partition type.
25946         (data_dep_in_cycle_p, update_type_for_merge): New functions.
25947         (build_rdg_partition_for_vertex): Compute partition type.
25948         (rdg_build_partitions): Dump partition type.
25949         (distribute_loop): Update calls to partition_merge_into.
25951 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25953         * tree-loop-distribution.c (struct ddr_hasher): New.
25954         (ddr_hasher::hash, ::equal, get_data_dependence): New function.
25955         (ddrs_table): New.
25956         (classify_partition): Call get_data_dependence.
25957         (pg_add_dependence_edges): Ditto.
25958         (distribute_loop): Release data dependence hash table.
25960 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25962         * tree-loop-distribution.c (ref_base_address): Delete.
25963         (similar_memory_accesses): Rename ...
25964         (share_memory_accesses): ... to this.  Check if partitions access
25965         the same memory reference.
25966         (distribute_loop): Call share_memory_accesses.
25968 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25970         * tree-loop-distribution.c (struct partition): New field recording
25971         its data reference.
25972         (partition_alloc, partition_free): Init and release data refs.
25973         (partition_merge_into): Merge data refs.
25974         (build_rdg_partition_for_vertex): Collect data refs for partition.
25975         (pg_add_dependence_edges): Change parameters from vector to bitmap.
25976         Update uses.
25977         (distribute_loop): Remve data refs from vertice data of partition
25978         graph.
25980 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25982         * tree-loop-distribution.c (params.h): Include header file.
25983         (MAX_DATAREFS_NUM, DR_INDEX): New macro.
25984         (datarefs_vec): New global var.
25985         (create_rdg_vertices): Use datarefs_vec directly.
25986         (free_rdg): Don't free data references.
25987         (build_rdg): Update use.  Don't free data references.
25988         (distribute_loop): Compute global variable for data references.
25989         Bail out if there are too many data references.
25991 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25993         * tree-loop-distribution.c (loop_nest): New global var.
25994         (build_rdg): Use loop directly, rather than loop nest.
25995         (pg_add_dependence_edges): Remove loop nest parameter.  Use global
25996         variable directly.
25997         (distribute_loop): Compute global variable loop nest.  Update use.
25999 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
26001         * tree-loop-distribution.c (enum fuse_type, fuse_message): New.
26002         (partition_merge_into): New parameter.  Dump reason for fusion.
26003         (distribute_loop): Update use of partition_merge_into.
26005 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
26007         * tree-loop-distribution.c (bb_top_order_index): New.
26008         (bb_top_order_index_size, bb_top_order_cmp): New.
26009         (stmts_from_loop): Use topological order.
26010         (pass_loop_distribution::execute): Compute and release topological
26011         order for basic blocks.
26013 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
26015         * tree-loop-distribution.c (pass_loop_distribution::execute): Skip
26016         if no loops.
26018 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
26020         * cfgloop.h (struct loop): Add comment.  New field orig_loop_num.
26021         * cfgloopmanip.c (lv_adjust_loop_entry_edge): Comment change.
26022         * internal-fn.c (expand_LOOP_DIST_ALIAS): New function.
26023         * internal-fn.def (LOOP_DIST_ALIAS): New.
26024         * tree-vectorizer.c (fold_loop_vectorized_call): Rename to ...
26025         (fold_loop_internal_call): ... this.
26026         (vect_loop_dist_alias_call): New function.
26027         (set_uid_loop_bbs): Call fold_loop_internal_call.
26028         (vectorize_loops): Fold IFN_LOOP_VECTORIZED and IFN_LOOP_DIST_ALIAS
26029         internal calls.
26031 2017-07-04  Uros Bizjak  <ubizjak@gmail.com>
26033         PR target/81300
26034         * config/i386/i386.md (setcc + movzbl/and to xor + setcc peepholes):
26035         Require dead FLAGS_REG at the beginning of a peephole.
26037 2017-07-04  Uros Bizjak  <ubizjak@gmail.com>
26039         PR target/81294
26040         * config/i386/adxintrin.h (_subborrow_u32): Swap _X and _Y
26041         arguments in the call to __builtin_ia32_sbb_u32.
26042         (_subborrow_u64): Swap _X and _Y arguments in the call to
26043         __builtin_ia32_sbb_u64.
26045 2017-07-04  Jakub Jelinek  <jakub@redhat.com>
26047         PR debug/81278
26048         * tree-vrp.c (compare_assert_loc): Turn into a function template
26049         with stable template parameter.  Only test if a->e is NULL,
26050         !a->e == !b->e has been verified already.  Use e == NULL or
26051         e != NULL instead of e or ! e tests.  If stable is true, don't use
26052         iterative_hash_expr, on the other side allow a or b or both NULL
26053         and sort the NULLs last.
26054         (process_assert_insertions): Sort using compare_assert_loc<false>
26055         instead of compare_assert_loc, later sort using
26056         compare_assert_loc<true> before calling process_assert_insertions_for
26057         in a loop.  Use break instead of continue once seen NULL pointer.
26059 2017-07-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
26061         * config/arm/driver-arm.c (arm_cpu_table): Add entry for ARM
26062         Cortex-R7 and Cortex-R8 processors.
26064 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
26066         * ipa-utils.c (ipa_merge_profiles): Fix merging when dst is
26067         uninitialized while src is not.
26069 2017-07-04  Richard Earnshaw  <rearnsha@arm.com>
26071         * common/config/arm/arm-common.c: Adjust include path for
26072         arm-cpu-cdata.h
26073         * config/arm/t-arm (TM_H): Adjust path for arm-cpu.h.
26074         (arm-cpu.h): Create in build directory.  Adjust dependency rules.
26075         (arm-cpu-data.h): Likewise.
26076         (arm-cpu-cdata.h): Likewise.
26077         * config/arm/arm-cpu.h: Delete.
26078         * config/arm/arm-cpu-cdata.h: Delete.
26079         * config/arm/arm-cpu-data.h: Delete.
26081 2017-07-04  James Greenhalgh  <james.greenhalgh@arm.com>
26083         * config/arm/arm-cpus.in (cortex-a55): New.
26084         (cortex-a75): Likewise.
26085         (cortex-a75.cortex-a55): Likewise.
26086         * config/arm/driver-arm.c (arm_cpu_table): Add cortex-a55 and
26087         cortex-a75.
26088         * doc/invoke.texi (-mcpu): Document cortex-a55 and cortex-a75.
26089         * config/arm/arm-cpu-cdata.h: Regenerate.
26090         * config/arm/arm-cpu-data.h: Regenerate.
26091         * config/arm/arm-cpu.h: Regenerate.
26092         * config/arm/arm-tables.opt: Regenerate.
26093         * config/arm/arm-tune.md: Regenerate.
26095 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
26097         * haifa-sched.c (sched_create_recovery_edges): Update profile.
26099 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
26101         * bb-reorder.c (better_edge_p): Fix handling of uninitialized
26102         probability.
26104 2017-07-04  Richard Sandiford  <richard.sandiford@linaro.org>
26106         PR tree-optimization/81292
26107         * tree-ssa-strlen.c (handle_builtin_strlen): When setting
26108         full_string_p, also call adjust_related_strinfos if the adjustment
26109         is simple, otherwise invalidate related strinfos.
26111 2017-07-04  Martin Liska  <mliska@suse.cz>
26113         PR sanitizer/81040
26114         * sanopt.c (sanitize_rewrite_addressable_params): Mark the
26115         newly created variable as DECL_IGNORED_P.
26117 2017-07-04  Martin Liska  <mliska@suse.cz>
26119         PR ipa/81293
26120         * ipa-inline.c (inline_small_functions):
26121         Use xstrdup_for_dump.
26123 2017-07-04  Tom de Vries  <tom@codesourcery.com>
26125         * graph.c (draw_cfg_edges): Save and restore EDGE_DFS_BACK.
26127 2017-07-03  Dominique d'Humieres  <dominiq@lps.ens.fr>
26129         PR target/81033
26130         * config/darwin.c (darwin_function_switched_text_sections):
26131         Fix spaces.
26133 2017-07-03  Jan Hubicka  <hubicka@ucw.cz>
26135         * tree-vect-loop-manip.c (vect_do_peeling): Fix scaling up.
26137 2017-07-03  Richard Earnshaw  <rearnsha@arm.com>
26139         * doc/invoke.texi (ARM Options): Add -mbe8 and -mbe32 to option summary.
26141 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26143         * tree-vect-loop.c (vect_analyze_loop_2): Treat min_scalar_loop_bound,
26144         min_profitable_iters, and th as inclusive lower bounds.
26145         Fix LOOP_VINFO_PEELING_FOR_GAPS condition.
26146         (vect_estimate_min_profitable_iters): Return inclusive lower bounds
26147         for min_profitable_iters and min_profitable_estimate.
26148         (vect_transform_loop): Treat th as an inclusive lower bound.
26149         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
26151 2017-07-03  Dominique d'Humieres  <dominiq@lps.ens.fr>
26153         PR target/81033
26154         * config/darwin.c (darwin_function_switched_text_sections):
26155         Replace DECL_NAME with DECL_ASSEMBLER_NAME, split assemble_name_raw
26156         in two pieces, and suppress the use of buf.
26158 2017-07-03  Nathan Sidwell  <nathan@acm.org>
26160         * hash-table.h (hash_table_mod1): Fix indentation.
26162 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26164         PR middle-end/81290
26165         * predict.c (force_edge_cold): Be more careful about propagation
26166         backward.
26167         * profile-count.h (profile_probability::guessed,
26168         profile_probability::fdo, profile_count::guessed, profile_count::fdo):
26169         New.
26170         * tree-ssa-threadupdate.c (recompute_probabilities): Result is guessed.
26172 2017-07-03  James Greenhalgh  <james.greenhalgh@arm.com>
26174         * doc/invoke.texi (rcpc architecture extension): Document it.
26176 2017-07-03  Richard Biener  <rguenther@suse.de>
26178         PR tree-optimization/60510
26179         * tree-vect-loop.c (vect_create_epilog_for_reduction): Pass in
26180         the scalar reduction PHI and use it.
26181         (vectorizable_reduction): Properly guard the single_defuse_cycle
26182         path for non-SLP reduction chains where we cannot use it.
26183         Rework reduc_def/index and vector type deduction.  Rework
26184         vector operand gathering during reduction op code-gen.
26185         * tree-vect-slp.c (vect_analyze_slp): For failed SLP reduction
26186         chains dissolve the chain and leave it to non-SLP reduction
26187         handling.
26189 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26191         * tree-data-ref.h (dr_alignment): Declare.
26192         * tree-data-ref.c (dr_alignment): New function.
26193         * tree-vectorizer.h (dataref_aux): Remove base_element_aligned.
26194         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
26195         set it.
26196         * tree-vect-stmts.c (vectorizable_store): Use dr_alignment.
26198 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26200         * tree-data-ref.h (innermost_loop_behavior): Add base_alignment
26201         and base_misalignment fields.
26202         (DR_BASE_ALIGNMENT, DR_BASE_MISALIGNMENT): New macros.
26203         * tree-data-ref.c: Include builtins.h.
26204         (dr_analyze_innermost): Set up the new innmost_loop_behavior fields.
26205         * tree-vectorizer.h (STMT_VINFO_DR_BASE_ALIGNMENT): New macro.
26206         (STMT_VINFO_DR_BASE_MISALIGNMENT): Likewise.
26207         * tree-vect-data-refs.c: Include tree-cfg.h.
26208         (vect_compute_data_ref_alignment): Use the new innermost_loop_behavior
26209         fields instead of calculating an alignment here.
26210         (vect_analyze_data_refs): Use dr_analyze_innermost.  Dump the new
26211         innermost_loop_behavior fields.
26213 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26215         * tree-data-ref.h (innermost_loop_behavior): Add a step_alignment
26216         field.
26217         (DR_STEP_ALIGNMENT): New macro.
26218         * tree-vectorizer.h (STMT_VINFO_DR_STEP_ALIGNMENT): Likewise.
26219         * tree-data-ref.c (dr_analyze_innermost): Initalize step_alignment.
26220         (create_data_ref): Print it.
26221         * tree-vect-stmts.c (vectorizable_load): Use the step alignment
26222         to tell whether the step preserves vector (mis)alignment.
26223         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Likewise.
26224         Move the check for an integer step and generalise to all INTEGER_CST.
26225         (vect_analyze_data_refs): Set DR_STEP_ALIGNMENT when setting DR_STEP.
26226         Print the outer step alignment.
26228 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26230         * tree-data-ref.h (innermost_loop_behavior): Replace aligned_to
26231         with offset_alignment.
26232         (DR_ALIGNED_TO): Delete.
26233         (DR_OFFSET_ALIGNMENT): New macro.
26234         * tree-vectorizer.h (STMT_VINFO_DR_ALIGNED_TO): Delete.
26235         (STMT_VINFO_DR_OFFSET_ALIGNMENT): New macro.
26236         * tree-data-ref.c (dr_analyze_innermost): Update after above changes.
26237         (create_data_ref): Likewise.
26238         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Likewise.
26239         (vect_analyze_data_refs): Likewise.
26240         * tree-if-conv.c (if_convertible_loop_p_1): Use memset before
26241         creating dummy innermost behavior.
26243 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26245         * tree-data-ref.h (dr_analyze_innermost): Replace the dr argument
26246         with a "innermost_loop_behavior *" and refeence tree.
26247         * tree-data-ref.c (dr_analyze_innermost): Likewise.
26248         (create_data_ref): Update call accordingly.
26249         * tree-predcom.c (find_looparound_phi): Likewise.
26251 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26253         * tree-vectorizer.h (_stmt_vec_info): Replace individual dr_*
26254         fields with dr_wrt_vec_loop.
26255         (STMT_VINFO_DR_BASE_ADDRESS, STMT_VINFO_DR_INIT, STMT_VINFO_DR_OFFSET)
26256         (STMT_VINFO_DR_STEP, STMT_VINFO_DR_ALIGNED_TO): Update accordingly.
26257         (STMT_VINFO_DR_WRT_VEC_LOOP): New macro.
26258         (vect_dr_behavior): New function.
26259         (vect_create_addr_base_for_vector_ref): Remove loop parameter.
26260         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
26261         vect_dr_behavior.  Use a step_preserves_misalignment_p boolean to
26262         track whether the step preserves the misalignment.
26263         (vect_create_addr_base_for_vector_ref): Remove loop parameter.
26264         Use vect_dr_behavior.
26265         (vect_setup_realignment): Update call accordingly.
26266         (vect_create_data_ref_ptr): Likewise.  Use vect_dr_behavior.
26267         * tree-vect-loop-manip.c (vect_gen_prolog_loop_niters): Update
26268         call to vect_create_addr_base_for_vector_ref.
26269         (vect_create_cond_for_align_checks): Likewise.
26270         * tree-vect-patterns.c (vect_recog_bool_pattern): Copy
26271         STMT_VINFO_DR_WRT_VEC_LOOP as a block.
26272         (vect_recog_mask_conversion_pattern): Likewise.
26273         * tree-vect-stmts.c (compare_step_with_zero): Use vect_dr_behavior.
26274         (new_stmt_vec_info): Remove redundant zeroing.
26276 2017-07-03  Richard Earnshaw  <rearnsha@arm.com>
26278         * common/config/arm/arm-common.c (arm_be8_option): New function.
26279         * config/arm/arm-isa.h (isa_feature): Add new feature bit isa_bit_be8.
26280         (ISA_ARMv6): Add isa_bit_be8.
26281         * config/arm/arm.h (arm_be8_option): Add prototype.
26282         (BE8_SPEC_FUNCTION): New define.
26283         (EXTRA_SPEC_FUNCTIONS): Add BE8_SPEC_FUNCTION.
26284         * config/arm/arm.opt (mbig-endian): Mark as Negative of mlittle-endian.
26285         (mlittle-endian): Similarly.
26286         (mbe8, mbe32): New options.
26287         * config/arm/bpabi.h (BE8_LINK_SPEC): Call arm_be8_option.
26288         * doc/invoke.texi (ARM Options): Document -mbe8 and -mbe32.
26290 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26292         * tree-cfg.c (gimple_find_sub_bbs): Fix profile updating.
26294 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26296         * tree-cfgcleanup.c (want_merge_blocks_p): New function.
26297         (cleanup_tree_cfg_bb): Use it.
26298         * profile-count.h (profile_count::of_for_merging, profile_count::merge):
26299         New functions.
26300         * tree-cfg.c (gimple_merge_blocks): Use profile_count::merge.
26302 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26304         PR bootstrap/81285
26305         * loop-doloop.c (add_test): Update profile.
26307 2017-07-03  Martin Liska  <mliska@suse.cz>
26309         PR sanitize/81040
26310         * sanopt.c (rewrite_usage_of_param): New function.
26311         (sanitize_rewrite_addressable_params): Likewise.
26312         (pass_sanopt::execute): Call rewrite_usage_of_param.
26314 2017-07-03  Richard Biener  <rguenther@suse.de>
26316         * tree-vect-loop.c (vect_create_epilog_for_reduction): Revert
26317         back to using VIEW_CONVERT_EXPR.
26319 2017-07-03  Martin Liska  <mliska@suse.cz>
26321         PR other/78366
26322         * doc/extend.texi: Document when a resolver function is
26323         generated for target_clones.
26325 2017-07-03  Martin Liska  <mliska@suse.cz>
26327         * asan.c (asan_emit_stack_protection): Unpoison just red zones
26328         and shadow memory of auto variables which are subject of
26329         use-after-scope sanitization.
26330         (asan_expand_mark_ifn): Add do set only when is_poison.
26332 2016-07-03  Richard Biener  <rguenther@suse.de>
26334         * tree-vect-loop.c (vect_analyze_loop_operations): Also analyze
26335         reduction PHIs.
26336         (vect_force_simple_reduction): Record reduction def -> phi mapping.
26337         (vectorizable_reduction): Perform reduction PHI creation when
26338         visiting a reduction PHI and adjust and simplify code generation
26339         phase of the reduction op.  Cache dts, use fold_binary, not fold_build2.
26340         (vect_transform_loop): Visit reduction PHIs.
26341         * tree-vect-slp.c (vect_get_and_check_slp_defs): Record reduction
26342         defs into the SLP tree.
26343         (vect_build_slp_tree): Reduction defs terminate the recursion.
26344         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Allow lookup
26345         of reduction defs.
26346         (vect_get_vec_defs_for_stmt_copy): Export.
26347         (vect_get_vec_defs): Likewise.
26348         * tree-vectorizer.h (struct _stmt_vec_info): Amend reduc_def
26349         purpose.
26350         (vect_get_vec_defs_for_stmt_copy): Declare.
26351         (vect_get_vec_defs): Likewise.
26353 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26355         * tree-data-ref.c (dr_analyze_innermost): Replace the "nest"
26356         parameter with a "loop" parameter and use it instead of the
26357         loop containing DR_STMT.  Don't check simple_iv when doing
26358         BB analysis.  Describe the two analysis modes in the comment.
26360 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26362         PR tree-optimization/69468
26363         * tree-ssa-tail-merge.c (ignore_edge_flags): New constant.
26364         (find_same_succ_bb): Handle ignore_edge_flags.
26366 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26368         PR tree-optimization/81192
26369         * tree-ssa-tail-merge.c (same_succ_hash): Use bb->loop_father->num in
26370         hash.
26371         (same_succ::equal): Don't find bbs to be equal if bb->loop_father
26372         differs.
26373         (find_same_succ_bb): Remove obsolete test on bb->loop_father->latch.
26375 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26377         PR tree-optimization/81192
26378         * tree-ssa-tail-merge.c (same_succ_flush_bb): Handle
26379         BB_SAME_SUCC (bb) == NULL.
26381 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26383         * cfgrtl.c (rtl_verify_edges): Enable checking of profile_probability
26384         consistency.
26386 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26388         * dumpfile.c: Include profile-count.h
26389         * tree-cfg.c (gimple_duplicate_sese_tail): Drop UNUSED attributes;
26390         update profile.
26391         (insert_cond_bb): Update profile.
26392         * tree-cfg.h (insert_cond_bb): Update prototype.
26393         * tree-chkp-opt.c (chkp_optimize_string_function_calls): Update.
26394         * tree-dump.c: Do not include tree-cfg.
26396 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26398         * bb-reorder.c (fix_up_crossing_landing_pad): Update profile.
26400 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26402         * expect.c (dw2_build_landing_pads): Update profile of the landing pad
26403         bb.
26405 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26407         * tree-complex.c (expand_complex_div_wide): update profile.
26409 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26410             Alan Hayward  <alan.hayward@arm.com>
26411             David Sherwood  <david.sherwood@arm.com>
26413         * Makefile.in (MACHMODE_H): Remove insn-modes.h
26414         (CORETYPES_H): New define.
26415         (MOSTLYCLEANFILES): Add insn-modes-inline.h.
26416         (insn-modes-inline.h, s-modes-inline-h): New rules.
26417         (generated_files): Add insn-modes-inline.h.
26418         (RTL_BASE_H, TREE_CORE_H): Use CORETYPES_H instead of coretypes.h.
26419         (build/gensupport.o, build/ggc-none.o, build/print-rtl.o): Likewise.
26420         (build/read-md.o, build/read-rtl.o, build/rtl.o): Likewise.
26421         (build/vec.o, build/hash-table.o, build/inchash.o): Likewise.
26422         (build/gencondmd.o, build/genattr.o, build/genattr-common.o): Likewise.
26423         (build/genattrtab.o, build/genautomata.o, build/gencheck.o): Likewise.
26424         (build/gencodes.o, build/genconditions.o): Likewise.
26425         (build/genconfig.o, build/genconstants.o, build/genemit.o): Likewise.
26426         (build/genenums.o, build/genextract.o, build/genflags.o): Likewise.
26427         (build/gentarget-def.o, build/genmddeps.o, build/genopinit.o)
26428         (build/genoutput.o, build/genpeep.o, build/genpreds.o): Likewise.
26429         (build/genrecog.o, build/genmddump.o, build/genmatch.o): Likewise.
26430         (build/gencfn-macros.o, build/gcov-iov.o): Likewise.
26431         * coretypes.h: Include everything up to real.h for generators.
26432         Include insn-modes.h first.  Include wide-int-print.h after
26433         wide-int.h.  Include insn-modes-inline.h and then machmode.h.
26434         * machmode.h: Don't include insn-modes.h here.
26435         * function-tests.c: Remove includes of signop.h, machmode.h,
26436         double-int.h and wide-int.h.
26437         * rtl.h: Likewise.
26438         * gcc-rich-location.c: Remove includes of machmode.h, double-int.h
26439         and wide-int.h.
26440         * optc-save-gen.awk: Likewise.
26441         * gencheck.c (BITS_PER_UNIT): Delete dummy definition.
26442         * godump.c: Remove include of wide-int-print.h.
26443         * pretty-print.h: Likewise.
26444         * wide-int-print.cc: Likewise.
26445         * wide-int.cc: Likewise.
26446         * hash-map-tests.c: Remove include of signop.h.
26447         * hash-set-tests.c: Likewise.
26448         * rtl-tests.c: Likewise.
26449         * mkconfig.sh: Remove include of machmode.h.
26450         * genmodes.c (emit_insn_modes_h): Split emission of inline functions
26451         into...
26452         (emit_insn_modes_inline_h): ...this new function.  Emit the code
26453         into an insn-modes-inline.h header file, adding appropriate
26454         include guards and end comments.
26455         (emit_insn_modes_c_header): Remove include of machmode.h.
26456         (emit_min_insn_modes_c_header): Include coretypes.h rather than
26457         machmode.h.
26458         (main): Handle -i flag and call emit_insn_modes_inline_h when
26459         it is passed.
26461 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26463         * tree-ssa-strlen.c (strinfo): Rename the length field to
26464         nonzero_chars.  Add a full_string_p field.
26465         (compare_nonzero_chars, zero_length_string_p): New functions.
26466         (get_addr_stridx): Add an offset_out parameter.
26467         Use compare_nonzero_chars.
26468         (get_stridx): Update accordingly.  Use compare_nonzero_chars.
26469         (new_strinfo): Update after above changes to strinfo.
26470         (set_endptr_and_length): Set full_string_p.
26471         (get_string_length): Update after above changes to strinfo.
26472         (unshare_strinfo): Update call to new_strinfo.
26473         (maybe_invalidate): Likewise.
26474         (get_stridx_plus_constant): Change off to unsigned HOST_WIDE_INT.
26475         Use compare_nonzero_chars and zero_string_p.  Treat nonzero_chars
26476         as a uhwi instead of an shwi.  Update after above changes to
26477         strinfo and new_strinfo.
26478         (zero_length_string): Assert that chainsi contains full strings.
26479         Use zero_length_string_p.  Update call to new_strinfo.
26480         (adjust_related_strinfos): Update after above changes to strinfo.
26481         Copy full_string_p from origsi.
26482         (adjust_last_stmt): Use zero_length_string_p.
26483         (handle_builtin_strlen): Update after above changes to strinfo and
26484         new_strinfo.  Install the lhs as the string length if the previous
26485         entry didn't describe a full string.
26486         (handle_builtin_strchr): Update after above changes to strinfo
26487         and new_strinfo.
26488         (handle_builtin_strcpy): Likewise.
26489         (handle_builtin_strcat): Likewise.
26490         (handle_builtin_malloc): Likewise.
26491         (handle_pointer_plus): Likewise.
26492         (handle_builtin_memcpy): Likewise.  Track nonzero characters
26493         that aren't necessarily followed by a nul terminator.
26494         (handle_char_store): Likewise.
26496 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26498         PR tree-optimization/80769
26499         * tree-ssa-strlen.c (strinfo): Document that "stmt" is also used
26500         for malloc and calloc.  Document the new invariant that all related
26501         strinfos have delayed lengths or none do.
26502         (verify_related_strinfos): Move earlier in file.
26503         (set_endptr_and_length): New function, split out from...
26504         (get_string_length): ...here.  Also set the lengths of related
26505         strinfos.
26506         (zero_length_string): Assert that chainsi has known (rather than
26507         delayed) lengths.
26508         (adjust_related_strinfos): Likewise.
26510 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26512         PR tree-optimization/81136
26513         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Only
26514         assert that two references with the same misalignment have the same
26515         compile-time misalignment if those compile-time misalignments
26516         are known.
26518 2017-07-01  Andi Kleen  <ak@linux.intel.com>
26520         * print-tree.c (print_node): Print all attributes.
26522 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26524         * cfg.c (scale_bbs_frequencies): New function.
26525         * cfg.h (scale_bbs_frequencies): Declare it.
26526         * cfgloopanal.c (single_likely_exit): Cleanup.
26527         * cfgloopmanip.c (scale_loop_frequencies): Take profile_probability
26528         as parameter.
26529         (scale_loop_profile): Likewise.
26530         (loop_version): Likewise.
26531         (create_empty_loop_on_edge): Update.
26532         * cfgloopmanip.h (scale_loop_frequencies, scale_loop_profile,
26533         scale_loop_frequencies, scale_loop_profile, loopify,
26534         loop_version): Update prototypes.
26535         * modulo-sched.c (sms_schedule): Update.
26536         * predict.c (unlikely_executed_edge_p): Also check probability.
26537         (probably_never_executed_edge_p): Fix typo.
26538         * tree-if-conv.c (version_loop_for_if_conversion): Update.
26539         * tree-parloops.c (gen_parallel_loop): Update.
26540         * tree-ssa-loop-ivcanon.c (try_peel_loop): Update.
26541         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
26542         * tree-ssa-loop-split.c (split_loop): Update.
26543         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Update.
26544         * tree-vect-loop-manip.c (vect_do_peeling): Update.
26545         (vect_loop_versioning): Update.
26546         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
26548 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26550         * trans-mem.c (split_bb_make_tm_edge): Update profile.
26552 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26554         * tree-if-conv.c (combine_blocks): Use make_single_succ_edge
26555         to keep profile consistent.
26557 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26559         * cfgrtl.c (rtl_flow_call_edges_add): Update profile.
26560         * tree-cfg.c (gimple_flow_call_edges_add): Likewise.
26561         * profile-count.h (max_safe_multiplier): Make unsigned.
26562         (profile_count::guessed_zero): New.
26564 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26566         * bb-reorder.c (fix_up_crossing_landing_pad,
26567         fix_crossing_conditional_branches): Use make_single_succ_edge
26568         to keep profile consistent.
26570 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26572         * tree-vect-loop.c (optimize_mask_stores): Use make_single_succ_edge
26573         to update profile.
26575 2017-07-01  Jakub Jelinek  <jakub@redhat.com>
26577         PR sanitizer/81262
26578         * bb-reorder.c (fix_up_fall_thru_edges): Move variable declarations to
26579         the right scopes, make sure cond_jump isn't preserved between multiple
26580         iterations.  Search for fallthru edge whenever there are 3+ edges and
26581         use find_fallthru_edge for it.
26583 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26585         Patch by Alexander Monakov <amonakov@ispras.ru>
26586         * sel-sched-ir.c (compute_succs_info): Handle uninitialized
26587         probabilities consistently.
26589 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26591         * pa.c (pa_expand_compare_and_swap_loop): Update call of
26592         emit_cmp_and_jump_insns.
26594 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26596         PR ipa/81261
26597         * tree-inline.c (expand_call_inline): Combine profile statuses.
26599 2017-06-30  Andrew Pinski  <apinski@cavium.com>
26601         * tree-if-conv.c (predicate_scalar_phi): Update new_stmt if
26602         fold_stmt returned true.
26604 2017-06-30  Nathan Sidwell  <nathan@acm.org>
26606         * ggc.h (empty_string): Delete.
26607         * cfgexpand.c (expand_asm_stmt): Use plain "".
26608         * optabs.c (expand_asm_memory_barrier): Likewise.
26609         * stringpool.c (empty_string): Delete.
26610         (digit_vector, digit_string): Delete.
26611         (ggc_alloc_string): Use plain "", don't optimize single digit
26612         strings.  Use ggc_alloc_atomic.
26614 2017-06-30  Richard Earnshaw  <rearnsha@arm.com>
26616         * rtlanal.c (insn_rtx_cost): If a parallel contains exactly one
26617         comparison set and one other set, use the cost of the non-comparison
26618         set.
26620 2017-06-30  Nathan Sidwell  <nathan@acm.org>
26622         * ggc.h: Replace all 'static inline' with plain 'inline'.  Fix
26623         some formatting.
26625 2017-06-30  Peter Bergner  <bergner@vnet.ibm.com>
26627         * tree-cfg.c (group_case_labels_stmt): Merge scanning and compressing
26628         loops.  Remove now unneeded calls to gimple_switch_set_label() that
26629         just set removed labels to NULL_TREE.
26631 2017-06-30  Aldy Hernandez  <aldyh@redhat.com>
26633         * tree-ssanames.c (set_range_info_raw): Abstract from ...
26634         (set_range_info): ...here.  Only call set_range_info_raw if domain
26635         is useful.
26636         (set_nonzero_bits): Call set_range_info_raw.
26637         * tree-ssanames.h (set_range_info_raw): New.
26639 2017-06-30  Jakub Jelinek  <jakub@redhat.com>
26641         PR target/81225
26642         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): For V8FI,
26643         V16FI and VI8F_256 iterators, use <store_mask_predicate> instead
26644         of nonimmediate_operand and <store_mask_constraint> instead of m
26645         for the input operand.  For V8FI iterator, always split if input
26646         is a MEM.  For V16FI and V8SF_256 iterators, don't test if both
26647         operands are MEM if <mask_applied>.  For VI4F_256 iterator, use
26648         <store_mask_predicate> instead of register_operand and
26649         <store_mask_constraint> instead of v for the input operand.  Make
26650         sure both operands aren't MEMs for if not <mask_applied>.
26652 2017-06-30  Sylvestre Ledru  <sylvestre@debian.org>
26654         * lto-wrapper.c (copy_file) Close both file descriptors before
26655         exiting normally.
26657 2017-06-30  Martin Liska  <mliska@suse.cz>
26659         PR ipa/81214
26660         * multiple_target.c (create_dispatcher_calls): Make ifunc
26661         also for function that don't have calls or are not referenced.
26663 2017-06-30  Richard Biener  <rguenther@suse.de>
26665         * tree-vect-slp.c (vect_slp_analyze_node_operations): Only
26666         analyze the first scalar stmt.  Move vector type computation
26667         for the BB case here from ...
26668         * tree-vect-stmts.c (vect_analyze_stmt): ... here.  Guard
26669         live operation processing in the SLP case properly.
26671 2017-06-30  Richard Biener  <rguenther@suse.de>
26673         * graph.c (draw_cfg_node_succ_edges): Fix broken dot syntax.
26675 2017-06-30  Martin Liska  <mliska@suse.cz>
26677         PR sanitizer/81021
26678         * tree-eh.c (lower_resx): Call BUILT_IN_ASAN_HANDLE_NO_RETURN
26679         before BUILT_IN_UNWIND_RESUME when ASAN is used.
26681 2017-06-30  Yvan Roux  <yvan.roux@linaro.org>
26683         * doc/invoke.texi (AArch64): Add missing options and remove redundant
26684         ones.
26686 2017-06-30  Richard Biener  <rguenther@suse.de>
26688         PR tree-optimization/81249
26689         * tree-vect-loop.c (vect_create_epilog_for_reduction): Convert
26690         condition reduction result to original scalar type.
26692 2017-06-30  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26694         * profile-count.h (enum profile_quality): Fix typos and whitespace
26695         issues.
26697 2017-06-30  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26699         * config/s390/s390.c (s390_expand_setmem): Adjust to the new data
26700         type for branch probabilities.
26702 2017-06-29  Julian Brown  <julian@codesourcery.com>
26703             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
26705         * config/aarch64/aarch64-fusion-pairs.def: Add ALU_BRANCH entry.
26706         * config/aarch64/aarch64.c (AARCH64_FUSE_ALU_BRANCH): New fusion type.
26707         (thunderx2t99_tunings): Set AARCH64_FUSE_ALU_BRANCH flag.
26708         (aarch_macro_fusion_pair_p): Add support for AARCH64_FUSE_ALU_BRANCH.
26710 2017-06-29  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
26712         * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Push the
26713         check for CC usage into AARCH64_FUSE_CMP_BRANCH.
26714         * config/i386/i386.c (ix86_macro_fusion_pair_p): Push the check for
26715         CC usage from generic code to here.
26716         * sched-deps.c (sched_macro_fuse_insns): Move the condition for
26717         CC usage into the target macros.
26719 2017-06-29  Maya Rashish  <coypu@sdf.org>
26721         * config/netbsd.h (NETBSD_LIB_SPEC): Add -lc when creating shared
26722         objects.
26724 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26726         * arm/arm-builtins.c: Include profile-count.h
26727         * except.c (sjlj_emit_function_enter): Use
26728         profile_probability::unlikely.
26730 2017-06-29  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
26732         * config/rs6000/rs6000.c (toc_relative_expr_p): Make tocrel_base
26733         and tocrel_offset be pointer args rather than implicitly using
26734         static versions.
26735         (legitimate_constant_pool_address_p, rs6000_emit_move,
26736         const_load_sequence_p, adjust_vperm): Add local tocrel_base and
26737         tocrel_offset and use in toc_relative_expr_p call.
26738         (print_operand, print_operand_address): Use static tocrel_base_oac
26739         and tocrel_offset_oac.
26740         (rs6000_output_addr_const_extra): Use static tocrel_base_oac and
26741         tocrel_offset_oac.
26743 2017-06-29  Maya Rashish  <coypu@sdf.org>
26745         * config/vax/builtins.md (ffssi2_internal): Correct constraint.
26747 2017-06-29  Eric Botcazou  <ebotcazou@adacore.com>
26749         * expr.c (expand_expr) <normal_inner_ref>: When testing for unaligned
26750         objects, take into account only the alignment of 'op0' and 'mode1' if
26751         'op0' is a MEM.
26753 2017-06-29  Steve Ellcey  <sellcey@cavium.com>
26755         * ccmp.c (ccmp_tree_comparison_p): New function.
26756         (ccmp_candidate_p): Update to use above function.
26757         (get_compare_parts): New function.
26758         (expand_ccmp_next): Update to use new functions.
26759         (expand_ccmp_expr_1): Take tree arg instead of gimple, update to use
26760         new functions.
26761         (expand_ccmp_expr): Pass tree instead of gimple to expand_ccmp_expr_1,
26762         take mode as argument.
26763         * ccmp.h (expand_ccmp_expr): Add mode as argument.
26764         * expr.c (expand_expr_real_1): Pass mode as argument.
26766 2017-06-29  Segher Boessenkool  <segher@kernel.crashing.org>
26768         * combine.c (combine_instructions): Print insns to dump_file, together
26769         with their costs.
26771 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26773         * asan.c (asan_emit_stack_protection): Update.
26774         (create_cond_insert_point): Update.
26775         * auto-profile.c (afdo_propagate_circuit): Update.
26776         * basic-block.h (struct edge_def): Turn probability to
26777         profile_probability.
26778         (EDGE_FREQUENCY): Update.
26779         * bb-reorder.c (find_traces_1_round): Update.
26780         (better_edge_p): Update.
26781         (sanitize_hot_paths): Update.
26782         * cfg.c (unchecked_make_edge): Initialize probability to uninitialized.
26783         (make_single_succ_edge): Update.
26784         (check_bb_profile): Update.
26785         (dump_edge_info): Update.
26786         (update_bb_profile_for_threading): Update.
26787         * cfganal.c (connect_infinite_loops_to_exit): Initialize new edge
26788         probabilitycount to 0.
26789         * cfgbuild.c (compute_outgoing_frequencies): Update.
26790         * cfgcleanup.c (try_forward_edges): Update.
26791         (outgoing_edges_match): Update.
26792         (try_crossjump_to_edge): Update.
26793         * cfgexpand.c (expand_gimple_cond): Update make_single_succ_edge.
26794         (expand_gimple_tailcall): Update.
26795         (construct_init_block): Use make_single_succ_edge.
26796         (construct_exit_block): Use make_single_succ_edge.
26797         * cfghooks.c (verify_flow_info): Update.
26798         (redirect_edge_succ_nodup): Update.
26799         (split_edge): Update.
26800         (account_profile_record): Update.
26801         * cfgloopanal.c (single_likely_exit): Update.
26802         * cfgloopmanip.c (scale_loop_profile): Update.
26803         (set_zero_probability): Remove.
26804         (duplicate_loop_to_header_edge): Update.
26805         * cfgloopmanip.h (loop_version): Update prototype.
26806         * cfgrtl.c (try_redirect_by_replacing_jump): Update.
26807         (force_nonfallthru_and_redirect): Update.
26808         (update_br_prob_note): Update.
26809         (rtl_verify_edges): Update.
26810         (purge_dead_edges): Update.
26811         (rtl_lv_add_condition_to_bb): Update.
26812         * cgraph.c: (cgraph_edge::redirect_call_stmt_to_calle): Update.
26813         * cgraphunit.c (init_lowered_empty_function): Update.
26814         (cgraph_node::expand_thunk): Update.
26815         * cilk-common.c: Include profile-count.h
26816         * dojump.c (inv): Remove.
26817         (jumpifnot): Update.
26818         (jumpifnot_1): Update.
26819         (do_jump_1): Update.
26820         (do_jump): Update.
26821         (do_jump_by_parts_greater_rtx): Update.
26822         (do_compare_rtx_and_jump): Update.
26823         * dojump.h (jumpifnot, jumpifnot_1, jumpif_1, jumpif, do_jump,
26824         do_jump_1. do_compare_rtx_and_jump): Update prototype.
26825         * dwarf2cfi.c: Include profile-count.h
26826         * except.c (dw2_build_landing_pads): Use make_single_succ_edge.
26827         (sjlj_emit_dispatch_table): Likewise.
26828         * explow.c: Include profile-count.h
26829         * expmed.c (emit_store_flag_force): Update.
26830         (do_cmp_and_jump): Update.
26831         * expr.c (compare_by_pieces_d::generate): Update.
26832         (compare_by_pieces_d::finish_mode): Update.
26833         (emit_block_move_via_loop): Update.
26834         (store_expr_with_bounds): Update.
26835         (store_constructor): Update.
26836         (expand_expr_real_2): Update.
26837         (expand_expr_real_1): Update.
26838         * expr.h (try_casesi, try_tablejump): Update prototypes.
26839         * gimple-pretty-print.c (dump_probability): Update.
26840         (dump_profile): New.
26841         (dump_gimple_label): Update.
26842         (dump_gimple_bb_header): Update.
26843         * graph.c (draw_cfg_node_succ_edges): Update.
26844         * hsa-gen.c (convert_switch_statements): Update.
26845         * ifcvt.c (cheap_bb_rtx_cost_p): Update.
26846         (find_if_case_1): Update.
26847         (find_if_case_2): Update.
26848         * internal-fn.c (expand_arith_overflow_result_store): Update.
26849         (expand_addsub_overflow): Update.
26850         (expand_neg_overflow): Update.
26851         (expand_mul_overflow): Update.
26852         (expand_vector_ubsan_overflow): Update.
26853         * ipa-cp.c (good_cloning_opportunity_p): Update.
26854         * ipa-split.c (split_function): Use make_single_succ_edge.
26855         * ipa-utils.c (ipa_merge_profiles): Update.
26856         * loop-doloop.c (add_test): Update.
26857         (doloop_modify): Update.
26858         * loop-unroll.c (compare_and_jump_seq): Update.
26859         (unroll_loop_runtime_iterations): Update.
26860         * lra-constraints.c (lra_inheritance): Update.
26861         * lto-streamer-in.c (input_cfg): Update.
26862         * lto-streamer-out.c (output_cfg): Update.
26863         * mcf.c (adjust_cfg_counts): Update.
26864         * modulo-sched.c (sms_schedule): Update.
26865         * omp-expand.c (expand_omp_for_init_counts): Update.
26866         (extract_omp_for_update_vars): Update.
26867         (expand_omp_ordered_sink): Update.
26868         (expand_omp_for_ordered_loops): Update.
26869         (expand_omp_for_generic): Update.
26870         (expand_omp_for_static_nochunk): Update.
26871         (expand_omp_for_static_chunk): Update.
26872         (expand_cilk_for): Update.
26873         (expand_omp_simd): Update.
26874         (expand_omp_taskloop_for_outer): Update.
26875         (expand_omp_taskloop_for_inner): Update.
26876         * omp-simd-clone.c (simd_clone_adjust): Update.
26877         * optabs.c (expand_doubleword_shift): Update.
26878         (expand_abs): Update.
26879         (emit_cmp_and_jump_insn_1): Update.
26880         (expand_compare_and_swap_loop): Update.
26881         * optabs.h (emit_cmp_and_jump_insns): Update prototype.
26882         * predict.c (predictable_edge_p): Update.
26883         (edge_probability_reliable_p): Update.
26884         (set_even_probabilities): Update.
26885         (combine_predictions_for_insn): Update.
26886         (combine_predictions_for_bb): Update.
26887         (propagate_freq): Update.
26888         (estimate_bb_frequencies): Update.
26889         (force_edge_cold): Update.
26890         * profile-count.c (profile_count::dump): Add missing space into dump.
26891         (profile_count::debug): Add newline.
26892         (profile_count::differs_from_p): Explicitly convert to unsigned.
26893         (profile_count::stream_in): Update.
26894         (profile_probability::dump): New member function.
26895         (profile_probability::debug): New member function.
26896         (profile_probability::differs_from_p): New member function.
26897         (profile_probability::differs_lot_from_p): New member function.
26898         (profile_probability::stream_in): New member function.
26899         (profile_probability::stream_out): New member function.
26900         * profile-count.h (profile_count_quality): Rename to ...
26901         (profile_quality): ... this one.
26902         (profile_probability): New.
26903         (profile_count): Update.
26904         * profile.c (compute_branch_probabilities): Update.
26905         * recog.c (peep2_attempt): Update.
26906         * sched-ebb.c (schedule_ebbs): Update.
26907         * sched-rgn.c (find_single_block_region): Update.
26908         (compute_dom_prob_ps): Update.
26909         (schedule_region): Update.
26910         * sel-sched-ir.c (compute_succs_info): Update.
26911         * stmt.c (struct case_node): Update.
26912         (do_jump_if_equal): Update.
26913         (get_outgoing_edge_probs): Update.
26914         (conditional_probability): Update.
26915         (emit_case_dispatch_table): Update.
26916         (expand_case): Update.
26917         (expand_sjlj_dispatch_table): Update.
26918         (emit_case_nodes): Update.
26919         * targhooks.c: Update.
26920         * tracer.c (better_p): Update.
26921         (find_best_successor): Update.
26922         * trans-mem.c (expand_transaction): Update.
26923         * tree-call-cdce.c: Update.
26924         * tree-cfg.c (gimple_split_edge): Upate.
26925         (move_sese_region_to_fn): Upate.
26926         * tree-cfgcleanup.c (cleanup_control_expr_graph): Upate.
26927         * tree-eh.c (lower_resx): Upate.
26928         (cleanup_empty_eh_move_lp): Upate.
26929         * tree-if-conv.c (version_loop_for_if_conversion): Update.
26930         * tree-inline.c (copy_edges_for_bb): Update.
26931         (copy_cfg_body): Update.
26932         * tree-parloops.c (gen_parallel_loop): Update.
26933         * tree-profile.c (gimple_gen_ic_func_profiler): Update.
26934         (gimple_gen_time_profiler): Update.
26935         * tree-ssa-dce.c (remove_dead_stmt): Update.
26936         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
26937         * tree-ssa-loop-im.c (execute_sm_if_changed): Update.
26938         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update.
26939         (unloop_loops): Update.
26940         (try_peel_loop): Update.
26941         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
26942         * tree-ssa-loop-split.c (connect_loops): Update.
26943         (split_loop): Update.
26944         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Update.
26945         (hoist_guard): Update.
26946         * tree-ssa-phionlycprop.c (propagate_rhs_into_lhs): Update.
26947         * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Update.
26948         (value_replacement): Update.
26949         * tree-ssa-reassoc.c (branch_fixup): Update.
26950         * tree-ssa-tail-merge.c (replace_block_by): Update.
26951         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges): Update.
26952         (create_edge_and_update_destination_phis): Update.
26953         (compute_path_counts): Update.
26954         (recompute_probabilities): Update.
26955         (update_joiner_offpath_counts): Update.
26956         (freqs_to_counts_path): Update.
26957         (duplicate_thread_path): Update.
26958         * tree-switch-conversion.c (hoist_edge_and_branch_if_true): Update.
26959         (struct switch_conv_info): Update.
26960         (gen_inbound_check): Update.
26961         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
26962         (vect_do_peeling): Update.
26963         (vect_loop_versioning): Update.
26964         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
26965         (optimize_mask_stores): Update.
26966         * ubsan.c (ubsan_expand_null_ifn): Update.
26967         * value-prof.c (gimple_divmod_fixed_value): Update.
26968         (gimple_divmod_fixed_value_transform): Update.
26969         (gimple_mod_pow2): Update.
26970         (gimple_mod_pow2_value_transform): Update.
26971         (gimple_mod_subtract): Update.
26972         (gimple_mod_subtract_transform): Update.
26973         (gimple_ic): Update.
26974         (gimple_stringop_fixed_value): Update.
26975         (gimple_stringops_transform): Update.
26976         * value-prof.h: Update.
26978 2017-06-29  Carl Love  <cel@us.ibm.com>
26980         * config/rs6000/rs6000-c.c: Add support for built-in functions
26981         vector signed int vec_signed (vector float);
26982         vector signed long long vec_signed (vector double);
26983         vector signed int vec_signed2 (vector double, vector double);
26984         vector signed int vec_signede (vector double);
26985         vector signed int vec_signedo (vector double);
26986         * config/rs6000/rs6000.c (rs6000_generate_vsigned2_code): Add
26987         instruction generator.
26988         * config/rs6000/vsx.md (UNSPEC_VSX_XVCVSPSXWS, UNSPEC_VSX_XVCVSPSXDS,
26989         UNSPEC_VSX_VSIGNED2): Add UNSPECS.
26990         (vsx_xvcvspsxws, vsx_xvcvdpuxds_scale, vsx_xvcvspuxws, vsigned2_v2df):
26991         Add define_insn.
26992         (vsignedo_v2df, vsignede_v2df, vunsigned2_v2df, vunsignedo_v2df,
26993         vunsignede_v2df): Add define_expands.
26994         * config/rs6000/rs6000-builtin.def (VEC_SIGNED, VEC_UNSIGNED,
26995         VEC_SIGNED2, VEC_UNSIGNED2, VEC_SIGNEDE, VEC_UNSIGNEDE, VEC_SIGNEDO,
26996         VEC_UNSIGNEDO): Add definitions.
26997         * config/vsx.md (UNSPEC_VSX_XVCVSPSXWS, UNSPEC_VSX_XVCVSPSXDS,
26998         UNSPEC_VSX_VSIGNED2): Add UNSPECs.
26999         (vsx_xvcvspsxws, vsx_xvcvspuxws): Add define_insn.
27000         (vsigned2_v2df, vsigendo_v2df, vsignede_v2df,
27001         vunsigned2_v2df, vunsignedo_v2df, vunsignede_v2df): Add define_expands.
27002         * config/rs6000/altivec.h (vec_signed, vec_signed2,
27003         vec_signede and vec_signedo, vec_unsigned, vec_unsigned2,
27004         vec_unsignede, vec_unsignedo): Add builtin defines.
27005         * config/rs6000-protos.h (rs6000_generate_vsigned2_code): Add extern
27006         declaration.
27007         * doc/extend.texi: Update the built-in documentation file for the
27008         new built-in functions.
27010 2017-06-29  Richard Biener  <rguenther@suse.de>
27012         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Do not add
27013         reduction chains to LOOP_VINFO_REDUCTIONS.
27014         * tree-vect-slp.c (vect_analyze_slp): Continue looking for
27015         SLP reductions after processing reduction chains.
27017 2017-06-29  Nathan Sidwell  <nathan@acm.org>
27019         * builtins.c (fold_builtin_FUNCTION): Use
27020         lang_hooks.decl_printable_name.
27022 2017-06-29  Peter Bergner  <bergner@vnet.ibm.com>
27024         PR middle-end/81194
27025         * cfgexpand.c (expand_gimple_stmt_1): Handle switch statements
27026         with only one label.
27027         * stmt.c (expand_case): Assert NCASES is greater than one.
27029 2017-06-29  Richard Biener  <rguenther@suse.de>
27031         * tree-cfg.c (group_case_labels_stmt): Return whether we changed
27032         anything.
27033         (group_case_labels): Likewise.
27034         (find_taken_edge): Push sanity checking on val to workers...
27035         (find_taken_edge_cond_expr): ... here
27036         (find_taken_edge_switch_expr): ... and here, handle cases
27037         with just a default label.
27038         * tree-cfg.h (group_case_labels_stmt): Adjust prototype.
27039         (group_case_labels): Likewise.
27040         * tree-cfgcleanup.c (execute_cleanup_cfg_post_optimizing): When
27041         group_case_labels does anything cleanup the CFG again.
27043 2017-06-29  Bin Cheng  <bin.cheng@arm.com>
27045         PR tree-optimization/81196
27046         * tree-ssa-loop-niter.c (number_of_iterations_cond): Handle loop
27047         exit condition comparing two IVs.
27049 2017-06-29  Richard Earnshaw  <rearnsha@arm.com>
27051         * config/arm/parsecpu.awk (gen_comm_data): Add initializer for
27052         profile to the dummy entry at the end of the list of architectures.
27053         * config/arm/arm-cpu-cdata.h: Regenerated.
27055 2017-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27056             Michael Collison <michael.collison@arm.com>
27058         PR target/70119
27059         * config/aarch64/aarch64.md (*aarch64_<optab>_reg_<mode>3_mask1):
27060         New pattern.
27061         (*aarch64_reg_<mode>3_neg_mask2): New pattern.
27062         (*aarch64_reg_<mode>3_minus_mask): New pattern.
27063         (*aarch64_<optab>_reg_di3_mask2): New pattern.
27064         * config/aarch64/aarch64.c (aarch64_rtx_costs): Account for cost
27065         of shift when the shift amount is masked with constant equal to
27066         the size of the mode.
27067         * config/aarch64/predicates.md (subreg_lowpart_operator): New
27068         predicate.
27070 2017-06-29  Martin Liska  <mliska@suse.cz>
27072         * config/i386/i386.opt: Change range from [1,5] to [0,5].
27074 2017-06-29  Yury Gribov  <tetra2005@gmail.com>
27076         PR bootstrap/80565
27077         * ipa-cp.c (allocate_and_init_ipcp_value): Add initialization
27078         code.
27079         * ipa-inline.h
27080         (edge_growth_cache_entry::edge_growth_cache_entry): New
27081         function.
27082         (reset_edge_growth_cache): Update to use constructor.
27084 2017-06-28  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
27086         * config/aarch64/aarch64.h (AARCH64_EXPAND_ALIGNMENT): New.
27087         (DATA_ALIGNMENT): Update to use AARCH64_EXPAND_ALIGNMENT.
27088         (LOCAL_ALIGNMENT): Update to use AARCH64_EXPAND_ALIGNMENT.
27090 2017-06-28  Sebastian Peryt  <sebastian.peryt@intel.com>
27092         * config/i386/avx512vlintrin.h (_mm256_permutexvar_epi64)
27093         (_mm256_permutexvar_epi32, _mm256_permutex_epi64): New intrinsics.
27095 2017-06-28  Szabolcs Nagy  <szabolcs.nagy@arm.com>
27097         * config.gcc (*-linux-musl*): Add t-musl tmake_file.
27098         (*-linux-uclibc*): Add t-uclibc tmake_file.
27099         * config/t-musl: New.
27100         * config/t-uclibc: New.
27102 2017-06-28  Richard Earnshaw  <rearnsha@arm.com>
27104         * config/arm/parsecpu.awk (profile): Parse new keyword in an arch
27105         context.
27106         (gen_comm_data): Emit architectural setting of arch_prof.
27107         * config/arm/arm-cpus.in (armv6-m, armv6s-m, armv7-a, armv7ve): Set the
27108         profile.
27109         (armv7-r, armv7-m, armv7e-m, armv8-a, armv8.1-a, armv8.2-a): Likewise.
27110         (armv8-m.base, armv8-m.main): Likewise.
27111         * arm-protos.h (arm_build_target): Add profile field.
27112         (arch_option): Likewise.
27113         * config/arm/arm.c (arm_configure_build_target): Copy the profile to
27114         the active target.
27115         * config/arm/arm.h (TARGET_ARM_ARCH_PROFILE): Use
27116         arm_active_target.profile.
27118 2017-06-28  Richard Biener  <rguenther@suse.de>
27120         PR middle-end/81227
27121         * fold-const.c (negate_expr_p): Use TYPE_UNSIGNED, not
27122         TYPE_OVERFLOW_WRAPS.
27123         * match.pd (negate_expr_p): Likewise.
27124         * tree-ssa-reassoc.c (optimize_range_tests_diff): Use
27125         fold_build2, not fold_binary.
27127 2017-06-28  Wilco Dijkstra  <wdijkstr@arm.com>
27129         * config/aarch64/aarch64 (aarch64_expand_mov_immediate):
27130         Convert memory address to Pmode.
27131         (aarch64_print_operand): Assert MEM operands are always Pmode.
27133 2017-06-28  Wilco Dijkstra  <wdijkstr@arm.com>
27135         PR target/79665
27136         * config/arm/aarch-common.c (arm_no_early_alu_shift_dep):
27137         Remove redundant if.
27138         (aarch_forward_to_shift_is_not_shifted_reg): Remove.
27139         * config/arm/aarch-common-protos.h
27140         (aarch_forward_to_shift_is_not_shifted_re): Remove.
27141         * config/arm/cortex-a53.md: Use arm_no_early_alu_shift_dep in bypass.
27143 2017-06-28  Michael Meissner  <meissner@linux.vnet.ibm.com>
27145         PR ipa/81238
27146         * multiple_target.c (create_dispatcher_calls): Set the default
27147         clone to be static, not public.
27149 2017-06-28  Richard Biener  <rguenther@suse.de>
27151         * tree-vect-loop.c (vectorizable_reduction): Move special
27152         cond reduction IV var creation ...
27153         (vect_create_epilog_for_reduction): ... here.  Remove induction_index
27154         parameter.  Use STMT_VINFO_VECTYPE.
27155         * tree-vect-slp.c (vect_get_constant_vectors): Properly reset
27156         constant_p.
27158 2017-06-28  Martin Liska  <mliska@suse.cz>
27160         PR ipa/81128
27161         * ipa-visibility.c (non_local_p): Handle visibility.
27163 2017-06-28  Martin Liska  <mliska@suse.cz>
27165         PR driver/79659
27166         * common.opt: Add IntegerRange to various options.
27167         * opt-functions.awk (integer_range_info): New function.
27168         * optc-gen.awk: Add integer_range_info to cl_options struct.
27169         * opts-common.c (decode_cmdline_option): Handle
27170         CL_ERR_INT_RANGE_ARG.
27171         (cmdline_handle_error): Likewise.
27172         * opts.c (print_filtered_help): Show valid interval in
27173         when --help is provided.
27174         * opts.h (struct cl_option): Add range_min and range_max fields.
27175         * config/i386/i386.opt: Add IntegerRange for -mbranch-cost.
27177 2017-06-28  Marc Glisse  <marc.glisse@inria.fr>
27179         * match.pd ((X & ~Y) | (~X & Y)): Generalize to + and ^.
27180         (x * C EQ/NE y * C): New transformation.
27182 2017-06-28  Christophe Lyon  <christophe.lyon@linaro.org>
27184         * genmultilib (combination_space): Accept '+' in option names.
27186 2017-06-28  Martin Liska  <mliska@suse.cz>
27188         PR sanitizer/81224
27189         * asan.c (instrument_derefs): Bail out inner references
27190         that are hard register variables.
27192 2017-06-28  Jakub Jelinek  <jakub@redhat.com>
27194         PR target/81175
27195         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use def_builtin
27196         rather than def_builtin_pure for __builtin_ia32_gatherpf*.
27198 2017-06-28  Richard Biener  <rguenther@suse.de>
27200         * tree-vectorizer.h (vect_get_vec_defs): Remove.
27201         (vect_get_slp_defs): Adjust.
27202         * tree-vect-loop.c (get_initial_defs_for_reduction): Split
27203         out from ...
27204         * tree-vect-slp.c (vect_get_constant_vectors): ... here and
27205         simplify.
27206         * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
27207         get_initial_defs_for_reduction instead of vect_get_vec_defs.
27208         (vectorizable_reduction): Adjust.
27209         * tree-vect-slp.c (vect_get_constant_vectors): Remove reduction
27210         handling.
27211         (vect_get_slp_defs): Likewise.
27212         * tree-vect-stmts.c (vect_get_vec_defs): Make static and adjust.
27213         (vectorizable_bswap): Adjust.
27214         (vectorizable_call): Likewise.
27215         (vectorizable_conversion): Likewise.
27216         (vectorizable_assignment): Likewise.
27217         (vectorizable_shift): Likewise.
27218         (vectorizable_operation): Likewise.
27219         (vectorizable_store): Likewise.
27220         (vectorizable_condition): Likewise.
27221         (vectorizable_comparison): Likewise.
27223 2017-06-28  Michael Collison  <michael.collison@arm.com>
27225         PR target/68535
27226         * config/arm/arm.c (gen_ldm_seq): Remove last unnecessary
27227         set of base_reg
27228         (arm_gen_movmemqi): Removed unused variable 'i'.
27229         Convert 'for' loop into 'while' loop.
27230         (arm_expand_prologue): Remove last unnecessary set of insn.
27231         (thumb_pop): Remove unused variable 'pushed_words'.
27232         (thumb_exit): Remove last unnecessary set of regs_to_pop.
27234 2017-06-28  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
27236         * config/s390/predicates.md: Use s390_rel_address_ok_p.
27237         * config/s390/s390-protos.h: Add prototype of
27238         s390_rel_address_ok_p.
27239         * config/s390/s390.c (s390_got_symbol): New function.
27240         (s390_rel_address_ok_p): New function.
27241         (legitimize_pic_address): Use s390_rel_address_ok_p.
27242         (s390_load_got): Use s390_got_symbol.
27243         (s390_option_override): Issue error if
27244         -mno-pic-data-is-text-relative is used without -fpic/-fPIC.
27245         * config/s390/s390.h (TARGET_DEFAULT_PIC_DATA_IS_TEXT_RELATIVE):
27246         New macro.
27247         * config/s390/s390.opt: New option mpic-data-is-text-relative.
27249 2017-06-27  Andrew Pinski  <apinski@cavium.com>
27251         * match.pd (X >/>=/</<= 0 ? 1.0 : -1.0): New patterns.
27252         (X * copysign (1.0, X)): New pattern.
27253         (X * copysign (1.0, -X)): New pattern.
27254         (copysign (-1.0, CST)): New pattern.
27256 2017-06-27  Joseph Myers  <joseph@codesourcery.com>
27258         * genmultilib (combination_space): Remove variable.
27259         Validate reuse rules against regular expression for any sequence
27260         of multilib options in any order.
27262 2017-06-27  Michael Collison  <michael.collison@arm.com>
27264         * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Directly
27265         call aarch64_split_simd_combine.
27266         * (aarch64_combine_internal<mode>): Delete pattern.
27267         * config/aarch64/aarch64.c (aarch64_split_simd_combine):
27268         Allow register and subreg operands.
27270 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27272         * config/i386/vxworks.h (ASM_SPEC): Remove definition. No target
27273         specific need, just fallback on defaults.
27274         (ASM_OUTPUT_ALIGNED_BSS): Add #undef before #define.
27276 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27277             Olivier Hainque  <hainque@adacore.com>
27279         * config/i386/vxworks.h (DBX_REGISTER_NUMBER): Pick distinct
27280         map for 64bits.
27281         (TARGET_OS_CPP_BUILTINS): builtin_define CPU to X86_64 for 64bit
27282         targets. Pick a default if no particular attempt applied.
27283         (STACK_CHECK_PROTECT): Double for 64bit targets, which have
27284         larger contexts.
27286 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27288         * config.gcc (i*86-wrs-vxworks7): Handle new acceptable triplet.
27289         (x86_64-wrs-vxworks7): Likewise.
27291 2017-06-27  Marek Polacek  <polacek@redhat.com>
27293         PR sanitizer/81223
27294         * ubsan.c (instrument_null): Check get_base_address's result for null.
27296 2017-06-27  Marc Glisse  <marc.glisse@inria.fr>
27298         * match.pd ((A+-B)+(C-A), (A+B)-(A-C)): New transformations.
27300 2017-06-27  Marc Glisse  <marc.glisse@inria.fr>
27302         * builtin-types.def (BT_FENV_T_PTR, BT_CONST_FENV_T_PTR,
27303         BT_FEXCEPT_T_PTR, BT_CONST_FEXCEPT_T_PTR): New primitive types.
27304         (BT_FN_INT_FENV_T_PTR, BT_FN_INT_CONST_FENV_T_PTR,
27305         BT_FN_INT_FEXCEPT_T_PTR_INT, BT_FN_INT_CONST_FEXCEPT_T_PTR_INT):
27306         New function types.
27307         * builtins.def (BUILT_IN_FECLEAREXCEPT, BUILT_IN_FEGETENV,
27308         BUILT_IN_FEGETEXCEPTFLAG, BUILT_IN_FEGETROUND,
27309         BUILT_IN_FEHOLDEXCEPT, BUILT_IN_FERAISEEXCEPT,
27310         BUILT_IN_FESETENV, BUILT_IN_FESETEXCEPTFLAG,
27311         BUILT_IN_FESETROUND, BUILT_IN_FETESTEXCEPT,
27312         BUILT_IN_FEUPDATEENV): New builtins.
27313         * tree-core.h (TI_FENV_T_PTR_TYPE, TI_CONST_FENV_T_PTR_TYPE,
27314         TI_FEXCEPT_T_PTR_TYPE, TI_CONST_FEXCEPT_T_PTR_TYPE): New entries.
27315         * tree.h (fenv_t_ptr_type_node, const_fenv_t_ptr_type_node,
27316         fexcept_t_ptr_type_node, const_fexcept_t_ptr_type_node): New
27317         macros.
27318         (builtin_structptr_types): Adjust size.
27319         * tree.c (builtin_structptr_types): Add four entries.
27321 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27322             Olivier Hainque  <hainque@adacore.com>
27324         * config/vxworks.h (VXWORKS_LIB_SPEC): Incorporate ...
27325         (TLS_SYM): New local macro, forcing reference to __tls__ on
27326         link command lines for VxWorks 7 RTPs, triggering initialization
27327         of tlsLib.
27328         (VXWORKS_HAVE_TLS): New macro. State whether the target VxWorks
27329         OS features TLS support, true for RTPs on VxWorks 7.
27330         * config/vxworks.c (vxworks_override_options): Setup emutls
27331         accordingly.
27333 2017-06-27  Jakub Jelinek  <jakub@redhat.com>
27335         * predict.c (test_prediction_value_range): Use -1U instead of -1
27336         to avoid narrowing conversion warning.
27337         * dumpfile.c (dump_options): Wrap all value into dump_flags_t cast
27338         to avoid narrowing conversion warning.
27339         * opt-functions.awk (var_ref): Return (unsigned short) -1 instead of
27340         -1.
27341         * optc-gen.awk (END): Expect (unsigned short) -1 instead of -1.
27343 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27345         * config/vxworks.h (VXWORKS_LIBS_RTP): Alternative definition for
27346         64bit configurations.
27347         (PTR_DIFF_TYPE): Alternative definition for TARGET_LP64.
27348         (SIZE_TYPE): Likewise.
27349         * config/vxworks.c (vxworks_emutls_var_fields): Use
27350         long_unsigned_type_node instead of unsigned_type_node as the offset
27351         field type, which is "pointer" mode in emutls.c.
27353 2017-06-27  Jakub Jelinek  <jakub@redhat.com>
27355         PR sanitizer/81209
27356         * ubsan.c (ubsan_encode_value): Initialize DECL_CONTEXT on var.
27358         PR middle-end/81207
27359         * gimple-fold.c (replace_call_with_call_and_fold): Handle
27360         gimple_vuse copying separately from gimple_vdef copying.
27362 2017-06-27  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
27364         * value-prof.c (free_hist): Remove call to memset and the enclosing if
27365         condition.
27367 2017-06-26  Jerome Lambourg  <lambourg@adacore.com>
27368             Olivier Hainque  <hainque@adacore.com>
27370         * config.gcc (*-*-vxworks*): Add TARGET_VXWORKS7=1 to tm_defines
27371         for all vxworks7 targets.
27372         * config/vxworks.h (TARGET_VXWORKS7): If not defined, define to 0.
27373         (VXWORKS_ADDITIONAL_CPP_SPEC): Alternative definition for VXWORKS7.
27374         (VXWORKS_LIBS_RTP, VXWORKS_LIBS_RTP_DIR): New macros, allowing
27375         variations for VX6/VX7 and 32/64bits later on in ...
27376         (VXWORKS_LIB_SPEC): Leverage new macros.
27377         (VXWORKS_OS_CPP_BUILTINS): Define _VSB_CONFIG_FILE for VXWORKS7,
27378         as well as _ALLOW_KEYWORD_MACROS when "inline" is not a keyword.
27380 2017-06-26  Jerome Lambourg  <lambourg@adacore.com>
27382         * config/vxworks.h (VXWORKS_OS_CPP_BUILTINS): builtin_define
27383         _VX_TOOL_FAMILY and _VX_TOOL to gnu.
27385 2017-06-26  Carl Love  <cel@us.ibm.com>
27387         * config/rs6000/rs6000-c.c: Add support for built-in functions
27388         vector bool char vec_reve (vector bool char);
27389         vector signed char vec_reve (vector signed char);
27390         vector unsigned char vec_reve (vector unsigned char);
27391         vector bool int vec_reve (vector bool int);
27392         vector signed int vec_reve (vector signed int);
27393         vector unsigned int vec_reve (vector unsigned int);
27394         vector bool long long vec_reve (vector bool long long);
27395         vector signed long long vec_reve (vector signed long long);
27396         vector unsigned long long vec_reve (vector unsigned long long);
27397         vector bool short vec_reve (vector bool short);
27398         vector signed short vec_reve (vector signed short);
27399         vector double vec_reve (vector double);
27400         vector float vec_reve (vector float);
27401         * config/rs6000/rs6000-builtin.def (VREVE_V2DI, VREVE_V4SI,
27402         VREVE_V8HI, VREVE_V16QI, VREVE_V2DF, VREVE_V4SF, VREVE): New builtin.
27403         * config/rs6000/altivec.md (UNSPEC_VREVEV): New UNSPEC.
27404         (altivec_vreve): New pattern.
27405         * config/rs6000/altivec.h (vec_reve): New define.
27406         * doc/extend.texi (vec_rev): Update the built-in documentation file
27407         for the new built-in functions.
27409 2016-06-26  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27411         PR tree-optimization/71815
27412         * gimple-ssa-strength-reduction.c (uses_consumed_by_stmt): New
27413         function.
27414         (find_basis_for_candidate): Call uses_consumed_by_stmt rather than
27415         has_single_use.
27416         (slsr_process_phi): Likewise.
27417         (replace_uncond_cands_and_profitable_phis): Don't replace a
27418         multiply candidate with a stride of 1 (copy or cast).
27419         (phi_incr_cost): Call uses_consumed_by_stmt rather than
27420         has_single_use.
27421         (lowest_cost_path): Likewise.
27422         (total_savings): Likewise.
27424 2017-06-26  Richard Biener  <rguenther@suse.de>
27426         PR target/81175
27427         * config/i386/i386.c (ix86_init_mmx_sse_builtins):
27428         Use def_builtin_pure for all gather builtins.
27430 2017-06-26  Richard Biener  <rguenther@suse.de>
27432         PR tree-optimization/81203
27433         * tree-tailcall.c (find_tail_calls): Do not move stmts into
27434         non-dominating BBs.
27436 2017-06-26  Marek Polacek  <polacek@redhat.com>
27438         PR c/80116
27439         * doc/invoke.texi: Document -Wmultistatement-macros.
27441 2017-06-26  Christophe Lyon  <christophe.lyon@linaro.org>
27443         * doc/sourcebuild.texi (ARM-specific attributes): Document new
27444         arm_neon_ok_no_float_abi effective target.
27446 2017-06-26  Richard Biener  <rguenther@suse.de>
27448         PR tree-optimization/80928
27449         * cfghooks.c (duplicate_block): Do not copy BB_DUPLICATED flag.
27450         (copy_bbs): Set BB_DUPLICATED flag early.
27451         (execute_on_growing_pred): Do not execute for BB_DUPLICATED
27452         marked blocks.
27453         (execute_on_shrinking_pred): Likewise.
27454         * tree-ssa.c (ssa_redirect_edge): Do not look for PHI args in
27455         BB_DUPLICATED blocks.
27456         * tree-ssa-phionlycoprop.c (eliminate_degenerate_phis_1): Properly
27457         iterate over all PHIs considering removal of *gsi.
27459 2017-06-23  Jim Wilson  <jim.wilson@linaro.org>
27461         * doc/invoke.texi (AArch64 Options, -mtune): Re-add falkor and
27462         qdf24xx.
27464 2017-06-23  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
27466         * config/rs6000/rs6000-string.c: (expand_block_clear,
27467         do_load_for_compare, select_block_compare_mode,
27468         compute_current_alignment, expand_block_compare,
27469         expand_strncmp_align_check, expand_strn_compare,
27470         expand_block_move, rs6000_output_load_multiple)
27471         Move functions related to string/block move/compare
27472         to a separate file.
27473         * config/rs6000/rs6000.c: Move above functions to rs6000-string.c.
27474         * config/rs6000/rs6000-protos.h (rs6000_emit_dot_insn): Add prototype
27475         for this function which is now used in two files.
27476         * config/rs6000/t-rs6000: Add rule to compile rs6000-string.o.
27477         * config.gcc: Add rs6000-string.o to extra_objs for
27478         targets powerpc*-*-* and rs6000*-*-*.
27480 2017-06-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
27482         PR target/80510
27483         * config/rs6000/rs6000.md (ALTIVEC_DFORM): Do not allow DImode in
27484         32-bit, since indexed is not valid for DImode.
27485         (mov<mode>_hardfloat32): Reorder ISA 2.07 load/stores before ISA
27486         3.0 d-form load/stores to be the same as mov<mode>_hardfloat64.
27487         (define_peephole2 for Altivec d-form load): Add 32-bit support.
27488         (define_peephole2 for Altivec d-form store): Likewise.
27490         PR ipa/81185
27491         * multiple_target.c (create_dispatcher_calls): Only create the
27492         dispatcher call if the function is the default clone of a
27493         versioned function.
27495 2017-06-23  Segher Boessenkool  <segher@kernel.crashing.org>
27497         PR middle-end/80902
27498         * builtins.c (expand_builtin_atomic_fetch_op): If emitting code after
27499         a call, force the call to not be a tail call.
27501 2017-06-23  Jeff Law  <law@redhat.com>
27503         * doc/contrib.texi: Add entry for Steven Pemberton's work on
27504         enquire.
27506 2017-06-23  Will Schmidt  <will_schmidt@vnet.ibm.com>
27508         * config/rs6000/rs6000.c: Add include of ssa-propagate.h for
27509         update_call_from_tree().  (rs6000_gimple_fold_builtin): Add
27510         handling for early expansion of vector shifts (sl,sr,sra,rl).
27511         (builtin_function_type): Add vector shift right instructions
27512         to the unsigned argument list.
27514 2017-06-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
27516         rtl-optimizatoin/79286
27517         * ira.c (update_equiv_regs): Revert to using may_trap_or_fault_p again.
27518         * rtlanal.c (rtx_addr_can_trap_p_1): SYMBOL_REF_FUNCTION_P can never
27519         trap.  PIC register plus a const unspec without offset can never trap.
27521 2017-06-23  Marc Glisse  <marc.glisse@inria.fr>
27523         * tree.h (builtin_structptr_type): New type.
27524         (builtin_structptr_types): Declare new array.
27525         * tree.c (builtin_structptr_types): New array.
27526         (free_lang_data, build_common_tree_nodes): Use it.
27528 2017-06-23  Jonathan Wakely  <jwakely@redhat.com>
27530         PR c++/81187
27531         * doc/invoke.texi (-Wnoexcept-type): Fix name of option, from
27532         -Wnoexcept.
27534 2017-06-22  Matt Turner  <mattst88@gmail.com>
27536         * config/i386/driver-i386.c (host_detect_local_cpu): Add Kaby
27537         Lake models to skylake case.  Assume skylake for unknown
27538         models with clflushopt.
27540 2017-06-22  Jeff Law  <law@redhat.com>
27542         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Handle
27543         frame sizes that do not satisfy aarch64_uimm12_shift.
27545 2017-06-22  Jan Hubicka <hubicka@ucw.cz>
27547         * profile-count.h (apply_probability,
27548         apply_scale, probability_in): Fix checks for zero.
27550 2017-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27552         * incpath.c (add_sysroot_to_chain): Allow for $SYSROOT prefix.
27553         * doc/cppdiropts.texi (-I @var{dir}): Document it.
27555 2016-06-22  Richard Biener  <rguenther@suse.de>
27557         * tree-vect-loop.c (vect_model_reduction_cost): Handle
27558         COND_REDUCTION and INTEGER_INDUC_COND_REDUCTION without
27559         REDUC_MAX_EXPR support.
27560         (vectorizable_reduction): Likewise.
27561         (vect_create_epilog_for_reduction): Likewise.
27563 2017-06-22  James Greenhalgh  <james.greenhalgh@arm.com>
27565         * match.pd (A / (1 << B) -> A >> B): New.
27566         * generic-match-head.c: Include optabs-tree.h.
27567         * gimple-match-head.c: Likewise.
27568         * optabs-tree.h (target_supports_op_p): New.
27569         * optabs-tree.c (target_supports_op_p): New.
27571 2017-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27573         * configure.ac (gcc_cv_ld_static_dynamic): Also check stderr for
27574         $gcc_cv_ld --help output.
27575         (gcc_cv_ld_demangle): Likewise.
27576         (gcc_cv_ld_eh_frame_hdr): Likewise.
27577         (gcc_cv_ld_pie): Likewise.
27578         (gcc_cv_ld_as_needed): Likewise.  Prefer native forms unless $gnu_ld.
27579         (gcc_cv_ld_buildid): Likewise.
27580         (gcc_cv_ld_sysroot): Likewise.
27581         (ld_bndplt_support): Likewise.
27582         (ld_pushpopstate_support): Likewise.
27583         * configure: Regenerate.
27584         * config/sol2.h [!USE_GLD] (SYSROOT_SPEC): Define.
27586 2017-06-21  Jakub Jelinek  <jakub@redhat.com>
27588         PR target/81151
27589         * config/i386/sse.md (round<mode>2): Renumber match_dup and
27590         operands indexes to avoid gap between operands and match_dups.
27592 2017-06-21  Andrew Pinski  <apinski@cavium.com>
27594         * config/aarch64/aarch64-cost-tables.h (thunderx_extra_costs):
27595         Increment Arith_shift and Arith_shift_reg by 1.
27596         * config/aarch64/aarch64-tuning-flags.def (cheap_shift_extend):
27597         New tuning flag.
27598         * config/aarch64/aarch64.c (thunderx_tunings): Enable
27599         AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
27600         (aarch64_strip_extend): Add new argument and test for it.
27601         (aarch64_cheap_mult_shift_p): New function.
27602         (aarch64_rtx_mult_cost): Call aarch64_cheap_mult_shift_p and don't
27603         add a cost if it is true.
27604         Update calls to aarch64_strip_extend.
27605         (aarch64_rtx_costs): Update calls to aarch64_strip_extend.
27607 2017-06-21  Andrew Pinski  <apinski@cavium.com>
27609         * config/aarch64/aarch64-cores.def (thunderxt88p1): Use thunderxt88
27610         tunings.
27611         (thunderxt88): Likewise.
27612         * config/aarch64/aarch64.c (thunderxt88_prefetch_tune): New variable.
27613         (thunderx_prefetch_tune): New variable.
27614         (thunderx2t99_prefetch_tune): Update for the correct values.
27615         (thunderxt88_tunings): New variable.
27616         (thunderx_tunings): Use thunderx_prefetch_tune instead of
27617         generic_prefetch_tune.
27618         (thunderx2t99_tunings): Use AUTOPREFETCHER_WEAK.
27620 2017-06-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27622         * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>_lse,
27623         SHORT): Relax operand 3 to aarch64_reg_or_zero and constraint to Z.
27624         (aarch64_compare_and_swap<mode>_lse, GPI): Likewise.
27625         (aarch64_atomic_cas<mode>, SHORT): Likewise for operand 2.
27626         (aarch64_atomic_cas<mode>, GPI): Likewise.
27628 2017-06-21  Martin Liska  <mliska@suse.cz>
27630         * gimplify.c (gimplify_label_expr): Insert GIMPLE_PREDICT
27631         statements on cold and hot labels.
27632         * predict.c (tree_estimate_probability_bb): Remove the
27633         prediction from this place.
27635 2017-06-21  Martin Liska  <mliska@suse.cz>
27637         PR tree-optimization/79489
27638         * gimplify.c (maybe_add_early_return_predict_stmt): New
27639         function.
27640         (gimplify_return_expr): Call the function.
27641         * predict.c (tree_estimate_probability_bb): Remove handling
27642         of early return.
27643         * predict.def: Update comment about early return predictor.
27644         * gimple-predict.h (is_gimple_predict): New function.
27645         * predict.def: Change default value of early return to 66.
27646         * tree-tailcall.c (find_tail_calls): Skip GIMPLE_PREDICT
27647         statements.
27648         * passes.def: Put pass_strip_predict_hints to the beginning of
27649         IPA passes.
27651 2017-06-21  Pierre-Marie de Rodat  <derodat@adacore.com>
27653         * dwarf2out.c (gen_decl_die): Remove the guard to skip file-scope
27654         FUNCTION_DECL declarations.
27655         (dwarf2out_early_global_decl): Remove the guard to skip FUNCTION_DECL
27656         declarations.
27657         (dwaf2out_decl): Likewise.
27658         * godump.c (go_early_global_decl): Skip call to the real debug hook
27659         for FUNCTION_DECL declarations.
27660         * passes.c (rest_of_decl_compilation): Skip call to the
27661         early_global_decl debug hook for FUNCTION_DECL declarations, unless
27662         -fdump-go-spec is passed.
27664 2017-06-21  Marc Glisse  <marc.glisse@inria.fr>
27666         * config/i386/i386.c (struct builtin_isa): New field pure_p.
27667         Reorder for compactness.
27668         (def_builtin, def_builtin2, ix86_add_new_builtins): Handle pure_p.
27669         (def_builtin_pure, def_builtin_pure2): New functions.
27670         (ix86_init_mmx_sse_builtins) [__builtin_ia32_stmxcsr]: Mark as pure.
27672 2017-06-21  Marc Glisse  <marc.glisse@inria.fr>
27674         * match.pd (nop_convert): New predicate.
27675         ((A +- CST1) +- CST2): Allow some NOP conversions.
27677 2017-06-21  Jakub Jelinek  <jakub@redhat.com>
27679         PR c++/81130
27680         * gimplify.c (omp_add_variable): Don't force GOVD_SEEN for types
27681         with ctors/dtors if GOVD_SHARED is set.
27683 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27685         * config/aarch64/aarch64.md (movti_aarch64):
27686         Emit mov rather than orr.
27687         (movtf_aarch64): Likewise.
27688         * config/aarch64/aarch64-simd.md (aarch64_simd_mov):
27689         Emit mov rather than orr.
27691 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27693         * config/aarch64/aarch64-simd.md (aarch64_simd_dup):
27694         Swap alternatives, make integer dup more expensive.
27696 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27698         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
27699         Return true for non-tls symbols.
27701 2017-06-21  James Greenhalgh  <james.greenhalgh@arm.com>
27703         * config/aarch64/aarch64-cores.def (cortex-a55): New.
27704         (cortex-a75): Likewise.
27705         (cortex-a75.cortex-a55): Likewise.
27706         * config/aarch64/aarch64-tune.md: Regenerate.
27707         * doc/invoke.texi (-mtune): Document new values for -mtune.
27709 2017-06-21  Tom de Vries  <tom@codesourcery.com>
27711         * doc/sourcebuild.texi (Add Options, Features for dg-add-options): Add
27712         stack_size feature.
27713         (Effective-Target Keywords, Other attributes): Suggest using
27714         dg-add-options stack_size feature to get stack limit in stack_size
27715         effective target documentation.
27717 2017-06-21  Julian Brown  <julian@codesourcery.com>
27718             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
27720         * config/aarch64/aarch64-simd.md (aarch64_crypto_pmulldi)
27721         (aarch64_crypto_pmullv2di): Change type attribute to crypto_pmull.
27722         * config/aarch64/thunderx2t99.md (thunderx2t99_pmull): New
27723         reservation.
27724         * config/arm/cortex-a53.md (cortex_a53_advsimd_type): Add crypto_pmull to
27725         attribute type list for neon_multiply.
27726         * config/arm/cortex-a57.md (cortex_a57_neon_type): Add crypto_pmull to
27727         attribute type list for neon_multiply.
27728         * config/arm/crypto.md (crypto_vmullp64): Change type to crypto_pmull.
27729         * config/arm/exynos-m1.md (exynos_m1_neon_type): Add crypto_pmull to
27730         attribute type list for neon_multiply.
27731         * config/arm/types.md (crypto_pmull): Add.
27732         * config/arm/xgene1.md (xgene1_neon_pmull): Add crypto_pmull to
27733         attribute type list.
27735 2017-06-20  Andreas Tobler  <andreast@gcc.gnu.org>
27737         * config.gcc (armv6*-*-freebsd*): Change the target_cpu_cname to
27738         arm1176jzf-s.
27740 2017-06-20  Jakub Jelinek  <jakub@redhat.com>
27742         * ira-costs.c (find_costs_and_classes): Initialize cost_classes later
27743         to make sure not to dereference a NULL cost_classes_ptr pointer.
27745 2017-06-20  Carl Love  <cel@us.ibm.com>
27747         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
27748         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
27749         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
27750         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
27751         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
27752         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
27753         VMULOSW): New enum "unspec" values.
27754         (vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
27755         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si,
27756         altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
27757         altivec_vmulosw): New patterns.
27758         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
27759         VMULOSW): Add definitions.
27761 2017-06-20  Julia Koval  <julia.koval@intel.com>
27763         * config/i386/i386.c: Fix rounding expand for new pattern.
27764         * config/i386/subst.md: Fix pattern (parallel -> unspec).
27766 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27768         * config/aarch64/aarch64-option-extensions.def (rcpc): New.
27769         * config/aarch64/aarch64.h (AARCH64_FL_RCPC): New.
27771 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27773         * config/aarch64/aarch64-option-extensions.def (fp16): Fix expected
27774         feature string.
27776 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27778         * config/aarch64/aarch64-cores.def: Rearrange to sort by
27779         architecture, then by implementer ID.
27780         * config/aarch64/aarch64-tune.md: Regenerate.
27782 2017-06-20  Richard Biener  <rguenther@suse.de>
27784         PR middle-end/81097
27785         * fold-const.c (split_tree): Fold to type before negating.
27787 2017-06-20  David Malcolm  <dmalcolm@redhat.com>
27789         * diagnostic-show-locus.c
27790         (selftest::test_fixit_deletion_affecting_newline): New function.
27791         (selftest::diagnostic_show_locus_c_tests): Call it.
27793 2017-06-20  Andreas Schwab  <schwab@suse.de>
27795         PR target/80970
27796         * config/m68k/m68k.md (bsetdreg, bchgdreg, bclrdreg): Use "=d"
27797         instead of "+d".
27799 2017-06-20  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
27801         * config/arm/arm-c.c (arm_cpu_builtins): New block to define
27802         __ARM_FEATURE_COPROC according to support.
27804 2017-06-20  Jakub Jelinek  <jakub@redhat.com>
27806         * tree-chkp.c (chkp_get_hard_register_var_fake_base_address):
27807         Rewritten to avoid overflow for > 32-bit pointers.
27809         PR sanitizer/81125
27810         * ubsan.h (ubsan_encode_value): Workaround buggy clang++ parser
27811         by removing enum keyword.
27812         (ubsan_type_descriptor): Likewise.  Formatting fix.
27814         PR target/81121
27815         * config/i386/i386.md (TARGET_USE_VECTOR_CONVERTS float si->{sf,df}
27816         splitter): Require TARGET_SSE2 in the condition.
27818 2017-06-20  Michael Meissner  <meissner@linux.vnet.ibm.com>
27820         PR target/79799
27821         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Add support
27822         for doing vector set of SFmode on ISA 3.0.
27823         * config/rs6000/vsx.md (vsx_set_v4sf_p9): Likewise.
27824         (vsx_set_v4sf_p9_zero): Special case setting 0.0f to a V4SF
27825         element.
27826         (vsx_insert_extract_v4sf_p9): Add an optimization for inserting a
27827         SFmode value into a V4SF variable that was extracted from another
27828         V4SF variable without converting the element to double precision
27829         and back to single precision vector format.
27830         (vsx_insert_extract_v4sf_p9_2): Likewise.
27832 2017-06-19  Jakub Jelinek  <jakub@redhat.com>
27834         * tree-ssa-structalias.c (get_constraint_for_ptr_offset): Multiply
27835         in UWHI to avoid undefined overflow.
27837         PR sanitizer/81125
27838         * ubsan.h (enum ubsan_encode_value_phase): New.
27839         (ubsan_encode_value): Change second argument to
27840         enum ubsan_encode_value_phase with default value of
27841         UBSAN_ENCODE_VALUE_GENERIC.
27842         * ubsan.c (ubsan_encode_value): Change second argument to
27843         enum ubsan_encode_value_phase PHASE from bool IN_EXPAND_P,
27844         adjust uses, for UBSAN_ENCODE_VALUE_GENERIC use just
27845         create_tmp_var_raw instead of create_tmp_var and use a
27846         TARGET_EXPR.
27847         (ubsan_expand_bounds_ifn, ubsan_build_overflow_builtin,
27848         instrument_bool_enum_load, ubsan_instrument_float_cast): Adjust
27849         ubsan_encode_value callers.
27851         PR sanitizer/81111
27852         * ubsan.c (ubsan_encode_value): If current_function_decl is NULL,
27853         use create_tmp_var_raw instead of create_tmp_var, mark it addressable
27854         just by setting TREE_ADDRESSABLE on the result and use a TARGET_EXPR.
27856 2017-06-19  Richard Biener  <rguenther@suse.de>
27858         PR middle-end/81118
27859         * tree-cfgcleanup.c (cleanup_tree_cfg_noloop): Clear niter
27860         estimates if we changed anything.
27862 2017-06-19  Richard Biener  <rguenther@suse.de>
27864         PR tree-optimization/80887
27865         * tree-ssa-sccvn.c (mprts_hook_cnt): New global.
27866         (vn_lookup_simplify_result): Allow only mprts_hook_cnt succesful
27867         simplified lookups, then reset mprts_hook.
27868         (vn_nary_build_or_lookup_1): Set mprts_hook_cnt to 9 before
27869         simplifying.
27870         (try_to_simplify): Likewise.
27872 2017-06-19  Martin Liska  <mliska@suse.cz>
27874         PR sanitizer/80879
27875         * gimplify.c (gimplify_switch_expr):
27876         Initialize live_switch_vars for SWITCH_BODY == STATEMENT_LIST.
27878 2017-06-19  Martin Liska  <mliska@suse.cz>
27880         * doc/install.texi: Document that PGO runs in 4 stages.
27882 2017-06-19  Martin Liska  <mliska@suse.cz>
27884         PR ipa/80732
27885         * attribs.c (make_dispatcher_decl): Do not append '.ifunc'
27886         to dispatcher function name.
27887         * multiple_target.c (replace_function_decl): New function.
27888         (create_dispatcher_calls): Redirect both edges and references.
27890 2017-06-19  Jan Hubicka <hubicka@ucw.cz>
27892         * profile-count.c (profile_count::dump): Dump quality.
27893         (profile_count::differs_from_p): Update for unsigned val.
27894         * profile-count.h (profile_count_quality): New enum.
27895         (profile_count): Turn m_val to 62bit unsigned, add quality tracking.
27897 2017-06-19  Richard Biener  <rguenther@suse.de>
27899         * tree-ssa-loop-niter.h (estimate_numbers_of_iterations): Take
27900         struct function as arg.
27901         (estimate_numbers_of_iterations): Export overload with loop arg.
27902         (free_numbers_of_iterations_estimates_loop): Use an overload of
27903         free_numbers_of_iterations_estimates instead.
27904         * tree-cfg.c (remove_bb): Adjust.
27905         * tree-cfgcleanup.c (remove_forwarder_block_with_phi): Likewise.
27906         * tree-parloops.c (gen_parallel_loop): Likewise.
27907         * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables):
27908         Likewise.
27909         (tree_unroll_loops_completely): Likewise.
27910         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations_loop):
27911         Use an overload instead and export.
27912         (estimated_loop_iterations): Adjust.
27913         (max_loop_iterations): Likewise.
27914         (likely_max_loop_iterations): Likewise.
27915         (estimate_numbers_of_iterations): Take struct function as arg
27916         and adjust.
27917         (loop_exits_before_overflow): Adjust.
27918         (free_numbers_of_iterations_estimates_loop): Use an overload.
27919         * tree-vect-loop.c (vect_analyze_loop_form): Adjust.
27920         * tree-vectorizer.c (vect_free_loop_info_assumptions): Likewise.
27922 2017-06-19  Richard Biener  <rguenther@suse.de>
27924         PR ipa/81112
27925         * ipa-prop.c (find_constructor_constant_at_offset): Handle
27926         RANGE_EXPR conservatively.
27928 2017-06-16  Carl Love  <cel@us.ibm.com>
27930         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
27931         definitions for vec_float, vec_float2, vec_floato,
27932         vec_floate built-ins.
27933         * config/rs6000/vsx.md (define_c_enum "unspec"): Add RTL code
27934         for instructions vsx_xvcvsxws vsx_xvcvuxwsp, float2, floato and
27935         floate.
27936         * config/rs6000/rs6000-builtin.def (FLOAT2_V2DI, FLOATE_V2DF,
27937         FLOATE_2DI, FLOATO_V2DF, FLOATEE_V2DI, XVCVSXWSP_V4SF,
27938         UNS_FLOATO_V2DI, UNS_FLOATE_V2DI): Add definitions.
27939         * config/altivec.md (define_insn "p8_vmrgew_<mode>",
27940         define_mode_attr VF_sxddp): Add V4SF type to p8_vmrgew.
27941         * config/rs6000/altivec.h (vec_float, vec_float2, vec_floate,
27942         vec_floato): Add builtin defines.
27943         * doc/extend.texi (vec_float, vec_float2, vec_floate, vec_floato):
27944         Update the built-in documentation file for the new built-in
27945         functions.
27947 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27949         * config/arm/arm.opt (marm): Mark as the negative of of -mthumb.
27950         (mthumb): Mark as the negative of -marm.
27952 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27954         * doc/invoke.texi (ARM Options, -mcpu): Document supported
27955         extension options.
27956         (ARM Options, -mtune): Document that this accepts the same
27957         extension options as -mcpu.
27958         (ARM Options, -mfpu): Document addition of -mfpu=auto.
27960 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27962         * doc/invoke.texi (ARM Options, -march=): Document new syntax and
27963         permitted extensions.
27965 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27967         * config/arm/arm-cpus.in (armv7): Add extension +nofp.
27968         (armv7-r): Add aliases vfpv3xd and vfpv3-d16.
27969         (armv8-m.main): Add option +nodsp.
27970         * config/arm/arm-cpu-cdata.h: Regenerated.
27972 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27974         * config/arm/t-fuchsia: New file.
27975         * config.gcc (arm*-*-fuchsia*): Use it.
27977 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27979         * config/arm/t-symbian: Rewrite for new option infrastructure.
27981 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27983         * config/arm/t-phoenix (MULTILIB_REUSE): Clear variable.
27984         (MULTILIB_REQUIRED): Likewise.
27986 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27988         * config/arm/t-linux-eabi (MULTILIB_EXCEPTIONS): Set to empty.
27989         (MULTILIB_RESUE): Likewise.
27990         (MULTILIB_MATCHES): Likewise.
27991         (MULTLIB_REQUIRED): Likewise.
27993 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27995         * config/arm/t-rtems: Rewrite for new option framework.
27997 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27999         * config/arm/t-aprofile (v7_a_nosimd_variants, v7_a_simd_variants)
28000         (v7ve_nosimd_variatns, v7ve_vfpv3_simd_variants)
28001         (v7ve_vfpv4_simd_variants, v8_a_nosimd_variants, v8_a_simd_variants)
28002         (v8_1_a_simd_variants, v8_2_a_simd_variants): Move to ...
28003         * config/arm/t-multilib: ... here.
28004         (MULTILIB_OPTIONS): Add armv7 and armv7+fp architectures.
28005         (MULTILIB_MATCHES): Use armv7 libraries for armv7-r.  Also use for
28006         armv7-a and armv8*-a when A-profile libraries have not been built.
28007         * config/arm/t-rmprofile: Rewrite.
28009 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28011         * genmultilib (multilib_reuse): Allow an explicit period to be escaped
28012         with a backslash.  Remove the backslash after substituting unescaped
28013         periods.
28014         * doc/fragments.texi (MULTILIB_REUSE): Document it.
28016 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28018         * config.gcc: (arm*-*-*): When building a-profile libraries, force
28019         the driver to pass through the default setting of -mfloat-abi.
28020         * common/config/arm/arm-common.c (arm_target_thumb_only): Return -marm
28021         rather than NULL.
28022         * config/arm/t-multilib (MULTILIB_REUSE): Initialize to empty.
28023         (all_feat_combs): New rule.
28024         (MULTILIB_OPTIONS): Use explicit ARM and Thumb directories.  Rework
28025         default libraries.
28026         * config/arm/t-aprofile: Rewrite.
28028 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28030         * config/arm/arm.h (FPUTYPE_AUTO): Define.
28031         * config/arm/arm.c (arm_option_override): Use FPUTYPE_AUTO if the
28032         fpu is not specified by the user/command-line.
28033         * config/arm/bpabi.h (FPUTYPE_DEFAULT): Delete.
28034         * config/arm/netbsd-elf.h (FPUTYPE_DEFAULT): Delete.
28035         * config/arm/linux-elf.h (FPUTYPE_DEFAULT): Delete.
28036         * config/arm/vxworks.h (FPUTYPE_DEFAULT): Delete.
28037         * common/config/arm/arm-common.c (arm_canon_arch_option): Use
28038         FPUTYPE_AUTO insted of FPUTYPE_DEFAULT.
28040 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28042         * config/arm/elf.h (MULTILIB_DEFAULTS): Delete.
28043         * config/arm/t-arm-elf: Rewritten.
28045 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28047         * config/arm/arm.h (TARGET_HARD_FLOAT): Also check that we
28048         have some floating-point instructions.
28049         (TARGET_SOFT_FLOAT): Define as inverse of TARGET_HARD_FLOAT.
28050         (TARGET_MAYBE_HARD_FLOAT): New macro.
28051         * config/arm/arm-builtins.c (arm_init_builtins): Use
28052         TARGET_MAYBE_HARD_FLOAT.
28053         * config/arm/arm.c (arm_option_override): Use TARGET_HARD_FLOAT_ABI.
28055 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28057         * common/config/arm/arm-common.c: Define INCLUDE_LIST.
28058         (configargs.h): Include it.
28059         (arm_print_hint_for_fpu_option): New function.
28060         (arm_parse_fpu_option): New function.
28061         (candidate_extension): New class.
28062         (arm_canon_for_multilib): New function.
28063         * config/arm/arm.h (CANON_ARCH_SPEC_FUNCTION): New macro.
28064         (EXTRA_SPEC_FUNCTIONS): Add CANON_ARCH_SPEC_FUNCTION.
28065         (ARCH_CANONICAL_SPECS): New macro.
28066         (DRIVER_SELF_SPECS): Add ARCH_CANONICAL_SPECS.
28068 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28070         * config.gcc (arm*-*-*): Ensure both target_cpu_cname and with_cpu
28071         are set after handling multilib fragments.  Set target_cpu_default2
28072         from with_cpu.
28074 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28076         * config.gcc (arm*-*-fucshia*): Set target_cpu_cname to the real
28077         cpu name.
28078         (arm*-*-*): Set target_cpu_default2 to a quoted string.
28079         * config/arm/parsecpu.awk (check_cpu): Validate any extension
28080         options.
28081         (check_arch): Likewise.
28082         * config/arm/arm.c (arm_configure_build_target): Handle
28083         TARGET_CPU_DEFAULT being a string constant.  Scan any feature
28084         options in the default.
28086 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28088         * config/arm/arm-protos.h (cpu_arch_extension): Add field to record
28089         when an option is an alias of another.
28090         * config/arm/parsecpu.awk (optalias): New parser token.
28091         (gen_comm_data): Mark non-alias options as such.  Emit entries
28092         for extension aliases.
28093         * config/arm/arm-cpus.in (armv5e): Make vfpv2 an alias.
28094         (armv5te, armv5tej, armv6, armv6j, armv6k, armv6z): Likewise.
28095         (armv6kz, armv6zk, armv6t2): Likewise.
28096         (armv7): Make vfpv3-d16 an alias.
28097         (armv7-a): Make vfpv3-d16, neon and neon-vfpv3 aliases.  Sort in
28098         canonical order.
28099         (armv7ve): Make vfpv4-d16, neon-vfpv3 and neon-vfpv4 aliases.
28100         Sort in canonical order.
28101         (armv8-a): Sort in canonical order.
28102         (armv8.1-a, armv8.2-a):  Likewise.
28103         (generic-armv7-a): Make neon and neon-vfpv3 aliases.  Sort in
28104         canonical order.
28105         (cortex-a9): Sort in canonical order.
28106         * config/arm/arm.c (selftests.h): Include it.
28107         (arm_test_cpu_arch_data): New function.
28108         (arm_run_self_tests): New function.
28109         (TARGET_RUN_TARGET_SELFTESTS): Redefine.
28110         (targetm): Move declaration to the end of the file.
28111         * arm-cpu-cdata.h: Regenerated.
28113 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28115         * config/arm/arm.h (TARGET_MODE_SPECS): Add additional parameter to
28116         call to target_mode_check describing the type of option passed.
28117         * common/config/arm/arm-common.c (arm_arch_core_flag): Delete.
28118         (arm_target_thumb_only): Use arm_parse_arch_option_name or
28119         arm_parse_cpu_option_name to match parameters against list of
28120         available targets.
28121         * config/arm/parsecpu.awk (gen_comm_data): Don't generate
28122         arm_arch_core_flags data structure.
28123         * config/arm/arm-cpu_cdata.h: Regenerated.
28125 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28127         * common/config/arm/arm-common.c (arm_initialize_isa): Moved here from
28128         config/arm/arm.c.
28129         (arm_print_hint_for_cpu_option): Likewise.
28130         (arm_print_hint_for_arch_option): Likewise.
28131         (arm_parse_cpu_option_name): Likewise.
28132         (arm_parse_arch_option_name): Likewise.
28133         * config/arm/arm.c (arm_identify_fpu_from_isa): Use the computed number
28134         of entries in the all_fpus list.
28135         * config/arm/arm-protos.h (all_architectures, all_cores): Declare.
28136         (arm_parse_cpu_option_name): Declare.
28137         (arm_parse_arch_option_name): Declare.
28138         (arm_parse_option_features): Declare.
28139         (arm_intialize_isa): Declare.
28140         * config/arm/parsecpu.awk (gen_data): Move CPU and architecture
28141         data tables to ...
28142         (gen_comm_data): ... here.  Make definitions non-static.
28143         * config/arm/arm-cpu-data.h: Regenerated.
28144         * config/arm/arm-cpu-cdata.h: Regenerated.
28146 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28148         * config/arm/arm-protos.h (arm_build_target): Remove arch_core.
28149         (cpu_arch_extension): New structure.
28150         (cpu_arch_option, arch_option, cpu_option): New structures.
28151         * config/arm/parsecpu.awk (gen_headers): Build an enumeration of
28152         architecture types.
28153         (gen_data): Generate new format data tables.
28154         * config/arm/arm.c (cpu_tune): New structure.
28155         (cpu_option, processors): Delete.
28156         (arm_print_hint_for_core_or_arch): Delete.  Replace with ...
28157         (arm_print_hint_for_cpu_option): ... this and ...
28158         (arm_print_hint_for_arch_option): ... this.
28159         (arm_parse_arch_cpu_name): Delete.  Replace with ...
28160         (arm_parse_cpu_option_name): ... this and ...
28161         (arm_parse_arch_option_name): ... this.
28162         (arm_unrecognized_feature): Change type of target parameter to
28163         cpu_arch_option.
28164         (arm_parse_arch_cpu_features): Delete.  Replace with ...
28165         (arm_parse_option_features): ... this.
28166         (arm_configure_build_target): Rework to use new configuration data
28167         tables.
28168         (arm_print_tune_info): Rework for new configuration data tables.
28169         * config/arm/arm-cpu-data.h: Regenerated.
28170         * config/arm/arm-cpu.h: Regenerated.
28172 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28174         * Makefile.in (OBJS): Move sbitmap.o from here ...
28175         (OBJS-libcommon): ... to here.
28177 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28179         * config/arm/arm-isa.h (ISA_ALL_FPU_INTERNAL): Renamed from ISA_ALL_FPU.
28180         (ISA_ALL_CRYPTO): New macro.
28181         (ISA_ALL_SIMD): New macro
28182         (ISA_ALL_FP): New macro.
28183         * config/arm/arm.c (fpu_bitlist): Update initializer.
28184         * config/arm/arm-cpus.in: Use new ISA_ALL macros to disable crypto,
28185         simd or fp.
28186         (arm9e): Add fpu.  Add option for nofp
28187         (arm946e-s, arm966e-s, arm968e-s, arm10e, arm1020e, arm1022e): Likewise.
28188         (arm926ej-s, arm1026ej-s): Likewise.
28189         (generic-armv7-a): Add fpu.  Add options for simd, vfpv3, vfpv3-d16,
28190         vfpv3-fp16, vfpv3-d16-fp16, vfpv4, vfpv4-d16, neon, neon-vfp3,
28191         neon-fp16, neon-vfpv4, nofp and nosimd.
28192         (cortex-a5, cortex-a7): Add fpu.  Add options for nosimd and nofp.
28193         (cortex-a8): Add fpu.  Add option for nofp.
28194         (cortex-a9): Add fpu.  Add options for nosimd and nofp.
28195         (cortex-a12, cortex-a15, cortex-a17): Add fpu.  Add option for nofp.
28196         (cortex-r4f): Add fpu.
28197         (cortex-r5): Add fpu.  Add options for nofp.dp and nofp.
28198         (cortex-r7): Use idiv option from architecture.  Add fpu.  Add option
28199         for nofp.
28200         (cortex-r8): Likewise.
28201         (cortex-m4): Add fpu.  Add option for nofp.
28202         (cortex-a15.cortex-a7): Add fpu.  Add option for nofp.
28203         (cortex-a17.cortex-a7): Likewise.
28204         (cortex-a32): Add fpu.  Add options for crypto and nofp.
28205         (cortex-a35, cortex-a53): Likewise.
28206         (cortex-a57): Add fpu.  Add option for crypto.
28207         (cortex-a72, cortex-a73): Likewise.
28208         (exynos-m1): Likewise.
28209         (cortex-a57.cortex-a53, cortex-a72.cortex-a53): Likewise.
28210         (cortex-a73.cortex-a35, cortex-a73.cortex-a53): Likewise.
28211         (cortex-m33): Add fpu.  Add option for nofp.
28212         * config/arm/arm-cpu-cdata.h: Regenerated
28213         * config/arm/arm-cpu-data.h: Regenerated.
28215 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28217         * arm-cpus.in (armv5e): Add options fp, vfpv2 and nofp.
28218         (armv5te, armv5tej): Likewise.
28219         (armv6, armv6j, armv6k, armv6z, armv6kz, armv6zk, armv6t2): Likewise.
28220         (armv7): Add options fp and vfpv3-d16.
28221         (armv7-a): Add options fp, simd, vfpv3, vfpv3-d16, vfpv3-d16-fp16,
28222         vfpv3-fp16, vfpv4, vfpv4-d16, neon, neon-vfpv3, neon-fp16, neon-vfpv4,
28223         nofp and nosimd.
28224         (armv7ve): Likewise.
28225         (armv7-r): Add options fp, fp.sp, idiv, nofp and noidiv.
28226         (armv7e-m): Add options fp, fpv5, fp.dp and nofp.
28227         (armv8-a): Add nocrypto option.
28228         (armv8.1-a, armv8.2-a): Likewise.
28229         (armv8-m.main): add options fp, fp.dp and nofp.
28231 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28233         * config/arm/arm-cpus.in (armv8-a): Add options crc, simd crypto and
28234         nofp.
28235         (armv8-a+crc): Delete.
28236         (armv8.1-a): Add options simd, crypto and nofp.
28237         (armv8.2-a): Add options fp16, simd, crypto and nofp.
28238         (armv8.2-a+fp16): Delete.
28239         (armv8-m.main): Add option dsp.
28240         (armv8-m.main+dsp): Delete.
28241         (cortex-a8): Add fpu.  Add nofp option.
28242         (cortex-a9): Add fpu.  Add nofp and nosimd options.
28243         * config/arm/parsecpu.awk (gen_data): Generate option tables and
28244         link to main cpu and architecture data structures.
28245         (gen_comm_data): Only put isa attributes from the main architecture
28246         in common tables.
28247         (option): New statement for architecture and CPU entries.
28248         * arm.c (struct cpu_option): New structure.
28249         (struct processors): Add entry for options.
28250         (arm_unrecognized_feature): New function.
28251         (arm_parse_arch_cpu_name): Ignore any characters after the first
28252         '+' character.
28253         (arm_parse_arch_cpu_feature): New function.
28254         (arm_configure_build_target): Separate out any CPU and architecture
28255         features and parse separately.  Don't error out if -mfpu=auto is
28256         used with only an architecture string.
28257         (arm_print_asm_arch_directives): New function.
28258         (arm_file_start): Call it.
28259         * config/arm/arm-cpu-cdata.h: Regenerated.
28260         * config/arm/arm-cpu-data.h: Likewise.
28261         * config/arm/arm-tables.opt: Likewise.
28263 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28265         * config/arm/elf.h (ASM_SPEC): Only pass -mfpu through to the
28266         assembler when it is not -mfpu=auto.
28268 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28270         * config/arm/arm.h (BIG_LITTLE_SPEC): Delete macro.
28271         (ASM_REWRITE_SPEC_FUNCTIONS): New macro.
28272         (BIG_LITTLE_CPU_SPEC_FUNCTIONS): Delete macro.
28273         (ASM_CPU_SPEC): Rewrite.
28274         (MCPU_MTUNE_NATIVE_FUNCTIONS): New macro.
28275         (EXTRA_SPEC_FUNCTIONS): Move outside of ifdef.  Use
28276         MCPU_MTUNE_NATIVE_FUNCTIONS and ASM_REWRITE_SPEC_FUNCTIONS.  Remove
28277         reference to BIG_LITTLE_CPU_SPEC_FUNCTIONS.
28278         * common/config/arm/arm-common.c (arm_rewrite_selected_cpu): Ensure
28279         copied string is NUL-terminated.  Also strip any characters prefixed
28280         by '+'.
28281         (arm_rewrite_selected_arch): New function.
28282         (arm_rewrite_march): New function.
28284 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28286         * config/arm/arm.opt (x_arm_arch_string): New TargetSave option.
28287         (x_arm_cpu_string, x_arm_tune_string): Likewise.
28288         (march, mcpu, mtune): Convert to string-based options.
28289         * config/arm/arm.c (arm_print_hint_for_core_or_arch): New function.
28290         (arm_parse_arch_cpu_name): New function.
28291         (arm_configure_build_target): Use arm_parse_arch_cpu_name to
28292         identify selected architecture or CPU.
28293         (arm_option_save): New function.
28294         (TARGET_OPTION_SAVE): Redefine.
28295         (arm_option_restore): Restore string options.
28296         (arm_option_print): Print string options.
28298 2017-06-16  Martin Sebor  <msebor@redhat.com>
28300         PR tree-optimization/80933
28301         PR tree-optimization/80934
28302         * builtins.c (fold_builtin_3): Do not handle bcmp here.
28303         * gimple-fold.c (gimple_fold_builtin_bcmp): New function.
28304         (gimple_fold_builtin_bcopy, gimple_fold_builtin_bzero): Likewise.
28305         (gimple_fold_builtin): Call them.
28307 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28309         * gimple-ssa-isolate-paths.c (isolate_path): Set edge leading to path
28310         as unlikely; update profile.
28312 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28314         * predict.c (force_edge_cold): Handle declaring edges impossible
28315         more aggresively.
28317 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28319         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update
28320         profile.
28321         (try_unroll_loop_completely): Fix reporting.
28323 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28325         * tree-ssa-tail-merge.c (replace_block_by): Fix profile updating.
28327 2017-06-16  James Greenhalgh  <james.greenhalgh@arm.com>
28329         PR target/71778
28330         * config/arm/arm-builtins.c (arm_expand_builtin_args): Return TARGET
28331         if given a non-constant argument for an intrinsic which requires a
28332         constant.
28334 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28336         * profile.c (compare_freqs): New function.
28337         (branch_prob): Sort edge list.
28338         (find_spanning_tree): Assume that the list is priority sorted.
28340 2017-06-16  Richard Biener  <rguenther@suse.de>
28342         PR tree-optimization/81090
28343         * passes.def (pass_record_bounds): Remove.
28344         * tree-pass.h (make_pass_record_bounds): Likewise.
28345         * tree-ssa-loop.c (pass_data_record_bounds, pass_record_bounds,
28346         make_pass_record_bounds): Likewise.
28347         * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables): Do
28348         not free niter estimates at the beginning but at the end.
28349         * tree-scalar-evolution.c (scev_finalize): Free niter estimates.
28351 2017-06-16  Richard Biener  <rguenther@suse.de>
28353         * tree-switch-conversion.c (emit_case_bit_tests): Adjust
28354         initializer to workaround ICE in host GCC 4.8.
28356 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28358         * ipa-inline-transform.c (update_noncloned_frequencies): Update also
28359         counts.
28360         (clone_inlined_nodes): Update.
28362 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28364         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune): Update
28365         prefetch settings, and enable prefetching by default at -O3.
28367 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28369         * config/aarch64/aarch64.c (aarch64_override_options_internal):
28370         Set flag_prefetch_loop_arrays according to tuning data.
28372 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28374         * config/aarch64/aarch64-protos.h (struct cpu_prefetch_tune):
28375         New tune structure.
28376         (struct tune_params): Use cpu_prefetch_tune instead of cache_line_size.
28377         [Unrelated to main purpose of the patch] Place the pointer field last
28378         to enable type checking errors when tune structure are wrongly merged.
28379         * config/aarch64/aarch64.c (generic_prefetch_tune,)
28380         (exynosm1_prefetch_tune, qdf24xx_prefetch_tune,)
28381         (thunderx2t99_prefetch_tune): New tune constants.
28382         (tune_params *_tunings): Update all tunings (no functional change).
28383         (aarch64_override_options_internal): Set PARAM_SIMULTANEOUS_PREFETCHES,
28384         PARAM_L1_CACHE_SIZE, PARAM_L1_CACHE_LINE_SIZE, and PARAM_L2_CACHE_SIZE
28385         from tunings structures.
28387 2017-06-16  Jakub Jelinek  <jakub@redhat.com>
28389         PR sanitizer/81094
28390         * ubsan.c (instrument_null): Add T argument, use it instead
28391         of computing it based on IS_LHS.
28392         (instrument_object_size): Likewise.
28393         (pass_ubsan::execute): Adjust instrument_null and
28394         instrument_object_size callers to pass gimple_get_lhs or
28395         gimple_assign_rhs1 result to it.  Use instrument_null instead of
28396         calling get_base_address and instrument_mem_ref.  Handle
28397         aggregate call arguments for object-size sanitization.
28399 2017-06-16  Yury Gribov  <tetra2005@gmail.com>
28401         PR tree-optimization/81089
28402         * tree-vrp.c (is_masked_range_test): Validate operands of
28403         subexpression.
28405 2017-06-15  Martin Sebor  <msebor@redhat.com>
28407         PR c++/80560
28408         * dumpfile.c (dump_register): Avoid calling memset to initialize
28409         a class with a default ctor.
28410         * gcc.c (struct compiler): Remove const qualification.
28411         * genattrtab.c (gen_insn_reserv): Replace memset with initialization.
28412         * hash-table.h: Ditto.
28413         * ipa-cp.c (allocate_and_init_ipcp_value): Replace memset with
28414           assignment.
28415         * ipa-prop.c (ipa_free_edge_args_substructures): Ditto.
28416         * omp-low.c (lower_omp_ordered_clauses): Replace memset with
28417         default ctor.
28418         * params.h (struct param_info): Make struct members non-const.
28419         * tree-switch-conversion.c (emit_case_bit_tests): Replace memset
28420         with default initialization.
28421         * vec.h (vec_copy_construct, vec_default_construct): New helper
28422         functions.
28423         (vec<T>::copy, vec<T>::splice, vec<T>::reserve): Replace memcpy
28424         with vec_copy_construct.
28425         (vect<T>::quick_grow_cleared): Replace memset with default ctor.
28426         (vect<T>::vec_safe_grow_cleared, vec_safe_grow_cleared): Same.
28427         * doc/invoke.texi (-Wclass-memaccess): Document.
28429 2017-06-15  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
28431         * emit-rtl.h (is_leaf): Update comment about local
28432         register allocator.
28434 2017-06-15  Jozef Lawrynowicz   <jozef.l@somniumtech.com>
28436         PR target/78818
28437         * config/msp430/msp430.c (msp430_data_attr): Check that it's possible
28438         for a variable to have a section before checking if the section has a
28439         name.
28440         Set section to.persistent if persistent attribute is set.
28441         Warn if .persistent attribute is used on an automatic variable.
28443 2017-06-15  Eric Botcazou  <ebotcazou@adacore.com>
28445         PR rtl-optimization/80474
28446         * reorg.c (update_block): Do not ignore instructions in a delay slot.
28448 2017-06-15  Segher Boessenkool  <segher@kernel.crashing.org>
28450         * config/rs6000/rs6000.md (add<mode>3): Use reg_or_subregno instead
28451         of REGNO.
28453 2017-06-14  Maciej W. Rozycki  <macro@imgtec.com>
28455         * config/mips/mips.md (MIPS16_T_REGNUM): Remove constant.
28456         (casesi): Emit bounds checking as RTL.
28457         (casesi_internal_mips16_<mode>): Remove bounds checking.
28459 2017-06-14  Max Filippov  <jcmvbkbc@gmail.com>
28461         * config/xtensa/xtensa.c (xtensa_option_override): Append
28462         MASK_CONST16 to target_flags in the absence of TARGET_L32R.
28463         (hwloop_optimize, hwloop_fail, hwloop_pattern_reg,
28464          xtensa_doloop_hooks): Define unconditionally.
28465         (xtensa_reorg_loops): Only call reorg_loops in the presence of
28466         TARGET_LOOPS.
28467         * config/xtensa/xtensa.h (TARGET_L32R): New definition.
28468         (TARGET_DEFAULT): Remove XCHAL_HAVE_L32R condition and account
28469         for it in xtensa_option_override.
28470         (HARD_FRAME_POINTER_IS_FRAME_POINTER,
28471          HARD_FRAME_POINTER_IS_ARG_POINTER): New definitions.
28473 2017-06-14  Boris Kolpackov  <boris@codesynthesis.com>
28475         * doc/cppopts.texi: Document '-' special value to -MF.
28477 2017-06-14  Wilco Dijkstra  <wdijkstr@arm.com>
28479         * config/arm/cortex-a53.md (cortex_a53_fpalu) Adjust latency.
28480         (cortex_a53_fconst): Likewise.
28481         (cortex_a53_fpmul): Likewise.
28482         (cortex_a53_f_load_64): Likewise.
28483         (cortex_a53_f_load_many): Likewise.
28484         (cortex_a53_advsimd_alu): Likewise.
28485         (cortex_a53_advsimd_alu_q): Likewise.
28486         (cortex_a53_advsimd_mul): Likewise.
28487         (cortex_a53_advsimd_mul_q): Likewise.
28488         (fpmac bypass): Add new bypass for fpmac-fpmac case.
28489         Add missing fmul, r2f_cvt and fconst cases.
28491 2017-06-14  Richard Biener  <rguenther@suse.de>
28493         PR middle-end/81088
28494         * fold-const.c (split_tree): Drop TREE_OVERFLOW flag from
28495         literal constants.
28496         (fold_binary_loc): When associating do not treat pre-existing
28497         TREE_OVERFLOW on literal constants as a reason to allow
28498         TREE_OVERFLOW on associated literal constants.
28500 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28502         * config/sparc/sparc.h (MASK_ISA): Add MASK_LEON and MASK_LEON3.
28503         (MASK_FEATURES): New macro.
28504         * config/sparc/sparc.c (sparc_option_override): Remove the special
28505         handling of -mfpu and generalize it to all MASK_FEATURES switches.
28507 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28509         * simplify-rtx.c (simplify_binary_operation_1) <UDIV>: Do not simplify
28510         a division of 0 if non-call exceptions are enabled.
28512 2017-06-14  Andrew Pinski  <apinski@cavium.com>
28513             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
28515         PR target/71663
28516         * config/aarch64/aarch64.c (aarch64_expand_vector_init):
28517         Improve vector initialization code gen for only variable case.
28519 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28521         * config/sparc/driver-sparc.c (cpu_names): Add SPARC-T5 entry.
28523 2017-06-14  Richard Biener  <rguenther@suse.de>
28525         PR tree-optimization/81083
28526         * tree-ssa-sccvn.c (vn_reference_lookup_3): Do not use abnormals
28527         as values.
28529 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28531         * config/rs6000/rs6000.c: Update all comments that mentioned SPE.
28532         (rs6000_expand_builtin): Remove RS6000_BTC_EVSEL.
28533         * config/rs6000/rs6000.h (RS6000_BTC_EVSEL): Delete.
28534         * config/rs6000/vxworks.h (VXCPU_FOR_8548): Delete.  Adjust former use.
28535         * config/rs6000/vxworksae.h (VXCPU_FOR_8548): Delete.
28536         * config/rs6000/vxworksmils.h (VXCPU_FOR_8548): Delete.
28538 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28540         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete VECTOR_SPE.
28541         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete VECTOR_SPE.
28543 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28545         * config/rs6000/rs6000.h (FIXED_SCRATCH): Delete.
28547 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28549         * config/rs6000/t-rtems: Don't handle SPE.
28551 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28553         * config/rs6000/t-linux: Don't handle SPE.
28555 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28557         * config/rs6000/eabispe.h: Delete file.
28559 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28561         * config/rs6000/t-spe: Delete file.
28563 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28565         * config/rs6000/rs6000.c (SPE_CONST_OFFSET_OK): Delete.
28566         (rs6000_legitimate_offset_address_p): Return false for anything in
28567         V2SImode or V2SFmode.
28569 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28571         * config/rs6000/rs6000-modes.def: Remove all 8-byte vector modes
28572         except V2SF and V2SI.  Rearrange the vector modes, and add comments.
28573         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove V8QImode
28574         and V4HImode.
28575         (reg_offset_addressing_ok_p): Remove V4HImode and V1DImode.
28576         (rs6000_legitimate_offset_address_p): Ditto.
28577         (rs6000_emit_move): Ditto.
28578         (rs6000_init_builtins): Remove V4HI_type_node.
28580 2017-06-13  Martin Liska  <mliska@suse.cz>
28582         PR sanitize/78204
28583         * asan.c (asan_sanitize_stack_p): Use sanitize_flags_p.
28584         (gate_asan): Likewise.
28585         * asan.h (asan_no_sanitize_address_p): Remove the function.
28586         (sanitize_flags_p): New function.
28587         * builtins.def: Fix coding style.
28588         * common.opt: Use renamed enum value.
28589         * convert.c (convert_to_integer_1): Use sanitize_flags_p.
28590         * doc/extend.texi: Document no_sanitize attribute.
28591         * flag-types.h (enum sanitize_code): Rename SANITIZE_NONDEFAULT
28592         to SANITIZE_UNDEFINED_NONDEFAULT.
28593         * gcc.c (sanitize_spec_function): Use the renamed enum value.
28594         * gimple-fold.c (optimize_atomic_compare_exchange_p):
28595         Use sanitize_flags_p.
28596         * gimplify.c (gimplify_function_tree): Likewise.
28597         * ipa-inline.c (sanitize_attrs_match_for_inline_p): Likewise.
28598         * opts.c (parse_no_sanitize_attribute): New function.
28599         (common_handle_option): Use renamed enum value.
28600         * opts.h (parse_no_sanitize_attribute): Declare.
28601         * tree.c (sanitize_flags_p): New function.
28602         * tree.h: Declared here.
28603         * tsan.c: Use sanitize_flags_p.
28604         * ubsan.c (ubsan_expand_null_ifn): Likewise.
28605         (instrument_mem_ref): Likewise.
28606         (instrument_bool_enum_load): Likewise.
28607         (do_ubsan_in_current_function): Remove the function.
28608         (pass_ubsan::execute): Use sanitize_flags_p.
28609         * ubsan.h: Remove do_ubsan_in_current_function
28610         * tree-cfg.c (print_no_sanitize_attr_value): New function.
28611         (dump_function_to_file): Use it here.
28613 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28615         PR tree-optimization/80803
28616         PR tree-optimization/81063
28617         * tree-sra.c (subtree_mark_written_and_enqueue): Move up in the file.
28618         (propagate_subaccesses_across_link): Enqueue subtree whenever
28619         necessary instead of relying on the caller.
28621 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28623         * tree-sra.c (add_access_to_work_queue): Only enqueue accesses
28624         that have a first_link.
28625         (sort_and_splice_var_accesses): Do not check first_link before
28626         enquing.
28627         (subtree_mark_written_and_enqueue): Likewise.
28628         (propagate_all_subaccesses): Likewise and do not stop at first
28629         parent with a first_link.
28631 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28633         * tree-sra.c (dump_access_tree_1): Fix accidental dumping to stderr
28634         instead of f.
28636 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28638         * match.pd: New pattern.
28640 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28642         * tree-vrp.c (is_masked_range_test): New function.
28643         (register_edge_assert_for): Determine ranges for
28644         some bit tests.
28646 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28648         PR tree-optimization/67328
28649         * fold-const.c (maskable_range_p): New function.
28650         (build_range_check): Generate bittests if possible.
28652 2017-06-13  Martin Liska  <mliska@suse.cz>
28654         * gimple-pretty-print.c (dump_probability): Add new argument.
28655         (dump_edge_probability): Dump both probability and count.
28656         (dump_gimple_label): Likewise.
28657         (dump_gimple_bb_header): Likewise.
28659 2017-06-13  Georg-Johann Lay  <avr@gjlay.de>
28661         PR target/81072
28662         * config/avr/avr-devices.c: Fix indentation.
28663         * config/avr/gen-avr-mmcu-specs.c: Dito.
28665 2017-06-13  Richard Biener  <rguenther@suse.de>
28667         * tree-vect-loop.c (vect_model_reduction_cost): Do not fail,
28668         instead get vector type from stmt_info.
28669         (vectorizable_reduction): Adjust.  Remove dead code.
28671 2017-06-13  Richard Biener  <rguenther@suse.de>
28673         PR middle-end/81065
28674         * fold-const.c (extract_muldiv_1): Remove bogus distribution
28675         case of C * (x * C2 + C3).
28676         (fold_addr_of_array_ref_difference): Properly fold index difference.
28678 2017-06-12  David S. Miller  <davem@davemloft.net>
28680         PR target/80968
28681         * config/sparc/sparc.md (return expander): Emit frame blockage if
28682         function uses alloca.
28684 2017-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
28686         * combine.c (make_field_assignment): Check len rather than the mode
28687         precision when calling force_to_mode.
28689 2017-06-12  Georg-Johann Lay  <avr@gjlay.de>
28691         Support multilibs and devices that see flash in RAM address range.
28693         PR target/81072
28694         * config/avr/avr-arch.h (avr_arch_id) <ARCH_AVRXMEGA3>: New enum.
28695         (avr_mcu_t) <flash_pm_offset>: New field.
28696         (avr_device_specific_features) <AVR_ISA_RCALL>: New enum.
28697         * config/avr/avr.h (AVR_SHORT_CALLS): New define.
28698         (AVR_HAVE_JMP_CALL): Don't set if AVR_SHORT_CALLS.
28699         (AVR_TINY_PM_OFFSET): Remove macro.
28700         * config/avr/avr.opt (-mshort-calls): New option.
28701         * config/avr/gen-avr-mmcu-specs.c (print_mcu)
28702         [*self_spec]: Add / remove -mshort-calls depending on AVR_ISA_RCALL.
28703         * config/avr/avr-c.c (avr_cpu_cpp_builtins)
28704         <__AVR_SHORT_CALLS__>: Built-in define if AVR_SHORT_CALLS.
28705         <__AVR_HAVE_JMP_CALL__>: Use AVR_HAVE_JMP_CALL as condition
28706         instead of avr_arch->have_jmp_call.
28707         <__AVR_PM_BASE_ADDRESS__>: Built-in define if avr_arch->flash_pm_offset.
28708         [AVR_TINY] <__AVR_TINY_PM_BASE_ADDRESS__>: Use
28709         avr_arch->flash_pm_offset to define.
28710         * config/avr/avr-devices.c (avr_arch_types): Add initializers for
28711         new field flash_pm_offset.  Add entry for avrxmega3.
28712         (avr_texinfo): Add entry for avrxmega3.
28713         * config/avr/avr-mcus.def: Add entries for: avrxmega3,
28714         attiny212, attiny214,
28715         attiny412, attiny414, attiny416, attiny417,
28716         attiny814, attiny816, attiny817,
28717         attiny1614, attiny1616, attiny1617,
28718         attiny3214, attiny3216, attiny3217.
28719         * config/avr/avr.c (avr_assemble_integer)[AVR_TINY]: Use
28720         avr_arch->flash_pm_offset instead of AVR_TINY_PM_OFFSET.
28721         (avr_print_operand_address) [AVR_TINY]: Same.
28722         (avr_asm_init_sections) <readonly_data_section>: Only patch
28723         callback if avr_arch->flash_pm_offset = 0.
28724         (avr_asm_named_section) <avr_need_copy_data_p>: Skip setting it
28725         for rodata if avr_arch->flash_pm_offset != 0.
28726         (avr_encode_section_info) [AVR_TINY]: Adjust comment.
28727         * config/avr/genmultilib.awk (dir_rcall, opt_rcall): New vars.
28728         (opts) [AVR_ISA_RCALL]: Append opt_rcall.
28729         (m_options): Append opt_rcall.
28730         (m_dirnames): Append dir_rcall.
28731         * config/avr/t-multilib: Regenerate.
28733         * configure.ac [target=avr]: Check whether avrxmega3 default
28734         linker description file works as needed.
28735         * configure: Regenerate.
28736         * doc/avr-mmcu.texi: Regenerate.
28737         * doc/invoke.texi (AVR Options) <-mshort-calls>: Document it.
28738         <__AVR_ARCH__>: Document avrxmega3 and 103.
28739         <__AVR_HAVE_JMP_CALL__>: Adjust documentation.
28740         <__AVR_SHORT_CALLS__>: Document it.
28741         <__AVR_PM_BASE_ADDRESS__>: Document it.
28742         * doc/extend.texi (AVR Options) <-mshort-calls>: Document it.
28743         (AVR Variable Attributes) <progmem>: Document this is
28744         not needed for avrxmega3.
28745         (AVR Named Address Spaces) <__flash>: Dito.
28747 2017-06-12  Jan Hubicka  <hubicka@ucw.cz>
28749         * cgraph.c (cgraph_node::dump): Complain about profile insanities.
28751 2017-06-12  Doug Rupp  <rupp@adacore.com>
28753         * config.gcc (*-*-vxworks*): Set use_gcc_stdint to "provide".
28754         Append vxworks-stdint.h to the tm_file list.
28755         * config/vxworks-stdint.h: New file.
28757 2017-06-12  Martin Liska  <mliska@suse.cz>
28759         PR tree-optimization/81041
28760         * tree-profile.c (gimple_gen_ic_func_profiler):
28761         Create an extra BB in profile-generate
28762         (gimple_gen_time_profiler): Likewise.
28764 2017-06-12  Jakub Jelinek  <jakub@redhat.com>
28766         PR tree-optimization/81003
28767         * tree-ssa-reassoc.c (force_into_ssa_name): New function.
28768         (update_range_test): Use it instead of force_gimple_operand_gsi.
28770 2017-06-12  Richard Biener  <rguenther@suse.de>
28772         PR tree-optimization/81053
28773         * tree-vect-loop.c (vect_is_simple_reduction): Handle PHI
28774         with backedge value not defined in loop.  Simplify def stmt
28775         compute.
28777 2017-06-11  Tom de Vries  <tom@codesourcery.com>
28779         PR target/79939
28780         * config/nvptx/nvptx.c (nvptx_cannot_force_const_mem): New function.
28781         Return true.
28782         (TARGET_CANNOT_FORCE_CONST_MEM): Redefine to
28783         nvptx_cannot_force_const_mem.
28785 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28787         * opts.c (finish_options): Move test for flag_split_stack after
28788         it has been initialized.
28790 2017-06-11  Jason Merrill  <jason@redhat.com>
28792         * tree.h (id_equal): New.
28793         * dwarf2out.c, hsa-gen.c, ipa-devirt.c, omp-expand.c,
28794         omp-simd-clone.c, read-rtl-function.c, tree-chkp.c, tree.c: Use it
28795         instead of strcmp of IDENTIFIER_POINTER.
28797 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28799         * ipa-inline-transform.c: Include function.h, cfg.h and basic-block.h
28800         (mark_all_inlined_calls_cdtor): Fix formating.
28801         (inline_transform): Rescale profile before inlining.
28803 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28805         * cgraph.h (cgraph_edge::clone): Update prototype.
28806         * cgraphclones.c (cgraph_edge::clone): Update profile scaling.
28807         (cgraph_node::create_clone): Update.
28808         (cgraph_node::create_version_clone): Update.
28809         * tree-inline.c (copy_bb): Update.
28810         (expand_call_inline): Update.
28812 2017-06-10  Segher Boessenkool  <segher@kernel.crashing.org>
28814         * config/rs6000/rs6000.c (emit_vrsave_prologue): New function,
28815         factored out from ...
28816         (rs6000_emit_prologue): ... here.
28818 2017-06-10  Segher Boessenkool  <segher@kernel.crashing.org>
28820         * config/rs6000/rs6000.c (emit_split_stack_prologue): New function,
28821         factored out from ...
28822         (rs6000_emit_prologue): ... here.
28824 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28826         * predict.c (drop_profile): Also drop individual bb/edge and cgraph
28827         edge counts.
28828         (handle_missing_profiles): Fix computation of tp_first_run.
28829         (counts_to_freqs): Do not touch freqs when count is 0.
28831 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28833         * cgraphbuild.c (cgraph_edge::rebuild_references): Do not touch
28834         profile.
28836 2017-06-10  Tom de Vries  <tom@codesourcery.com>
28838         * doc/sourcebuild.texi (Effective-Target Keywords, Environment
28839         attributes): Document signal effective target.
28841 2017-06-10  Tom de Vries  <tom@codesourcery.com>
28843         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
28844         Document effective target stack_size.
28846 2017-06-09  David Malcolm  <dmalcolm@redhat.com>
28848         * diagnostic.c (diagnostic_report_diagnostic): Only add fixits
28849         to the edit_context if they can be auto-applied.
28851 2017-06-9  Ian Lance Taylor  <iant@golang.org>
28853         * opts.c (finish_options): If -fsplit-stack, disable implicit
28854         -forder-blocks-and-partition.
28855         * doc/invoke.texi (Optimize Options): Document that when using
28856         -fsplit-stack -forder-blocks-and-partition is not implicitly
28857         enabled.
28859 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28861         * builtin-attrs.def (ATTR_NORETURN_NOTHROW_LEAF_COLD_LIST,
28862         ATTR_CONST_NORETURN_NOTHROW_LEAF_COLD_LIST,
28863         ATTR_TMPURE_NORETURN_NOTHROW_LEAF_COLD_LIST): New.
28864         * builtins.def (abort, trap, unreachable): Declare cold.
28865         * calls.c (flags_from_decl_or_type): Lookup ECF_COLD.
28866         * tree-core.h (ECF_COLD): New.
28867         * tree.c (set_call_expr_flags): Handle ECF_COLD.
28868         (build_common_builtin_nodes): Mark unreachable and abort as cold.
28870 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28872         * predict.c (unlikely_executed_stmt_p): Cleanup.
28874 2017-06-09  Richard Biener  <rguenther@suse.de>
28876         * tree-ssa-loop-im.c (execute_sm): Do not force multi-threaded
28877         model if the ref is always written to.
28879 2017-06-09  Tamar Christina  <tamar.christina@arm.com>
28881         * config/aarch64/aarch64.md (lrint<GPF:mode><GPI:mode>2): New.
28883 2017-06-09  Tamar Christina  <tamar.christina@arm.com>
28885         * config/arm/arm.c (arm_rtx_costs_internal): Make sdiv more expensive
28886         than udiv.
28888 2017-06-09  Tom de Vries  <tom@codesourcery.com>
28890         PR target/80855
28891         * config/nvptx/nvptx.md (define_expand "mov<QHSDISDFM>"): Error out with
28892         "target cannot support label values" when encountering LABEL_REF.
28894 2017-06-09  Martin Liska  <mliska@suse.cz>
28896         * tree-profile.c (gimple_gen_ic_profiler): Update comment.
28897         (gimple_gen_ic_func_profiler): Emit direct comparison
28898         of __gcov_indirect_call_callee with NULL.
28899         (gimple_gen_time_profiler): Change probability from
28900         PROB_VERY_UNLIKELY to PROB_UNLIKELY.
28902 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28904         * profile.c (edge_gcov_counts): Turn to pointer.
28905         (compute_branch_probabilities, compute_branch_probabilities): Update.
28906         (branch_prob): Do not clear edge_gcov_count.
28907         * profile.h (edge_gcov_counts): Turn to pointer.
28908         (edge_gcov_count): Update.
28910 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28912         * gimple.h (gimple_check_failed): Mark cold.
28914 2017-06-09  Richard Biener  <rguenther@suse.de>
28916         PR tree-optimization/66623
28917         * tree-vect-loop.c (vect_is_simple_reduction): Cleanup,
28918         refactor check_reduction into two parts, properly computing
28919         whether we have to check reduction validity for outer loop
28920         vectorization.
28922 2017-06-09  Richard Biener  <rguenther@suse.de>
28924         PR tree-optimization/79483
28925         * graphite-scop-detection.c (order): New global.
28926         (get_order): Compute bb to order mapping that satisfies code
28927         generation constraints.
28928         (cmp_pbbs): New helper.
28929         (build_scops): Start domwalk at entry block, sort generated
28930         pbbs.
28932 2017-06-09  Richard Biener  <rguenther@suse.de>
28934         PR middle-end/81007
28935         * ipa-polymorphic-call.c
28936         (ipa_polymorphic_call_context::restrict_to_inner_class):
28937         Skip FIELD_DECLs with error_mark_node type.
28938         * passes.def (all_lowering_passes): Run pass_build_cgraph_edges
28939         last again.
28941 2017-06-09  Martin Liska  <mliska@suse.cz>
28943         * predict.c (struct branch_predictor): New struct.
28944         (test_prediction_value_range): New test.
28945         (predict_c_tests): New function.
28946         * selftest-run-tests.c (selftest::run_tests): Run the function.
28947         * selftest.h: Declare new tests.
28949 2017-06-09  Segher Boessenkool  <segher@kernel.crashing.org>
28951         PR target/80966
28952         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack): Assert that
28953         gen_add3_insn did not fail.
28954         * config/rs6000/rs6000.md (add<mode>3): If asked to add a constant to
28955         r0, construct that number in a temporary reg and add that reg to r0.
28956         If asked to put the result in r0 as well, fail.
28958 2017-06-08  Will Schmidt  <will_schmidt@vnet.ibm.com>
28960         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
28961         for early expansion of vec_eqv.
28963 2017-06-08  Jakub Jelinek  <jakub@redhat.com>
28965         PR middle-end/81005
28966         * ubsan.c (instrument_null): Avoid pointless code temporary.
28967         (pass_ubsan::execute): Instrument aggregate arguments of calls.
28969 2017-06-08  Uros Bizjak  <ubizjak@gmail.com>
28971         PR target/81015
28972         Revert:
28973         2016-12-14  Uros Bizjak  <ubizjak@gmail.com>
28975         PR target/59874
28976         * config/i386/i386.md (*ctzhi2): New insn_and_split pattern.
28977         (*clzhi2): Ditto.
28979 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28981         * predict.c (unlikely_executed_edge_p): Move ahead.
28982         (probably_never_executed_edge_p): Use it.
28984 2017-06-08  Alexander Ivchenko  <alexander.ivchenko@intel.com>
28986         PR middle-end/79988
28987         * tree-chkp.c (chkp_gimple_call_builtin_p): Remove
28988         gimple_call_builtin_p call.
28990 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28992         * system.h (fancy_abort): Annotate by ATTRIBUTE_COLD.
28993         * rtl.h (rtl_check_failed_bounds, rtl_check_failed_type1,
28994         rtl_check_failed_type2, rtl_check_failed_code1,
28995         rtl_check_failed_code2, rtl_check_failed_code_mode,
28996         rtl_check_failed_block_symbol, cwi_check_failed_bounds,
28997         rtvec_check_failed_bounds, rtl_check_failed_flag,
28998         _fatal_insn_not_found, _fatal_insn): Likewise.
28999         * tree.h (tree_contains_struct_check_failed,
29000         tree_check_failed, tree_not_check_failed,
29001         tree_class_check_failed, tree_range_check_failed,
29002         tree_not_class_check_failed, tree_int_cst_elt_check_failed,
29003         tree_vec_elt_check_failed, phi_node_elt_check_failed,
29004         tree_operand_check_failed, omp_clause_check_failed,
29005         omp_clause_operand_check_failed, omp_clause_range_check_failed):
29006         Likewise.
29008 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
29010         * cgraph.c (cgraph_edge::maybe_hot_p): Do not check
29011         flag_branch_probabilities.
29012         * ipa-inline.c (edge_badness): Likewise.
29013         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
29014         * postreload-gcse.c (eliminate_partially_redundant_load): Likewise.
29015         * predict.c (maybe_hot_frequency_p): Likewise.
29016         (probably_never_executed): Likewise.
29017         * sched-ebb.c (schedule_ebbs): Likewise.
29018         * sched-rgn.c (find_single_block_region): Likewise.
29019         * tracer.c (tail_duplicate): Likewise.
29021 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
29023         * opts.c (finish_options): x_flag_reorder_blocks_and_partition no
29024         longer requires x_flag_profile_use.
29026 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
29028         * cfgrtl.c (cfg_layout_initialize): Check crtl->has_bb_partition
29029         instead of flag_reorder_blocks_and_partition.
29030         * dbxout.c (dbxout_function_end): Likewise.
29031         * dwarf2out.c (gen_subprogram_die): Likewise.
29032         * haifa-sched.c (sched_create_recovery_edges): Likewise.
29033         * hw-doloop.c (reorg_loops): Likewise.
29034         * varasm.c (assemble_start_function,
29035         assemble_end_function): Likewise.
29036         (decide_function_section): Do not check for
29037         flag_reorder_blocks_and_partition.
29039 2017-06-08  Alexander Ivchenko  <alexander.ivchenko@intel.com>
29041         * tree-chkp.c (chkp_get_hard_register_var_fake_base_address):
29042         New function.
29043         (chkp_get_hard_register_fake_addr_expr): Ditto.
29044         (chkp_build_addr_expr): Add check for hard reg case.
29045         (chkp_parse_array_and_component_ref): Ditto.
29046         (chkp_find_bounds_1): Ditto.
29047         (chkp_process_stmt): Don't generate bounds store for
29048         hard reg case.
29050 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
29052         * predict.c (maybe_hot_bb_p): Do not check profile status.
29053         (maybe_hot_edge_p): Likewise.
29054         (probably_never_executed): Check for zero counts even if profile
29055         is not read.
29056         (unlikely_executed_edge_p): New function.
29057         (unlikely_executed_stmt_p): New function.
29058         (unlikely_executed_bb_p): New function.
29059         (set_even_probabilities): Use unlikely predicates.
29060         (combine_predictions_for_bb): Likewise.
29061         (predict_paths_for_bb): Likewise.
29062         (predict_paths_leading_to_edge): Likewise.
29063         (determine_unlikely_bbs): New function.
29064         (estimate_bb_frequencies): Use it.
29065         (compute_function_frequency): Use zero counts even if profile is
29066         not read.
29067         * profile-count.h: Fix typo.
29069 2017-08-08  Julia Koval  <julia.koval@intel.com>
29071         * config/i386/avx512bwintrin.h (_mm512_mask_cvtepi16_storeu_epi8,
29072         _mm512_mask_cvtsepi16_storeu_epi8,
29073         _mm512_mask_cvtusepi16_storeu_epi8): New intrinsics.
29074         * config/i386/avx512vlbwintrin.h (_mm256_mask_cvtepi16_storeu_epi8,
29075         _mm_mask_cvtsepi16_storeu_epi8, _mm256_mask_cvtsepi16_storeu_epi8,
29076         _mm_mask_cvtusepi16_storeu_epi8, _mm256_mask_cvtusepi16_storeu_epi8,
29077         _mm_mask_cvtepi16_storeu_epi8): New intrinsics.
29078         * config/i386/i386-builtin-types.def (PV8Q, V8QI): New pointer type.
29079         (VOID_FTYPE_PV32QI_V32HI_USI, VOID_FTYPE_PV8QI_V8HI_UQI,
29080         VOID_FTYPE_PV16QI_V16HI_UHI): New function types.
29081         * config/i386/i386-builtin.def (__builtin_ia32_pmovwb128mem_mask,
29082         __builtin_ia32_pmovwb256mem_mask, __builtin_ia32_pmovswb128mem_mask,
29083         __builtin_ia32_pmovswb256mem_mask, __builtin_ia32_pmovuswb128mem_mask,
29084         __builtin_ia32_pmovuswb256mem_mask,
29085         __builtin_ia32_pmovuswb512mem_mask, __builtin_ia32_pmovswb512mem_mask)
29086         __builtin_ia32_pmovwb512mem_mask): New builtins.
29088 2017-08-08  Julia Koval  <julia.koval@intel.com>
29090         PR target/73350,80862
29091         * config/i386/subst.md (round): Fix round pattern.
29092         * config/i386/i386.c (ix86_erase_embedded_rounding):
29093         Fix erasing rounding for the fixed pattern.
29095 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
29097         * cfgbuild.c (find_many_sub_basic_blocks): Fix thinko.
29099 2017-06-08  Martin Liska  <mliska@suse.cz>
29101         PR gcov-profile/80911
29102         * gcov.c (block_info::block_info): New constructor.
29104 2017-06-07  Carl Love  <cel@us.ibm.com>
29106         * config/rs6000/rs6000-c: The return type of the following
29107         built-in functions was implemented as int not long long.  Fix sign
29108         of return value for the unsigned version of vec_mulo and vec_mule.
29109         vector unsigned long long vec_bperm (vector unsigned long long,
29110                                              vector unsigned char)
29111         vector signed long long vec_mule (vector signed int,
29112                                           vector signed int)
29113         vector unsigned long long vec_mule (vector unsigned int,
29114                                             vector unsigned int)
29115         vector signed long long vec_mulo (vector signed int,
29116                                           vector signed int)
29117         vector unsigned long long vec_mulo (vector unsigned int,
29118                                             vector unsigned int)
29119         * doc/extend.texi: Fix the documentation for the built-in
29120         functions.
29122 2017-06-07  Carl Love  <cel@us.ibm.com>
29124         PR target/80982
29125         * config/rs6000/altivec.md (double<mode>2): Fix the implementation of
29126         for BE.
29128 2017-06-07  Carl Love  <cel@us.ibm.com>
29130         * config/rs6000/altivec.md: Fix argument swizzle in vec_doublel
29131         support, Generate       doublehv for signed int/float for BE case only.
29133 2017-06-07  Alexander Monakov  <amonakov@ispras.ru>
29135         * doc/invoke.texi (mcx16): Rewrite.
29137 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29139         * config/rs6000/predicates.md (rs6000_nonimmediate_operand): Delete.
29140         * config/rs6000/rs6000.md (*movsi_internal1, movsi_from_sf,
29141         *mov<mode>_softfloat, and an anonymous splitter): Use
29142         nonimmediate_operand instead of rs6000_nonimmediate_operand.
29144 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29146         * config/rs6000/darwin.h (REGISTER_NAMES): Delete the SPE_ACC and
29147         SPEFSCR registers.
29148         * config/rs6000/rs6000.c (rs6000_reg_names, alt_reg_names): Ditto.
29149         (enum rs6000_reg_type): Delete SPE_ACC_TYPE and SPEFSCR_REG_TYPE.
29150         (rs6000_debug_reg_global): Adjust.
29151         (rs6000_init_hard_regno_mode_ok): Adjust.
29152         (rs6000_dbx_register_number): Adjust.
29153         * config/rs6000/rs6000.h (FIRST_PSEUDO_REGISTER): Change to 115.
29154         (FIXED_REGISTERS, CALL_USED_REGISTERS, CALL_REALLY_USED_REGISTERS):
29155         Remove SPE_ACC and SPEFSCR.
29156         (REG_ALLOC_ORDER): Ditto.
29157         (FRAME_POINTER_REGNUM): Change to 111.
29158         (enum reg_class): Remove the SPE_ACC and SPEFSCR registers.
29159         (REG_CLASS_NAMES): Ditto.
29160         (REG_CLASS_CONTENTS): Delete the SPE_ACC and SPEFSCR registers.
29161         (REGISTER_NAMES): Ditto.
29162         (ADDITIONAL_REG_NAMES): Ditto.
29163         (rs6000_reg_names): Ditto.
29164         * config/rs6000/rs6000.md: Renumber some register number
29165         define_constants.
29167 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29169         * config/rs6000/darwin.h (REGISTER_NAMES): Delete the SPE high
29170         registers.
29171         * config/rs6000/rs6000.c (rs6000_reg_names, alt_reg_names): Ditto.
29172         * config/rs6000/rs6000.h (FIRST_PSEUDO_REGISTER): Change from 149
29173         to 117.
29174         (DWARF_REG_TO_UNWIND_COLUMN): Do not define.
29175         (FIXED_REGISTERS, CALL_USED_REGISTERS, CALL_REALLY_USED_REGISTERS):
29176         Delete the SPE high registers.
29177         (REG_ALLOC_ORDER): Ditto.
29178         (enum reg_class): Remove SPE_HIGH_REGS.
29179         (REG_CLASS_NAMES): Ditto.
29180         (REG_CLASS_CONTENTS): Delete the SPE high registers.
29181         (REGISTER_NAMES): Ditto.
29182         (rs6000_reg_names): Ditto.
29183         * doc/tm.texi.in: Remove SPE as example.
29184         * doc/tm.texi: Regenerate.
29186 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29188         * config/rs6000/8540.md (ppc8540_brinc): Delete.
29189         * config/rs6000/e500mc.md (e500mc_brinc): Delete.
29190         * config/rs6000/e500mc64.md (e500mc64_brinc): Delete.
29191         * config/rs6000/rs6000.md (type): Remove "brinc".
29193 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29195         * config.gcc (powerpc*-*-*): Don't add spe.h to extra_headers.
29196         (powerpc*-linux*spe*): Use ${cpu_type} instead of rs6000.
29197         * config/rs6000/linuxspe.h: Delete file.
29198         * config/rs6000/rs6000.md: Don't include spe.md.
29199         * config/rs6000/spe.h: Delete file.
29200         * config/rs6000/spe.md: Delete file.
29201         * config/rs6000/t-rs6000: Remove spe.md.
29203 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29205         * config/rs6000/predicates.md (reg_or_mem_operand): Reformat.
29206         (reg_or_none500mem_operand): Delete.
29207         * config/rs6000/rs6000.md (extendsfdf2): Use reg_or_mem_operand
29208         instead of reg_or_none500mem_operand.
29210 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29212         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
29213         handling of SPE flags.
29214         * config/rs6000/rs6000.opt (-mspe, -mspe=no, -mspe=yes): Delete.
29216 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29218         * config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
29219         SPE ABI handling.
29220         * config/rs6000/paired.md (paired_negv2sf2): Rename to negv2sf2.
29221         (paired_absv2sf2, paired_addv2sf3, paired_subv2sf3, paired_mulv2sf3,
29222         paired_divv2sf3): Similar.
29223         * config/rs6000/predicates.md: Replace TARGET_SPE, TARGET_SPE_ABI,
29224         SPE_VECTOR_MODE and SPE_HIGH_REGNO_P by 0; simplify.
29225         * config/rs6000/rs6000-builtin.def: Delete RS6000_BUILTIN_E and
29226         RS6000_BUILTIN_S.
29227         Delete BU_SPE_1, BU_SPE_2, BU_SPE_3, BU_SPE_E, BU_SPE_P, and BU_SPE_X.
29228         Rename the paired_* instruction patterns.
29229         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Do not
29230         define __SPE__.
29231         * config/rs6000/rs6000-protos.h (invalid_e500_subreg): Delete.
29232         * config/rs6000/rs6000.c: Delete RS6000_BUILTIN_E and RS6000_BUILTIN_S.
29233         (struct rs6000_stack): Delete fields spe_gp_save_offset, spe_gp_size,
29234         spe_padding_size, and spe_64bit_regs_used.  Replace TARGET_SPE and
29235         TARGET_SPE_ABI with 0, simplify.  Replace SPE_VECTOR_MODE with
29236         PAIRED_VECTOR_MODE.
29237         (struct machine_function): Delete field spe_insn_chain_scanned_p.
29238         (spe_func_has_64bit_regs_p): Delete.
29239         (spe_expand_predicate_builtin): Delete.
29240         (spe_expand_evsel_builtin): Delete.
29241         (TARGET_DWARF_REGISTER_SPAN): Do not define.
29242         (TARGET_MEMBER_TYPE_FORCES_BLK): Do not define.
29243         (invalid_e500_subreg): Delete.
29244         (rs6000_legitimize_address): Always force_reg op2 as well, for
29245         paired single memory accesses.
29246         (rs6000_member_type_forces_blk): Delete.
29247         (rs6000_spe_function_arg): Delete.
29248         (rs6000_expand_unop_builtin): Delete SPE handling.
29249         (rs6000_expand_binop_builtin): Ditto.
29250         (spe_expand_stv_builtin): Delete.
29251         (bdesc_2arg_spe): Delete.
29252         (spe_expand_builtin): Delete.
29253         (spe_expand_predicate_builtin): Delete.
29254         (spe_expand_evsel_builtin): Delete.
29255         (rs6000_invalid_builtin): Remove RS6000_BTM_SPE handling.
29256         (spe_init_builtins): Delete.
29257         (spe_func_has_64bit_regs_p): Delete.
29258         (savres_routine_name): Delete "info" parameter.  Adjust callers.
29259         (rs6000_emit_stack_reset): Ditto.
29260         (rs6000_dwarf_register_span): Delete.
29261         * config/rs6000/rs6000.h (TARGET_SPE_ABI, TARGET_SPE,
29262         UNITS_PER_SPE_WORD, SPE_HIGH_REGNO_P, SPE_SIMD_REGNO_P,
29263         SPE_VECTOR_MODE, RS6000_BTM_SPE, RS6000_BUILTIN_E, RS6000_BUILTIN_S):
29264         Delete.
29265         * config/rs6000/rs6000.md (FIRST_SPE_HIGH_REGNO, LAST_SPE_HIGH_REGNO):
29266         Delete.
29267         * config/rs6000/rs6000.opt (-mabi=spe, -mabi=no-spe): Delete.
29268         * config/rs6000/spe.md: Delete every pattern that uses TARGET_SPE.
29269         * config/rs6000/vector.md (absv2sf2, negv2sf2, addv2sf3, subv2sf3,
29270         mulv2sf3, divv2sf3): Delete expanders.
29272 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29274         config/rs6000/rs6000.md (UNSPEC_MV_CR_GT): Delete.
29276 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29278         * config/rs6000/rs6000-protos.h (output_e500_flip_gt_bit): Delete.
29279         * config/rs6000/rs6000.c: Ditto.
29281 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29283         * config/rs6000/predicated.md (rs6000_cbranch_operator): Delete.
29284         * config/rs6000/rs6000.md: Replace rs6000_cbranch_operator by
29285         comparison_operator.
29287 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29289         * config/rs6000/rs6000.c: Remove everything related to -mfloat-gprs.
29290         * config/rs6000/rs6000.opt: Ditto.
29291         * config/rs6000/t-rtems: Ditto.
29293 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29295         * config/rs6000/predicates.md: Replace TARGET_E500_DOUBLE and
29296         TARGET_E500_SINGLE by 0, simplify.
29297         * config/rs6000/rs6000.c: Ditto.
29298         (rs6000_option_override_internal): Delete CHECK_E500_OPTIONS.
29299         (spe_build_register_parallel): Delete.
29300         * config/rs6000/rs6000.h: Delete TARGET_E500_SINGLE,
29301         TARGET_E500_DOUBLE, and CHECK_E500_OPTIONS.
29302         * config/rs6000/rs6000.md: Replace TARGET_E500_DOUBLE,
29303         TARGET_E500_SINGLE, and <E500_CONVERT> by 0, simplify.
29304         (E500_CONVERT): Delete.
29305         * config/rs6000/spe.md: Remove many patterns and all define_constants.
29307 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29309         * config/rs6000/darwin.md: Replace TARGET_FPRS by 1 and simplify.
29310         * config/rs6000/dfp.md: Ditto.
29311         (negdd2, *negdd2_fpr): Merge.
29312         (absdd2, *absdd2_fpr): Merge.
29313         (negtd2, *negtd2_fpr): Merge.
29314         (abstd2, *abstd2_fpr): Merge.
29315         * config/rs6000/e500.h: Delete file.
29316         * config/rs6000/predicates.md (rs6000_cbranch_operator): Replace
29317         TARGET_FPRS by 1 and simplify.
29318         * config/rs6000/rs6000-c.c: Ditto.
29319         * config/rs6000/rs6000.c: Ditto.  Also replace TARGET_SF_SPE and
29320         TARGET_DF_SPE by 0.
29321         * config/rs6000/rs6000.h: Ditto.  Delete TARGET_SF_SPE and
29322         TARGET_DF_SPE.
29323         * config/rs6000/rs6000.md: Ditto.
29324         (floatdidf2, *floatdidf2_fpr): Merge.
29325         (move_from_CR_gt_bit): Delete.
29326         * config/rs6000/spe.md: Replace TARGET_FPRS by 1 and simplify.
29327         (E500_CR_IOR_COMPARE): Delete.
29328         (All patterns that require !TARGET_FPRS): Delete.
29329         * config/rs6000/vsx.md: Replace TARGET_FPRS by 1 and simplify.
29331 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29333         * passes.def (pass_iv_canon): Move before pass_loop_distribution.
29335 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29337         * graphds.c (add_edge): Intitialize edge's attached data.
29338         (foll_in_subgraph, dfs_fst_edge, dfs_next_edge): New function
29339         pointer parameter.  Call pointed function on each edge during
29340         graph traversing.  Skip traversing the edge when the function
29341         returns true.
29342         (graphds_dfs, graphds_scc): Ditto.
29343         (for_each_edge): New parameter.  Pass the new parameter to callback
29344         function.
29345         * graphds.h (skip_edge_callback): New function pointer type.
29346         (graphds_dfs, graphds_scc): New function pointer parameter.
29347         (graphds_edge_callback, for_each_edge): New parameter.
29349 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29351         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test): Factor
29352         out code checking if runtime alias check is possible to below ...
29353         Call the new function.
29354         * tree-data-ref.c (runtime_alias_check_p): ... to new function.
29355         * tree-data-ref.h (runtime_alias_check_p): New decalaration.
29357 2017-06-07  Marek Polacek  <polacek@redhat.com>
29359         PR sanitizer/80932
29360         * fold-const.c (extract_muldiv_1) <case MINUS_EXPR>: Add
29361         TYPE_OVERFLOW_WRAPS check.
29363 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29365         * tree-vect-loop-manip.c (vect_do_peeling): Don't skip vector loop
29366         if versioning is required.
29367         * tree-vect-loop.c (vect_analyze_loop_2): Merge niter check for loop
29368         peeling with the check for versioning.
29370 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29372         * tree-vectorizer.h (vect_build_loop_niters): New parameter.
29373         * tree-vect-loop-manip.c (vect_build_loop_niters): New parameter.
29374         Set true to new parameter if new ssa variable is defined.
29375         (vect_gen_vector_loop_niters): Refactor.  Set range information
29376         for the new vector loop bound variable.
29377         (vect_do_peeling): Ditto.
29379 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29381         * tree-affine.c (ssa.h): Include header file.
29382         (tree_to_aff_combination): Handle (T1)(X - CST) when inner type
29383         has wrapping overflow behavior.
29385 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29387         * tree-affine.c (tree_to_aff_combination): Handle (T1)(X + X).
29389 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29391         (aff_combination_expand): Move (T1)(X *+- CST) simplification to ...
29392         (tree_to_aff_combination): ... here.
29394 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29396         * tree-ssa-loop-ivopts.c (ivopts_estimate_reg_pressure): New
29397         reg_pressure model function.
29398         (ivopts_global_cost_for_size): Delete.
29399         (determine_set_costs, iv_ca_recount_cost): Call new model function
29400         ivopts_estimate_reg_pressure.
29402 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29404         * config/aarch64/aarch64.c (aarch64_rtx_costs): Make sdiv more
29405         expensive than udiv.  Remove floating point cases from mod.
29407 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29409         * config/arm/aarch-cost-tables.h (cortexa53_extra_cost):
29410         Increase idiv cost.
29412 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29414         * config/aarch64/aarch64.md
29415         (copysignsf3): Fix mask generation.
29417 2017-06-07  Jakub Jelinek  <jakub@redhat.com>
29419         * dumpfile.h (enum tree_dump_index): Rename TDI_generic to
29420         TDI_gimple.
29421         (class dump_manager): Add register_dumps method.
29422         * dumpfile.c: Include langhooks.h.
29423         (dump_files): Use 0 instead of 3/4/5 for TDI_{original,gimple,nested}.
29424         (FIRST_AUTO_NUMBERED_DUMP): Decrease to 1.
29425         (FIRST_ME_AUTO_NUMBERED_DUMP): Define.
29426         (dump_manager::dump_register): Start with 512 entries instead of 32.
29427         (dump_manager::register_dumps): New method.
29428         * toplev.c (general_init): Instead of invoking register_dumps
29429         langhook, invoke register_dumps method on the dump manager.
29430         * gimplify.c (gimplify_function_tree): Use TDI_gimple instead of
29431         TDI_generic.
29433 2017-06-07  Richard Sandiford  <richard.sandiford@linaro.org>
29435         * doc/md.texi: Clarify the restrictions on a define_insn condition.
29436         Say that # requires an associated define_split to exist, and that
29437         the define_split must be suitable for use after register allocation.
29439 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29441         * cfgbuild.c (find_bb_boundaries): Initialize profile of split blocks.
29442         (compute_outgoing_frequencies): Also initialize zero counts.
29443         (find_many_sub_basic_blocks): Do not produce uninitialized profile
29444         around loops; preserve more of profile when nothing changes.
29446 2017-06-06  Jim Wilson  <jim.wilson@linaro.org>
29448         * config/aarch64/aarch64-cost-tables.h (qdf24xx_extra_costs): Move to
29449         here.
29450         * config/arm/aarch-cost-tables.h (qdf24xx_extra_costs): From here.
29451         * config/arm/arm-cpu-cdata.h: Regenerate.
29452         * config/arm/arm-cpu-data.h, config/arm/arm-cpu.h: Likewise.
29453         * config/arm/arm-tables.opt, config/arm/arm-tune.md: Likewise.
29454         * config/arm/arm-cpus.in: Delete falkor and qdf24xx entries.
29455         * config/arm/arm.c (arm_qdf24xx_tune): Delete.
29456         * config/arm/bpabi.h (BE8_LINK_SPEC): Delete falkor and qdf24xx
29457         support.
29458         * config/arm/t-aprofile (MULTILIB_MATCHES): Delete falkor and qdf24xx
29459         support.
29460         * config/arm/t-rmprofile: Likewise.
29461         * doc/invoke.texi (ARM Options): Drop falkor and qdf24xx support.
29463 2017-06-06  David S. Miller  <davem@davemloft.net>
29465         PR target/80968
29466         * config/sparc/sparc.c (sparc_expand_prologue): Emit frame
29467         blockage if function uses alloca.
29469 2017-06-06  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
29471         * tree-ssa-loop-prefetch.c (struct mem_ref_group, struct mem_ref):
29472         New "uid" fields to hold pretty-print IDs of group and ref.
29473         Memory references are now identified as <group_id>:<ref_id>
29474         instead of using [random] addresses.
29475         (dump_mem_details): Simplify, no functional change.
29476         (dump_mem_ref): Simplify and make output more concise.
29477         Replace couple of fprintf's throughout code with calls to dump_mem_ref.
29478         (find_or_create_group): Initialize group uid.
29479         (record_ref): Initialize ref uid.  Improve debug output.
29480         (prune_group_by_reuse, should_issue_prefetch_p,)
29481         (should_issue_prefetch_p, schedule_prefetches, issue_prefetch_ref,)
29482         (mark_nontemporal_store, determine_loop_nest_reuse):
29483         Improve debug output.
29485 2017-06-06  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
29487         * dbgcnt.def (prefetch): New debug counter.
29488         * tree-ssa-loop-prefetch.c (dbgcnt.h): New include.
29489         (schedule_prefetches): Stop issueing prefetches if debug counter
29490         tripped.
29492 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29494         * doc/sourcebuild.texi (Testsuites, C Language Testsuites,
29495         gcc.c-torture/compile): Remove mention of NO_LABEL_VALUES in fixme.
29497 2017-06-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29499         * config/aarch64/atomics.md (atomic_compare_and_swap<mode> expander):
29500         Use aarch64_reg_or_zero predicate for operand 4.
29501         (aarch64_compare_and_swap<mode> define_insn_and_split):
29502         Use aarch64_reg_or_zero predicate for operand 3.  Add 'Z' constraint.
29503         (aarch64_store_exclusive<mode>): Likewise for operand 2.
29505 2017-06-06  Thomas Preud'homme  <thomas.preudhomme@arm.com>
29507         * config/arm/arm.c (arm_compute_save_reg_mask): Rename into ...
29508         (arm_compute_save_core_reg_mask): This.
29509         (thumb1_compute_save_reg_mask): Rename into ...
29510         (thumb1_compute_save_core_reg_mask): This.
29511         (arm_compute_save_reg0_reg12_mask): Adapt comment.
29512         (arm_compute_frame_layout): Likewise.
29514 2017-06-06  Richard Biener  <rguenther@suse.de>
29516         PR tree-optimization/80974
29517         * tree-ssa-sccvn.c (set_ssa_val_to): Do not change but only
29518         keep or clear leaders SSA info.
29520 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29522         * config/nvptx/nvptx.c (split_mode_p): New function.
29523         (nvptx_declare_function_name, nvptx_print_operand): Use split_mode_p.
29525 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29527         * config/nvptx/nvptx.c (nvptx_print_operand): Use maybe_split_mode.
29529 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29531         PR bootstrap/80978
29532         * tree-cfg.c (execute_fixup_cfg): Fix condition on when to rescale
29533         profile.
29535 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29537         * shrink-wrap.c (handle_simple_exit): Update profile.
29538         (try_shrink_wrapping): Upate profile.
29540 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29542         * predict.c (tree_estimate_probability_bb): Add LOCAL_ONLY.
29543         (tree_guess_outgoing_edge_probabilities): New.
29544         * predict.h (tree_guess_outgoing_edge_probabilities): Declare.
29545         * tree-cfg.c (gimple_find_sub_bbs): Propagate profile.
29547 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29549         * ipa-split.c (split_function): Initialize return bb profile.
29551 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29553         * profile.c (compute_branch_probabilities): Also initialize
29554         EXIT_BLOCK profile.
29556 2017-06-06  Richard Biener  <rguenther@suse.de>
29558         PR tree-optimization/80928
29559         * tree-vect-loop.c (vect_update_vf_for_slp): Amend dumps.
29560         (vect_analyze_loop_operations): Properly guard analysis for
29561         pure SLP case.
29562         (vect_transform_loop): Likewise.
29563         (vect_analyze_loop_2): Also reset SLP type on PHIs.
29564         (vect_model_induction_cost): Do not cost for pure SLP.
29565         (vectorizable_induction): Pass in SLP node, implement SLP vectorization
29566         of induction in inner loop vectorization.
29567         * tree-vect-slp.c (vect_create_new_slp_node): Handle PHIs.
29568         (vect_get_and_check_slp_defs): Handle vect_induction_def.
29569         (vect_build_slp_tree): Likewise.  Handle PHIs as terminating the
29570         recursion.
29571         (vect_analyze_slp_cost_1): Cost induction.
29572         (vect_detect_hybrid_slp_stmts): Handle PHIs.
29573         (vect_get_slp_vect_defs): Likewise.
29574         * tree-vect-stmts.c (vect_analyze_stmt): Handle induction.
29575         (vect_transform_stmt): Handle SLP reductions.
29576         * tree-vectorizer.h (vectorizable_induction): Adjust.
29578 2017-06-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
29580         * config/rs6000/rs6000.c (make_resolver_func): Update
29581         init_lowered_empty_function call.
29583 2017-06-05  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29585         * doc/invoke.texi: Document the -fprofile-abs-path option.
29586         * common.opt (fprofile-abs-path): New option.
29587         * gcov-io.h (gcov_write_filename): Declare.
29588         * gcov-io.c (gcov_write_filename): New function.
29589         * coverage.c (coverage_begin_function): Use gcov_write_filename.
29590         * profile.c (output_location): Likewise.
29592 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29594         * shring-wrap.c: Revert accidental commit.
29596 2017-06-05  Volker Reichelt  <v.reichelt@netcologne.de>
29598         * doc/invoke.texi (-Wduplicated-branches): Add to warning list.
29600 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29602         * cfgexpand.c (expand_gimple_tailcall): Initialize profile of
29603         new edge.
29604         * ipa-inline.c (want_inline_self_recursive_call_p): Watch for missing
29605         profile in callgraph edge.
29606         * profile-count.h (apply_probability): If THIS is 0, then result is 0
29607         (apply_scale): Likewise.
29608         * tree-inline.c (copy_bb, copy_edges_for_bb, copy_cfg_body):
29609         Also scale profile when inlining function with zero profile.
29610         (initialize_cfun): Update exit block profile even when it is zero.
29611         * tree-ssa-threadupdate.c (clear_counts_path): Handle correctly case
29612         when profile is read.
29614 2017-06-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
29616         * config/rs6000/rs6000.c (toplevel): Include attribs.h.
29617         (CLONE_*): New constants to define the processors we can generate
29618         code for with the target_clone attribute.
29619         (rs6000_clone_map): New array to identify which clone processors
29620         the current program is running on.
29621         (TARGET_COMPARE_VERSION_PRIORITY): Define to enable the
29622         target_clone attribute.
29623         (TARGET_GENERATE_VERSION_DISPATCHER_BODY): Likewise.
29624         (TARGET_GET_FUNCTION_VERSIONS_DISPATCHER): Likewise.
29625         (TARGET_OPTION_FUNCTION_VERSIONS): Likewise.
29626         (cpu_expand_builtin): Add support for target_clone attribute.
29627         (rs6000_valid_attribute_p): Allow "default" attribute.
29628         (get_decl_name): New debug function to simplify printing the
29629         current function name in debugging statements.
29630         (rs6000_clone_priority): New functions to support the target_clone
29631         attribute, and be able to generate code to switch between ISA 2.05
29632         through ISA 3.0 (power6 through power9).
29633         (rs6000_compare_version_priority): Likewise.
29634         (rs6000_get_function_versions_dispatcher): Likewise.
29635         (make_resolver_func): Likewise.
29636         (add_condition_to_bb): Likewise.
29637         (dispatch_function_versions): Likewise.
29638         (rs6000_generate_version_dispatcher_body): Likewise.
29639         (rs6000_can_inline_p): Call get_decl_name for debugging usage.
29640         (fusion_gpr_load_p): Fix a spacing issue.
29641         * doc/extend.texi (Common Function Attributes): Document that the
29642         PowerPC supports the target_clone attribute.
29644 2017-06-05  Thomas Preud'homme  <thomas.preudhomme@arm.com>
29646         * config/arm/arm.h: explain F symbol found in description of ARM
29647         register allocation in its legend.
29649 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29651         * config/mips/frame-header-opt.c: Include profile-count.h.
29652         * config/riscv/riscv.c: Include profile-count.h
29654 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29656         * tree-ssa-loop-im.c (execute_sm_if_changed): Add FLAG_BBS parameter;
29657         update profile.
29658         (sm_set_flag_if_changed): Add bbs field.
29659         (execute_sm_if_changed_flag_set): Pass BBS.
29660         (execute_sm): Update.
29662 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29664         * config/aarch64/aarch64-simd.md (aarch64_store_lane0<mode>):
29665         New pattern.
29667 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29669         * config/aarch64/aarch64.md (sub<mode>3_compare1_imm): New define_insn.
29670         (peephole2): New peephole2 to emit the above.
29671         * config/aarch64/predicates.md (aarch64_sub_immediate): New predicate.
29673 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29675         * config/aarch64/aarch64.c (define_peephole2 above
29676         *sub_<shift>_<mode>): New peephole.
29678 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
29680         * config/i386/i386.c (make_resolver_func): Update.
29681         * Makefile.in: Add profile-count.h and profile-count.o
29682         * auto-profile.c (afdo_indirect_call): Update to new API.
29683         (afdo_set_bb_count): Update.
29684         (afdo_propagate_edge): Update.
29685         (afdo_propagate_circuit): Update.
29686         (afdo_calculate_branch_prob): Update.
29687         (afdo_annotate_cfg): Update.
29688         * basic-block.h: Include profile-count.h
29689         (struct edge_def): Turn count to profile_count.
29690         (struct basic_block_def): Likewie.
29691         (REG_BR_PROB_BASE): Move to profile-count.h
29692         (RDIV): Move to profile-count.h
29693         * bb-reorder.c (max_entry_count): Turn to profile_count.
29694         (find_traces): Update.
29695         (rotate_loop):Update.
29696         (connect_traces):Update.
29697         (sanitize_hot_paths):Update.
29698         * bt-load.c (migrate_btr_defs): Update.
29699         * cfg.c (RDIV): Remove.
29700         (init_flow): Use alloc_block.
29701         (alloc_block): Uninitialize count.
29702         (unchecked_make_edge): Uninitialize count.
29703         (check_bb_profile): Update.
29704         (dump_edge_info): Update.
29705         (dump_bb_info): Update.
29706         (update_bb_profile_for_threading): Update.
29707         (scale_bbs_frequencies_int): Update.
29708         (scale_bbs_frequencies_gcov_type): Update.
29709         (scale_bbs_frequencies_profile_count): New.
29710         * cfg.h (update_bb_profile_for_threading): Update.
29711         (scale_bbs_frequencies_profile_count): Declare.
29712         * cfgbuild.c (compute_outgoing_frequencies): Update.
29713         (find_many_sub_basic_blocks): Update.
29714         * cfgcleanup.c (try_forward_edges): Update.
29715         (try_crossjump_to_edge): Update.
29716         * cfgexpand.c (expand_gimple_tailcall): Update.
29717         (construct_exit_block): Update.
29718         * cfghooks.c (verify_flow_info): Update.
29719         (dump_bb_for_graph): Update.
29720         (split_edge): Update.
29721         (make_forwarder_block): Update.
29722         (duplicate_block): Update.
29723         (account_profile_record): Update.
29724         * cfgloop.c (find_subloop_latch_edge_by_profile): Update.
29725         (get_estimated_loop_iterations): Update.
29726         * cfgloopanal.c (expected_loop_iterations_unbounded): Update.
29727         (single_likely_exit): Update.
29728         * cfgloopmanip.c (scale_loop_profile): Update.
29729         (loopify): Update.
29730         (set_zero_probability): Update.
29731         (lv_adjust_loop_entry_edge): Update.
29732         * cfgrtl.c (force_nonfallthru_and_redirect): Update.
29733         (purge_dead_edges): Update.
29734         (rtl_account_profile_record): Update.
29735         * cgraph.c (cgraph_node::create): Uninitialize count.
29736         (symbol_table::create_edge): Uninitialize count.
29737         (cgraph_update_edges_for_call_stmt_node): Update.
29738         (cgraph_edge::dump_edge_flags): Update.
29739         (cgraph_node::dump): Update.
29740         (cgraph_edge::maybe_hot_p): Update.
29741         * cgraph.h: Include profile-count.h
29742         (create_clone), create_edge, create_indirect_edge): Update.
29743         (cgraph_node): Turn count to profile_count.
29744         (cgraph_edge0: Likewise.
29745         (make_speculative, clone): Update.
29746         (create_edge): Update.
29747         (init_lowered_empty_function): Update.
29748         * cgraphclones.c (cgraph_edge::clone): Update.
29749         (duplicate_thunk_for_node): Update.
29750         (cgraph_node::create_clone): Update.
29751         * cgraphunit.c (cgraph_node::analyze): Update.
29752         (cgraph_node::expand_thunk): Update.
29753         * final.c (dump_basic_block_info): Update.
29754         * gimple-streamer-in.c (input_bb): Update.
29755         * gimple-streamer-out.c (output_bb): Update.
29756         * graphite.c (print_global_statistics): Update.
29757         (print_graphite_scop_statistics): Update.
29758         * hsa-brig.c: Include basic-block.h.
29759         * hsa-dump.c: Include basic-block.h.
29760         * hsa-gen.c (T sum_slice): Update.
29761         (convert_switch_statements):Update.
29762         * hsa-regalloc.c: Include basic-block.h.
29763         * ipa-chkp.c (chkp_produce_thunks): Update.
29764         * ipa-cp.c (struct caller_statistics): Update.
29765         (init_caller_stats): Update.
29766         (gather_caller_stats): Update.
29767         (ipcp_cloning_candidate_p): Update.
29768         (good_cloning_opportunity_p): Update.
29769         (get_info_about_necessary_edges): Update.
29770         (dump_profile_updates): Update.
29771         (update_profiling_info): Update.
29772         (update_specialized_profile): Update.
29773         (perhaps_add_new_callers): Update.
29774         (decide_about_value): Update.
29775         (ipa_cp_c_finalize): Update.
29776         * ipa-devirt.c (struct odr_type_warn_count): Update.
29777         (struct decl_warn_count): Update.
29778         (struct final_warning_record): Update.
29779         (possible_polymorphic_call_targets): Update.
29780         (ipa_devirt): Update.
29781         * ipa-fnsummary.c (redirect_to_unreachable): Update.
29782         * ipa-icf.c (sem_function::merge): Update.
29783         * ipa-inline-analysis.c (do_estimate_edge_time): Update.
29784         * ipa-inline.c (compute_uninlined_call_time): Update.
29785         (compute_inlined_call_time): Update.
29786         (want_inline_small_function_p): Update.
29787         (want_inline_self_recursive_call_p): Update.
29788         (edge_badness): Update.
29789         (lookup_recursive_calls): Update.
29790         (recursive_inlining): Update.
29791         (inline_small_functions): Update.
29792         (dump_overall_stats): Update.
29793         (dump_inline_stats): Update.
29794         * ipa-profile.c (ipa_profile_generate_summary): Update.
29795         (ipa_propagate_frequency): Update.
29796         (ipa_profile): Update.
29797         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
29798         * ipa-utils.c (ipa_merge_profiles): Update.
29799         * loop-doloop.c (doloop_modify): Update.
29800         * loop-unroll.c (report_unroll): Update.
29801         (unroll_loop_runtime_iterations): Update.
29802         * lto-cgraph.c (lto_output_edge): Update.
29803         (lto_output_node): Update.
29804         (input_node): Update.
29805         (input_edge): Update.
29806         (merge_profile_summaries): Update.
29807         * lto-streamer-in.c (input_cfg): Update.
29808         * lto-streamer-out.c (output_cfg): Update.
29809         * mcf.c (create_fixup_graph): Update.
29810         (adjust_cfg_counts): Update.
29811         (sum_edge_counts): Update.
29812         * modulo-sched.c (sms_schedule): Update.
29813         * postreload-gcse.c (eliminate_partially_redundant_load): Update.
29814         * predict.c (maybe_hot_count_p): Update.
29815         (probably_never_executed): Update.
29816         (dump_prediction): Update.
29817         (combine_predictions_for_bb): Update.
29818         (propagate_freq): Update.
29819         (handle_missing_profiles): Update.
29820         (counts_to_freqs): Update.
29821         (rebuild_frequencies): Update.
29822         (force_edge_cold): Update.
29823         * predict.h: Include profile-count.h
29824         (maybe_hot_count_p, counts_to_freqs): UPdate.
29825         * print-rtl-function.c: Do not include cfg.h
29826         * print-rtl.c: Include basic-block.h
29827         * profile-count.c: New file.
29828         * profile-count.h: New file.
29829         * profile.c (is_edge_inconsistent): Update.
29830         (correct_negative_edge_counts): Update.
29831         (is_inconsistent): Update.
29832         (set_bb_counts): Update.
29833         (read_profile_edge_counts): Update.
29834         (compute_frequency_overlap): Update.
29835         (compute_branch_probabilities): Update; Initialize and deinitialize
29836         gcov_count tables.
29837         (branch_prob): Update.
29838         * profile.h (bb_gcov_counts, edge_gcov_counts): New.
29839         (edge_gcov_count): New.
29840         (bb_gcov_count): New.
29841         * shrink-wrap.c (try_shrink_wrapping): Update.
29842         * tracer.c (better_p): Update.
29843         * trans-mem.c (expand_transaction): Update.
29844         (ipa_tm_insert_irr_call): Update.
29845         (ipa_tm_insert_gettmclone_call): Update.
29846         * tree-call-cdce.c: Update.
29847         * tree-cfg.c (gimple_duplicate_sese_region): Update.
29848         (gimple_duplicate_sese_tail): Update.
29849         (gimple_account_profile_record): Update.
29850         (execute_fixup_cfg): Update.
29851         * tree-inline.c (copy_bb): Update.
29852         (copy_edges_for_bb): Update.
29853         (initialize_cfun): Update.
29854         (freqs_to_counts): Update.
29855         (copy_cfg_body): Update.
29856         (expand_call_inline): Update.
29857         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
29858         * tree-ssa-loop-ivcanon.c (unloop_loops): Update.
29859         (try_unroll_loop_completely): Update.
29860         (try_peel_loop): Update.
29861         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
29862         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations_loop): Update.
29863         * tree-ssa-loop-split.c (connect_loops): Update.
29864         * tree-ssa-loop-unswitch.c (hoist_guard): Update.
29865         * tree-ssa-reassoc.c (branch_fixup): Update.
29866         * tree-ssa-tail-merge.c (replace_block_by): Update.
29867         * tree-ssa-threadupdate.c (create_block_for_threading): Update.
29868         (compute_path_counts): Update.
29869         (update_profile): Update.
29870         (recompute_probabilities): Update.
29871         (update_joiner_offpath_counts): Update.
29872         (estimated_freqs_path): Update.
29873         (freqs_to_counts_path): Update.
29874         (clear_counts_path): Update.
29875         (ssa_fix_duplicate_block_edges): Update.
29876         (duplicate_thread_path): Update.
29877         * tree-switch-conversion.c (case_bit_test_cmp): Update.
29878         (struct switch_conv_info): Update.
29879         * tree-tailcall.c (decrease_profile): Update.
29880         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
29881         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
29882         * value-prof.c (check_counter): Update.
29883         (gimple_divmod_fixed_value): Update.
29884         (gimple_mod_pow2): Update.
29885         (gimple_mod_subtract): Update.
29886         (gimple_ic_transform): Update.
29887         (gimple_stringop_fixed_value): Update.
29888         * value-prof.h (gimple_ic): Update.
29890 2017-06-02  Carl Love  <cel@us.ibm.com>
29892         * config/rs6000/rs6000-c: Add support for built-in functions
29893         vector double vec_doublee (vector signed int);
29894         vector double vec_doublee (vector unsigned int);
29895         vector double vec_doublee (vector float);
29896         vector double vec_doubleh (vector signed int);
29897         vector double vec_doubleh (vector unsigned int);
29898         vector double vec_doubleh (vector float);
29899         vector double vec_doublel (vector signed int);
29900         vector double vec_doublel (vector unsigned int);
29901         vector double vec_doublel (vector float);
29902         vector double vec_doubleo (vector signed int);
29903         vector double vec_doubleo (vector unsigned int);
29904         vector double vec_doubleo (vector float);.
29905         * config/rs6000/rs6000-builtin.def: Add definitions for DOUBLEE,
29906         DOUBLEO, DOUBLEH, DOUBLEL, UNS_DOUBLEO, UNS_DOUBLEE, UNS_DOUBLEH,
29907         UNS_DOUBLEL.
29908         * config/rs6000/altivec.md: Add code generator for doublee<mode>2,
29909         unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2, doubleh<mode>2,
29910         unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2, add mode attribute
29911         VS_sxwsp.
29912         * config/rs6000/altivec.h: Add define for vec_doublee, vec_doubleo,
29913         vec_doublel, vec_doubleh.
29914         * doc/extend.texi: Update the built-in documentation file for the
29915         new built-in functions.
29917 2017-06-02  David Malcolm  <dmalcolm@redhat.com>
29919         PR jit/80954
29920         * ipa-inline-analysis.c (free_growth_caches): Set
29921         edge_removal_hook_holder to NULL after removing it.
29923 2017-06-02  Sudakshina Das  <sudi.das@arm.com>
29925         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Return CC_SWP for
29926         comparision with zero.
29928 2017-06-02  Will Schmidt  <will_schmidt@vnet.ibm.com>
29929         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
29930         for early expansion of vec_min and vec_max builtins.
29931         (builtin_function_type): Add min/max unsigned variants to those
29932         identified as having unsigned arguments.
29934 2017-06-02  Olivier Hainque  <hainque@adacore.com>
29936         * config/vx-common.h (DWARF_UNWIND_INFO): Switch #define to 1.
29938 2017-06-02  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29940         * config/aarch64/aarch64-simd.md (*aarch64_simd_vec_copy_lane<mode>):
29941         Use VALL_F16 iterator rather than VALL.
29943 2017-06-02  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29945         * config/aarch64/aarch64.c (aarch64_split_compare_and_swap):
29946         Emit CBNZ inside loop when doing a strong exchange and comparing
29947         against zero.  Generate the CC flags after the loop.
29949 2017-06-02  David Edelsohn  <dje.gcc@gmail.com>
29951         * dwarf2out.c (DWARF_INITIAL_LENGTH_SIZE_STR): New.
29952         (dl_section_ref): New.
29953         (dwarf2out_finish): Copy debug_line_section_label to dl_section_ref.
29954         On AIX, append an expression to subtract the size of the
29955         section length to dl_section_ref.
29957 2017-06-02  Will Schmidt  <will_schmidt@vnet.ibm.com>
29959         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
29960         for early expansion of vector absolute builtins.
29962 2017-06-02  Richard Biener  <rguenther@suse.de>
29964         * tree-vect-slp.c (vect_detect_hybrid_slp_2): Match up
29965         what we consider a relevant use stmt with vect_detect_hybrid_slp_stmts.
29967 2017-06-02  Richard Biener  <rguenther@suse.de>
29969         PR tree-optimization/80948
29970         * tree-tailcall.c (find_tail_calls): Track stmts to move in
29971         stmt order as well.
29973 2017-06-02  Richard Biener  <rguenther@suse.de>
29975         * tree-vect-loop.c (vect_analyze_loop_operations): Not relevant
29976         PHIs are ok.
29977         * tree-vect-stmts.c (process_use): Do not mark backedge defs
29978         for inductions as relevant.
29980 2017-06-02  Richard Biener  <rguenther@suse.de>
29982         * tree-vect-loop.c (get_initial_def_for_induction): Inline into ...
29983         (vectorizable_induction): ... this.  Remove dead code.
29985 2017-06-02  Eric Botcazou  <ebotcazou@adacore.com>
29987         * builtins. (expand_builtin_alloca): Remove second parameter and
29988         infer its value from the first parameter instead.
29989         (expand_builtin) <BUILT_IN_ALLOCA>: Adjust call to above.
29991 2017-06-02  Jakub Jelinek  <jakub@redhat.com>
29993         PR rtl-optimization/80903
29994         * loop-doloop.c (add_test): Unshare sequence.
29996 2017-06-02  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29998         * doc/invoke.texi: Document the -Wsizeof-pointer-div warning.
30000 2017-06-01  Bernd Edlinger  <bernd.edlinger@hotmail.de>
30002         * config/i386/i386.c (x86_64_ms_sysv_extra_clobbered_registers): Make
30003         static.
30004         (xlogue_layout::get_stack_space_used, xlogue_layout::s_instances,
30005         xlogue_layout::get_instance, logue_layout::xlogue_layout,
30006         sp_valid_at, fp_valid_at, choose_basereg): Formatting.
30007         (xlogue_layout::get_stub_rtx): Make static.
30008         (xlogue_layout::get_stub_name): Avoid const-cast, make static.
30009         (xlogue_layout::compute_stub_managed_regs): Rename to...
30010         (xlogue_layout::count_stub_managed_regs): ...this.
30011         (xlogue_layout::is_stub_managed_reg): New function.
30012         (xlogue_layout::m_stub_names): Rename to...
30013         (xlogue_layout::s_stub_names): ...this, make static.
30014         (xlogue_layout::STUB_INDEX_OFFSET, xlogue_layout::MIN_REGS,
30015         xlogue_layout::MAX_REGS, xlogue_layout::MAX_EXTRA_REGS,
30016         xlogue_layout::VARIANT_COUNT, xlogue_layout::STUB_NAME_MAX_LEN,
30017         xlogue_layout::s_stub_names): Instantiate statics.
30018         (stub_managed_regs): Remove.
30019         (ix86_save_reg): Use xlogue_layout::compute_stub_managed_regs.
30020         (disable_call_ms2sysv_xlogues): Rename to...
30021         (warn_once_call_ms2sysv_xlogues): ...this, and warn only once.
30022         (ix86_initial_elimination_offset, ix86_expand_call): Fix call_ms2sysv
30023         warning logic.
30024         (ix86_static_chain): Make sure that ix86_static_chain_on_stack can't
30025         change after reload_completed.
30026         (ix86_can_use_return_insn_p): Use the ix86_frame data structure
30027         directly.
30028         (ix86_expand_prologue): Likewise.
30029         (ix86_expand_epilogue): Likewise.
30030         (ix86_expand_split_stack_prologue): Likewise.
30031         (ix86_compute_frame_layout): Remove frame parameter ...
30032         (TARGET_COMPUTE_FRAME_LAYOUT): ... and export it as a target hook.
30033         (ix86_finalize_stack_realign_flags): Call ix86_compute_frame_layout
30034         only if necessary.
30035         (ix86_init_machine_status): Don't set use_fast_prologue_epilogue_nregs.
30036         (ix86_frame): Move from here ...
30037         * config/i386/i386.h (ix86_frame): ... to here.
30038         (machine_function): Remove use_fast_prologue_epilogue_nregs, cache the
30039         complete ix86_frame data structure instead.  Remove some_ld_name.
30041 2017-06-01  Pierre-Marie de Rodat  <derodat@adacore.com>
30043         * dwarf2out.c (dwarf2out_late_global_decl): Add locations for
30044         symbols that hold a DECL_VALUE_EXPR.
30046 2017-06-01  Martin Jambor  <mjambor@suse.cz>
30048         PR tree-optimization/80898
30049         * tree-sra.c (process_subtree_disqualification): Removed.
30050         (disqualify_candidate): Do not acll
30051         process_subtree_disqualification.
30052         (subtree_mark_written_and_enqueue): New function.
30053         (propagate_all_subaccesses): Set grp_write of LHS subtree if the
30054         RHS has been disqualified and re-queue LHS if necessary.  Apart
30055         from that, ignore disqualified RHS.
30057 2017-06-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
30059         * config/s390/s390.c (s390_emit_epilogue): Disable early return
30060         address fetch for z10 or later.
30062 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30064         * config/arc/arc.md (tst_movb): Add guard when splitting.
30066 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30068         * config/arc/arc.c (arc_can_eliminate): Test against
30069         arc_frame_pointer_needed.
30071 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30073         * config/arc/arc.c (arc_expand_prologue): Emit a special barrier
30074         to prevent store reordering.
30075         * config/arc/arc.md (UNSPEC_ARC_STKTIE): Define.
30076         (type): Add block type.
30077         (stack_tie): Define special instruction to be used in
30078         expand_prologue.
30080 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30082         * config/arc/arc.md (commutative_binary_comparison): Remove 'I'
30083         constraint. It is not valid for the pattern.
30084         (noncommutative_binary_comparison): Likewise.
30086 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30088         * config/arc/simdext.md (movv2hi_insn): Change predicate to avoid
30089         scaled addresses.
30091 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30093         * config/arc/arc.c (arc_conditional_register_usage): Allow r30 to
30094         be used by the reg-alloc.
30096 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30098         * config/arc/arc.md (mulsi3): Avoid use of hard registers before
30099         reg-alloc when having mul64 or mul32x16 instructions.
30100         (mulsidi3): Likewise.
30101         (umulsidi3): Likewise.
30102         (mulsi32x16): New pattern.
30103         (mulsi64): Likewise.
30104         (mulsidi64): Likewise.
30105         (umulsidi64): Likewise.
30106         (MUL32x16_REG): Define.
30107         (mul64_600): Use MUL32x16_REG.
30108         (mac64_600): Likewise.
30109         (umul64_600): Likewise.
30110         (umac64_600): Likewise.
30112 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30114         * config/arc/arc.md (mulsi3_700): Make it commutative.
30116 2017-06-01  Jose E. Marchesi  <jose.marchesi@oracle.com>
30118         * config/sparc/sparc.md (*zero_extendsidi2_insn_sp64): Set insn
30119         type for movstouw.
30120         (*sign_extendsidi2_insn): Likewise for movstosw.
30122 2017-06-01  Pierre-Marie de Rodat  <derodat@adacore.com>
30124         * dwarf2out.c (get_discr_value): Call the get_debug_type hook on
30125         the type of the input discriminant value.  Convert the
30126         discriminant value of signedness vary.
30128 2017-06-01  Volker Reichelt  <v.reichelt@netcologne.de>
30130         * doc/invoke.texi (-Wcatch-value): Document new shortcut.
30131         Add to -Wall section.
30133 2017-06-01  Richard Biener  <rguenther@suse.de>
30135         PR middle-end/66313
30136         * fold-const.c (fold_plusminus_mult_expr): If the factored
30137         factor may be zero use a wrapping type for the inner operation.
30138         * tree-tailcall.c (independent_of_stmt_p): Pass in to_move bitmap
30139         and handle moved defs.
30140         (process_assignment): Properly guard the unary op case.  Return a
30141         tri-state indicating that moving the stmt before the call may allow
30142         to continue.  Pass through to_move.
30143         (find_tail_calls): Handle moving unrelated defs before
30144         the call.
30146 2017-05-31  Segher Boessenkool  <segher@kernel.crashing.org>
30148         PR target/80618
30149         * config/rs6000/vector.md (*vector_uneq<mode>): Write the nor in the
30150         splitter result in the canonical way.
30152 2017-05-31  Uros Bizjak  <ubizjak@gmail.com>
30154         * config/i386/i386.md (*zero_extendsidi2): Enable alternative (?r, *Yj)
30155         also for 32bit target.  Update insn attributes.
30156         (zero-extendsidi2 splitter): Allow all registers for operand 1.
30158 2017-05-31  Sebastian Peryt  <sebastian.peryt@intel.com>
30160         * config/i386/avx512fintrin.h (_mm_mask_max_sd)
30161         (_mm_maskz_max_sd, _mm_mask_max_ss, _mm_maskz_max_ss)
30162         (_mm_mask_min_sd, _mm_maskz_min_sd, _mm_mask_min_ss)
30163         (_mm_maskz_min_ss): New intrinsics.
30165 2017-05-31  Martin Liska  <mliska@suse.cz>
30167         * tree-vect-loop.c (vect_create_epilog_for_reduction):
30168         Change comment style to one we normally use.
30169         (vectorizable_reduction): Likewise.
30170         (vectorizable_induction): Likewise.
30171         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
30172         (vectorizable_call): Likewise.
30173         (vectorizable_simd_clone_call): Likewise.
30174         (vectorizable_conversion): Likewise.
30175         (vectorizable_assignment): Likewise.
30176         (vectorizable_shift): Likewise.
30177         (vectorizable_operation): Likewise.
30178         (vectorizable_store): Likewise.
30179         (vectorizable_load): Likewise.
30180         * tree-vectorizer.h: Likewise.
30182 2017-05-31  Alexander Monakov  <amonakov@ispras.ru>
30184         * passes.c (emergency_dump_function): New.
30185         * tree-pass.h (emergency_dump_function): Declare.
30186         * plugin.c (plugins_internal_error_function): Remove.
30187         * plugin.h (plugins_internal_error_function): Remove declaration.
30188         * toplev.c (internal_error_function): New static function.  Use it...
30189         (general_init): ...here.
30191 2017-05-31  Graham Markall  <graham.markall@embecosm.com>
30193         * config/arc/arc.c (arc_print_operand): Handle constant operands.
30194         (arc_rtx_costs): Add costs for new patterns.
30195         * config/arc/arc.md: Additional *add_n and *sub_n patterns.
30196         * config/arc/predicates.md: Add _1_2_3_operand predicate.
30198 2017-05-31  Richard Sandiford  <richard.sandiford@linaro.org>
30200         * tree-ssa-strlen.c (get_next_strinfo): New function.
30201         (get_stridx_plus_constant): Use it.
30202         (zero_length_string): Likewise.
30203         (adjust_related_strinfos): Likewise.
30204         (adjust_last_stmt): Likewise.
30206 2017-05-31  Richard Biener  <rguenther@suse.de>
30208         PR target/80880
30209         * config/i386/i386.c (ix86_expand_builtin): Remove assert
30210         for arg being an SSA name when expanding IX86_BUILTIN_BNDRET.
30212 2017-05-31  Richard Sandiford  <richard.sandiford@linaro.org>
30214         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Remove
30215         loop_vinfo argument and use of dependence distance vectors.
30216         Check instead whether the two references differ only in their
30217         initial value and assume that they have the same alignment if the
30218         difference is a multiple of the vector alignment.
30219         (vect_analyze_data_refs_alignment): Update call accordingly.
30221 2017-05-31  Martin Liska  <mliska@suse.cz>
30223         PR target/79155
30224         * config/i386/cpuid.h: Fix typo in a comment in cpuid.h.
30226 2017-05-31  Bin Cheng  <bin.cheng@arm.com>
30228         * tree-vect-loop-manip.c (create_intersect_range_checks_index)
30229         (create_intersect_range_checks): Move from ...
30230         * tree-data-ref.c (create_intersect_range_checks_index)
30231         (create_intersect_range_checks): ... to here.
30232         (create_runtime_alias_checks): New function factored from ...
30233         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks): ...
30234         here.  Call above function.
30235         * tree-data-ref.h (create_runtime_alias_checks): New function.
30237 2017-05-31  Bin Cheng  <bin.cheng@arm.com>
30239         * tree-data-ref.c (prune_runtime_alias_test_list): Relax minimal
30240         segment length for dr_b and compute it in wide_int.
30242 2017-05-31  Richard Biener  <rguenther@suse.de>
30244         PR tree-optimization/80906
30245         * graphite-isl-ast-to-gimple.c (copy_loop_close_phi_nodes): Get
30246         and pass through iv_map.
30247         (copy_bb_and_scalar_dependences): Adjust.
30248         (translate_pending_phi_nodes): Likewise.
30249         (copy_loop_close_phi_args): Handle code-generating IVs instead
30250         of ICEing.
30252 2017-05-30  David Malcolm  <dmalcolm@redhat.com>
30254         * diagnostic-color.c (color_dict): Add "type-diff".
30255         (parse_gcc_colors): Update comment.
30256         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
30257         -fdiagnostics-show-template-tree and -fno-elide-type.
30258         (GCC_COLORS): Add type-diff to example.
30259         (type-diff=): New.
30260         (-fdiagnostics-show-template-tree): New.
30261         (-fno-elide-type): New.
30262         * pretty-print.c (pp_format): Pass quote and formatters[argno] to
30263         the pp_format_decoder callback.  Call any m_format_postprocessor's
30264         "handle" method.
30265         (pretty_printer::pretty_printer): Initialize
30266         m_format_postprocessor.
30267         (pretty_printer::~pretty_printer): Delete any
30268         m_format_postprocessor.
30269         * pretty-print.h (printer_fn): Add bool and const char ** parameters.
30270         (class format_postprocessor): New class.
30271         (struct pretty_printer::format_decoder): Document the new parameters.
30272         (struct pretty_printer::m_format_postprocessor): New field.
30273         * tree-diagnostic.c (default_tree_printer): Update for new
30274         bool and const char ** params.
30275         * tree-diagnostic.h (default_tree_printer): Likewise.
30277 2017-05-30  Segher Boessenkool  <segher@kernel.crashing.org>
30279         * config/rs6000/predicates.md (cc_reg_not_micro_cr0_operand): Delete.
30280         (lwa_operand): Delete rs6000_gen_cell_microcode test.
30281         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
30282         rs6000_gen_cell_microcode code.
30283         (rs6000_final_prescan_insn): Delete.
30284         (rs6000_opt_vars): Delete the "gen-cell-microcode" and
30285         "warn-cell-microcode" entries.
30286         * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Delete.
30287         * config/rs6000/rs6000.md: Delete rs6000_gen_cell_microcode tests
30288         throughout.  Change cc_reg_not_micro_cr0_operand to
30289         cc_reg_not_cr0_operand throughout.
30290         (*extendhi<mode>2_noload): Delete.
30291         * config/rs6000/rs6000.opt (mgen-cell-microcode): Replace by stub.
30292         (mwarn-cell-microcode): Delete.
30293         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete
30294         -mgen-cell-microcode and -mwarn-cell-microcode.
30296 2017-05-30  Uros Bizjak  <ubizjak@gmail.com>
30298         PR target/80833
30299         * config/i386/constraints.md (Yd): New constraint.
30300         (Ye): Ditto.
30301         * config/i386/i386.md (*movti_internal): Add (?r, Ye)
30302         and (?Yd, r) alternatives.  Update insn attributes.
30303         * config/i386/i386.md (*movti_internal): Add (?r, *Ye)
30304         and (?*Yd, r) alternatives.  Update insn attributes.
30305         (double-mode inter-unit splitters): Add new GR<->XMM splitters.
30307 2017-05-30  Pierre-Marie de Rodat  <derodat@adacore.com>
30309         * gimplify.c (gimplify_modify_expr): Don't create a
30310         DECL_DEBUG_EXPR link if *FROM_P does not belong to the current
30311         function.
30313 2017-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
30315         * config/arm/arm-builtins.c (arm_expand_builtin): Remove const.
30317 2017-05-30  Richard Biener  <rguenther@suse.de>
30319         * tree-vectorizer.h (struct _stmt_vec_info): Add reduc_type
30320         and reduc_def fields.
30321         (STMT_VINFO_REDUC_TYPE): New define.
30322         (STMT_VINFO_REDUC_DEF): Likewise.
30323         (vect_force_simple_reduction): Adjust prototype.
30324         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Adjust.
30325         (vect_is_simple_reduction): Remove check_reduction argument.
30326         (vect_force_simple_reduction): Adjust and set
30327         STMT_VINFO_REDUC_TYPE and STMT_VINFO_REDUC_DEF.
30328         (vectorizable_reduction): Do not re-do reduction analysis
30329         but use STMT_VINFO_REDUC_TYPE and STMT_VINFO_REDUC_DEF.
30330         * tree-parloops.c (gather_scalar_reductions): Adjust.
30332 2017-05-30  Richard Biener  <rguenther@suse.de>
30334         PR middle-end/80901
30335         * cfgexpand.c (expand_gimple_cond): Match up loop fixup with
30336         split_edge code.
30338 2017-05-24  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30340         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs):
30341         Introduce unknown_misalignment parameter and remove vf.
30342         (vect_peeling_hash_get_lowest_cost):
30343         Pass unknown_misalignment parameter.
30344         (vect_enhance_data_refs_alignment):
30345         Fix unsupportable data ref treatment.
30347 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30349         * tree-vect-data-refs.c (vect_get_data_access_cost):
30350         Workaround for SLP handling.
30351         (vect_enhance_data_refs_alignment):
30352         Compute costs for doing no peeling at all, compare to the best
30353         peeling costs so far and avoid peeling if cheaper.
30355 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30357         * tree-vect-data-refs.c (vect_peeling_hash_choose_best_peeling):
30358         Return peeling info and set costs to zero for unlimited cost
30359         model.
30360         (vect_enhance_data_refs_alignment): Also inspect all datarefs
30361         with unknown misalignment. Compute and costs for unknown
30362         misalignment, compare them to the costs for known misalignment
30363         and choose the cheapest for peeling.
30365 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30367         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Rename.
30368         (vect_get_peeling_costs_all_drs): Create function.
30369         (vect_peeling_hash_get_lowest_cost):
30370         Use vect_get_peeling_costs_all_drs.
30371         (vect_peeling_supportable): Create function.
30372         (vect_enhance_data_refs_alignment): Use vect_peeling_supportable.
30374 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30376         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Create
30377         DR_HAS_NEGATIVE_STEP.
30378         (vect_update_misalignment_for_peel): Define DR_MISALIGNMENT.
30379         (vect_enhance_data_refs_alignment): Use.
30380         (vect_duplicate_ssa_name_ptr_info): Use.
30381         * tree-vectorizer.h (dr_misalignment): Use.
30382         (known_alignment_for_access_p): Use.
30384 2017-05-30  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
30386         PR target/78838
30387         * config/msp430/msp430.c (gen_prefix): Return NULL when section name is
30388         .lowtext.
30389         (has_section_name): New function.
30391 2017-05-30  Martin Liska  <mliska@suse.cz>
30393         PR other/80909
30394         * auto-profile.c (get_function_decl_from_block): Fix
30395         parenthesis.
30397 2017-05-30  Richard Biener  <rguenther@suse.de>
30399         PR middle-end/80876
30400         * cfgexpand.c (expand_gimple_cond): Fixup preserving loops again.
30402 2017-05-30  Martin Liska  <mliska@suse.cz>
30404         * dumpfile.c: Use newly added macro DUMP_FILE_INFO.
30405         * dumpfile.h (struct dump_file_info): Remove ctors.
30407 2017-05-30  Martin Liska  <mliska@suse.cz>
30409         * predict.def: Fix GNU coding style.
30411 2017-05-29  Max Filippov  <jcmvbkbc@gmail.com>
30413         * config/xtensa/xtensa.c (xtensa_initial_elimination_offset):
30414         Mark 'to' argument with ATTRIBUTE_UNUSED.
30416 2017-05-29  Max Filippov  <jcmvbkbc@gmail.com>
30418         * config/xtensa/xtensa.c (xtensa_emit_call): Use
30419         HOST_WIDE_INT_PRINT_HEX instead of 0x%lx format string.
30420         (print_operand): Use HOST_WIDE_INT_PRINT_DEC instead of %ld
30421         format string.
30423 2017-05-29  Eric Botcazou  <ebotcazou@adacore.com>
30425         * doc/install.texi (Options specification): Restore entry of
30426         --enable-sjlj-exceptions.
30428 2017-05-27  Michael Eager  <eager@eagercon.com>
30430         Revert:
30431         2016-01-21  Ajit Agarwal  <ajitkum@xilinx.com>
30433         See https://gcc.gnu.org/ml/gcc/2017-05/msg00221.html.
30435         * config/microblaze/microblaze.h
30436         (FIXED_REGISTERS): Update in macro.
30437         (CALL_USED_REGISTERS): Update in macro.
30439 2017-05-27  François-Xavier Coudett  <fxcoudert@gcc.gnu.org>
30441         * doc/install.texi: Add links to macOS binary distributions.
30443 2017-05-27  Jakub Jelinek  <jakub@redhat.com>
30445         PR bootstrap/80887
30446         Revert:
30447         2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30449         * match.pd ((A +- CST1) +- CST2): Allow some conversions.
30451 2017-05-26  Martin Liska  <mliska@suse.cz>
30453         * dumpfile.h (enum dump_kind): Renumber TDF_* flags to be contiguous.
30455 2017-05-26  Martin Liska  <mliska@suse.cz>
30457         * cfg.c (check_bb_profile): Do not use TDF_COMMENT and print
30458         always leading ';; '.
30459         (dump_bb_info): Likewise.
30460         (brief_dump_cfg): Likewise.
30461         * cfgrtl.c (print_rtl_with_bb): Do not use TDF_COMMENT.
30462         * dumpfile.c: Remove usage of TDF_VERBOSE.
30463         * dumpfile.h (enum dump_kind): Likewise.
30464         (dump_gimple_bb_header): Do not use TDF_COMMENT.
30465         * print-tree.c (debug_verbose): Remove.
30466         * tree-cfg.c (gimple_dump_cfg): Do not use TDF_COMMENT.
30467         (dump_function_to_file): Remove dumps guarded with TDF_VERBOSE.
30468         * tree-diagnostic.c (default_tree_printer): Replace
30469         TDF_DIAGNOSTIC with TDF_SLIM.
30471 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30473         * tree-vect-loop-manip.c (create_intersect_range_checks_index): Pass
30474         in parameter loop, rather than loop_vinfo.
30475         (create_intersect_range_checks): Ditto.
30476         (vect_create_cond_for_alias_checks): Update call to above functions.
30478 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30480         PR tree-optimization/80815
30481         * tree-data-ref.c (prune_runtime_alias_test_list): Simplify condition
30482         for merging runtime alias checks.  Handle negative DR_STEPs.
30484 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30486         * tree-vect-data-refs.c (Operator==, comp_dr_with_seg_len_pair):
30487         Move from ...
30488         * tree-data-ref.c (Operator==, comp_dr_with_seg_len_pair): To here.
30489         * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list): Factor
30490         out code pruning runtime alias checks.
30491         * tree-data-ref.c (prune_runtime_alias_test_list): New function
30492         factored out from above.
30493         * tree-vectorizer.h (struct dr_with_seg_len, dr_with_seg_len_pair_t):
30494         Move from ...
30495         * tree-data-ref.h (struct dr_with_seg_len, dr_with_seg_len_pair_t):
30496         ... to here.
30497         (prune_runtime_alias_test_list): New decalaration.
30499 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30501         * tree-vect-data-refs.c (compare_tree): Rename and move ...
30502         * tree-data-ref.c (data_ref_compare_tree): ... to here.
30503         * tree-data-ref.h (data_ref_compare_tree): New decalaration.
30504         * tree-vect-data-refs.c (dr_group_sort_cmp): Update uses.
30505         (operator==, comp_dr_with_seg_len_pair): Ditto.
30506         (vect_prune_runtime_alias_test_list): Ditto.
30508 2017-05-26  Martin Liska  <mliska@suse.cz>
30510         PR ipa/80663
30511         * params.def: Bound partial-inlining-entry-probability param.
30513 2017-05-26  Marek Polacek  <polacek@redhat.com>
30515         PR sanitizer/80875
30516         * fold-const.c (fold_binary_loc) <case MULT_EXPR>: Check if OP1
30517         can be negated.
30519 2017-05-26  Richard Biener  <rguenther@suse.de>
30521         PR tree-optimization/80842
30522         * tree-ssa-ccp.c (set_lattice_value): Always meet with the old
30523         value.
30525 2017-05-26  Richard Biener  <rguenther@suse.de>
30527         PR tree-optimization/80844
30528         * tree-vectorizer.c (adjust_simduid_builtins): Propagate results.
30530 2017-05-25  Sebastian Peryt  <sebastian.peryt@intel.com>
30532         * doc/md.texi (Machine Constraints): Update x86 family
30533         machine constraints section to match 'config/i386/constraints.md'.
30535 2017-05-25  Volker Reichelt  <v.reichelt@netcologne.de>
30537         * doc/invoke.texi (-Wcatch-value=): Document new warning option.
30539 2017-05-25  Nathan Sidwell  <nathan@acm.org>
30541         * doc/invoke.texi (--enable-languages): Update documentation.
30543 2017-05-25  Martin Liska  <mliska@suse.cz>
30545         * dumpfile.c: Add TDF_FOLDING.
30546         * dumpfile.h (enum dump_kind): Likewise.
30547         * genmatch.c (dt_simplify::gen_1): Use it.
30549 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30551         * match.pd (view_convert (convert@0 @1)): Handle zero-extension.
30553 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30555         * match.pd ((A +- CST1) +- CST2): Allow some conversions.
30556         * tree.c (drop_tree_overflow): Handle COMPLEX_CST and VECTOR_CST.
30558 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30560         * fold-const.c (fold_binary_loc) [(A & C) == D]: Remove transformation.
30561         * match.pd (X == C): Rewrite it here.
30562         (with_possible_nonzero_bits, with_possible_nonzero_bits2,
30563         with_certain_nonzero_bits2): New predicates.
30564         * tree-ssanames.c (get_nonzero_bits): Handle INTEGER_CST.
30566 2017-05-24  Nathan Sidwell  <nathan@acm.org>
30568         * lto-streamer-in.c (lto_input_data_block): Adjust T const cast to
30569         avoid warning.
30571         * auto-profile.c (afdo_propagate): Adjust T const cast to avoid
30572         warning.
30574 2017-05-24  Segher Boessenkool  <segher@kernel.crashing.org>
30576         * config/powerpcspe: New port.  Files are copied from the rs6000
30577         port, with "rs6000" in filenames replaced by "powerpcspe".
30579 2017-05-24  Wilco Dijkstra  <wdijkstr@arm.com>
30581         PR rtl-optimization/80754
30582         * lra-remat.c (do_remat): Add overlap checks for dst_regno.
30584 2017-05-24  Sheldon Lobo  <smlobo@sheldon.us.oracle.com>
30586         * config/sparc/sparc.md (length): Return the correct value for -mflat
30587         sibcalls to match output_sibcall.
30589 2017-05-24  Segher Boessenkool  <segher@kernel.crashing.org>
30591         PR bootstrap/80860
30592         PR bootstrap/80843
30593         * config/rs6000/rs6000.c (struct machine_function): Add new field
30594         n_components.
30595         (rs6000_get_separate_components): Init that field, use it.
30596         (rs6000_components_for_bb): Use the field.
30598 2017-05-24  Thomas Preud'homme  <thomas.preudhomme@arm.com>
30600         * config/arm/arm.c (arm_expand_prologue): Fix typo in comment.
30602 2017-05-24  Peter Bergner  <bergner@vnet.ibm.com>
30604         PR middle-end/80823
30605         * tree-cfg.c (group_case_labels_stmt): Delete increment of "i";
30607 2017-05-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
30609         PR target/80725
30610         * config/s390/s390.c (s390_check_qrst_address): Check incoming
30611         address against address_operand predicate.
30612         * config/s390/s390.md ("*indirect_jump"): Swap alternatives.
30614 2017-05-24  Eric Botcazou  <ebotcazou@adacore.com>
30616         * var-tracking.c (track_expr_p): Do not return 0 for tracked record
30617         parameters passed indirectly.
30619 2017-05-23  Uros Bizjak  <ubizjak@gmail.com>
30621         * config/i386/i386.md (*movdi_internal): Remove SSE4
30622         alternative 18 (?r, *v).  Update insn attributes.
30623         (*movsi_internal): Remove SSE4 alternative 13 (?r, *v).
30624         Update insn attributes.
30625         (*zero_extendsidi2): Remove SSE4 alternative (?r, *x).
30626         Update insn attributes.
30627         * config/i386/sse.md (vec_extract<ssevecmodelower>_0): Remove SSE4
30628         alternative 1 (r, v). Remove isa attribute.
30629         * config/i386/i386.c (dimode_scalar_chain::make_vector_copies):
30630         Always move value through stack for !TARGET_INTER_UNIT_MOVES_TO_VEC
30631         and !TARGET_INTER_UNIT_MOVES_TO_VEC targets.
30633 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30635         * doc/sourcebuild.texi (Directives, Verify compiler message): Document
30636         dg-line directive.
30638 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30640         * cgraphunit.c (symbol_table::process_new_functions): Update.
30641         * ipa-fnsummary.c (pass_data_inline_parameters): Remove.
30642         (inline_generate_summary): Rename to ...
30643         (ipa_fn_summary_generate): ... this one.
30644         (inline_read_summary): Rename to ...
30645         (ipa_fn_summary_read): ... this one.
30646         (inline_write_summary): Rename to ...
30647         (ipa_fn_summary_write): ... this one.
30648         (inline_free_summary): Rename to ...
30649         (ipa_free_fn_summary): ... this one.
30650         (pass_data_local_fn_summary, pass_local_fn_summary,
30651         make_pass_local_fn_summary, pass_data_ipa_free_fn_summary,
30652         pass_ipa_free_fn_summary, make_pass_ipa_free_fn_summary,
30653         pass_data_ipa_fn_summary, pass_ipa_fn_summary,
30654         make_pass_ipa_fn_summary): New.
30655         * ipa-fnsummary.h (inline_generate_summary, inline_read_summary,
30656         inline_write_summary, inline_free_summary): Remove.
30657         (ipa_free_fn_summary) : New.
30658         * ipa-inline.c (ipa_inline): Update.
30659         (pass_ipa_inline): Do not generate summaries.
30660         * ipa.c (pass_data_ipa_free_fn_summary, pass_ipa_free_fn_summary):
30661         Remove.
30662         * passes.def: Replace pass_inline_parameters by pass_local_fn_summary
30663         and add pass_ipa_fn_summary.
30664         * tree-pass.h (make_pass_ipa_fn_summary, make_pass_local_fn_summary):
30665         New.
30666         (make_pass_inline_parameters): Remove.
30668 2017-05-23  Thomas Schwinge  <thomas@codesourcery.com>
30670         * omp-low.c (struct omp_context): Remove "default_kind" member.
30671         Adjust all users.
30673         * omp-offload.c (execute_oacc_device_lower): Remove the
30674         parallelism dimensions function attributes for unparallelized
30675         OpenACC kernels constructs.
30677 2017-05-23  Martin Liska  <mliska@suse.cz>
30679         * cgraph.c (cgraph_node::get_create): Use symtab_node::dump_{asm_,}name
30680         functions.
30681         (cgraph_edge::make_speculative): Likewise.
30682         (cgraph_edge::resolve_speculation): Likewise.
30683         (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
30684         (cgraph_node::dump): Likewise.
30685         * cgraph.h: Likewise.
30686         * cgraphunit.c (analyze_functions): Likewise.
30687         (symbol_table::compile): Likewise.
30688         * ipa-cp.c (print_all_lattices): Likewise.
30689         (determine_versionability): Likewise.
30690         (initialize_node_lattices): Likewise.
30691         (ipcp_verify_propagated_values): Likewise.
30692         (estimate_local_effects): Likewise.
30693         (update_profiling_info): Likewise.
30694         (create_specialized_node): Likewise.
30695         (perhaps_add_new_callers): Likewise.
30696         (decide_about_value): Likewise.
30697         (decide_whether_version_node): Likewise.
30698         (identify_dead_nodes): Likewise.
30699         (ipcp_store_bits_results): Likewise.
30700         * ipa-devirt.c (dump_targets): Likewise.
30701         (ipa_devirt): Likewise.
30702         * ipa-icf.c (sem_item::dump): Likewise.
30703         (sem_function::equals): Likewise.
30704         (sem_variable::equals): Likewise.
30705         (sem_item_optimizer::read_section): Likewise.
30706         (sem_item_optimizer::execute): Likewise.
30707         (congruence_class::dump): Likewise.
30708         * ipa-inline-analysis.c (dump_ipa_call_summary): Likewise.
30709         (dump_inline_summary): Likewise.
30710         (estimate_node_size_and_time): Likewise.
30711         (inline_analyze_function): Likewise.
30712         * ipa-inline-transform.c (inline_call): Likewise.
30713         * ipa-inline.c (report_inline_failed_reason): Likewise.
30714         (want_early_inline_function_p): Likewise.
30715         (edge_badness): Likewise.
30716         (update_edge_key): Likewise.
30717         (inline_small_functions): Likewise.
30718         * ipa-profile.c (ipa_profile): Likewise.
30719         * ipa-prop.c (ipa_print_node_jump_functions): Likewise.
30720         (ipa_make_edge_direct_to_target): Likewise.
30721         (remove_described_reference): Likewise.
30722         (ipa_impossible_devirt_target): Likewise.
30723         (propagate_controlled_uses): Likewise.
30724         (ipa_print_node_params): Likewise.
30725         (ipcp_transform_function): Likewise.
30726         * ipa-pure-const.c (pure_const_read_summary): Likewise.
30727         (propagate_pure_const): Likewise.
30728         * ipa-reference.c (generate_summary): Likewise.
30729         (read_write_all_from_decl): Likewise.
30730         (propagate): Likewise.
30731         (ipa_reference_read_optimization_summary): Likewise.
30732         * ipa-utils.c (ipa_merge_profiles): Likewise.
30733         * ipa.c (walk_polymorphic_call_targets): Likewise.
30734         (symbol_table::remove_unreachable_nodes): Likewise.
30735         (ipa_single_use): Likewise.
30736         * passes.c (execute_todo): Likewise.
30737         * predict.c (drop_profile): Likewise.
30738         * symtab.c (symtab_node::get_dump_name): New function.
30739         (symtab_node::dump_name): Likewise.
30740         (symtab_node::dump_asm_name): Likewise.
30741         (symtab_node::dump_references): Likewise.
30742         (symtab_node::dump_referring): Likewise.
30743         (symtab_node::dump_base): Likewise.
30744         (symtab_node::debug_symtab): Likewise.
30745         * tree-sra.c (convert_callers_for_node): Likewise.
30746         * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
30747         * value-prof.c (init_node_map): Likewise.
30749 2017-05-23  Martin Liska  <mliska@suse.cz>
30751         * cgraph.h: Move symtab_node::dump_table to symbol_table::dump
30752         and symtab_node::debug_symtab to symbol_table::debug.
30753         * cgraphunit.c (analyze_functions): Use the renamed function.
30754         (symbol_table::compile): Likewise.
30755         * ipa-cp.c (ipcp_verify_propagated_values): Likewise.
30756         * ipa-icf.c (sem_item_optimizer::execute): Likewise.
30757         * passes.c (execute_todo): Likewise.
30758         * symtab.c (symbol_table::dump): New function.
30759         * tree-ssa-structalias.c (ipa_pta_execute): Use the renamed function.
30761 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30763         * ipa-fnsummary.c (estimate_node_size_and_time): Do not sanity check
30764         that nonconst implies exec.
30766 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30768         * ipa-inline-analysis.c (cgraph_2edge_hook_list, cgraph_edge_hook_list,
30769         inline_edge_removal_hook, inline_edge_duplication_hook): Remove.
30770         (inline_edge_summary_vec): Turn into ...
30771         (ipa_call_summaries): ... this one.
30772         (redirect_to_unreachable, edge_set_predicate,
30773         evaluate_properties_for_edge, inline_summary_alloc,
30774         reset_ipa_call_summary, reset_inline_summary,
30775         inline_summary_t::duplicate): Update.
30776         (inline_edge_duplication_hook): Turn to ...
30777         (ipa_call_summary_t::duplicate): ... this one.
30778         (inline_edge_removal_hook): Turn to ...
30779         (ipa_call_summary_t::remove): ... this one.
30780         (dump_inline_edge_summary): Turn to ...
30781         (dump_ipa_call_summary): ... this one.
30782         (estimate_function_body_sizes): Update.
30783         (inline_update_callee_summaries): Update.
30784         (remap_edge_change_prob): Update.
30785         (remap_edge_summaries): Update.
30786         (inline_merge_summary): Update.
30787         (do_estimate_edge_time): Update.
30788         (inline_generate_summary): Update.
30789         (inline_read_section): Update.
30790         (inline_read_summary): Update.
30791         (inline_free_summary): Update.
30792         * ipa-inline.c (can_inline_edge_p): Update.
30793         (compute_inlined_call_time): Update.
30794         (want_inline_small_function_p): Update.
30795         (edge_badness): Update.
30796         (early_inliner): Update.
30797         * ipa-inline.h (inline_edge_summary): Turn to ...
30798         (ipa_call_summary): ... this one.
30799         (ipa_call_summary_t): New class.
30800         (inline_edge_summary_t, inline_edge_summary_vec): Remove.
30801         (ipa_call_summaries): New.
30802         (inline_edge_summary): Remove.
30803         (estimate_edge_growth): Update.
30804         * ipa-profile.c (ipa_propagate_frequency_1): Update.
30805         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
30806         * ipa-split.c (execute_split_functions): Update.
30807         * ipa.c (symbol_table::remove_unreachable_nodes): Update.
30809 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30811         * doc/sourcebuild.texi (Effective-Target Keywords, Other hardware
30812         attributes): Document rdrand effective target.
30814 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30816         * doc/sourcebuild.texi (Effective-Target Keywords, Other hardware
30817         attributes): Sort alphabetically.
30819 2017-05-23  Georg-Johann Lay  <avr@gjlay.de>
30821         * config/avr/genmultilib.awk: Use gsub instead of gensub.
30823 2017-05-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
30825         PR target/80718
30826         * config/rs6000/vsx.md (vsx_splat_<mode>, VSX_D iterator): Split
30827         V2DF/V2DI splat into two separate patterns, one that handles
30828         registers, and the other that only handles memory.  Drop support
30829         for splatting from a GPR on ISA 2.07 and then splitting the
30830         splat into direct move and splat.
30831         (vsx_splat_<mode>_reg): Likewise.
30832         (vsx_splat_<mode>_mem): Likewise.
30834 2017-05-22  Segher Boessenkool  <segher@kernel.crashing.org>
30836         * cfgcleanup.c (bb_is_just_return): Allow CLOBBERs.
30838 2017-05-22  Jakub Jelinek  <jakub@redhat.com>
30840         PR middle-end/80809
30841         * omp-low.c (finish_taskreg_remap): New function.
30842         (finish_taskreg_scan): If unit size of ctx->record_type
30843         is non-constant, unshare the size expression and replace
30844         decls in it with possible outer var refs.
30846         PR middle-end/80809
30847         * gimplify.c (omp_add_variable): For GOVD_DEBUG_PRIVATE use
30848         GOVD_SHARED rather than GOVD_PRIVATE with it.
30849         (gimplify_adjust_omp_clauses_1, gimplify_adjust_omp_clauses): Expect
30850         GOVD_SHARED rather than GOVD_PRIVATE with GOVD_DEBUG_PRIVATE.
30852         PR middle-end/80853
30853         * omp-low.c (lower_reduction_clauses): Pass OMP_CLAUSE_PRIVATE
30854         as last argument to build_outer_var_ref for pointer bases of array
30855         section reductions.
30857 2017-05-19  Martin Sebor  <msebor@redhat.com>
30859         * print-tree.c (print_node): Print DECL_READ_P flag.
30861 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30863         * Makefile.in: Add ipa-fnsummary.o and ipa-fnsummary.h
30864         * auto-profile.c: Replace ipa-inline.h by ipa-fnsummary.h
30865         * cgraph.c: Likewise.
30866         * cgraphunit.c: Likewise.
30867         * gengtype.c: Likewise.
30868         * ipa-cp.c: Likewise.
30869         * ipa-devirt.c: Likewise.
30870         * ipa-icf.c: Likewise.
30871         * ipa-predicate.c: Likewise.
30872         * ipa-profile.c: Likewise.
30873         * ipa-prop.c: Likewise.
30874         * ipa-split.c: Likewise.
30875         * ipa.c: Likewise.
30876         * ipa-inline-analysis.c (inline_summaries, ipa_call_summaries,
30877         edge_predicate_pool, dump_inline_hints,
30878         inline_summary::account_size_time, redirect_to_unreachable,
30879         edge_set_predicate, set_hint_predicate,
30880         evaluate_conditions_for_known_args, evaluate_properties_for_edge,
30881         inline_summary_alloc, ipa_call_summary::reset, inline_summary::reset,
30882         inline_summary_t::remove, remap_hint_predicate_after_duplication,
30883         inline_summary_t::duplicate, ipa_call_summary_t::duplicate,
30884         ipa_call_summary_t::remove, initialize_growth_caches,
30885         free_growth_caches, dump_ipa_call_summary, dump_inline_summary,
30886         debug_inline_summary, dump_inline_summaries, initialize_inline_failed,
30887         mark_modified, unmodified_parm_1, unmodified_parm,
30888         unmodified_parm_or_parm_agg_item, eliminated_by_inlining_prob,
30889         set_cond_stmt_execution_predicate, set_switch_stmt_execution_predicate,
30890         compute_bb_predicates, will_be_nonconstant_expr_predicate,
30891         will_be_nonconstant_predicate, record_modified_bb_info,
30892         get_minimal_bb, record_modified, param_change_prob,
30893         phi_result_unknown_predicate, predicate_for_phi_result,
30894         array_index_predicate, clobber_only_eh_bb_p, fp_expression_p,
30895         estimate_function_body_sizes, compute_inline_parameters,
30896         compute_inline_parameters_for_curren, pass_data_inline_parameters,
30897         estimate_node_size_and_time, estimate_ipcp_clone_size_and_time,
30898         inline_update_callee_summaries, remap_edge_change_prob,
30899         remap_edge_summaries, remap_hint_predicate, inline_merge_summary,
30900         inline_update_overall_summary, inline_indirect_intraprocedural_analysis,
30901         inline_analyze_function, inline_summary_t::insert,
30902         inline_generate_summary, read_ipa_call_summary, inline_read_section,
30903         inline_read_summary, write_ipa_call_summary, inline_write_summary,
30904         inline_free_summary): Move to ipa-fnsummary.h
30905         (predicate_t): Remove.
30906         * ipa-fnsummary.c: New file.
30907         * ipa-inline.h:  Do not include sreal.h and ipa-predicate.h
30908         (enum inline_hints_vals, inline_hints, agg_position_info,
30909         INLINE_SIZE_SCALE, size_time_entry, inline_summary, inline_summary_t,
30910         inline_summaries, ipa_call_summary, ipa_call_summary_t,
30911         ipa_call_summaries, debug_inline_summary, dump_inline_summaries,
30912         dump_inline_summary, dump_inline_hints, inline_generate_summary,
30913         inline_read_summary, inline_write_summary, inline_free_summary,
30914         inline_analyze_function, initialize_inline_failed,
30915         inline_merge_summary, inline_update_overall_summary,
30916         compute_inline_parameters): Move to ipa-fnsummary.h
30917         * ipa-fnsummary.h: New file.
30918         * ipa-inline-transform.h: Include ipa-inline.h.
30919         * ipa-inline.c: LIkewise.
30921 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30923         * ipa-inline.c (edge_badness): Use inlined_time instead of
30924         inline_summaries->get.
30926 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30928         * ipa-inline.c (edge_badness): Use estimate_size_after_inlining.
30930 2017-05-22  Nathan Sidwell  <nathan@acm.org>
30932         * doc/invoke.texi (fdump-translation-unit): Delete documentation.
30933         (fdump-lang): Document 'raw' option.
30934         * dumpfile.h (TDI_tu): Delete.
30935         * dumpfile.c (dump_files): Remove translation-unit.
30936         (FIRST_AUTO_NUMBERED_DUMP): Decrement.
30938 2017-05-22  Georg-Johann Lay  <avr@gjlay.de>
30940         * config/avr/t-avr (AWK) [t-multilib]: Remove "-v FORMAT=Makefile"
30941         command option from $(AWK) call.
30942         * config/avr/genmultilib.awk: Simplify and rewrite so that it
30943         generates MULTILIB_REQUIRED instead of MULTILIB_EXCEPTIONS.
30944         [FORMAT]: Remove handling of variable.
30945         * config/avr/t-multilib: Regenerate.
30947 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30949         * ipa-inline-analysis.c (inline_summary::reset): Do not reset
30950         self_time.
30951         (dump_inline_summary): Do not print self_time.
30952         (estimate_function_body_sizes): Do not set self_time.
30953         (compute_inline_parameters): Likewise.
30954         (inline_read_section, inline_write_summary): Do not stream self_time.
30955         * ipa-inline.h (inline_summary): Drop self_time.
30957 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30959         * ipa-inline-analysis.c (account_size_time): Rename to ...
30960         (inline_summary::account_size_time): ... this one.
30961         (reset_ipa_call_summary): Turn to ...
30962         (ipa_call_summary::reset): ... this one.
30963         (reset_inline_summary): Turn to ...
30964         (inline_summary::reset): ... this one.
30965         (inline_summary_t::remove): Update.
30966         (inline_summary_t::duplicate): Update.
30967         (ipa_call_summary_t::remove): Update.
30968         (dump_inline_summary): Update.
30969         (estimate_function_body_sizes): Update.
30970         (compute_inline_parameters): Update.
30971         (estimate_node_size_and_time): Update.
30972         (inline_merge_summary): Update.
30973         (inline_update_overall_summary): Update.
30974         (inline_read_section): Update.
30975         (inline_write_summary): Update.
30976         * ipa-inline.h (inline_summary): Rename entry to size_time_table;
30977         add account_size_time and reset member functions.
30978         (ipa_call_summary): Add reset function.
30979         * ipa-predicate.h (predicate::operator &): Constify.
30981 2017-05-22  Richard Biener  <rguenther@suse.de>
30983         * df-scan.c (df_insn_refs_verify): Speedup when not verifying.
30985 2017-05-19  Jason Merrill  <jason@redhat.com>
30987         * tree.c (make_tree_vec_stat, grow_tree_vec_stat): Use size_t.
30989 2017-05-19  Marek Polacek  <polacek@redhat.com>
30991         PR sanitizer/80800
30992         * fold-const.c (extract_muldiv_1) <case TRUNC_DIV_EXPR>: Add
30993         TYPE_OVERFLOW_WRAPS checks.
30995 2017-05-19  Thomas Schwinge  <thomas@codesourcery.com>
30997         * tree-core.h (enum omp_clause_default_kind): Add
30998         "OMP_CLAUSE_DEFAULT_PRESENT".
30999         * tree-pretty-print.c (dump_omp_clause): Handle it.
31000         * gimplify.c (enum gimplify_omp_var_data): Add
31001         "GOVD_MAP_FORCE_PRESENT".
31002         (gimplify_adjust_omp_clauses_1): Map it to
31003         "GOMP_MAP_FORCE_PRESENT".
31004         (oacc_default_clause): Handle "OMP_CLAUSE_DEFAULT_PRESENT".
31006         * gimplify.c (oacc_default_clause): Clarify.
31008 2017-05-19  Nathan Sidwell  <nathan@acm.org>
31010         LANG_HOOK_REGISTER_DUMPS
31011         * toplev.c (general_init): Call register dump lang hook.
31012         * doc/invoke.texi: Document -fdump-lang option family.
31013         * dumpfile.c (dump_files): Remove class dump here.
31014         (FIRST_AUTO_NUMBERED_DUMP): Adjust.
31015         * dumpfile.h (tree_dump_index): Remove TDI_class.
31016         * langhooks-def.h (lhd_register_dumps): Declare.
31017         (LANG_HOOKS_REGISTER_DUMPS): Define.
31018         (LANG_HOOKS_INITIALIZER): Add it.
31019         * langhooks.c (lhd_register_dumps): Define.
31020         * langhooks.h (struct lang_hooks): Add register_dumps.
31022 2017-05-19  Nathan Sidwell  <nathan@acm.org>
31024         * context.h (context::set_passes): New.
31025         * context.c (context::context): Do not create pass manager.
31026         * toplev.c (general_init): Create pass manager here.
31028 2017-05-19  Segher Boessenkool  <segher@kernel.crashing.org>
31030         * config/rs6000/rs6000.md (splitter to load of -1 and mask): Don't
31031         use this splitter if two add or or instructions would also work for
31032         the constant we want to generate.
31034 2017-05-19  Richard Biener  <rguenther@suse.de>
31036         PR build/80821
31037         * genmatch.c (dt_node::gen_kids_1): Add missing scope around
31038         predicate evaluation.
31040 2017-05-19  Jan Hubicka  <hubicka@ucw.cz>
31042         * ipa-inline.h (ipa_call_summary): Turn sizes into signed;
31043         add ctor.
31044         * ipa-inline.c (want_inline_small_function_p): Do not cast to
31045         unsigned.
31047 2017-05-19  Jan Hubicka  <hubicka@ucw.cz>
31049         * ipa-inline-analysis.c (cgraph_2edge_hook_list, cgraph_edge_hook_list,
31050         inline_edge_removal_hook, inline_edge_duplication_hook): Remove.
31051         (inline_edge_summary_vec): Turn into ...
31052         (ipa_call_summaries): ... this one.
31053         (redirect_to_unreachable, edge_set_predicate,
31054         evaluate_properties_for_edge, inline_summary_alloc,
31055         reset_ipa_call_summary, reset_inline_summary,
31056         inline_summary_t::duplicate): Update.
31057         (inline_edge_duplication_hook): Turn to ...
31058         (ipa_call_summary_t::duplicate): ... this one.
31059         (inline_edge_removal_hook): Turn to ...
31060         (ipa_call_summary_t::remove): ... this one.
31061         (dump_inline_edge_summary): Turn to ...
31062         (dump_ipa_call_summary): ... this one.
31063         (estimate_function_body_sizes): Update.
31064         (inline_update_callee_summaries): Update.
31065         (remap_edge_change_prob): Update.
31066         (remap_edge_summaries): Update.
31067         (inline_merge_summary): Update.
31068         (do_estimate_edge_time): Update.
31069         (inline_generate_summary): Update.
31070         (inline_read_section): Update.
31071         (inline_read_summary): Update.
31072         (inline_free_summary): Update.
31073         * ipa-inline.c (can_inline_edge_p): Update.
31074         (compute_inlined_call_time): Update.
31075         (want_inline_small_function_p): Update.
31076         (edge_badness): Update.
31077         (early_inliner): Update.
31078         * ipa-inline.h (inline_edge_summary): Turn to ...
31079         (ipa_call_summary): ... this one.
31080         (ipa_call_summary_t): New class.
31081         (inline_edge_summary_t, inline_edge_summary_vec): Remove.
31082         (ipa_call_summaries): New.
31083         (inline_edge_summary): Remove.
31084         (estimate_edge_growth): Update.
31085         * ipa-profile.c (ipa_propagate_frequency_1): Update.
31086         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
31087         * ipa-split.c (execute_split_functions): Update.
31088         * ipa.c (symbol_table::remove_unreachable_nodes): Update.
31090 2017-05-19  Richard Biener  <rguenther@suse.de>
31092         PR middle-end/80764
31093         * cfgexpand.c (expand_gimple_cond): Fix loop fixup.
31095 2017-05-18  Segher Boessenkool  <segher@kernel.crashing.org>
31097         * config/rs6000/rs6000.c (struct machine_function): Add field
31098         fpr_is_wrapped_separately.
31099         (rs6000_get_separate_components): Use 64 components.  Handle the
31100         new FPR components.
31101         (rs6000_components_for_bb): Handle the FPR components.
31102         (rs6000_emit_prologue_components): Handle the FPR components.
31103         (rs6000_emit_epilogue_components): Handle the FPR components.
31104         (rs6000_set_handled_components): Handle the FPR components.
31105         (rs6000_emit_prologue): Don't output prologue code for those FPRs
31106         that are already separately shrink-wrapped.
31107         (rs6000_emit_epilogue): Don't output epilogue code for those FPRs
31108         that are already separately shrink-wrapped.
31110 2017-05-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
31112         PR target/80510
31113         * config/rs6000/predicates.md (simple_offsettable_mem_operand):
31114         New predicate.
31116         * config/rs6000/rs6000.md (ALTIVEC_DFORM): New iterator.
31117         (define_peephole2 for Altivec d-form load): Add peepholes to catch
31118         cases where the register allocator uses a move and an offsettable
31119         memory operation to/from a FPR register on ISA 2.06/2.07.
31120         (define_peephole2 for Altivec d-form store): Likewise.
31122 2017-05-18  Uros Bizjak  <ubizjak@gmail.com>
31124         PR target/80799
31125         * config/i386/mmx.md (*mov<mode>_internal): Enable
31126         alternatives 11, 12, 13 and 14 also for 32bit targets.
31127         Remove alternatives 15, 16, 17 and 18.
31128         * config/i386/sse.md (vec_concatv2di): Change
31129         alternative (!x, *y) to (x, ?!*Yn).
31131 2017-05-18  Paolo Carlini  <paolo.carlini@oracle.com>
31133         * dumpfile.h (enum dump_kind): Remove stray comma.
31135 2017-05-18  Jan Hubicka  <hubicka@ucw.cz>
31137         * Makefile.in: Add ipa-predicate.o and ipa-predicate.h
31138         * ipa-inline-analysis.c (NUM_CONDITIONS): turn into
31139         predicate::num_conditions
31140         (IS_NOT_CONSTANT): turn into predicate::is_not_constant.
31141         (CHANGED): turn into predicate::changed.
31142         (agg_position_info): Move to ipa-predicate.h
31143         (add_condition, predicate::add_clause, predicate::operator &=,
31144         predicate::or_with, predicate::evaluate, predicate::probability,
31145         dump_condition, dump_clause, predicate::dump,
31146         predicate::remap_after_duplication, predicate::remap_after_inlining,
31147         predicate::stream_in, predicate::stream_out): Move to ipa-predicate.c
31148         (evaluate_conditions_for_known_args): Update.
31149         (set_cond_stmt_execution_predicate): Update.
31150         * ipa-inline.h: Include ipa-predicate.h
31151         (condition, inline_param_summary, conditions, agg_position_info,
31152         predicate): Move to ipa-predicate.h
31153         * ipa-predicate.c: New file.
31154         * ipa-predicate.h: New file.
31156 2017-05-18  Wilco Dijkstra  <wdijkstr@arm.com>
31158         * final.c (leaf_function_p): Check we are not in a sequence.
31160 2017-05-18  Martin Liska  <mliska@suse.cz>
31162         * cfgrtl.c (rtl_verify_edges): Remove usage of TDF_RTL.
31163         * dumpfile.c (dump_register): Use new enum dump_kind.
31164         (get_dump_file_name): Likewise.
31165         (dump_enable_all): Likewise.
31166         (dump_switch_p_1): Likewise.
31167         (enable_rtl_dump_file): Remove usage of TDF_RTL.
31168         * dumpfile.h (enum dump_kind): New enum type.
31169         (struct dump_file_info): Create constructor and
31170         format fields and comments.
31171         * passes.c (pass_manager::register_one_dump_file):
31172         Use num dump_kind.
31173         * statistics.c (statistics_early_init): Likewise.
31174         * tree-ssa-loop-prefetch.c (dump_mem_details): Replace
31175         TDF_TREE with TDF_SLIM.
31176         (gather_memory_references_ref): Likewise.
31178 2017-05-18  Martin Liska  <mliska@suse.cz>
31180         * vec.h (struct vnull): Use it.
31182 2017-05-18  Jan Hubicka  <hubicka@ucw.cz>
31184         * ipa-inline-analysis.c (predicate_conditions): Move to ipa-inline.h
31185         (true_predicate, false_predicate, true_predicate_p,
31186         false_predicate_p): Remove.
31187         (single_cond_predicate, not_inlined_predicate): Turn to member function
31188         in ipa-inline.h
31189         (add_condition): Update.
31190         (add_clause): Turn to...
31191         (predicate::add_clause): ... this one; update; allow passing NULL
31192         as parameter.
31193         (and_predicates): Turn to ...
31194         (predicate::operator &=): ... this one.
31195         (predicates_equal_p): Move to predicate::operator == in ipa-inline.h
31196         (or_predicates): Turn to ...
31197         (predicate::or_with): ... this one.
31198         (evaluate_predicate): Turn to ...
31199         (predicate::evaluate): ... this one.
31200         (predicate_probability): Turn to ...
31201         (predicate::probability): ... this one.
31202         (dump_condition): Update.
31203         (dump_predicate): Turn to ...
31204         (predicate::dump): ... this one.
31205         (account_size_time): Update.
31206         (edge_set_predicate): Update.
31207         (set_hint_predicate): UPdate.
31208         (evaluate_conditions_for_known_args): Update.
31209         (evaluate_properties_for_edge): Update.
31210         (remap_predicate_after_duplication): Turn to...
31211         (predicate::remap_after_duplication): ... this one.
31212         (remap_hint_predicate_after_duplication): Update.
31213         (inline_summary_t::duplicate): UPdate.
31214         (dump_inline_edge_summary): Update.
31215         (dump_inline_summary): Update.
31216         (set_cond_stmt_execution_predicate): Update.
31217         (set_switch_stmt_execution_predicate): Update.
31218         (compute_bb_predicates): Update.
31219         (will_be_nonconstant_expr_predicate): Update.
31220         (will_be_nonconstant_predicate): Update.
31221         (phi_result_unknown_predicate): Update.
31222         (predicate_for_phi_result): Update.
31223         (array_index_predicate): Update.
31224         (estimate_function_body_sizes): Update.
31225         (estimate_node_size_and_time): Update.
31226         (estimate_ipcp_clone_size_and_time): Update.
31227         (remap_predicate): Rename to ...
31228         (predicate::remap_after_inlining): ... this one.
31229         (remap_hint_predicate): Update.
31230         (inline_merge_summary): Update.
31231         (inline_update_overall_summary): Update.
31232         (estimate_size_after_inlining): Update.
31233         (read_predicate): Rename to ...
31234         (predicate::stream_in): ... this one.
31235         (read_inline_edge_summary): Update.
31236         (write_predicate): Rename to ...
31237         (predicate::stream_out): ... this one.
31238         (write_inline_edge_summary): Update.
31239         * ipa-inline.h (MAX_CLAUSES): Turn to predicate::max_clauses.
31240         (clause_t): Turn to uint32_t
31241         (predicate): Turn to class; implement constructor and operators
31242         ==, !=, &
31243         (size_time_entry): Update.
31244         (inline_summary): Update.
31245         (inline_edge_summary): Update.
31247 2017-05-18  Marc Glisse  <marc.glisse@inria.fr>
31249         * fold-const.c (fold_binary_loc): Move transformation...
31250         * match.pd (C - X CMP X): ... here.
31252 2017-05-18  Sheldon Lobo  <sheldon.lobo@oracle.com>
31254         * config/sparc/sparc.c (sparc_option_override): Set function
31255         alignment for -mcpu=niagara7 to 64 to match the I$ line.
31256         * config/sparc/sparc.h (BRANCH_COST): Set the SPARC M7 branch
31257         latency to 1.
31258         * config/sparc/sparc.h (BRANCH_COST): Set the SPARC T4 branch
31259         latency to 2.
31260         * config/sparc/sol2.h: Fix a ASM_CPU32_DEFAULT_SPEC typo.
31262 2017-05-18  Marek Polacek  <polacek@redhat.com>
31264         PR sanitizer/80797
31265         * ubsan.c (instrument_null): Unwrap ADDR_EXPRs.
31266         (pass_ubsan::execute): Call gimple_assign_single_p instead of
31267         gimple_assign_load_p.
31269 2017-05-17  Segher Boessenkool  <segher@kernel.crashing.org>
31271         PR middle-end/80692
31272         * real.c (do_compare): Give decimal_do_compare preference over
31273         comparing just the signs.
31275 2017-05-17  Uros Bizjak  <ubizjak@gmail.com>
31277         * doc/md.texi (Canonicalization of Instructions): Describe the
31278         canonical form of instructions that inherently set a condition
31279         code register.
31281 2017-05-17  Peter Bergner  <bergner@vnet.ibm.com>
31283         PR middle-end/80775
31284         * tree-cfg.c: Move deletion of unreachable case statements to after
31285         the merging of consecutive case labels.
31287 2017-05-17  Thomas Preud'homme  <thomas.preudhomme@arm.com>
31289         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Refer
31290         readers to __gnu_cmse_nonsecure_call libcall for saving, clearing and
31291         restoring of callee-saved registers.
31293 2017-05-17  Eric Botcazou  <ebotcazou@adacore.com>
31295         * compare-elim.c (equivalent_reg_at_start): Adjust outdated comment.
31296         * config/visium/visium.c (single_set_and_flags): Likewise.
31297         * config/visium/visium.md (Substitutions): Likewise.
31299 2017-05-17  Martin Liska  <mliska@suse.cz>
31301         * cfg.c: Introduce dump_flags_t type and
31302         use it instead of int type.
31303         * cfg.h: Likewise.
31304         * cfghooks.c: Likewise.
31305         * cfghooks.h (struct cfg_hooks): Likewise.
31306         * cfgrtl.c: Likewise.
31307         * cfgrtl.h: Likewise.
31308         * cgraph.c (cgraph_node::get_body): Likewise.
31309         * coretypes.h: Likewise.
31310         * domwalk.c: Likewise.
31311         * domwalk.h: Likewise.
31312         * dumpfile.c (struct dump_option_value_info): Likewise.
31313         (dump_enable_all): Likewise.
31314         (dump_switch_p_1): Likewise.
31315         (opt_info_switch_p): Likewise.
31316         * dumpfile.h (enum tree_dump_index): Likewise.
31317         (struct dump_file_info): Likewise.
31318         * genemit.c: Likewise.
31319         * generic-match-head.c: Likewise.
31320         * gengtype.c (open_base_files): Likewise.
31321         * gimple-pretty-print.c: Likewise.
31322         * gimple-pretty-print.h: Likewise.
31323         * graph.c (print_graph_cfg): Likewise.
31324         * graphite-scop-detection.c (dot_all_sese): Likewise.
31325         * ipa-devirt.c (build_type_inheritance_graph): Likewise.
31326         * loop-unroll.c (report_unroll): Likewise.
31327         * passes.c (pass_manager::register_one_dump_file): Likewise.
31328         * print-tree.c: Likewise.
31329         * statistics.c: Likewise.
31330         * tree-cfg.c: Likewise.
31331         * tree-cfg.h: Likewise.
31332         * tree-dfa.c: Likewise.
31333         * tree-dfa.h: Likewise.
31334         * tree-dump.c (dump_function): Likewise.
31335         * tree-dump.h (struct dump_info): Likewise.
31336         * tree-pretty-print.c: Likewise.
31337         * tree-pretty-print.h: Likewise.
31338         * tree-ssa-live.c: Likewise.
31339         * tree-ssa-live.h: Likewise.
31340         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Likewise.
31341         * tree-vect-loop.c: Likewise.
31342         * tree-vect-slp.c: Likewise.
31344 2017-05-16  James Greenhalgh  <james.greenhalgh@arm.com>
31345             Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
31347         PR tree-optimization/80457
31348         * tree-vect-stmts.c (vect_model_simple_cost): Model the cost
31349         of all arguments to a statement as scalar_to_vec operations.
31350         (vectorizable_call): Adjust call to vect_model_simple_cost for
31351         new parameter.
31352         (vectorizable_conversion): Likewise.
31353         (vectorizable_assignment): Likewise.
31354         (vectorizable_shift): Likewise.
31355         (vectorizable_operation): Likewise.
31356         (vectorizable_comparison): Likewise.
31357         (vect_is_simple_cond): Record the def types for operands.
31358         (vectorizable_condition): Likewise, call vect_model_simple_cost.
31359         * tree-vectorizer.h (vect_model_simple_cost): Add new parameter
31360         for statement argument count.
31362 2017-05-16  Carl Love  <cel@us.ibm.com>
31364         * config/rs6000/rs6000-c: Add support for built-in functions
31365         vector unsigned long long vec_bperm (vector unsigned long long,
31366                                              vector unsigned char)
31367         vector signed long long vec_mule (vector signed int,
31368                                           vector signed int)
31369         vector unsigned long long vec_mule (vector unsigned int,
31370                                             vector unsigned int)
31371         vector signed long long vec_mulo (vector signed int,
31372                                           vector signed int)
31373         vector unsigned long long vec_mulo (vector unsigned int,
31374                                             vector unsigned int)
31375         vector signed char vec_sldw (vector signed char,
31376                                      vector signed char,
31377                                      const int)
31378         vector unsigned char vec_sldw (vector unsigned char,
31379                                        vector unsigned char,
31380                                        const int)
31381         vector signed short vec_sldw (vector signed short,
31382                                       vector signed short,
31383                                       const int)
31384         vector unsigned short vec_sldw (vector unsigned short,
31385                                         vector unsigned short,
31386                                         const int)
31387         vector signed int vec_sldw (vector signed int,
31388                                     vector signed int,
31389                                     const int)
31390         vector unsigned int vec_sldw (vector unsigned int,
31391                                       vector unsigned int,
31392                                       const int)
31393         vector signed long long vec_sldw (vector signed long long,
31394                                           vector signed long long,
31395                                           const int)
31396         vector unsigned long long vec_sldw (vector unsigned long long,
31397                                             vector unsigned long long,
31398                                             const int)
31399         * config/rs6000/rs6000-c: Add support for built-in functions
31400         * config/rs6000/rs6000-builtin.def: Add definition for SLDW.
31401         * config/rs6000/altivec.h: Add defintion for vec_sldw.
31402         * doc/extend.texi: Update the built-in documentation for the
31403         new built-in functions.
31405 2017-05-16  Marek Polacek  <polacek@redhat.com>
31407         PR sanitizer/80536
31408         PR sanitizer/80386
31409         * tree.c (save_expr): Don't fold the expression.
31411 2017-05-16  Uros Bizjak  <ubizjak@gmail.com>
31413         * config/i386/i386.md (*movsi_internal): Split (?rm,*y) alternative
31414         to (?r,*Yn) and (?m,*y) alternatives, and (?*y,rm) to (?*Ym,r)
31415         and (?*y,m).  Update insn attributes.
31417 2017-05-16  Martin Liska  <mliska@suse.cz>
31419         * cgraph.c (cgraph_edge::resolve_speculation): Add default value for
31420         flags argument of print_gimple_stmt, print_gimple_expr,
31421         print_generic_stmt and print_generic_expr.
31422         * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
31423         * coretypes.h: Likewise.
31424         * except.c (dump_eh_tree): Likewise.
31425         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
31426         * gimple-pretty-print.h: Likewise.
31427         * gimple-ssa-backprop.c (dump_usage_prefix): Likewise.
31428         (backprop::push_to_worklist): Likewise.
31429         (backprop::pop_from_worklist): Likewise.
31430         (backprop::process_use): Likewise.
31431         (backprop::intersect_uses): Likewise.
31432         (note_replacement): Likewise.
31433         * gimple-ssa-store-merging.c
31434         (pass_store_merging::terminate_all_aliasing_chains): Likewise.
31435         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
31436         (pass_store_merging::execute): Likewise.
31437         * gimple-ssa-strength-reduction.c (dump_candidate): Likewise.
31438         (ssa_base_cand_dump_callback): Likewise.
31439         (dump_incr_vec): Likewise.
31440         (replace_refs): Likewise.
31441         (replace_mult_candidate): Likewise.
31442         (create_add_on_incoming_edge): Likewise.
31443         (create_phi_basis): Likewise.
31444         (insert_initializers): Likewise.
31445         (all_phi_incrs_profitable): Likewise.
31446         (introduce_cast_before_cand): Likewise.
31447         (replace_one_candidate): Likewise.
31448         * gimplify.c (gimplify_expr): Likewise.
31449         * graphite-isl-ast-to-gimple.c (is_valid_rename): Likewise.
31450         (set_rename): Likewise.
31451         (rename_uses): Likewise.
31452         (copy_loop_phi_nodes): Likewise.
31453         (add_close_phis_to_merge_points): Likewise.
31454         (copy_loop_close_phi_args): Likewise.
31455         (copy_cond_phi_args): Likewise.
31456         (graphite_copy_stmts_from_block): Likewise.
31457         (translate_pending_phi_nodes): Likewise.
31458         * graphite-poly.c (print_pdr): Likewise.
31459         (dump_gbb_cases): Likewise.
31460         (dump_gbb_conditions): Likewise.
31461         (print_scop_params): Likewise.
31462         * graphite-scop-detection.c (build_cross_bb_scalars_def): Likewise.
31463         (build_cross_bb_scalars_use): Likewise.
31464         (gather_bbs::before_dom_children): Likewise.
31465         * hsa-dump.c (dump_hsa_immed): Likewise.
31466         * ipa-cp.c (print_ipcp_constant_value): Likewise.
31467         (get_replacement_map): Likewise.
31468         * ipa-inline-analysis.c (dump_condition): Likewise.
31469         (estimate_function_body_sizes): Likewise.
31470         * ipa-polymorphic-call.c (check_stmt_for_type_change): Likewise.
31471         (ipa_polymorphic_call_context::get_dynamic_type): Likewise.
31472         * ipa-prop.c (ipa_dump_param): Likewise.
31473         (ipa_print_node_jump_functions_for_edge): Likewise.
31474         (ipa_modify_call_arguments): Likewise.
31475         (ipa_modify_expr): Likewise.
31476         (ipa_dump_param_adjustments): Likewise.
31477         (ipa_dump_agg_replacement_values): Likewise.
31478         (ipcp_modif_dom_walker::before_dom_children): Likewise.
31479         * ipa-pure-const.c (check_stmt): Likewise.
31480         (pass_nothrow::execute): Likewise.
31481         * ipa-split.c (execute_split_functions): Likewise.
31482         * omp-offload.c (dump_oacc_loop_part): Likewise.
31483         (dump_oacc_loop): Likewise.
31484         * trans-mem.c (tm_log_emit): Likewise.
31485         (tm_memopt_accumulate_memops): Likewise.
31486         (dump_tm_memopt_set): Likewise.
31487         (dump_tm_memopt_transform): Likewise.
31488         * tree-cfg.c (gimple_verify_flow_info): Likewise.
31489         (print_loop): Likewise.
31490         * tree-chkp-opt.c (chkp_print_addr): Likewise.
31491         (chkp_gather_checks_info): Likewise.
31492         (chkp_get_check_result): Likewise.
31493         (chkp_remove_check_if_pass): Likewise.
31494         (chkp_use_outer_bounds_if_possible): Likewise.
31495         (chkp_reduce_bounds_lifetime): Likewise.
31496         * tree-chkp.c (chkp_register_addr_bounds): Likewise.
31497         (chkp_mark_completed_bounds): Likewise.
31498         (chkp_register_incomplete_bounds): Likewise.
31499         (chkp_mark_invalid_bounds): Likewise.
31500         (chkp_maybe_copy_and_register_bounds): Likewise.
31501         (chkp_build_returned_bound): Likewise.
31502         (chkp_get_bound_for_parm): Likewise.
31503         (chkp_build_bndldx): Likewise.
31504         (chkp_get_bounds_by_definition): Likewise.
31505         (chkp_generate_extern_var_bounds): Likewise.
31506         (chkp_get_bounds_for_decl_addr): Likewise.
31507         * tree-chrec.c (chrec_apply): Likewise.
31508         * tree-data-ref.c (dump_data_reference): Likewise.
31509         (dump_subscript): Likewise.
31510         (dump_data_dependence_relation): Likewise.
31511         (analyze_overlapping_iterations): Likewise.
31512         * tree-inline.c (expand_call_inline): Likewise.
31513         (tree_function_versioning): Likewise.
31514         * tree-into-ssa.c (dump_defs_stack): Likewise.
31515         (dump_currdefs): Likewise.
31516         (dump_names_replaced_by): Likewise.
31517         (dump_update_ssa): Likewise.
31518         (update_ssa): Likewise.
31519         * tree-object-size.c (pass_object_sizes::execute): Likewise.
31520         * tree-parloops.c (build_new_reduction): Likewise.
31521         (try_create_reduction_list): Likewise.
31522         (ref_conflicts_with_region): Likewise.
31523         (oacc_entry_exit_ok_1): Likewise.
31524         (oacc_entry_exit_single_gang): Likewise.
31525         * tree-pretty-print.h: Likewise.
31526         * tree-scalar-evolution.c (set_scalar_evolution): Likewise.
31527         (get_scalar_evolution): Likewise.
31528         (add_to_evolution): Likewise.
31529         (get_loop_exit_condition): Likewise.
31530         (analyze_evolution_in_loop): Likewise.
31531         (analyze_initial_condition): Likewise.
31532         (analyze_scalar_evolution): Likewise.
31533         (instantiate_scev): Likewise.
31534         (number_of_latch_executions): Likewise.
31535         (gather_chrec_stats): Likewise.
31536         (final_value_replacement_loop): Likewise.
31537         (scev_const_prop): Likewise.
31538         * tree-sra.c (dump_access): Likewise.
31539         (disqualify_candidate): Likewise.
31540         (create_access): Likewise.
31541         (reject): Likewise.
31542         (maybe_add_sra_candidate): Likewise.
31543         (create_access_replacement): Likewise.
31544         (analyze_access_subtree): Likewise.
31545         (analyze_all_variable_accesses): Likewise.
31546         (sra_modify_assign): Likewise.
31547         (initialize_constant_pool_replacements): Likewise.
31548         (find_param_candidates): Likewise.
31549         (decide_one_param_reduction): Likewise.
31550         (replace_removed_params_ssa_names): Likewise.
31551         * tree-ssa-ccp.c (ccp_fold_stmt): Likewise.
31552         * tree-ssa-copy.c (dump_copy_of): Likewise.
31553         (copy_prop_visit_cond_stmt): Likewise.
31554         * tree-ssa-dce.c (mark_operand_necessary): Likewise.
31555         * tree-ssa-dom.c (pass_dominator::execute): Likewise.
31556         (record_equivalences_from_stmt): Likewise.
31557         * tree-ssa-dse.c (compute_trims): Likewise.
31558         (delete_dead_call): Likewise.
31559         (delete_dead_assignment): Likewise.
31560         * tree-ssa-forwprop.c (forward_propagate_into_gimple_cond): Likewise.
31561         (forward_propagate_into_cond): Likewise.
31562         (pass_forwprop::execute): Likewise.
31563         * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
31564         * tree-ssa-loop-im.c (invariantness_dom_walker::before_dom_children):
31565         Likewise.
31566         (move_computations_worker): Likewise.
31567         (execute_sm): Likewise.
31568         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Likewise.
31569         (remove_exits_and_undefined_stmts): Likewise.
31570         (remove_redundant_iv_tests): Likewise.
31571         * tree-ssa-loop-ivopts.c (dump_use): Likewise.
31572         (adjust_iv_update_pos): Likewise.
31573         * tree-ssa-math-opts.c (bswap_replace): Likewise.
31574         * tree-ssa-phiopt.c (factor_out_conditional_conversion): Likewise.
31575         (value_replacement): Likewise.
31576         * tree-ssa-phiprop.c (phiprop_insert_phi): Likewise.
31577         * tree-ssa-pre.c (print_pre_expr): Likewise.
31578         (get_representative_for): Likewise.
31579         (create_expression_by_pieces): Likewise.
31580         (insert_into_preds_of_block): Likewise.
31581         (eliminate_insert): Likewise.
31582         (eliminate_dom_walker::before_dom_children): Likewise.
31583         (eliminate): Likewise.
31584         (remove_dead_inserted_code): Likewise.
31585         * tree-ssa-propagate.c (substitute_and_fold): Likewise.
31586         * tree-ssa-reassoc.c (get_rank): Likewise.
31587         (eliminate_duplicate_pair): Likewise.
31588         (eliminate_plus_minus_pair): Likewise.
31589         (eliminate_not_pairs): Likewise.
31590         (undistribute_ops_list): Likewise.
31591         (eliminate_redundant_comparison): Likewise.
31592         (update_range_test): Likewise.
31593         (optimize_range_tests_var_bound): Likewise.
31594         (optimize_vec_cond_expr): Likewise.
31595         (rewrite_expr_tree): Likewise.
31596         (rewrite_expr_tree_parallel): Likewise.
31597         (linearize_expr): Likewise.
31598         (break_up_subtract): Likewise.
31599         (linearize_expr_tree): Likewise.
31600         (attempt_builtin_powi): Likewise.
31601         (attempt_builtin_copysign): Likewise.
31602         (transform_stmt_to_copy): Likewise.
31603         (transform_stmt_to_multiply): Likewise.
31604         (dump_ops_vector): Likewise.
31605         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Likewise.
31606         (print_scc): Likewise.
31607         (set_ssa_val_to): Likewise.
31608         (visit_reference_op_store): Likewise.
31609         (visit_use): Likewise.
31610         (sccvn_dom_walker::before_dom_children): Likewise.
31611         (run_scc_vn): Likewise.
31612         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
31613         Likewise.
31614         (expr_hash_elt::print): Likewise.
31615         (const_and_copies::pop_to_marker): Likewise.
31616         (const_and_copies::record_const_or_copy_raw): Likewise.
31617         * tree-ssa-structalias.c (compute_dependence_clique): Likewise.
31618         * tree-ssa-uninit.c (collect_phi_def_edges): Likewise.
31619         (dump_predicates): Likewise.
31620         (find_uninit_use): Likewise.
31621         (warn_uninitialized_phi): Likewise.
31622         (pass_late_warn_uninitialized::execute): Likewise.
31623         * tree-ssa.c (verify_vssa): Likewise.
31624         (verify_ssa): Likewise.
31625         (maybe_optimize_var): Likewise.
31626         * tree-vrp.c (dump_value_range): Likewise.
31627         (dump_all_value_ranges): Likewise.
31628         (dump_asserts_for): Likewise.
31629         (register_edge_assert_for_2): Likewise.
31630         (vrp_visit_cond_stmt): Likewise.
31631         (vrp_visit_switch_stmt): Likewise.
31632         (vrp_visit_stmt): Likewise.
31633         (vrp_visit_phi_node): Likewise.
31634         (simplify_cond_using_ranges_1): Likewise.
31635         (fold_predicate_in): Likewise.
31636         (evrp_dom_walker::before_dom_children): Likewise.
31637         (evrp_dom_walker::push_value_range): Likewise.
31638         (evrp_dom_walker::pop_value_range): Likewise.
31639         (execute_early_vrp): Likewise.
31641 2017-05-16  Richard Biener  <rguenther@suse.de>
31643         * dwarf2out.c (loc_list_from_tree_1): Do not create
31644         DW_OP_GNU_variable_value for DECL_IGNORED_P decls.
31646 2017-05-16  Richard Biener  <rguenther@suse.de>
31648         * dwarf2out.c (resolve_variable_value_in_expr): Lookup DIE
31649         just generated.
31650         (note_variable_value_in_expr): If we resolved the decl ref
31651         do not push to the stack.
31653 2017-05-16  Matthew Wahab  <matthew.wahab@arm.com>
31655         * config/arm/arm_neon.h (vadd_f16): Use standard arithmetic
31656         operations in fast-math mode.
31657         (vaddq_f16): Likewise.
31658         (vmul_f16): Likewise.
31659         (vmulq_f16): Likewise.
31660         (vsub_f16): Likewise.
31661         (vsubq_f16): Likewise.
31662         * config/arm/neon.md (add<mode>3): New.
31663         (sub<mode>3): New.
31664         (fma:<VH:mode>3): New.  Also remove outdated comment.
31665         (mul<mode>3): New.
31667 2017-05-16  Martin Liska  <mliska@suse.cz>
31669         PR ipa/79849.
31670         PR ipa/79850.
31671         * ipa-devirt.c (warn_types_mismatch): Fix typo.
31672         (odr_types_equivalent_p): Likewise.
31674 2017-05-15  Sylvestre Ledru  <sylvestre@debian.org>
31676         * plugin.c (try_init_one_plugin): Fix ressource leaks (CID 726637).
31678 2017-05-15  Uros Bizjak  <ubizjak@gmail.com>
31680         PR target/80425
31681         * config/i386.i386.md (*zero_extendsidi2): Do not penalize
31682         non-interunit SSE move alternatives with '?'.
31683         (zero-extendsidi peephole2): New peephole to skip intermediate
31684         general register in SSE zero-extend sequence.
31686 2017-05-15  Jeff Law  <law@redhat.com>
31688         * reorg.c (relax_delay_slots): Create a new variable to hold
31689         the temporary target rather than clobbering TARGET_LABEL.
31691         * config/tilegx/tilegx.c (tilegx_expand_unaligned_load): Add
31692         missing argument to extract_bit_field call.
31693         * config/tilepro/tilepro.c (tilepro_expand_unaligned_load): Likewise.
31695 2017-05-15  Martin Liska  <mliska@suse.cz>
31697         PR driver/31468
31698         * gcc.c (process_command): Do not allow empty argument of -o option.
31700 2017-05-15  Renlin Li  <renlin.li@arm.com>
31702         * config/aarch64/aarch64-protos.h (aarch64_expand_call): Declare.
31703         * config/aarch64/aarch64.c (aarch64_expand_call): Define.
31704         * config/aarch64/constraints.md (Usf): Add long call check.
31705         * config/aarch64/aarch64.md (call): Use aarch64_expand_call.
31706         (call_value): Likewise.
31707         (sibcall): Likewise.
31708         (sibcall_value): Likewise.
31709         (call_insn): New.
31710         (call_value_insn): New.
31711         (sibcall_insn): Update rtx pattern.
31712         (sibcall_value_insn): Likewise.
31713         (call_internal): Remove.
31714         (call_value_internal): Likewise.
31715         (sibcall_internal): Likewise.
31716         (sibcall_value_internal): Likewise.
31717         (call_reg): Likewise.
31718         (call_symbol): Likewise.
31719         (call_value_reg): Likewise.
31720         (call_value_symbol): Likewise.
31722 2017-05-14  Krister Walfridsson  <krister.walfridsson@gmail.com>
31724         PR target/80600
31725         * config/netbsd.h (NETBSD_LIBGCC_SPEC): Always add -lgcc.
31727 2017-05-14  Uros Bizjak  <ubizjak@gmail.com>
31729         * config/i386.i386.c (ix86_cc_modes_compatible): CCNOmode is
31730         compatible with CCGOCmode and with CCZmode.
31732 2017-05-14  Martin Sebor  <msebor@redhat.com>
31734         PR middle-end/77671
31735         * gimple-fold.c (gimple_fold_builtin_sprintf): Make extern.
31736         (gimple_fold_builtin_snprintf): Same.
31737         * gimple-fold.h (gimple_fold_builtin_sprintf): Declare.
31738         (gimple_fold_builtin_snprintf): Same.
31739         * gimple-ssa-sprintf.c (get_format_string): Correct the detection
31740         of character types.
31741         (is_call_safe): New function.
31742         (try_substitute_return_value): Call it.
31743         (try_simplify_call): New function.
31744         (pass_sprintf_length::handle_gimple_call): Call it.
31746 2017-05-14  Martin Sebor  <msebor@redhat.com>
31748         PR middle-end/80669
31749         * builtins.c (expand_builtin_stpncpy): Simplify.
31751 2017-05-14  Daniel Santos  <daniel.santos@pobox.com>
31753         * config/i386/i386.opt: Add option -mcall-ms2sysv-xlogues.
31754         * config/i386/i386.h
31755         (x86_64_ms_sysv_extra_clobbered_registers): Change type to unsigned.
31756         (NUM_X86_64_MS_CLOBBERED_REGS): New macro.
31757         (struct machine_function): Add new members call_ms2sysv,
31758         call_ms2sysv_pad_in, call_ms2sysv_pad_out and call_ms2sysv_extra_regs.
31759         (struct machine_frame_state): New fields sp_realigned and
31760         sp_realigned_offset.
31761         * config/i386/i386.c
31762         (enum xlogue_stub): New enum.
31763         (enum xlogue_stub_sets): New enum.
31764         (class xlogue_layout): New class.
31765         (struct ix86_frame): New fields stack_realign_allocate_offset,
31766         stack_realign_offset and outlined_save_offset.  Modify comments to
31767         detail stack layout when using out-of-line stubs.
31768         (ix86_target_string): Add -mcall-ms2sysv-xlogues option.
31769         (ix86_option_override_internal): Add sorry() for TARGET_SEH and
31770         -mcall-ms2sysv-xlogues.
31771         (stub_managed_regs): New static variable.
31772         (ix86_save_reg): Add new parameter ignore_outlined to optionally omit
31773         registers managed by out-of-line stub.
31774         (disable_call_ms2sysv_xlogues): New function.
31775         (ix86_compute_frame_layout): Modify re-alignment calculations, disable
31776         m->call_ms2sysv when appropriate and compute frame layout for
31777         out-of-line stubs.
31778         (sp_valid_at, fp_valid_at): New inline functions.
31779         (choose_basereg): New function.
31780         (choose_baseaddr): Add align parameter, use choose_basereg and modify
31781         all callers.
31782         (ix86_emit_save_reg_using_mov, ix86_emit_restore_sse_regs_using_mov):
31783         Use align parameter of choose_baseaddr to generated aligned SSE movs
31784         when possible.
31785         (pro_epilogue_adjust_stack): Modify to track
31786         machine_frame_state::sp_realigned.
31787         (ix86_nsaved_regs): Modify to accommodate changes to ix86_save_reg.
31788         (ix86_nsaved_sseregs): Likewise.
31789         (ix86_emit_save_regs): Likewise.
31790         (ix86_emit_save_regs_using_mov): Likewise.
31791         (ix86_emit_save_sse_regs_using_mov): Likewise.
31792         (get_scratch_register_on_entry): Likewise.
31793         (gen_frame_set): New function.
31794         (gen_frame_load): Likewise.
31795         (gen_frame_store): Likewise.
31796         (emit_outlined_ms2sysv_save): Likewise.
31797         (emit_outlined_ms2sysv_restore): Likewise.
31798         (ix86_expand_prologue): Modify stack re-alignment code and call
31799         emit_outlined_ms2sysv_save when appropriate.
31800         (ix86_emit_leave): Clear machine_frame_state::sp_realigned.  Add
31801         parameter rtx_insn *insn, which allows the function to be used to only
31802         generate the notes.
31803         (ix86_expand_epilogue): Modify validity checks of frame and stack
31804         pointers, and call emit_outlined_ms2sysv_restore when appropriate.
31805         (ix86_expand_call): Modify to enable m->call_ms2sysv when appropriate.
31806         * config/i386/predicates.md
31807         (save_multiple): New predicate.
31808         (restore_multiple): Likewise.
31809         * config/i386/sse.md
31810         (save_multiple<mode>): New pattern.
31811         (save_multiple_realign<mode>): Likewise.
31812         (restore_multiple<mode>): Likewise.
31813         (restore_multiple_and_return<mode>): Likewise.
31814         (restore_multiple_leave_return<mode>): Likewise.
31815         * Makefile.in: Export HOSTCXX and HOSTCXXFLAGS to site.exp
31817 2017-05-14  Julia Koval  <julia.koval@intel.com>
31819         * config/i386/i386-builtin-types.def (VOID_FTYPE_INT_INT64): New type.
31820         * config/i386/i386-builtin.def (__builtin_ia32_xgetbv)
31821         (__builtin_ia32_xsetbv): New builtins.
31822         * config/i386/i386.c (ix86_expand_special_args_builtin):
31823         Process new types.
31824         (ix86_expand_builtin): Special expand for new intrinsics.
31825         * config/i386/i386.md (UNSPECV_XGETBV, UNSPECV_XSETBV): New.
31826         (xsetbv, xsetbv_rex64, xgetbv, xgetbv_rex64): New insn patterns.
31827         * config/i386/xsaveintrin.h (_xsetbv, _getbv): New intrinsics.
31829 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31831         * cfganal.c (inverted_post_order_compute): Change argument type
31832         to vec *.
31833         * cfganal.h (inverted_post_order_compute): Adjust prototype.
31834         * df-core.c (rest_of_handle_df_initialize): Adjust.
31835         (rest_of_handle_df_finish): Likewise.
31836         (df_analyze_1): Likewise.
31837         (df_analyze): Likewise.
31838         (loop_inverted_post_order_compute): Change argument to be a vec *.
31839         (df_analyze_loop): Adjust.
31840         (df_get_n_blocks): Likewise.
31841         (df_get_postorder): Likewise.
31842         * df.h (struct df_d): Change field to be a vec.
31843         * lcm.c (compute_laterin): Adjust.
31844         (compute_available): Likewise.
31845         * lra-lives.c (lra_create_live_ranges_1): Likewise.
31846         * tree-ssa-dce.c (remove_dead_stmt): Likewise.
31847         * tree-ssa-pre.c (compute_antic): Likewise.
31849 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31851         * cfganal.c (connect_infinite_loops_to_exit): Adjust.
31852         (depth_first_search::depth_first_search): Change structure init
31853         function to this constructor.
31854         (depth_first_search::add_bb): Rename function to this member.
31855         (depth_first_search::execute): Likewise.
31856         (flow_dfs_compute_reverse_finish): Adjust.
31858 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31860         * ddg.c (find_nodes_on_paths): Use auto_sbitmap.
31861         (longest_simple_path): Likewise.
31862         * shrink-wrap.c (spread_components): Likewise.
31863         (disqualify_problematic_components): Likewise.
31864         (emit_common_heads_for_components): Likewise.
31865         (emit_common_tails_for_components): Likewise.
31866         (insert_prologue_epilogue_for_components): Likewise.
31868 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31870         * tree-ssa-dse.c (dse_dom_walker): Make m_live_byes a
31871         auto_sbitmap.
31873 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31875         * df-core.c (df_set_blocks): Start using auto_bitmap.
31876         (df_compact_blocks): Likewise.
31877         * df-problems.c (df_rd_confluence_n): Likewise.
31878         * df-scan.c (df_insn_rescan_all): Likewise.
31879         (df_process_deferred_rescans): Likewise.
31880         (df_update_entry_block_defs): Likewise.
31881         (df_update_exit_block_uses): Likewise.
31882         (df_entry_block_bitmap_verify): Likewise.
31883         (df_exit_block_bitmap_verify): Likewise.
31884         (df_scan_verify): Likewise.
31885         * lra-constraints.c (lra_constraints): Likewise.
31886         (undo_optional_reloads): Likewise.
31887         (lra_undo_inheritance): Likewise.
31888         * lra-remat.c (calculate_gen_cands): Likewise.
31889         (do_remat): Likewise.
31890         * lra-spills.c (assign_spill_hard_regs): Likewise.
31891         (spill_pseudos): Likewise.
31892         * tree-ssa-pre.c (bitmap_set_and): Likewise.
31893         (bitmap_set_subtract_values): Likewise.
31895 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31897         * haifa-sched.c (estimate_shadow_tick): Replace manual bitmap
31898         management with auto_bitmap.
31899         (fix_inter_tick): Likewise.
31900         (fix_recovery_deps): Likewise.
31901         * ira.c (add_store_equivs): Likewise.
31902         (find_moveable_pseudos): Likewise.
31903         (split_live_ranges_for_shrink_wrap): Likewise.
31904         * print-rtl.c (rtx_reuse_manager::rtx_reuse_manager): Likewise.
31905         (rtx_reuse_manager::seen_def_p): Likewise.
31906         (rtx_reuse_manager::set_seen_def): Likewise.
31907         * print-rtl.h (class rtx_reuse_manager): Likewise.
31909 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31911         * bt-load.c (combine_btr_defs): Use auto_bitmap to manage bitmap
31912         lifetime.
31913         (migrate_btr_def): Likewise.
31914         * cfgloop.c (get_loop_body_in_bfs_order): Likewise.
31915         * df-core.c (loop_post_order_compute): Likewise.
31916         (loop_inverted_post_order_compute): Likewise.
31917         * hsa-common.h: Likewise.
31918         * hsa-gen.c (hsa_bb::~hsa_bb): Likewise.
31919         * init-regs.c (initialize_uninitialized_regs): Likewise.
31920         * ipa-inline.c (resolve_noninline_speculation): Likewise.
31921         (inline_small_functions): Likewise.
31922         * ipa-reference.c (ipa_reference_write_optimization_summary): Likewise.
31923         * ira.c (combine_and_move_insns): Likewise.
31924         (build_insn_chain): Likewise.
31925         * loop-invariant.c (find_invariants): Likewise.
31926         * lower-subreg.c (propagate_pseudo_copies): Likewise.
31927         * predict.c (tree_predict_by_opcode): Likewise.
31928         (predict_paths_leading_to): Likewise.
31929         (predict_paths_leading_to_edge): Likewise.
31930         (estimate_loops_at_level): Likewise.
31931         (estimate_loops): Likewise.
31932         * shrink-wrap.c (try_shrink_wrapping): Likewise.
31933         (spread_components): Likewise.
31934         * tree-cfg.c (remove_edge_and_dominated_blocks): Likewise.
31935         * tree-loop-distribution.c (rdg_build_partitions): Likewise.
31936         * tree-predcom.c (tree_predictive_commoning_loop): Likewise.
31937         * tree-ssa-coalesce.c (coalesce_ssa_name): Likewise.
31938         * tree-ssa-phionlycprop.c (pass_phi_only_cprop::execute): Likewise.
31939         * tree-ssa-pre.c (remove_dead_inserted_code): Likewise.
31940         * tree-ssa-sink.c (nearest_common_dominator_of_uses): Likewise.
31941         * tree-ssa-threadupdate.c (compute_path_counts): Likewise.
31942         (mark_threaded_blocks): Likewise.
31943         (thread_through_all_blocks): Likewise.
31944         * tree-ssa.c (verify_ssa): Likewise.
31945         (execute_update_addresses_taken): Likewise.
31946         * tree-ssanames.c (verify_ssaname_freelists): Likewise.
31948 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31950         * cfganal.c (mark_dfs_back_edges): Replace manual stack with
31951         auto_vec.
31952         (post_order_compute): Likewise.
31953         (inverted_post_order_compute): Likewise.
31954         (pre_and_rev_post_order_compute_fn): Likewise.
31956 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31958         * genrecog.c (int_set::int_set): Explicitly construct our
31959         auto_vec base class.
31960         * vec.h (auto_vec::auto_vec): New constructor.
31962 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31964         * bitmap.h (class auto_bitmap): New constructor taking
31965         bitmap_obstack * argument.
31967 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31969         * bitmap.h (class auto_bitmap): Change type of m_bits to
31970         bitmap_head, and adjust ctor / dtor and member operators.
31972 2017-05-13  Uros Bizjak  <ubizjak@gmail.com>
31974         * compare-elim.c (equivalent_reg_at_start): Return NULL_RTX
31975         when returned register mode doesn't match original mode.
31977 2017-05-12  Jeff Law  <law@redhat.com>
31978             Jakub Jelinek  <jakub@redhat.com>
31980         * config/mn10300/mn10300.c (mn10300_match_ccmode): Fix where
31981         we look for cc setter after the compare-elim changes.
31982         * config/mn10300/mn10300.md (addsi3_flags): Fix order of patterns
31983         within the vector to match what compare-elim now expects.
31984         (subsi3_flags, andsi3_flags, iorsi3_flags): Likewise.
31985         (xorsi3_flags, one_cmplsi2_flags): Likewise.
31987         * config/rx/rx.c (rx_match_ccmode): Fix where we look cc setter
31988         after the compare-elim changes.
31989         * config/rx/rx.md (abssi2_flags): Fix order of patterns within
31990         the vector to match what compare-elim now expects.
31991         (addsi3_flags, adc_flags, addsi3_flags peepholes): Likewise.
31992         (andsi3_flags, negsi2_flags, one_cmplsi2_flags): Likewise.
31993         (iorsi3_flags, rotlsi3_flags, rotrsi3_flags): Likewise.
31994         (ashrsi3_flags, lshrsi3_flags, ashlsi3_flags): Likewise.
31995         (ssaddsi3, subsi3_flags, sbb_flags, xorsi3_flags): Likewise.
31997         * config/visium/visium.c (single_set_and_flags): Fix where
31998         we look for cc setter after the compare-elim changes.
31999         * config/visium/visium.md (flags_subst_logic): Fix order of patterns
32000         with the vector to match what compare-elim now expects.
32001         (flags_subst_arith, add<mode>3_insn_set_carry): Likewise.
32002         (add<mode>3_insn_set_overflow, addsi3_insn_set_carry): Likewise.
32003         (addsi3_insn_set_overflow, sub<mode>3_insn_set_carry): Likewise.
32004         (sub<mode>3_insn_set_overflow, subsi3_insn_set_carry): Likewise.
32005         (subsi3_insn_set_overflow, negsi2_insn_set_carry): Likewise.
32006         (neg<mode>2_insn_set_overflow): Likewise.
32008 2017-05-12  Jim Wilson  <jim.wilson@linaro.org>
32010         PR middle-end/79794
32011         * expmed.c (extract_bit_field_1): Add alt_rtl argument.  Before
32012         maybe_expand_insn call, set ops[0].target.  If still set after call,
32013         set alt_rtl.  Add extra arg to recursive calls.
32014         (extract_bit_field): Add alt_rtl argument.  Pass to
32015         extract_bit_field.
32016         * expmed.h (extract_bit_field): Fix prototype.
32017         * expr.c (emit_group_load_1, copy_blkmode_from_reg)
32018         (copy_blkmode_to_reg, read_complex_part, store_field): Pass extra NULL
32019         to extract_bit_field_calls.
32020         (expand_expr_real_1): Pass alt_rtl to expand_expr_real instead of 0.
32021         Pass alt_rtl to extract_bit_field calls.
32022         * calls.c (store_unaligned_arguments_into_psuedos)
32023         load_register_parameters): Pass extra NULL to extract_bit_field calls.
32024         * optabs.c (maybe_legitimize_operand): Clear op->target when call
32025         gen_reg_rtx.
32026         * optabs.h (struct expand_operand): Add target bitfield.
32028 2017-05-12  Uros Bizjak  <ubizjak@gmail.com>
32030         * compare-elim.c (try_eliminate_compare): Canonicalize
32031         operation with embedded compare to
32032         [(set (reg:CCM) (compare:CCM (operation) (immediate)))
32033          (set (reg) (operation)].
32035         * config/i386/i386.c (TARGET_FLAGS_REGNUM): New define.
32037 2017-05-12  Uros Bizjak  <ubizjak@gmail.com>
32039         PR target/80723
32040         * config/i386/i386.c (ix86_rtx_cost) [case PLUS]: Ignore the
32041         cost of adding a carry flag for ADC instruction.
32042         [case MINUS]: Ignore the cost of subtracting a carry flag
32043         for SBB instruction.
32045 2017-05-12  Steven Munroe  <munroesj@gcc.gnu.org>
32047         * config.gcc (powerpc*-*-*): Add bmi2intrin.h, bmiintrin.h,
32048         and x86intrin.h
32049         * config/rs6000/bmiintrin.h: New file.
32050         * config/rs6000/bmi2intrin.h: New file.
32051         * config/rs6000/x86intrin.h: New file.
32053 2017-05-12  Jeff Law  <law@redhat.com>
32055         * tree-vrp.c (vrp_dom_walker::before_dom_children): Push unwinding
32056         markers.
32058 2017-05-12  Peter Bergner  <bergner@vnet.ibm.com>
32060         PR middle-end/80707
32061         * tree-cfg.c: Remove cfg edges of unreachable case statements.
32063 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
32065         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling for
32066         early expansion of vector divide builtins.
32067         (builtin_function_type): Add VSX_BUILTIN_UDIV_V2DI to the list of
32068         builtins identified as having unsigned arguments.
32070 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
32072         * config/rs6000/rs6000.c (gimple-fold.h): New #include.
32073         (rs6000_gimple_fold_builtin): Add handling for early GIMPLE
32074         expansion of vector logical operations (and, andc, or, xor,
32075         nor, orc, nand).
32077 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
32079         * gimple-fold.c (create_tmp_reg_or_ssa_name): Remove static declaration.
32080         * gimple-fold.h (create_tmp_reg_or_ssa_name): New prototype.
32082 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
32084         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling for
32085         early GIMPLE expansion of vector multiplies.
32087 2017-05-12  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
32089         * config/arm/arm.md (movsi): Add TARGET_32BIT in addition to the
32090         TARGET_HAVE_MOVT conditional.
32091         (movt splitter): Likewise.
32093 2017-05-12  Richard Biener  <rguenther@suse.de>
32095         * tree-ssa-sccvn.h (has_VN_INFO): Declare.
32096         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
32097         Fold all stmts not inplace.
32099 2017-05-12  Richard Biener  <rguenther@suse.de>
32101         PR tree-optimization/80713
32102         * tree-ssa-pre.c (remove_dead_inserted_code): Clear
32103         inserted_exprs bit for not removed stmts.
32105 2017-05-12  Thomas Schwinge  <thomas@codesourcery.com>
32107         PR middle-end/69921
32108         * tree-parloops.c (create_parallel_loop): Set "oacc kernels
32109         parallelized" attribute for parallelized OpenACC kernels.
32110         * omp-offload.c (execute_oacc_device_lower): Use it.
32112         * omp-expand.c (expand_omp_target) <GF_OMP_TARGET_KIND_OACC_KERNELS>:
32113         Set "oacc kernels" attribute.
32114         * omp-general.c (oacc_set_fn_attrib): Remove is_kernel formal
32115         parameter.  Adjust all users.
32116         (oacc_fn_attrib_kernels_p): Remove function.
32117         * omp-offload.c (execute_oacc_device_lower): Look for "oacc
32118         kernels" attribute instead of calling oacc_fn_attrib_kernels_p.
32119         * tree-ssa-loop.c (gate_oacc_kernels): Likewise.
32120         * tree-parloops.c (create_parallel_loop): If oacc_kernels_p,
32121         assert "oacc kernels" attribute is set.
32123 2017-05-11  Carl Love  <cel@us.ibm.com>
32125         * config/rs6000/rs6000-c: Add support for built-in functions
32126         vector unsigned char vec_popcnt (vector signed char)
32127         vector unsigned char vec_popcnt (vector unsigned char)
32128         vector unsigned short vec_popcnt (vector signed short)
32129         vector unsigned short vec_popcnt (vector unsigned short)
32130         vector unsigned int vec_popcnt (vector signed int)
32131         vector unsigned int vec_popcnt (vector unsigned int)
32132         vector unsigned long long vec_popcnt (vector signed long long)
32133         vector unsigned long long vec_popcnt (vector unsigned long long)
32134         vector signed long long vec_slo (vector signed long long,
32135                                          vector signed char)
32136         vector signed long long vec_slo (vector signed long long,
32137                                          vector unsigned char)
32138         vector unsigned long long vec_slo (vector unsigned long long,
32139                                            vector signed char)
32140         vector unsigned long long vec_slo (vector unsigned long long,
32141                                            vector unsigned char)
32142         * config/rs6000/rs6000-builtin.def: Add definitions for VPOPCNTUB,
32143         VPOPCNTUH, VPOPCNTUW, and VPOPCNTUD overloads.
32144         * config/rs6000/altivec.h: Add define for vec_popcnt, vec_popcntb,
32145         vec_popcnth, vec_popcntw and vec_popcntd built-in functions.
32146         * doc/extend.texi: Update the built-in documentation file for the
32147         new built-in functions.
32149 2017-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
32151         * attribs.h (sorted_attr_string): Move machine independent
32152         functions for target clone support from the i386 port to common
32153         code.  Rename ix86_function_versions to common_function_versions.
32154         Rename make_name to make_unique_name.
32155         (common_function_versions): Likewise.
32156         (make_unique_name): Likewise.
32157         (make_dispatcher_decl): Likewise.
32158         (is_function_default_version): Likewise.
32159         * attribs.c (attr_strcmp): Likewise.
32160         (sorted_attr_string): Likewise.
32161         (common_function_versions): Likewise.
32162         (make_unique_name): Likewise.
32163         (make_dispatcher_decl): Likewise.
32164         (is_function_default_version): Likewise.
32165         * config/i386/i386.c (attr_strcmp): Likewise.
32166         (sorted_attr_string): Likewise.
32167         (ix86_function_versions): Likewise.
32168         (make_name): Likewise.
32169         (make_dispatcher_decl): Likewise.
32170         (is_function_default_version): Likewise.
32171         (TARGET_OPTION_FUNCTION_VERSIONS): Update target function hook.
32173 2017-05-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32175         PR target/80695
32176         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost):
32177         Account for direct move costs for vec_construct of integer
32178         vectors.
32180 2017-05-11  Uros Bizjak  <ubizjak@gmail.com>
32182         PR target/80706
32183         * config/i386/sync.md (UNSPEC_LDX_ATOMIC): New unspec.
32184         (UNSPEC_STX_ATOMIC): Ditto.
32185         (loaddi_via_sse): New insn.
32186         (storedi_via_sse): Ditto.
32187         (atomic_loaddi_fpu): Emit loaddi_via_sse and storedi_via_sse.
32188         Update corresponding peephole2 patterns.
32189         (atomic_storedi_fpu): Ditto.
32191 2017-05-11  Julia Koval  <julia.koval@intel.com>
32193         * config/i386/avx512fintrin.h (_mm_mask_rsqrt14_sd)
32194         (_mm_maskz_rsqrt14_sd, _mm_mask_rsqrt14_ss, _mm_maskz_rsqrt14_ss):
32195         New intrinsics.
32196         * config/i386/i386-builtin.def (__builtin_ia32_rsqrt14sd_mask)
32197         (__builtin_ia32_rsqrt14ss_mask): New builtins.
32198         * config/i386/sse.md (rsqrt14_<mode>_mask): New pattern.
32200 2017-05-11  Nathan Sidwell  <nathan@acm.org>
32202         * graphite-poly.c: Include dumpfile.h.
32204         * dumpfle.h (dump_function): Declare here ...
32205         * tree-dump.h (dump_function): ... not here.
32206         * dumpfile.c: #include tree-cfg.h.
32207         (dump_function): Move here from ...
32208         * tree-dump.c (dump_function): ... here.
32209         * gimplify.c: #include splay-tree.h, not tree-dump.h.
32210         * graphite-poly.c: Don't include tree-dump.h.
32211         * cgraphclones.c: Include dumpfile.h not tree-dump.h.
32212         * print-tree.c: Likewise.
32213         * stor-layout.c: Likewise.
32214         * tree-nested.c: Likewise.
32216         * dumpfile.c (dump_start): Use TDF_FLAGS.
32217         (dump_enable_all): Fix TDF_KIND check thinko.
32219 2017-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
32221         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
32222         array entries to represent two legal parameterizations of the
32223         overloaded __builtin_cmpb function, as represented by the
32224         P6_OV_BUILTIN_CMPB constant.
32225         (altivec_resolve_overloaded_builtin): Add special case handling
32226         for the __builtin_cmpb function, as represented by the
32227         P6_OV_BUILTIN_CMPB constant.
32228         * config/rs6000/rs6000-builtin.def (BU_P6_2): New macro.
32229         (BU_P6_64BIT_2): New macro.
32230         (BU_P6_OVERLOAD_2): New macro
32231         (CMPB_32): Add 32-bit compare-bytes support for 32-bit only targets.
32232         (CMPB): Add 64-bit compare-bytes support for 32-bit and 64-bit targets.
32233         (CMPB): Add overload support to represent both 32-bit and 64-bit
32234         compare-bytes function.
32235         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add
32236         support for TARGET_CMPB.
32237         * config/rs6000/rs6000.h: Add support for RS6000_BTM_CMPB.
32238         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
32239         documentation of the __builtin_cmpb overloaded built-in function.
32241 2017-05-11  Richard Biener  <rguenther@suse.de>
32243         PR tree-optimization/80705
32244         * tree-vect-data-refs.c (vect_analyze_data_refs): DECL_NONALIASED
32245         bases are not vectorizable.
32247 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32249         * tree-ssa-loop-ivopts.c (determine_set_costs): Skip non-interger
32250         when counting register pressure.
32252 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32254         * tree-ssa-loop-ivopts.c (compare_cost_pair): New.
32255         (iv_ca_more_deps): Renamed to ...
32256         (iv_ca_compare_deps): ... this.
32257         (iv_ca_extend): Extend iv_ca if NEW_CP is cheaper than OLD_CP.
32259 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32261         * tree-ssa-loop-ivopts.c (find_interesting_uses): Move inv vars dump
32262         to ...
32263         (determine_group_iv_costs): ... here.
32264         (find_inv_vars_cb): Record inv var if it's not recorded before.
32266 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32268         * tree-ssa-loop-ivopts.c (allow_ip_end_pos_p): Refine comments.
32269         (get_shiftadd_cost): Ditto.
32271 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32273         * tree-ssa-address.c: Include header file.
32274         (move_hint_to_base): Return TRUE if BASE_HINT is moved to memory
32275         address.
32276         (add_to_parts): Refactor.
32277         (addr_to_parts): New parameter.  Update use of move_hint_to_base.
32278         (create_mem_ref): Update use of addr_to_parts.  Re-associate addr
32279         in new order.
32281 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32283         PR tree-optimization/53090
32284         * tree-ssa-loop-ivopts.c (enum comp_iv_rewrite): New enum value
32285         COMP_IV_EXPR_2.
32286         (extract_cond_operands): Detect condition with IV on both sides
32287         and return COMP_IV_EXPR_2.
32288         (find_interesting_uses_cond): Add iv_use for both IVs in condition.
32289         (rewrite_use_compare): Simplify by removing call to function
32290         extract_cond_operands.
32292 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32294         * tree-ssa-loop-ivopts.c (enum comp_iv_rewrite): New.
32295         (extract_cond_operands): Detect condition comparing against non-
32296         invariant bound and return appropriate enum value.
32297         (find_interesting_uses_cond): Update use of extract_cond_operands.
32298         Handle its return value accordingly.
32299         (determine_group_iv_cost_cond, rewrite_use_compare): Ditto.
32301 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32303         * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Re-associate
32304         nonlinear iv_use computation in loop invariant sensitive way.
32306 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32308         * tree-ssa-loop-ivopts.c (relate_compare_use_with_all_cands): New.
32309         (find_iv_candidates): Call relate_compare_use_with_all_cands.
32311 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32313         * tree-ssa-loop-ivopts.c (struct iv_cand): New field inv_exprs.
32314         (dump_cand): Support iv_cand.inv_exprs.
32315         (add_candidate_1): Record invariant exprs in iv_cand.inv_exprs
32316         for candidates.
32317         (iv_ca_set_no_cp, iv_ca_set_cp, free_loop_data): Support
32318         iv_cand.inv_exprs.
32320 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32322         * tree-ssa-loop-ivopts.c (multiplier_allowed_in_address_p): Move
32323         from ...
32324         * tree-ssa-address.c (multiplier_allowed_in_address_p): ... to here
32325         as local function.  Include necessary header files.
32326         * tree-ssa-loop-ivopts.h (multiplier_allowed_in_address_p): Delete.
32328 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32330         * tree-ssa-loop-ivopts.c (autoinc_possible_for_pair): Simplify.
32332 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32334         * tree-ssa-loop-ivopts.c (force_expr_to_var_cost): Handle more
32335         operators: TRUNC_DIV_EXPR, BIT_AND_EXPR, BIT_IOR_EXPR, LSHIFT_EXPR,
32336         RSHIFT_EXPR and BIT_NOT_EXPR.
32338 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32340         * tree-ssa-loop-ivopts.c (get_loop_invariant_expr): Simplify.
32341         (adjust_setup_cost): New parameter supporting round up adjustment.
32342         (struct address_cost_data): Delete.
32343         (force_expr_to_var_cost): Don't bound cost with spill_cost.
32344         (split_address_cost, ptr_difference_cost): Delete.
32345         (difference_cost, compare_aff_trees, record_inv_expr): Delete.
32346         (struct ainc_cost_data): New struct.
32347         (get_address_cost_ainc): New function.
32348         (get_address_cost, get_computation_cost): Reimplement.
32349         (determine_group_iv_cost_address): Record inv_expr for all uses of
32350         a group.
32351         (determine_group_iv_cost_cond): Call get_loop_invariant_expr.
32352         (iv_ca_has_deps): Reimplemented to ...
32353         (iv_ca_more_deps): ... this.  Check if NEW_CP introduces more deps
32354         than OLD_CP.
32355         (iv_ca_extend): Call iv_ca_more_deps.
32357 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32359         * tree-ssa-address.c (struct mem_address): Move to header file.
32360         (valid_mem_ref_p, move_fixed_address_to_symbol): Make it global.
32361         * tree-ssa-address.h (struct mem_address): Move from C file.
32362         (valid_mem_ref_p, move_fixed_address_to_symbol): Declare.
32364 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32366         * tree-affine.h (aff_combination_type): New interface.
32367         (aff_combination_zero_p): Remove static.
32368         (aff_combination_const_p): New interface.
32369         (aff_combination_singleton_var_p): New interfaces.
32371 2017-05-11  Richard Biener  <rguenther@suse.de>
32373         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
32374         Skip unreachable blocks and destinations.
32375         (eliminate): Move stmt removal and fixup ...
32376         (fini_eliminate): ... here.  Skip inserted exprs.
32377         (pass_pre::execute): Move fini_pre after fini_eliminate.
32378         * tree-ssa-tailmerge.c: Include tree-cfgcleanup.h.
32379         (tail_merge_optimize): Run cleanup_tree_cfg if requested by
32380         PRE to get rid of dead code that has invalid SSA form and
32381         split critical edges again.
32383 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32385         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
32387 2017-05-11  Richard Biener  <rguenther@suse.de>
32389         * passes.c (execute_function_todo): Verify loops if they are
32390         said to be up-to-date.
32391         * cfgexpand.c (pass_expand::execute): Discard loops for -dx.
32392         * trans-mem.c (pass_tm_edges::execute): Mark loops for fixup.
32394 2017-05-10  John David Anglin  <danglin@gcc.gnu.org>
32396         PR target/80090
32397         * config/pa/pa.c (pa_assemble_integer): When outputting a SYMBOL_REF,
32398         handle calling assemble_external ourself.
32400         PR target/79027
32401         * config/pa/pa.c (pa_cannot_change_mode_class): Reject changes to/from
32402         modes with zero size.  Enhance comment.
32404 2017-05-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32406         * config/rs6000/rs6000.c (altivec_init_builtins): Define POWER8
32407         built-ins for vec_xl and vec_xst with short and char pointer
32408         arguments.
32410 2017-05-10  Sebastian Peryt  <sebastian.peryt@intel.com>
32412         * config/i386/avx512fintrin.h (_mm_mask_max_round_sd)
32413         (_mm_maskz_max_round_sd, _mm_mask_max_round_ss)
32414         (_mm_maskz_max_round_ss, _mm_mask_min_round_sd)
32415         (_mm_maskz_min_round_sd, _mm_mask_min_round_ss)
32416         (_mm_maskz_min_round_ss): New intrinsics.
32417         * config/i386/i386-builtin-types.def (V2DF, V2DF, V2DF, V2DF, UQI, INT)
32418         (V4SF, V4SF, V4SF, V4SF, UQI, INT): New function type aliases.
32419         * config/i386/i386-builtin.def (__builtin_ia32_maxsd_mask_round)
32420         (__builtin_ia32_maxss_mask_round, __builtin_ia32_minsd_mask_round)
32421         (__builtin_ia32_minss_mask_round): New builtins.
32422         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32423         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
32424         * config/i386/sse.md (<sse>_vm<code><mode>3<round_saeonly_name>):
32425         Rename to ...
32426         (<sse>_vm<code><mode>3<mask_name><round_saeonly_name>): ... this.
32427         (v<maxmin_float><ssescalarmodesuffix>\t{<round_saeonly_op3>%2, %1, %0|%0, %1, %<iptr>2<round_saeonly_op3>}):
32428         Change to ...
32429         (v<maxmin_float><ssescalarmodesuffix>\t{<round_saeonly_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_saeonly_mask_op3>}):
32430         ... this.
32432 2017-05-10  Sebastian Peryt  <sebastian.peryt@intel.com>
32434         * config/i386/avx512fintrin.h (_mm_mask_mul_round_sd)
32435         (_mm_maskz_mul_round_sd, _mm_mask_mul_round_ss)
32436         (_mm_maskz_mul_round_ss, _mm_mask_div_round_sd)
32437         (_mm_maskz_div_round_sd, _mm_mask_div_round_ss)
32438         (_mm_maskz_div_round_ss, _mm_mask_mul_sd, _mm_maskz_mul_sd)
32439         (_mm_mask_mul_ss, _mm_maskz_mul_ss, _mm_mask_div_sd)
32440         (_mm_maskz_div_sd, _mm_mask_div_ss, _mm_maskz_div_ss): New intrinsics.
32441         * config/i386/i386-builtin-types.def (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32442         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): New function type aliases.
32443         * config/i386/i386-builtin.def (__builtin_ia32_divsd_mask_round)
32444         (__builtin_ia32_divss_mask_round, __builtin_ia32_mulsd_mask_round)
32445         (__builtin_ia32_mulss_mask_round): New builtins.
32446         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32447         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
32448         * config/i386/sse.md (<sse>_vm<multdiv_mnemonic><mode>3<round_name>):
32449         Rename to ...
32450         (<sse>_vm<multdiv_mnemonic><mode>3<mask_name><round_name>): ... this.
32451         (v<multdiv_mnemonic><ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|%0, %1, %<iptr>2<round_op3>}):
32452         Change to ...
32453         (v<multdiv_mnemonic><ssescalarmodesuffix>\t{<round_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}):
32454         ... this.
32456 2017-05-10  Julia Koval  <julia.koval@intel.com>
32458         * config/i386/avxintrin.h (_mm256_set_m128, _mm256_set_m128d)
32459         (_mm256_set_m128i, _mm256_setr_m128, _mm256_setr_m128d)
32460         (_mm256_setr_m128i): New intrinsics.
32462 2017-05-10  Julia Koval  <julia.koval@intel.com>
32464         * config/i386/avx512fintrin.h (_mm_mask_rcp14_sd)
32465         (_mm_maskz_rcp14_sd, _mm_mask_rcp14_ss)
32466         (_mm_maskz_rcp14_ss): New intrinsics.
32467         * config/i386/i386-builtin.def (__builtin_ia32_rcp14sd_mask)
32468         (__builtin_ia32_rcp14ss_mask): New builtins.
32469         * config/i386/sse.md (srcp14<mode>_mask): New pattern.
32471 2017-05-10  Peter Bergner  <bergner@vnet.ibm.com>
32473         PR tree-optimization/51513
32474         * tree-cfg.c (gimple_seq_unreachable_p): New function.
32475         (assert_unreachable_fallthru_edge_p): Use it.
32476         (group_case_labels_stmt): Likewise.
32477         * tree-cfg.h: Prototype it.
32478         * stmt.c: Include cfghooks.h and tree-cfg.h.
32479         (emit_case_dispatch_table) <gap_label>: New local variable.
32480         Use it to fill dispatch table gaps.
32481         Test for default_label before updating probabilities.
32482         (expand_case) <default_label>: Remove unneeded initialization.
32483         Test for unreachable default case statement and remove its edge.
32484         Set default_label accordingly.
32485         * tree-ssa-ccp.c (optimize_unreachable): Update comment.
32487 2017-05-10  Carl Love  <cel@us.ibm.com>
32489         * config/rs6000/rs6000-c: Add support for built-in functions
32490         vector signed char      vec_neg (vector signed char)
32491         vector signed short int vec_neg (vector short int)
32492         vector signed int       vec_neg (vector signed int)
32493         vector signed long long vec_neg (vector signed long long)
32494         vector float            vec_neg (vector float)
32495         vector double           vec_neg (vector double)
32496         * config/rs6000/rs6000-builtin.def: Add definitions for NEG function
32497         overload.
32498         * config/rs6000/altivec.h: Add define for vec_neg
32499         * doc/extend.texi: Update the built-in documentation for the
32500         new built-in functions.
32502 2017-05-10  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
32504         PR tree-optimization/77644
32505         * match.pd (sqrt(x) cmp sqrt(y) -> x cmp y): New pattern.
32507 2017-05-10  Nathan Sidwell  <nathan@acm.org>
32509         * dumpfile.h (TDI_lang_all): New.
32510         (TDF_KIND): New. Renumber others
32511         (TDF_LANG, TDF_TREE, TDF_RTL, TDF_IPA): Enumerate value, rather
32512         than bits.
32513         * dumpfile.c (dump_files): Mark language dumps as TDF_LANG.  add
32514         lang-all.
32515         (get_dump_file_name): Adjust suffix generation.
32516         (dump_enable_all): Use TDF_KIND.
32517         * doc/invoke.texi (-fdump-lang-all): Document.
32519         * dumpfile.h: Tabify.
32521 2017-05-10  Wilco Dijkstra  <wdijkstr@arm.com>
32523         PR target/80671
32524         * config/aarch64/cortex-a57-fma-steering.c (merge_forest):
32525         Move member access before delete.
32527 2017-05-10  Alexandre Oliva  <aoliva@redhat.com>
32529         * tree-inline.c (expand_call_inline): Split block at stmt
32530         before the call.
32532 2017-05-09  Michael Meissner  <meissner@linux.vnet.ibm.com>
32534         PR target/68163
32535         * config/rs6000/rs6000.md (f32_lr): Delete mode attributes that
32536         are now unused after splitting mov{sf,sd}_hardfloat.
32537         (f32_lr2): Likewise.
32538         (f32_lm): Likewise.
32539         (f32_lm2): Likewise.
32540         (f32_li): Likewise.
32541         (f32_li2): Likewise.
32542         (f32_lv): Likewise.
32543         (f32_sr): Likewise.
32544         (f32_sr2): Likewise.
32545         (f32_sm): Likewise.
32546         (f32_sm2): Likewise.
32547         (f32_si): Likewise.
32548         (f32_si2): Likewise.
32549         (f32_sv): Likewise.
32550         (f32_dm): Likewise.
32551         (f32_vsx): Likewise.
32552         (f32_av): Likewise.
32553         (mov<mode>_hardfloat): Split into separate movsf and movsd pieces.
32554         For movsf, order stores so the VSX stores occur before the GPR
32555         store which encourages the register allocator to use a traditional
32556         FPR instead of a GPR.  For movsd, order the stores so that the GPR
32557         store comes before the VSX stores to allow the power6 to work.
32558         This is due to the power6 not having a 32-bit integer store
32559         instruction from a FPR.
32560         (movsf_hardfloat): Likewise.
32561         (movsd_hardfloat): Likewise.
32563 2017-05-09  Martin Sebor  <msebor@redhat.com>
32565         PR translation/80280
32566         * config/sol2-c.c (cmn_err_flag_specs): Initialize new data member
32567         added in r247778.
32569         PR translation/80280
32570         * config/i386/msformat-c.c (ms_printf_flag_specs): Initialize new
32571         data member added in r247778.
32572         (ms_scanf_flag_specs, ms_strftime_flag_specs): Same.
32574 2017-05-09  Nathan Sidwell  <nathan@acm.org>
32576         * tree.h (tree_fits_shwi_p, tree_fits_uhwi_p): Unconditionally pure.
32578         * ipa-devirt.c (default_hash_traits<type_pair>): Skip struct-scope
32579         typedefs.
32581 2017-05-09  Marek Polacek  <polacek@redhat.com>
32583         * doc/invoke.texi: Fix typo.
32585 2017-05-09  Richard Biener  <rguenther@suse.de>
32587         * tree-vrp.c (vrp_val_is_max): Adjust comment.
32588         (vrp_val_is_min): Likewise.
32589         (set_value_range_to_value): Likewise.
32590         (set_value_range_to_nonnegative): Likewise.
32591         (gimple_assign_nonzero_p): Likewise.
32592         (gimple_stmt_nonzero_p): Likewise.
32593         (vrp_int_const_binop): Likewise.  Remove unreachable case.
32594         (adjust_range_with_scev): Adjust comments.
32595         (compare_range_with_value): Likewise.
32596         (extract_range_from_phi_node): Likewise.
32597         (test_for_singularity): Likewise.
32599 2017-05-09  Richard Biener  <rguenther@suse.de>
32601         * tree-vrp.c (get_single_symbol): Add assert that we don't
32602         get overflowed constants as invariant part.
32603         (compare_values_warnv): Add comment before the TREE_NO_WARNING
32604         checks.  Use wi::cmp instead of recursing for integer constants.
32605         (compare_values): Just ignore whether we assumed undefined
32606         overflow instead of failing the compare.
32607         (extract_range_for_var_from_comparison_expr): Add comment before the
32608         TREE_NO_WARNING sets.
32609         (test_for_singularity): Likewise.
32610         (extract_range_from_comparison): Do not disable optimization
32611         when we assumed undefined overflow.
32612         (extract_range_basic): Remove init of unused var.
32614 2017-05-09  Richard Biener  <rguenther@suse.de>
32616         * tree-vrp.c (vrp_int_const_binop): Use wide-ints and simplify.
32617         (extract_range_from_multiplicative_op_1): Adjust.
32618         (extract_range_from_binary_expr_1): Use int_const_binop.
32620 2017-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
32622         PR target/80101
32623         * config/rs6000/power6.md: Replace store_data_bypass_p calls with
32624         rs6000_store_data_bypass_p in seven define_bypass directives and
32625         in several comments.
32626         * config/rs6000/rs6000-protos.h: Add prototype for
32627         rs6000_store_data_bypass_p function.
32628         * config/rs6000/rs6000.c (rs6000_store_data_bypass_p): New
32629         function implements slightly different (rs6000-specific) semantics
32630         than store_data_bypass_p, returning false rather than aborting
32631         with assertion error when arguments do not satisfy the
32632         requirements of store data bypass.
32633         (rs6000_adjust_cost): Replace six calls of store_data_bypass_p with
32634         rs6000_store_data_bypass_p.
32636 2017-05-08  Max Filippov  <jcmvbkbc@gmail.com>
32638         * config/xtensa/xtensa-protos.h
32639         (xtensa_initial_elimination_offset): New declaration.
32640         * config/xtensa/xtensa.c (xtensa_initial_elimination_offset):
32641         New function. Move its body from the INITIAL_ELIMINATION_OFFSET
32642         macro definition, add case for FRAME_POINTER_REGNUM when
32643         FRAME_GROWS_DOWNWARD.
32644         * config/xtensa/xtensa.h (FRAME_GROWS_DOWNWARD): New macro definition.
32645         (INITIAL_ELIMINATION_OFFSET): Replace body with call to
32646         xtensa_initial_elimination_offset.
32648 2017-05-08  Nathan Sidwell  <nathan@acm.org>
32650         * doc/invoke.texi: Alphabetize -fdump options.
32652 2017-05-08  Martin Sebor  <msebor@redhat.com>
32654         PR translation/80280
32655         * config/sol2-c.c (solaris_pragma_align): Correct quoting.
32657 2017-05-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
32659         * target.def (compute_frame_layout): New optional target hook.
32660         * doc/tm.texi.in (TARGET_COMPUTE_FRAME_LAYOUT): Add hook.
32661         * doc/tm.texi (TARGET_COMPUTE_FRAME_LAYOUT): Add documentation.
32662         * lra-eliminations.c (update_reg_eliminate): Call compute_frame_layout
32663         target hook.
32664         * reload1.c (verify_initial_elim_offsets): Likewise.
32665         * config/arm/arm.c (TARGET_COMPUTE_FRAME_LAYOUT): Define.
32666         (use_simple_return_p): Call arm_compute_frame_layout if needed.
32667         (arm_get_frame_offsets): Split up into this ...
32668         (arm_compute_frame_layout): ... and this function.
32670 2017-05-08  Richard Sandiford  <richard.sandiford@arm.com>
32672         * config/aarch64/constraints.md (Usa): New constraint.
32673         * config/aarch64/aarch64.md (*movsi_aarch64, *movdi_aarch64): Use it.
32675 2017-05-08  Thomas Preud'homme  <thomas.preudhomme@arm.com>
32677         * config.gcc (arm*-*-*): Set TM_MULTILIB_CONFIG from
32678         with_multilib_list after it has been checked.
32680 2017-05-08  Richard Biener  <rguenther@suse.de>
32682         * tree-ssa-pre.c (bitmap_set_and): Avoid bitmap copy.
32683         (bitmap_set_subtract_values): Likewise.
32685 2017-05-08  Richard Biener  <rguenther@suse.de>
32687         * tree-vrp.c (gimple_assign_nonzero_warnv_p): Rename to ...
32688         (gimple_assign_nonzero): ... this and remove strict_overflow_p
32689         argument.
32690         (gimple_stmt_nonzero_warnv_p): Rename to ...
32691         (gimple_stmt_nonzero_p): ... this and remove strict_overflow_p
32692         argument.
32693         (vrp_stmt_computes_nonzero): Remove strict_overflow_p argument.
32694         (extract_range_basic): Adjust, do not disable propagation on
32695         strict overflow sensitive simplification.
32696         (vrp_visit_cond_stmt): Likewise.
32698 2017-05-05  Jan Hubicka  <hubicka@ucw.cz>
32700         * ipa-inline-analysis.c (estimate_function_body_sizes): Recompute
32701         body size unconditionally.
32703 2017-05-07  Jeff Law  <law@redhat.com>
32705         Revert:
32706         2017-05-06  Jeff Law  <law@redhat.com>
32707         PR tree-optimization/78496
32708         * tree-vrp.c (simplify_assert_expr_using_ranges): Remove debugging
32709         code.
32711         PR tree-optimization/78496
32712         * tree-vrp.c (simplify_assert_expr_using_ranges): New function.
32713         (simplify_stmt_using_ranges): Call it.
32714         (vrp_dom_walker::before_dom_children): Extract equivalences
32715         from an ASSERT_EXPR with an equality comparison against a
32716         constant.
32718 2017-05-06  Jeff Law  <law@redhat.com>
32720         PR tree-optimization/78496
32721         * tree-vrp.c (simplify_assert_expr_using_ranges): Remove debugging
32722         code.
32724         PR tree-optimization/78496
32725         * tree-vrp.c (simplify_assert_expr_using_ranges): New function.
32726         (simplify_stmt_using_ranges): Call it.
32727         (vrp_dom_walker::before_dom_children): Extract equivalences
32728         from an ASSERT_EXPR with an equality comparison against a
32729         constant.
32731 2017-05-06  Richard Sandiford  <richard.sandiford@linaro.org>
32733         * lra-constraints.c (lra_copy_reg_equiv): New function.
32734         (split_reg): Use it to copy equivalence information from the
32735         original register to the spill register.
32737 2017-05-06  Richard Sandiford  <richard.sandiford@linaro.org>
32739         PR rtl-optimization/75964
32740         * simplify-rtx.c (simplify_const_relational_operation): Remove
32741         invalid handling of comparisons of integer ABS.
32743 2017-05-06  Uros Bizjak  <ubizjak@gmail.com>
32745         * config/i386/i386.c (ext_80387_constant_init): Do not explicitly
32746         initialize to zero.
32747         (init_regs): Remove declaration.
32748         (function_arg_advance_32): Initialize error_p as boolean variable.
32750 2017-05-05  Nathan Sidwell  <nathan@acm.org>
32752         * store-motion.c (remove_reachable_equiv_notes): Reformat long
32753         lines.  Use for (;;).
32755 2017-05-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32757         * config/rs6000/rs6000.c (rs6000_vect_nonmem): New static var.
32758         (rs6000_init_cost): Initialize rs6000_vect_nonmem.
32759         (rs6000_add_stmt_cost): Update rs6000_vect_nonmem.
32760         (rs6000_finish_cost): Avoid vectorizing simple copy loops with
32761         VF=2 that require versioning.
32763 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32765         * diagnostic.h (CARET_LINE_MARGIN): Convert from macro to const
32766         int.
32768 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32770         * diagnostic.h (diagnostic_override_option_index): Convert from
32771         macro to inline function.
32773 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32775         * diagnostic.c (last_module_changed_p): New function.
32776         (set_last_module): New function.
32777         (diagnostic_report_current_module): Convert macro usage to
32778         the above functions.
32779         * diagnostic.h (diagnostic_context::last_module): Strengthen
32780         from const line_map * to const line_map_ordinary *.
32781         (diagnostic_last_module_changed): Delete macro.
32782         (diagnostic_set_last_module): Delete macro.
32784 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32786         * diagnostic.c (diagnostic_impl): Replace report_diagnostic
32787         with diagnostic_report_diagnostic.
32788         (diagnostic_n_impl_richloc): Likewise.
32789         * diagnostic.h (report_diagnostic): Delete macro.
32790         * rtl-error.c (diagnostic_for_asm): Replace report_diagnostic
32791         with diagnostic_report_diagnostic.
32792         * substring-locations.c (format_warning_va): Likewise.
32794 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32796         * diagnostic.c (diagnostic_report_diagnostic): Eliminate
32797         save/restor of format_spec.  Move option-printing code to...
32798         (print_option_information): ...this new function, and
32799         reimplement by simply printing to the pretty_printer,
32800         rather than appending to the format string.
32802 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32804         * diagnostic.c (diagnostic_report_diagnostic): Split out pragma
32805         handling logic into...
32806         (update_effective_level_from_pragmas): ...this new function.
32808 2017-05-04  Andrew Waterman  <andrew@sifive.com>
32810         * config/riscv/riscv.opt (mstrict-align): New option.
32811         * config/riscv/riscv.h (STRICT_ALIGNMENT): Use it.  Update comment.
32812         (SLOW_UNALIGNED_ACCESS): Define.
32813         (riscv_slow_unaligned_access): Declare.
32814         * config/riscv/riscv.c (riscv_tune_info): Add slow_unaligned_access
32815         field.
32816         (riscv_slow_unaligned_access): New variable.
32817         (rocket_tune_info): Set slow_unaligned_access to true.
32818         (optimize_size_tune_info): Set slow_unaligned_access to false.
32819         (riscv_cpu_info_table): Add entry for optimize_size_tune_info.
32820         (riscv_valid_lo_sum_p): Use TARGET_STRICT_ALIGN.
32821         (riscv_option_override): Set riscv_slow_unaligned_access.
32822         * doc/invoke.texi: Add -mstrict-align to RISC-V.
32824 2017-05-04  Kito Cheng  <kito.cheng@gmail.com>
32826         * config/riscv/riscv.md: Unify indentation.
32828 2017-05-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
32830         PR target/79038
32831         PR target/79202
32832         PR target/79203
32833         * config/rs6000/rs6000.md (u code attribute): Add FIX and
32834         UNSIGNED_FIX.
32835         (extendsi<mode>2): Add support for doing sign extension via
32836         VUPKHSW and XXPERMDI if the value is in Altivec registers and we
32837         don't have ISA 3.0 instructions.
32838         (extendsi<mode>2 splitter): Likewise.
32839         (fix_trunc<mode>si2): If we are at ISA 2.07 (VSX small integer),
32840         generate the normal insns since SImode can now go in vector
32841         registers.  Disallow the special UNSPECs needed for previous
32842         machines to hide SImode being used.  Add new insns
32843         fctiw{,w}_<mode>_smallint if SImode can go in vector registers.
32844         (fix_trunc<mode>si2_stfiwx): Likewise.
32845         (fix_trunc<mode>si2_internal): Likewise.
32846         (fixuns_trunc<mode>si2): Likewise.
32847         (fixuns_trunc<mode>si2_stfiwx): Likewise.
32848         (fctiw<u>z_<mode>_smallint): Likewise.
32849         (fctiw<u>z_<mode>_mem): New combiner pattern to prevent conversion
32850         of floating point to 32-bit integer from doing a direct move to
32851         the GPR registers to do a store.
32852         (fctiwz_<mode>): Break long line.
32854 2017-05-05  Bin Cheng  <bin.cheng@arm.com>
32856         * Makefile.in (GTFILES): Add tree-ssa-loop-ivopts.c.
32857         * tree-ssa-loop-ivopts.c (compute_max_addr_offset): Delete.
32858         (addr_list, addr_offset_valid_p): New.
32859         (split_address_groups): Check offset validity with above function.
32860         (gt-tree-ssa-loop-ivopts.h): Include header file.
32862 2017-05-05  Nathan Sidwell  <nathan@acm.org>
32864         * config.gcc (arm*-*-*): Add missing 'fi'.
32866 2017-05-05  Steve Ellcey  <sellcey@cavium.com>
32868         * doc/invoke.texi (-fopt-info): Explicitly say order of options
32869         included in -fopt-info does not matter.
32870         * doc/optinfo.texi (-fopt-info): Fix description of default
32871         behavour. Explicitly say order of options included in -fopt-info
32872         does not matter.
32874 2017-05-05  Thomas Preud'homme  <thomas.preudhomme@arm.com>
32876         * config.gcc: Allow combinations of aprofile and rmprofile values for
32877         --with-multilib-list.
32878         * config/arm/t-multilib: New file.
32879         * config/arm/t-aprofile: Remove initialization of MULTILIB_*
32880         variables.  Remove setting of ISA and floating-point ABI in
32881         MULTILIB_OPTIONS and MULTILIB_DIRNAMES.  Set architecture and FPU in
32882         MULTI_ARCH_OPTS_A and MULTI_ARCH_DIRS_A rather than MULTILIB_OPTIONS
32883         and MULTILIB_DIRNAMES respectively.  Add comment to introduce all
32884         matches.  Add architecture matches for marvel-pj4 and generic-armv7-a
32885         CPU options.
32886         * config/arm/t-rmprofile: Likewise except for the matches changes.
32887         * doc/install.texi (--with-multilib-list): Document the combination of
32888         aprofile and rmprofile values and warn about pitfalls in doing that.
32890 2017-05-05  Wilco Dijkstra  <wdijkstr@arm.com>
32892         * config/aarch64/aarch64.md (movsi_aarch64): Remove '*' from r=w.
32893         (movdi_aarch64): Likewise.
32895 2017-05-05  Jakub Jelinek  <jakub@redhat.com>
32897         PR tree-optimization/80632
32898         * tree-switch-conversion.c (struct switch_conv_info): Add target_vop
32899         field.
32900         (build_arrays): Initialize it for virtual phis.
32901         (fix_phi_nodes): Use it for virtual phis.
32903         PR tree-optimization/80558
32904         * tree-vrp.c (extract_range_from_binary_expr_1): Optimize
32905         [x, y] op z into [x op, y op z] for op & or | if conditions
32906         are met.
32908 2017-05-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
32909             Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
32911         PR target/71607
32912         * config/arm/arm.md (use_literal_pool): Remove.
32913         (64-bit immediate split): No longer takes cost into consideration
32914         if arm_disable_literal_pool is enabled.
32915         * config/arm/arm.c (arm_tls_referenced_p): Add diagnostic if TLS is
32916         used when arm_disable_literal_pool is enabled.
32917         (arm_max_const_double_inline_cost): Remove use of
32918         arm_disable_literal_pool.
32919         (push_minipool_fix): Add assert.
32920         (arm_reorg): Add return if arm_disable_literal_pool is enabled.
32921         * config/arm/vfp.md (no_literal_pool_df_immediate): New.
32922         (no_literal_pool_sf_immediate): New.
32924 2017-05-05  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
32926         PR tree-optimization/80613
32927         * tree-ssa-dce.c (propagate_necessity): Remove cases for
32928         BUILT_IN_STRDUP and BUILT_IN_STRNDUP.
32930 2017-05-05  Richard Biener  <rguenther@suse.de>
32932         * tree-ssa-pre.c (get_or_alloc_expr_for): Simplify.
32934 2017-05-05  Georg-Johann Lay  <avr@gjlay.de>
32936         * config/avr/avr.md [flag_strict_overflow]: Remove any occurence
32937         of this flag from insn conditions due to removal from r247495.
32939 2017-05-05  Wilco Dijkstra  <wdijkstr@arm.com>
32941         * config/arm/aarch-common.c (arm_early_load_addr_dep_ptr):
32942         New function.
32943         (arm_early_store_addr_dep_ptr): Likewise.
32944         * config/arm/aarch-common-protos.h
32945         (arm_early_load_addr_dep_ptr): Add prototype.
32946         (arm_early_store_addr_dep_ptr): Likewise.
32947         * config/arm/cortex-a53.md: Add new bypasses.
32949 2017-05-05  Jakub Jelinek  <jakub@redhat.com>
32951         * tree.c (next_type_uid): Change type to unsigned.
32952         (type_hash_canon): Decrement back next_type_uid if
32953         freeing a type node with the highest TYPE_UID.  For INTEGER_TYPEs
32954         also ggc_free TYPE_MIN_VALUE, TYPE_MAX_VALUE and TYPE_CACHED_VALUES
32955         if possible.
32957 2017-05-04  Martin Sebor  <msebor@redhat.com>
32959         * builtins.c: Fix a trivial typo in a comment.
32961         PR middle-end/79234
32962         * builtins.c (check_sizes): Adjust to handle reading past the end.
32963         Avoid printing excessive upper bound of ranges.  Use %E to print
32964         tree nodes instead of converting them to %wu.
32965         (expand_builtin_memchr): New function.
32966         (compute_dest_size): Rename...
32967         (compute_objsize): ...to this.
32968         (expand_builtin_memcpy): Adjust.
32969         (expand_builtin_mempcpy): Adjust.
32970         (expand_builtin_strcat): Adjust.
32971         (expand_builtin_strcpy): Adjust.
32972         (check_strncat_sizes): Adjust.
32973         (expand_builtin_strncat): Adjust.
32974         (expand_builtin_strncpy): Adjust and simplify.
32975         (expand_builtin_memset): Adjust.
32976         (expand_builtin_bzero): Adjust.
32977         (expand_builtin_memcmp): Adjust.
32978         (expand_builtin): Handle memcmp.
32979         (maybe_emit_chk_warning): Check strncat just once.
32981 2017-05-04  Martin Sebor  <msebor@redhat.com>
32983         PR preprocessor/79214
32984         PR middle-end/79222
32985         PR middle-end/79223
32986         * builtins.c (check_sizes): Add inlining context and issue
32987         warnings even when -Wno-system-headers is set.
32988         (check_strncat_sizes): Same.
32989         (expand_builtin_strncat): Same.
32990         (expand_builtin_memmove): New function.
32991         (expand_builtin_stpncpy): Same.
32992         (expand_builtin): Handle memmove and stpncpy.
32994 2017-05-04  Bin Cheng  <bin.cheng@arm.com>
32996         * tree-ssa-loop-ivopts.c (struct cost_pair): Remove field inv_expr
32997         which is not used any more.
32999 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
33001         * config/aarch64/aarch64.c (generic_tunings): Update prefetch model.
33003 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
33005         * config/aarch64/aarch64.c (cortexa35_tunings): Set jump alignment to 4.
33006         (cortexa53_tunings): Likewise.
33007         (cortexa57_tunings): Likewise.
33008         (cortexa72_tunings): Likewise.
33009         (cortexa73_tunings): Likewise.
33011 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
33013         * config/aarch64/aarch64.c (generic_tunings): Set jump alignment to 4.
33014         Set loop alignment to 8.
33016 2017-05-04  Martin Sebor  <msebor@redhat.com>
33018         PR translation/80280
33019         * builtins.c (expand_builtin_object_size): Add missing quoting to
33020         %D and like directives.
33021         * hsa-gen.c (hsa_type_for_scalar_tree_type): Same.
33022         (hsa_type_for_tree_type): Same.
33023         (verify_function_arguments): Same.
33024         * symtab.c (symbol_table::change_decl_assembler_name): Same.
33025         * varasm.c (get_section): Same.
33026         (mark_weak): Same.
33028 2017-05-04  Martin Sebor  <msebor@redhat.com>
33030         PR translation/80280
33031         * config/i386/i386.c (ix86_function_versions): Quote a %D directive.
33033 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
33035         * config/aarch64/aarch64.c (generic_addrcost_table):
33036         Change HI/TI mode setting.
33038 2017-05-04  Martin Jambor  <mjambor@suse.cz>
33040         PR tree-optimization/80622
33041         * tree-sra.c (comes_initialized_p): New function.
33042         (build_accesses_from_assign): Only set write lazily when
33043         comes_initialized_p is false.
33044         (analyze_access_subtree): Use comes_initialized_p.
33045         (propagate_subaccesses_across_link): Assert !comes_initialized_p
33046         instead of testing for PARM_DECL.
33048 2017-05-04  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
33050         * config/aarch64/aarch64.md (prefetch); Adjust predicate and
33051         constraint on operand 0 to allow more general addressing modes.
33052         Adjust output template.
33053         * config/aarch64/aarch64.c (aarch64_address_valid_for_prefetch_p):
33054         New function.
33055         * config/aarch64/aarch64-protos.h
33056         (aarch64_address_valid_for_prefetch_p): Declare prototype.
33057         * config/aarch64/constraints.md (Dp): New address constraint.
33058         * config/aarch64/predicates.md (aarch64_prefetch_operand): New
33059         predicate.
33061 2017-05-04  Jan Hubicka  <hubicka@ucw.cz>
33063         * ipa-cp.c (perform_estimation_of_a_value): Drop base_time parameter;
33064         update use of estimate_ipcp_clone_size_and_time.
33065         (estimate_local_effects): Update use of
33066         estimate_ipcp_clone_size_and_time and perform_estimation_of_a_value.
33067         * ipa-inline.h (estimate_ipcp_clone_size_and_time): Update prototype.
33068         * ipa-inline-analysis.c (estimate_ipcp_clone_size_and_time):
33069         Return nonspecialized time.
33071 2017-05-04  Richard Biener  <rguenther@suse.de>
33073         * tree-ssa-alias.c (get_continuation_for_phi): Improve looking
33074         for the last VUSE which def dominates the PHI.  Directly call
33075         maybe_skip_until.
33076         (get_continuation_for_phi_1): Remove.
33078 2017-05-04  Richard Sandiford  <richard.sandiford@linaro.org>
33080         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): Add commentary
33081         to explain the use of truncating division.  Cap the number of
33082         iterations to the maximum given by nb_iterations_upper_bound,
33083         if defined.
33085 2017-05-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
33087         * configure.ac (--enable-mingw-wildcard): Add new configurable feature.
33088         * configure: Regenerate.
33089         * config.in: Regenerate.
33090         * config/i386/driver-mingw32.c: new file.
33091         * config/i386/x-mingw32: Add rule to build driver-mingw32.o.
33092         * config.host: Link driver-mingw32.o on MinGW host.
33093         * doc/install.texi: Document new --enable-mingw-wildcard configure
33094         option.
33096 2017-05-04  Marek Polacek  <polacek@redhat.com>
33098         PR tree-optimization/80612
33099         * calls.c (get_size_range): Check for INTEGRAL_TYPE_P.
33101 2017-05-04  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
33102             Andre Simoes Dias Vieira  <andre.simoesdiasvieira@arm.com>
33104         * config/arm/arm.md (movsi): Change TARGET_32BIT to TARGET_HAVE_MOVT.
33105         (movt splitter): Likewise.
33106         * config/arm/arm.c (arm_option_check_internal): Change arm_arch_thumb2
33107         to TARGET_HAVE_MOVT, and merge with -mslow-flash-data check.
33108         (const_ok_for_arm): Change else to else if (TARGET_THUMB2) and add else
33109         block for Thumb-1 with MOVT.
33110         (thumb2_legitimate_address_p): Move code block ...
33111         (can_avoid_literal_pool_for_label_p): ... into this new function.
33112         (thumb1_legitimate_address_p): Add check for TARGET_HAVE_MOVT and
33113         literal pool.
33114         (thumb_legitimate_constant_p): Add conditional on TARGET_HAVE_MOVT
33115         * doc/invoke.texi (-mpure-code): Change "ARMv7-M targets" for
33116         "M-profile targets with the MOVT instruction".
33118 2017-05-04  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
33120         * config/arm/arm-builtins.c (arm_init_builtins): Rename
33121         __builtin_arm_ldfscr to __builtin_arm_get_fpscr, and rename
33122         __builtin_arm_stfscr to __builtin_arm_set_fpscr.
33124 2017-05-04  Martin Liska  <mliska@suse.cz>
33126         * tree-vrp.c (simplify_cond_using_ranges_2): Remove unused
33127         variable cond_code.
33129 2017-05-04  Richard Biener  <rguenther@suse.de>
33131         * tree.c (array_at_struct_end_p): Handle arrays at struct
33132         end with flexarrays more conservatively.  Refactor and treat
33133         arrays of arrays or aggregates more strict.  Fix
33134         VIEW_CONVERT_EXPR handling.  Remove allow_compref argument.
33135         * tree.c (array_at_struct_end_p): Adjust prototype.
33136         * emit-rtl.c (set_mem_attributes_minus_bitpos): Adjust.
33137         * gimple-fold.c (get_range_strlen): Likewise.
33138         * tree-chkp.c (chkp_may_narrow_to_field): Likewise.
33140 2017-05-04  Richard Biener  <rguenther@suse.de>
33142         PR tree-optimization/31130
33143         * tree-vrp.c (needs_overflow_infinity): Remove as always returning
33144         false.
33145         (supports_overflow_infinity): Likewise.
33146         (is_negative_overflow_infinity): Likewise.
33147         (is_positive_overflow_infinity): Likewise.
33148         (is_overflow_infinity): Likewise.
33149         (stmt_overflow_infinity): Likewise.
33150         (overflow_infinity_range_p): Likewise.
33151         (usable_range_p): Remove as always returning true.
33152         (make_overflow_infinity): Remove.
33153         (negative_overflow_infinity): Likewise.
33154         (positive_overflow_infinity): Likewise.
33155         (avoid_overflow_infinity): Likewise.
33156         (set_value_range): Adjust accordingly.
33157         (set_value_range_to_nonnegative): Likewise, remove now unused
33158         overflow_infinity arg.
33159         (vrp_operand_equal_p): Adjust.
33160         (update_value_range): Likewise.
33161         (range_int_cst_singleton_p): Likewise.
33162         (operand_less_p): Likewise.
33163         (compare_values_warnv): Likewise.
33164         (extract_range_for_var_from_comparison_expr): Likewise.
33165         (vrp_int_const_binop): Likewise.
33166         (zero_nonzero_bits_from_vr): Likewise.
33167         (extract_range_from_multiplicative_op_1): Likewise.
33168         (extract_range_from_binary_expr_1): Likewise.
33169         (extract_range_from_unary_expr): Likewise.
33170         (extract_range_from_comparison): Likewise.
33171         (extract_range_basic): Likewise.
33172         (adjust_range_with_scev): Likewise.
33173         (compare_ranges): Likewise.
33174         (compare_range_with_value): Likewise.
33175         (dump_value_range): Likewise.
33176         (test_for_singularity): Likewise, remove strict_overflow_p parameter
33177         never used.
33178         (simplify_cond_using_ranges): Adjust.
33180 2017-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
33182         * brig-builtins.def: Added a builtin for class_f64.
33183         * builtin-types.def: Added a builtin type needed by class_f64.
33185 2017-05-03  Jason Merrill  <jason@redhat.com>
33187         * timevar.def: Add TV_CONSTEXPR.
33189 2017-05-03  David Malcolm  <dmalcolm@redhat.com>
33191         * common.opt (fdiagnostics-parseable-fixits): Fix typo.
33193 2017-05-03  Martin Jambor  <mjambor@suse.cz>
33195         * ipa-prop.c (ipa_update_after_lto_read): Removed.
33196         * ipa-prop.h (ipa_update_after_lto_read): Remove declaration.
33197         * ipa-cp.c (ipcp_propagate_stage): Do not call
33198         ipa_update_after_lto_read.
33199         * ipa-inline.c (ipa_inline): Likewise.
33201 2017-05-03  Martin Jambor  <mjambor@suse.cz>
33203         * ipa-prop.h (ipa_edge_args): Make a class.  Mark with for_user GTY
33204         tag.  Added a default constructor and a destructor.
33205         (ipa_edge_args_sum_t): New class;
33206         (ipa_edge_args_sum): Declare.
33207         (ipa_edge_args_vector): Remove declaration.
33208         (IPA_EDGE_REF): Use ipa_edge_args_sum.
33209         (ipa_free_edge_args_substructures): Remove declaration.
33210         (ipa_check_create_edge_args): Use ipa_edge_args_sum.
33211         (ipa_edge_args_info_available_for_edge_p): Likewise.
33212         * ipa-prop.c (ipa_edge_args_vector): Removed.
33213         (edge_removal_hook_holder): Likewise.
33214         (edge_duplication_hook_holder): Likewise.
33215         (ipa_edge_args_sum): New variable.
33216         (ipa_propagate_indirect_call_infos): Test ipa_edge_args_sum instead of
33217         ipa_edge_args_vector.
33218         (ipa_free_edge_args_substructures): Likewise.
33219         (ipa_free_all_edge_args): Free ipa_edge_args_sum instead of
33220         ipa_edge_args_vector.
33221         (ipa_edge_removal_hook): Turned into method
33222         ipa_edge_args_sum_t::remove.
33223         (ipa_edge_duplication_hook): Turned into method
33224         ipa_edge_args_sum_t::duplicate.
33225         (ipa_register_cgraph_hooks): Create ipa_edge_args_sum instead of
33226         registering edge hooks.
33227         (ipa_unregister_cgraph_hooks): Do not unregister edge hooks.
33228         * ipa-inline-analysis.c (estimate_function_body_sizes): Test
33229         ipa_edge_args_sum instead of ipa_edge_args_vector.
33230         * ipa-profile.c (ipa_profile): Likewise.
33232 2017-05-03  Martin Jambor  <mjambor@suse.cz>
33234         * symbol-summary.h (function_summary): New method exists.
33235         (function_summary::symtab_removal): Deallocate through release.
33236         (call_summary): New class.
33237         (gt_ggc_mx): New overload.
33238         (gt_pch_nx): Likewise.
33239         (gt_pch_nx): Likewise.
33241 2017-05-03  Jeff Law  <law@redhat.com>
33243         PR tree-optimization/78496
33244         * tree-vrp.c (simplify_cond_using_ranges_1): Renamed
33245         from simplify_cond_using_ranges.  Split off code to walk
33246         backwards through casts into ...
33247         (simplify_cond_using_ranges_2): New function.
33248         (simplify_stmt_using_ranges): Call simplify_cond_using_ranges_1.
33249         (execute_vrp): After identifying jump threads, call
33250         simplify_cond_using_ranges_2.
33252 2017-05-03  Jan Hubicka  <hubicka@ucw.cz>
33254         PR bootstrap/80609
33255         * ipa-inline.h (inline_summary): Add ctor.
33256         (create_ggc): Do not use ggc_cleared_alloc.
33258 2017-05-03  Jeff Downs  <heydowns@somuchpressure.net>
33259             Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
33261         * gcc.c (handle_braces): Support escaping in switch matching
33262         text.
33263         * doc/invoke.texi (Spec Files): Document it.
33264         Remove superfluous @code markup in items.
33266 2017-05-03  David Malcolm  <dmalcolm@redhat.com>
33268         * diagnostic-show-locus.c (struct column_range): New struct.
33269         (get_affected_columns): New function.
33270         (get_printed_columns): New function.
33271         (struct correction): New struct.
33272         (correction::ensure_capacity): New function.
33273         (correction::ensure_terminated): New function.
33274         (struct line_corrections): New struct.
33275         (line_corrections::~line_corrections): New dtor.
33276         (line_corrections::add_hint): New function.
33277         (layout::print_trailing_fixits): Reimplement in terms of the new
33278         classes.
33279         (selftest::test_overlapped_fixit_printing): New function.
33280         (selftest::diagnostic_show_locus_c_tests): Call it.
33282 2017-05-03  Nathan Sidwell  <nathan@acm.org>
33284         Canonicalize canonical type hashing
33285         * tree.h (type_hash_canon_hash): Declare.
33286         * tree.c (type_hash_list, attribute_hash_list): Move into
33287         type_hash_canon_hash.
33288         (build_type_attribute_qual_variant): Break out hash code calc into
33289         type_hash_canon_hash.
33290         (type_hash_canon_hash): New.  Generic type hash computation.
33291         (build_range_type_1, build_array_type_1, build_function_type,
33292         build_method_type_directly, build_offset_type, build_complex_type,
33293         make_vector_type): Call it.
33295 2017-05-03  Richard Biener  <rguenther@suse.de>
33297         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
33298         When all DRs have unknown misaligned do not always peel
33299         when there is a store but apply the same costing model as if
33300         there were only loads.
33302 2017-05-03  Richard Biener  <rguenther@suse.de>
33304         Revert
33305         PR tree-optimization/80492
33306         * tree-ssa-alias.c (decl_refs_may_alias_p): Handle
33307         compare_base_decls returning dont-know properly.
33309 2017-05-03  Thomas Preud'homme  <thomas.preudhomme@arm.com>
33311         * config/arm/iterators.md (CCSI): New mode iterator.
33312         (arch): New mode attribute.
33313         * config/arm/sync.md (atomic_compare_and_swap<mode>_1): Rename into ...
33314         (atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1): This and ...
33315         (atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): This.  Use CCSI
33316         code iterator for success result mode.
33317         * config/arm/arm.c (arm_expand_compare_and_swap): Adapt code to use
33318         the corresponding new insn generators.
33320 2017-05-03  Bin Cheng  <bin.cheng@arm.com>
33322         Revert r247509
33323         2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33324         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
33326 2017-05-03  Richard Sandiford  <richard.sandiford@linaro.org>
33328         * tree-data-ref.h (SUB_CONFLICTS_IN_A): Wrap SUB argument in brackets.
33329         (SUB_CONFLICTS_IN_B, SUB_LAST_CONFLICT, SUB_DISTANCE): Likewise.
33330         (DDR_A): Wrap DDR argument in brackets.
33331         (DDR_B, DDR_AFFINE_P, DDR_ARE_DEPENDENT, DDR_SUBSCRIPTS): Likewise.
33332         (DDR_LOOP_NEST, DDR_INNER_LOOP, DDR_SELF_REFERENCE): Likewise.
33333         (DDR_REVERSED_P): Likewise.
33335 2017-05-03  Jakub Jelinek  <jakub@redhat.com>
33337         PR tree-optimization/79472
33338         * tree-switch-conversion.c (struct switch_conv_info): Add
33339         contiguous_range and default_case_nonstandard fields.
33340         (collect_switch_conv_info): Compute contiguous_range and
33341         default_case_nonstandard fields, don't clear final_bb if
33342         contiguous_range and only the default case doesn't have the required
33343         structure.
33344         (check_all_empty_except_final): Set default_case_nonstandard instead
33345         of failing if contiguous_range and the default case doesn't have empty
33346         block.
33347         (check_final_bb): Add SWTCH argument, don't fail if contiguous_range
33348         and only the default case doesn't have the required constants.  Skip
33349         virtual phis.
33350         (gather_default_values): Skip virtual phis.  Allow non-NULL CASE_LOW
33351         if default_case_nonstandard.
33352         (build_constructors): Build constant 1 just once.  Assert that default
33353         values aren't inserted in between cases if contiguous_range.  Skip
33354         virtual phis.
33355         (build_arrays): Skip virtual phis.
33356         (prune_bbs): Add DEFAULT_BB argument, don't remove that bb.
33357         (fix_phi_nodes): Don't add e2f phi arg if default_case_nonstandard.
33358         Handle virtual phis.
33359         (gen_inbound_check): Handle default_case_nonstandard case.
33360         (process_switch): Adjust check_final_bb caller.  Call
33361         gather_default_values with the first non-default case instead of
33362         default case if default_case_nonstandard.
33364 2017-05-02  Nathan Sidwell  <nathan@acm.org>
33366         * ggc-page.c (move_ptes_to_front): Replace unsigned >0 with i--
33367         check.  Fix formatting.
33369 2017-05-02  Jan Hubicka  <hubicka@ucw.cz>
33371         * ipa-inline-analysis.c (estimate_node_size_and_time): Allow roundoff
33372         errors when comparing specialized and unspecialized times.
33374 2017-05-02  David Malcolm  <dmalcolm@redhat.com>
33376         * diagnostic-show-locus.c
33377         (layout::should_print_annotation_line_p): Make private.
33378         (layout::print_annotation_line): Make private.
33379         (layout::annotation_line_showed_range_p): Make private.
33380         (layout::show_ruler): Make private.
33381         (layout::print_source_line): Make private.  Pass in line and
33382         line_width, rather than calling location_get_source_line.  Drop
33383         returned value.
33384         (layout::print_leading_fixits): New method.
33385         (layout::print_any_fixits): Rename to...
33386         (layout::print_trailing_fixits): ...this, and make private.
33387         Don't print newline fixits.
33388         (diagnostic_show_locus): Move logic for printing one row into...
33389         (layout::print_line): ...this new function.  Move the
33390         location_get_source_line call and error-handling from
33391         print_source_line to here.  Call print_leading_fixits, and rename
33392         print_any_fixits to print_trailing_fixits.
33393         (selftest::test_fixit_insert_containing_newline): Update now that
33394         newlines are partially supported.
33395         (selftest::test_fixit_insert_containing_newline_2): New test.
33396         (selftest::test_fixit_replace_containing_newline): Update comments.
33397         (selftest::diagnostic_show_locus_c_tests): Call the new test.
33398         * edit-context.c (class added_line): New class.
33399         (class edited_line): Describe newline handling in comment.
33400         (edited_line::actually_edited_p): New method.
33401         (edited_line::print_content): Delete redundant decl.
33402         (edited_line::m_predecessors): New field.
33403         (edited_file::print_content): Call edited_line::print_content.
33404         (edited_file::print_diff): Update to support newlines.
33405         (edited_file::print_diff_hunk): Likewise.
33406         (edited_file::print_run_of_changed_lines): New function.
33407         (edited_file::print_diff_line): Convert to...
33408         (print_diff_line): ...this.
33409         (edited_file::get_effective_line_count): New function.
33410         (edited_line::edited_line): Initialize new field m_predecessors.
33411         (edited_line::~edited_line): Clean up m_predecessors.
33412         (edited_line::apply_fixit): Handle newlines.
33413         (edited_line::get_effective_line_count): New function.
33414         (edited_line::print_content): New function.
33415         (edited_line::print_diff_lines): New function.
33416         (selftest::test_applying_fixits_insert_containing_newline): New
33417         test.
33418         (selftest::test_applying_fixits_replace_containing_newline): New
33419         test.
33420         (selftest::insert_line): New function.
33421         (selftest::test_applying_fixits_multiple_lines): Add example of
33422         inserting a line.
33423         (selftest::edit_context_c_tests): Call the new tests.
33425 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33427         * tree-ssa-loop-ivopts.c (get_scaled_computation_cost_at): Delete
33428         parameter cand.  Update dump information.
33429         (get_computation_cost): Update uses.
33431 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33433         * tree-ssa-loop-ivopts.c (get_computation_aff_1): New.
33434         (get_computation_aff): Reorder parameters.  Use get_computation_aff_1.
33435         (get_computation_at, rewrite_use_address): Update use of
33436         get_computation_aff.
33438 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33440         * tree-ssa-loop-ivopts.c (get_computation_at): Reorder parameters.
33441         (get_computation): Delete.
33442         (get_computation_cost): Implement like get_computation_cost_at.
33443         Use get_computation_at.
33444         (get_computation_cost_at): Delete.
33445         (rewrite_use_nonlinear_expr): Use get_computation_at.
33446         (rewrite_use_compare, remove_unused_ivs): Ditto.
33448 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33450         * tree-ssa-loop-ivopts.c (rewrite_use_address): Simple refactor.
33452 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33454         * tree-ssa-loop-ivopts.c (struct iv_ca): Rename n_regs to n_invs.
33455         (ivopts_global_cost_for_size): Rename parameter and update uses.
33456         (iv_ca_recount_cost): Update uses.
33457         (iv_ca_set_remove_invs, iv_ca_set_no_cp): Record invariants and
33458         candidates seperately in n_invs and n_cands.
33459         (iv_ca_set_add_invs, iv_ca_set_cp, iv_ca_new): Ditto.
33461 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33463         * tree-ssa-loop-ivopts.c (struct walk_tree_data): New.
33464         (find_inv_vars_cb): New.
33465         (find_depends): Renamed to ...
33466         (find_inv_vars): ... this.
33467         (add_candidate_1, force_var_cost): Call find_inv_vars.
33468         (split_address_cost, determine_group_iv_cost_cond): Ditto.
33470 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33472         * tree-ssa-loop-ivopts.c (struct cost_pair): Rename depends_on to
33473         inv_vars.  Add inv_exprs.
33474         (struct iv_cand): Rename depends_on to inv_vars.
33475         (struct ivopts_data): Rename max_inv_id/n_invariant_uses to
33476         max_inv_var_id/n_inv_var_uses.  Move max_inv_expr_id around.
33477         Refactor field used_inv_exprs from has_map to array n_inv_expr_uses.
33478         (dump_cand): Dump inv_vars.
33479         (tree_ssa_iv_optimize_init): Support inv_vars and inv_exprs.
33480         (record_invariant, find_depends, add_candidate_1): Ditto.
33481         (set_group_iv_cost, force_var_cost): Ditto.
33482         (split_address_cost, ptr_difference_cost, difference_cost): Ditto.
33483         (get_computation_cost_at, get_computation_cost): Ditto.
33484         (determine_group_iv_cost_generic): Ditto.
33485         (determine_group_iv_cost_address): Ditto.
33486         (determine_group_iv_cost_cond, autoinc_possible_for_pair): Ditto.
33487         (determine_group_iv_costs): Ditto.
33488         (iv_ca_recount_cost): Update call to ivopts_global_cost_for_size.
33489         (iv_ca_set_remove_invariants): Renamed to ...
33490         (iv_ca_set_remove_invs): ... this.  Support inv_vars and inv_exprs.
33491         (iv_ca_set_no_cp): Use iv_ca_set_remove_invs.
33492         (iv_ca_set_add_invariants):  Renamed to ...
33493         (iv_ca_set_add_invs): ... this.  Support inv_vars and inv_exprs.
33494         (iv_ca_set_cp): Use iv_ca_set_add_invs.
33495         (iv_ca_has_deps): Support inv_vars and inv_exprs.
33496         (iv_ca_new, iv_ca_free, iv_ca_dump, free_loop_data): Ditto.
33497         (create_new_ivs): Remove useless dump.
33499 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33501         * tree-ssa-loop-ivopts.c (get_computation_cost_at): Remove pseudo
33502         iv_cand code.
33503         (determine_group_iv_cost_cond, determine_iv_cost): Ditto.
33504         (iv_ca_set_no_cp, create_new_iv): Ditto.
33506 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33508         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
33510 2017-05-02  Xi Ruoyao  <ryxi@stu.xidian.edu.cn>
33512         * tree.h (EXPR_CILK_SPAWN): Use macro TREE_CHECK2 instead of
33513         function tree_check2.
33515 2017-05-02  Martin Liska  <mliska@suse.cz>
33517         * doc/gcov.texi: Add missing preposition.
33518         * gcov.c (function_info::function_info): Properly fill up
33519         all member variables.
33521 2017-05-02  Tamar Christina  <tamar.christina@arm.com>
33523         * expr.c (expand_expr_real_2): Re-cost if previous costs are the same.
33525 2017-05-02  Tamar Christina  <tamar.christina@arm.com>
33527         * simplify-rtx.c (simplify_binary_operation_1): Add LSHIFTRT case.
33529 2017-05-02  Martin Liska  <mliska@suse.cz>
33531         PR lto/77954.
33532         * lto-streamer-in.c (lto_read_tree_1): Remove
33533         LTO_STREAMER_DEBUG.
33534         * lto-streamer.c (struct tree_hash_entry): Likewise.
33535         (struct tree_entry_hasher): Likewise.
33536         (tree_entry_hasher::hash): Likewise.
33537         (tree_entry_hasher::equal): Likewise.
33538         (lto_streamer_init): Likewise.
33539         (lto_orig_address_map): Likewise.
33540         (lto_orig_address_get): Likewise.
33541         (lto_orig_address_remove): Likewise.
33542         * lto-streamer.h: Likewise.
33543         * tree-streamer-in.c (streamer_alloc_tree): Likewise.
33544         * tree-streamer-out.c (streamer_write_tree_header): Likewise.
33546 2017-05-02  Sebastian Peryt  <sebastian.peryt@intel.com>
33548         * config/i386/avx512fintrin.h (_mm_mask_add_round_sd)
33549         (_mm_maskz_add_round_sd, _mm_mask_add_round_ss)
33550         (mm_maskz_add_round_ss, _mm_mask_sub_round_sd)
33551         (mm_maskz_sub_round_sd, _mm_mask_sub_round_ss)
33552         (mm_maskz_sub_round_ss, _mm_mask_add_sd)
33553         (mm_maskz_add_sd, _mm_mask_add_ss, _mm_maskz_add_ss)
33554         (mm_mask_sub_sd, _mm_maskz_sub_sd, _mm_mask_sub_ss)
33555         (mm_maskz_sub_ss): New intrinsics.
33556         * config/i386/i386-builtin-types.def (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
33557         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): New function type aliases.
33558         * config/i386/i386-builtin.def (__builtin_ia32_addsd_mask_round)
33559         (__builtin_ia32_addss_mask_round, __builtin_ia32_subsd_mask_round)
33560         (__builtin_ia32_subss_mask_round): New builtins.
33561         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
33562         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
33563         * config/i386/sse.md (<sse>_vm<plusminus_insn><mode>3<round_name>):
33564         Renamed to ...
33565         (<sse>_vm<plusminus_insn><mode>3<mask_name><round_name>): ... this.
33566         (v<plusminus_mnemonic><ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|%0, %1, %<iptr>2<round_op3>}):
33567         Changed to ...
33568         (v<plusminus_mnemonic><ssescalarmodesuffix>\t{<round_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}):
33569         ... this.
33571 2017-05-02  Martin Jambor  <mjambor@suse.cz>
33573         PR tree-optimization/78687
33574         * tree-sra.c (access): New field parent.
33575         (process_subtree_disqualification): New function.
33576         (disqualify_candidate): Call it.
33577         (build_accesses_from_assign): Reset write flag if creating an
33578         assighnment link.
33579         (build_access_subtree): Fill in parent field and also prpagate
33580         down grp_write flag.
33581         (create_artificial_child_access): New parameter set_grp_write, set
33582         grp_write to its value.
33583         (propagate_subaccesses_across_link): Also propagate grp_write flag
33584         values.
33585         (propagate_all_subaccesses): Push the closest parent back to work
33586         queue if add_access_to_work_queue returned true.
33588 2017-05-02  Richard Biener  <rguenther@suse.de>
33590         * common.opt (fstrict-overflow): Alias negative to fwrapv.
33591         * doc/invoke.texi (fstrict-overflow): Remove all traces of
33592         -fstrict-overflow documentation.
33593         * tree.h (TYPE_OVERFLOW_UNDEFINED): Do not test flag_strict_overflow.
33594         (POINTER_TYPE_OVERFLOW_UNDEFINED): Test !flag_wrapv instead of
33595         flag_strict_overflow.
33596         * ipa-inline.c (can_inline_edge_p): Do not test flag_strict_overflow.
33597         * lto-opts.c (lto_write_options): Do not stream it.
33598         * lto-wrapper.c (merge_and_complain): Do not handle it.
33599         * opts.c (default_options_table): Do not set -fstrict-overflow.
33600         (finish_options): Likewise do not clear it when sanitizing.
33601         * simplify-rtx.c (simplify_const_relational_operation): Do not
33602         test flag_strict_overflow.
33604 2017-05-02  Uros Bizjak  <ubizjak@gmail.com>
33606         * config/alpha/alpha.md (*add<mode>3_ieee): Merge to add<mode>3
33607         using enabled attribute.
33608         (*sub<mode>3_ieee): Merge to sub<mode>3 using enabled attribute.
33609         (*mul<mode>3_ieee): Merge to mul<mode>3 using enabled attribute.
33610         (*div<mode>3_ieee): Merge to div<mode>3 using enabled attribute.
33611         (*sqrt<mode>2_ieee): Merge to sqrt<mode>2 using enabled attribute.
33612         (*fix_truncdfdi_ieee): Merge to *fix_truncdfdi2 using enabled attribute.
33613         (*fix_truncsfdi_ieee): Merge to *fix_truncsfdi2 using enabled attribute.
33614         (*floatdisf_ieee): Merge to floatdisf2 using enabled attribute.
33615         (*floatdidf_ieee): Merge to floatdidf2 using enabled attribute.
33616         (*truncdfsf2_ieee): Merge to truncdfsf2 using enabled attribute.
33617         (*cmpdf_ieee): Merge to *cmpdf_internal using enabled attribute.
33619 2017-05-02  Uros Bizjak  <ubizjak@gmail.com>
33621         * config/i386/i386.c (ix86_code_end): Use {FIRST,LAST}_INT_REG.
33623 2017-05-02  Richard Biener  <rguenther@suse.de>
33625         PR tree-optimization/80591
33626         Revert
33627         2017-04-10  Richard Biener  <rguenther@suse.de>
33629         * tree-ssa-structalias.c (find_func_aliases): Properly handle
33630         asm inputs.
33632 2017-05-02  Richard Biener  <rguenther@suse.de>
33634         PR tree-optimization/80549
33635         * tree-cfgcleanup.c (mfb_keep_latches): New helper.
33636         (cleanup_tree_cfg_noloop): Create forwarders to known loop
33637         headers if they do not have a preheader.
33639 2017-05-02  Martin Liska  <mliska@suse.cz>
33641         PR other/80589
33642         * common.opt: Fix typo.
33643         * doc/invoke.texi: Likewise.
33645 2017-05-01  Jan Beulich  <jbeulich@suse.com>
33647         * config/i386/sse.md (xop_vpermil2<mode>3): Do not allow operand
33648         swapping, add (x,x,m,x,n) alternative.
33650 2017-05-01  Nathan Sidwell  <nathan@acm.org>
33652         * calls.c (combine_pending_stack_adjustment_and_call): Remove
33653         unnecessary unadjusted_alignment check.
33655 2017-05-01  Xi Ruoyao  <ryxi@stu.xidian.edu.cn>
33657         PR c++/80038
33658         * cilk_common.c (expand_builtin_cilk_detach): Move pedigree
33659         operations here.
33660         * gimplify.c (gimplify_cilk_detach): New function.
33661         (gimplify_call_expr, gimplify_modify_expr): Call it as needed.
33662         * tree-core.h: Document EXPR_CILK_SPAWN.
33663         * tree.h (EXPR_CILK_SPAWN): Define.
33665 2017-05-01  David Malcolm  <dmalcolm@redhat.com>
33667         * diagnostic-show-locus.c (layout::get_expanded_location): Rewrite
33668         to use new fixit_hint representation, using the "replace" logic.
33669         (get_line_span_for_fixit_hint): Likewise.
33670         (layout::print_any_fixits): Likewise.
33671         (selftest::test_one_liner_many_fixits): Rename to...
33672         (selftest::test_one_liner_many_fixits_1): ...this, and update
33673         comment and expected output to reflect that the multiple fix-it
33674         hints are now consolidated into one insertion.
33675         (selftest::test_one_liner_many_fixits_2): New test.
33676         (selftest::test_diagnostic_show_locus_one_liner): Update for
33677         above.
33678         (selftest::test_fixit_consolidation): Update for fix-it API
33679         change.
33680         * diagnostic.c (print_parseable_fixits): Likewise.
33681         * edit-context.c (edited_line::m_line_events): Convert from
33682         auto_vec <line_event *> to auto_vec <line_event>.
33683         (class line_event): Convert from abstract base class to a concrete
33684         class, taking over the role of replace_event.
33685         (class insert_event): Delete.
33686         (class replace_event): Rename to class line_event.  Convert to
33687         half-open range.
33688         (edit_context::add_fixits): Reimplement.
33689         (edit_context::apply_insert): Delete.
33690         (edit_context::apply_replace): Rename to...
33691         (edit_context::apply_fixit): ...this.  Convert to half-open range.
33692         (edited_file::apply_insert): Delete.
33693         (edited_file::apply_replace): Rename to...
33694         (edited_file::apply_fixit): ...this.
33695         (edited_line::~edited_line): Drop deletion of events.
33696         (edited_line::apply_insert): Delete.
33697         (edited_line::apply_replace): Rename to...
33698         (edited_line::apply_fixit): ...this.  Convert to half-open range.
33699         Update for change to type of m_line_events.
33700         * edit-context.h (edit_context::apply_insert): Delete.
33701         (edit_context::apply_replace): Rename to...
33702         (edit_context::apply_fixit): ...this.
33704 2017-05-01  Martin Sebor  <msebor@redhat.com>
33706         * gimple-ssa-sprintf.c (format_integer): Set knownrange when it's
33707         known.
33709 2017-05-01  Uros Bizjak  <ubizjak@gmail.com>
33711         PR target/68491
33712         * config/i386/cpuid.h (__get_cpuid): Always return 0 when
33713         __get_cpuid_max returns 0.
33714         (__get_cpuid_count): Ditto.
33716 2017-05-01  Eric Botcazou  <ebotcazou@adacore.com>
33718         * tree.c (substitute_in_expr) <tcc_vl_exp>: Also inline a call if the
33719         replacement expression is another instance of one of its arguments.
33721 2017-05-01  Jakub Jelinek  <jakub@redhat.com>
33723         PR target/79430
33724         * rtlanal.c (reg_set_p): If reg is a stack_pointer_rtx, also
33725         check for stack push/pop autoinc.
33726         * config/i386/i386.c (ix86_agi_dependent): Return false
33727         if the only reason why modified_in_p returned true is that
33728         addr is SP based and set_insn is a push or pop.
33730 2017-04-29  Jan Hubicka  <hubicka@ucw.cz>
33732         * ipa-inline.c (compute_inlined_call_time): Remove now unnecesary
33733         overflow check.
33735 2017-04-29  Jan Hubicka  <hubicka@ucw.cz>
33737         PR ipa/79224
33738         * ipa-inline-analysis.c (dump_predicate): Add optional parameter NL.
33739         (account_size_time): Use two predicates - exec_pred and
33740         nonconst_pred_ptr.
33741         (evaluate_conditions_for_known_args): Compute both clause and
33742         nonspec_clause.
33743         (evaluate_properties_for_edge): Evaulate both clause and nonspec_clause.
33744         (inline_summary_t::duplicate): Update.
33745         (estimate_function_body_sizes): Caluculate exec and nonconst predicates
33746         separately.
33747         (compute_inline_parameters): Likewise.
33748         (estimate_edge_size_and_time): Update caluclation of time.
33749         (estimate_node_size_and_time): Compute both time and nonspecialized
33750         time.
33751         (estimate_ipcp_clone_size_and_time): Update.
33752         (inline_merge_summary): Update.
33753         (do_estimate_edge_time): Update.
33754         (do_estimate_edge_size): Update.
33755         (do_estimate_edge_hints): Update.
33756         (inline_read_section, inline_write_summary): Stream both new predicates.
33757         * ipa-inline.c (compute_uninlined_call_time): Take uninlined_call_time
33758         as argument.
33759         (compute_inlined_call_time): Cleanup.
33760         (big_speedup_p): Update.
33761         (edge_badness): Update.
33762         * ipa-inline.h (INLINE_TIME_SCALE): Remove.
33763         (size_time_entry): Replace predicate by exec_predicate and
33764         nonconst_predicate.
33765         (edge_growth_cache_entry): Cache both time nad nonspecialized time.
33766         (estimate_edge_time): Return also nonspec_time.
33767         (reset_edge_growth_cache): Update.
33769 2017-04-29  Jakub Jelinek  <jakub@redhat.com>
33771         PR rtl-optimization/80491
33772         * ifcvt.c (noce_process_if_block): When looking for x setter
33773         with missing else_bb, don't check only the insn right before
33774         cond_earliest, but look for the last insn that x is modified in
33775         within the same bb.
33777         PR rtl-optimization/80491
33778         * alias.c (memory_modified_in_insn_p): Return true for CALL_INSNs.
33780 2017-04-29  Marc Glisse  <marc.glisse@inria.fr>
33782         PR tree-optimization/80487
33783         * tree-ssa-alias.c (stmt_kills_ref_p): Handle stpncpy and strncpy.
33785 2017-04-29  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
33787         PR tree-optimization/79697
33788         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Check if callee
33789         is BUILT_IN_STRDUP, BUILT_IN_STRNDUP, BUILT_IN_REALLOC.
33790         (propagate_necessity): Check if def_callee is BUILT_IN_STRDUP or
33791         BUILT_IN_STRNDUP.
33792         * gimple-fold.c (gimple_fold_builtin_realloc): New function.
33793         (gimple_fold_builtin): Call gimple_fold_builtin_realloc.
33795 2017-04-28  Martin Sebor  <msebor@redhat.com>
33797         PR tree-optimization/80523
33798         * gimple-ssa-sprintf.c (target_to_host_charmap): New global variable.
33799         (init_target_to_host_charmap, target_to_host, target_strtol10): New
33800         functions.
33801         (maybe_warn, format_directive, parse_directive): Use new functions.
33802         (pass_sprintf_length::execute): Call init_target_to_host_charmap.
33804 2017-04-28  Marc Glisse  <marc.glisse@inria.fr>
33806         * match.pd (X+Z OP Y+Z, X-Z OP Y-Z, Z-X OP Z-Y): New transformations.
33808 2017-04-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
33810         * configure.ac (SYSTEM_HEADER_DIR, BUILD_SYSTEM_HEADER_DIR,
33811         target_header_dir): Set correctly.
33812         * configure: Regenerated.
33813         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
33814         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
33815         instead of SYSTEM_HEADER_DIR.
33817 2017-04-28  Jan Hubicka  <hubicka@ucw.cz>
33819         * ipa-cp.c (perform_estimation_of_a_value): Turn time to sreal.
33820         (estimate_local_effects): Likewise.
33821         * ipa-inline.c (compute_inlined_call_time, want_inline_small_function_p,
33822         edge_badness, inline_small_functions, dump_overall_stats): LIkewise.
33823         * ipa-inline.h (edge_growth_cache_entry, estimate_time_after_inlining,
33824         estimate_ipcp_clone_size_and_time, do_estimate_edge_time,
33825         do_estimate_edge_time, estimate_edge_time): Likewise.
33826         * ipa-inline-analysis.c (estimate_node_size_and_time,
33827         estimate_ipcp_clone_size_and_time, do_estimate_edge_time): Likewise.
33828         (estimate_time_after_inlining): Remove.
33830 2017-04-28  Martin Liska  <mliska@suse.cz>
33832         * doc/gcov.texi: Enhance documentation of gcov.
33834 2017-04-28  Martin Liska  <mliska@suse.cz>
33836         * doc/gcov.texi: Sort options in alphabetic order.
33837         * doc/gcov-dump.texi: Likewise.
33838         * doc/gcov-tool.texi: Likewise.
33839         * gcov.c (print_usage): Likewise.
33840         * gcov-dump.c (print_usage): Likewise.
33841         * gcov-tool.c (print_merge_usage_message): Likewise.
33842         (print_rewrite_usage_message): Likewise.
33843         (print_overlap_usage_message): Likewise.
33845 2017-04-28  Martin Liska  <mliska@suse.cz>
33847         PR gcov-profile/53915
33848         * gcov.c (format_gcov): Print 'NAN %' when top > bottom.
33850 2017-04-28  Martin Liska  <mliska@suse.cz>
33852         PR gcov-profile/79891
33853         * gcov.c (add_line_counts): Assign BBs to lines just if the BB
33854         is marked by compiler as living on a line.
33855         (get_cycles_count): Remove usage of the union.
33856         (output_intermediate_file): Likewise.
33857         (find_source): Fix GNU coding style.
33858         (accumulate_line_counts): Remove old non-all block mode.
33859         (output_lines): Remove usage of the union.
33860         * profile.c (output_location): Include all BBs, even if
33861         belonging to a same line (and file) as a previous BB.
33863 2017-04-28  Martin Liska  <mliska@suse.cz>
33865         * gcov.c (process_args): Handle new argument 'w'.
33866         (read_graph_file): Assign ID to BBs.
33867         (output_branch_count): Display BB # if verbose flag is set.
33868         (output_lines): Likewise for arcs.
33869         (print_usage): Add '--verbose' option help.
33870         * doc/gcov.texi: Document --verbose (-w) option.
33872 2017-04-28  Martin Liska  <mliska@suse.cz>
33874         * gcov.c (struct block_location_info): New struct.
33875         (process_file): Fill up the new structure.
33876         (read_graph_file): Replace usage of encoding by the newly added
33877         struct.
33878         (add_line_counts): Likewise.
33879         (accumulate_line_counts): Remove usage of the union.
33880         (function_info::function_info): New function.
33881         (function_info::~function_info): Likewise.
33882         (process_file): Call delete instead of release_function.
33883         (release_function): Release the function.
33884         (release_structures): Call delete instead of release_function.
33885         (solve_flow_graph): Replace usage of num_blocks.
33886         (find_exception_blocks): Likewise.
33887         (output_lines): Fix GNU coding style.
33889 2017-04-28  Martin Liska  <mliska@suse.cz>
33891         PR driver/56469
33892         * coverage.c (coverage_remove_note_file): New function.
33893         * coverage.h: Declare the function.
33894         * toplev.c (finalize): Clean if an error has been seen.
33896 2017-04-28  Martin Liska  <mliska@suse.cz>
33898         PR gcov-profile/80031
33899         * gcov-dump.c (tag_blocks): Just print number of basic blocks.
33900         * gcov-io.h (GCOV_TAG_BLOCKS_NUM): Remove unused macro.
33901         * gcov.c (read_graph_file): Read just number of blocks.
33902         * profile.c (branch_prob): Do not stream 0 flags per a basic
33903         block.
33905 2017-04-28  Martin Liska  <mliska@suse.cz>
33907         * gcov-dump.c (tag_*): Add new argument to declarations.
33908         (dump_gcov_file): Likewise.
33909         (tag_blocks): Add and use new argument depth.
33910         (tag_arcs): Likewise.
33911         (tag_lines): Likewise.
33912         (tag_counters): Likewise.
33913         (tag_summary): Likewise.
33914         (dump_working_sets): Use depth to do a proper indentation.
33916 2017-04-28  Jakub Jelinek  <jakub@redhat.com>
33918         PR bootstrap/80531
33919         * cgraph.h (symtab_node::debug_symtab): No longer inline.
33920         * symtab.c (symtab_node::debug_symtab): Move definition here.
33922 2017-04-28  Richard Biener  <rguenther@suse.de>
33924         * lto-streamer.h (LTO_major_version): Bump to 7.
33926 2017-04-28  Richard Biener  <rguenther@suse.de>
33928         * tree-vrp.c (assert_info): New struct.
33929         (add_assert_info): New helper.
33930         (register_edge_assert_for_2): Refactor to add asserts to a vector
33931         of assert_info.
33932         (register_edge_assert_for_1): Likewise.
33933         (register_edge_assert_for): Likewise.
33934         (finish_register_edge_assert_for): New helper actually registering
33935         asserts where live on edge.
33936         (find_conditional_asserts): Adjust.
33937         (find_switch_asserts): Likewise.
33938         (evrp_dom_walker::try_find_new_range): Generalize.
33939         (evrp_dom_walker::before_dom_children): Use register_edge_assert_for.
33941 2017-04-27  Marek Polacek  <polacek@redhat.com>
33943         PR sanitizer/80349
33944         * fold-const.c (fold_binary_loc) <case EQ_EXPR, NE_EXPR>: Convert
33945         arg10 and arg11 to itype.
33947 2017-04-27  Jonathan Wakely  <jwakely@redhat.com>
33949         * doc/extend.texi (Object Size Checking): Improve grammar.
33951 2017-04-27  Richard Earnshaw  <rearnsha@arm.com>
33953         PR target/80530
33954         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Ensure
33955         that the logic for permitting reciprocal estimates matches that
33956         in use_rsqrt_p.
33958 2017-04-27  Jakub Jelinek  <jakub@redhat.com>
33960         PR c++/80534
33961         * tree.c (type_cache_hasher::equal): Only compare
33962         TYPE_TYPELESS_STORAGE flag on non-aggregate element types.
33963         (build_array_type_1): Only hash TYPE_TYPELESS_STORAGE flag on
33964         non-aggregate element types.
33965         * tree.h (TYPE_TYPELESS_STORAGE): Fix comment typo, add more details
33966         about the flag on ARRAY_TYPEs in the comment, formatting fix.
33968 2017-04-27  Richard Biener  <rguenther@suse.de>
33970         PR middle-end/80533
33971         * emit-rtl.c (set_mem_attributes_minus_bitpos): When
33972         stripping ARRAY_REFs from MEM_EXPR make sure we're not
33973         keeping a reference to a trailing array.
33975 2017-04-27  Richard Biener  <rguenther@suse.de>
33977         PR middle-end/80539
33978         * tree-chrec.c (chrec_fold_plus_poly_poly): Deal with not
33979         being in loop-closed SSA form conservatively.
33980         (chrec_fold_multiply_poly_poly): Likewise.
33982 2017-04-27  Tamar Christina  <tamar.christina@arm.com>
33984         PR middle-end/79665
33985         * expr.c (expand_expr_real_2): Move TRUNC_MOD_EXPR, FLOOR_MOD_EXPR,
33986         CEIL_MOD_EXPR, ROUND_MOD_EXPR cases.
33988 2017-04-27  Jakub Jelinek  <jakub@redhat.com>
33990         PR target/77728
33991         * config/aarch64/aarch64.c (struct aarch64_fn_arg_alignment): Remove.
33992         (aarch64_function_arg_alignment): Return unsigned int again, but still
33993         ignore TYPE_FIELDS chain decls other than FIELD_DECLs.
33994         (aarch64_layout_arg): Adjust aarch64_function_arg_alignment caller.
33995         Don't emit -Wpsabi note.
33996         (aarch64_function_arg_boundary): Likewise.
33997         (aarch64_gimplify_va_arg_expr): Adjust aarch64_function_arg_alignment
33998         caller.
34000 2017-04-26  Nathan Sidwell  <nathan@acm.org>
34002         * tree.h (crc32_unsigned_n): Declare.
34003         (crc32_unsigned, crc32_unsigned): Make inline.
34004         * tree.c (crc32_unsigned_bits): Replace with ...
34005         (crc32_unsigned_n): ... this.
34006         (crc32_unsigned, crc32_byte): Remove.
34007         (crc32_string): Remove unnecessary braces.
34009 2017-04-25  Jan Hubicka  <hubicka@ucw.cz>
34011         * ipa-cp.c (estimate_local_effects): Convert sreal to int.
34012         * ipa-inline-analysis.c (MAX_TIME): Remove.
34013         (account_size_time): Use sreal for time.
34014         (dump_inline_summary): Update.
34015         (estimate_function_body_sizes): Update.
34016         (estimate_edge_size_and_time): Update.
34017         (estimate_calls_size_and_time): Update.
34018         (estimate_node_size_and_time): Update.
34019         (inline_merge_summary): Update.
34020         (inline_update_overall_summary): Update.
34021         (estimate_time_after_inlining): Update.
34022         (inline_read_section): Update.
34023         (inline_write_summary): Update.
34024         * ipa-inline.c (compute_uninlined_call_time): Update.
34025         (compute_inlined_call_time): Update.
34026         (recursive_inlining): Update.
34027         (inline_small_functions): Update.
34028         (dump_overall_stats): Update.
34029         * ipa-inline.h: Include sreal.h.
34030         (size_time_entry): Turn time to sreal.
34031         (inline_summary): Turn self_time nad time to sreal.
34033 2017-04-25  Jan Hubicka  <hubicka@ucw.cz>
34035         * sreal.c: Include backend.h, tree.h, gimple.h, cgraph.h and
34036         data-streamer.h
34037         (sreal::stream_out, sreal::stream_in): New.
34038         * sreal.h (sreal::stream_out, sreal::stream_in): Declare.
34040 2017-04-25  Jakub Jelinek  <jakub@redhat.com>
34042         * Makefile.in (s-options): Invoke opt-gather.awk with LC_ALL=C in the
34043         environment.
34045 2017-04-25  Uros Bizjak  <ubizjak@gmail.com>
34047         PR target/70799
34048         * config/i386/i386.c (dimode_scalar_to_vector_candidate_p):
34049         Handle ASHIFTRT.
34050         (dimode_scalar_chain::compute_convert_gain): Ditto.
34051         (dimode_scalar_chain::make_vector_copies): Ditto.
34052         (dimode_scalar_chain::convert_reg): Ditto.
34053         (dimode_scalar_chain::convert_insn): Ditto.
34054         * config/i386/sse.md (VI24_AVX512BW_1): Remove mode iterator.
34055         (VI248_AVX512BW_1): New mode iterator.
34056         (<mask_codefor>ashr<mode>3<mask_name>): Merge insn pattern with
34057         <mask_codefor>ashrv2di3<mask_name> insn using VI248_AVX512BW_1
34058         mode iterator.
34060 2017-04-25  Martin Sebor  <msebor@redhat.com>
34062         PR tree-optimization/80497
34063         * gimple-ssa-sprintf.c (get_int_range): Avoid assuming all integer
34064         constants are representable in HOST_WIDE_INT.
34065         (parse_directive): Ditto.
34067 2017-04-25  Martin Sebor  <msebor@redhat.com>
34069         PR bootstrap/80486
34070         * dominance.c (dom_info::m_n_basic_blocks): Change type to unsigned.
34071         (new_zero_array): Adjust signature.
34072         (dom_info::dom_init): Used unsigned rather that size_t.
34073         (dom_info::dom_info): Same.
34075 2017-04-25  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
34076             Jakub Jelinek  <jakub@redhat.com>
34078         PR target/77728
34079         * config/arm/arm.c: Include gimple.h.
34080         (aapcs_layout_arg): Emit -Wpsabi note if arm_needs_doubleword_align
34081         returns negative, increment ncrn only if it returned positive.
34082         (arm_needs_doubleword_align): Return int instead of bool,
34083         ignore DECL_ALIGN of non-FIELD_DECL TYPE_FIELDS chain
34084         members, but if there is any such non-FIELD_DECL
34085         > PARM_BOUNDARY aligned decl, return -1 instead of false.
34086         (arm_function_arg): Emit -Wpsabi note if arm_needs_doubleword_align
34087         returns negative, increment nregs only if it returned positive.
34088         (arm_setup_incoming_varargs): Likewise.
34089         (arm_function_arg_boundary): Emit -Wpsabi note if
34090         arm_needs_doubleword_align returns negative, return
34091         DOUBLEWORD_ALIGNMENT only if it returned positive.
34093 2017-04-25  Marek Polacek  <polacek@redhat.com>
34095         PR sanitizer/80349
34096         * fold-const.c (fold_binary_loc) <case BIT_IOR_EXPR>: Convert arg0's
34097         first argument to type.
34099 2017-04-25  Bill Seurer  <seurer@linux.vnet.ibm.com>
34101         PR target/80482
34102         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Change
34103         type checks to test for compatibility instead of equality.
34105 2017-04-25  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
34106             Jakub Jelinek  <jakub@redhat.com>
34108         PR target/77728
34109         * config/aarch64/aarch64.c (struct aarch64_fn_arg_alignment): New
34110         type.
34111         (aarch64_function_arg_alignment): Return aarch64_fn_arg_alignment
34112         struct.  Ignore DECL_ALIGN of decls other than FIELD_DECL for
34113         the alignment computation, but return their maximum in warn_alignment.
34114         (aarch64_layout_arg): Adjust aarch64_function_arg_alignment caller.
34115         Emit a -Wpsabi note if warn_alignment is 16 bytes, but alignment
34116         is smaller.
34117         (aarch64_function_arg_boundary): Likewise.  Simplify using MIN/MAX.
34118         (aarch64_gimplify_va_arg_expr): Adjust aarch64_function_arg_alignment
34119         caller.
34121 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34123         * config/arc/simdext.md (dmpyh): Fix typo.
34125 2017-04-25  Richard Biener  <rguenther@suse.de>
34127         PR tree-optimization/80492
34128         * alias.c (compare_base_decls): Handle registers with asm
34129         specification conservatively.
34130         * tree-ssa-alias.c (decl_refs_may_alias_p): Handle
34131         compare_base_decls returning dont-know properly.
34133 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34135         * config/arc/arc.c (LEGITIMATE_OFFSET_ADDRESS_P): Delete macro.
34136         (legitimate_offset_address_p): New function.
34137         (arc_legitimate_address_p): Use above function.
34139 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34141         * config/arc/arc.c (arc_output_mi_thunk): Emit PIC calls.
34143 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34145         * config/arc/arc.c (arc_conditional_register_usage): Use ACCL,
34146         ACCH registers whenever they are available.
34148 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34150         * config/arc/arc.c (arc_conditional_register_usage): Make D0, D1
34151         double regs fix when not used.
34153 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34155         * config/arc/arc.h (REGNO_OK_FOR_BASE_P): Consider also extension
34156         core registers.
34157         (REG_OK_FOR_INDEX_P_NONSTRICT): Likewise.
34158         (REG_OK_FOR_BASE_P_NONSTRICT): Likewise.
34160 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34162         * config/arc/arc.c (arc_output_addsi): Check for h-register class
34163         when emitting short ADD instructions.
34165 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34167         * config/arc/arc.md (cmpsi_cc_insn_mixed): Use 'h' register
34168         constraint.
34169         (cmpsi_cc_c_insn): Likewise.
34170         (cbranchsi4_scratch): Compute proper instruction length using
34171         compact_hreg_operand.
34172         * config/arc/predicates.md (compact_hreg_operand): New predicate.
34174 2017-04-25  Richard Biener  <rguenther@suse.de>
34176         PR middle-end/80509
34177         * passes.c (pass_manager::pass_manager): Initialize
34178         m_name_to_pass_map.
34180 2017-04-25  Richard Biener  <rguenther@suse.de>
34182         PR tree-optimization/79201
34183         * tree-ssa-sink.c (statement_sink_location): Handle calls.
34185 2017-04-25  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
34187         PR target/80464
34188         * config/s390/vector.md: Split MEM->GPR vector moves for
34189         non-s_operand addresses.
34191 2017-04-25  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
34193         PR target/79895
34194         * config/s390/predicates.md (reload_const_wide_int_operand): New
34195         predicate.
34196         * config/s390/s390.md ("movti"): Remove d/P alternative.
34197         ("movti_bigconst"): New pattern definition.
34199 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
34201         PR target/80080
34202         * s390-protos.h (s390_expand_cs_hqi): Removed.
34203         (s390_expand_cs, s390_expand_atomic_exchange_tdsi): New prototypes.
34204         * config/s390/s390.c (s390_emit_compare_and_swap): Handle all integer
34205         modes as well as CCZ1mode and CCZmode.
34206         (s390_expand_atomic_exchange_tdsi, s390_expand_atomic): Adapt to new
34207         signature of s390_emit_compare_and_swap.
34208         (s390_expand_cs_hqi): Likewise, make static.
34209         (s390_expand_cs_tdsi): Generate an explicit compare before trying
34210         compare-and-swap, in some cases.
34211         (s390_expand_cs): Wrapper function.
34212         (s390_expand_atomic_exchange_tdsi): New backend specific expander for
34213         atomic_exchange.
34214         (s390_match_ccmode_set): Allow CCZmode <-> CCZ1 mode.
34215         * config/s390/s390.md ("atomic_compare_and_swap<mode>"): Merge the
34216         patterns for small and large integers.  Forbid symref memory operands.
34217         Move expander to s390.c.  Require cc register.
34218         ("atomic_compare_and_swap<DGPR:mode><CCZZ1:mode>_internal")
34219         ("*atomic_compare_and_swap<TDI:mode><CCZZ1:mode>_1")
34220         ("*atomic_compare_and_swapdi<CCZZ1:mode>_2")
34221         ("*atomic_compare_and_swapsi<CCZZ1:mode>_3"): Use s_operand to forbid
34222         symref memory operands.  Remove CC mode and call s390_match_ccmode
34223         instead.
34224         ("atomic_exchange<mode>"): Allow and implement all integer modes.
34226 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
34228         * config/s390/s390.md (define_peephole2): New peephole to help
34229         combining the load-and-test pattern with volatile memory.
34231 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
34233         * config/s390/s390.md ("cstorecc4"): Use load-on-condition and deal
34234         with CCZmode for TARGET_Z196.
34236 2017-04-25  Jakub Jelinek  <jakub@redhat.com>
34238         PR rtl-optimization/80501
34239         * combine.c (make_compound_operation_int): Set subreg_code to SET
34240         even for AND with mask of the sign bit of mode.
34242         PR rtl-optimization/80500
34243         * loop-unroll.c (combine_var_copies_in_loop_exit): Call copy_rtx on
34244         sum's initial value.
34246 2017-04-25  Julian Brown  <julian@codesourcery.com>
34247             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
34249         * config/aarch64/thunderx2t99.md (thunderx2t99_crc): New Reservation.
34251 2017-04-25  Marc Glisse  <marc.glisse@inria.fr>
34253         * fold-const.c (tree_single_nonzero_warnv_p): Handle SSA_NAME.
34255 2017-04-25  Julian Brown  <julian@codesourcery.com>
34256             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
34258         * config/aarch64/thunderx2t99.md (thunderx2t99_aes): New Reservation.
34259         (thunderx2t99_sha): New Reservation.
34261 2017-04-25  Julian Brown  <julian@codesourcery.com>
34262             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
34264         * config/aarch64/aarch64-simd.md (aarch64_simd_vec_set<mode>): Fix
34265         type for 1-element load.
34267 2017-04-24  Marc Glisse  <marc.glisse@inria.fr>
34269         * match.pd (X/[ex]C CMP Y/[ex]C): New transformation.
34271 2017-04-24  Martin Jambor  <mjambor@suse.cz>
34273         PR tree-optimization/80293
34274         * tree-sra.c (scalarizable_type_p): New parameter const_decl, make
34275         char arrays not totally scalarizable if it is false.
34276         (analyze_all_variable_accesses): Pass correct value in the new
34277         parameter.  Add a statistics counter.
34279 2017-04-24  Jan Hubicka  <hubicka@ucw.cz>
34281         PR middle-end/79931
34282         * ipa-devirt.c (dump_possible_polymorphic_call_targets): Fix ICE.
34284 2017-04-24  Richard Biener  <rguenther@suse.de>
34286         PR tree-optimization/80494
34287         * tree-scalar-evolution.c (analyze_scalar_evolution_1): Bail
34288         out for complex types.
34290 2017-04-24  Richard Biener  <rguenther@suse.de>
34292         * tree-ssa-sccvn.h (run_scc_vn): Adjust prototype.
34293         * tree-ssa-sccvn.c (print_scc): Print SCC size.
34294         (extract_and_process_scc_for_name): Never fail but drop SCC to varying.
34295         (DFS): Adjust and never fail.
34296         (sccvn_dom_walker::fail): Remove.
34297         (sccvn_dom_walker::before_dom_children): Adjust.
34298         (run_scc_vn): Likewise and never fail.
34299         * tree-ssa-pre.c (pass_pre::execute): Adjust.
34300         (pass_fre::execute): Likewise.
34302 2017-04-24  Richard Biener  <rguenther@suse.de>
34304         PR tree-optimization/79725
34305         * tree-ssa-sink.c (statement_sink_location): Return whether
34306         failure reason was zero uses.  Move that check later.
34307         (sink_code_in_bb): Deal with zero uses by removing the stmt
34308         if possible.
34310 2017-04-24  Richard Biener  <rguenther@suse.de>
34312         PR c++/2972
34313         * tree-ssa-uninit.c (warn_uninitialized_vars): Handle some
34314         pointer-based references.
34316 2017-04-24  Richard Biener  <rguenther@suse.de>
34318         PR bootstrap/79814
34319         * pass_manager.h (pass_manager::operator new): Remove.
34320         (pass_manager::operator delete): Likewise.
34321         * passes.c (pass_manager::operator new): Remove.
34322         (pass_manager::operator delete): Likewise.
34323         (pass_manager::pass_manager): Zero individual pass members.
34325 2017-04-23  Uros Bizjak  <ubizjak@gmail.com>
34327         PR target/70799
34328         * config/i386/i386.c (dimode_scalar_to_vector_candidate_p)
34329         <case ASHIFT, case LSHIFTRT>: Also consider variable shifts.
34330         Check "XEXP (src, 1)" operand here.
34331         <case PLUS, case MINUS, case IOR, case XOR, case AND>:
34332         Check "XEXP (src, 1)" operand here.
34333         (dimode_scalar_chain::make_vector_copies): Detect count register
34334         of a shift instruction.  Zero extend count register from QImode
34335         to DImode to satisfy vector shift pattern count operand predicate.
34336         Substitute vector shift count operand with a DImode copy.
34337         (dimode_scalar_chain::convert_reg): Ditto, zero-extend from
34338         vector register.
34340 2017-04-21  Uros Bizjak  <ubizjak@gmail.com>
34342         * config/i386/i386.md (*extzvqi_mem_rex64): Move above *extzv<mode>.
34343         Remove UNSPEC_NOREX_MEM tag.  Update corresponding peephole2 pattern.
34344         (*insvqi_1_mem_rex64): Move above insv<mode>_1.  Remove
34345         UNSPEC_NOREX_MEM tag.  Update corresponding peephole2 pattern.
34346         (UNSPEC_NOREX_MEM): Remove definition.
34348 2017-04-21  Richard Biener  <rguenther@suse.de>
34350         PR tree-optimization/79547
34351         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
34352         Handle strlen, strcmp, strncmp, strcasecmp, strncasecmp, memcmp,
34353         bcmp, strspn, strcspn, __builtin_object_size and __builtin_constant_p
34354         without any constraints.
34356 2017-04-21  Richard Biener  <rguenther@suse.de>
34358         PR tree-optimization/78847
34359         * fold-const.c (split_tree): Handle POINTER_PLUS_EXPR.
34361 2017-04-21  Richard Biener  <rguenther@suse.de>
34363         * tree.h (build_qualified_type): Annotate with CXX_MEM_STAT_INFO.
34364         (build_distinct_type_copy): Likewise.
34365         (build_variant_type_copy): Likewise.
34366         * tree.c (build_qualified_type): Pass down mem-stat info.
34367         (build_distinct_type_copy): Likewise.
34368         (build_variant_type_copy): Likewise.
34370 2017-04-21  Richard Biener  <rguenther@suse.de>
34372         PR tree-optimization/80237
34373         * tree-ssa-pre.c (find_leader_in_sets): Add third set argument,
34374         defaulted to NULL.
34375         (phi_translate_1): Also allow a leader in AVAIL_OUT of pred
34376         for a simplified result.
34378 2016-04-21  Richard Biener  <rguenther@suse.de>
34380         * tree-ssa-loop-ivcanon.c (constant_after_peeling): Do not require
34381         sth as strict as a simple_iv but a chrec without symbols and an
34382         operand defined in the loop we are peeling (and not some subloop).
34383         (propagate_constants_for_unrolling): Propagate all constants.
34385 2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
34387         PR target/79804
34388         * config/i386/i386.c (print_reg): Remove assert for disalowed
34389         regno values, call output_operand_lossage instead.
34391 2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
34393         PR target/78090
34394         * config/i386/constraints.md (Yc): New register constraint.
34395         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
34396         Use Yc constraint for alternative 2 of operand 0.  Remove
34397         preferred_for_speed attribute.
34399 2017-04-20  Alexander Monakov  <amonakov@ispras.ru>
34401         * omp-low.c (lower_lastprivate_clauses): Correct handling of linear and
34402         lastprivate clauses in SIMT case.
34404 2017-04-20  Volker Reichelt  <v.reichelt@netcologne.de>
34406         * doc/invoke.texi (-Wextra-semi): Document new warning option.
34408 2017-04-20  Richard Biener  <rguenther@suse.de>
34410         PR tree-optimization/57796
34411         * tree-vect-stmts.c (vect_model_store_cost): Cost scatters
34412         as N scalar stores.
34413         (vect_model_load_cost): Cost gathers as N scalar loads.
34415 2017-04-20  Richard Biener  <rguenther@suse.de>
34417         * ggc-page.c (ggc_allocated_p): Rename to ...
34418         (safe_lookup_page_table_entry): ... this and return the lookup
34419         result.
34420         (gt_ggc_m_S): Use safe_lookup_page_table_entry.
34422 2017-04-20  Richard Biener  <rguenther@suse.de>
34424         PR tree-optimization/80453
34425         * tree-ssa-sccvn.h (struct vn_phi_s): Add cclhs and ccrhs members.
34426         * tree-ssa-sccvn.c (cond_stmts_equal_p): Use recorded lhs and rhs
34427         from the conditions.
34428         (vn_phi_eq): Pass them down.
34429         (vn_phi_lookup): Record them.
34430         (vn_phi_insert): Likewise.
34432 2017-04-20  Matthew Fortune  <matthew.fortune@imgtec.com>
34434         * config/mips/mips.c (mips_expand_vec_perm_const): Re-fix
34435         uninitialized variable warning to avoid buffer overrun.
34437 2017-04-20  Alexander Monakov  <amonakov@ispras.ru>
34439         PR other/71250
34440         * doc/invoke.texi (-Wmissing-field-initializers): Mention that warning
34441         is suppressed for '{ 0 }' in C.
34443 2017-04-20  Jakub Jelinek  <jakub@redhat.com>
34445         * BASE-VER: Set to 8.0.0.
34447 2017-04-20  Thomas Preud'homme  <thomas.preudhomme@arm.com>
34449         * config/arm/arm.c (arm_elf_asm_cdtor): Create non-default
34450         priority .init_array and .fini_array section with SECTION_NOTYPE
34451         flag.
34453 2017-04-20  Jakub Jelinek  <jakub@redhat.com>
34455         PR middle-end/80423
34456         * tree.h (build_array_type): Add typeless_storage default argument.
34457         * tree.c (type_cache_hasher::equal): Also compare
34458         TYPE_TYPELESS_STORAGE flag for ARRAY_TYPEs.
34459         (build_array_type): Add typeless_storage argument, set
34460         TYPE_TYPELESS_STORAGE to it, if shared also hash it, and pass to
34461         recursive call.
34462         (build_nonshared_array_type): Adjust build_array_type_1 caller.
34463         (build_array_type): Likewise.  Add typeless_storage argument.
34465 2017-04-19  Eric Botcazou  <ebotcazou@adacore.com>
34466             Jakub Jelinek  <jakub@redhat.com>
34468         PR tree-optimization/80426
34469         * tree-vrp.c (extract_range_from_binary_expr_1): For an additive
34470         operation on symbolic operands, also compute the overflow for the
34471         invariant part when the operation degenerates into a negation.
34473 2017-04-19  Jakub Jelinek  <jakub@redhat.com>
34475         PR debug/80461
34476         * dwarf2out.c (modified_type_die, gen_type_die_with_usage):
34477         Check for t with zero TYPE_QUALS_NO_ADDR_SPACE.
34479         PR debug/80436
34480         * tree-ssa-loop-manip.c (find_uses_to_rename_def): Ignore debug uses.
34482 2017-04-19  Georg-Johann Lay  <avr@gjlay.de>
34484         PR target/80462
34485         * config/avr/avr.c (tree.h): Include it.
34486         (cgraph.h): Include it.
34487         (avr_encode_section_info): Don't warn for uninitialized progmem
34488         variable if it's just an alias.
34490 2017-04-19  Richard Biener  <rguenther@suse.de>
34492         PR ipa/65972
34493         * auto-profile.c (afdo_vpt_for_early_inline): Update SSA
34494         when needed by AutoPGO.
34496 2017-04-19  Paulo J. Matos  <paulo@matos-sorge.com>
34498         PR lto/50345
34499         * doc/lto.texi: Remove an extra 'that'.
34501 2017-04-19  Segher Boessenkool  <segher@kernel.crashing.org>
34503         PR rtl-optimization/80429
34504         * ira.c (split_live_ranges_for_shrink_wrap): Don't split regs that
34505         are only used in debug insns.
34507 2017-04-19  Eric Botcazou  <ebotcazou@adacore.com>
34508             Vladimir Makarov  <vmakarov@redhat.com>
34510         * config/sparc/predicates.md (input_operand): Add comment.  Return
34511         true for any memory operand when LRA is in progress.
34512         * config/sparc/sparc.c (sparc_expand_move): Minor formatting fix.
34514 2017-04-18  Jeff Law  <law@redhat.com>
34516         PR target/74563
34517         * mips.md ({return,simple_return}_internal): Do not overwrite
34518         operands[0].
34520 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34522         PR tree-optimization/80443
34523         * tree-vrp.c (intersect_ranges): For signed 1-bit precision type,
34524         instead of adding 1, subtract -1 and similarly instead of subtracting
34525         1 add -1.
34527 2017-04-18  Richard Sandiford  <richard.sandiford@arm.com>
34529         PR rtl-optimization/80357
34530         * haifa-sched.c (tmp_bitmap): New variable.
34531         (model_recompute): Handle duplicate use records.
34532         (alloc_global_sched_pressure_data): Initialize tmp_bitmap.
34533         (free_global_sched_pressure_data): Free it.
34535 2017-04-18  Bernd Edlinger  <bernd.edlinger@hotmail.de>
34537         Revert:
34538         2017-02-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
34539         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
34540         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
34541         instead of SYSTEM_HEADER_DIR.
34543 2017-04-18  Jeff Law  <law@redhat.com>
34545         PR middle-end/80422
34546         * cfgcleanup.c (try_crossjump_to_edge): Verify SRC1 and SRC2 have
34547         predecessors after walking up the insn chain.
34549 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34551         PR debug/80263
34552         * dwarf2out.c (modified_type_die): Try harder not to emit internal
34553         sizetype type into debug info.
34555 2017-04-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
34557         PR target/80099
34558         * config/rs6000/rs6000.c (rs6000_expand_vector_extract): Eliminate
34559         unneeded test for TARGET_UPPER_REGS_SF.
34560         * config/rs6000/vsx.md (vsx_extract_v4sf_var): Likewise.
34562 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34564         PR sanitizer/80444
34565         * sancov.c (sancov_pass): Use gsi_start_nondebug_after_labels_bb
34566         instead of gsi_after_labels.
34568 2017-04-18  Jeff Law  <law@redhat.com>
34570         * regcprop.c (maybe_mode_change): Avoid creating copies of the
34571         stack pointer.
34573         Revert:
34574         2017-04-13  Jeff Law  <law@redhat.com>
34575         * config/mips.mips.md (zero_extendsidi2): Do not allow SP to appear
34576         in operands[1] if it is a MEM and TARGET_MIPS16 is active.
34578 2017-04-18  Georg-Johann Lay  <avr@gjlay.de>
34580         PR target/79453
34581         * config/avr/avr.c (intl.h): Include it.
34582         (avr_pgm_check_var_decl) [reason]: Wrap diagnostic snippets into _().
34584 2017-04-18  Martin Liska  <mliska@suse.cz>
34586         PR gcov-profile/78783
34587         * gcov-tool.c (gcov_output_files): Validate that destination
34588         file is either removed by the tool or by a user.
34590 2017-04-14  Andrew Burgess  <andrew.burgess@embecosm.com>
34591             Guy Benyei  <guybe@mellanox.com>
34593         * config/arc/arc.c (arc_reorg): Move loop_end_id into a more local
34594         block, and do not negate it, the stored id is already negative.
34596 2017-04-14  Andrew Burgess  <andrew.burgess@embecosm.com>
34598         * config/arc/arc.md (doloop_begin_i): Use @pcl assembler syntax.
34600 2017-04-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
34602         PR target/80098
34603         * config/rs6000/rs6000-cpus.def (OTHER_P9_VECTOR_MASKS): Define
34604         masks of options that should be turned off if the VSX vector
34605         options are turned off.
34606         (OTHER_P8_VECTOR_MASKS): Likewise.
34607         (OTHER_VSX_VECTOR_MASKS): Likewise.
34608         * config/rs6000/rs6000.c (rs6000_option_override_internal): Call
34609         rs6000_disable_incompatible_switches to validate no type switches
34610         like -mvsx.
34611         (rs6000_incompatible_switch): New function to disallow turning on
34612         other vector options if -mno-vsx, -mno-power8-vector, or
34613         -mno-power9-vector are specified.
34615 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34617         * config/arc/arc.h (CRT_CALL_STATIC_FUNCTION): Use long calls.
34619 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34621         * config/arc/arc-protos.h (arc_decl_pretend_args): Remove.
34622         * config/arc/arc.c (arc_decl_pretend_args): Likewise.
34623         * config/arc/arc.h (CFA_FRAME_BASE_OFFSET): Likewise.
34624         (ARG_POINTER_CFA_OFFSET): Likewise.
34626 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34628         * config/arc/arc.c (arc_mode_dependent_address_p): Relax
34629         conditions to take advantage of various optimizations.
34631 2017-04-13  Jeff Law  <law@redhat.com>
34633         * config/mips.mips.md (zero_extendsidi2): Do not allow SP to appear
34634         in operands[1] if it is a MEM and TARGET_MIPS16 is active.
34635         (zero_extendsidi2_dext): Likewise.
34637 2017-04-13  Jakub Jelinek  <jakub@redhat.com>
34639         PR sanitizer/80403
34640         * fold-const.c (fold_ternary_loc): Revert
34641         use op0 instead of fold_convert_loc (loc, type, arg0) part of
34642         2017-04-12 change.
34644 2017-04-13  Vladimir Makarov  <vmakarov@redhat.com>
34646         PR rtl-optimization/80343
34647         * lra-remat.c (update_scratch_ops): Assign original hard reg to
34648         new scratch pseudo.
34650 2017-04-13  Denis Khalikov <d.khalikov@partner.samsung.com>
34652         PR sanitizer/80414
34653         * ubsan.c (ubsan_expand_bounds_ifn): Pass original index
34654         to ubsan_encode_value.
34656 2017-04-13  Jeff Law  <law@redhat.com>
34658         * reload1.c (eliminate_regs_1): Call gen_rtx_raw_SUBREG for SUBREGs
34659         appearing in DEBUG_INSNs.
34661 2017-04-13  Martin Liska  <mliska@suse.cz>
34663         PR gcov-profile/80413
34664         * gcov-io.c (gcov_write_string): Copy to buffer just when
34665         allocated size is greater than zero.
34667 2017-04-13  Jakub Jelinek  <jakub@redhat.com>
34669         PR debug/80321
34670         * dwarf2out.c (decls_for_scope): Ignore declarations of
34671         current_function_decl in BLOCK_NONLOCALIZED_VARS.
34673 2017-04-12  Jan Hubicka  <hubicka@ucw.cz>
34675         PR lto/69953
34676         * ipa-visibility.c (non_local_p): Fix typos.
34677         (localize_node): When localizing symbol in same comdat group,
34678         dissolve the group only when we know external symbols are going
34679         to be privatized.
34680         (function_and_variable_visibility): Do not localize DECL_EXTERNAL.
34682 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34684         PR tree-optimization/79390
34685         * optabs.c (emit_conditional_move): If the preferred op2/op3 operand
34686         order does not result in usable sequence, retry with reversed operand
34687         order.
34689         PR sanitizer/80403
34690         PR sanitizer/80404
34691         PR sanitizer/80405
34692         * fold-const.c (fold_ternary_loc): Use op1 instead of arg1 as argument
34693         to fold_build2_loc.  Convert TREE_OPERAND (tem, 0) to type.  Use
34694         op0 instead of fold_convert_loc (loc, type, arg0).
34696 2017-04-12  Jeff Law  <law@redhat.com>
34698         * genattrtab.c (write_eligible_delay): Verify DELAY_INSN still
34699         has a delay slot in the generated code.
34701         * config/cris/cris.md (cris_preferred_reload_class): Return
34702         GENNONACR_REGS rather than GENERAL_REGS.
34704 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34706         PR c/80163
34707         * expr.c <CASE_CONVERT>: For EXPAND_INITIALIZER determine SIGN_EXTEND
34708         vs. ZERO_EXTEND based on signedness of treeop0's type rather than
34709         signedness of the result type.
34711 2017-04-12  Richard Biener  <rguenther@suse.de>
34712             Jeff Law  <law@redhat.com>
34714         PR tree-optimization/80359
34715         * tree-ssa-dse.c (maybe_trim_partially_dead_store): Do not
34716         trim stores to TARGET_MEM_REFs.
34718 2017-04-12  Richard Biener  <rguenther@suse.de>
34720         PR tree-optimization/79390
34721         * gimple-ssa-split-paths.c (is_feasible_trace): Restrict
34722         threading case even more.
34724 2017-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
34726         PR target/80382
34727         * config/rs6000/sync.md (atomic_load<mode>, atomic_store<mode): Test
34728         for quad_address_p for TImode, instead of just not indexed_address.
34730 2017-04-12  Richard Biener  <rguenther@suse.de>
34731             Bernd Edlinger  <bernd.edlinger@hotmail.de>
34733         PR middle-end/79671
34734         * alias.c (component_uses_parent_alias_set_from): Handle
34735         TYPE_TYPELESS_STORAGE.
34736         (get_alias_set): Likewise.
34737         * tree-core.h (tree_type_common): Add typeless_storage flag.
34738         * tree.h (TYPE_TYPELESS_STORAGE): New macro.
34739         * stor-layout.c (place_union_field): Set TYPE_TYPELESS_STORAGE
34740         for types containing members with TYPE_TYPELESS_STORAGE.
34741         (place_field): Likewise.
34742         (layout_type): Likewise for ARRAY_TYPE.
34743         * lto-streamer-out.c (hash_tree): Hash TYPE_TYPELESS_STORAGE.
34744         * tree-streamer-in.c (unpack_ts_type_common_value_fields): Stream
34745         TYPE_TYPELESS_STORAGE.
34746         * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
34748 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34750         PR sanitizer/80349
34751         * fold-const.c (fold_binary_loc) <case BIT_IOR_EXPR>: Convert arg0's
34752         first argument to type.
34754 2017-04-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
34756         PR target/80376
34757         PR target/80315
34758         * config/rs6000/rs6000.c (rs6000_expand_unop_builtin): Return
34759         CONST0_RTX (mode) rather than const0_rtx where appropriate.
34760         (rs6000_expand_binop_builtin): Likewise.
34761         (rs6000_expand_ternop_builtin): Likewise; also add missing
34762         vsx_xxpermdi_* variants; also fix typo (arg1 => arg2) for
34763         vshasigma built-ins.
34764         * doc/extend.texi: Document that vec_xxpermdi's third argument
34765         must be a constant.
34767 2017-04-11  Uros Bizjak  <ubizjak@gmail.com>
34769         * config/i386/i386.c (dimode_scalar_chain::compute_convert_gain):
34770         Use shift_const cost parameter when calculating gain of STV shifts.
34772 2017-04-11  Vladimir Makarov  <vmakarov@redhat.com>
34774         PR rtl-optimization/70478
34775         * lra-constraints.c (process_alt_operands): Check memory for
34776         disfavoring memory insn operand.
34778 2017-04-11  Jakub Jelinek  <jakub@redhat.com>
34780         PR middle-end/80100
34781         * simplify-rtx.c (simplify_binary_operation_1) <case IOR>: Perform
34782         left shift in unsigned HOST_WIDE_INT type.
34784         PR rtl-optimization/80385
34785         * simplify-rtx.c (simplify_unary_operation_1): Don't transform
34786         (not (neg X)) into (plus X -1) for complex or non-integral modes.
34788         PR libgomp/80394
34789         * omp-low.c (scan_omp_task): Don't optimize away empty tasks
34790         if they have any depend clauses.
34792 2017-04-11  Martin Liska  <mliska@suse.cz>
34794         PR ipa/80212
34795         * cgraph.c (cgraph_node::dump): Dump calls_comdat_local.
34796         * ipa-split.c (split_function): Create a local comdat symbol
34797         if caller is in a comdat group.
34799 2017-04-11  Martin Liska  <mliska@suse.cz>
34801         PR ipa/80212
34802         * ipa-cp.c (determine_versionability): Handle calls_comdat_local
34803         flags.
34805 2017-04-11  Martin Sebor  <msebor@redhat.com>
34807         PR middle-end/80364
34808         * gimple-ssa-sprintf.c (get_int_range): Remove second argument and
34809         always use the int type.  Use INTEGRAL_TYPE_P() rather than testing
34810         for INTEGER_TYPE.
34811         (directive::set_width, directive::set_precision, format_character):
34812         Adjust.
34813         (parse_directive): Use INTEGRAL_TYPE_P() rather than testing for
34814         INTEGER_TYPE.
34816 2017-04-11  Richard Earnshaw  <rearnsha@arm.com>
34818         PR target/80389
34819         * config/arm/arm.c (arm_configure_build_target): When -mcpu and -arch
34820         conflict, set target->arch_name instead of target->cpu_name.
34822 2017-04-11  Richard Biener  <rguenther@suse.de>
34824         PR tree-optimization/80374
34825         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Use
34826         build_zero_cst, remove fold_convertible_p check again.
34828 2017-04-11  Martin Liska  <mliska@suse.cz>
34830         PR sanitizer/70878
34831         * ubsan.c (instrument_object_size): Do not instrument register
34832         variables.
34834 2017-04-11  Jakub Jelinek  <jakub@redhat.com>
34836         PR target/80381
34837         * config/i386/i386-builtin-types.def
34838         (V16HI_FTYPE_V16HI_INT_V16HI_UHI_COUNT,
34839         V16HI_FTYPE_V16HI_V8HI_V16HI_UHI_COUNT,
34840         V16SI_FTYPE_V16SI_INT_V16SI_UHI_COUNT,
34841         V16SI_FTYPE_V16SI_V4SI_V16SI_UHI_COUNT,
34842         V2DI_FTYPE_V2DI_INT_V2DI_UQI_COUNT,
34843         V2DI_FTYPE_V2DI_V2DI_V2DI_UQI_COUNT,
34844         V32HI_FTYPE_V32HI_INT_V32HI_USI_COUNT,
34845         V32HI_FTYPE_V32HI_V8HI_V32HI_USI_COUNT,
34846         V4DI_FTYPE_V4DI_INT_V4DI_UQI_COUNT,
34847         V4DI_FTYPE_V4DI_V2DI_V4DI_UQI_COUNT,
34848         V4SI_FTYPE_V4SI_INT_V4SI_UQI_COUNT,
34849         V4SI_FTYPE_V4SI_V4SI_V4SI_UQI_COUNT,
34850         V8DI_FTYPE_V8DI_INT_V8DI_UQI_COUNT,
34851         V8DI_FTYPE_V8DI_V2DI_V8DI_UQI_COUNT,
34852         V8HI_FTYPE_V8HI_INT_V8HI_UQI_COUNT,
34853         V8HI_FTYPE_V8HI_V8HI_V8HI_UQI_COUNT,
34854         V8SI_FTYPE_V8SI_INT_V8SI_UQI_COUNT,
34855         V8SI_FTYPE_V8SI_V4SI_V8SI_UQI_COUNT): New function type aliases.
34856         * config/i386/i386-builtin.def (__builtin_ia32_pslld512_mask,
34857         __builtin_ia32_pslldi512_mask, __builtin_ia32_psllq512_mask,
34858         __builtin_ia32_psllqi512_mask, __builtin_ia32_psrad512_mask,
34859         __builtin_ia32_psradi512_mask, __builtin_ia32_psraq512_mask,
34860         __builtin_ia32_psraqi512_mask, __builtin_ia32_psrld512_mask,
34861         __builtin_ia32_psrldi512_mask, __builtin_ia32_psrlq512_mask,
34862         __builtin_ia32_psrlqi512_mask, __builtin_ia32_psllwi128_mask,
34863         __builtin_ia32_pslldi128_mask, __builtin_ia32_psllqi128_mask,
34864         __builtin_ia32_psllw128_mask, __builtin_ia32_pslld128_mask,
34865         __builtin_ia32_psllq128_mask, __builtin_ia32_psllwi256_mask,
34866         __builtin_ia32_psllw256_mask, __builtin_ia32_pslldi256_mask,
34867         __builtin_ia32_pslld256_mask, __builtin_ia32_psllqi256_mask,
34868         __builtin_ia32_psllq256_mask, __builtin_ia32_psradi128_mask,
34869         __builtin_ia32_psrad128_mask, __builtin_ia32_psradi256_mask,
34870         __builtin_ia32_psrad256_mask, __builtin_ia32_psraqi128_mask,
34871         __builtin_ia32_psraq128_mask, __builtin_ia32_psraqi256_mask,
34872         __builtin_ia32_psraq256_mask, __builtin_ia32_psrldi128_mask,
34873         __builtin_ia32_psrld128_mask, __builtin_ia32_psrldi256_mask,
34874         __builtin_ia32_psrld256_mask, __builtin_ia32_psrlqi128_mask,
34875         __builtin_ia32_psrlq128_mask, __builtin_ia32_psrlqi256_mask,
34876         __builtin_ia32_psrlq256_mask, __builtin_ia32_psrawi256_mask,
34877         __builtin_ia32_psraw256_mask, __builtin_ia32_psrawi128_mask,
34878         __builtin_ia32_psraw128_mask, __builtin_ia32_psrlwi256_mask,
34879         __builtin_ia32_psrlw256_mask, __builtin_ia32_psrlwi128_mask,
34880         __builtin_ia32_psrlw128_mask, __builtin_ia32_psllwi512_mask,
34881         __builtin_ia32_psllw512_mask, __builtin_ia32_psrawi512_mask,
34882         __builtin_ia32_psraw512_mask, __builtin_ia32_psrlwi512_mask,
34883         __builtin_ia32_psrlw512_mask): Use _COUNT suffixed function type
34884         aliases.
34885         * config/i386/i386.c (ix86_expand_args_builtin): Rename last_arg_count
34886         flag to second_arg_count, handle 4 argument function type _COUNT
34887         aliases, handle second_arg_count on second argument rather than last.
34889 2017-04-10  Jeff Law  <law@redhat.com>
34891         PR tree-optimization/80374
34892         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Do not try to
34893         record anything if we can not convert integer_zero_node to the
34894         desired type.
34896 2017-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
34898         PR target/80108
34899         * config/rs6000/rs6000.c (rs6000_option_override_internal):
34900         Enhance special handling given to the TARGET_P9_MINMAX option in
34901         relation to certain other options.
34903 2017-04-10  Bin Cheng  <bin.cheng@arm.com>
34905         PR tree-optimization/80153
34906         * tree-ssa-loop-ivopts.c (add_iv_candidate_for_use): Check and
34907         remove POINTER_PLUS_EXPR's base part directly, rather than through
34908         aff_tree.
34910 2017-04-10  Richard Biener  <rguenther@suse.de>
34911             Bin Cheng  <bin.cheng@arm.com>
34913         PR tree-optimization/80153
34914         * tree-affine.c (aff_combination_to_tree): Get base pointer from
34915         the first element of pointer type aff_tree.  Build result expr in
34916         aff_tree's type.
34917         (add_elt_to_tree): Convert to type unconditionally.  Remove other
34918         fold_convert calls.
34919         * tree-ssa-loop-ivopts.c (alloc_iv): Pass in consistent types.
34920         (rewrite_use_nonlinear_expr): Check invariant using iv information.
34922 2017-04-10  Richard Biener  <rguenther@suse.de>
34924         * tree-ssa-structalias.c (find_func_aliases): Properly handle
34925         asm inputs.
34927 2017-04-10  Vladimir Makarov  <vmakarov@redhat.com>
34929         PR rtl-optimization/70478
34930         * lra-constraints.c (curr_small_class_check): New.
34931         (update_and_check_small_class_inputs): New.
34932         (process_alt_operands): Update curr_small_class_check.  Disfavor
34933         alternative insn memory operands.  Check available regs for small
34934         class operands.
34936 2017-03-31  Matthew Fortune  <matthew.fortune@imgtec.com>
34938         PR target/80057
34939         * config/mips/mips.opt (-mvirt): Update description.
34940         * doc/invoke.texi (-mvirt): Likewise.
34942 2017-04-10  Richard Biener  <rguenther@suse.de>
34944         PR middle-end/80362
34945         * fold-const.c (fold_binary_loc): Look at unstripped ops when
34946         looking for NEGATE_EXPR in -A / -B to A / B folding.
34948 2017-04-10  Martin Liska  <mliska@suse.cz>
34950         PR gcov-profile/80224
34951         * gcov.c (print_usage): Fix usage string.
34952         (get_gcov_intermediate_filename): Remove.
34953         (output_gcov_file): Use both for normal and intermediate format.
34954         (generate_results): Do not initialize special file for
34955         intermediate format.
34957 2017-04-10  Richard Biener  <rguenther@suse.de>
34959         PR tree-optimization/80304
34960         * tree-ssa-loop-im.c (ref_indep_loop_p_1): Also recurse
34961         for safelen.
34963 2017-04-10  Nathan Sidwell  <nathan@acm.org>
34965         PR target/79905
34966         * config/rs6000/rs6000.c (rs6000_vector_type): New.
34967         (rs6000_init_builtins): Use it.
34969 2016-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
34971         * config/arm/arm.md (<mrc>): Add mode to SET source.
34972         (<mrrc>): Likewise.
34974 2017-04-10  Richard Biener  <rguenther@suse.de>
34976         PR middle-end/80344
34977         * gimplify.c (is_gimple_mem_rhs_or_call): Allow CLOBBERs.
34979 2017-04-10  Jakub Jelinek  <jakub@redhat.com>
34981         PR target/80324
34982         * config/i386/avx512fintrin.h (_mm512_reduce_add_epi32,
34983         _mm512_reduce_mul_epi32, _mm512_reduce_and_epi32,
34984         _mm512_reduce_or_epi32, _mm512_mask_reduce_add_epi32,
34985         _mm512_mask_reduce_mul_epi32, _mm512_mask_reduce_and_epi32,
34986         _mm512_mask_reduce_or_epi32, _mm512_reduce_min_epi32,
34987         _mm512_reduce_max_epi32, _mm512_reduce_min_epu32,
34988         _mm512_reduce_max_epu32, _mm512_mask_reduce_min_epi32,
34989         _mm512_mask_reduce_max_epi32, _mm512_mask_reduce_min_epu32,
34990         _mm512_mask_reduce_max_epu32, _mm512_reduce_add_ps,
34991         _mm512_reduce_mul_ps, _mm512_mask_reduce_add_ps,
34992         _mm512_mask_reduce_mul_ps, _mm512_reduce_min_ps, _mm512_reduce_max_ps,
34993         _mm512_mask_reduce_min_ps, _mm512_mask_reduce_max_ps,
34994         _mm512_reduce_add_epi64, _mm512_reduce_mul_epi64,
34995         _mm512_reduce_and_epi64, _mm512_reduce_or_epi64,
34996         _mm512_mask_reduce_add_epi64, _mm512_mask_reduce_mul_epi64,
34997         _mm512_mask_reduce_and_epi64, _mm512_mask_reduce_or_epi64,
34998         _mm512_reduce_min_epi64, _mm512_reduce_max_epi64,
34999         _mm512_mask_reduce_min_epi64, _mm512_mask_reduce_max_epi64,
35000         _mm512_reduce_min_epu64, _mm512_reduce_max_epu64,
35001         _mm512_mask_reduce_min_epu64, _mm512_mask_reduce_max_epu64,
35002         _mm512_reduce_add_pd, _mm512_reduce_mul_pd, _mm512_mask_reduce_add_pd,
35003         _mm512_mask_reduce_mul_pd, _mm512_reduce_min_pd, _mm512_reduce_max_pd,
35004         _mm512_mask_reduce_min_pd, _mm512_mask_reduce_max_pd): New intrinsics.
35006 2017-04-08  Vladimir Makarov  <vmakarov@redhat.com>
35008         PR rtl-optimization/70478
35009         * lra-constraints.c: Reverse the last patch.
35011 2017-04-08  Andreas Tobler  <andreast@gcc.gnu.org>
35013         * config/aarch64/aarch64-freebsd.h: Define MCOUNT_NAME.
35014         Add comment for WCHAR_T.
35016 2017-04-08  Martin Liska  <mliska@suse.cz>
35018         Revert:
35019         2017-04-07  Martin Liska  <mliska@suse.cz>
35021         PR ipa/80212
35022         * ipa-split.c (split_function): Add function part to a same comdat
35023         group.
35025 2017-04-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
35027         PR target/80358
35028         * config/rs6000/rs6000.c (expand_block_compare): Fix boundary check.
35030 2017-04-07  Pat Haugen  <pthaugen@us.ibm.com>
35032         * rs6000/rs6000.c (vec_load_pendulum): Rename...
35033         (vec_pairing): ...to this.
35034         (power9_sched_reorder2): Rewrite code for pairing vector/vecload insns.
35035         (rs6000_sched_init): Adjust for name change.
35036         (struct rs6000_sched_context): Likewise.
35037         (rs6000_init_sched_context): Likewise.
35038         (rs6000_set_sched_context): Likewise.
35040 2017-04-07  Jakub Jelinek  <jakub@redhat.com>
35042         PR target/80322
35043         PR target/80323
35044         PR target/80325
35045         PR target/80326
35046         * config/i386/avxintrin.h (_mm256_cvtsd_f64, _mm256_cvtss_f32): New
35047         intrinsics.
35048         * config/i386/avx512fintrin.h (_mm512_int2mask, _mm512_mask2int,
35049         _mm512_abs_ps, _mm512_mask_abs_ps, _mm512_abs_pd, _mm512_mask_abs_pd,
35050         _mm512_cvtsd_f64, _mm512_cvtss_f32): Likewise.
35052 2017-04-07  Andreas Tobler  <andreast@gcc.gnu.org>
35054         * config/aarch64/aarch64-freebsd.h: Define WCHAR_TYPE.
35056 2017-04-07  Vladimir Makarov  <vmakarov@redhat.com>
35058         PR rtl-optimization/70703
35059         * ira-color.c (update_conflict_hard_regno_costs): Use
35060         int64_t instead of HOST_WIDE_INT.
35062 2017-04-07  Vladimir Makarov  <vmakarov@redhat.com>
35064         PR rtl-optimization/70478
35065         * lra-constraints.c (process_alt_operands): Disfavor alternative
35066         insn memory operands.
35068 2017-04-07  Jeff Law  <law@redhat.com>
35070         * config/iq2000/iq2000.c (final_prescan_insn): Do not separate a
35071         CALL and NOTE_INSN_CALL_ARG_LOCATION.
35073 2017-04-07  Martin Liska  <mliska@suse.cz>
35075         PR target/79889
35076         * config/aarch64/aarch64.c (aarch64_process_target_attr):
35077         Show error message instead of an ICE.
35079 2017-04-07  Martin Liska  <mliska@suse.cz>
35081         PR ipa/80212
35082         * ipa-split.c (split_function): Add function part to a same comdat
35083         group.
35085 2017-04-07  Richard Biener  <rguenther@suse.de>
35087         PR middle-end/80341
35088         * tree.c (get_unwidened): Also handle ! for_type case for
35089         INTEGER_CSTs.
35090         * convert.c (do_narrow): Split out from ...
35091         (convert_to_integer_1): ... here.  Do not pass final truncation
35092         type to get_unwidened for TRUNC_DIV_EXPR.
35094 2017-04-07  Richard Biener  <rguenther@suse.de>
35096         * tree-affine.c (wide_int_ext_for_comb): Take type rather
35097         than aff_tree.
35098         (aff_combination_const): Adjust.
35099         (aff_combination_scale): Likewise.
35100         (aff_combination_add_elt): Likewise.
35101         (aff_combination_add_cst): Likewise.
35102         (aff_combination_convert): Likewise.
35103         (add_elt_to_tree): Likewise.  Remove unused argument.
35104         (aff_combination_to_tree): Adjust calls to add_elt_to_tree.
35106 2017-04-07  Sebastian Huber  <sebastian.huber@embedded-brains.de>
35108         * config/arm/arm.h (ARM_DEFAULT_SHORT_ENUMS): Provide default
35109         definition.
35110         * config/arm/arm.c (arm_default_short_enums): Use
35111         ARM_DEFAULT_SHORT_ENUMS.
35112         * config/arm/rtems.h (ARM_DEFAULT_SHORT_ENUMS): Define.
35114 2017-04-06  Jakub Jelinek  <jakub@redhat.com>
35116         PR debug/80234
35117         * dwarf2out.c (gen_member_die): Handle C++17 inline static data
35118         members with redundant out-of-class redeclaration.
35120 2017-04-06  Uros Bizjak  <ubizjak@gmail.com>
35122         PR target/80286
35123         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): New pattern.
35124         * config/i386/i386.md (*zero_extendsidi2):
35125         Add (?*x,*x) and (?*v,*v) alternatives.
35127 2017-04-06  Uros Bizjak  <ubizjak@gmail.com>
35129         PR target/79733
35130         * config/i386/i386.c (ix86_expand_builtin)
35131         <case IX86_BUILTIN_K{,OR}TEST{C,Z}{8,16,32,64}>: Determine insn operand
35132         mode from insn data. Convert operands to insn operand mode.
35133         Copy operands that don't satisfy insn predicate to a register.
35135 2017-04-06  Sam Thursfield  <sam.thursfield@codethink.co.uk>
35137         * config/rs6000/x-aix: Increase memory limit for genautomata on AIX.
35138         Update comments.
35140 2017-04-06  Richard Biener  <rguenther@suse.de>
35142         PR tree-optimization/80334
35143         * tree-ssa-loop-ivopts.c (rewrite_use_address): Properly
35144         preserve alignment of accesses.
35146 2017-04-06  Richard Biener  <rguenther@suse.de>
35148         PR tree-optimization/80262
35149         * tree-sra.c (build_ref_for_offset): Preserve address-space
35150         information.
35151         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address):
35152         Drop useless address-space information on MEM_REF offsets.
35154 2017-04-05  Andreas Schwab  <schwab@linux-m68k.org>
35156         * builtins.def (BUILT_IN_UPDATE_SETJMP_BUF): Fix type.
35158 2017-04-05  Vladimir Makarov  <vmakarov@redhat.com>
35160         PR rtl-optimization/70703
35161         * ira-color.c (update_conflict_hard_regno_costs): Use
35162         HOST_WIDE_INT instead of long.
35164 2017-04-05  Uros Bizjak  <ubizjak@gmail.com>
35166         PR target/80298
35167         * config/i386/mmintrin.h: Add -msse target option when __SSE__ is
35168         not defined for x86_64 target.  Add -mmmx target option when __SSE2__
35169         is not defined.
35170         * config/i386/mm3dnow.h: Add -msse target when __SSE__ is not defined
35171         for x86_64 target.  Handle -m3dnowa option.
35173 2017-04-05  Vladimir Makarov  <vmakarov@redhat.com>
35175         PR rtl-optimization/70703
35176         * ira-color.c (update_costs_from_allocno): Use the smallest mode.
35177         (update_conflict_hard_regno_costs): Use long instead of unsigned
35178         arithmetic for cost calculation.
35180 2017-04-05  Jakub Jelinek  <jakub@redhat.com>
35181             Bernd Edlinger  <bernd.edlinger@hotmail.de>
35183         PR sanitizer/80308
35184         * asan.c (asan_store_shadow_bytes): Fix location of last_chunk_value
35185         for big endian.
35187 2017-04-05  Eric Botcazou  <ebotcazou@adacore.com>
35189         PR target/78002
35190         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Replace
35191         ptr_mode with Pmode throughout.
35192         * config/aarch64/aarch64.md (probe_stack_range_<PTR:mode): Rename
35193         into probe_stack_range and use DImode.
35195 2017-04-05  Dominik Vogt  <vogt@linux.vnet.ibm.com>
35197         PR target/79890
35198         * config/s390/s390.c (s390_register_info_gprtofpr): Return if
35199         call_eh_return is true.
35201 2017-04-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35203         * config/s390/s390-c.c (s390_resolve_overloaded_builtin):
35204         Initialize last_match_fntype_index.
35206 2017-04-05  Jakub Jelinek  <jakub@redhat.com>
35208         PR target/80310
35209         * tree-nvr.c: Include internal-fn.h.
35210         (pass_return_slot::execute): Ignore internal calls without
35211         direct optab.
35213 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
35214             Richard Biener  <rguenther@suse.de>
35216         PR c++/80297
35217         * genmatch.c (capture::gen_transform): For GENERIC unshare_expr
35218         captures used multiple times, except for the last use.
35219         * generic-match-head.c: Include gimplify.h.
35221 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
35223         PR tree-optimization/79390
35224         * target.h (struct noce_if_info): Declare.
35225         * targhooks.h (default_noce_conversion_profitable_p): Declare.
35226         * target.def (noce_conversion_profitable_p): New target hook.
35227         * ifcvt.h (struct noce_if_info): New type, moved from ...
35228         * ifcvt.c (struct noce_if_info): ... here.
35229         (noce_conversion_profitable_p): Renamed to ...
35230         (default_noce_conversion_profitable_p): ... this.  No longer
35231         static nor inline.
35232         (noce_try_store_flag_constants, noce_try_addcc,
35233         noce_try_store_flag_mask, noce_try_cmove, noce_try_cmove_arith,
35234         noce_convert_multiple_sets): Use targetm.noce_conversion_profitable_p
35235         instead of noce_conversion_profitable_p.
35236         * config/i386/i386.c: Include ifcvt.h.
35237         (ix86_option_override_internal): Don't override
35238         PARAM_MAX_RTL_IF_CONVERSION_INSNS default.
35239         (ix86_noce_conversion_profitable_p): New function.
35240         (TARGET_NOCE_CONVERSION_PROFITABLE_P): Redefine.
35241         * config/i386/x86-tune.def (X86_TUNE_ONE_IF_CONV_INSN): Adjust comment.
35242         * doc/tm.texi.in (TARGET_NOCE_CONVERSION_PROFITABLE_P): Add.
35243         * doc/tm.texi: Regenerated.
35245 2017-04-04  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35247         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Grammar
35248         correction.
35250 2017-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
35252         PR target/80307
35253         * config/arm/arm.c (thumb1_rtx_costs): Give a cost of 32
35254         instructions for small multiply cores.
35256 2017-04-04  Jeff Law  <law@redhat.com>
35258         * config/mips/mips.c (mips_multi_add): Zero initialize the newly
35259         added member.
35260         (mips_expand_vec_perm_const): Initialize elements in orig_perm
35261         that are not set by the loop over the elements.
35263 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
35265         PR target/80286
35266         * config/i386/i386.c (ix86_expand_args_builtin): If op has scalar
35267         int mode, convert_modes it to mode as unsigned, otherwise use
35268         lowpart_subreg to mode rather than SImode.
35269         * config/i386/sse.md (<mask_codefor>ashr<mode>3<mask_name>,
35270         ashr<mode>3, ashr<mode>3<mask_name>, <shift_insn><mode>3<mask_name>):
35271         Use DImode instead of SImode for the shift count operand.
35272         * config/i386/mmx.md (mmx_ashr<mode>3, mmx_<shift_insn><mode>3):
35273         Likewise.
35275 2017-04-04  Richard Biener  <rguenther@suse.de>
35277         PR middle-end/80281
35278         * match.pd (A + (-B) -> A - B): Make sure to preserve unsigned
35279         arithmetic done for the negate or the plus.  Simplify.
35280         (A - (-B) -> A + B): Likewise.
35281         * fold-const.c (split_tree): Make sure to not negate pointers.
35283 2017-04-04  Segher Boessenkool  <segher@kernel.crashing.org>
35285         PR rtl-optimization/60818
35286         * simplify-rtx.c (simplify_binary_operation_1): Do not replace
35287         a compare of comparisons with the thing compared if this results
35288         in a different machine mode.
35290 2017-04-03  Jonathan Wakely  <jwakely@redhat.com>
35292         * alias.c (base_alias_check): Fix typo in comment.
35293         * cgraph.h (class ipa_polymorphic_call_context): Likewise.
35294         * cgraphunit.c (symbol_table::compile): Likewise.
35295         * collect2.c (maybe_run_lto_and_relink): Likewise.
35296         * config/arm/arm.c (arm_thumb1_mi_thunk): Likewise.
35297         * config/avr/avr-arch.h (avr_arch_info_t): Likewise.
35298         * config/avr/avr.c (avr_map_op_t): Likewise.
35299         * config/cr16/cr16.h (DATA_ALIGNMENT): Likewise.
35300         * config/epiphany/epiphany.c (TARGET_ARG_PARTIAL_BYTES): Likewise.
35301         * config/epiphany/epiphany.md (movcc): Likewise.
35302         * config/i386/i386.c (legitimize_pe_coff_extern_decl): Likewise.
35303         * config/m68k/m68k.c (struct _sched_ib, m68k_sched_variable_issue):
35304         Likewise.
35305         * config/mips/mips.c (mips_save_restore_reg): Likewise.
35306         * config/rx/rx.c (rx_is_restricted_memory_address): Likewise.
35307         * config/s390/s390.c (Z10_EARLYLOAD_DISTANCE): Likewise.
35308         * config/sh/sh.c (sh_rtx_costs): Likewise.
35309         * fold-const.c (fold_truth_andor): Likewise.
35310         * genautomata.c (collapse_flag): Likewise.
35311         * gengtype.h (struct type::u::s): Likewise.
35312         * gensupport.c (has_subst_attribute, add_mnemonic_string): Likewise.
35313         * input.c (FORMAT_AMOUNT): Likewise.
35314         * ipa-cp.c (class ipcp_lattice, agg_replacements_to_vector)
35315         (known_aggs_to_agg_replacement_list): Likewise.
35316         * ipa-inline-analysis.c: Likewise.
35317         * ipa-inline.h (estimate_edge_time, estimate_edge_hints): Likewise.
35318         * ipa-polymorphic-call.c
35319         (ipa_polymorphic_call_context::restrict_to_inner_class): Likewise.
35320         * loop-unroll.c (analyze_insn_to_expand_var): Likewise.
35321         * lra.c (lra_optional_reload_pseudos, lra_subreg_reload_pseudos):
35322         Likewise.
35323         * modulo-sched.c (apply_reg_moves): Likewise.
35324         * omp-expand.c (build_omp_regions_1): Likewise.
35325         * trans-mem.c (struct tm_wrapper_hasher): Likewise.
35326         * tree-ssa-loop-ivopts.c (may_eliminate_iv): Likewise.
35327         * tree-ssa-loop-niter.c (maybe_lower_iteration_bound): Likewise.
35328         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
35329         * value-prof.c: Likewise.
35330         * var-tracking.c (val_reset): Likewise.
35332 2017-04-03  Richard Biener  <rguenther@suse.de>
35334         PR tree-optimization/80275
35335         * fold-const.c (split_address_to_core_and_offset): Handle
35336         POINTER_PLUS_EXPR.
35338 2017-04-03  Eric Botcazou  <ebotcazou@adacore.com>
35340         * tree-nested.c (get_descriptor_type): Make sure that the alignment of
35341         descriptors is at least equal to that of functions.
35343 2017-04-02  Uros Bizjak  <ubizjak@gmail.com>
35345         * config/i386/sse.md (movdi_to_sse): Add missing DONE.
35347 2017-04-02  Uros Bizjak  <ubizjak@gmail.com>
35349         PR target/80250
35350         * config/i386/sse.md (mov<IMOD4:mode>): Remove insn pattern.
35351         (mov<IMOD4:mode>): New expander.
35352         (*mov<IMOD4:mode>_internal): New insn and split pattern.
35354 2017-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
35356         PR rtl-optimization/79405
35357         * fwprop.c (propagations_left): New variable.
35358         (forward_propagate_into): Decrement it.
35359         (fwprop_init): Initialize it.
35360         (fw_prop): If the variable has reached zero, stop propagating.
35361         (fwprop_addr): Ditto.
35363 2017-03-31  Jakub Jelinek  <jakub@redhat.com>
35365         PR debug/79255
35366         * dwarf2out.c (decls_for_scope): If BLOCK_NONLOCALIZED_VAR is
35367         a FUNCTION_DECL, pass it as decl instead of origin to
35368         process_scope_var.
35370 2017-03-31  Alexander Monakov  <amonakov@ispras.ru>
35372         * config/nvptx/nvptx.c (nvptx_output_softstack_switch): Correct format
35373         string.
35375 2017-03-31  Pat Haugen  <pthaugen@us.ibm.com>
35377         PR target/80107
35378         * config/rs6000/rs6000.md (extendhi<mode>2): Add test for
35379         TARGET_VSX_SMALL_INTEGER.
35381 2017-03-31  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35383         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
35384         reference to the OpenPOWER 64-Bit ELF V2 ABI Specification.
35386 2017-03-31  Matthew Fortune  <matthew.fortune@imgtec.com>
35388         * config/mips/mips-msa.md (msa_vec_extract_<msafmt_f>): Update
35389         extraction from odd-numbered MSA register.
35391 2017-03-31  Jakub Jelinek  <jakub@redhat.com>
35393         PR middle-end/80173
35394         * expmed.c (store_bit_field_1): Don't attempt to create
35395         a word subreg out of hard registers wider than word if they
35396         have HARD_REGNO_NREGS of 1 for their mode.
35398         PR middle-end/80163
35399         * varasm.c (initializer_constant_valid_p_1): Disallow sign-extending
35400         conversions to integer types wider than word and pointer.
35402         PR debug/80025
35403         * cselib.h (rtx_equal_for_cselib_1): Add depth argument.
35404         (rtx_equal_for_cselib_p): Pass 0 to it.
35405         * cselib.c (cselib_hasher::equal): Likewise.
35406         (rtx_equal_for_cselib_1): Add depth argument.  If depth
35407         is 128, don't look up VALUE locs and punt.  Increment
35408         depth in recursive calls when walking VALUE locs.
35410 2017-03-31  Bernd Edlinger  <bernd.edlinger@hotmail.de>
35412         * gcov.c (md5sum_to_hex): Fix output of MD5 hex bytes.
35413         (make_gcov_file_name): Use the canonical path name for generating
35414         the MD5 value.
35415         (read_line): Fix handling of files with ascii null bytes.
35417 2017-03-30  Matthew Fortune  <matthew.fortune@imgtec.com>
35419         * config/mips/mips.c (mips_expand_vector_init): Create a const_vector
35420         to initialise a vector register instead
35421         of using a const_int.
35423 2017-03-30  Jakub Jelinek  <jakub@redhat.com>
35425         PR translation/80189
35426         * gimplify.c (omp_default_clause): Use %qs instead of %s in
35427         diagnostic messages.
35429 2017-03-30  Peter Bergner  <bergner@vnet.ibm.com>
35431         PR target/80246
35432         * config/rs6000/dfp.md (dfp_dxex_<mode>): Update mode of operand 0.
35433         (dfp_diex_<mode>): Update mode of operand 1.
35434         * doc/extend.texi (dxex, dxexq): Document change to return type.
35435         (diex, diexq): Document change to argument type.
35437 2017-03-30  Martin Jambor  <mjambor@suse.cz>
35439         PR ipa/77333
35440         * cgraph.h (cgraph_build_function_type_skip_args): Declare.
35441         * cgraph.c (redirect_call_stmt_to_callee): Set gimple fntype so that
35442         it reflects the signature changes performed at the callee side.
35443         * cgraphclones.c (build_function_type_skip_args): Make public, renamed
35444         to cgraph_build_function_type_skip_args.
35445         (build_function_decl_skip_args): Adjust call to the above function.
35447 2017-03-30  Jakub Jelinek  <jakub@redhat.com>
35449         PR target/80206
35450         * config/i386/sse.md
35451         (<extract_type>_vextract<shuffletype><extract_suf>_mask): Use
35452         register as dest whenever it is a MEM not rtx_equal_p to the
35453         corresponding dup operand, and when forcing into reg move the
35454         reg into the memory afterwards.
35455         (<extract_type_2>_vextract<shuffletype><extract_suf_2>_mask):
35456         Likewise.  Use <ssehalfvecmode> instead of <ssequartermode>
35457         for the force_reg mode.
35458         (avx512vl_vextractf128<mode>): Use register as dest either
35459         always when a MEM, or when it is a MEM not rtx_equal_p to the
35460         corresponding dup operand, or even not when it is a CONST_VECTOR
35461         depending on the mode and lo vs. hi.
35462         (avx512dq_vextract<shuffletype>64x2_1_maskm): Remove extraneous
35463         parens.
35464         (avx512f_vextract<shuffletype>32x4_1_maskm): Likewise.
35465         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name>):
35466         Likewise.  Require that operands[2] is even.
35467         (<mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name>):
35468         Remove extraneous parens.  Require that operands[2] is a multiple
35469         of 4.
35470         (vec_extract_lo_<mode><mask_name>): Don't bother testing if
35471         operands[0] is a MEM if <mask_applied>, the predicates/constraints
35472         disallow memory then.
35474 2017-03-30  Richard Biener  <rguenther@suse.de>
35476         PR tree-optimization/77498
35477         * tree-ssa-pre.c (phi_translate_1): Do not allow simplifications
35478         to non-constants over backedges.
35480 2017-03-29  Segher Boessenkool  <segher@kernel.crashing.org>
35482         PR rtl-optimization/80233
35483         * combine.c (combine_instructions): Only take NONDEBUG_INSN_P insns
35484         as last_combined_insn.  Do not test for BARRIER_P separately.
35486 2017-03-29  Andreas Schwab  <schwab@suse.de>
35488         PR ada/80146
35489         * calls.c (prepare_call_address): Convert funexp to Pmode before
35490         copying to temp reg.
35492 2017-03-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35494         PR tree-optimization/80158
35495         * gimple-ssa-strength-reduction.c (replace_mult_candidate):
35496         Handle possible future case of more than one alternate
35497         interpretation.
35498         (replace_rhs_if_not_dup): Likewise.
35499         (replace_one_candidate): Likewise.
35501 2017-03-28  Vladimir Makarov  <vmakarov@redhat.com>
35503         PR rtl-optimization/80193
35504         * ira.c (ira): Do not check allocation for LRA.
35506 2017-03-28  Alexander Monakov  <amonakov@ispras.ru>
35508         * config/nvptx/nvptx-protos.h (nvptx_output_simt_enter): Declare.
35509         (nvptx_output_simt_exit): Declare.
35510         * config/nvptx/nvptx.c (nvptx_init_unisimt_predicate): Use
35511         cfun->machine->unisimt_location.  Handle NULL unisimt_predicate.
35512         (init_softstack_frame): Move initialization of crtl->is_leaf to...
35513         (nvptx_declare_function_name): ...here.  Emit declaration of local
35514         memory space buffer for omp_simt_enter insn.
35515         (nvptx_output_unisimt_switch): New.
35516         (nvptx_output_softstack_switch): New.
35517         (nvptx_output_simt_enter): New.
35518         (nvptx_output_simt_exit): New.
35519         * config/nvptx/nvptx.h (struct machine_function): New fields
35520         has_simtreg, unisimt_location, simt_stack_size, simt_stack_align.
35521         * config/nvptx/nvptx.md (UNSPECV_SIMT_ENTER): New unspec.
35522         (UNSPECV_SIMT_EXIT): Ditto.
35523         (omp_simt_enter_insn): New insn.
35524         (omp_simt_enter): New expansion.
35525         (omp_simt_exit): New insn.
35526         * config/nvptx/nvptx.opt (msoft-stack-reserve-local): New option.
35528         * internal-fn.c (expand_GOMP_SIMT_ENTER): New.
35529         (expand_GOMP_SIMT_ENTER_ALLOC): New.
35530         (expand_GOMP_SIMT_EXIT): New.
35531         * internal-fn.def (GOMP_SIMT_ENTER): New internal function.
35532         (GOMP_SIMT_ENTER_ALLOC): Ditto.
35533         (GOMP_SIMT_EXIT): Ditto.
35534         * target-insns.def (omp_simt_enter): New insn.
35535         (omp_simt_exit): Ditto.
35536         * omp-low.c (struct omplow_simd_context): New fields simt_eargs,
35537         simt_dlist.
35538         (lower_rec_simd_input_clauses): Implement SIMT privatization.
35539         (lower_rec_input_clauses): Likewise.
35540         (lower_lastprivate_clauses): Handle SIMT privatization.
35542         * omp-offload.c: Include langhooks.h, tree-nested.h, stor-layout.h.
35543         (ompdevlow_adjust_simt_enter): New.
35544         (find_simtpriv_var_op): New.
35545         (execute_omp_device_lower): Handle IFN_GOMP_SIMT_ENTER,
35546         IFN_GOMP_SIMT_ENTER_ALLOC, IFN_GOMP_SIMT_EXIT.
35548         * tree-inline.h (struct copy_body_data): New field dst_simt_vars.
35549         * tree-inline.c (expand_call_inline): Handle SIMT privatization.
35550         (copy_decl_for_dup_finish): Ditto.
35552         * tree-ssa.c (execute_update_addresses_taken): Handle GOMP_SIMT_ENTER.
35554 2017-03-28  Uros Bizjak  <ubizjak@gmail.com>
35556         PR target/53383
35557         * config/i386/i386.c (ix86_option_override_internal): Always
35558         allow -mpreferred-stack-boundary=3 for 64-bit targets.
35560 2017-03-28  Bin Cheng  <bin.cheng@arm.com>
35562         * tree-vect-loop.c (optimize_mask_stores): Add bb to the right loop.
35564 2017-03-28  Bin Cheng  <bin.cheng@arm.com>
35566         * tree-vect-loop-manip.c (slpeel_add_loop_guard): New param and
35567         mark new edge's irreducible flag accordign to it.
35568         (vect_do_peeling): Check loop preheader edge's irreducible flag
35569         and pass it to function slpeel_add_loop_guard.
35571 2017-03-28  Richard Sandiford  <richard.sandiford@arm.com>
35573         PR tree-optimization/80218
35574         * tree-call-cdce.c (shrink_wrap_one_built_in_call_with_conds):
35575         Update block frequencies and counts.
35577 2017-03-28  Richard Biener  <rguenther@suse.de>
35579         PR tree-optimization/78644
35580         * tree-ssa-ccp.c (evaluate_stmt): When we may not use the value
35581         of a simplification result we may not use it at all.
35583 2017-03-28  Richard Biener  <rguenther@suse.de>
35585         PR ipa/80205
35586         * tree-inline.c (copy_phis_for_bb): Do not create PHI node
35587         without arguments, generate default definition of a SSA name.
35589 2017-03-28  Richard Biener  <rguenther@suse.de>
35591         PR middle-end/80222
35592         * gimple-fold.c (gimple_fold_indirect_ref): Do not touch
35593         TYPE_REF_CAN_ALIAS_ALL references.
35594         * fold-const.c (fold_indirect_ref_1): Likewise.
35596 2017-03-28  Martin Liska  <mliska@suse.cz>
35598         PR ipa/80104
35599         * cgraphunit.c (cgraph_node::expand_thunk): Mark argument of a
35600         thunk call as DECL_GIMPLE_REG_P when vector or complex type.
35602 2017-03-28  Claudiu Zissulescu  <claziss@synopsys.com>
35603             Thomas Petazzoni  <thomas.petazzoni@free-electrons.com>
35605         * config/arc/arc.h (CPP_SPEC): Add subtarget_cpp_spec.
35606         (EXTRA_SPECS): Define.
35607         (SUBTARGET_EXTRA_SPECS): Likewise.
35608         (SUBTARGET_CPP_SPEC): Likewise.
35609         * config/arc/elf.h (EXTRA_SPECS): Renamed to
35610         SUBTARGET_EXTRA_SPECS.
35611         * config/arc/linux.h (SUBTARGET_CPP_SPEC): Define.
35613 2017-03-28  Claudiu Zissulescu  <claziss@synopsys.com>
35615         * config/arc/simdext.md (vst64_insn): Update pattern.
35616         (vld32wh_insn): Likewise.
35617         (vld32wl_insn): Likewise.
35618         (vld64_insn): Likewise.
35619         (vld32_insn): Likewise.
35621 2017-03-28  Marek Polacek  <polacek@redhat.com>
35623         PR sanitizer/80067
35624         * fold-const.c (fold_comparison): Use protected_set_expr_location
35625         instead of SET_EXPR_LOCATION.
35627 2017-03-28  Markus Trippelsdorf  <markus@trippelsdorf.de>
35629         * tree.c (add_expr): Avoid name lookup warning.
35631 2017-03-27  Jeff Law  <law@redhat.com>
35633         PR tree-optimization/80216
35634         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Fix typo in
35635         function name.  Limit recursion depth.
35636         (record_temporary_equivalences): Corresponding changes.
35638 2017-03-27  Jonathan Wakely  <jwakely@redhat.com>
35640         * doc/invoke.texi (-Wno-narrowing): Reorder so default behavior is
35641         covered first.
35643 2017-03-27  Jakub Jelinek  <jakub@redhat.com>
35645         PR target/80102
35646         * reg-notes.def (REG_CFA_NOTE): Define.  Use it for CFA related
35647         notes.
35648         * cfgcleanup.c (reg_note_cfa_p): New array.
35649         (insns_have_identical_cfa_notes): New function.
35650         (old_insns_match_p): Don't cross-jump in between /f
35651         and non-/f instructions.  If both i1 and i2 are frame related,
35652         verify all CFA notes, their order and content.
35654 2017-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
35656         PR target/78543
35657         * config/rs6000/rs6000.md (bswaphi2_extenddi): Combine bswap
35658         HImode and SImode with zero extend to DImode to one insn.
35659         (bswap<mode>2_extenddi): Likewise.
35660         (bswapsi2_extenddi): Likewise.
35661         (bswaphi2_extendsi): Likewise.
35662         (bswaphi2): Combine bswap HImode and SImode into one insn.
35663         Separate memory insns from swapping register.
35664         (bswapsi2): Likewise.
35665         (bswap<mode>2): Likewise.
35666         (bswaphi2_internal): Delete, no longer used.
35667         (bswapsi2_internal): Likewise.
35668         (bswap<mode>2_load): Split bswap HImode/SImode into separate load,
35669         store, and gpr<-gpr swap insns.
35670         (bswap<mode>2_store): Likewise.
35671         (bswaphi2_reg): Register only splitter, combine with the splitter.
35672         (bswaphi2 splitter): Likewise.
35673         (bswapsi2_reg): Likewise.
35674         (bswapsi2 splitter): Likewise.
35675         (bswapdi2): If we have the LDBRX and STDBRX instructions, split
35676         the insns into load, store, and register/register insns.
35677         (bswapdi2_ldbrx): Likewise.
35678         (bswapdi2_load): Likewise.
35679         (bswapdi2_store): Likewise.
35680         (bswapdi2_reg): Likewise.
35682 2017-03-27  Gunther Nikl  <gnikl@users.sourceforge.net>
35684         * system.h (HAVE_DESIGNATED_INITIALIZERS): Fix non C++ case.
35685         (HAVE_DESIGNATED_UNION_INITIALIZERS): Likewise.
35687 2017-03-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
35689         PR target/80103
35690         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Edit and
35691         add comments.
35692         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
35693         special handling for target option conflicts between dform
35694         options (-mpower9-dform, -mpower9-dform-vector,
35695         -mpower9-dform-scalar) and -mno-direct-move.
35697 2017-03-27  Richard Biener  <rguenther@suse.de>
35699         PR tree-optimization/80181
35700         * tree-ssa-ccp.c (likely_value): UNDEFINED ^ X is UNDEFINED.
35702 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35704         * config/arc/predicates.md (move_double_src_operand): Replace the
35705         call to move_double_src_operand with a call to address_operand.
35707 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35709         * config/arc/elf.h (ARGET_ARC_TP_REGNO_DEFAULT): Define.
35710         * config/arc/linux.h (ARGET_ARC_TP_REGNO_DEFAULT): Likewise.
35711         * config/arc/arc.opt (mtp-regno): Use ARGET_ARC_TP_REGNO_DEFAULT.
35713 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35715         * config/arc/predicates.md (long_immediate_loadstore_operand):
35716         Consider scaled addresses cases.
35718 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35720         * config/arc/arc.c (arc_epilogue_uses): BLINK should be also
35721         restored when in interrupt.
35722         * config/arc/arc.md (simple_return): ARCv2 rtie instruction
35723         doesn't have delay slot.
35725 2017-03-27  Richard Biener  <rguenther@suse.de>
35727         PR ipa/79776
35728         * tree-ssa-structalias.c (associate_varinfo_to_alias): Skip
35729         inlined thunk clones.
35731 2017-03-27  Jakub Jelinek  <jakub@redhat.com>
35733         PR sanitizer/80168
35734         * asan.c (instrument_derefs): Copy over last operand from
35735         original COMPONENT_REF to the new COMPONENT_REF with
35736         DECL_BIT_FIELD_REPRESENTATIVE.
35737         * ubsan.c (instrument_object_size): Likewise.
35739 2017-03-27  Richard Biener  <rguenther@suse.de>
35741         PR tree-optimization/80170
35742         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Make
35743         sure DR/SCEV didnt fold in constants we do not see when looking
35744         at the reference base alignment.
35746 2017-03-27  Richard Biener  <rguenther@suse.de>
35748         PR middle-end/80171
35749         * gimple-fold.c (fold_ctor_reference): Properly guard against
35750         NULL return value from canonicalize_constructor_val.
35752 2017-03-25  Uros Bizjak  <ubizjak@gmail.com>
35754         PR target/80180
35755         * config/i386/i386.c (ix86_expand_builtin)
35756         <IX86_BUILTIN_RDSEED{16,32,64}_STEP>: Do not expand arg0 between
35757         flags reg setting and flags reg using instructions.
35758         <IX86_BUILTIN_RDRAND{16,32,64}_STEP>: Ditto.  Use non-flags reg
35759         clobbering instructions to zero extend op2.
35761 2017-03-25  Gerald Pfeifer  <gerald@pfeifer.com>
35763         * doc/install.texi (Configuration) <--with-aix-soname>:
35764         Update link to AIX ld.
35766 2017-03-25  Bernd Schmidt  <bschmidt@redhat.com>
35768         PR rtl-optimization/80160
35769         PR rtl-optimization/80159
35770         * lra-assigns.c (must_not_spill_p): Tighten new test to also take
35771         reg_alternate_class into account.
35773 2017-03-24  Vladimir Makarov  <vmakarov@redhat.com>
35775         PR target/80148
35776         * lra-assigns.c (assign_by_spills): Add spilled non-reload pseudos
35777         to consider in curr_insn_transform.
35779 2017-03-24  Jakub Jelinek  <jakub@redhat.com>
35781         * genrecog.c (validate_pattern): Add VEC_SELECT validation.
35782         * genmodes.c (emit_min_insn_modes_c): Call emit_mode_nunits
35783         and emit_mode_inner.
35785 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35787         * config/s390/s390-builtins.def: Add VXE builtins.  Add a flags
35788         argument to the overloaded builtin variants.  Use the new flag to
35789         deprecate certain builtin variants.
35790         * config/s390/s390-builtin-types.def: Add new builtin types.
35791         * config/s390/s390-builtins.h: Support new flags field for
35792         overloaded builtins.
35793         * config/s390/s390-c.c (OB_DEF_VAR): New flags field.
35794         (s390_macro_to_expand): Enable vector float data type.
35795         (s390_cpu_cpp_builtins_internal): Indicate support of the new
35796         builtins by incrementing the __VEC__ version number.
35797         (s390_expand_overloaded_builtin): Support expansion of vec_xl and
35798         vec_xst.
35799         (s390_resolve_overloaded_builtin): Emit error messages depending
35800         on the builtin flags.
35801         * config/s390/s390.c (s390_expand_builtin): Support additional
35802         flags argument.  Change error message to match the messages
35803         emitted in s390-c.c.
35804         * config/s390/s390.md: New UNSPEC_* constants.
35805         (op_type): Add new instruction types.
35806         * config/s390/vecintrin.h: Add new builtins and test data class
35807         constants.
35808         * config/s390/vx-builtins.md (V_HW_32_64): Add V4SF.
35809         (V_HW_4, VEC_HW, VECF_HW): New mode iterators.
35810         (VEC_INEXACT, VEC_NOINEXACT): New constants.
35811         ("vec_splats<mode>", "vec_insert<mode>", "vec_promote<mode>")
35812         ("vec_insert_and_zero<mode>", "vec_mergeh<mode>")
35813         ("vec_mergel<mode>"): V_HW -> VEC_HW.
35815         ("vlrlrv16qi", "vstrlrv16qi", "vbpermv16qi", "vec_msumv2di")
35816         ("vmslg", "*vftci<mode>_cconly", "vftci<mode>_intcconly")
35817         ("*vftci<mode>", "vftci<mode>_intcc", "vec_double_s64")
35818         ("vec_double_u64", "vfmin<mode>", "vfmax<mode>"): New definition.
35820         ("and_av2df3", "and_cv2df3", "vec_andc_av2df3")
35821         ("vec_andc_cv2df3", "xor_av2df3", "xor_cv2df3", "vec_nor_av2df3")
35822         ("vec_nor_cv2df3", "ior_av2df3", "ior_cv2df3", "vec_nabs")
35823         ("*vftcidb", "*vftcidb_cconly", "vftcidb"): Remove definition.
35825         ("vec_all_<fpcmpcc:code>v2df", "vec_any_<fpcmpcc:code>v2df")
35826         ("vec_scatter_elementv4si_DI", "vec_cmp<fpcmp:code>v2df")
35827         ("vec_di_to_df_s64", "vec_di_to_df_u64", "vec_df_to_di_u64")
35828         ("vfidb", "*vldeb", "*vledb", "*vec_cmp<insn_cmp>v2df_cconly")
35829         ("vec_cmpeqv2df_cc", "vec_cmpeqv2df_cc", "vec_cmphv2df_cc")
35830         ("vec_cmphev2df_cc", "*vec_cmpeqv2df_cc")
35831         ("*vec_cmphv2df_cc", "*vec_cmphev2df_cc"): Enable new modes as ...
35833         ("vec_all_<fpcmpcc:code><mode>", "vec_any_<fpcmpcc:code><mode>")
35834         ("vec_scatter_element<V_HW_4:mode>_DI")
35835         ("vec_cmp<fpcmp:code><mode>", "vcdgb", "vcdlgb", "vclgdb")
35836         ("vec_fpint<mode>", "vflls")
35837         ("vflrd", "*vec_cmp<insn_cmp><mode>_cconly", "vec_cmpeq<mode>_cc")
35838         ("vec_cmpeq<mode>_cc", "vec_cmph<mode>_cc", "vec_cmphe<mode>_cc")
35839         ("*vec_cmpeq<mode>_cc", "*vec_cmph<mode>_cc")
35840         ("*vec_cmphe<mode>_cc"): ... these.
35842         ("vec_ctd_s64", "vec_ctsl", "vec_ctul", "vec_st2f"): Use rounding
35843         mode constant instead of magic value.
35845 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35847         * config/s390/s390.c (s390_expand_vec_compare): Support other
35848         vector floating point modes than just V2DF.
35849         (s390_expand_vcond): Likewise.
35850         (s390_hard_regno_mode_ok): Allow SFmode values in VRs.
35851         (s390_cannot_change_mode_class): Prevent mode changes between TF
35852         and V1TF in vector registers.
35853         * config/s390/s390.md (DF, SF): New mode attributes.
35854         ("*cmp<mode>_ccs", "add<mode>3", "sub<mode>3", "mul<mode>3")
35855         ("fma<mode>4", "fms<mode>4", "div<mode>3", "*neg<mode>2"): Add
35856         SFmode support for VRs.
35857         * config/s390/vector.md (V_HW, V_HW2, VT_HW, ti*, nonvec): Add new
35858         vector fp modes.
35859         (VFT, VF_HW): New mode iterators.
35860         (vw, sdx): New mode attributes.
35861         ("addv2df3", "subv2df3", "mulv2df3", "divv2df3", "sqrtv2df2")
35862         ("fmav2df4","fmsv2df4", "negv2df2", "absv2df2", "*negabsv2df2")
35863         ("smaxv2df3", "sminv2df3", "*vec_cmp<VFCMP_HW_OP:code>v2df_nocc")
35864         ("vec_cmpuneqv2df", "vec_cmpltgtv2df", "vec_orderedv2df")
35865         ("vec_unorderedv2df"): Adjust the v2df only patterns to support
35866         also the new vector floating point modes.  Renaming to ...
35868         ("add<mode>3", "sub<mode>3", "mul<mode>3", "div<mode>3")
35869         ("sqrt<mode>2", "fma<mode>4", "fms<mode>4", "neg<mode>2")
35870         ("abs<mode>2", "negabs<mode>2", "smax<mode>3")
35871         ("smin<mode>3", "*vec_cmp<VFCMP_HW_OP:code><mode>_nocc")
35872         ("vec_cmpuneq<mode>", "vec_cmpltgt<mode>", "vec_ordered<mode>")
35873         ("vec_unordered<mode>"): ... these.
35875         ("neg_fma<mode>4", "neg_fms<mode>4", "*smax<mode>3_vxe")
35876         ("*smin<mode>3_vxe", "*sminv2df3_vx", "*vec_extendv4sf")
35877         ("*vec_extendv2df"): New insn definitions.
35879 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35881         * config/s390/s390.md ("*adddi3_sign", "*subdi3_sign", "mulditi3")
35882         ("mulditi3_2", "*muldi3_sign"): New patterns.
35883         ("muldi3", "*muldi3", "mulsi3", "*mulsi3"): Add an expander and
35884         rename the pattern definition.
35886 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35888         * config/s390/s390.md ("indirect_jump"): Turn insn definition into
35889         expander.
35890         ("*indirect_jump", "*indirect2_jump"): New pattern definitions.
35892 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35894         * config/s390/s390.c (s390_expand_vec_init): Use vllezl
35895         instruction if possible.
35896         * config/s390/vector.md (vec_halfnumelts): New mode
35897         attribute.
35898         ("*vec_vllezlf<mode>"): New pattern.
35900 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35902         * config/s390/vector.md ("popcountv16qi2", "popcountv8hi2")
35903         ("popcountv4si2", "popcountv2di2"): Rename to ...
35904         ("popcount<mode>2", "popcountv8hi2_vx", "popcountv4si2_vx")
35905         ("popcountv2di2_vx"): ... these and add !TARGET_VXE to the
35906         condition.
35907         ("popcount<mode>2_vxe"): New pattern.
35909 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35911         * common/config/s390/s390-common.c (processor_flags_table): Add
35912         arch12.
35913         * config.gcc: Add arch12.
35914         * config/s390/driver-native.c (s390_host_detect_local_cpu):
35915         Default to arch12 for unknown CPU model numbers.
35916         * config/s390/s390-builtins.def: Add B_VXE builtin flag.
35917         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Adjust
35918         PROCESSOR_max sanity check.
35919         * config/s390/s390-opts.h (enum processor_type): Add
35920         PROCESSOR_ARCH12.
35921         * config/s390/s390.c (processor_table): Add arch12.
35922         (s390_expand_builtin): Add check for B_VXE flag.
35923         (s390_issue_rate): Add PROCESSOR_ARCH12.
35924         (s390_get_sched_attrmask): Likewise.
35925         (s390_get_unit_mask): Likewise.
35926         (s390_sched_score): Enable z13 scheduling for arch12.
35927         (s390_sched_reorder): Likewise.
35928         (s390_sched_variable_issue): Likewise.
35929         * config/s390/s390.h (enum processor_flags): Add PF_ARCH12 and
35930         PF_VXE.
35931         (s390_tune_attr): Use z13 scheduling also for arch12.
35932         (TARGET_CPU_ARCH12, TARGET_CPU_ARCH12_P, TARGET_CPU_VXE)
35933         (TARGET_CPU_VXE_P, TARGET_ARCH12, TARGET_ARCH12_P, TARGET_VXE)
35934         (TARGET_VXE_P): New macros.
35935         * config/s390/s390.md: Add arch12 to cpu attribute.  Add arch12
35936         and vxe to cpu_facility.  Add arch12 and vxe to enabled attribute.
35937         * config/s390/s390.opt: Add arch12 as processor_type.
35939 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35941         * config/s390/s390.md
35942         ("fixuns_truncdddi2", "fixuns_trunctddi2")
35943         ("fixuns_trunc<BFP:mode><GPR:mode>2"): Merge into ...
35944         ("fixuns_trunc<FP:mode><GPR:mode>2"): New expander.
35946         ("fixuns_trunc<BFP:mode><GPR:mode>2", "fixuns_trunc<mode>si2"):
35947         Rename expanders to ...
35949         ("fixuns_trunc<BFP:mode><GPR:mode>2_emu")
35950         ("fixuns_truncdddi2_emu"): ... these.
35952         ("fixuns_trunc<mode>si2_emu"): New expander.
35954         ("*fixuns_truncdfdi2_z13"): Rename to ...
35955         ("*fixuns_truncdfdi2_vx"): ... this.
35957 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35959         * config/s390/2964.md: Remove the single element vector compare
35960         instructions which are no longer used.
35961         * config/s390/s390.c (s390_select_ccmode): Remove handling of
35962         vector CCmodes.
35963         (s390_canonicalize_comparison): Remove handling of DFmode
35964         compares.
35965         (s390_expand_vec_compare_scalar): Remove function.
35966         (s390_emit_compare): Don't call s390_expand_vec_compare_scalar.
35967         * config/s390/s390.md ("*vec_cmp<insn_cmp>df_cconly"): Remove
35968         pattern.
35969         ("*cmp<mode>_ccs"): Add wfcdb instruction.
35971 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35973         * config/s390/s390.md ("mov<mode>_64dfp" DD_DF): Use vleig for loading a
35974         FP zero.
35975         ("*mov<mode>_64" DD_DF): Remove the vector instructions. These
35976         will anyway by matched by mov<mode>_64dfp.
35978 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35980         * config/s390/s390.md ("mov<mode>" SD_SF): Change vleg/vsteg to
35981         vlef/vstef.  Add missing operand to vleif.
35983 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35985         * config/s390/s390.c (s390_expand_vec_init): Enable vector load
35986         pair for all vector types with 64 bit elements.
35987         * config/s390/vx-builtins.md (V_HW_64): Move mode iterator to ...
35988         * config/s390/vector.md (V_HW_64): ... here.
35989         (V_128_NOSINGLE): New mode iterator.
35990         ("vec_init<V_HW:mode>"): Use V_128 as mode iterator.
35991         ("*vec_splat<mode>"): Use V_128_NOSINGLE mode iterator.
35992         ("*vec_tf_to_v1tf", "*vec_ti_to_v1ti"): New pattern definitions.
35993         ("*vec_load_pairv2di"): Change to ...
35994         ("*vec_load_pair<mode>"): ... this one.
35996 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35998         * config/s390/constraints.md: Add comments.
35999         (jKK): Reject element sizes > 8 bytes.
36000         * config/s390/s390.c (s390_split_ok_p): Enable splitting also for
36001         s_operands.
36002         * config/s390/s390.md: Add the s_operand checks formerly in
36003         s390_split_ok_p to various splitters where they are still
36004         required.
36005         * config/s390/vector.md ("mov<mode>" V_128): Add GPR alternatives
36006         for 128 bit vectors.  Plus two splitters.
36008 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
36010         * config/s390/s390.md: Rename the cpu facilty vec to vx throughout
36011         the file.
36013 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
36015         PR target/79893
36016         * config/s390/s390-c.c (s390_adjust_builtin_arglist): Issue an
36017         error if the boundary argument is not constant.
36019 2017-03-24  Jakub Jelinek  <jakub@redhat.com>
36021         PR rtl-optimization/80112
36022         * loop-doloop.c (doloop_condition_get): Don't check condition
36023         if cmp isn't SET with IF_THEN_ELSE src.
36025 2017-03-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36027         PR tree-optimization/80158
36028         * gimple-ssa-strength-reduction.c (replace_mult_candidate): When
36029         replacing a candidate statement, also replace it for the
36030         candidate's alternate interpretation.
36031         (replace_rhs_if_not_dup): Likewise.
36032         (replace_one_candidate): Likewise.
36034 2017-03-24  Richard Biener  <rguenther@suse.de>
36036         PR tree-optimization/80167
36037         * graphite-isl-ast-to-gimple.c
36038         (translate_isl_ast_to_gimple::is_valid_rename): Handle default-defs
36039         properly.
36040         (translate_isl_ast_to_gimple::get_rename): Likewise.
36042 2017-03-23  Kelvin Nilsen  <kelvin@gcc.gnu.org>
36044         * config/rs6000/rs6000.c (rs6000_option_override_internal): Change
36045         handling of certain combinations of target options, including the
36046         combinations -mpower8-vector vs. -mno-vsx, -mpower9-vector vs.
36047         -mno-power8-vector, and -mpower9_dform vs. -mno-power9-vector.
36049 2017-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36051         PR target/71436
36052         * config/arm/arm.md (*load_multiple): Add reload_completed to
36053         matching condition.
36055 2017-03-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36056             Richard Biener  <rguenth@suse.de>
36058         PR tree-optimization/79908
36059         PR tree-optimization/80136
36060         * tree-stdarg.c (expand_ifn_va_arg_1): For a VA_ARG whose LHS has
36061         been cast away, gimplify_and_add suffices.
36063 2017-03-23  Markus Trippelsdorf  <markus@trippelsdorf.de>
36065         * tree-vrp.c (identify_jump_threads): Delete avail_exprs.
36067 2017-03-23  Richard Biener  <rguenther@suse.de>
36069         PR tree-optimization/80032
36070         * gimplify.c (gimple_push_cleanup): Forced unconditional
36071         cleanups still have to go to the conditional_cleanups
36072         sequence.
36074 2017-03-22  Jakub Jelinek  <jakub@redhat.com>
36076         PR tree-optimization/80072
36077         * tree-ssa-reassoc.c (struct operand_entry): Change id field type
36078         to unsigned int.
36079         (next_operand_entry_id): Change type to unsigned int.
36080         (sort_by_operand_rank): Make sure to return the right return value
36081         even if unsigned fields are bigger than INT_MAX.
36082         (struct oecount): Change cnt and id type to unsigned int.
36083         (oecount_hasher::equal): Formatting fix.
36084         (oecount_cmp): Make sure to return the right return value
36085         even if unsigned fields are bigger than INT_MAX.
36086         (undistribute_ops_list): Change next_oecount_id type to unsigned int.
36088         PR c++/80129
36089         * gimplify.c (gimplify_modify_expr_rhs) <case COND_EXPR>: Clear
36090         TREE_READONLY on result if writing it more than once.
36092         PR sanitizer/80110
36093         * doc/invoke.texi (-fsanitize=thread): Document that with
36094         -fnon-call-exceptions atomics are not able to throw
36095         exceptions.
36097         PR sanitizer/80110
36098         * tsan.c: Include tree-eh.h.
36099         (instrument_builtin_call): Call maybe_clean_eh_stmt or
36100         maybe_clean_or_replace_eh_stmt where needed.
36101         (instrument_memory_accesses): Add cfg_changed argument.
36102         Call gimple_purge_dead_eh_edges on each block and set *cfg_changed
36103         if it returned true.
36104         (tsan_pass): Adjust caller.  Return TODO_cleanup_cfg if cfg_changed.
36106         PR rtl-optimization/63191
36107         * config/i386/i386.c (ix86_delegitimize_address): Turn into small
36108         wrapper function, moved the whole old content into ...
36109         (ix86_delegitimize_address_1): ... this.  New inline function.
36110         (ix86_find_base_term): Use ix86_delegitimize_address_1 with
36111         true as last argument instead of ix86_delegitimize_address.
36113 2017-03-22  Wilco Dijkstra  <wdijkstr@arm.com>
36115         * config/aarch64/aarch64.c (generic_branch_cost): Copy
36116         cortexa57_branch_cost.
36118 2017-03-22  Wilco Dijkstra  <wdijkstr@arm.com>
36120         * config/aarch64/aarch64.c (generic_tunings): Add AES fusion.
36122 2017-03-21  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
36124         PR target/80123
36125         * doc/md.texi (Constraints): Document wA constraint.
36126         * config/rs6000/constraints.md (wA): New.
36127         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Add wA reg_class.
36128         (rs6000_init_hard_regno_mode_ok): Init wA constraint.
36129         * config/rs6000/rs6000.h (RS6000_CONSTRAINT_wA): New.
36130         * config/rs6000/vsx.md (vsx_splat_<mode>): Use wA constraint.
36132 2017-03-22  Cesar Philippidis  <cesar@codesourcery.com>
36134         PR c++/80029
36135         * gimplify.c (is_oacc_declared): New function.
36136         (oacc_default_clause): Use it to set default flags for acc declared
36137         variables inside parallel regions.
36138         (gimplify_scan_omp_clauses): Strip firstprivate pointers for acc
36139         declared variables.
36140         (gimplify_oacc_declare): Gimplify the declare clauses.  Add the
36141         declare attribute to any decl as necessary.
36143 2017-03-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
36145         PR target/80082
36146         * config/arm/arm-isa.h (isa_bit_lpae): New feature bit.
36147         (ISA_ARMv7ve): Add isa_bit_lpae to the definition.
36148         * config/arm/arm-protos.h (arm_arch7ve): Rename into ...
36149         (arm_arch_lpae): This.
36150         * config/arm/arm.c (arm_arch7ve): Rename into ...
36151         (arm_arch_lpae): This.  Define it in term of isa_bit_lpae.
36152         * config/arm/arm.h (TARGET_HAVE_LPAE): Redefine in term of
36153         arm_arch_lpae.
36155 2017-03-22  Martin Liska  <mliska@suse.cz>
36157         PR target/79906
36158         * config/rs6000/rs6000.c (rs6000_inner_target_options): Show
36159         error message instead of an ICE.
36161 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36163         * doc/extend.texi (6.11 Additional Floating Types): Revise.
36165 2017-03-21  Kelvin Nilsen  <kelvin@gcc.gnu.org>
36167         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Add
36168         comments.
36169         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
36170         comments.
36172 2017-03-21  Martin Sebor  <msebor@redhat.com>
36174         * doc/extend.texi: Use "cannot" instead of "can't."
36175         * doc/hostconfig.texi: Same.
36176         * doc/install.texi: Same.
36177         * doc/invoke.texi: Same.
36178         * doc/loop.texi: Same.
36179         * doc/md.texi: Same.
36180         * doc/objc.texi: Same.
36181         * doc/rtl.texi: Same.
36182         * doc/tm.texi: Same.
36183         * doc/tm.texi.in: Same.
36184         * doc/trouble.texi: Same.
36186 2017-03-21  Alexandre Oliva  <aoliva@redhat.com>
36188         PR debug/63238
36189         * dwarf2out.c (struct checksum_attributes): Add at_alignment.
36190         (collect_checksum_attributes): Set it.
36191         (die_checksum_ordered): Use it.
36193 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36195         PR tree-optimization/79908
36196         * tree-stdarg.c (expand_ifn_va_arg_1): Revert the following
36197         change: For a VA_ARG whose LHS has been cast away, use
36198         force_gimple_operand to construct the side effects.
36200 2017-03-21  David Malcolm  <dmalcolm@redhat.com>
36202         PR translation/80001
36203         * omp-offload.c (oacc_loop_fixed_partitions): Make diagnostics
36204         more amenable to translation.
36205         (oacc_loop_auto_partitions): Likewise.
36207 2017-03-21  Marek Polacek  <polacek@redhat.com>
36208             Martin Sebor  <msebor@redhat.com>
36210         PR tree-optimization/80109
36211         * gimple-ssa-warn-alloca.c (alloca_call_type): Only call get_range_info
36212         on INTEGRAL_TYPE_P.
36214 2017-03-21  Jakub Jelinek  <jakub@redhat.com>
36215             Segher Boessenkool  <segher@kernel.crashing.org>
36217         PR target/80125
36218         * combine.c (can_combine_p): Revert the 2017-03-20 change, only
36219         check reg_used_between_p between insn and one of succ or succ2
36220         depending on if succ is artificial insn not inserted into insn
36221         stream.
36223 2017-03-21  Martin Liska  <mliska@suse.cz>
36225         PR gcov-profile/80081
36226         * Makefile.in: Add gcov-dump and fix installation of gcov-tool.
36227         * doc/gcc.texi: Include gcov-dump stuff.
36228         * doc/gcov-dump.texi: New file.
36230 2017-03-21  Toma Tabacu  <toma.tabacu@imgtec.com>
36232         PR rtl-optimization/79150
36233         * config/mips/mips.c (mips_block_move_loop): Emit a NOP after the
36234         conditional jump, if the jump is the last insn of the loop.
36236 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36237             Richard Biener  <rguenth@suse.de>
36239         PR tree-optimization/79908
36240         * tree-stdarg.c (expand_ifn_va_arg_1): For a VA_ARG whose LHS has
36241         been cast away, use force_gimple_operand to construct the side
36242         effects.
36244 2017-03-21  Martin Liska  <mliska@suse.cz>
36246         PR libfortran/79956
36247         * simplify-rtx.c (simplify_immed_subreg): Initialize a variable
36248         to NULL.
36250 2017-03-21  Brad Spengler <spender@grsecurity.net>
36252         PR plugins/80094
36253         * plugin.c (htab_hash_plugin): New function.
36254         (add_new_plugin): Use it and adjust.
36255         (parse_plugin_arg_opt): Adjust.
36256         (init_one_plugin): Likewise.
36258 2017-03-21  Richard Biener  <rguenther@suse.de>
36260         PR tree-optimization/80032
36261         * gimplify.c (gimple_push_cleanup): Add force_uncond parameter,
36262         if set force the cleanup to happen unconditionally.
36263         (gimplify_target_expr): Push inserted clobbers with force_uncond
36264         to avoid them being removed by control-dependent DCE.
36266 2017-03-21  Richard Biener  <rguenther@suse.de>
36268         PR tree-optimization/80122
36269         * tree-inline.c (copy_bb): Do not expans va-arg packs or
36270         va_arg_pack_len when the inlined call stmt requires pack
36271         expansion itself.
36272         * tree-inline.h (struct copy_body_data): Make call_stmt a gcall *.
36274 2017-03-21  Jakub Jelinek  <jakub@redhat.com>
36276         PR sanitizer/78158
36277         * tsan.c (instrument_builtin_call): If the memory model argument
36278         is not a constant, assume it is valid.
36280         PR c/67338
36281         * fold-const.c (round_up_loc): Negate divisor in unsigned type to
36282         avoid UB.
36284 2017-03-20  Segher Boessenkool  <segher@kernel.crashing.org>
36286         PR rtl-optimization/79910
36287         * combine.c (can_combine_p): Do not allow combining an I0 or I1
36288         if its dest is used by an insn before I2 (other than the combined
36289         insns themselves, which are properly handled already).
36291 2017-03-20  Segher Boessenkool  <segher@kernel.crashing.org>
36293         Revert:
36294         2017-03-17  Bernd Schmidt  <bschmidt@redhat.com>
36296         * combine.c (record_used_regs): New static function.
36297         (try_combine): Handle situations where there is an additional
36298         instruction between I2 and I3 which needs to have a LOG_LINK
36299         updated.
36301         Revert:
36302         2017-03-17  Jim Wilson  <jim.wilson@linaro.org>
36304         * combine.c (try_combine): Delete redundant i1 test.  Call
36305         prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
36307 2017-03-20  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
36309         PR target/80083
36310         * config/rs6000/rs6000.md (*movsi_internal1): Fix constraints for
36311         alternatives 13/14.
36313 2017-03-20  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36315         PR tree-optimization/80054
36316         * gimple-ssa-strength-reduction.c (all_phi_incrs_profitable): Fail
36317         the optimization if a PHI or any of its arguments is not dominated
36318         by the candidate's basis.  Use gphi* rather than gimple* as
36319         appropriate.
36320         (replace_profitable_candidates): Clean up a gimple* variable that
36321         should be a gphi* variable.
36323 2017-03-20  Martin Sebor  <msebor@redhat.com>
36325         PR c++/52477
36326         * doc/extend.texi (attribute constructor): Document present limitation.
36328 2017-03-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
36330         PR target/79963
36331         * config/rs6000/altivec.h (vec_all_ne): Under __cplusplus__ and
36332         __POWER9_VECTOR__ #ifdef control, change template definition to
36333         use Power9-specific built-in function.
36334         (vec_any_eq): Likewise.
36335         * config/rs6000/vector.md (vector_ae_v2di_p): Change the flag used
36336         to control outcomes from this test.
36337         (vector_ae_<mode>p): For VEC_F modes, likewise.
36339 2017-03-20  Ian Lance Taylor  <iant@google.com>
36341         * config/i386/i386.c (ix86_function_regparm): Save an extra
36342         register for -fsplit-stack with DECL_STATIC_CHAIN.
36344 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com>
36346         PR target/79912
36347         * config/riscv/riscv.c (riscv_preferred_reload_class): Remove.
36348         (TARGET_PREFERRED_RELOAD_CLASS): Likewise.
36350 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com>
36352         * config/riscv/riscv.c (riscv_print_operand): Use "fence
36353         iorw,ow".
36354         * config/riscv/sync.mc (mem_thread_fence_1): Use "fence
36355         iorw,iorw".
36357 2017-03-20  Marek Polacek  <polacek@redhat.com>
36359         PR sanitizer/80063
36360         * asan.c (DEF_SANITIZER_BUILTIN): Use do { } while (0).
36362 2017-03-20  Richard Biener  <rguenther@suse.de>
36364         PR tree-optimization/80113
36365         * graphite-isl-ast-to-gimple.c (copy_loop_phi_nodes): Do not
36366         allocate extra SSA name for PHI def.
36367         (add_close_phis_to_outer_loops): Likewise.
36368         (add_close_phis_to_merge_points): Likewise.
36369         (copy_loop_close_phi_args): Likewise.
36370         (copy_cond_phi_nodes): Likewise.
36372 2017-03-20  Martin Liska  <mliska@suse.cz>
36374         PR middle-end/79753
36375         * tree-chkp.c (chkp_build_returned_bound): Do not build
36376         returned bounds for a LHS that's not a BOUNDED_P type.
36378 2017-03-20  Martin Liska  <mliska@suse.cz>
36380         PR target/79769
36381         PR target/79770
36382         * tree-chkp.c (chkp_find_bounds_1): Handle REAL_CST,
36383         COMPLEX_CST and VECTOR_CST.
36385 2017-03-20  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
36387         PR target/78857
36388         * config/s390/s390.md ("cmp<mode>_ccs_0"): Add a clobber of the
36389         target operand.  A new splitter adds the clobber statement in case
36390         the target operand is dead anyway.
36392 2017-03-19  Gerald Pfeifer  <gerald@pfeifer.com>
36394         * doc/install.texi (Specific) <sparc-*-linux*>: No longer refer
36395         to age-old versions of binutils and glibc.
36397 2017-03-18  Segher Boessenkool  <segher@kernel.crashing.org>
36399         * doc/contrib.texi (Contributors): Remove duplicate entry for myself.
36401 2017-03-18  Gerald Pfeifer  <gerald@pfeifer.com>
36403         * doc/contrib.texi (Contributors): Add Segher Boessenkool.
36405 2017-03-18  Gerald Pfeifer  <gerald@pfeifer.com>
36407         * doc/install.texi (Specific) <arm-*-eabi>: Remove old
36408         requirement for binutils 2.13.
36410 2017-03-17  Jim Wilson  <jim.wilson@linaro.org>
36412         * combine.c (try_combine): Delete redundant i1 test.  Call
36413         prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
36415 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com
36417         * doc/install.texi (Specific) <riscv32-*-elf>: Add riscv32-*-elf,
36418         riscv32-*-linux, riscv64-*-elf, riscv64-*-linux to the table of
36419         contents.
36420         <riscv64-*-elf>: Re-arrange section
36421         <riscv32-*-elf>: Add a note about requiring binutils 2.28.
36422         <riscv32-*-linux>: Likewise.
36423         <riscv64-*-elf>: Likewise
36424         <riscv64-*-linux>: Likewise.
36426 2017-03-17  Richard Earnshaw  <rearnsha@arm.com>
36428         PR target/80052
36429         * aarch64.opt(verbose-cost-dump): Fix typo.
36431 2017-03-17  Pat Haugen  <pthaugen@us.ibm.com>
36433         PR target/79951
36434         * config/rs6000/rs6000.md (copysign<mode>3_fcpsgn): Test
36435         for VECTOR_UNIT_VSX_P (<MODE>mode) too.
36437 2017-03-17  Bernd Schmidt  <bschmidt@redhat.com>
36439         * reload.c (find_reloads): When reloading a nonoffsettable address,
36440         use RELOAD_OTHER for it and its address reloads.
36442         PR rtl-optimization/79910
36443         * combine.c (record_used_regs): New static function.
36444         (try_combine): Handle situations where there is an additional
36445         instruction between I2 and I3 which needs to have a LOG_LINK
36446         updated.
36448 2017-03-17  Jeff Law  <law@redhat.com>
36450         PR tree-optimization/71437
36451         * tree-vrp.c (simplify_stmt_for_jump_threading): Lookup the
36452         conditional in the hash table first.
36453         (vrp_dom_walker::before_dom_children): Extract condition from
36454         ASSERT_EXPR.  Record condition, its inverion and any implied
36455         conditions as well.
36457 2017-03-17  Marek Polacek  <polacek@redhat.com>
36458             Markus Trippelsdorf  <markus@trippelsdorf.de>
36460         PR tree-optimization/80079
36461         * gimple-ssa-store-merging.c (class pass_store_merging): Initialize
36462         m_stores_head.
36464 2017-03-17  Richard Biener  <rguenther@suse.de>
36466         PR middle-end/80075
36467         * tree-eh.c (stmt_could_throw_1_p): Only handle gimple assigns.
36468         Properly verify the LHS before the RHS possibly claims to be
36469         handled.
36470         (stmt_could_throw_p): Hande gimple conds fully here.  Clobbers
36471         do not throw.
36473 2017-03-17  Martin Jambor  <mjambor@suse.cz>
36475         * doc/invoke.texi (Option Options): Include -fipa-vrp in the list.
36476         (List of -O2 options): Likewise.
36477         (-fipa-bit-cp): Replace "ipa" with "interprocedural."
36478         (-fipa-vrp) New.
36480 2017-03-17  Tom de Vries  <tom@codesourcery.com>
36482         * gcov-dump.c (print_usage): Print bug_report_url.
36484 2017-03-17  Richard Biener  <rguenther@suse.de>
36486         PR middle-end/80050
36487         * genmatch.c (parser::next): Remove pointless check for CPP_EOF.
36488         (parser::peek): Likewise.
36490 2017-03-17  Richard Biener  <rguenther@suse.de>
36492         PR tree-optimization/80048
36493         * sese.c (free_sese_info): Properly release rename_map and
36494         copied_bb_map elements.
36496 2017-03-16  Alexandre Oliva  <aoliva@redhat.com>
36498         * gimple-ssa-store-merging.c (struct imm_store_chain_info):
36499         Add linked-list forward and backlinks.  Insert on
36500         construction, remove on destruction.
36501         (class pass_store_merging): Add m_stores_head field.
36502         (pass_store_merging::terminate_and_process_all_chains):
36503         Iterate over m_stores_head list.
36504         (pass_store_merging::terminate_all_aliasing_chains):
36505         Likewise.
36506         (pass_store_merging::execute): Check for debug stmts first.
36507         Push new chains onto the m_stores_head stack.
36509 2017-03-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
36511         PR target/71294
36512         * config/rs6000/vsx.md (vsx_splat_<mode>, VSX_D iterator): Allow a
36513         SPLAT operation on ISA 2.07 64-bit systems that have direct move,
36514         but no MTVSRDD support, by doing MTVSRD and XXPERMDI.
36516 2017-03-16  Jeff Law  <law@redhat.com>
36518         PR tree-optimization/71437
36519         * tree-ssa-dom.c (dom_opt_dom_walker): Remove thread_across_edge
36520         member function.  Implementation moved into after_dom_children
36521         member function and into the threader's thread_outgoing_edges
36522         function.
36523         (dom_opt_dom_walker::after_dom_children): Simplify by moving
36524         some code into new thread_outgoing_edges.
36525         * tree-ssa-threadedge.c (thread_across_edge): Make static and simplify
36526         definition.  Simplify marker handling (do it here).   Assume we always
36527         have the available expression and the const/copies tables.
36528         (thread_outgoing_edges): New function extracted from tree-ssa-dom.c
36529         and tree-vrp.c
36530         * tree-ssa-threadedge.h (thread_outgoing_edges): Declare.
36531         * tree-vrp.c (equiv_stack): No longer file scoped.
36532         (vrp_dom_walker): New class.
36533         (vrp_dom_walker::before_dom_children): New member function.
36534         (vrp_dom_walker::after_dom_children): Likewise.
36535         (identify_jump_threads):  Setup domwalker.  Use it rather than
36536         walking edges in a random order by hand.  Simplify setup/finalization.
36537         (finalize_jump_threads): Remove.
36538         (vrp_finalize): Do not call identify_jump_threads here.
36539         (execute_vrp): Do it here instead and call thread_through_all_blocks
36540         here too.
36542         PR tree-optimization/71437
36543         * tree-ssa-dom.c (pfn_simplify): Add basic_block argument.  All
36544         callers changed.
36545         (simplify_stmt_for_jump_threading): Add basic_block argument.  All
36546         callers changed.
36547         (lhs_of_dominating_assert): Moved from here into tree-vrp.c.
36548         (dom_opt_dom_walker::thread_across_edge): Remove
36549         handle_dominating_asserts argument.  All callers changed.
36550         (record_temporary_equivalences_from_stmts_at_dest): Corresponding
36551         changes.  Remove calls to lhs_of_dominating_assert.  Other
36552         uses of handle_dominating_asserts turn into unconditional code
36553         (simplify_control_stmt_condition_1): Likewise.
36554         (simplify_control_stmt_condition): Likewise.
36555         (thread_through_normal_block, thread_across_edge): Likewise.
36556         * tree-ssa-threadedge.h (thread_across_edge): Corresponding changes.
36557         * tree-vrp.c (lhs_of_dominating_assert): Move here.  Return original
36558         object if it is not an SSA_NAME.
36559         (simplify_stmt_for_jump_threading): Call lhs_of_dominating_assert
36560         before calling into the VRP specific simplifiers.
36561         (identify_jump_threads): Remove handle_dominating_asserts
36562         argument.
36564 2017-03-16  Jakub Jelinek  <jakub@redhat.com>
36566         PR fortran/79886
36567         * tree-diagnostic.c (default_tree_printer): No longer static.
36568         * tree-diagnostic.h (default_tree_printer): New prototype.
36570 2017-03-16  Tamar Christina  <tamar.christina@arm.com>
36572         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>)
36573         Change ins into fmov.
36575 2017-03-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36577         * config/aarch64/iterators.md (h_con): Return "x" for V4HF and V8HF.
36578         * config/aarch64/aarch64-simd.md (*aarch64_fma4_elt_from_dup<mode>):
36579         Use h_con constraint for operand 1.
36580         (*aarch64_fnma4_elt_from_dup<mode>): Likewise.
36581         (*aarch64_mulx_elt_from_dup<mode>): Likewise for operand 2.
36583 2017-03-15  Jeff Law  <law@redhat.com>
36585         PR tree-optimization/71437
36586         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): New function.
36587         (record_temporary_equivalences): Use it.
36589         PR tree-optimization/71437
36590         * tree-ssa-dom.c (struct cond_equivalence): Moved from here into
36591         tree-ssa-scopedtables.
36592         (lookup_avail_expr, build_and_record_new_cond): Likewise.
36593         (record_conditions, record_cond, vuse_eq): Likewise.
36594         (record_edge_info): Adjust to API tweak of record_conditions.
36595         (simplify_stmt_for_jump_threading): Similarly for lookup_avail_expr.
36596         (record_temporary_equivalences, optimize_stmt): Likewise.
36597         (eliminate_redundant_computations): Likewise.
36598         (record_equivalences_from_stmt): Likewise.
36599         * tree-ssa-scopedtables.c: Include options.h and params.h.
36600         (vuse_eq): New function, moved from tree-ssa-dom.c
36601         (build_and_record_new_cond): Likewise.
36602         (record_conditions): Likewise.  Accept vector of conditions rather
36603         than edge_equivalence structure for first argument.
36604         for the first argument.
36605         (avail_exprs_stack::lookup_avail_expr): New member function, moved
36606         from tree-ssa-dom.c.
36607         (avail_exprs_stack::record_cond): Likewise.
36608         * tree-ssa-scopedtables.h (struct cond_equivalence): Moved here
36609         from tree-ssa-dom.c.
36610         (avail_exprs_stack): Add new member functions lookup_avail_expr
36611         and record_cond.
36612         (record_conditions): Declare.
36614 2017-03-15  Vladimir Makarov  <vmakarov@redhat.com>
36616         PR target/80017
36617         * lra-constraints.c (process_alt_operands): Increase reject for
36618         reloading an input/output operand.
36620 2017-03-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
36622         PR target/79038
36623         * config/rs6000/rs6000.md (float<QHI:mode><IEEE128:mode>2): Define
36624         insns to convert from signed/unsigned char/short to IEEE 128-bit
36625         floating point.
36626         (floatuns<QHI:mode><IEEE128:mode>2): Likewise.
36628 2017-03-15  Uros Bizjak  <ubizjak@gmail.com>
36630         PR target/80019
36631         * config/i386/i386.c (ix86_vector_duplicate_value): Create
36632         subreg of inner mode for values already in registers.
36634 2017-03-15  Bernd Schmidt  <bschmidt@redhat.com>
36636         * config/c6x/c6x.c (hwloop_optimize): Handle case where the old
36637         iteration reg is used after the loop.
36639 2017-03-14  Martin Sebor  <msebor@redhat.com>
36641         PR tree-optimization/79800
36642         * gimple-ssa-sprintf.c (format_floating: Add argument.  Handle
36643         precision in negative-positive range.
36644         (format_floating): Call non-const overload with adjusted precision.
36646 2017-03-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
36648         PR target/79947
36649         * config/rs6000/rs6000.h (TARGET_FRSQRTES): Add check for
36650         -mpowerpc-gfxopt.
36652 2017-03-14  Martin Sebor  <msebor@redhat.com>
36654         PR middle-end/80020
36655         * builtin-attrs.def (ATTR_ALLOC_SIZE_2_NOTHROW_LIST): New macro.
36656         * builtins.def (aligned_alloc): Use it.
36658         PR c/79936
36659         * Makefile.in (GTFILES): Add calls.c.
36660         * calls.c: Include "gt-calls.h".
36662 2017-03-14  Bernd Schmidt  <bschmidt@redhat.com>
36664         PR rtl-optimization/79728
36665         * regs.h (struct target_regs): New field
36666         x_contains_allocatable_regs_of_mode.
36667         (contains_allocatable_regs_of_mode): New macro.
36668         * reginfo.c (init_reg_sets_1): Initialize it, and change
36669         contains_reg_of_mode so it includes global regs as well.
36670         * reload.c (push_reload): Use contains_allocatable_regs_of_mode
36671         rather than contains_regs_of_mode.
36673 2017-03-14  Martin Liska  <mliska@suse.cz>
36675         * doc/invoke.texi: Document options that can't be combined with
36676         -fcheck-pointer-bounds.
36678 2017-03-14  Martin Liska  <mliska@suse.cz>
36680         PR middle-end/79831
36681         * doc/invoke.texi (-Wchkp): Document the option.
36683 2017-03-14  Martin Liska  <mliska@suse.cz>
36685         * Makefile.in: Install gcov-dump.
36687 2017-03-14  Martin Liska  <mliska@suse.cz>
36689         * multiple_target.c (expand_target_clones): Bail out for
36690         an invalid attribute.
36692 2017-03-14  Richard Biener  <rguenther@suse.de>
36694         * alias.c (struct alias_set_entry): Pack properly.
36695         * cfgloop.h (struct loop): Likewise.
36696         * cse.c (struct set): Likewise.
36697         * ipa-utils.c (struct searchc_env): Likewise.
36698         * loop-invariant.c (struct invariant): Likewise.
36699         * lra-remat.c (struct cand): Likewise.
36700         * recog.c (struct change_t): Likewise.
36701         * rtl.h (struct address_info): Likewise.
36702         * symbol-summary.h (function_summary): Likewise.
36703         * tree-loop-distribution.c (struct partition): Likewise.
36704         * tree-object-size.c (struct object_size_info): Likewise.
36705         * tree-ssa-loop-ivopts.c (struct cost_pair): Likewise.
36706         * tree-ssa-threadupdate.c (struct ssa_local_info_t): Likewise.
36707         * tree-vect-data-refs.c (struct _vect_peel_info): Likewise.
36708         * tree-vect-slp.c (struct _slp_oprnd_info): Likewise.
36709         * tree-vect-stmts.c (struct simd_call_arg_info): Likewise.
36710         * tree-vectorizer.h (struct _loop_vec_info): Likewise.
36711         (struct _stmt_vec_info): Likewise.
36713 2017-03-14  Martin Liska  <mliska@suse.cz>
36715         PR target/79892
36716         * multiple_target.c (create_dispatcher_calls): Check that
36717         a target can create a function dispatcher.
36719 2017-03-14  Martin Liska  <mliska@suse.cz>
36721         PR lto/66295
36722         * multiple_target.c (expand_target_clones): Drop local.local
36723         flag for default implementation.
36725 2017-03-14  Richard Biener  <rguenther@suse.de>
36727         PR tree-optimization/80030
36728         * tree-vect-stmts.c (vectorizable_store): Plug memleak.
36730 2017-03-13  Kito Cheng  <kito.cheng@gmail.com>
36732         * config/riscv/riscv.c (riscv_emit_float_compare>: Use
36733         gcc_fallthrough() instead of __attribute__((fallthrough));
36735 2017-03-13  Gerald Pfeifer  <gerald@pfeifer.com>
36737         * doc/gcc.texi: Remove "up" link to (DIR).
36738         * doc/gccint.texi: Ditto.
36740 2017-03-13  Gerald Pfeifer  <gerald@pfeifer.com>
36742         * doc/install.texi (Specific) <avr>: Remove reference to
36743         binutils 2.13.
36745 2017-03-13  Jeff Law  <law@redhat.com>
36747         * config/riscv/riscv.c (riscv_emit_float_compare): Use fallthru
36748         attribute rather than comments.
36750         * config/pdp11/pdp11.md (movmemhi): Adjust operand numbers to
36751         match_scratch operand is highest.
36753 2017-03-13  Martin Liska  <mliska@suse.cz>
36755         PR middle-end/78339
36756         * ipa-pure-const.c (warn_function_noreturn): If the declarations
36757         is a CHKP clone, use original declaration.
36759 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36761         * config/arc/arc.c (arc_init): Use multiplier whenever we have it.
36762         (arc_conditional_register_usage): Use a different allocation order
36763         when optimizing for size.
36764         * common/config/arc/arc-common.c (arc_option_optimization_table):
36765         Section anchors default on when optimizing for size.
36767 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36769         * config/arc/arc.md (*tst_bitfield_tst): Fix pattern.
36771 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36773         * config/arc/arc.c (arc_output_addsi): Emit code density adds.
36774         * config/arc/arc.md (cpu_facility): Add cd variant.
36775         (*movqi_insn): Add code density variant.
36776         (*movhi_insn): Likewise.
36777         (*movqi_insn): Likewise.
36778         (*addsi3_mixed): Likewise.
36779         (subsi3_insn): Likewise.
36781 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36783         * config/arc/arc.md (movsi_cond_exec): Update constraint.
36785 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36787         * config/arc/arc.c (arc_legitimize_pic_address): Handle PIC
36788         expressions with MINUS and UNARY ops.
36790 2017-03-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36792         PR target/79911
36793         * config/arm/neon.md (vec_sel_widen_ssum_lo<VQI:mode><VW:mode>3):
36794         Rename to...
36795         (vec_sel_widen_ssum_lo<mode><V_half>3): ... This. Avoid mismatch
36796         between vec_select and vector argument.
36797         (vec_sel_widen_ssum_hi<VQI:mode><VW:mode>3): Rename to...
36798         (vec_sel_widen_ssum_hi<mode><V_half>3): ... This. Likewise.
36799         (vec_sel_widen_usum_lo<VQI:mode><VW:mode>3): Rename to...
36800         (vec_sel_widen_usum_lo<mode><V_half>3): ... This.
36801         (vec_sel_widen_usum_hi<VQI:mode><VW:mode>3): Rename to...
36802         (vec_sel_widen_usum_hi<mode><V_half>3): ... This.
36804 2017-03-13  Richard Biener  <rguenther@suse.de>
36806         PR other/79991
36807         * params.def (vect-max-peeling-for-alignment): Fix typo.
36809 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36811         * doc/install.texi (Specific) <mips-*-*>: Remove description of
36812         issue that only occurred with binutils below 2.18.
36814 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36816         * doc/install.texi (Specific) <cris-axis-elf>: No longer
36817         refer to binutils 2.11/2.12 minimum.
36819 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36821         * doc/install.texi (Specific) <powerpc-*-*>: Remove link to
36822         ftp.kernel.org and simplify binutils requirement.
36824 2017-03-11  Gerald Pfeifer  <gerald@pfeifer.com>
36826         * doc/invoke.texi (Warning Options): Fix spelling of link-time
36827         optimization.
36828         (Optimize Options): Ditto.  Also remove redundancy.
36830 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36832         PR translation/79848
36833         * ipa-devirt.c (warn_types_mismatch): Simplify uses of "%<%s%>" to
36834         "%qs".
36835         * ipa-pure-const.c (suggest_attribute): Likewise.  Convert _
36836         to G_ to avoid double translation.
36838 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36840         PR translation/79923
36841         * auto-profile.c (get_combined_location): Convert leading
36842         character of diagnostics to lower case and remove trailing period.
36843         (read_profile): Likewise for various diagnostics.
36844         * config/arm/arm.c (arm_option_override): Remove trailing period
36845         from various diagnostics.
36846         * config/msp430/msp430.c (msp430_expand_delay_cycles): Likewise.
36847         (msp430_expand_delay_cycles): Likewise.
36849 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36851         PR target/79925
36852         * config/aarch64/aarch64.c (aarch64_validate_mcpu): Quote the
36853         full command-line argument, rather than just "str".
36854         (aarch64_validate_march): Likewise.
36855         (aarch64_validate_mtune): Likewise.
36857 2017-03-10  Bernd Schmidt  <bschmidt@redhat.com>
36859         PR rtl-optimization/78911
36860         * lra-assigns.c (must_not_spill_p): New function.
36861         (spill_for): Use it.
36863 2017-03-10  Jakub Jelinek  <jakub@redhat.com>
36865         PR tree-optimization/79981
36866         * tree-vrp.c (extract_range_basic): Handle IMAGPART_EXPR of
36867         ATOMIC_COMPARE_EXCHANGE ifn result.
36868         (stmt_interesting_for_vrp, vrp_visit_stmt): Handle
36869         IFN_ATOMIC_COMPARE_EXCHANGE.
36871 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36873         PR driver/79875
36874         * opts.c (parse_sanitizer_options): Add missing question mark to
36875         "did you mean" message.
36877 2017-03-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36879         * config/rs6000/rs6000-builtin.def (VMULEUB_UNS): Remove orphaned
36880         built-in.
36881         (VMULEUH_UNS): Likewise.
36882         (VMULOUB_UNS): Likewise.
36883         (VMULOUH_UNS): Likewise.
36884         * config/rs6000/rs6000.c (builtin_function_type): Remove
36885         references to ALTIVEC_BUILTIN_VMUL[EO]U[BH]_UNS.
36887 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36889         PR bootstrap/79952
36890         * read-rtl-function.c (function_reader::read_rtx_operand): Update
36891         x with result of extra_parsing_for_operand_code_0.
36892         (function_reader::extra_parsing_for_operand_code_0): Convert
36893         return type from void to rtx, returning x.  When reading
36894         SYMBOL_REF with SYMBOL_FLAG_HAS_BLOCK_INFO, reallocate x to the
36895         larger size containing struct block_symbol.
36897 2017-03-10  Segher Boessenkool  <segher@kernel.crashing.org>
36899         * config/rs6000/rs6000.c (rs6000_option_override_internal): Disallow
36900         -mfloat128-hardware without -m64.
36902 2017-03-10  Will Schmidt <will_schmidt@vnet.ibm.com>
36904         PR target/79941
36905         * config/rs6000/rs6000.c (builtin_function_type): Add VMUL*U[HB]
36906         entries to the case statement that marks unsigned arguments to
36907         overloaded functions.
36909 2017-03-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
36911         * config/rs6000/rs6000.c (rs6000_option_override_internal): Fix
36912         two typographic errors in the handling of TARGET_UPPER_REGS_DI.
36914 2017-03-10  Pat Haugen  <pthaugen@us.ibm.com>
36916         PR target/79907
36917         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Test
36918         TARGET_UPPER_REGS_DI when setting 'wi' constraint regclass.
36920 2017-03-10  Martin Liska  <mliska@suse.cz>
36922         PR target/65705
36923         PR target/69804
36924         * toplev.c (process_options): Enable MPX with LSAN and UBSAN.
36925         * tree-chkp.c (chkp_walk_pointer_assignments): Verify that
36926         FIELD != NULL.
36928 2017-03-10  Olivier Hainque  <hainque@adacore.com>
36930         * tree-switch-conversion (array_value_type): Start by resetting
36931         candidate type to it's main variant.
36933 2017-03-10  Jakub Jelinek  <jakub@redhat.com>
36935         PR rtl-optimization/79909
36936         * combine.c (try_combine): Use simplify_replace_rtx on individual
36937         CALL_INSN_FUNCTION_USAGE elements instead of replace_rtx on copy_rtx
36938         of the whole CALL_INSN_FUNCTION_USAGE.
36940         PR tree-optimization/79972
36941         * gimple-ssa-warn-alloca.c (alloca_call_type): Only call
36942         get_range_info on SSA_NAMEs.  Formatting fixes.
36944 2017-03-10  Richard Biener  <rguenther@suse.de>
36945             Jakub Jelinek  <jakub@redhat.com>
36947         PR tree-optimization/77975
36948         * tree-ssa-loop-niter.c (get_base_for): Allow phi argument from latch
36949         edge to be constant.
36950         (get_val_for): For constant x return it.  Formatting fix.
36951         (loop_niter_by_eval): Avoid pointless looping if the next iteration
36952         would use the same bases as the current one.
36954 2017-03-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36956         * config/rs6000/rs6000.c (rs6000_gen_le_vsx_permute): Use rotate
36957         instead of vec_select for V1TImode.
36958         * conifg/rs6000/vsx.md (VSX_LE): Remove mode iterator that is no
36959         longer needed.
36960         (VSX_LE_128): Add V1TI to this mode iterator.
36961         (*vsx_le_perm_load_<mode>): Change to use VSX_D mode iterator.
36962         (*vsx_le_perm_store_<mode>): Likewise.
36963         (pre-reload splitter for VSX stores): Likewise.
36964         (post-reload splitter for VSX stores): Likewise.
36965         (*vsx_xxpermdi2_le_<mode>): Likewise.
36966         (*vsx_lxvd2x2_le_<mode>): Likewise.
36967         (*vsx_stxvd2x2_le_<mode>): Likewise.
36969 2017-03-09  Michael Eager  <eager@eagercon.com>
36971         Correct failures with --enable-checking=yes,rtl.
36973         * config/microblaze/microblaze.c (microblaze_expand_shift):
36974         Replace GET_CODE test with CONST_INT_P and INTVAL test with
36975         test for const0_rtx.
36976         * config/microblaze/microblaze.md (ashlsi3_byone, ashrsi3_byone,
36977         lshrsi3_byone): Replace INTVAL with test for const1_rtx.
36979 2017-03-09  Richard Biener  <rguenther@suse.de>
36981         PR tree-optimization/79977
36982         * graphite-scop-detection.c (scop_detection::merge_sese):
36983         Handle the case of extra exits to blocks dominating the entry.
36985 2017-03-09  Toma Tabacu  <toma.tabacu@imgtec.com>
36987         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
36988         Document rdynamic.
36990 2017-03-09  Vladimir Makarov  <vmakarov@redhat.com>
36992         PR rtl-optimization/79949
36993         * lra-constraints.c (process_alt_operands): Check memory when
36994         trying to predict a cycle.  Print about the overall increase.
36996 2017-03-09  Richard Biener  <rguenther@suse.de>
36998         PR middle-end/79971
36999         * gimple-expr.c (useless_type_conversion_p): Preserve
37000         TYPE_SATURATING for fixed-point types.
37002 2017-03-09  Richard Biener  <rguenther@suse.de>
37004         PR ipa/79970
37005         * ipa-prop.c (ipa_modify_formal_parameters): Avoid changing
37006         alignment of BLKmode params.
37008 2017-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
37010         PR target/79913
37011         * config/aarch64/iterators.md (VALL_F16_NO_V2Q): New mode iterator.
37012         (VALL_NO_V2Q): Likewise.
37013         (VDQF_DF): Delete.
37014         * config/aarch64/aarch64-simd.md
37015         (aarch64_dup_lane_<vswap_width_name><mode>): Use VALL_F16_NO_V2Q
37016         iterator.
37017         (*aarch64_simd_vec_copy_lane_<vswap_width_name><mode>): Use
37018         VALL_NO_V2Q mode iterator.
37019         (*aarch64_vgetfmulx<mode>): Use VDQF iterator.
37021 2017-03-09  Martin Liska  <mliska@suse.cz>
37023         PR tree-optimization/79631
37024         * tree-chkp-opt.c (chkp_is_constant_addr): Call
37025         tree_int_cst_sign_bit just for INTEGER constants.
37027 2017-03-09  Martin Liska  <mliska@suse.cz>
37029         PR target/65705
37030         PR target/69804
37031         * toplev.c (process_options): Disable -fcheck-pointer-bounds with
37032         sanitizers.
37034 2017-03-09  Marek Polacek  <polacek@redhat.com>
37036         PR c++/79672
37037         * tree.c (inchash::add_expr): Handle TREE_VEC.
37039 2017-03-09  Martin Liska  <mliska@suse.cz>
37041         PR ipa/79764
37042         (chkp_narrow_size_and_offset): New function.
37043         (chkp_parse_array_and_component_ref): Support BIT_FIELD_REF.
37044         (void chkp_parse_bit_field_ref): New function.
37045         (chkp_make_addressed_object_bounds): Add case for BIT_FIELD_REF.
37046         (chkp_process_stmt): Use chkp_parse_bit_field_ref.
37048 2017-03-09  Martin Liska  <mliska@suse.cz>
37050         PR ipa/79761
37051         * tree-chkp.c (chkp_get_bound_for_parm): Get bounds for a param.
37052         (chkp_find_bounds_1): Remove gcc_unreachable.
37054 2017-03-09  Jakub Jelinek  <jakub@redhat.com>
37056         PR sanitizer/79944
37057         * asan.c (get_mem_refs_of_builtin_call): For BUILT_IN_ATOMIC* and
37058         BUILT_IN_SYNC*, determine the access type from the size suffix and
37059         always build a MEM_REF with that type.  Handle forgotten
37060         BUILT_IN_SYNC_FETCH_AND_NAND_16 and BUILT_IN_SYNC_NAND_AND_FETCH_16.
37062         PR target/79932
37063         * config/i386/avx512vlintrin.h (_mm256_cmpge_epi32_mask,
37064         _mm256_cmpge_epi64_mask, _mm256_cmpge_epu32_mask,
37065         _mm256_cmpge_epu64_mask, _mm256_cmple_epi32_mask,
37066         _mm256_cmple_epi64_mask, _mm256_cmple_epu32_mask,
37067         _mm256_cmple_epu64_mask, _mm256_cmplt_epi32_mask,
37068         _mm256_cmplt_epi64_mask, _mm256_cmplt_epu32_mask,
37069         _mm256_cmplt_epu64_mask, _mm256_cmpneq_epi32_mask,
37070         _mm256_cmpneq_epi64_mask, _mm256_cmpneq_epu32_mask,
37071         _mm256_cmpneq_epu64_mask, _mm256_mask_cmpge_epi32_mask,
37072         _mm256_mask_cmpge_epi64_mask, _mm256_mask_cmpge_epu32_mask,
37073         _mm256_mask_cmpge_epu64_mask, _mm256_mask_cmple_epi32_mask,
37074         _mm256_mask_cmple_epi64_mask, _mm256_mask_cmple_epu32_mask,
37075         _mm256_mask_cmple_epu64_mask, _mm256_mask_cmplt_epi32_mask,
37076         _mm256_mask_cmplt_epi64_mask, _mm256_mask_cmplt_epu32_mask,
37077         _mm256_mask_cmplt_epu64_mask, _mm256_mask_cmpneq_epi32_mask,
37078         _mm256_mask_cmpneq_epi64_mask, _mm256_mask_cmpneq_epu32_mask,
37079         _mm256_mask_cmpneq_epu64_mask, _mm_cmpge_epi32_mask,
37080         _mm_cmpge_epi64_mask, _mm_cmpge_epu32_mask, _mm_cmpge_epu64_mask,
37081         _mm_cmple_epi32_mask, _mm_cmple_epi64_mask, _mm_cmple_epu32_mask,
37082         _mm_cmple_epu64_mask, _mm_cmplt_epi32_mask, _mm_cmplt_epi64_mask,
37083         _mm_cmplt_epu32_mask, _mm_cmplt_epu64_mask, _mm_cmpneq_epi32_mask,
37084         _mm_cmpneq_epi64_mask, _mm_cmpneq_epu32_mask, _mm_cmpneq_epu64_mask,
37085         _mm_mask_cmpge_epi32_mask, _mm_mask_cmpge_epi64_mask,
37086         _mm_mask_cmpge_epu32_mask, _mm_mask_cmpge_epu64_mask,
37087         _mm_mask_cmple_epi32_mask, _mm_mask_cmple_epi64_mask,
37088         _mm_mask_cmple_epu32_mask, _mm_mask_cmple_epu64_mask,
37089         _mm_mask_cmplt_epi32_mask, _mm_mask_cmplt_epi64_mask,
37090         _mm_mask_cmplt_epu32_mask, _mm_mask_cmplt_epu64_mask,
37091         _mm_mask_cmpneq_epi32_mask, _mm_mask_cmpneq_epi64_mask,
37092         _mm_mask_cmpneq_epu32_mask, _mm_mask_cmpneq_epu64_mask): Move
37093         definitions outside of __OPTIMIZE__ guarded section.
37095         PR target/79932
37096         * config/i386/avx512bwintrin.h (_mm512_packs_epi32,
37097         _mm512_maskz_packs_epi32, _mm512_mask_packs_epi32,
37098         _mm512_packus_epi32, _mm512_maskz_packus_epi32,
37099         _mm512_mask_packus_epi32): Move definitions outside of __OPTIMIZE__
37100         guarded section.
37102 2017-03-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
37104         * config/s390/vx-builtins.md ("vfee<mode>", "vfeez<mode>")
37105         ("vfenez<mode>"): Add missing constraints.
37107 2017-03-08  Martin Sebor  <msebor@redhat.com>
37109         PR target/79928
37110         * config/nds32/nds32.c (nds32_option_override):
37111         Fix misspelled diagnostic.
37113 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
37115         PR c/79940
37116         * gimplify.c (gimplify_omp_for): Replace index var in outer
37117         taskloop statement with an artificial variable and add
37118         OMP_CLAUSE_PRIVATE clause for it.
37120 2017-03-08  Richard Biener  <rguenther@suse.de>
37122         PR tree-optimization/79955
37123         * tree-ssa-uninit.c (warn_uninitialized_vars): Do not warn
37124         for accesses that are completely outside of the variable.
37126 2017-03-08  Andrew Haley  <aph@redhat.com>
37128         PR tree-optimization/79943
37129         * tree-ssa-loop-split.c (compute_new_first_bound): When
37130         calculating the new upper bound, (END-BEG) should be added, not
37131         subtracted.
37133 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
37135         * config/avr/avr.md (setmemhi): Make sure match_dup
37136         operand number comes before match_scratch.
37138 2017-03-08  Richard Biener  <rguenther@suse.de>
37140         PR tree-optimization/79920
37141         * tree-vect-slp.c (vect_create_mask_and_perm): Remove and inline
37142         with ncopies == 1 to ...
37143         (vect_transform_slp_perm_load): ... here.  Properly compute
37144         all element loads by iterating VF times over the group.  Do
37145         not handle ncopies (computed in a broken way) in
37146         vect_create_mask_and_perm.
37148 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
37150         PR sanitizer/79904
37151         * internal-fn.c (expand_vector_ubsan_overflow): If arg0 or arg1
37152         is a uniform vector, use uniform_vector_p return value instead of
37153         building ARRAY_REF on folded VIEW_CONVERT_EXPR to array type.
37155 2017-03-07  Marek Polacek  <polacek@redhat.com>
37157         PR middle-end/79809
37158         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Use HOST_WIDE_INT.
37159         (alloca_call_type): Likewise.
37161 2017-03-07  Martin Liska  <mliska@suse.cz>
37163         * gcov.c (process_args): Put comment to correct location.
37165 2017-03-07  Martin Liska  <mliska@suse.cz>
37167         PR middle-end/68270
37168         * tree-chkp.c (chkp_may_narrow_to_field): Add new argument ref.
37169         Use array_at_struct_end_p instead of DECL_CHAIN (field).
37170         (chkp_narrow_bounds_for_field): Likewise.
37171         (chkp_parse_array_and_component_ref): Pass one more argument to
37172         call.
37174 2017-03-07  Richard Biener  <rguenther@suse.de>
37176         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Preserve
37177         preheaders.
37179 2017-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
37181         * config/i386/i386.c (ix86_local_alignment): Align most aggregates
37182         of 16 bytes and more to 16 bytes, not those of 16 bits and more.
37184 2017-03-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
37186         PR c/79855
37187         * params.def (PARAM_STORE_MERGING_ALLOW_UNALIGNED): Add full stop
37188         to end of description.
37189         (PARAM_MAX_STORES_TO_MERGE): Likewise.
37191 2017-03-07  Jakub Jelinek  <jakub@redhat.com>
37193         PR rtl-optimization/79901
37194         * config/i386/sse.md (*avx512bw_<code><mode>3<mask_name>): Renamed to
37195         ...
37196         (*avx512f_<code><mode>3<mask_name>): ... this.
37197         (<code><mode>3 with maxmin code iterator): Use VI8_AVX2_AVX512F
37198         iterator instead of VI8_AVX2_AVX512BW.
37200         PR rtl-optimization/79901
37201         * expr.c (expand_expr_real_2): For vector MIN/MAX, if there is no
37202         min/max expander, expand it using expand_vec_cond_expr.
37204         PR sanitizer/79897
37205         * ubsan.c (ubsan_encode_value): Call mark_addressable on the
37206         temporary.
37208 2017-03-06  Jakub Jelinek  <jakub@redhat.com>
37210         PR c++/79821
37211         * dwarf2out.h (dw_vec_const): Change array type from unsigned char *
37212         to void * for PCH reasons.
37213         * dwarf2out.c (output_loc_operands, output_die): Cast
37214         v.val_vec.array to unsigned char *.
37216 2017-03-06  John David Anglin  <danglin@gcc.gnu.org>
37218         PR target/77850
37219         * config/pa/pa-64.h (PAD_VARARGS_DOWN): Don't pad down complex and
37220         vector types.
37222 2017-03-06  Vladimir Makarov  <vmakarov@redhat.com>
37224         PR rtl-optimization/79571
37225         * lra-constraints.c (process_alt_operands): Calculate static
37226         reject and subtract it from overall when only addresses will be
37227         reloaded.
37229 2017-03-06  Julia Koval  <julia.koval@intel.com>
37231         PR target/79793
37232         * config/i386/i386.c (ix86_minimum_incoming_stack_boundary): Set
37233         incoming stack boundary to 128 for 64-bit targets.
37235 2017-03-06  Richard Biener  <rguenther@suse.de>
37237         PR tree-optimization/79894
37238         * tree-vectorizer.c (vectorize_loops): Set loop_vectorized_call
37239         to NULL after folding it.
37241 2017-03-06  Richard Biener  <rguenther@suse.de>
37243         PR tree-optimization/79824
37244         * tree-vect-stmts.c (get_group_load_store_type): Fix alignment
37245         check disabling peeling for gaps.
37247 2017-03-06  Toma Tabacu  <toma.tabacu@imgtec.com>
37249         * doc/sourcebuild.texi (Effective-Target Keywords, Environment
37250         attributes): Document gettimeofday.
37252 2017-03-06  Robin Dapp  <rdapp@linux.vnet.ibm.com>
37254         * config/s390/s390.c (s390_option_override_internal): Set
37255         PARAM_MIN_VECT_LOOP_BOUND
37257 2017-03-06  Robin Dapp  <rdapp@linux.vnet.ibm.com>
37259         * config/s390/s390.c (s390_asm_output_function_label): Use nopr %r0.
37260         * config/s390/s390.md: Likewise.
37262 2017-03-06  Jakub Jelinek  <jakub@redhat.com>
37264         PR target/79812
37265         * config/i386/sse.md (VI8F_256_512): Remove mode iterator.
37266         (<avx2_avx512>_perm<mode>): Rename to ...
37267         (avx2_perm<mode>): ... this.  Use VI8F_256 iterator instead
37268         of VI8F_256_512.
37269         (<avx512>_perm<mode>_mask): Rename to ...
37270         (avx512vl_perm<mode>_mask): ... this.  Use VI8F_256 iterator instead
37271         of VI8F_256_512.
37272         (<avx2_avx512>_perm<mode>_1<mask_name>): Rename to ...
37273         (avx2_perm<mode>_1<mask_name): ... this.  Use VI8F_256 iterator
37274         instead of VI8F_256_512.
37275         (avx512f_perm<mode>): New define_expand.
37276         (avx512f_perm<mode>_mask): Likewise.
37277         (avx512f_perm<mode>_1<mask_name>): New define_insn.
37278         (<avx512>_vec_dup<mode>_1): Fix up vec_select mode.
37280 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37282         * config/mips/mips-msa.md (msa_fmax_a_<msafmt>, msa_fmin_a_<msafmt>,
37283         msa_max_a_<msafmt>, msa_min_a_<msafmt>): Introduce mode interator for
37284         if_then_else.
37285         (smin<mode>3, smax<mode>3): Change operand print code from 'B' to 'E'.
37287 2017-03-06  Martin Liska  <mliska@suse.cz>
37289         PR sanitize/79783
37290         * asan.c (asan_expand_poison_ifn): Do not expand ASAN_POISON
37291         when having a SSA NAME w/o VAR_DECL assigned to it.
37293 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37295         * config/mips/mips-msa.md (msa_dotp_<su>_d, msa_dpadd_<su>_d,
37296         msa_dpsub_<su>_d): Fix MODE for vec_select.
37298 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37300         * config/mips/mips.c (mips_gen_const_int_vector): Change type of last
37301         argument.
37302         * config/mips/mips-protos.h (mips_gen_const_int_vector): Likewise.
37304 2017-03-06  Richard Biener  <rguenther@suse.de>
37306         * lto-streamer.c (lto_check_version): Use %qs in diagnostics.
37307         * plugin.c (register_plugin_info): Likewise.
37308         * tree-chkp.c (chkp_make_static_const_bounds): Likewise.
37310 2017-03-05  Jakub Jelinek  <jakub@redhat.com>
37312         * config/i386/sse.md (sse_storehps, sse_storelps,
37313         avx_<castmode><avxsizesuffix>_<castmode>,
37314         avx512f_<castmode><avxsizesuffix>_<castmode>,
37315         avx512f_<castmode><avxsizesuffix>_256<castmode>): Require
37316         in condition that at least one operand is not a MEM.
37318 2017-03-03  Jakub Jelinek  <jakub@redhat.com>
37320         PR middle-end/79805
37321         * internal-fn.def (ATOMIC_BIT_TEST_AND_SET, ATOMIC_BIT_TEST_AND_RESET,
37322         ATOMIC_BIT_TEST_AND_COMPLEMENT, ATOMIC_COMPARE_EXCHANGE): Remove
37323         ECF_NOTHROW.
37324         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Set
37325         gimple_call_nothrow_p flag based on whether original builtin can throw.
37326         If it can, emit following stmts on the fallthrough edge.
37327         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Similarly, except
37328         don't create new bb if inserting just debug stmts on the edge, try to
37329         insert them on the fallthru bb or just reset debug stmts.
37331 2017-03-03  Segher Boesssenkool  <segher@kernel.crashing.org>
37333         PR target/43763
37334         * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Save and
37335         restore recog_data (including the operand rtxes inside it) around
37336         the call to get_insn_template.
37338 2017-03-03  Martin Sebor  <msebor@redhat.com>
37340         PR tree-optimization/79699
37341         * context.c (context::~context): Free MPFR caches to avoid
37342         a memory leak on program exit.
37344 2017-03-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
37346         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
37347         Use wide_int::ulow () instead of .elt (0).
37349 2017-03-03  Uros Bizjak  <ubizjak@gmail.com>
37351         * config/i386/i386.md (*pushtf): Change *roF constraint to *roC.
37352         (*pushxf): Limit oF constraint to 32bit targets and add oC
37353         constraint for 64bit targets.
37354         (pushxf splitter): Use PUSH_ROUNDING to calculate stack adjustment.
37355         (*pushdf): Change rmF constraint to rmC.
37357 2017-03-03  Martin Liska  <mliska@suse.cz>
37359         * tree-ssa-loop-prefetch.c (pass_loop_prefetch::execute):
37360         Remove unused variable.
37362 2017-03-03  Jakub Jelinek  <jakub@redhat.com>
37364         PR target/79807
37365         * config/i386/i386.c (ix86_expand_multi_arg_builtin): If target
37366         is a memory operand, increase num_memory.
37367         (ix86_expand_args_builtin): Likewise.
37369 2017-03-03  Jan Hubicka  <jh@suse.cz>
37371         PR lto/79760
37372         * ipa-devirt.c (maybe_record_node): Properly handle
37373         __cxa_pure_virtual visibility.
37375 2017-03-03  Martin Liska  <mliska@suse.cz>
37377         PR tree-optimization/79803
37378         * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Remove
37379         assert.
37380         (pass_loop_prefetch::execute): Disabled optimization if an
37381         assumption about L1 cache size is not met.
37383 2017-03-03  Martin Liska  <mliska@suse.cz>
37385         PR rtl-optimization/79574
37386         * gcse.c (struct gcse_expr): Use HOST_WIDE_INT instead of int.
37387         (hash_scan_set): Likewise.
37388         (dump_hash_table): Likewise.
37389         (hoist_code): Likewise.
37391 2017-03-03  Richard Biener  <rguenther@suse.de>
37393         * fixed-value.c (fixed_from_string): Restore use of elt (1)
37394         in place of uhigh ().
37395         (fixed_convert_from_real): Likewise.
37397 2017-03-03  Uros Bizjak  <ubizjak@gmail.com>
37399         PR target/79514
37400         * config/i386/i386.md (*pushxf_rounded): Use Pmode instead of DImode.
37402 2017-03-03  Richard Biener  <rguenther@suse.de>
37404         PR middle-end/79818
37405         * match.pd ( X +- C1 CMP C2 -> X CMP C2 -+ C1): Add missing
37406         TYPE_OVERFLOW_UNDEFINED check.
37408 2017-03-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
37410         * config/rs6000/vector.md (vector_ne_<mode>_p): Correct operand
37411         numbers.
37412         (vector_ae_<mode>_p): Likewise.
37413         (vector_nez_<mode>_p): Likewise.
37414         (vector_ne_v2di_p): Likewise.
37415         (vector_ae_v2di_p): Likewise.
37416         (vector_ne_<mode>_p): Likewise.
37417         * config/rs6000/vsx.md (vsx_tsqrt<mode>2_fg): Correct operand
37418         numbers.
37419         (vsx_tsqrt<mode>2_fe): Likewise.
37421 2017-03-02  Uros Bizjak  <ubizjak@gmail.com>
37423         PR target/79514
37424         * config/i386/i386.md (*pushxf_rounded): New insn_and_split pattern.
37426 2017-03-02  Jakub Jelinek  <jakub@redhat.com>
37428         PR rtl-optimization/79780
37429         * cprop.c (one_cprop_pass): When second and further conditional trap
37430         in a single basic block is turned into an unconditional trap, turn it
37431         into a deleted note to avoid RTL verification failures.
37433 2017-03-02  Richard Biener  <rguenther@suse.de>
37435         * fold-const.c (const_binop): Use ulow () instead of elt (0).
37437 2017-03-02  Richard Biener  <rguenther@suse.de>
37439         PR tree-optimization/79345
37440         PR c++/42000
37441         * tree-ssa-alias.c (walk_aliased_vdefs_1): Take a limit
37442         param and abort the walk, returning -1 if it is hit.
37443         (walk_aliased_vdefs): Take a limit param and pass it on.
37444         * tree-ssa-alias.h (walk_aliased_vdefs): Add a limit param,
37445         defaulting to 0 and return a signed int.
37446         * tree-ssa-uninit.c (struct check_defs_data): New struct.
37447         (check_defs): New helper.
37448         (warn_uninitialized_vars): Use walk_aliased_vdefs to warn
37449         about uninitialized memory.
37450         * fixed-value.c (fixed_from_string): Use ulow/uhigh to avoid
37451         bogus uninitialized warning.
37452         (fixed_convert_from_real): Likewise.
37454 2017-03-02  Bin Cheng  <bin.cheng@arm.com>
37456         PR tree-optimization/66768
37457         * tree-ssa-loop-ivopts.c (find_interesting_uses_address): Skip addr
37458         iv_use if base object can't be determined.
37460 2017-03-02  Jakub Jelinek  <jakub@redhat.com>
37462         PR tree-optimization/79345
37463         * gensupport.h (struct pattern_stats): Add min_scratch_opno field.
37464         * gensupport.c (get_pattern_stats_1) <case MATCH_SCRATCH>: Update it.
37465         (get_pattern_stats): Initialize it.
37466         * genemit.c (gen_expand): Verify match_scratch numbers come after
37467         match_operand/match_dup numbers.
37468         * config/i386/i386.md (<s>mul<mode>3_highpart): Swap match_dup and
37469         match_scratch numbers.
37470         * config/i386/sse.md (avx2_gathersi<mode>, avx2_gatherdi<mode>):
37471         Likewise.
37472         * config/s390/s390.md (trunctdsd2): Likewise.
37474 2017-03-02  Richard Biener  <rguenther@suse.de>
37476         * wide-int.h (wide_int_storage::operator=): Implement in terms
37477         of wi::copy.
37479 2017-03-02  Richard Biener  <rguenther@suse.de>
37481         PR tree-optimization/79777
37482         * tree-ssa-pre.c (eliminate_insert): Give up if we simplify
37483         the to insert expression to sth existing.
37485 2017-03-01  Martin Sebor  <msebor@redhat.com>
37487         PR middle-end/79692
37488         * gimple-ssa-sprintf.c
37489         (directive::known_width_and_precision): New function.
37490         (format_integer): Use it.
37491         (get_mpfr_format_length): Consider the full range of precision
37492         when computing %g output with the # flag.  Set the likely byte
37493         count to 3 rather than 1 when precision is indeterminate.
37494         (format_floating): Correct the lower bound of precision.
37496 2017-03-01  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
37498         * doc/invoke.texi: Document default code model for 64-bit Linux.
37500 2017-03-01  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
37502         PR target/79752
37503         * config/rs6000/rs6000.md (peephole2 for udiv/umod): Should emit
37504         udiv rather than div since input pattern is unsigned.
37506 2017-03-01  Uros Bizjak  <ubizjak@gmail.com>
37508         * config/i386/i386.c (print_reg): Warn for values of
37509         unsupported size in integer register.
37511 2017-03-01  Michael Meissner  <meissner@linux.vnet.ibm.com>
37513         PR target/79439
37514         * config/rs6000/predicates.md (current_file_function_operand): Do
37515         not allow self calls to be local if the function is replaceable.
37517 2017-03-01  Kelvin Nilsen  <kelvin@gcc.gnu.org>
37519         PR target/79395
37520         * config/rs6000/altivec.h (vec_ctz and others): Change the
37521         preprocessor macro that controls conditional compilation from
37522         _ARCH_PWR9 to __POWER9_VECTOR__.
37523         (vec_all_ne): Change parameterization of __altivec_scalar_pred
37524         macro expansion under preprocessor #ifdef __POWER9_VECTOR__
37525         control (instead of _ARCH_PWR9 control) so that template
37526         definition uses power9-specific function.
37527         (vec_any_eq): Likewise.
37528         (vec_all_ne): Change macro definition to use a power9-specific
37529         expansion under #ifdef __POWER9_VECTOR__ control (instead of
37530         _ARCH_PWR9 control).
37531         (vec_any_eq) Likewise.
37532         * config/rs6000/rs6000-builtin.def (CMPNEF): Remove BU_P9V_AV_2
37533         expansion for CMPNEF to remove support for xvcmpnesp instruction.
37534         (CMPNED): Remove BU_P9V_AV2 expansion for CMPNED to remove
37535         support for xvcmpnedp instruction.
37536         (VCMPNEB_P): Replace BU_P9V_AV_P macro expansion with BU_P9V_AV_2
37537         macro expansion so that Power9 implementation of vec_all_ne does
37538         not use the AltiVec predicate framework.
37539         (VCMPNEH_P): Likewise.
37540         (VCMPNEW_P): Likewise.
37541         (VCMPNED_P): Likewise.
37542         (VCMPNEFP_P): Likewise.
37543         (VCMPNEDP_P): Likewise.
37544         (VCMPAEB_P): Add BU_P9V_AV_2 macro expansion to change
37545         implementation of vec_any_eq to not use AltiVec predicate
37546         framework.
37547         (VCMPAEH_P): Likewise.
37548         (VCMPAEW_P): Likewise.
37549         (VCMPAED_P): Likewise.
37550         (VCMPAEFP_P): Likewise.
37551         (VCMPAEDP_P): Likewise.
37552         (VCMPNE_P): Replace BU_P9V_OVERLOAD_P macro expansion with
37553         BU_P9V_OVERLOAD_2 so that Power9 implementation of vec_all_ne does
37554         not use the AltiVec predicate framework.
37555         (VCMPAE_P): Add BU_P9V_OVERLOAD_2 macro to change implementation
37556         of vec_any_eq to not use AltiVec predicate framework.
37557         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Add
37558         support for predefined __POWER9_VECTOR__ macro to indicate that
37559         Power9 instruction selection is enabled.
37560         (altivec_overloaded_builtins): Remove extraneous
37561         ALTIVEC_BUILTIN_VEC_CMPNE entry for overloaded
37562         function argument types RS6000_BTI_bool_V16QI and
37563         RS6000_BTI_bool_V16QI.  Remove erroneous ALTIVEC_BUILTIN_VEC_CMPNE
37564         entry for overloaded function argument types RS6000_BTI_bool_V4SI
37565         andRS6000_BTI_bool_V4SI, mapping to P9V_BUILTIN_CMPNEB.  Remove
37566         two entries mapping to P9V_BUITIN_CMPNED and one entry mapping to
37567         P9V_BUILTIN_CMPNEF to force use of instructions not specific to
37568         Power9 for implementations of vec_cmpne.  Change the signature for
37569         all definitions of the overloaded P9V_BUILTIN_VEC_CMPNE_P function
37570         (representing vec_all_ne) to remove the previously described first
37571         argument of type RS6000_BTI_INTSI, as this was an artifact of
37572         reliance on the AltiVec predicate framework, which is no longer
37573         used in the implementation of these functions.  Add
37574         P9V_BUILTIN_VEC_VCMPAE_P entries (representing the vec_anyeq
37575         function) to match all of the P9V_BUILTIN_VEC_VCMNE_P entries
37576         since, unlike the AltiVec predicate framework implementation, we
37577         do not share function descriptors between vec_alle and vec_anyeq.
37578         (altivec_resolve_overloaded_builtin): Add SFmode and DFmode to the
37579         set of modes that receive special treatment even when
37580         TARGET_P9_VECTOR is true.  The special treatment emits code that
37581         does not depend on Power9 instructions.
37582         * config/rs6000/vector.md (vector_ne_<mode>_p): Change this
37583         define_expand to not rely on AltiVec predicate framework.
37584         (vector_ae_<mode>p): New define_expand to represent vec_any_eq
37585         function.
37586         (vector_ne_v2di_p): Change this define_expand to not rely on
37587         AltiVec predicate framework.
37588         (vector_ae_v2di_p): New define_expand to represent vec_any_eq
37589         function.
37590         (vector_ne_<mode>_p): Change this define_expand to not rely on
37591         AltiVec predicate framework.
37592         (vector_ae_<mode>p): New define_expand to represent vec_any_eq
37593         function.
37594         * config/rs6000/vsx.md (*vsx_ne_<mode>_p): For modes VSX_EXTRACT_I
37595         (V16QI, V8HI, V4SI), correct a typo in the code emitted for this
37596         define_insn pattern.
37597         (*vsx_ne_<mode>_p): For modes VSX_F (V4SF and V2DF), remove this
37598         define_insn pattern because the xvcmpne<VSs>. instruction is not
37599         supported.
37600         (vcmpne<VSs>): Remove this define_insn because xvcmpne<VSs>
37601         instruction is not supported.
37603 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37605         * config/nvptx/nvptx.c: Include intl.h.
37607 2017-03-01  Martin Jambor  <mjambor@suse.cz>
37609         PR lto/78140
37610         * ipa-prop.h (ipa_bits): Removed field known.
37611         (ipa_jump_func): Removed field vr_known.  Changed fields bits and m_vr
37612         to pointers.  Adjusted their comments to warn about their sharing.
37613         (ipcp_transformation_summary): Change bits to a vector of pointers.
37614         (ipa_check_create_edge_args): Moved to ipa-prop.c, declare.
37615         (ipa_get_ipa_bits_for_value): Declare.
37616         * tree-vrp.h (value_range): Mark as GTY((for_user)).
37617         * ipa-prop.c (ipa_bit_ggc_hash_traits): New.
37618         (ipa_bits_hash_table): Likewise.
37619         (ipa_vr_ggc_hash_traits): Likewise.
37620         (ipa_vr_hash_table): Likewise.
37621         (ipa_print_node_jump_functions_for_edge): Adjust for bits and m_vr
37622         being pointers and vr_known being removed.
37623         (ipa_set_jf_unknown): Likewise.
37624         (ipa_get_ipa_bits_for_value): New function.
37625         (ipa_set_jfunc_bits): Likewise.
37626         (ipa_get_value_range): New overloaded functions.
37627         (ipa_set_jfunc_vr): Likewise.
37628         (ipa_compute_jump_functions_for_edge): Use the above functions to
37629         construct bits and vr parts of jump functions.
37630         (ipa_check_create_edge_args): Move here from ipa-prop.h, also allocate
37631         ipa_bits_hash_table and ipa_vr_hash_table if they do not already
37632         exist.
37633         (ipcp_grow_transformations_if_necessary): Also allocate
37634         ipa_bits_hash_table and ipa_vr_hash_table if they do not already
37635         exist.
37636         (ipa_node_params_t::duplicate): Do not copy bits, just pointers to
37637         them.  Fix too long lines.
37638         (ipa_write_jump_function): Adjust for bits and m_vr being pointers and
37639         vr_known being removed.
37640         (ipa_read_jump_function): Use new setter functions to construct bits
37641         and vr parts of jump functions or set them to NULL.
37642         (write_ipcp_transformation_info): Adjust for bits being pointers.
37643         (read_ipcp_transformation_info): Likewise.
37644         (ipcp_update_bits): Likewise.  Fix excessively long lines a trailing
37645         space.
37646         Include gt-ipa-prop.h.
37647         * ipa-cp.c (propagate_bits_across_jump_function): Adjust for bits
37648         being pointers.
37649         (ipcp_store_bits_results): Likewise.
37650         (propagate_vr_across_jump_function): Adjust for m_vr being a pointer.
37651         Do not write to existing jump functions but use a temporary instead.
37653 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37655         PR c++/79681
37656         * fold-const.c (make_bit_field_ref): If orig_inner is COMPONENT_REF,
37657         attempt to use its first operand as BIT_FIELD_REF base.
37659 2017-03-01  Richard Biener  <rguenther@suse.de>
37661         PR middle-end/79721
37662         * tree-chrec.c (chrec_evaluate): Perform computation of Newtons
37663         interpolating formula in wrapping arithmetic.
37664         (chrec_apply): Convert chrec_evaluate return value to wanted type.
37666 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37668         PR tree-optimization/79734
37669         * tree-vect-generic.c (expand_vector_condition): Optimize
37670         AVX512 vector boolean VEC_COND_EXPRs into bitwise operations.
37671         Handle VEC_COND_EXPR where comparison has different inner width from
37672         type's inner width.
37674 2017-02-28  Sandra Loosemore  <sandra@codesourcery.com>
37676         * doc/invoke.texi (ARC Options): Copy-edit to fix punctuation,
37677         markup, and similar issues.  Remove @opindex entries for things
37678         that aren't options.  Add missing -mmpy-option entries.
37680 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37682         PR tree-optimization/79737
37683         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): If bitlen is
37684         a multiple of BITS_PER_UNIT and !BYTES_BIG_ENDIAN, clear
37685         tmpbuf[byte_size - 1].  Call natice_encode_expr with byte_size - 1
37686         instead of byte_size.  Formatting fix.
37687         (shift_bytes_in_array_right): Formatting fix.
37689 2017-02-28  Eric Botcazou  <ebotcazou@adacore.com>
37691         PR target/79749
37692         * config/sparc/sparc.c (sparc_frame_pointer_required): Add missing
37693         condition on optimize for the leaf function test.
37695 2017-02-28  Martin Liska  <mliska@suse.cz>
37697         PR lto/79625
37698         * read-rtl-function.c (function_reader::handle_unknown_directive):
37699         Bail out when one uses -flto.
37701 2017-02-28  Martin Liska  <mliska@suse.cz>
37703         * common.opt: Replace space with tabular for options of <number>
37704         type.
37705         * config/i386/i386.opt: Show <number> value for
37706         -mlarge-data-threshold.
37707         * opts.c (print_filtered_help): Do not display number in hexadecimal
37708         format.
37710 2017-02-28  Martin Liska  <mliska@suse.cz>
37712         * common.opt: Fix --help=option -Q for options which are of
37713         an enum type.
37715 2017-02-28  Uros Bizjak  <ubizjak@gmail.com>
37717         * config/i386/i386.c (print_reg): Error out for values
37718         of 8-bit size in invalid integer register.
37720 2017-02-28  Martin Sebor  <msebor@redhat.com>
37722         PR tree-optimization/79691
37723         * passes.def (pass_all_optimizations_g): Enable pass_sprintf_length.
37725 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37727         PR target/79729
37728         * config/i386/i386.c (ix86_print_operand) <case 'R'>: Replace
37729         gcc_unreachable with output_operand_lossage.
37731 2017-02-28  Richard Biener  <rguenther@suse.de>
37733         PR tree-optimization/79740
37734         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Allow redundant
37735         inserts.
37736         (visit_nary_op): Insert the nary into the hashtable if we
37737         pattern-matched sth.
37738         * tree-ssa-pre.c (eliminate_insert): Robustify.
37740 2017-02-28  Richard Biener  <rguenther@suse.de>
37742         PR middle-end/79731
37743         * fold-const.c (decode_field_reference): Reject out-of-bound
37744         accesses.
37746 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37748         * config/i386/i386.c: Include intl.h.
37749         (ix86_option_override_internal): Use cond ? G_("...") : G_("...")
37750         instead of just cond ? "..." : "...".
37751         * config/nvptx/nvptx.c (nvptx_goacc_validate_dims): Likewise.
37752         * coverage.c (read_counts_file): Likewise.
37753         * omp-offload.c: Include intl.h.
37754         (oacc_loop_fixed_partitions): Use cond ? G_("...") : G_("...") instead
37755         of just cond ? "..." : "...".
37756         * gcov.c (read_count_file): Use cond ? N_("...") : N_("...") instead
37757         of just cond ? "..." : "...".
37759 2017-02-28  Richard Earnshaw  <rearnsha@arm.com>
37761         PR target/79742
37762         * config/arm/parsecpu.awk (gen_data): Set tuning target to 'tune for'
37763         entry, if present.
37764         * config/arm/arm-cpus.in (cortex-m0plus.small-multiply): Correct
37765         'tune for' CPU name.
37766         * config/arm/arm-cpu-data.h: Regenerated.
37768 2017-02-28  Richard Biener  <rguenther@suse.de>
37770         PR tree-optimization/79732
37771         * tree-inline.c (expand_call_inline): Do not shadow var.
37773 2017-02-28  Richard Biener  <rguenther@suse.de>
37775         PR tree-optimization/79723
37776         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Preserve
37777         address-space properly.
37779 2017-02-28  Thomas Schwinge  <thomas@codesourcery.com>
37781         * doc/optinfo.texi (Optimization groups): Fix option used for
37782         OPTGROUP_ALL.
37783         * doc/invoke.texi (-fopt-info): Document "omp".
37784         * dumpfile.h: Sort OPTGROUP_OMP before OPTGROUP_VEC.
37785         (OPTGROUP_ALL): Add OPTGROUP_OMP.
37786         * hsa-gen.c (pass_data_gen_hsail): Use OPTGROUP_OMP.
37787         * ipa-hsa.c (pass_data_ipa_hsa): Likewise.
37788         * omp-simd-clone.c (pass_data_omp_simd_clone): Likewise.
37790         * dumpfile.h (OPTGROUP_OPENMP): Rename to OPTGROUP_OMP.  Adjust
37791         all users.
37792         * dumpfile.c (optgroup_options): Instead of "openmp", associate
37793         OPTGROUP_OMP with "omp".
37795 2017-02-27  Pat Haugen  <pthaugen@us.ibm.com>
37797         PR target/79544
37798         * config/rs6000/rs6000-c.c (struct altivec_builtin_types): Use VSRAD
37799         for arithmetic shift of unsigned V2DI.
37801 2017-02-27  Claudiu Zissulescu  <claziss@synopsys.com>
37803         * config.gcc (arc*-): Clean up, use arc/big.h, arc/elf.h, and
37804         arc/linux.h headers.
37805         * config/arc/arc.h (TARGET_OS_CPP_BUILTINS): Remove.
37806         (LINK_SPEC): Likewise.
37807         (ARC_TLS_EXTRA_START_SPEC): Likewise.
37808         (EXTRA_SPECS): Likewise.
37809         (STARTFILE_SPEC): Likewise.
37810         (ENDFILE_SPEC): Likewise.
37811         (LIB_SPEC): Likewise.
37812         (TARGET_SDATA_DEFAULT): Likewise.
37813         (TARGET_MMEDIUM_CALLS_DEFAULT): Likewise.
37814         (MULTILIB_DEFAULTS): Likewise.
37815         (DWARF2_UNWIND_INFO): Likewise.
37816         * config/arc/big.h: New file.
37817         * config/arc/elf.h: Likewise.
37818         * config/arc/linux.h: Likewise.
37819         * config/arc/t-uClibc: Remove.
37821 2017-02-27  Bin Cheng  <bin.cheng@arm.com>
37823         PR tree-optimization/77536
37824         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): New function.
37825         (tree_transform_and_unroll_loop): Use above function to compute the
37826         estimated niter of unrolled loop and use it when scaling profile.
37827         Also use count info rather than frequency if it's non-zero.
37828         * tree-ssa-loop-manip.h niter_for_unrolled_loop(): New declaration.
37829         * tree-vect-loop.c (scale_profile_for_vect_loop): New function.
37830         (vect_transform_loop): Call above function.
37832 2017-02-27  Richard Biener  <rguenther@suse.de>
37834         PR tree-optimization/45397
37835         * tree-ssa-pre.c (eliminate_insert): Handle BIT_AND_EXPR.
37836         * tree-ssa-sccvn.c (valueized_wider_op): New helper.
37837         (visit_nary_op): Add pattern matching for CSEing sign-changed
37838         or truncated operations with wider ones.
37840 2017-02-27  Richard Biener  <rguenther@suse.de>
37842         PR tree-optimization/79690
37843         * tree-vect-stmts.c (vectorizable_store): Use vector type
37844         built from the DR with address-space.
37846 2017-02-26  Gerald Pfeifer  <gerald@pfeifer.com>
37848         * doc/invoke.texi (Optimize Options): Refine the description
37849         of asan-use-after-return.
37851 2017-02-25  Alan Modra  <amodra@gmail.com>
37853         PR rtl-optimization/79584
37854         * lra-constraints.c (base_to_reg): Reload ad->base, the entire
37855         base, not ad->base_term, the reg within base.  Remove assertion
37856         that ad->base == ad->base_term.  Replace gen_int_mode using
37857         bogus mode with const0_rtx.
37859 2017-02-25  Jakub Jelinek  <jakub@redhat.com>
37861         PR middle-end/79396
37862         * tree-eh.c (operation_could_trap_p, stmt_could_throw_1_p): Handle
37863         FMA_EXPR like tcc_binary or tcc_unary.
37865         * tree-ssa-loop-niter.c (number_of_iterations_exit): Simplify warning.
37867         PR debug/77589
37868         * dwarf2out.c (struct dw_loc_list_struct): Add noted_variable_value
37869         bitfield.
37870         (size_of_loc_descr): Handle DW_OP_GNU_variable_value.
37871         (output_loc_operands): Handle DW_OP_call_ref and
37872         DW_OP_GNU_variable_value.
37873         (struct variable_value_struct): New type.
37874         (struct variable_value_hasher): Likewise.
37875         (variable_value_hash): New variable.
37876         (string_types): Remove.
37877         (copy_loc_descr): New function.
37878         (add_loc_descr_to_each): Clarify comment.  Use copy_loc_descr.
37879         (prepend_loc_descr_to_each): New function.
37880         (add_loc_list): Fix comment typo.  Use prepend_loc_descr_to_each
37881         instead of add_loc_descr_to_each if the first argument is single
37882         location list and the second has multiple.
37883         (resolve_args_picking_1): Handle DW_OP_GNU_variable_value.
37884         (loc_list_from_tree_1): For early_dwarf, emit DW_OP_GNU_variable_value
37885         when looking for variable value which doesn't have other location info.
37886         (loc_list_from_tree): Formatting fix.
37887         (gen_array_type_die): Simplify DW_AT_string_length handling.
37888         (adjust_string_types): Remove.
37889         (gen_subprogram_die): Don't call adjust_string_types nor test/set
37890         string_types.  Call resolve_variable_values.
37891         (prune_unused_types_walk_loc_descr): Handle DW_OP_GNU_variable_value.
37892         (resolve_addr_in_expr): Likewise.  Add A argument.
37893         (copy_deref_exprloc): Remove deref argument.  Adjust for the
37894         original expression being DW_OP_GNU_variable_value with optionally
37895         DW_OP_stack_value after it instead of DW_OP_call4 with DW_OP_deref
37896         optionally after it.
37897         (optimize_string_length): Rework for DW_OP_GNU_variable_value.
37898         (resolve_addr): Adjust optimize_string_length and resolve_addr_in_expr
37899         callers.  Set remove_AT_byte_size if removing DW_AT_string_length.
37900         (variable_value_hasher::hash, variable_value_hasher::equal): New
37901         methods.
37902         (resolve_variable_value_in_expr, resolve_variable_value,
37903         resolve_variable_values, note_variable_value_in_expr,
37904         note_variable_value): New functions.
37905         (dwarf2out_early_finish): Call note_variable_value on all toplevel
37906         DIEs.
37908 2017-02-24  Jakub Jelinek  <jakub@redhat.com>
37910         PR c/79677
37911         * opts.h (handle_generated_option): Add GENERATED_P argument.
37912         * opts-common.c (handle_option): Adjust function comment.
37913         (handle_generated_option): Add GENERATED_P argument, pass it to
37914         handle_option.
37915         (control_warning_option): Pass false to handle_generated_option
37916         GENERATED_P.
37917         * opts.c (maybe_default_option): Pass true to handle_generated_option
37918         GENERATED_P.
37919         * optc-gen.awk: Likewise.
37921 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37923         * config/sh/sh.md (tstsi_t): If operands[0] is a SUBREG instead of
37924         a REG, look at the REG it is a SUBREG of.
37925         (splitter for cmpeqsi_t): Ditto.
37927 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37929         * config/pa/pa.c (pa_combine_instructions): Do not share RTL.  Make
37930         the special USEs with the pattern of the insn, not the insn itself.
37932 2017-02-24  Matthew Fortune  <matthew.fortune@imgtec.com>
37934         PR target/79473
37935         * doc/invoke.texi: Document -mload-store-pairs.
37937 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37938             Sandra Loosemore  <sandra@codesourcery.com>
37940         * config/nios2/nios2.c (nios2_simple_const_p): Returns false if the
37941         argument isn't a CONST_INT.
37942         (nios2_alternate_compare_const): Assert op is a CONST_INT.
37943         (nios2_valid_compare_const_p): Assert op is a CONST_INT.
37944         (nios2_validate_compare): Bypass alternate compare logic if *op2
37945         is not a CONST_INT.
37946         (ldstwm_operation_p): Return false if first_base is not a REG or
37947         if first_offset is not a CONST_INT.
37949 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37951         * config/cris/cris.md: Use correct operand in a define_peephole2.
37953 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37955         * config/c6x/c6x.c (predicate_insn): Do not incorrectly share RTL.
37957 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37959         * config/arc/arc.c (arc_ccfsm_advance): Only take the PATTERN of
37960         this_insn if it is an INSN or JUMP_INSN.
37961         (force_offsettable): Look at base, not at addr.
37962         * config/arc/predicates.md (brcc_nolimm_operator): Don't call INTVAL
37963         on things that aren't necessarily CONST_INTs.
37965 2017-02-24  Uros Bizjak  <ubizjak@gmail.com>
37967         * doc/invoke.texi (x86 Options, -mfpmath=sse): Mention that
37968         -mfpmath=sse is the default also for x86-32 targets with SSE2
37969         instruction set when @option{-ffast-math} is enabled
37971 2017-02-24  Jeff Law  <law@redhat.com>
37973         PR rtl-optimizatoin/79286
37974         * ira.c (update_equiv_regs): Drop may_trap_p exception to
37975         dominance test.
37977 2017-02-24  Richard Biener  <rguenther@suse.de>
37979         PR tree-optimization/79389
37980         * gimple-ssa-split-paths.c (is_feasible_trace): Properly skip
37981         debug insns.
37983 2017-02-24  Aldy Hernandez  <aldyh@redhat.com>
37985         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update
37986         function comment to reflect reality.
37987         (loop_exits_before_overflow): Fix typo in function description.
37989 2017-02-24  Richard Biener  <rguenther@suse.de>
37991         PR tree-optimization/79389
37992         * gimple-ssa-split-paths.c (is_feasible_trace): Verify more
37993         properly that a threading opportunity exists.  Detect conditional
37994         copy/constant propagation opportunities.
37996 2017-02-23  Eric Botcazou  <ebotcazou@adacore.com>
37998         * config/visium/visium.md (type): Add trap.
37999         (b): New mode attribute.
38000         (*btst): Rename into...
38001         (*btst<mode>): ...this and adjust.
38002         (*cbranchsi4_btst_insn): Rename into...
38003         (*cbranch<mode>4_btst_insn): ...this and adjust.
38004         (trap): New define_insn.
38006 2017-02-23  Jakub Jelinek  <jakub@redhat.com>
38008         PR tree-optimization/79389
38009         * ifcvt.c (struct noce_if_info): Add rev_cond field.
38010         (noce_reversed_cond_code): New function.
38011         (noce_emit_store_flag): Use rev_cond if non-NULL instead of
38012         reversed_comparison_code.  Formatting fix.
38013         (noce_try_store_flag): Test rev_cond != NULL in addition to
38014         reversed_comparison_code.
38015         (noce_try_store_flag_constants): Likewise.
38016         (noce_try_store_flag_mask): Likewise.
38017         (noce_try_addcc): Use rev_cond if non-NULL instead of
38018         reversed_comparison_code.
38019         (noce_try_cmove_arith): Likewise.  Formatting fixes.
38020         (noce_try_minmax, noce_try_abs): Clear rev_cond.
38021         (noce_find_if_block): Initialize rev_cond.
38022         (find_cond_trap): Call noce_get_condition with then_bb == trap_bb
38023         instead of false as last argument never attempt to reverse it
38024         afterwards.
38026 2017-02-23  Bin Cheng  <bin.cheng@arm.com>
38028         PR tree-optimization/79663
38029         * tree-predcom.c (combine_chains): Process refs in reverse order
38030         only for ZERO length chains, and add explaining comment.
38032 2017-02-23  Jeff Law  <law@redhat.com>
38034         PR tree-optimization/79578
38035         * tree-ssa-dse.c (clear_bytes_written_by): Use OEP_ADDRESS_OF
38036         in call to operand_equal_p.
38038 2017-01-23  Dominique d'Humieres  <dominiq@lps.ens.fr>
38040         PR target/71017
38041         * config/i386/cpuid.h: Fix another undefined behavior.
38043 2017-02-23  Richard Biener  <rguenther@suse.de>
38045         PR tree-optimization/79683
38046         * tree-vect-stmts.c (vect_analyze_stmt): Do not overwrite
38047         vector types for data-refs.
38049 2017-02-23  Martin Liska  <mliska@suse.cz>
38051         * params.def (PARAM_MIN_NONDEBUG_INSN_UID): Change default to 0.
38053 2017-02-23  Jakub Jelinek  <jakub@redhat.com>
38055         PR middle-end/79665
38056         * internal-fn.c (get_range_pos_neg): Moved to ...
38057         * tree.c (get_range_pos_neg): ... here.  No longer static.
38058         * tree.h (get_range_pos_neg): New prototype.
38059         * expr.c (expand_expr_real_2) <case TRUNC_DIV_EXPR>: If both arguments
38060         are known to be in between 0 and signed maximum inclusive, try to
38061         expand both unsigned and signed divmod and use the cheaper one from
38062         those.
38064 2017-02-22  Jeff Law  <law@redhat.com>
38066         PR tree-optimization/79578
38067         * tree-ssa-dse.c (clear_bytes_written_by): Use operand_equal_p
38068         to compare base operands.
38070 2017-02-22  Segher Boessenkool  <segher@kernel.crashing.org>
38072         PR target/79211
38073         * config/rs6000/rs6000.md (*fsel<SFDF:mode><SFDF2:mode>4): Use
38074         gpc_reg_operand instead of fpr_reg_operand.
38076 2017-02-22  Sameera Deshpande  <sameera.deshpande@imgtec.com>
38078         * config/mips/mips.c (mips_return_in_memory): Force FP
38079         vector types to be returned in memory for o32 ABI.
38081 2017-02-22  Jakub Jelinek  <jakub@redhat.com>
38083         * dwarf2out.c (gen_variable_die): For -gdwarf-5, use DW_TAG_variable
38084         instead of DW_TAG_member for static data member declarations and don't
38085         set no_linkage_name for static inline data members.
38086         (gen_member_die): For -gdwarf-5 don't change DW_TAG_variable
38087         to DW_TAG_member.
38089 2017-02-22  Martin Liska  <mliska@suse.cz>
38091         * doc/invoke.texi: Replace inequality signs with square brackets
38092         for -Wnormalized.
38094 2017-02-22  Matthew Fortune  <matthew.fortune@imgtec.com>
38096         PR target/78660
38097         * lra-constraints.c (simplify_operand_subreg): Handle
38098         WORD_REGISTER_OPERATIONS targets.
38100 2017-02-22  Jakub Jelinek  <jakub@redhat.com>
38102         PR target/70465
38103         * reg-stack.c (emit_swap_insn): Treat (float_extend:?F (mem:?F))
38104         and (const_double:?F) like (mem:?F) for the purpose of fxch %st(1)
38105         elimination by swapping fld*.
38107 2017-02-22  Richard Biener  <rguenther@suse.de>
38109         PR tree-optimization/79673
38110         * tree-ssa-pre.c (compute_avail): Use wide_int_to_tree to
38111         convert the [TARGET_]MEM_REF offset INTEGER_CST, scrapping off
38112         irrelevant address-space qualifiers and avoiding a
38113         ADDR_SPACE_CONVERT_EXPR from fold_convert.
38115 2017-02-22  Richard Biener  <rguenther@suse.de>
38117         PR tree-optimization/79666
38118         * tree-vrp.c (extract_range_from_binary_expr_1): Make sure
38119         to not symbolically negate if that may introduce undefined
38120         overflow.
38122 2017-02-22  Martin Liska  <mliska@suse.cz>
38124         PR lto/79587
38125         * data-streamer-in.c (streamer_read_gcov_count): Remove assert.
38126         * data-streamer-out.c (streamer_write_gcov_count_stream):
38127         Likewise.
38128         * value-prof.c (stream_out_histogram_value): Make assert more
38129         precise based on type of counter.
38131 2017-02-21  Uros Bizjak  <ubizjak@gmail.com>
38133         PR target/79593
38134         * config/i386/i386.md (standard_x87sse_constant_load splitter):
38135         Use nonimmediate_operand instead of memory_operand for operand 1.
38136         (float-extend standard_x87sse_constant_load splitter): Ditto.
38138 2017-02-21  Jeff Law  <law@redhat.com>
38140         PR tree-optimization/79621
38141         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior): Ignore
38142         blocks with edges to themselves.
38144 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
38146         PR target/79633
38147         * tree-chkp-opt.c (chkp_optimize_string_function_calls): Use
38148         is_gimple_call instead of comparing gimple_code with GIMPLE_CALL.
38149         Use gimple_call_builtin_p.
38151         PR target/79570
38152         * sel-sched.c (moveup_expr_cached): Don't call sel_bb_head
38153         on temporarily removed DEBUG_INSNs.
38155         PR tree-optimization/79649
38156         * tree-loop-distribution.c (classify_partition): Give up on
38157         non-generic address space loads/stores.
38159 2017-02-21  Aldy Hernandez  <aldyh@redhat.com>
38161         * doc/loop.texi (Loop manipulation): Remove nonexistent
38162         tree_ssa_loop_version from the documentation.
38163         * cfgloopmanip.c (loop_version): Document CONDITION_BB argument.
38165 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
38167         PR target/79494
38168         * config/i386/i386.c (ix86_expand_split_stack_prologue): Call
38169         make_reg_eh_region_note_nothrow_nononlocal on call_insn.
38170         * config/rs6000/rs6000.c: Include except.h.
38171         (rs6000_expand_split_stack_prologue): Call
38172         make_reg_eh_region_note_nothrow_nononlocal on the call insn.
38174 2017-02-21  Martin Jambor  <mjambor@suse.cz>
38176         PR lto/79579
38177         * ipa-prop.c (ipa_prop_write_jump_functions): Bail out if no edges
38178         have been analyzed.
38180 2017-02-21  Martin Jambor  <mjambor@suse.cz>
38182         * common.opt (-fipa-cp-alignment): Mark as ignored and preserved
38183         for backward compatibility only.
38184         * doc/invoke.texi (Option Summary): Remove all references to
38185         -fipa-cp-alignment.
38187 2017-02-21  Matthew Fortune  <matthew.fortune@imgtec.com>
38189         PR target/78660
38190         Revert:
38191         2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38193         * lra-constraints.c (curr_insn_transform): Handle
38194         WORD_REGISTER_OPERATIONS requirements when reloading SUBREGs.
38196 2017-02-21  Martin Liska  <mliska@suse.cz>
38198         * config/i386/i386.opt: Replace -masm-dialect with -masm.
38200 2017-02-21  Thomas Schwinge  <thomas@codesourcery.com>
38202         PR translation/79638
38203         * config/nvptx/nvptx.c (ENTRY_TEMPLATE): Single out "%ntid.y".
38205 2017-02-21  Eric Botcazou  <ebotcazou@adacore.com>
38207         PR ada/67205
38208         * config/arm/arm.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): Define.
38209         (arm_function_ok_for_sibcall): Return false for an indirect call by
38210         descriptor if all the argument registers are used.
38211         (arm_relayout_function): Use FUNCTION_ALIGNMENT macro to adjust the
38212         alignment of the function.
38214 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
38216         PR tree-optimization/61441
38217         * simplify-rtx.c (simplify_const_unary_operation): For
38218         -fsignaling-nans and sNaN operand, return NULL_RTX rather than
38219         the sNaN unmodified.
38221 2017-02-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
38223         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
38224         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
38225         instead of SYSTEM_HEADER_DIR.
38227 2017-02-20  Gerald Pfeifer  <gerald@pfeifer.com>
38228             Martin LiÅ¡ka  <mliska@suse.cz>
38230         * doc/invoke.texi (use-after-scope-direct-emission-threshold):
38231         Fix typos and grammar, use active voice, and clarify.
38233 2017-02-20  Marek Polacek  <polacek@redhat.com>
38235         PR middle-end/79537
38236         * gimplify.c (gimplify_expr): Handle unused *&&L;.
38238         PR sanitizer/79558
38239         * ubsan.c (ubsan_type_descriptor): Check if TYPE_MAX_VALUE is null.
38241 2017-02-20  Jakub Jelinek  <jakub@redhat.com>
38243         PR target/79568
38244         * config/i386/i386.c (ix86_expand_builtin): Handle
38245         OPTION_MASK_ISA_AVX512VL and OPTION_MASK_ISA_64BIT in
38246         ix86_builtins_isa[fcode].isa as a requirement of those
38247         flags and any other flag in the bitmask.
38248         (ix86_init_mmx_sse_builtins): Use 0 instead of
38249         ~OPTION_MASK_ISA_64BIT as mask.
38250         * config/i386/i386-builtin.def (__builtin_ia32_rdtsc,
38251         __builtin_ia32_rdtscp, __builtin_ia32_pause, __builtin_ia32_bsrsi,
38252         __builtin_ia32_rdpmc, __builtin_ia32_rolqi, __builtin_ia32_rolhi,
38253         __builtin_ia32_rorqi, __builtin_ia32_rorhi): Likewise.
38255 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38257         PR target/78012
38258         * lra-constraints.c (split_reg): Check requested split mode
38259         is supported by the register.
38261 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38263         * lra-constraints.c (simplify_operand_subreg): Remove early
38264         return false.
38266 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38268         PR target/78660
38269         * lra-constraints.c (curr_insn_transform): Tighten condition
38270         for converting SUBREG reloads from OP_OUT to OP_INOUT.
38272 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38274         PR target/78660
38275         * lra-constraints.c (curr_insn_transform): Handle
38276         WORD_REGISTER_OPERATIONS requirements when reloading SUBREGs.
38278 2017-02-19  Uros Bizjak  <ubizjak@gmail.com>
38280         Revert:
38281         2016-05-30  Uros Bizjak  <ubizjak@gmail.com>
38283         * config/i386/sync.md (mfence_nosse): Use "lock orl $0, -4(%esp)".
38285 2017-02-19  Jonathan Wakely  <jwakely@redhat.com>
38287         PR c++/69523
38288         * doc/invoke.texi (C++ Dialect Options) [-Wliteral-suffix]: Update
38289         description.
38291 2017-02-19  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
38293         * gimple-pretty-print.c (dump_ternary_rhs): Adjust gimple dump format
38294         for FMA_EXPR.
38296 2017-02-18  Jakub Jelinek  <jakub@redhat.com>
38298         * final.c (last_columnnum, override_columnnum): New variables.
38299         (final_start_function): Set last_columnnum, pass it to begin_prologue
38300         hook and pass 0 to dwarf2out_begin_prologue.
38301         (final_scan_insn): Update override_columnnum.  Pass last_columnnum
38302         to source_line debug hook.
38303         (notice_source_line): Compute last_columnnum and for debug_column_info
38304         return true on column changes.
38305         * debug.h (struct gcc_debug_hooks): Add column argument to
38306         source_line and begin_prologue hooks.
38307         (debug_nothing_int_charstar_int_bool): Remove prototype.
38308         (debug_nothing_int_int_charstar,
38309         debug_nothing_int_int_charstar_int_bool): New prototypes.
38310         (dwarf2out_begin_prologue): Add column argument.
38311         * debug.c (do_nothing_debug_hooks): Adjust source_line and
38312         begin_prologue hooks.
38313         (debug_nothing_int_charstar_int_bool): Remove.
38314         (debug_nothing_int_int_charstar,
38315         debug_nothing_int_int_charstar_int_bool): New functions.
38316         * dwarf2out.c (dwarf2out_begin_prologue): Add column argument, pass it
38317         through to dwarf2out_source_line.
38318         (dwarf2_lineno_debug_hooks): Adjust begin_prologue hook.
38319         (dwarf2out_source_line): Add column argument, emit it if requested.
38320         * sdbout.c (sdbout_source_line, sdbout_begin_prologue): Add column
38321         arguments.
38322         * xcoffout.h (xcoffout_begin_prologue, xcoffout_source_line): Likewise.
38323         * xcoffout.c (xcoffout_begin_prologue, xcoffout_source_line): Likewise.
38324         * vmsdbgout.c (vmsdbgout_begin_prologue): Add column argument, pass it
38325         through to dwarf2out_begin_prologue.
38326         (vmsdbgout_source_line): Add column argument, pass it through to
38327         dwarf2out_source_line.
38328         * dbxout.c (dbxout_begin_prologue): Add column argument, adjust
38329         dbxout_source_line caller.
38330         (dbxout_source_line): Add column argument.
38332         * common.opt (gno-column-info, gcolumn-info): New options.
38333         * dwarf2out.c (dwarf2_lineno_debug_hooks): Formatting fix.
38334         (check_die): Also test for multiple DW_AT_decl_column attributes.
38335         (add_src_coords_attributes, dwarf2out_imported_module_or_decl_1): Add
38336         DW_AT_decl_column if requested.
38337         (gen_subprogram_die): Compare and/or add also DW_AT_decl_column
38338         if requested.
38339         (gen_variable_die): Likewise.
38340         (add_call_src_coords_attributes): Add DW_AT_call_column if requested.
38341         * doc/invoke.texi (-gcolumn-info, -gno-column-info): Document.
38343         PR target/79569
38344         * config/i386/i386.opt (m3dnowa): Replace Undocumented with Report.
38345         * common/config/i386/i386-common.c (OPTION_MASK_ISA_3DNOW_A_SET): Define.
38346         (ix86_handle_option): Handle OPT_m3dnowa.
38347         * doc/invoke.texi (-m3dnowa): Document.
38348         * doc/extend.texi (__builtin_ia32_pmulhuw, __builtin_ia32_pf2iw): Use
38349         -m3dnowa instead of -m3dnow -march=athlon.
38351         PR target/79559
38352         * config/i386/i386.c (ix86_print_operand): Use output_operand_lossage
38353         instead of gcc_assert for K, r and R code checks.  Formatting fixes.
38355 2017-02-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
38357         PR target/79261
38358         * config/rs6000/rs6000.c (rs6000_expand_ternop_builtin): Add
38359         support for CODE_FOR_vsx_xxpermdi_v2d[fi]_be.
38360         * config/rs6000/rs6000.md (reload_gpr_from_vsx<mode>): Call
38361         generator for vsx_xxpermdi_<mode>_be.
38362         * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Remove logic to
38363         force big-endian semantics.
38364         (vsx_xxpermdi_<mode>_be): New define_expand with same
38365         implementation as previous version of vsx_xxpermdi_<mode>.
38367 2017-02-17  Jakub Jelinek  <jakub@redhat.com>
38369         PR tree-optimization/79327
38370         * gimple-ssa-sprintf.c (format_integer): Remove likely_adjust
38371         variable, its initialization and use.
38373 2017-02-17  Julia Koval  <julia.koval@intel.com>
38375         * common/config/i386/i386-common.c (OPTION_MASK_ISA_RDPID_SET): New.
38376         (OPTION_MASK_ISA_PKU_UNSET): New.
38377         (ix86_handle_option): Handle -mrdpid.
38378         * config/i386/cpuid.h (bit_RDPID): New.
38379         * config/i386/driver-i386.c (host_detect_local_cpu):
38380         Detect RDPID feature.
38381         * config/i386/i386-builtin.def (__builtin_ia32_rdpid): New.
38382         * config/i386/i386-c.c (ix86_target_macros_internal):
38383         Handle RDPID flag.
38384         * config/i386/i386.c (ix86_target_string): Add -mrdpid to isa2_opts.
38385         (ix86_valid_target_attribute_inner_p): Add "rdpid".
38386         (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
38387         * config/i386/i386.h (TARGET_RDPID, TARGET_RDPID_P): New.
38388         * config/i386/i386.md (define_insn "rdpid"): New.
38389         * config/i386/i386.opt Add -mrdpid.
38390         * config/i386/immintrin.h (_rdpid_u32): New.
38392 2017-02-17  Vladimir Makarov  <vmakarov@redhat.com>
38394         PR rtl-optimization/79541
38395         * lra-constraints.c (curr_insn_transform): Remove wrong asm insn
38396         instead of transforming it into USE.
38398 2017-02-17  Segher Boessenkool  <segher@kernel.crashing.org>
38400         * config/rs6000/rs6000.md (extendsfdf2): Remove default arguments.
38401         If HONOR_SNANS (SFmode) force the input to a register.
38402         (*extendsfdf2_fpr): Add !HONOR_SNANS (SFmode) condition.
38403         (*extendsfdf2_snan): New pattern, used when using SNaNs; it generates
38404         an frsp or similar insn.
38406 2017-02-17  Martin Liska  <mliska@suse.cz>
38408         PR rtl-optimization/79577
38409         * params.def (selsched-max-sched-times): Increase minimum to 1.
38411 2017-02-17  Martin Liska  <mliska@suse.cz>
38413         PR rtl-optimization/79574
38414         * gcse.c (want_to_gcse_p): Prevent integer overflow.
38416 2017-02-17  Martin Liska  <mliska@suse.cz>
38418         PR tree-optimization/79529
38419         * tree-ssa-loop-unswitch.c (is_maybe_undefined): Use
38420         ssa_defined_default_def_p to handle cases which are implicitly
38421         defined.
38422         * tree-ssa.c (ssa_defined_default_def_p): New function.
38423         (ssa_undefined_value_p): Use ssa_defined_default_def_p to handle cases
38424         which are implicitly defined.
38425         * tree-ssa.h (ssa_defined_default_def_p): Declare.
38427 2017-02-17  Richard Biener  <rguenther@suse.de>
38429         PR middle-end/79576
38430         * params.def (max-ssa-name-query-depth): Limit to 10.
38432 2017-02-17  Richard Biener  <rguenther@suse.de>
38434         PR tree-optimization/79552
38435         * tree-ssa-structalias.c (visit_loadstore): Properly verify
38436         default defs.
38438 2017-02-17  Richard Biener  <rguenther@suse.de>
38440         PR bootstrap/79567
38441         * genmatch.c (output_line_directive): Handle DIR_SEPARATOR_2.
38443 2017-02-17  Marek Polacek  <polacek@redhat.com>
38445         PR middle-end/79536
38446         * fold-const.c (fold_negate_expr_1): Renamed from fold_negate_expr.
38447         (fold_negate_expr): New wrapper.
38449 2017-02-16  Sandra Loosemore  <sandra@codesourcery.com>
38451         * doc/invoke.texi (C++ Dialect Options) [-Wno-non-template-friend]: 
38452         Correct terminology and de-emphasize pre-standard behavior.
38454 2017-02-16  Alan Modra  <amodra@gmail.com>
38456         PR rtl-optimization/79286
38457         * ira.c (def_dominates_uses): New function.
38458         (update_equiv_regs): Don't create an equivalence for insns that
38459         may trap where the register def does not dominate the use.
38461 2017-02-16  Vladimir Makarov  <vmakarov@redhat.com>
38463         PR rtl-optimization/78127
38464         * lra.c (lra): Call lra_eliminate before finish the loop after
38465         lra_constraint.
38467 2017-02-16  Richard Biener  <rguenther@suse.de>
38469         * graphite.h: Do not include isl/isl_val_gmp.h, instead include
38470         isl/isl_val.h.
38471         * graphite-isl-ast-to-gimple.c (gmp_cst_to_tree): Remove.
38472         (gcc_expression_from_isl_expr_int): Use generic isl_val interface.
38473         * graphite-sese-to-poly.c: Do not include isl/isl_val_gmp.h.
38474         (isl_val_int_from_wi): New function.
38475         (extract_affine_gmp): Rename to ...
38476         (extract_affine_wi): ... this, take a widest_int.
38477         (extract_affine_int): Just wrap extract_affine_wi.
38478         (add_param_constraints): Use isl_val_int_from_wi.
38479         (add_loop_constraints): Likewise, and extract_affine_wi.
38481 2017-02-15  Jeff Law  <law@redhat.com>
38483         PR middle-end/79521
38484         * ira-costs.c (scan_one_insn): Check have_regs_of_mode before calling
38485         ira_init_register_move_cost_if_necessary.
38487 2017-02-15  Martin Sebor  <msebor@redhat.com>
38489         PR middle-end/32003
38490         * doc/invoke.texi (-fdump-final-insns): Replace option accidentally
38491         removed in a prior commit.
38493 2017-02-15  Bin Cheng  <bin.cheng@arm.com>
38495         PR tree-optimization/79347
38496         * tree-vect-loop-manip.c (vect_do_peeling): Maintain profile
38497         counters during peeling.
38499 2017-02-15  Thomas Schwinge  <thomas@codesourcery.com>
38501         * Makefile.in (site.exp): Remove "set ISLVER".
38503 2017-02-15  Jakub Jelinek  <jakub@redhat.com>
38505         PR target/79487
38506         * real.c (real_from_integer): Call real_convert even for decimal.
38508 2017-02-15  Dominik Vogt  <vogt@linux.vnet.ibm.com>
38510         PR target/79421
38511         * config/s390/s390.c: define TARGET_CUSTOM_FUNCTION_DESCRIPTORS.
38513 2017-02-14  Andrew Pinski  <apinski@cavium.com>
38515         * config/aarch64/aarch64-cores.def (thunderx2t99): Move to under 'C"
38516         cores and change the partno/implementer to be correct.
38517         (thunderx2t99p1): New core which replaces thunderx2t99 and still has
38518         the 'B" as the implementer.
38519         * config/aarch64/aarch64-tune.md: Regenerate.
38521 2017-02-14  Carl Love  <cel@us.ibm.com>
38523         * config/rs6000/rs6000.c: Add case statement entry to make the
38524         xvcvuxdsp built-in argument unsigned.
38525         * config/rs6000/vsx.md: Fix the source and return operand types so they
38526         match the instruction definitions from the ISA document.  Fix typo
38527         in the instruction generation for the (define_insn "vsx_xvcvuxdsp"
38528         statement.
38530 2017-02-14  Vladimir Makarov  <vmakarov@redhat.com>
38532         PR target/79282
38533         * lra-int.h (struct lra_operand_data, struct lra_insn_reg): Add
38534         member early_clobber_alts.
38535         * lra-lives.c (reg_early_clobber_p): New.
38536         (process_bb_lives): Use it.
38537         * lra.c (new_insn_reg): New arg early_clobber_alts.  Use it.
38538         (debug_operand_data): Initialize early_clobber_alts.
38539         (setup_operand_alternative): Set up early_clobber_alts.
38540         (collect_non_operand_hard_regs): Ditto.  Pass early clobber
38541         alternatives to new_insn_reg.
38542         (add_regs_to_insn_regno_info): Add arg early_clobber_alts.  Use
38543         it.
38544         (lra_update_insn_regno_info): Pass the new arg.
38546 2017-02-14  Jakub Jelinek  <jakub@redhat.com>
38548         PR middle-end/79505
38549         * omp-offload.c (free_oacc_loop): Release loop->ifns vector.
38550         (new_oacc_loop_raw): Don't clear already cleared fields.
38552         PR target/79481
38553         * config/i386/avx512pfintrin.h (_mm512_prefetch_i32gather_pd,
38554         _mm512_prefetch_i32gather_ps, _mm512_prefetch_i64gather_pd,
38555         _mm512_prefetch_i64gather_ps): New inline functions and macros.
38557 2017-02-14  Uros Bizjak  <ubizjak@gmail.com>
38559         PR target/79495
38560         * config/i386/i386.md (*movxf_internal): Add (o,rC) alternative.
38562 2017-02-14  H.J. Lu  <hongjiu.lu@intel.com>
38564         PR target/79498
38565         * config/i386/i386.c (timode_scalar_chain::convert_insn): Insert
38566         the extra instruction to the right place to store 128-bit constant
38567         when needed.
38569 2017-02-14  Martin Sebor  <msebor@redhat.com>
38571         PR middle-end/79448
38572         * gimple-ssa-sprintf.c (format_directive): Avoid issuing INT_MAX
38573           warning for strings of unknown length.
38575 2017-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
38577         * config.gcc (supported_defaults) [powerpc*-*-*]: Update.
38579 2017-02-14  Jeff Law  <law@redhat.com>
38581         PR target/79404
38582         * ira-costs.c (scan_one_insn): Initialize register move costs
38583         for pseudos seen in USE/CLOBBER insns.
38585         PR tree-optimization/79095
38586         * tree-vrp.c (extract_range_from_binary_expr_1): For EXACT_DIV_EXPR,
38587         if the numerator has the range ~[0,0] make the resultant range ~[0,0].
38588         (extract_range_from_binary_expr): For MINUS_EXPR with no derived range,
38589         if the operands are known to be not equal, then the resulting range
38590         is ~[0,0].
38591         (intersect_ranges): If the new range is ~[0,0] and the old range is
38592         wide, then prefer ~[0,0].
38593         * tree-vrp.c (overflow_comparison_p_1): New function.
38594         (overflow_comparison_p): New function.
38595         * tree-vrp.c (register_edge_assert_for_2): Register additional asserts
38596         if NAME is used in an overflow test.
38597         (vrp_evaluate_conditional_warnv_with_ops): If the ops represent an
38598         overflow check that can be expressed as an equality test, then adjust
38599         ops to be that equality test.
38601 2017-02-14  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
38603         * config/s390/s390-builtin-types.def: Remove flags argument.
38604         * config/s390/s390.c (s390_init_builtins): Likewise.
38606 2017-02-14  Martin Liska  <mliska@suse.cz>
38608         * tree-ssa-loop-unswitch.c (hoist_guard): Release get_loop_body
38609         vector.  Fix trailing white spaces.
38611 2017-02-14  James Greenhalgh  <james.greenhalgh@arm.com>
38613         * config/aarch64/aarch64.c (aarch64_simd_container_mode): Handle
38614         HFmode.
38616 2017-02-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38618         PR rtl-optimization/68664
38619         * config/arm/arm.c (arm_sched_can_speculate_insn):
38620         New function.  Declare prototype.
38621         (TARGET_SCHED_CAN_SPECULATE_INSN): Define.
38623 2017-02-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38625         PR rtl-optimization/68664
38626         * config/aarch64/aarch64.c (aarch64_sched_can_speculate_insn):
38627         New function.
38628         (TARGET_SCHED_CAN_SPECULATE_INSN): Define.
38630 2017-02-14  Amit Pawar  <amit.pawar@amd.com>
38632         * config/i386/i386.c (znver1_cost): Fix the alignment for function and
38633         max skip bytes for function, loop and jump.
38635 2017-02-14  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
38637         * gimple-pretty-print.c (dump_unary_rhs): Adjust dump format for
38638         ABS_EXPR for gimple dump.
38640 2017-02-14  Jakub Jelinek  <jakub@redhat.com>
38642         PR target/79462
38643         * config/sh/sh.c (expand_cbranchdi4): Don't clear operands[4].
38645         PR tree-optimization/79408
38646         * tree-vrp.c (simplify_div_or_mod_using_ranges): Handle also the
38647         case when on TRUNC_MOD_EXPR op0 is INTEGER_CST.
38648         (simplify_stmt_using_ranges): Call simplify_div_or_mod_using_ranges
38649         also if rhs1 is INTEGER_CST.
38651 2017-02-14  Richard Biener  <rguenther@suse.de>
38653         PR middle-end/79432
38654         * tree-into-ssa.c (insert_phi_nodes): When the function can
38655         have abnormal edges rewrite SSA names with broken use-def
38656         dominance out of SSA and register them for PHI insertion.
38658 2017-02-13  Martin Sebor  <msebor@redhat.com>
38660         PR middle-end/79496
38661         * gimple-ssa-sprintf.c (pass_sprintf_length::handle_gimple_call): Avoid
38662         clearing info.nowrite flag when snprintf size argument is a range.
38664 2017-02-13  Jakub Jelinek  <jakub@redhat.com>
38666         * cprop.c (cprop_jump): Add missing space in string literal.
38667         * tree-ssa-structalias.c (rewrite_constraints): Likewise.
38668         (get_constraint_for_component_ref): Likewise.
38669         * df-core.c (df_worklist_dataflow_doublequeue): Likewise.
38670         * tree-outof-ssa.c (insert_partition_copy_on_edge): Likewise.
38671         * lra-constraints.c (process_alt_operands): Likewise.
38672         * ipa-inline.c (inline_small_functions): Likewise.
38673         * tree-ssa-sccvn.c (visit_reference_op_store): Likewise.
38674         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
38675         * trans-mem.c (diagnose_tm_1_op): Likewise.
38676         * omp-grid.c (grid_find_single_omp_among_assignments): Likewise.
38677         (grid_parallel_clauses_gridifiable): Likewise.
38679         * config/nvptx/mkoffload.c (process): Add space in between
38680         , and %d.
38682         * config/i386/i386.h (REG_CLASS_NAMES): Add , in between
38683         "MOD4_SSE_REGS" and "ALL_REGS".
38685         * spellcheck.c (test_data): Add , in between "foo" and "food".
38687 2017-02-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
38689         PR target/79449
38690         * config/rs6000/rs6000.c (expand_block_compare): Make sure runtime
38691         boundary crossing check and subsequent code generation agree.
38693 2017-02-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38695         * config/aarch64/aarch64.c (has_memory_op): Delete.
38696         (aarch64_madd_needs_nop): Use contains_mem_rtx_p instead of
38697         has_memory_op.
38699 2017-02-13  Jakub Jelinek  <jakub@redhat.com>
38701         PR rtl-optimization/79388
38702         PR rtl-optimization/79450
38703         * combine.c (distribute_notes): When removing TEM_INSN for which
38704         corresponding dest has last value recorded, invalidate that last
38705         value.
38707 2017-02-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38709         * config/arm/arm.c (arm_print_tune_info): Use ASM_COMMENT_START instead
38710         of explicit '@'.  Add missing assembly comment marker on branch costs
38711         printout.
38713 2017-02-13  Nathan Sidwell  <nathan@acm.org>
38715         * gengtype-lex.l (<in_struct>): Add '/'.
38717 2017-02-13  Martin Liska  <mliska@suse.cz>
38719         PR c/79471
38720         * calls.c (expand_call): Replace XALLOCAVEC with XCNEWVEC.
38722 2017-02-13  Richard Biener  <rguenther@suse.de>
38724         * configure.ac (HAVE_ISL_OPTIONS_SET_SCHEDULE_SERIALIZE_SCCS):
38725         Remove.
38726         * configure: Re-generate.
38727         * config.in: Likewise.
38728         * graphite-dependences.c: Simplify as if
38729         HAVE_ISL_OPTIONS_SET_SCHEDULE_SERIALIZE_SCCS was defined.
38730         * graphite-isl-ast-to-gimple.c: Likewise.
38731         * graphite-optimize-isl.c: Likewise.
38732         * graphite-poly.c: Likewise.
38733         * graphite-sese-to-poly.c: Likewise.
38734         * graphite.h: Likewise.
38735         * toplev.c: Include isl/version.h and use isl_version () for
38736         printing the ISL version.
38737         * doc/install.texi: Update ISL requirement.
38739 2017-02-12  Gerald Pfeifer  <gerald@pfeifer.com>
38741         * doc/standards.texi (Standards): Update reference to
38742         Objective-C 2.0.
38744 2017-02-12  Gerald Pfeifer  <gerald@pfeifer.com>
38746         * doc/extend.texi (Named Address Spaces): sourceware.org now
38747         defaults to https.
38748         * doc/install.texi (Binaries): Ditto.
38749         (Specific): Ditto.
38751 2017-02-11  Sandra Loosemore  <sandra@codesourcery.com>
38753         * doc/cpp.texi: Replace "stringify"/"stringification" with C
38754         standard terminology "stringize"/"stringizing" throughout.
38755         * doc/cppinternals.texi: Likewise.
38757 2017-02-11  Sandra Loosemore  <sandra@codesourcery.com>
38759         * doc/extend.texi: Fix some spelling mistakes and typos.
38760         * doc/invoke.texi: Likewise.
38762 2017-02-11  Jan Hubicka  <hubicka@ucw.cz>
38764         PR ipa/79224
38765         * params.def (inline-min-speedup) Change from 10 to 8.
38767 2017-02-11  Jakub Jelinek  <jakub@redhat.com>
38769         * doc/invoke.texi (fopenmp): Bump OpenMP version from 4.0 to
38770         4.5.
38772 2017-02-11  Jan Hubicka  <hubicka@ucw.cz>
38774         PR ipa/79224
38775         * ipa-inline-analysis.c (get_minimal_bb): New function.
38776         (record_modified): Use it.
38777         (remap_edge_change_prob): Handle also ancestor functions.
38779 2017-02-11  Gerald Pfeifer  <gerald@pfeifer.com>
38781         * doc/contrib.texi (Contributors): Remove broken link into
38782         the Mauve CVS repository.
38784 2017-02-11  Jakub Jelinek  <jakub@redhat.com>
38786         PR middle-end/79454
38787         * internal-fn.c (expand_vector_ubsan_overflow): Use piece-wise
38788         result computation whenever lhs doesn't have vector mode, not
38789         just when it has BLKmode.
38791 2017-02-10  Gerald Pfeifer  <gerald@pfeifer.com>
38793         * doc/makefile.texi (profiledbootstrap): Refer to the
38794         installation instructions only in textual form.
38796 2017-02-10  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
38798         PR target/79295
38799         * config/rs6000/altivec.md (bcd<bcd_add_sub>): Fix constraints.
38801 2017-02-10  Gerald Pfeifer  <gerald@pfeifer.com>
38803         * doc/install.texi (Specific): Use https for blackfin.uclinux.org.
38804         (Specific): Update mingw-w64 reference.
38805         (Binaries): Ditto.
38806         (Specific): Remove broken link to Renesas RX processor.
38808 2017-02-10  Richard Biener  <rguenther@suse.de>
38810         * toplev.c (process_options): Do not mention obsolete graphite
38811         options when printing sorry message about missing graphite support.
38812         Mention -floop-nest-optimize.
38814 2017-02-10  Christophe Lyon  <christophe.lyon@linaro.org>
38816         * config/aarch64/arm_neon.h (vtst_p8): Rewrite without asm.
38817         (vtst_p16): Likewise.
38818         (vtstq_p8): Likewise.
38819         (vtstq_p16): Likewise.
38820         (vtst_p64): New.
38821         (vtstq_p64): Likewise.
38822         * config/arm/arm_neon.h (vgetq_lane_p64): New.
38823         (vset_lane_p64): New.
38824         (vsetq_lane_p64): New.
38826 2017-02-10  Jakub Jelinek  <jakub@redhat.com>
38828         PR tree-optimization/79411
38829         * tree-ssa-reassoc.c (is_reassociable_op): Return false if
38830         stmt operands are SSA_NAMEs used in abnormal phis.
38831         (can_reassociate_p): Return false if op is SSA_NAME used in abnormal
38832         phis.
38834 2017-02-09  Jan Hubicka  <hubicka@ucw.cz>
38836         PR ipa/70795
38837         * cgraphunit.c (cgraph_node::add_new_function): Set externally_visible
38838         flag if needed.
38840 2017-02-09  Jan Hubicka  <hubicka@ucw.cz>
38842         * tree-ssa-loop-unswitch.c (hoist_guard): Update profile.
38844 2017-02-09  Jakub Jelinek  <jakub@redhat.com>
38846         * omp-offload.c (oacc_loop_auto_partitions): Use || instead of |
38847         to avoid warning.
38849         PR c/79413
38850         * gimplify.h (is_gimple_sizepos): Only test for INTEGER_CST constants,
38851         not arbitrary TREE_CONSTANT.
38853         PR c/79431
38854         * gimplify.c (gimplify_adjust_omp_clauses): Ignore
38855         "omp declare target link" attribute unless is_global_var.
38856         * omp-offload.c (find_link_var_op): Likewise.
38858 2017-02-09  Nathan Sidwell  <nathan@codesourcery.com>
38859             Chung-Lin Tang  <cltang@codesourcery.com>
38861         * gimplify.c (gimplify_scan_omp_clauses): No special handling for
38862         OMP_CLAUSE_TILE.
38863         (gimplify_adjust_omp_clauses): Don't delete TILE.
38864         (gimplify_omp_for): Deal with TILE.
38865         * internal-fn.c (expand_GOACC_TILE): New function.
38866         * internal-fn.def (GOACC_DIM_POS): Comment may be overly conservative.
38867         (GOACC_TILE): New.
38868         * omp-expand.c (struct oacc_collapse): Add tile and outer fields.
38869         (expand_oacc_collapse_init): Add LOC paramter.  Initialize tile
38870         element fields.
38871         (expand_oacc_collapse_vars): Add INNER parm, adjust for tiling,
38872         avoid DIV for outermost collapse var.
38873         (expand_oacc_for): Insert tile element loop as needed.  Adjust.
38874         Remove out of date comments, fix whitespace.
38875         * omp-general.c (omp_extract_for_data): Deal with tiling.
38876         * omp-general.h (enum oacc_loop_flags): Add OLF_TILE flag,
38877         adjust OLF_DIM_BASE value.
38878         (struct omp_for_data): Add tiling field.
38879         * omp-low.c (scan_sharing_clauses): Allow OMP_CLAUSE_TILE.
38880         (lower_oacc_head_mark): Add OLF_TILE as appropriate.  Ensure 2 levels
38881         for auto loops.  Remove default auto determining, moved to
38882         oacc_loop_fixed_partitions.
38883         * omp-offload.c (struct oacc_loop): Change 'ifns' to vector of call
38884         stmts, add e_mask field.
38885         (oacc_dim_call): New function, abstracted out from oacc_thread_numbers.
38886         (oacc_thread_numbers): Use oacc_dim_call.
38887         (oacc_xform_tile): New.
38888         (new_oacc_loop_raw): Initialize e_mask, adjust for ifns vector.
38889         (finish_oacc_loop): Adjust for ifns vector.
38890         (oacc_loop_discover_walk): Append loop abstraction sites to list,
38891         add case for GOACC_TILE fns.
38892         (oacc_loop_xform_loop): Delete.
38893         (oacc_loop_process): Iterate over call list directly, and add
38894         handling for GOACC_TILE fns.
38895         (oacc_loop_fixed_partitions): Determine default auto, deal with TILE,
38896         dump partitioning.
38897         (oacc_loop_auto_partitions): Add outer_assign parm. Assign all but
38898         vector partitioning to outer loops.  Assign 2 partitions to loops
38899         when available. Add TILE handling.
38900         (oacc_loop_partition): Adjust oacc_loop_auto_partitions call.
38901         (execite_oacc_device_lower): Process GOACC_TILE fns,
38902         ignore unknown specs.
38903         * tree-nested.c (convert_nonlocal_omp_clauses): Allow OMP_CLAUSE_TILE.
38904         * tree.c (omp_clause_num_ops): Adjust TILE ops.
38905         * tree.h (OMP_CLAUSE_TILE_ITERVAR, OMP_CLAUSE_TILE_COUNT): New.
38907 2017-02-09  Gerald Pfeifer  <gerald@pfeifer.com>
38909         * configure.ac (ACX_BUGURL): Update.
38910         * configure: Regenerate.
38912 2017-02-09  Richard Biener  <rguenther@suse.de>
38914         PR tree-optimization/69823
38915         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
38916         Properly enumerate all BBs in the region.  Use auto_vec/auto_bitmap.
38918 2017-02-09  Andrew Burgess  <andrew.burgess@embecosm.com>
38920         * config/arc/arc-c.def: Add __NPS400__ definition.
38921         * config/arc/arc.h (CPP_SPEC): Don't define __NPS400__ here.
38922         (TARGET_NPS400): Define.
38924 2017-02-09  Andrew Burgess  <andrew.burgess@embecosm.com>
38926         * config/arc/arc-arch.h (arc_arch_t): Move unchanged to earlier in
38927         file.
38928         (arc_cpu_t): Change base_architecture field, arch, to a arc_arc_t
38929         pointer, arch_info.
38930         (arc_cpu_types): Fill the arch_info field with a pointer into the
38931         arc_arch_types table.
38932         (arc_selected_cpu): Declare.
38933         * config/arc/arc.c (arc_selected_cpu): Make global.
38934         (arc_selected_arch): Delete.
38935         (arc_base_cpu): Delete.
38936         (arc_override_options): Remove references to deleted variables,
38937         update access to arch information.
38938         (ARC_OPT): Update access to arch information.
38939         (ARC_OPTX): Likewise.
38940         * config/arc/arc.h (arc_base_cpu): Remove declaration.
38941         (TARGET_ARC600): Update access to arch information.
38942         (TARGET_ARC601): Likewise.
38943         (TARGET_ARC700): Likewise.
38944         (TARGET_EM): Likewise.
38945         (TARGET_HS): Likewise.
38946         * config/arc/driver-arc.c (arc_cpu_to_as): Update access to arch
38947         information.
38949 2017-02-08  Pat Haugen  <pthaugen@us.ibm.com>
38951         PR target/78604
38952         * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Invert
38953         condition/operands for integer GE/LE/GEU/LEU operations.
38955 2017-02-08  Segher Boessenkool  <segher@kernel.crashing.org>
38957         PR translation/79397
38958         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Fix spelling
38959         of AltiVec.
38961 2017-02-08  Martin Jambor  <mjambor@suse.cz>
38963         PR ipa/79375
38964         * ipa-prop.c (ipa_alloc_node_params): Make static, return bool
38965         whether allocation happened.
38966         (ipa_initialize_node_params): Do not call ipa_alloc_node_params if
38967         nothing was allocated.
38969 2017-02-08  Jakub Jelinek  <jakub@redhat.com>
38971         PR tree-optimization/79408
38972         * tree-vrp.c (simplify_div_or_mod_using_ranges): If op1 is not
38973         constant, but SSA_NAME with a known integer range, use the minimum
38974         of that range instead of op1 to determine if modulo can be replaced
38975         with its first operand.
38977 2016-02-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38979         * config/riscv/riscv.c (riscv_build_integer_1): Avoid use of INT16_MAX.
38981 2017-02-08  Richard Biener  <rguenther@suse.de>
38983         PR tree-optimization/71824
38984         * graphite-scop-detection.c (scop_detection::build_scop_breadth):
38985         Check all loops contained in the merged region.
38987 2017-02-07  Andrew Pinski  <apinski@cavium.com>
38989         * config/aarch64/aarch64.md (popcount<mode>2): New pattern.
38991 2017-02-07  Andrew Pinski  <apinski@cavium.com>
38993         * config/aarch64/aarch64-cores.def (thunderx): Disable LSE.
38994         (thunderxt88): Likewise.
38995         (thunderxt81): Disable LSE and change v8.1 to v8.
38996         (thunderxt83): Likewise.
38998 2017-02-07  Jakub Jelinek  <jakub@redhat.com>
38999             Richard Biener  <rguenther@suse.de>
39001         PR middle-end/79399
39002         * ira-int.h (struct target_ira_int): Change x_max_struct_costs_size
39003         type from int to size_t.
39004         * ira-costs.c (struct_costs_size): Change type from int to size_t.
39006 2017-02-07  Jakub Jelinek  <jakub@redhat.com>
39008         PR rtl-optimization/79386
39009         * cprop.c (bypass_conditional_jumps): Initialize
39010         bypass_last_basic_block already before splitting bbs after
39011         unconditional traps...
39012         (bypass_conditional_jumps): ... rather than here.
39014         PR target/79299
39015         * config/i386/sse.md (xtg_mode, gatherq_mode): New mode attrs.
39016         (*avx512f_gathersi<mode>, *avx512f_gathersi<mode>_2,
39017         *avx512f_gatherdi<mode>, *avx512f_gatherdi<mode>_2): Use them,
39018         fix -masm=intel patterns.
39020 2017-02-07  Richard Biener  <rguenther@suse.de>
39022         PR tree-optimization/79256
39023         PR middle-end/79278
39024         * builtins.c (get_object_alignment_2): Use min_align_of_type
39025         to extract alignment for MEM_REFs to honor BIGGEST_FIELD_ALIGNMENT
39026         and ADJUST_FIELD_ALIGN.
39028         * doc/tm.texi.in (ADJUST_FIELD_ALIGN): Adjust to take additional
39029         type parameter.
39030         * doc/tm.texi: Regenerate.
39031         * stor-layout.c (layout_decl): Adjust.
39032         (update_alignment_for_field): Likewise.
39033         (place_field): Likewise.
39034         (min_align_of_type): Likewise.
39035         * config/arc/arc.h (ADJUST_FIELD_ALIGN): Adjust.
39036         * config/epiphany/epiphany.h (ADJUST_FIELD_ALIGN): Likewise.
39037         * config/epiphany/epiphany.c (epiphany_adjust_field_align): Likewise.
39038         * config/frv/frv.h (ADJUST_FIELD_ALIGN): Likewise.
39039         * config/frv/frv.c (frv_adjust_field_align): Likewise.
39040         * config/i386/i386.h (ADJUST_FIELD_ALIGN): Likewise.
39041         * config/i386/i386.c (x86_field_alignment): Likewise.
39042         * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Likewise.
39043         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Likewise.
39044         * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Likewise.
39045         * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
39046         * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Likewise.
39047         * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
39048          Likewise.
39050         Revert
39051         2017-01-30  Richard Biener  <rguenther@suse.de>
39053         PR tree-optimization/79256
39054         * targhooks.c (default_builtin_vector_alignment_reachable): Honor
39055         BIGGEST_FIELD_ALIGNMENT and ADJUST_FIELD_ALIGN to fix up bogus
39056         alignment on TYPE.
39058 2017-02-07  Toma Tabacu  <toma.tabacu@imgtec.com>
39060         * config/mips/mips.c (mips_expand_builtin_insn): Convert the QImode
39061         argument of the pshufh, psllh, psllw, psrah, psraw, psrlh, psrlw
39062         builtins to SImode and emit a zero-extend, if necessary.
39064 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
39066         * docs/invoke.texi (RISC-V Options): Alphabetize.
39068 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
39070         * doc/invoke.texi (RISC-V Options): Use two spaces to separate
39071         options.
39073 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
39075         * config/riscv/riscv.c: New file.
39076         * common/config/riscv/riscv-common.c: Likewise.
39077         * config.gcc: Likewise.
39078         * config/riscv/constraints.md: Likewise.
39079         * config/riscv/elf.h: Likewise.
39080         * config/riscv/generic.md: Likewise.
39081         * config/riscv/linux.h: Likewise.
39082         * config/riscv/multilib-generator: Likewise.
39083         * config/riscv/peephole.md: Likewise.
39084         * config/riscv/pic.md: Likewise.
39085         * config/riscv/predicates.md: Likewise.
39086         * config/riscv/riscv-builtins.c: Likewise.
39087         * config/riscv/riscv-c.c: Likewise.
39088         * config/riscv/riscv-ftypes.def: Likewise.
39089         * config/riscv/riscv-modes.def: Likewise.
39090         * config/riscv/riscv-opts.h: Likewise.
39091         * config/riscv/riscv-protos.h: Likewise.
39092         * config/riscv/riscv.h: Likewise.
39093         * config/riscv/riscv.md: Likewise.
39094         * config/riscv/riscv.opt: Likewise.
39095         * config/riscv/sync.md: Likewise.
39096         * config/riscv/t-elf-multilib: Likewise.
39097         * config/riscv/t-linux: Likewise.
39098         * config/riscv/t-linux-multilib: Likewise.
39099         * config/riscv/t-riscv: Likewise.
39100         * configure.ac: Likewise.
39101         * doc/contrib.texi: Add Kito Cheng, Palmer Dabbelt, and Andrew
39102         Waterman as RISC-V maintainers.
39103         * doc/install.texi: Add RISC-V entries.
39104         * doc/invoke.texi: Add RISC-V options section.
39105         * doc/md.texi: Add RISC-V constraints section.
39106         * configure: Regenerated.
39108 2017-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
39110         PR target/66144
39111         * config/rs6000/vector.md (vcond<mode><mode>): Allow the true and
39112         false values to be constant vectors with all 0 or all 1 bits set.
39113         (vcondu<mode><mode>): Likewise.
39114         * config/rs6000/predicates.md (vector_int_reg_or_same_bit): New
39115         predicate.
39116         (fpmask_comparison_operator): Update comment.
39117         (vecint_comparison_operator): New predicate.
39118         * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Optimize
39119         vector conditionals when the true and false values are constant
39120         vectors with all 0 bits or all 1 bits set.
39122 2017-02-06  Martin Sebor  <msebor@redhat.com>
39124         PR  tree-optimization/79376
39125         * gimple-fold.c (get_range_strlen): Set the minimum length to zero.
39127 2017-02-06  Uros Bizjak  <ubizjak@gmail.com>
39129         * config/i386/sse.md (vector modes -> vec_extract* splitter): Use
39130         explicit subreg RTX with operand 1.  Use VECTOR_MODE_P predicate
39131         to simplify split condition.
39133 2017-02-06  Jakub Jelinek  <jakub@redhat.com>
39135         * omp-expand.c (oxpand_omp_atomic_fetch_op,
39136         expand_omp_atomic_pipeline): Return false if can_atomic_load_p is
39137         false.
39139 2017-02-06  Segher Boessenkool  <segher@kernel.crashing.org>
39141         PR rtl-optimization/68664
39142         * target.def (can_speculate_insn): New hook.
39143         * doc/tm.texi.in (TARGET_SCHED_CAN_SPECULATE_INSN): New hook.
39144         * doc/tm.texi: Regenerate.
39145         * sched-rgn.c (can_schedule_ready_p): Use the new hook.
39146         * config/rs6000/rs6000.c (TARGET_SCHED_CAN_SPECULATE_INSN): New macro.
39147         (rs6000_sched_can_speculate_insn): New function.
39149 2017-02-06  Jakub Jelinek  <jakub@redhat.com>
39151         PR tree-optimization/79284
39152         * tree-vectorizer.h (VECT_SCALAR_BOOLEAN_TYPE_P): Define.
39153         * tree-vect-stmts.c (vect_get_vec_def_for_operand,
39154         vectorizable_mask_load_store, vectorizable_operation,
39155         vect_is_simple_cond, get_same_sized_vectype): Use it instead
39156         of comparing TREE_CODE of a type against BOOLEAN_TYPE.
39157         * tree-vect-patterns.c (check_bool_pattern, search_type_for_mask_1,
39158         vect_recog_bool_pattern, vect_recog_mask_conversion_pattern): Likewise.
39159         * tree-vect-slp.c (vect_get_constant_vectors): Likewise.
39160         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
39161         Remove redundant gimple_code (stmt) == GIMPLE_ASSIGN test after
39162         is_gimple_assign (stmt).  Replace another such test with
39163         is_gimple_assign (stmt).
39165 2017-02-06  Georg-Johann Lay  <avr@gjlay.de>
39167         PR target/78883
39168         * config/avr/avr.c (rtl-iter.h): Include it.
39169         (TARGET_LEGITIMATE_COMBINED_INSN): New hook define...
39170         (avr_legitimate_combined_insn): ...and implementation.
39172 2017-02-06  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39174         * config/s390/predicates.md ("larl_operand"): Use macros from hwint.h.
39175         * config/s390/s390.c (s390_const_operand_ok)
39176         (s390_canonicalize_comparison, s390_extract_part)
39177         (s390_single_part, s390_contiguous_bitmask_nowrap_p)
39178         (s390_contiguous_bitmask_p, s390_rtx_costs)
39179         (legitimize_pic_address): Likewise.
39180         * config/s390/s390.md ("clzdi2", "clztidi2"): Likewise.
39181         * config/s390/vx-builtins.md ("vec_genbytemaskv16qi")
39182         ("vec_permi<mode>", "vfae<mode>", "*vfaes<mode>", "vstrc<mode>")
39183         ("*vstrcs<mode>"): Use UINTVAL() to set unsigned HOST_WIDE_INT.
39184         * config/s390/vector.md ("vec_vfenes<mode>"): Likewise.
39186 2017-02-06  Georg-Johann Lay  <avr@gjlay.de>
39188         * config/avr/avr.md (*addhi3_zero_extend): Add alternative where
39189         REGNO($0) == REGNO($1).
39191 2017-02-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39193         * config/s390/linux.h(SIZE_TYPE): Add comment.
39195 2017-02-06  Julian Brown  <julian@codesourcery.com>
39196             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
39197             Virendra Pathak  <virendra.pathak@broadcom.com>
39199         * config/aarch64/aarch64-cores.def: Change the scheduler
39200         to Thunderx2t99.
39201         * config/aarch64/aarch64.md: Include thunderx2t99.md.
39202         * config/aarch64/thunderx2t99.md: New file.
39204 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
39206         * doc/standards.texi (Go Language): Update link to language
39207         standard.
39209 2017-02-05  Jan Hubicka  <hubicka@ucw.cz>
39211         * tree-eh.c (lower_resx): Sanitize profile.
39212         (cleanup_empty_eh_move_lp): Likewise.
39214 2017-02-05  Jan Hubicka  <hubicka@ucw.cz>
39216         PR tree-ssa/79347
39217         * cfgloopmanip.c (lv_adjust_loop_entry_edge, loop_version): Add
39218         ELSE_PROB.
39219         * cfgloopmanip.h (loop_version): Update prototype.
39220         * modulo-sched.c (sms_schedule): Update call of loop_version.
39221         * tree-if-conv.c(version_loop_for_if_conversion): Likewise.
39222         * tree-parloops.c (gen_parallel_loop): Likewise.
39223         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Likewise.
39224         * tree-ssa-loop-split.c (split_loop): Likewise.
39225         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Likewise.
39226         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
39228 2017-02-05  Martin Liska  <mliska@suse.cz>
39230         PR bootstrap/78985
39231         * config/s390/s390.c (s390_gimplify_va_arg): Initialize local
39232         variable to NULL.
39233         (print_operand_address): Initialize a struct to zero.
39235 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
39237         * doc/contrib.texi (Contributors): Refer to Hans Boehm's
39238         garbage collector only in textual form.
39240 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
39242         * doc/extend.texi (x86 specific memory model extensions for
39243         transactional memory): Simplify a phrase.
39245 2017-02-05  Eric Botcazou  <ebotcazou@adacore.com>
39247         PR target/79353
39248         * config/sparc/sync.md (atomic_loaddi_1): Replace 'U' constraint with
39249         'r', 'm' constraint with 'T' and !TARGET_ARCH64 with TARGET_ARCH32.
39250         (atomic_storedi_1): Likewise.
39252 2017-02-04  Jakub Jelinek  <jakub@redhat.com>
39254         PR tree-optimization/79338
39255         * tree-parloops.c (gather_scalar_reductions): Don't call
39256         vect_analyze_loop_form for loop->inner before destroying loop's
39257         loop_vinfo.
39259 2017-02-03  Martin Sebor  <msebor@redhat.com>
39261         PR tree-optimization/79327
39262         * gimple-ssa-sprintf.c (tree_digits): Avoid adding the base prefix
39263         when precision has resulted in leading zeros.
39264         (format_integer): Adjust the likely counter to assume an unknown
39265         argument that may be zero is non-zero.
39267 2017-02-03  Jason Merrill  <jason@redhat.com>
39269         PR c++/78689
39270         * tree-inline.c (copy_tree_body_r) [COND_EXPR]: Revert change to
39271         avoid copying non-taken branch.
39273 2017-02-03  Jakub Jelinek  <jakub@redhat.com>
39275         PR tree-optimization/79340
39276         * tree-vect-loop.c (vectorizable_reduction): Release
39277         vec_defs elements after safe_splicing them into other vectors.
39278         Formatting fixes.
39280         PR tree-optimization/79327
39281         * gimple-ssa-sprintf.c (adjust_range_for_overflow): If returning
39282         true, always set *argmin and *argmax to TYPE_{MIN,MAX}_VALUE of
39283         dirtype.
39284         (format_integer): Use wide_int_to_tree instead of build_int_cst
39285         + to_?hwi.  If argmin is NULL, just set argmin and argmax to
39286         TYPE_{MIN,MAX}_VALUE of argtype.  Simplify and fix computation
39287         of shortest and longest sequence.
39289 2017-02-03  Uros Bizjak  <ubizjak@gmail.com>
39291         * config/i386/i386.c (dimode_scalar_chain::convert_reg):
39292         Use pextrd for TARGET_SSE4_1 when creating scalar copy.
39294 2017-02-03  Walter Lee  <walt@tilera.com>
39296         PR target/78862
39297         * config/tilegx/tilegx.md (tilegx_expand_prologue): Add blockage
39298         after initial stackframe link reg save.
39299         * config/tilepro/tilepro.md (tilepro_expand_prologue): Likewise.
39301 2017-02-03  Jakub Jelinek  <jakub@redhat.com>
39303         PR target/79354
39304         * config/rs6000/rs6000.md (movsi_from_sf): Use wb constraint instead of
39305         wu for stxssp alternative.
39307 2017-02-03  Martin Sebor  <msebor@redhat.com>
39309         PR tree-optimization/79352
39310         * gimple-fold.c (get_range_strlen): Add argument.
39311         (get_range_strlen): Change return type to bool.
39312         (get_maxval_strlen): Pass in a dummy argument.
39313         * gimple-fold.h (get_range_strlen): Change return type to bool.
39314         * gimple-ssa-sprintf.c (get_string_length): Set unlikely counter.
39315         * tree.h (array_at_struct_end_p): Add argument.
39316         * tree.c (array_at_struct_end_p): Handle it.
39318 2017-02-03  Martin Liska  <mliska@suse.cz>
39320         PR lto/66295
39321         * multiple_target.c (create_dispatcher_calls): Redirect edge
39322         from a caller of a dispatcher.
39323         (expand_target_clones): Make the clones local.
39324         (ipa_target_clone): Do both target clones and resolvers.
39325         (ipa_dispatcher_calls): Remove the pass.
39326         (pass_dispatcher_calls::gate): Likewise.
39327         (make_pass_dispatcher_calls): Likewise.
39328         * passes.def (pass_target_clone): Put as very first IPA early
39329         pass.
39331 2017-02-03  Martin Liska  <mliska@suse.cz>
39333         * symtab.c (symtab_node::binds_to_current_def_p): Bail out
39334         in case of a function with ifunc attribute.
39336 2017-02-03  Martin Liska  <mliska@suse.cz>
39338         * cgraph.c (cgraph_node::dump): Dump function version info.
39339         * symtab.c (symtab_node::dump_base): Add missing new line.
39341 2017-02-02  Jan Hubicka  <hubicka@ucw.cz>
39343         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): New function.
39344         (ifcombine_ifandif): Use it.
39346 2017-02-03  Martin Liska  <mliska@suse.cz>
39348         * doc/invoke.texi: Document default value for
39349         use-after-scope-direct-emission-threshold.
39351 2017-02-03  Martin Liska  <mliska@suse.cz>
39353         PR tree-optimization/79339
39354         * gimple-ssa-sprintf.c (format_floating_max): Call mpfr_clear.
39355         (format_floating): Likewise.
39357 2017-02-03  Martin Liska  <mliska@suse.cz>
39359         PR ipa/79337
39360         * ipa-prop.c (ipa_node_params_t::insert): Remove current
39361         implementation.
39362         (ipa_node_params_t::remove): Likewise.
39363         * ipa-prop.h (ipa_node_params::ipa_node_params): Make default
39364         initialization from removed ipa_node_params_t::insert.
39365         (ipa_node_params::~ipa_node_params): Move from removed
39366         ipa_node_params_t::release.
39367         * symbol-summary.h (symbol_summary::m_released): New member.
39368         Do not release a summary twice.  Do not allow to call finalizer
39369         for types of a summary that live in GGC memory.
39371 2017-02-02  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
39373         * config/aarch64/aarch64.c (thunderx2t99_tunings): Enable AES and
39374         cmp_branch fusion.
39376 2017-02-02  Martin Sebor  <msebor@redhat.com>
39378         PR middle-end/79275
39379         * gimple-ssa-sprintf.c (get_string_length): Set lower bound to zero.
39380         (format_string): Tighten up the range of output for non-constant
39381         strings and correct the expected range for wide non-constant strings.
39383 2017-02-02  Martin Sebor  <msebor@redhat.com>
39385         * doc/invoke.texi (-maccumulate-args): Fix bad grammar.
39387         PR middle-end/32003
39388         * doc/invoke.texi (-fdump-tree-): Remove pass-specific options from
39389         index.
39390         (-fdump-tree-@var): Add to index and document how to come up
39391         with pass-specific option and dump file names.
39392         (-fdump-passes): Clarify where to look for output.
39394 2017-02-02  Jan Hubicka  <hubicka@ucw.cz>
39396         PR middle-end/77445
39397         * tree-ssa-threadbackward.c (profitable_jump_thread_path): Dump
39398         statistics of the analyzed path; allow threading for speed when
39399         any of BBs along the path are optimized for speed.
39401 2017-02-02  Eric Botcazou  <ebotcazou@adacore.com>
39403         PR middle-end/78468
39404         * emit-rtl.c (init_emit): Add ??? comment for problematic alignment
39405         settings of the virtual registers.
39407         Revert again
39408         2016-08-23  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39410         * explow.c (get_dynamic_stack_size): Take known alignment of stack
39411         pointer + STACK_DYNAMIC_OFFSET into account when calculating the size
39412         needed.
39414 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39416         * config/s390/vx-builtins.md ("vec_ceil", "vec_floor")
39417         ("vec_trunc", "vec_roundc", "vec_round"): Remove expanders.
39419 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39421         * config/s390/s390.md: Add missing comments with the expanded
39422         mnemonics.
39423         * config/s390/vector.md: Likewise.
39424         * config/s390/vx-builtins.md: Likewise.
39426 2017-02-02  Jakub Jelinek  <jakub@redhat.com>
39428         PR target/79197
39429         * config/rs6000/rs6000.md (*fixuns_trunc<mode>di2_fctiduz): Rename to ...
39430         (fixuns_trunc<mode>di2): ... this, remove previous expander.  Put all
39431         conditions on a single line.
39433 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39435         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Rename
39436         __S390_VX__ to __VX__.
39438 2017-02-01  Andrew Pinski  <apinski@cavium.com>
39440         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost): Pass
39441         stmt_info to record_stmt_cost.
39442         (vect_get_known_peeling_cost): Pass stmt_info if known to
39443         record_stmt_cost.
39444         * config/aarch64/aarch64-protos.h (cpu_vector_cost): Split
39445         cpu_vector_cost field into
39446         scalar_int_stmt_cost and scalar_fp_stmt_cost.  Split vec_stmt_cost
39447         field into vec_int_stmt_cost and vec_fp_stmt_cost.
39448         * config/aarch64/aarch64.c (generic_vector_cost): Update for the
39449         splitting of scalar_stmt_cost and vec_stmt_cost.
39450         (thunderx_vector_cost): Likewise.
39451         (cortexa57_vector_cost): LIkewise.
39452         (exynosm1_vector_cost): Likewise.
39453         (xgene1_vector_cost): Likewise.
39454         (thunderx2t99_vector_cost): Improve after the splitting of the two
39455         fields.
39456         (aarch64_builtin_vectorization_cost): Update for the splitting of
39457         scalar_stmt_cost and vec_stmt_cost.
39459 2017-02-01  Torvald Riegel  <triegel@redhat.com>
39460             Richard Henderson  <rth@redhat.com>
39462         * builtins.c (fold_builtin_atomic_always_lock_free): Make "lock-free"
39463         conditional on existance of a fast atomic load.
39464         * optabs-query.c (can_atomic_load_p): New function.
39465         * optabs-query.h (can_atomic_load_p): Declare it.
39466         * optabs.c (expand_atomic_exchange): Always delegate to libatomic if
39467         no fast atomic load is available for the particular size of access.
39468         (expand_atomic_compare_and_swap): Likewise.
39469         (expand_atomic_load): Likewise.
39470         (expand_atomic_store): Likewise.
39471         (expand_atomic_fetch_op): Likewise.
39472         * testsuite/lib/target-supports.exp
39473         (check_effective_target_sync_int_128): Remove x86 because it provides
39474         no fast atomic load.
39475         (check_effective_target_sync_int_128_runtime): Likewise.
39477 2017-02-01  Richard Biener  <rguenther@suse.de>
39479         * graphite.c: Include tree-vectorizer.h for find_loop_location.
39480         (graphite_transform_loops): Provide opt-info for optimized nests.
39481         * tree-parloop.c (parallelize_loops): Provide opt-info for
39482         parallelized loops.
39484 2017-02-01  Richard Biener  <rguenther@suse.de>
39486         PR middle-end/79315
39487         * tree-cfg.c (move_stmt_op): Never set TREE_BLOCK when it
39488         was not set before.
39490 2017-02-01  Richard Biener  <rguenther@suse.de>
39492         PR tree-optimization/71824
39493         * graphite-scop-detection.c (scop_detection::build_scop_breadth):
39494         Verify the loops are valid in the merged SESE region.
39495         (scop_detection::can_represent_loop_1): Check analyzing the
39496         evolution of the number of iterations in the region succeeds.
39498 2017-01-31  Ian Lance Taylor  <iant@golang.org>
39500         * config/i386/i386.c (ix86_expand_split_stack_prologue): Add
39501         REG_ARGS_SIZE note to 32-bit push insns and call insn.
39503 2017-01-31  David Malcolm  <dmalcolm@redhat.com>
39505         PR preprocessor/79210
39506         * input.c (get_substring_ranges_for_loc): Replace line_width
39507         assertion with error-handling.
39509 2017-01-31  Richard Biener  <rguenther@suse.de>
39511         PR tree-optimization/77318
39512         * graphite-sese-to-poly.c (extract_affine): Fix assert.
39513         (create_pw_aff_from_tree): Take loop parameter.
39514         (add_condition_to_pbb): Pass loop of the condition to
39515         create_pw_aff_from_tree.
39517 2017-01-31  Jakub Jelinek  <jakub@redhat.com>
39519         * config/s390/s390.c (s390_asan_shadow_offset): New function.
39520         (TARGET_ASAN_SHADOW_OFFSET): Redefine.
39522 2017-01-31  Michael Meissner  <meissner@linux.vnet.ibm.com>
39524         PR target/78597
39525         PR target/79038
39526         * config/rs6000/rs6000-protos.h (convert_float128_to_int): Delete,
39527         no longer used.
39528         (convert_int_to_float128): Likewise.
39529         * config/rs6000/rs6000.c (convert_float128_to_int): Likewise.
39530         (convert_int_to_float128): Likewise.
39531         * config/rs6000/rs6000.md (UNSPEC_IEEE128_MOVE): Likewise.
39532         (UNSPEC_IEEE128_CONVERT): Likewise.
39533         (floatsi<mode>2, FLOAT128 iterator): Bypass calling
39534         rs6000_expand_float128_convert if we have IEEE 128-bit hardware.
39535         Use local variables for IBM extended format.
39536         (fix_trunc<mode>si2, FLOAT128 iterator): Likewise.
39537         (fix_trunc<mode>si2_fprs): Likewise.
39538         (fixuns_trunc<IEEE128:mode><SDI:mode>2): Likewise.
39539         (floatuns<IEEE128:mode>2, IEEE128 iterator): Likewise.
39540         (fix<uns>_<mode>si2_hw): Rework the IEEE 128-bt hardware support
39541         to know that we can now have integers of all sizes in vector
39542         registers.
39543         (fix<uns>_<mode>di2_hw): Likewise.
39544         (float<uns>_<mode>si2_hw): Likewise.
39545         (fix_<mode>si2_hw): Likewise.
39546         (fixuns_<mode>si2_hw): Likewise.
39547         (float<uns>_<mode>di2_hw): Likewise.
39548         (float_<mode>di2_hw): Likewise.
39549         (float_<mode>si2_hw): Likewise.
39550         (floatuns_<mode>di2_hw): Likewise.
39551         (floatuns_<mode>si2_hw): Likewise.
39552         (xscvqp<su>wz_<mode>): Delete, no longer used.
39553         (xscvqp<su>dz_<mode>): Likewise.
39554         (xscv<su>dqp_<mode>): Likewise.
39555         (ieee128_mfvsrd_64bit): Likewise.
39556         (ieee128_mfvsrd_32bit): Likewise.
39557         (ieee128_mfvsrwz): Likewise.
39558         (ieee128_mtvsrw): Likewise.
39559         (ieee128_mtvsrd_64bit): Likewise.
39560         (ieee128_mtvsrd_32bit): Likewise.
39562 2017-01-31  Martin Liska  <mliska@suse.cz>
39564         PR ipa/79285
39565         * ipa-prop.c (ipa_free_all_node_params): Call release method
39566         instead of ~sumbol_summary to not to trigger double times
39567         dtor of hash_map.
39569 2017-01-31  Aldy Hernandez  <aldyh@redhat.com>
39571         PR tree-optimization/71691
39572         * bitmap.h (class auto_bitmap): New.
39573         * tree-ssa-loop-unswitch.c (tree_may_unswitch_on): Call
39574         is_maybe_undefined instead of ssa_undefined_value_p.
39576 2017-01-31  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39578         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Rename
39579         __S390_ARCH_LEVEL__ to __ARCH__.
39581 2017-01-31  Jakub Jelinek  <jakub@redhat.com>
39583         PR tree-optimization/79267
39584         * value-prof.c (gimple_ic): Only drop lhs for noreturn calls
39585         if should_remove_lhs_p is true.
39587 2017-01-30  Alexandre Oliva  <aoliva@redhat.com>
39589         PR debug/63238
39590         * dwarf2out.c (clone_as_declaration): Drop DW_AT_alignment.
39591         (add_alignment_attribute): New.
39592         (base_type_die): Add alignment attribute.
39593         (subrange_type_die): Likewise.
39594         (modified_type_die): Likewise.
39595         (gen_array_type_die): Likewise.
39596         (gen_descr_array_type_die: Likewise.
39597         (gen_enumeration_type_die): Likewise.
39598         (gen_subprogram_die): Likewise.
39599         (gen_variable_die): Likewise.
39600         (gen_field_die): Likewise.
39601         (gen_ptr_to_mbr_type_die): Likewise.
39602         (gen_struct_or_union_type_die): Likewise.
39603         (gen_subroutine_type_die): Likewise.
39604         (gen_typedef_die): Likewise.
39605         (base_type_cmp): Compare alignment attribute.
39607 2017-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
39609         PR target/79170
39610         * config/rs6000/altivec.md (*setb_internal): Rename to setb_signed.
39611         (setb_unsigned) New pattern for setb with CCUNS.
39612         * config/rs6000/rs6000.c (expand_block_compare): Use a different
39613         subfc./subfe sequence to avoid overflow problems.  Generate a
39614         shorter sequence with cmpld/setb for power9.
39615         * config/rs6000/rs6000.md (subf<mode>3_carry_dot2): Add a new pattern
39616         for generating subfc. instruction.
39617         (cmpstrsi): Add TARGET_POPCNTD predicate as the generate sequence
39618         now uses this instruction.
39620 2017-01-30  Ian Lance Taylor  <iant@google.com>
39622         PR debug/79289
39623         * dwarf2out.c (gen_type_die_with_usage): When picking a variant
39624         for FUNCTION_TYPE/METHOD_TYPE, use the first matching one.
39626 2017-01-30  Martin Sebor  <msebor@redhat.com>
39628         * gimple-ssa-sprintf.c (fmtresult::adjust_for_width_or_precision):
39629         Move constant to the right of a relational operator.
39630         (get_mpfr_format_length, format_character, format_string): Ditto.
39631         (should_warn_p, maybe_warn): Same.
39633         * doc/invoke.texi (-Wformat-truncation=1): Fix typo.
39635 2017-01-30  Maxim Ostapenko  <m.ostapenko@samsung.com>
39637         PR lto/79061
39638         * asan.c (get_translation_unit_decl): Remove function.
39639         (asan_add_global): Force has_dynamic_init to zero in LTO mode.
39641 2017-01-30  Martin Liska  <mliska@suse.cz>
39643         PR gcov-profile/79259
39644         * opts.c (common_handle_option): Enable flag_ipa_bit_cp w/
39645         -fprofile-generate.
39647 2017-01-30  Martin Liska  <mliska@suse.cz>
39649         PR bootstrap/78985
39650         * config/aarch64/cortex-a57-fma-steering.c (func_fma_steering::analyze):
39651         Initialize variables with NULL value.
39653 2017-01-30  Richard Earnshaw  <rearnsh@arm.com>
39655         PR target/79260
39656         * config.gcc (arm*-*-*): Add arm/arm-flags.h and arm/arm-isa.h to
39657         tm_p_file.
39658         * arm/arm-protos.h: Don't directly include arm-flags.h and arm-isa.h.
39660 2017-01-30  Richard Biener  <rguenther@suse.de>
39662         PR tree-optimization/79276
39663         * tree-vrp.c (process_assert_insertions): Properly adjust common
39664         when removing a duplicate.
39666 2017-01-30  Richard Biener  <rguenther@suse.de>
39668         PR tree-optimization/79256
39669         * targhooks.c (default_builtin_vector_alignment_reachable): Honor
39670         BIGGEST_FIELD_ALIGNMENT and ADJUST_FIELD_ALIGN to fix up bogus
39671         alignment on TYPE.
39672         * tree.c (build_aligned_type): Set TYPE_USER_ALIGN.
39674 2017-01-30  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39676         PR target/79240
39677         * config/s390/s390.md ("*r<noxa>sbg_<mode>_srl_bitmask")
39678         ("*r<noxa>sbg_<mode>_sll_bitmask")
39679         ("*extzv_<mode>_srl<clobbercc_or_nocc>")
39680         ("*extzv_<mode>_sll<clobbercc_or_nocc>"):
39681         Use contiguous_bitmask_nowrap_operand.
39683 2017-01-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
39685         PR target/79268
39686         * config/rs6000/altivec.h (vec_xl): Revise #define.
39687         (vec_xst): Likewise.
39689 2017-01-27  Uros Bizjak  <ubizjak@gmail.com>
39691         * config/i386/i386.c (print_reg): Use REGNO instead of true_regnum.
39693 2017-01-27  Bernd Schmidt  <bschmidt@redhat.com>
39695         PR rtl-optimization/79194
39696         * cprop.c (one_cprop_pass): Move deletion of code after unconditional
39697         traps before call to bypass_conditional_jumps.
39699 2017-01-27  Vladimir Makarov  <vmakarov@redhat.com>
39701         PR tree-optimization/71374
39702         * lra-constraints.c (check_conflict_input_operands): New.
39703         (match_reload): Use it.
39705 2017-01-27  Vladimir Makarov  <vmakarov@redhat.com>
39707         PR target/79131
39708         * lra-assigns.c (find_hard_regno_for_1): Take endianess for into
39709         account to calculate conflict_set.
39711 2017-01-27  Bin Cheng  <bin.cheng@arm.com>
39713         PR rtl-optimization/78559
39714         * combine.c (try_combine): Discard REG_EQUAL and REG_EQUIV for
39715         other_insn in combine.
39717 2017-01-27  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
39719         * builtin-types.def: Use unsigned_char_type_node for BT_UINT8.  Use
39720         uint16_type_node for BT_UINT16.
39722 2017-01-27  David Malcolm  <dmalcolm@redhat.com>
39724         * doc/sourcebuild.texi (Testsuites): Add "GIMPLE Tests" and
39725         "RTL Tests" to menu.
39726         (GIMPLE Tests): New node.
39727         (RTL Tests): New node.
39729 2017-01-27  Richard Biener  <rguenther@suse.de>
39731         PR tree-optimization/79245
39732         * tree-loop-distribution.c (distribute_loop): Apply cost
39733         modeling also to detected patterns.
39735 2017-01-27  Richard Biener  <rguenther@suse.de>
39737         PR tree-optimization/71433
39738         * tree-vrp.c (register_new_assert_for): Revert earlier changes.
39739         (compare_assert_loc): New function.
39740         (process_assert_insertions): Sort and optimize assert locations
39741         to remove duplicates and push down identical assertions on
39742         edges to their destination block.
39744 2017-01-27  Richard Biener  <rguenther@suse.de>
39746         PR tree-optimization/79244
39747         * tree-vrp.c (remove_range_assertions): Forcefully propagate
39748         out SSA names even if abnormal.
39750 2017-01-27  Jakub Jelinek  <jakub@redhat.com>
39752         * realmpfr.h: Poison MPFR_RND{N,Z,U,D}.
39753         * gimple-ssa-sprintf.c (format_floating_max): Use GMP_RNDN
39754         instead of MPFR_RNDN.
39756 2017-01-27  Richard Earnshaw  <rearnsha@arm.com>
39758         PR target/79239
39759         * arm.c (arm_option_override): Don't call build_target_option_node
39760         until after doing all option overrides.
39761         (arm_valid_target_attribute_tree): Likewise.
39763 2017-01-27  Martin Liska  <mliska@suse.cz>
39765         * doc/invoke.texi (-fprofile-arcs): Document profiling support
39766         for {cd}tors and C++ {cd}tors.
39768 2017-01-27  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39770         * config/s390/s390.md ("*setmem_long_and")
39771         ("*setmem_long_and_31z"): Use zero_extend instead of and.
39773 2017-01-26  Martin Sebor  <msebor@redhat.com>
39775         * gimple-ssa-sprintf.c (format_floating): Simplify the computation
39776         of precision.
39778 2017-01-26  Martin Sebor  <msebor@redhat.com>
39780         * gimple-ssa-sprintf.c (format_floating): Test HAVE_XFmode and
39781         HAVE_DFmode before using XFmode or DFmode.
39782         (parse_directive): Avoid using the z length modifier to avoid
39783         the ISO C++98 does not support the â€˜z’ gnu_printf length modifier.
39785         PR middle-end/78703
39786         * gimple-ssa-sprintf.c (adjust_for_width_or_precision): Change
39787         to accept adjustment as an array.
39788         (get_int_range): New function.
39789         (struct directive): Make width and prec arrays.
39790         (directive::set_width, directive::set_precision): Call get_int_range.
39791         (format_integer, format_floating): Handle width and precision ranges.
39792         (format_string, parse_directive): Same.
39794 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39796         PR debug/79129
39797         * dwarf2out.c (generate_skeleton_bottom_up): For children with
39798         comdat_type_p set, just clone them, but keep the children in the
39799         original DIE.
39801         PR debug/78835
39802         * dwarf2out.c (prune_unused_types): Mark all functions with DIEs
39803         which have direct callers with -fvar-tracking-assignments enabled
39804         in the current TU.
39805         (resolve_addr): Avoid adding skeleton DIEs for DW_AT_call_origin
39806         inside of type units.
39808 2017-01-26  Martin Sebor  <msebor@redhat.com>
39810         PR middle-end/78703
39811         * gimple-ssa-sprintf.c (struct result_range): Add likely and
39812         unlikely counters.
39813         (struct format_result): Replace number_chars, number_chars_min,
39814         and number_chars_max with a single member of struct result_range.
39815         Remove bounded.
39816         (format_result::operator+=): Adjust.
39817         (struct fmtresult): Remove bounded.  Handle likely and unlikely
39818         counters.
39819         (fmtresult::adjust_for_width_or_precision): New function.
39820         (fmtresult:type_max_digits): New function.
39821         (bytes_remaining): Handle likely and unlikely counters.
39822         (min_bytes_remaining): Remove.
39823         (format_percent): Simplify.
39824         (format_integer, format_floating): Set likely and unlikely counters.
39825         (get_string_length, format_character, format_string): Same.
39826         (format_plain, should_warn_p): New function.
39827         (maybe_warn): Call should_warn_p.  Update diagnostic messages
39828         and handle those for all directives, including plain strings.
39829         (format_directive): Handle likely and unlikely counters.
39830         Remove unnecessary quoting from diagnostics.  Add an informational
39831         note.
39832         (add_bytes): Remove.
39833         (pass_sprintf_length::compute_format_length): Simplify.
39834         (try_substitute_return_value): Handle likely and unlikely counters.
39836 2017-01-26  Carl Love  <cel@us.ibm.com>
39838         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Remove
39839         bogus entries for the P8V_BUILTIN_VEC_VGBBD built-ins
39841 2017-01-26  Vladimir Makarov  <vmakarov@redhat.com>
39843         PR target/79131
39844         * lra-assigns.c (setup_live_pseudos_and_spill_after_risky): Take
39845         endianess for subregs into account.
39846         * lra-constraints.c (lra_constraints): Do risky transformations
39847         always on the first iteration.
39848         * lra-lives.c (check_pseudos_live_through_calls): Add arg
39849         last_call_used_reg_set.
39850         (process_bb_lives): Define and use last_call_used_reg_set.
39851         * lra.c (lra): Always continue after lra_constraints on the first
39852         iteration.
39854 2017-01-26  Kirill Yukhin  <kirill.yukhin@gmail.com>
39856         * gcc.target/i386/avx512bw-kshiftlq-2.c: Use unsigned long long
39857         constant.
39858         * gcc.target/i386/avx512bw-kshiftrq-2.c: Ditto.
39860 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39862         * config/i386/avx512fintrin.h (_ktest_mask16_u8,
39863         _ktestz_mask16_u8, _ktestc_mask16_u8, _kadd_mask16): Move to ...
39864         * config/i386/avx512dqintrin.h (_ktest_mask16_u8,
39865         _ktestz_mask16_u8, _ktestc_mask16_u8, _kadd_mask16): ... here.
39866         * config/i386/i386-builtin.def (__builtin_ia32_ktestchi,
39867         __builtin_ia32_ktestzhi, __builtin_ia32_kaddhi): Use
39868         OPTION_MASK_ISA_AVX512DQ instead of OPTION_MASK_ISA_AVX512F.
39869         * config/i386/sse.md (SWI1248_AVX512BWDQ2): New mode iterator.
39870         (kadd<mode>, ktest<mode>): Use it instead of SWI1248_AVX512BWDQ.
39872 2017-01-26  Marek Polacek  <polacek@redhat.com>
39874         PR c/79199
39875         * fold-const.c (operand_equal_p) [COND_EXPR]: Use OP_SAME_WITH_NULL
39876         for the third operand.
39878 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39880         PR middle-end/79236
39881         * omp-low.c (struct omp_context): Add simt_stmt field.
39882         (scan_omp_for): Return omp_context *.
39883         (scan_omp_simd): Set simt_stmt on the non-_simt_ SIMD
39884         context to the _simt_ SIMD stmt.
39885         (lower_omp_for): For combined SIMD with sibling _simt_
39886         SIMD, make sure to use the same decls in _looptemp_
39887         clauses as in the sibling.
39889 2017-01-26  David Sherwood  <david.sherwood@arm.com>
39891         PR middle-end/79212
39892         * gimplify.c (omp_notice_variable): Add GOVD_SEEN flag to variables in
39893         all contexts.
39895 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39897         PR target/70465
39898         * reg-stack.c (emit_swap_insn): Instead of fld a; fld b; fxchg %st(1);
39899         emit fld b; fld a; if possible.
39901         * brig-builtins.def: Update copyright years.
39902         * config/arm/arm_acle_builtins.def: Update copyright years.
39904 2017-01-25  Michael Meissner  <meissner@linux.vnet.ibm.com>
39906         PR target/79179
39907         * config/rs6000/vsx.md (vsx_extract_<mode>_store): Use wY
39908         constraint instead of o for the stxsd instruction.
39910 2017-01-25  Carl Love  <cel@us.ibm.com>
39912         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Fix order
39913         of entries for ALTIVEC_BUILTIN_VEC_PACKS and P8V_BUILTIN_VEC_VGBBD.
39915 2017-01-25  Jonathan Wakely  <jwakely@redhat.com>
39917         * doc/invoke.texi (C++ Dialect Options): Fix typo.
39919 2017-01-25  Richard Biener  <rguenther@suse.de>
39921         PR tree-optimization/69264
39922         * target.def (vector_alignment_reachable): Improve documentation.
39923         * doc/tm.texi: Regenerate.
39924         * targhooks.c (default_builtin_vector_alignment_reachable): Simplify
39925         and add a comment.
39926         * tree-vect-data-refs.c (vect_supportable_dr_alignment): Revert
39927         earlier changes with respect to TYPE_USER_ALIGN.
39928         (vector_alignment_reachable_p): Likewise.  Improve dumping.
39930 2016-01-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
39932         PR target/79145
39933         * config/arm/arm.md (xordi3): Force constant operand into a register
39934         for TARGET_IWMMXT.
39936 2016-01-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
39938         * doc/invoke.texi (-fstore-merging): Correct default optimization
39939         levels at which it is enabled.
39940         (-O): Move -fstore-merging from list to...
39941         (-O2): ... Here.
39943 2017-01-25  Richard Biener  <rguenther@suse.de>
39945         PR debug/78363
39946         * omp-expand.c: Include debug.h.
39947         (expand_omp_taskreg): Make sure to generate early debug before
39948         outlining anything from a function.
39949         (expand_omp_target): Likewise.
39950         (grid_expand_target_grid_body): Likewise.
39952 2017-01-25  Maxim Ostapenko  <m.ostapenko@samsung.com>
39954         PR lto/79061
39955         * asan.c (get_translation_unit_decl): New function.
39956         (asan_add_global): Extract modules file name from globals
39957         TRANSLATION_UNIT_DECL name.
39959 2017-01-24  Eric Botcazou  <ebotcazou@adacore.com>
39961         PR target/77439
39962         * config/arm/arm.c (arm_function_ok_for_sibcall): Add back restriction
39963         for long calls with APCS frame and VFP.
39965 2017-01-24  David Malcolm  <dmalcolm@redhat.com>
39967         * cfg.c (original_copy_tables_initialized_p): New function.
39968         * cfg.h (original_copy_tables_initialized_p): New decl.
39969         * cfgrtl.c (relink_block_chain): Guard the call to
39970         free_original_copy_tables with a call to
39971         original_copy_tables_initialized_p.
39972         * cgraph.h (symtab_node::native_rtl_p): New decl.
39973         * cgraphunit.c (symtab_node::native_rtl_p): New function.
39974         (symtab_node::needed_p): Don't assert for early assembly output
39975         for __RTL functions.
39976         (cgraph_node::finalize_function): Set "force_output" for __RTL
39977         functions.
39978         (cgraph_node::analyze): Bail out early for __RTL functions.
39979         (analyze_functions): Update assertion to support __RTL functions.
39980         (cgraph_node::expand): Bail out early for __RTL functions.
39981         * final.c (rest_of_clean_state): Don't call delete_tree_ssa for
39982         __RTL functions.
39983         * function.h (struct function): Update comment for field
39984         "pass_startwith".
39985         * gimple-expr.c: Include "tree-pass.h".
39986         (gimple_has_body_p): Return false for __RTL functions.
39987         * Makefile.in (OBJS): Add run-rtl-passes.o.
39988         * pass_manager.h (gcc::pass_manager::get_rest_of_compilation): New
39989         accessor.
39990         (gcc::pass_manager::get_clean_slate): New accessor.
39991         * passes.c: Include "insn-addr.h".
39992         (should_skip_pass_p): Add logging.  Update logic for running
39993         "expand" to be compatible with both __GIMPLE and __RTL.  Guard
39994         property-provider override so it is only done for gimple passes.
39995         Don't skip dfinit.
39996         (skip_pass): New function.
39997         (execute_one_pass): Call skip_pass when skipping passes.
39998         * read-md.c (md_reader::read_char): Support filtering
39999         the input to a subset of line numbers.
40000         (md_reader::md_reader): Initialize fields
40001         m_first_line and m_last_line.
40002         (md_reader::read_file_fragment): New function.
40003         * read-md.h (md_reader::read_file_fragment): New decl.
40004         (md_reader::m_first_line): New field.
40005         (md_reader::m_last_line): New field.
40006         * read-rtl-function.c (function_reader::create_function): Only
40007         create cfun if it doesn't already exist.  Set PROP_rtl on cfun's
40008         curr_properties.  Set DECL_INITIAL to a dummy block.
40009         (read_rtl_function_body_from_file_range): New function.
40010         * read-rtl-function.h (read_rtl_function_body_from_file_range):
40011         New decl.
40012         * run-rtl-passes.c: New file.
40013         * run-rtl-passes.h: New file.
40015 2017-01-24  Jeff Law  <law@redhat.com>
40017         * config/microblaze/microblaze.h (ASM_FORMAT_PRIVATE_NAME): Increase
40018         buffer size.
40020 2017-01-24  Bin Cheng  <bin.cheng@arm.com>
40022         PR tree-optimization/79159
40023         * tree-ssa-loop-niter.c (get_cst_init_from_scev): New function.
40024         (record_nonwrapping_iv): Improve boundary using above function if no
40025         value range information.
40027 2017-01-24  Pekka Jääskeläinen  <pekka@parmance.com>
40028             Martin Jambor  <mjambor@suse.cz>
40030         * brig-builtins.def: New file.
40031         * builtins.def (DEF_HSAIL_BUILTIN): New macro.
40032         (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
40033         (DEF_HSAIL_SAT_BUILTIN): Likewise.
40034         (DEF_HSAIL_INTR_BUILTIN): Likewise.
40035         (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
40036         * builtin-types.def (BT_INT8): New.
40037         (BT_INT16): Likewise.
40038         (BT_UINT8): Likewise.
40039         (BT_UINT16): Likewise.
40040         (BT_FN_ULONG): Likewise.
40041         (BT_FN_UINT_INT): Likewise.
40042         (BT_FN_UINT_ULONG): Likewise.
40043         (BT_FN_UINT_LONG): Likewise.
40044         (BT_FN_UINT_PTR): Likewise.
40045         (BT_FN_ULONG_PTR): Likewise.
40046         (BT_FN_INT8_FLOAT): Likewise.
40047         (BT_FN_INT16_FLOAT): Likewise.
40048         (BT_FN_UINT32_FLOAT): Likewise.
40049         (BT_FN_UINT16_FLOAT): Likewise.
40050         (BT_FN_UINT8_FLOAT): Likewise.
40051         (BT_FN_UINT64_FLOAT): Likewise.
40052         (BT_FN_UINT16_UINT32): Likewise.
40053         (BT_FN_UINT32_UINT16): Likewise.
40054         (BT_FN_UINT16_UINT16_UINT16): Likewise.
40055         (BT_FN_INT_PTR_INT): Likewise.
40056         (BT_FN_UINT_PTR_UINT): Likewise.
40057         (BT_FN_LONG_PTR_LONG): Likewise.
40058         (BT_FN_ULONG_PTR_ULONG): Likewise.
40059         (BT_FN_VOID_UINT64_UINT64): Likewise.
40060         (BT_FN_UINT8_UINT8_UINT8): Likewise.
40061         (BT_FN_INT8_INT8_INT8): Likewise.
40062         (BT_FN_INT16_INT16_INT16): Likewise.
40063         (BT_FN_INT_INT_INT): Likewise.
40064         (BT_FN_UINT_FLOAT_UINT): Likewise.
40065         (BT_FN_FLOAT_UINT_UINT): Likewise.
40066         (BT_FN_ULONG_UINT_UINT): Likewise.
40067         (BT_FN_ULONG_UINT_PTR): Likewise.
40068         (BT_FN_ULONG_ULONG_ULONG): Likewise.
40069         (BT_FN_UINT_UINT_UINT): Likewise.
40070         (BT_FN_VOID_UINT_PTR): Likewise.
40071         (BT_FN_UINT_UINT_PTR: Likewise.
40072         (BT_FN_UINT32_UINT64_PTR): Likewise.
40073         (BT_FN_INT_INT_UINT_UINT): Likewise.
40074         (BT_FN_UINT_UINT_UINT_UINT): Likewise.
40075         (BT_FN_UINT_UINT_UINT_PTR): Likewise.
40076         (BT_FN_UINT_ULONG_ULONG_UINT): Likewise.
40077         (BT_FN_ULONG_ULONG_ULONG_ULONG): Likewise.
40078         (BT_FN_LONG_LONG_UINT_UINT): Likewise.
40079         (BT_FN_ULONG_ULONG_UINT_UINT): Likewise.
40080         (BT_FN_VOID_UINT32_UINT64_PTR): Likewise.
40081         (BT_FN_VOID_UINT32_UINT32_PTR): Likewise.
40082         (BT_FN_UINT_UINT_UINT_UINT_UINT): Likewise.
40083         (BT_FN_UINT_FLOAT_FLOAT_FLOAT_FLOAT): Likewise.
40084         (BT_FN_ULONG_ULONG_ULONG_UINT_UINT): Likewise.
40085         * doc/frontends.texi: List BRIG FE.
40086         * doc/install.texi (Testing): Add BRIG tesring requirements.
40087         * doc/invoke.texi (Overall Options): Mention BRIG.
40088         * doc/standards.texi (Standards): Doucment BRIG HSA version.
40090 2017-01-24  Richard Biener  <rguenther@suse.de>
40092         PR translation/79208
40093         * ipa-devirt.c (odr_types_equivalent_p): Fix typo in diagnostic.
40095 2017-01-24  Martin Jambor  <mjambor@suse.cz>
40097         PR bootstrap/79198
40098         * ipa-prop.c (ipa_free_all_node_params): Call summary destructor.
40099         * ipa-prop.c (ipa_node_params_t::insert): Initialize fields known_csts
40100         and known_contexts.
40102 2017-01-24  Aldy Hernandez  <aldyh@redhat.com>
40104         PR middle-end/79123
40105         * gimple-ssa-warn-alloca.c (alloca_call_type): Make sure
40106         casts from signed to unsigned really don't have a range.
40108 2017-01-24  Markus Trippelsdorf  <markus@trippelsdorf.de>
40110         * gimple-ssa-sprintf.c (format_floating): Change MPFR_RNDx to
40111         GMP_RNDx for compatiblity.
40113 2017-01-24  Martin Liska  <mliska@suse.cz>
40115         PR bootstrap/79132
40116         * tree-ssa-reassoc.c (rewrite_expr_tree_parallel): Insert assert
40117         that would prevent us to call alloca with -1 as argument.
40119 2017-01-24  Jakub Jelinek  <jakub@redhat.com>
40121         * dwarf2out.c (output_compilation_unit_header, output_file_names):
40122         Avoid -Wformat-security warning.
40124 2017-01-23  Andrew Pinski  <apinski@cavium.com>
40126         * config/aarch64/aarch64.c (thunderx2t99_addrcost_table): Improve
40127         cost table.
40129 2017-01-23  Martin Sebor  <msebor@redhat.com>
40131         PR middle-end/78703
40132         * gimple-ssa-sprintf.c (warn_level): New global.
40133         (format_integer): Use it here and throughout the rest of the file.
40134         Use the same switch to compute sign as base.
40135         (maybe_warn): New function.
40136         (format_directive): Factor out warnings into maybe_warn.
40137         Add debugging output.  Use warn_level.
40138         (add_bytes): Use warn_level.
40139         (pass_sprintf_length::compute_format_length): Add debugging output.
40140         (try_substitute_return_value): Same.
40141         (pass_sprintf_length::handle_gimple_call): Set and use warn_level.
40143         PR middle-end/78703
40144         * gimple-ssa-sprintf.c (struct format_result): Remove constant member.
40145         (struct fmtresult, format_integer, format_floating): Adjust.
40146         (fmtresult::fmtresult): Set max correctly in two argument ctor.
40147         (get_string_length, format_string,format_directive): Same.
40148         (pass_sprintf_length::compute_format_length): Same.
40149         (try_substitute_return_value): Simplify slightly.
40151         PR middle-end/78703
40152         * gimple-ssa-sprintf.c (pass_sprintf_length::gate): Adjust formatting.
40153         (fmtresult::operator+=): Outlined.
40154         (struct fmtresult): Add ctors.
40155         (struct conversion_spec): Rename...
40156         (struct directive): ...to this.  Add and remove data members.
40157         (directive::set_width, directive::set_precision): New functions.
40158         (format_percent): Use fmtresult ctor.
40159         (get_width_and_precision): Remove.
40160         (format_integer): Make naming changes.  Avoid computing width and
40161         precision.
40162         (format_floating): Same.  Adjust indentation.
40163         (format_character, format_none): New functions.
40164         (format_string): Moved character handling to format_character.
40165         (format_directive): Remove arguments, change return type.
40166         (parse_directive): New function.
40167         (pass_sprintf_length::compute_format_length): Move directive
40168         parsing to parse_directive.
40170 2017-01-23  Jakub Jelinek  <jakub@redhat.com>
40172         * tree.h (assign_assembler_name_if_neeeded): Rename to ...
40173         (assign_assembler_name_if_needed): ... this.
40174         * tree.c (assign_assembler_name_if_neeeded): Rename to ...
40175         (assign_assembler_name_if_needed): ... this.
40176         (free_lang_data_in_cgraph): Adjust callers.
40177         * cgraphunit.c (cgraph_node::analyze): Likewise.
40178         * omp-expand.c (expand_omp_taskreg, expand_omp_target):
40179         Likewise.
40181 2017-01-23  Richard Biener  <rguenther@suse.de>
40183         PR tree-optimization/79088
40184         PR tree-optimization/79188
40185         * tree-ssa-threadupdate.c (mark_threaded_blocks): Move code
40186         resetting loop bounds after last path deletion.  Reset loop
40187         bounds of the target loop, make code match the comments.
40188         * tree-ssa-threadbackwards.c (pass_early_thread_jumps::execute):
40189         Make sure loops need no fixups.
40191 2017-01-23  Kelvin Nilsen  <kelvin@gcc.gnu.org>
40193         * config/rs6000/rs6000-builtin.def (VSIEDPF): Add scalar insert
40194         exponent support with double type for first argument.
40195         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Changed
40196         type returned by __builtin_vec_extract_sig,
40197         __builtin_vec_extract_sig_sp, and __builtin_vec_extract_sig_dp
40198         functions from "vector int" to "vector unsigned int" or from
40199         "vector long long int" to "vector unsigned long long int".
40200         Changed type returned by __builtin_vec_extract_exp,
40201         __builtin_vec_extract_exp_sp, and __builtin_vec_extract_exp_dp
40202         functions from "vector int" to "vector unsigned int" or from
40203         "vector long long int" to "vector unsigned long long int".
40204         Changed return type of __builtin_vec_test_data_class,
40205         __builtin_vec_test_data_class_sp, and
40206         __builtin_vec_test_data_class_dp from "vector int" to
40207         "vector bool int" or from "vector long long int" to "vector bool
40208         long long int" and changed second argument type from "unsigned
40209         int" to "int".  Added new overloaded function forms "vector float
40210         __builtin_vec_insert_exp (vector float, vector unsigned int)" and
40211         "vector float __builtin_vec_insert_exp_sp (vector float, vector
40212         unsigned int)" and "vector double __builtin_vec_insert_exp (vector
40213         double, vector unsigned long long int)" and "vector double
40214         __builtin_vec_insert_exp_dp (vector double, vector unsigned long
40215         long int)".  Changed return type of
40216         __builtin_scalar_test_data_class and
40217         __builtin_scalar_test_data_class_sp and
40218         __builtin_scalar_test_data_class_dp from "unsigned int" to "bool
40219         int" and changed second argument from "unsigned int" to "int".
40220         Changed type returned by __builtin_scalar_test_neg,
40221         __builtin_scalar_test_neg_sp, and __builtin_scalar_test_neg_dp
40222         from "int" to "bool int".  Added new overloaded function form
40223         "double __builtin_scalar_insert_exp (double, unsigned long long int)".
40224         * config/rs6000/vsx.md (xsiexpdpf): New insn for scalar insert
40225         exponent double-precision with floating point first argument.
40226         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
40227         documentation of scalar_test_data_class, scalar_test_neg,
40228         scalar_extract_sig, scalar_extract_exp, scalar_insert_exp,
40229         vector_extract_exp, vec_extract_sig, vec_insert_exp, and
40230         vec_test_data_class built-in functions to reflect refinements in
40231         their type signatures.
40233 2017-01-23  Andreas Tobler  <andreast@gcc.gnu.org>
40235         * config/aarch64/aarch64.c (aarch64_elf_asm_constructor): Increase
40236         size of buf.
40237         (aarch64_elf_asm_destructor): Likewise.
40239 2017-01-23  Bernd Schmidt  <bschmidt@redhat.com>
40241         PR rtl-optimization/78634
40242         * config/i386/i386.c (ix86_max_noce_ifcvt_seq_cost): New function.
40243         (TARGET_MAX_NOCE_IFCVT_SEQ_COST): Define.
40244         * ifcvt.c (noce_try_cmove): Add missing cost check.
40246         PR rtl-optimization/71724
40247         * combine.c (if_then_else_cond): Look for situations where it is
40248         beneficial to undo the work of one of the recursive calls.
40250 2017-01-23  Bin Cheng  <bin.cheng@arm.com>
40252         PR tree-optimization/70754
40253         * tree-predcom.c (stmt_combining_refs): New parameter INSERT_BEFORE.
40254         (reassociate_to_the_same_stmt): New parameter INSERT_BEFORE.  Insert
40255         combined stmt before it if not NULL.
40256         (combine_chains): Process refs reversely and compute dominance point
40257         for root ref.
40259 2017-01-23  Martin Liska  <mliska@suse.cz>
40261         PR tree-optimization/79196
40262         * tree-ssa-strlen.c (fold_strstr_to_memcmp): Rename to ...
40263         (fold_strstr_to_strncmp): ... this.  Fold the pattern to strncmp
40264         instead of memcmp.
40265         (strlen_optimize_stmt): Call the renamed function.
40267 2017-01-23  Michael Matz  <matz@suse.de>
40269         PR tree-optimization/78384
40270         * tree-ssa-loop-split.c (patch_loop_exit): Use correct edge.
40272 2017-01-23  Richard Biener  <rguenther@suse.de>
40274         PR tree-optimization/79186
40275         * tree-vrp.c (register_new_assert_for): Make sure we've seen
40276         both incoming edges before moving an assert.
40278 2017-01-23  Martin Jambor  <mjambor@suse.cz>
40280         * ipa-prop.c (load_from_param_1): Removed.
40281         (load_from_unmodified_param): Bits from load_from_param_1 put back
40282         here.
40283         (load_from_param): Removed.
40284         (compute_complex_assign_jump_func): Removed stmt2 and just replaced it
40285         with stmt.  Reverted back to use of load_from_unmodified_param.
40287 2017-01-23  Martin Jambor  <mjambor@suse.cz>
40289         PR ipa/79108
40290         * ipa-prop.h (ipa_param_descriptor): Anotate with with GTY(()).
40291         (ipa_node_params): Annotate with GTY((for_user)).  Make descriptors
40292         field a pointer to garbage collected vector, mark lattices and
40293         ipcp_orig_node with GTY((skip)).
40294         (ipa_get_param_count): Adjust to descriptors being a pointer.
40295         (ipa_get_param): Likewise.
40296         (ipa_get_type): Likewise.
40297         (ipa_get_param_move_cost): Likewise.
40298         (ipa_set_param_used): Likewise.
40299         (ipa_get_controlled_uses): Likewise.
40300         (ipa_set_controlled_uses): Likewise.
40301         (ipa_is_param_used): Likewise.
40302         (ipa_node_params_t): Move into garbage collector.  New methods insert
40303         and remove.
40304         (ipa_node_params_sum): Annotate wth GTY(()).
40305         (ipa_check_create_node_params): Adjust to ipa_node_params_sum being
40306         garbage collected.
40307         (ipa_load_from_parm_agg): Adjust declaration.
40308         * ipa-icf.c (param_used_p): Adjust to descriptors being a pointer.
40309         * ipa-profile.c (ipa_profile): Likewise.
40310         * ipa-prop.c (ipa_get_param_decl_index_1): Likewise.
40311         (ipa_populate_param_decls): Make descriptors parameter garbage
40312         collected.
40313         (ipa_dump_param): Adjust to descriptors being a pointer.
40314         (ipa_alloc_node_params): Likewise.
40315         (ipa_initialize_node_params): Likewise.
40316         (load_from_param_1): Make descriptors parameter garbage collected.
40317         (load_from_unmodified_param): Likewise.
40318         (load_from_param): Likewise.
40319         (ipa_load_from_parm_agg): Likewise.
40320         (ipa_node_params::~ipa_node_params): Removed.
40321         (ipa_free_all_node_params): Remove call to delete operator.
40322         (ipa_node_params_t::insert): New.
40323         (ipa_node_params_t::remove): Likewise.
40324         (ipa_node_params_t::duplicate): Adjust to descriptors being a pointer,
40325         copy known_csts and known_contexts vectors.
40326         (ipa_read_node_info): Adjust to descriptors being a pointer.
40327         (ipcp_modif_dom_walker): Make m_descriptors field garbage
40328         collected.
40329         (ipcp_transform_function): Make descriptors variable garbage
40330         collected.
40332 2017-01-23  Andrew Senkevich  <andrew.senkevich@intel.com>
40334         * config/i386/avx512bwintrin.h: Add k-mask test, kortest intrinsics.
40335         * config/i386/avx512dqintrin.h: Ditto.
40336         * config/i386/avx512fintrin.h: Ditto.
40337         * config/i386/i386.c: Handle new builtins.
40338         * config/i386/i386-builtin.def: Add new builtins.
40339         * config/i386/sse.md (ktest<mode>, kortest<mode>): New.
40340         (UNSPEC_KORTEST, UNSPEC_KTEST): New.
40342 2017-01-23  Jakub Jelinek  <jakub@redhat.com>
40343             Martin Liska  <mliska@suse.cz>
40345         * asan.h: Define ASAN_USE_AFTER_SCOPE_ATTRIBUTE.
40346         * asan.c (asan_expand_poison_ifn): Support stores and use
40347         appropriate ASAN report function.
40348         * internal-fn.c (expand_ASAN_POISON_USE): New function.
40349         * internal-fn.def (ASAN_POISON_USE): Declare.
40350         * tree-into-ssa.c (maybe_add_asan_poison_write): New function.
40351         (maybe_register_def): Create ASAN_POISON_USE when sanitizing.
40352         * tree-ssa-dce.c (eliminate_unnecessary_stmts): Remove
40353         ASAN_POISON calls w/o LHS.
40354         * tree-ssa.c (execute_update_addresses_taken): Create clobber
40355         for ASAN_MARK (UNPOISON, &x, ...) in order to prevent usage of a LHS
40356         from ASAN_MARK (POISON, &x, ...) coming to a PHI node.
40357         * gimplify.c (asan_poison_variables): Add attribute
40358         use_after_scope_memory to variables that really needs to live
40359         in memory.
40360         * tree-ssa.c (is_asan_mark_p): Do not rewrite into SSA when
40361         having the attribute.
40363 2017-01-23  Martin Liska  <mliska@suse.cz>
40365         * asan.c (create_asan_shadow_var): New function.
40366         (asan_expand_poison_ifn): Likewise.
40367         * asan.h (asan_expand_poison_ifn): New declaration.
40368         * internal-fn.c (expand_ASAN_POISON): Likewise.
40369         * internal-fn.def (ASAN_POISON): New builtin.
40370         * sanopt.c (pass_sanopt::execute): Expand
40371         asan_expand_poison_ifn.
40372         * tree-inline.c (copy_decl_for_dup_finish): Make function
40373         external.
40374         * tree-inline.h (copy_decl_for_dup_finish): Likewise.
40375         * tree-ssa.c (is_asan_mark_p): New function.
40376         (execute_update_addresses_taken): Rewrite local variables
40377         (identified just by use-after-scope as addressable) into SSA.
40379 2017-01-22  Gerald Pfeifer  <gerald@pfeifer.com>
40381         * doc/install.texi (Specific): opensource.apple.com uses https
40382         now. Remove trailing slash.
40384 2017-01-22  Gerald Pfeifer  <gerald@pfeifer.com>
40386         * README.Portability: Remove note on an Irix compatibility issue.
40388 2017-01-22  Dimitry Andric <dim@FreeBSD.org>
40390         * gcov.c (INCLUDE_ALGORITHM): Define.
40391         (INCLUDE_VECTOR): Define.
40392         No longer include <vector> and <algorithm> directly.
40394 2017-01-21  Gerald Pfeifer  <gerald@pfeifer.com>
40396         * doc/extend.texi (Thread-Local): Change www.akkadia.org reference
40397         to https.
40398         * doc/invoke.texi (Code Gen Options): Ditto.
40400 2017-01-21  Jan Hubicka  <hubicka@ucw.cz>
40402         PR lto/78407
40403         * cfg.c (update_bb_profile_for_threading): Fix updating of probablity.
40405 2017-01-21  Bernd Schmidt  <bschmidt@redhat.com>
40407         rtl-optimization/79125
40408         * cprop.c (local_cprop_pass): Handle cases where we make an
40409         unconditional trap.
40411 2017-01-20  Segher Boessenkool  <segher@kernel.crashing.org>
40413         PR target/61729
40414         PR target/77850
40415         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Adjust address to
40416         read from, for big endian.
40418 2017-01-20  Jiong Wang  <jiong.wang@arm.com>
40420         * config/aarch64/aarch64-builtins.c (aarch64_init_builtins): Register
40421         register pauth builtins for LP64 only.
40423 2017-01-20  Marek Polacek  <polacek@redhat.com>
40425         PR c/79152
40426         * gimplify.c (should_warn_for_implicit_fallthrough): Handle consecutive
40427         non-case labels.
40429 2017-01-20  Alexander Monakov  <amonakov@ispras.ru>
40431         * omp-expand.c (expand_omp_simd): Clear PROP_gimple_lomp_dev regardless
40432         of safelen status.
40433         * omp-offload.c (pass_omp_device_lower::gate): Use PROP_gimple_lomp_dev.
40434         * passes.c (dump_properties): Handle PROP_gimple_lomp_dev.
40435         * tree-inline.c (expand_call_inline): Propagate PROP_gimple_lomp_dev.
40437 2017-01-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
40439         PR target/71270
40440         * config/arm/arm.c (neon_valid_immediate): Reject vector constants
40441         in big-endian mode when they are not a single duplicated value.
40443 2017-01-20  Richard Biener  <rguenther@suse.de>
40445         * BASE-VER: Bump to 7.0.1.
40447 2017-01-20  Alexander Monakov  <amonakov@ispras.ru>
40449         * omp-low.c (omplow_simd_context): New struct.  Use it...
40450         (lower_rec_simd_input_clauses): ...here and...
40451         (lower_rec_input_clauses): ...here to hold common data.  Adjust all
40452         references to idx, lane, max_vf, is_simt.
40454 2017-01-20  Graham Markall  <graham.markall@embecosm.com>
40456         * config/arc/arc.h (LINK_SPEC): Use arclinux_nps emulation when
40457         mcpu=nps400.
40459 2017-01-20  Martin Jambor  <mjambor@suse.cz>
40461         * hsa.h: Renaed to hsa-common.h.  Adjusted a comment.
40462         * hsa.c: Renaed to hsa-common.c.  Change include of gt-hsa.h to
40463         gt-hsa-common.h.
40464         * Makefile.in (OBJS): Rename hsa.o to hsa-common.o.
40465         (GTFILES): Rename hsa.c to hsa-common.c.
40466         * hsa-brig.c: Change include of hsa.h to hsa-common.h.
40467         * hsa-dump.c: Likewise.
40468         * hsa-gen.c: Likewise.
40469         * hsa-regalloc.c: Likewise.
40470         * ipa-hsa.c: Likewise.
40471         * omp-expand.c: Likewise.
40472         * omp-low.c: Likewise.
40473         * toplev.c: Likewise.
40475 2017-01-20  Marek Polacek  <polacek@redhat.com>
40477         PR c/64279
40478         * doc/invoke.texi: Document -Wduplicated-branches.
40479         * fold-const.c (operand_equal_p): Handle MODIFY_EXPR, INIT_EXPR,
40480         COMPOUND_EXPR, PREDECREMENT_EXPR, PREINCREMENT_EXPR,
40481         POSTDECREMENT_EXPR, POSTINCREMENT_EXPR, CLEANUP_POINT_EXPR, EXPR_STMT,
40482         STATEMENT_LIST, and RETURN_EXPR.  For non-pure non-const functions
40483         return 0 only when not OEP_LEXICOGRAPHIC.
40484         (fold_build_cleanup_point_expr): Use the expression
40485         location when building CLEANUP_POINT_EXPR.
40486         * tree-core.h (enum operand_equal_flag): Add OEP_LEXICOGRAPHIC.
40487         * tree.c (add_expr): Handle error_mark_node.
40489 2017-01-20  Martin Liska  <mliska@suse.cz>
40491         PR lto/69188
40492         * tree-profile.c (init_ic_make_global_vars): Do not call
40493         finalize_decl.
40494         (gimple_init_gcov_profiler): Likewise.
40496 2017-01-20  Martin Liska  <mliska@suse.cz>
40498         PR ipa/71190
40499         * cgraph.h (maybe_create_reference): Remove argument and
40500         update comment.
40501         * cgraphclones.c (cgraph_node::create_virtual_clone): Remove one
40502         argument.
40503         * ipa-cp.c (create_specialized_node): Likewise.
40504         * symtab.c (symtab_node::maybe_create_reference): Handle
40505         VAR_DECLs and ADDR_EXPRs and select ipa_ref_use type.
40507 2017-01-20  Martin Liska  <mliska@suse.cz>
40509         * read-rtl-function.c (function_reader::create_function): Use
40510         build_decl instread of build_decl_stat.
40512 2017-01-20  Andrew Senkevich  <andrew.senkevich@intel.com>
40514         * config/i386/avx512bwintrin.h: Add k-mask registers shift intrinsics.
40515         * config/i386/avx512dqintrin.h: Ditto.
40516         * config/i386/avx512fintrin.h: Ditto.
40517         * config/i386/i386-builtin-types.def: Add new types.
40518         * config/i386/i386.c: Handle new types.
40519         * config/i386/i386-builtin.def (__builtin_ia32_kshiftliqi)
40520         (__builtin_ia32_kshiftlihi, __builtin_ia32_kshiftlisi)
40521         (__builtin_ia32_kshiftlidi, __builtin_ia32_kshiftriqi)
40522         (__builtin_ia32_kshiftrihi, __builtin_ia32_kshiftrisi)
40523         (__builtin_ia32_kshiftridi): New.
40524         * config/i386/sse.md (k<code><mode>): Rename *k<code><mode>.
40526 2017-01-19  Segher Boessenkool  <segher@kernel.crashing.org>
40528         PR target/78875
40529         PR target/79140
40530         * config/rs6000/rs6000.c (TARGET_STACK_PROTECT_GUARD): Unconditionally
40531         define to rs6000_init_stack_protect_guard.
40532         (rs6000_init_stack_protect_guard): New function.
40534 2017-01-19  Matthew Fortune  <matthew.fortune@imgtec.com>
40535             Yunqiang Su  <yunqiang.su@imgtec.com>
40537         * config.gcc (supported_defaults): Add madd4.
40538         (with_madd4): Add validation.
40539         (all_defaults): Add madd4.
40540         * config/mips/mips.opt (mmadd4): New option.
40541         * config/mips/mips.h (OPTION_DEFAULT_SPECS): Add a default for
40542         mmadd4.
40543         (TARGET_CPU_CPP_BUILTINS): Add builtin_define for
40544         __mips_no_madd4.
40545         (ISA_HAS_UNFUSED_MADD4): Gate with mips_madd4.
40546         (ISA_HAS_FUSED_MADD4): Likewise.
40547         * doc/invoke.texi (-mmadd4): Document the new option.
40548         * doc/install.texi (--with-madd4): Document the new option.
40550 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40552         * config/aarch64/aarch64-builtins.c (enum aarch64_builtins): New
40553         entries for AARCH64_PAUTH_BUILTIN_XPACLRI,
40554         AARCH64_PAUTH_BUILTIN_PACIA1716, AARCH64_PAUTH_BUILTIN_AUTIA1716.
40555         (aarch64_init_pauth_hint_builtins): New.
40556         (aarch64_init_builtins): Call aarch64_init_pauth_hint_builtins.
40557         (aarch64_expand_builtin): Expand new builtins.
40559 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40561         * reg-notes.def (CFA_TOGGLE_RA_MANGLE): New reg-note.
40562         * combine-stack-adj.c (no_unhandled_cfa): Handle
40563         REG_CFA_TOGGLE_RA_MANGLE.
40564         * dwarf2cfi.c (dwarf2out_frame_debug): Handle REG_CFA_TOGGLE_RA_MANGLE.
40565         * config/aarch64/aarch64.c (aarch64_expand_prologue): Generates DWARF
40566         info for return address signing.
40567         (aarch64_expand_epilogue): Likewise.
40569 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40571         * config/aarch64/aarch64-opts.h (aarch64_function_type): New enum.
40572         * config/aarch64/aarch64-protos.h
40573         (aarch64_return_address_signing_enabled): New declaration.
40574         * config/aarch64/aarch64.c (aarch64_return_address_signing_enabled):
40575         New function.
40576         (aarch64_expand_prologue): Sign return address before it's pushed onto
40577         stack.
40578         (aarch64_expand_epilogue): Authenticate return address fetched from
40579         stack.
40580         (aarch64_override_options): Sanity check for ILP32 and ISA level.
40581         (aarch64_attributes): New function attributes for "sign-return-address".
40582         * config/aarch64/aarch64.md (UNSPEC_AUTI1716, UNSPEC_AUTISP,
40583         UNSPEC_PACI1716, UNSPEC_PACISP, UNSPEC_XPACLRI): New unspecs.
40584         ("*do_return"): Generate combined instructions according to key index.
40585         ("<pauth_mnem_prefix>sp", "<pauth_mnem_prefix1716", "xpaclri"): New.
40586         * config/aarch64/iterators.md (PAUTH_LR_SP, PAUTH_17_16): New integer
40587         iterators.
40588         (pauth_mnem_prefix, pauth_hint_num_a): New integer attributes.
40589         * config/aarch64/aarch64.opt (msign-return-address=): New.
40590         * doc/extend.texi (AArch64 Function Attributes): Documents
40591         "sign-return-address=".
40592         * doc/invoke.texi (AArch64 Options): Documents "-msign-return-address=".
40594 2017-01-19  Matthew Fortune  <matthew.fortune@imgtec.com>
40596         * doc/invoke.texi: Add missing -mlxc1-sxc1 options to
40597         overall option summary.
40599 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40601         * config/aarch64/aarch64-arches.def: New entry for "armv8.3-a".
40602         * config/aarch64/aarch64.h (AARCH64_FL_V8_3, AARCH64_FL_FOR_ARCH8_3,
40603         AARCH64_ISA_V8_3, TARGET_ARMV8_3): New.
40604         * doc/invoke.texi (AArch64 Options): Document "armv8.3-a".
40606 2017-01-19  Michael Meissner  <meissner@linux.vnet.ibm.com>
40608         * config/rs6000/rs6000-cpus.def (ISA_3_0_MASKS_SERVER): Enable
40609         -mpower9-minmax by default for -mcpu=power9.
40610         (ISA_3_MASKS_IEEE): Require -mvsx-small-integer to enable IEEE
40611         128-bit floating point.
40613 2017-01-20  Alan Modra  <amodra@gmail.com>
40615         * config/rs6000/rs6000.md (cmpstrnsi, cmpstrsi): Fail if
40616         optimizing for size.
40618 2017-01-20  Alan Modra  <amodra@gmail.com>
40620         PR target/79144
40621         * config/rs6000/rs6000.c (expand_strn_compare): Get the asm name
40622         for strcmp and strncmp from corresponding builtin decl.
40624 2017-01-19  Uros Bizjak  <ubizjak@gmail.com>
40626         * config.gcc (x86_64-*-rtems*): Use i386/rtemself.h
40627         instead of i386/rtems-64.h.
40628         * config/i386/rtems-64.h: Remove.
40630 2017-01-19  Uros Bizjak  <ubizjak@gmail.com>
40632         PR target/78478
40633         Revert:
40634         2013-11-05  Uros Bizjak  <ubizjak@gmail.com>
40636         * config/i386/rtemself.h (LONG_DOUBLE_TYPE_SIZE): New define.
40638 2017-01-19  Tamar Christina  <tamar.christina@arm.com>
40640         * config/aarch64/aarch64.c (aarch64_simd_gen_const_vector_dup):
40641         Change int to HOST_WIDE_INT.
40642         * config/aarch64/aarch64-protos.h
40643         (aarch64_simd_gen_const_vector_dup): Likewise.
40644         * config/aarch64/aarch64-simd.md: Add copysign<mode>3.
40646 2017-01-19  David Malcolm  <dmalcolm@redhat.com>
40648         * langhooks-def.h (lhd_type_for_size): New decl.
40649         (LANG_HOOKS_TYPE_FOR_SIZE): Define as lhd_type_for_size.
40650         * langhooks.c (lhd_type_for_size): New function, taken from
40651         lto_type_for_size.
40653 2017-01-19  Pat Haugen  <pthaugen@us.ibm.com>
40655         * config/rs6000/power9.md (power9-alu): Remove 'cmp' type and add
40656         define_bypass for CR latency.
40657         (power9-cracked-alu): Update bypass latency and remove power9-branch.
40658         (power9-alu2): Add define_bypass for CR latency.
40659         (power9-cmp): New.
40660         (power9-mul): Update insn latency.
40661         (power9-mul-compare): Update insn latency, bypass latency and remove
40662         power9-branch.
40664 2016-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
40666         * config/aarch64/aarch64-protos.h (aarch64_nopcrelative_literal_loads):
40667         Delete.
40668         * config/aarch64/aarch64.md
40669         (aarch64_reload_movcp<GPF_TF:mode><P:mode>): Delete reference to
40670         aarch64_nopcrelative_literal_loads.
40671         (aarch64_reload_movcp<VALL:mode><P:mode>): Likewise.
40673 2017-01-19  Chenghua Xu  <paul.hua.gm@gmail.com>
40675         * config/mips/mips.h (ISA_HAS_FUSED_MADD4): Enable for
40676         TARGET_LOONGSON_3A.
40677         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_LOONGSON_3A.
40679 2017-01-19  Doug Gilmore  <doug.gilmore@imgtec.com>
40681         PR target/78176
40682         * config.gcc (supported_defaults): Add lxc1-sxc1.
40683         (with_lxc1_sxc1): Add validation.
40684         (all_defaults): Add lxc1-sxc1.
40685         * config/mips/mips.opt (mlxc1-sxc1): New option.
40686         * config/mips/mips.h (OPTION_DEFAULT_SPECS): Add a default for
40687         mlxc1-sxc1.
40688         (TARGET_CPU_CPP_BUILTINS): Add builtin_define for
40689         __mips_no_lxc1_sxc1.
40690         (ISA_HAS_LXC1_SXC1): Gate with mips_lxc1_sxc1.
40691         * doc/invoke.texi (-mlxc1-sxc1): Document the new option.
40692         * doc/install.texi (--with-lxc1-sxc1): Document the new option.
40694 2017-01-19  Richard Biener  <rguenther@suse.de>
40696         PR tree-optimization/72488
40697         * tree-ssa-sccvn.c (run_scc_vn): When we abort the VN make
40698         sure to restore SSA info.
40699         * tree-ssa.c (verify_ssa): Verify SSA info is not shared.
40701 2017-01-19  Richard Earnshaw  <rearnsha@arm.com>
40703         PR rtl-optimization/79121
40704         * expr.c (expand_expr_real_2, case LSHIFT_EXPR): Look at the signedness
40705         of the inner type when shifting an extended value.
40707 2017-01-17  Jan Hubicka  <hubicka@ucw.cz>
40709         PR lto/78407
40710         * symtab.c (symtab_node::equal_address_to): Fix comparing of
40711         interposable aliases.
40713 2017-01-18  Peter Bergner  <bergner@vnet.ibm.com>
40715         PR target/78516
40716         * config/rs6000/spe.md (mov_si<mode>_e500_subreg0): Fix constraints.
40717         Use the evmergelohi instruction.
40718         (mov_si<mode>_e500_subreg4_2_le): Likewise.
40719         (mov_sitf_e500_subreg8_2_be): Likewise.
40720         (mov_sitf_e500_subreg12_2_le): Likewise.
40721         (mov_si<mode>_e500_subreg0_2_le): Fix constraints.
40722         (mov_si<mode>_e500_subreg4_2_be): Likewise.
40723         (mov_sitf_e500_subreg8_2_le): Likewise.
40724         (mov_sitf_e500_subreg12_2_be): Likewise.
40726 2017-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40728         * config/rs6000/altivec.md (altivec_vbpermq): Change "type"
40729         attribute from vecsimple to vecperm.
40730         (altivec_vbpermq2): Likewise.
40732 2017-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40734         PR target/79040
40735         * config/rs6000/altivec.h: Fix typo of vec_cntlz to vec_cnttz.
40737 2017-01-18  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
40738         * config/rs6000/rs6000-protos.h (expand_strn_compare): Add arg.
40739         * config/rs6000/rs6000.c (expand_strn_compare): Add ability to expand
40740         strcmp. Fix bug where comparison didn't stop with zero byte. Fix
40741         case where N arg is SIZE_MAX.
40742         * config/rs6000/rs6000.md (cmpstrnsi): Args to expand_strn_compare.
40743         (cmpstrsi): Add pattern.
40745 2017-01-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
40747         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
40748         __builtin_vec_revb builtins.
40749         * config/rs6000/rs6000-builtins.def (P9V_BUILTIN_XXBRQ_V16QI): Add
40750         built-in functions to support generation of the ISA 3.0 XXBR<x>
40751         vector byte reverse instructions.
40752         (P9V_BUILTIN_XXBRQ_V1TI): Likewise.
40753         (P9V_BUILTIN_XXBRD_V2DI): Likewise.
40754         (P9V_BUILTIN_XXBRD_V2DF): Likewise.
40755         (P9V_BUILTIN_XXBGW_V4SI): Likewise.
40756         (P9V_BUILTIN_XXBGW_V4SF): Likewise.
40757         (P9V_BUILTIN_XXBGH_V8HI): Likewise.
40758         (P9V_BUILTIN_VEC_REVB): Likewise.
40759         * config/rs6000/vsx.md (p9_xxbrq_v1ti): New insns/expanders to
40760         generate the ISA 3.0 XXBR<x> vector byte reverse instructions.
40761         (p9_xxbrq_v16qi): Likewise.
40762         (p9_xxbrd_<mode>, VSX_D iterator): Likewise.
40763         (p9_xxbrw_<mode>, VSX_W iterator): Likewise.
40764         (p9_xxbrh_v8hi): Likewise.
40765         * config/rs6000/altivec.h (vec_revb): Define if ISA 3.0.
40766         * doc/extend.texi (RS/6000 Altivec Built-ins): Document the
40767         vec_revb built-in functions.
40769 2017-01-18  Uros Bizjak  <ubizjak@gmail.com>
40771         PR rtl-optimization/78952
40772         * config/i386/i386.md (any_extract): New code iterator.
40773         (*insvqi_2): Use any_extract for source operand.
40774         (*insvqi_3): Use any_shiftrt for source operand.
40776 2017-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
40778         * config/aarch64/aarch64.c (aarch64_sched_adjust_priority)
40779         New function.
40780         (TARGET_SCHED_ADJUST_PRIORITY): Define target hook.
40782 2017-01-18  Matthias Klose  <doko@ubuntu.com>
40784         * doc/install.texi: Allow default for --with-target-bdw-gc-include.
40786 2016-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40788         * config/rs6000/altivec.h (vec_bperm): Change #define.
40789         * config/rs6000/altivec.md (UNSPEC_VBPERMD): New enum constant.
40790         (altivec_vbpermq2): New define_insn.
40791         (altivec_vbpermd): Likewise.
40792         * config/rs6000/rs6000-builtin.def (VBPERMQ2): New monomorphic
40793         function interface.
40794         (VBPERMD): Likewise.
40795         (VBPERM): New polymorphic function interface.
40796         * config/rs6000/r6000-c.c (altivec_overloaded_builtins_table):
40797         Add entries for P9V_BUILTIN_VEC_VBPERM.
40798         * doc/extend.texi: Add interfaces for vec_bperm.
40800 2017-01-18  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
40802         * config/s390/s390-c.c (s390_expand_overloaded_builtin): Downcase
40803         first letter of error messages.
40804         (s390_resolve_overloaded_builtin): Likewise.
40805         * config/s390/s390.c (s390_expand_builtin): Likewise.
40806         (s390_invalid_arg_for_unprototyped_fn): Likewise.
40807         (s390_valid_target_attribute_inner_p): Likewise.
40808         * config/s390/s390.md ("tabort"): Likewise.
40810 2017-01-18  Toma Tabacu  <toma.tabacu@imgtec.com>
40812         * config/mips/mips.h (ISA_HAS_DIV3): Remove unused macro.
40813         (ISA_AVOID_DIV_HILO): New macro.
40814         (ISA_HAS_DIV): Use new ISA_AVOID_DIV_HILO macro.
40815         (ISA_HAS_DDIV): Likewise.
40817 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40819         * doc/invoke.texi (fabi-version): Correct number of occurrences.
40821 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40823         * doc/invoke.texi (fabi-version): Spelling fix.
40825 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40827         PR c++/70182
40828         * doc/invoke.texi (fabi-version): Mention mangling fix for
40829         operator names.
40831 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40833         PR c++/77489
40834         * doc/invoke.texi (fabi-version): Document discriminator mangling.
40836 2017-01-17  Segher Boessenkool  <segher@kernel.crashing.org>
40838         PR target/78875
40839         * config/rs6000/rs6000-opts.h (stack_protector_guard): New enum.
40840         * config/rs6000/rs6000.c (rs6000_option_override_internal): Handle
40841         the new options.
40842         * config/rs6000/rs6000.md (stack_protect_set): Handle the new more
40843         flexible settings.
40844         (stack_protect_test): Ditto.
40845         * config/rs6000/rs6000.opt (mstack-protector-guard=,
40846         mstack-protector-guard-reg=, mstack-protector-guard-offset=): New
40847         options.
40848         * doc/invoke.texi (Option Summary) [RS/6000 and PowerPC Options]:
40849         Add -mstack-protector-guard=, -mstack-protector-guard-reg=, and
40850         -mstack-protector-guard-offset=.
40851         (RS/6000 and PowerPC Options): Ditto.
40853 2017-01-17  Uros Bizjak  <ubizjak@gmail.com>
40855         * config/i386/i386.h (MASK_CLASS_P): New define.
40856         * config/i386/i386.c (inline_secondary_memory_needed): Ensure that
40857         there are no registers from different register sets also when
40858         mask registers are used.  Update function comment.
40859         * config/i386/i386.md (*movsi_internal): Split (*k/*krm) alternative
40860         to (*k/*r) and (*k/*km) alternatives.
40862 2017-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
40864         * config/aarch64/aarch64.md (eh_return): Remove pattern and splitter.
40865         * config/aarch64/aarch64.h (AARCH64_EH_STACKADJ_REGNUM): Remove.
40866         (EH_RETURN_HANDLER_RTX): New define.
40867         * config/aarch64/aarch64.c (aarch64_frame_pointer_required):
40868         Force frame pointer in EH return functions.
40869         (aarch64_expand_epilogue): Add barrier for eh_return.
40870         (aarch64_final_eh_return_addr): Remove.
40871         (aarch64_eh_return_handler_rtx): New function.
40872         * config/aarch64/aarch64-protos.h (aarch64_final_eh_return_addr):
40873         Remove.
40874         (aarch64_eh_return_handler_rtx): New prototype.
40876 2017-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40878         * config/rs6000/altivec.h (vec_rlmi): New #define.
40879         (vec_vrlnm): Likewise.
40880         (vec_rlnm): Likewise.
40881         * config/rs6000/altivec.md (UNSPEC_VRLMI): New UNSPEC enum value.
40882         (UNSPEC_VRLNM): Likewise.
40883         (VIlong): New mode iterator.
40884         (altivec_vrl<VI_char>mi): New define_insn.
40885         (altivec_vrl<VI_char>nm): Likewise.
40886         * config/rs6000/rs6000-builtin.def (VRLWNM): New monomorphic
40887         function entry.
40888         (VRLDNM): Likewise.
40889         (RLNM): New polymorphic function entry.
40890         (VRLWMI): New monomorphic function entry.
40891         (VRLDMI): Likewise.
40892         (RLMI): New polymorphic function entry.
40893         * config/rs6000/r6000-c.c (altivec_overloaded_builtin_table): Add
40894         new entries for P9V_BUILTIN_VEC_RLMI and P9V_BUILTIN_VEC_RLNM.
40895         * doc/extend.texi: Add description of vec_rlmi, vec_rlnm, and
40896         vec_vrlnm.
40898 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
40900         PR debug/78839
40901         * dwarf2out.c (field_byte_offset): Restore the
40902         PCC_BITFIELD_TYPE_MATTERS behavior for INTEGER_CST DECL_FIELD_OFFSET
40903         and DECL_FIELD_BIT_OFFSET.  Use fold_build2 instead of build2 + fold.
40904         (analyze_variants_discr, gen_variant_part): Use fold_build2 instead
40905         of build2 + fold.
40907 2017-01-17  Eric Botcazou  <ebotcazou@adacore.com>
40909         PR ada/67205
40910         * config/aarch64/aarch64.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): Define
40912 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
40914         PR debug/71669
40915         * dwarf2out.c (add_data_member_location_attribute): For constant
40916         offset bitfield emit for -gdwarf-5 DW_AT_data_bit_offset attribute
40917         instead of DW_AT_data_member_location, DW_AT_bit_offset and
40918         DW_AT_byte_size attributes.
40920 2017-01-17  Eric Botcazou  <ebotcazou@adacore.com>
40922         * config/rs6000/rs6000.c (rs6000_emit_move): Also use a TOC reference
40923         after forcing to constant memory when the code model is medium.
40925 2017-01-17  Julia Koval  <julia.koval@intel.com>
40927         PR target/76731
40928         * config/i386/avx512fintrin.h
40929         (_mm512_i32gather_ps): Change __addr type to void const*.
40930         (_mm512_mask_i32gather_ps): Ditto.
40931         (_mm512_i32gather_pd): Ditto.
40932         (_mm512_mask_i32gather_pd): Ditto.
40933         (_mm512_i64gather_ps): Ditto.
40934         (_mm512_mask_i64gather_ps): Ditto.
40935         (_mm512_i64gather_pd): Ditto.
40936         (_mm512_mask_i64gather_pd): Ditto.
40937         (_mm512_i32gather_epi32): Ditto.
40938         (_mm512_mask_i32gather_epi32): Ditto.
40939         (_mm512_i32gather_epi64): Ditto.
40940         (_mm512_mask_i32gather_epi64): Ditto.
40941         (_mm512_i64gather_epi32): Ditto.
40942         (_mm512_mask_i64gather_epi32): Ditto.
40943         (_mm512_i64gather_epi64): Ditto.
40944         (_mm512_mask_i64gather_epi64): Ditto.
40945         (_mm512_i32scatter_ps): Change __addr type to void*.
40946         (_mm512_mask_i32scatter_ps): Ditto.
40947         (_mm512_i32scatter_pd): Ditto.
40948         (_mm512_mask_i32scatter_pd): Ditto.
40949         (_mm512_i64scatter_ps): Ditto.
40950         (_mm512_mask_i64scatter_ps): Ditto.
40951         (_mm512_i64scatter_pd): Ditto.
40952         (_mm512_mask_i64scatter_pd): Ditto.
40953         (_mm512_i32scatter_epi32): Ditto.
40954         (_mm512_mask_i32scatter_epi32): Ditto.
40955         (_mm512_i32scatter_epi64): Ditto.
40956         (_mm512_mask_i32scatter_epi64): Ditto.
40957         (_mm512_i64scatter_epi32): Ditto.
40958         (_mm512_mask_i64scatter_epi32): Ditto.
40959         (_mm512_i64scatter_epi64): Ditto.
40960         (_mm512_mask_i64scatter_epi64): Ditto.
40961         * config/i386/avx512pfintrin.h
40962         (_mm512_mask_prefetch_i32gather_pd): Change __addr type to void const*.
40963         (_mm512_mask_prefetch_i32gather_ps): Ditto.
40964         (_mm512_mask_prefetch_i64gather_pd): Ditto.
40965         (_mm512_mask_prefetch_i64gather_ps): Ditto.
40966         (_mm512_prefetch_i32scatter_pd): Change __addr type to void*.
40967         (_mm512_prefetch_i32scatter_ps): Ditto.
40968         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
40969         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
40970         (_mm512_prefetch_i64scatter_pd): Ditto.
40971         (_mm512_prefetch_i64scatter_ps): Ditto.
40972         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
40973         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
40974         * config/i386/avx512vlintrin.h
40975         (_mm256_mmask_i32gather_ps): Change __addr type to void const*.
40976         (_mm_mmask_i32gather_ps): Ditto.
40977         (_mm256_mmask_i32gather_pd): Ditto.
40978         (_mm_mmask_i32gather_pd): Ditto.
40979         (_mm256_mmask_i64gather_ps): Ditto.
40980         (_mm_mmask_i64gather_ps): Ditto.
40981         (_mm256_mmask_i64gather_pd): Ditto.
40982         (_mm_mmask_i64gather_pd): Ditto.
40983         (_mm256_mmask_i32gather_epi32): Ditto.
40984         (_mm_mmask_i32gather_epi32): Ditto.
40985         (_mm256_mmask_i32gather_epi64): Ditto.
40986         (_mm_mmask_i32gather_epi64): Ditto.
40987         (_mm256_mmask_i64gather_epi32): Ditto.
40988         (_mm_mmask_i64gather_epi32): Ditto.
40989         (_mm256_mmask_i64gather_epi64): Ditto.
40990         (_mm_mmask_i64gather_epi64): Ditto.
40991         (_mm256_i32scatter_ps): Change __addr type to void*.
40992         (_mm256_mask_i32scatter_ps): Ditto.
40993         (_mm_i32scatter_ps): Ditto.
40994         (_mm_mask_i32scatter_ps): Ditto.
40995         (_mm256_i32scatter_pd): Ditto.
40996         (_mm256_mask_i32scatter_pd): Ditto.
40997         (_mm_i32scatter_pd): Ditto.
40998         (_mm_mask_i32scatter_pd): Ditto.
40999         (_mm256_i64scatter_ps): Ditto.
41000         (_mm256_mask_i64scatter_ps): Ditto.
41001         (_mm_i64scatter_ps): Ditto.
41002         (_mm_mask_i64scatter_ps): Ditto.
41003         (_mm256_i64scatter_pd): Ditto.
41004         (_mm256_mask_i64scatter_pd): Ditto.
41005         (_mm_i64scatter_pd): Ditto.
41006         (_mm_mask_i64scatter_pd): Ditto.
41007         (_mm256_i32scatter_epi32): Ditto.
41008         (_mm256_mask_i32scatter_epi32): Ditto.
41009         (_mm_i32scatter_epi32): Ditto.
41010         (_mm_mask_i32scatter_epi32): Ditto.
41011         (_mm256_i32scatter_epi64): Ditto.
41012         (_mm256_mask_i32scatter_epi64): Ditto.
41013         (_mm_i32scatter_epi64): Ditto.
41014         (_mm_mask_i32scatter_epi64): Ditto.
41015         (_mm256_i64scatter_epi32): Ditto.
41016         (_mm256_mask_i64scatter_epi32): Ditto.
41017         (_mm_i64scatter_epi32): Ditto.
41018         (_mm_mask_i64scatter_epi32): Ditto.
41019         (_mm256_i64scatter_epi64): Ditto.
41020         (_mm256_mask_i64scatter_epi64): Ditto.
41021         (_mm_i64scatter_epi64): Ditto.
41022         (_mm_mask_i64scatter_epi64): Ditto.
41023         * config/i386/i386-builtin-types.def (V16SF_V16SF_PCFLOAT_V16SI_HI_INT)
41024         (V8DF_V8DF_PCDOUBLE_V8SI_QI_INT, V8SF_V8SF_PCFLOAT_V8DI_QI_INT)
41025         (V8DF_V8DF_PCDOUBLE_V8DI_QI_INT, V16SI_V16SI_PCINT_V16SI_HI_INT)
41026         (V8DI_V8DI_PCINT64_V8SI_QI_INT, V8SI_V8SI_PCINT_V8DI_QI_INT)
41027         (V8DI_V8DI_PCINT64_V8DI_QI_INT, V2DF_V2DF_PCDOUBLE_V4SI_QI_INT)
41028         (V4DF_V4DF_PCDOUBLE_V4SI_QI_INT, V2DF_V2DF_PCDOUBLE_V2DI_QI_INT)
41029         (V4DF_V4DF_PCDOUBLE_V4DI_QI_INT, V4SF_V4SF_PCFLOAT_V4SI_QI_INT)
41030         (V8SF_V8SF_PCFLOAT_V8SI_QI_INT, V4SF_V4SF_PCFLOAT_V2DI_QI_INT)
41031         (V4SF_V4SF_PCFLOAT_V4DI_QI_INT, V2DI_V2DI_PCINT64_V4SI_QI_INT)
41032         (V4DI_V4DI_PCINT64_V4SI_QI_INT, V2DI_V2DI_PCINT64_V2DI_QI_INT)
41033         (V4DI_V4DI_PCINT64_V4DI_QI_INT, V4SI_V4SI_PCINT_V4SI_QI_INT)
41034         (V8SI_V8SI_PCINT_V8SI_QI_INT, V4SI_V4SI_PCINT_V2DI_QI_INT)
41035         (V4SI_V4SI_PCINT_V4DI_QI_INT, VOID_PFLOAT_HI_V16SI_V16SF_INT)
41036         (VOID_PFLOAT_QI_V8SI_V8SF_INT, VOID_PFLOAT_QI_V4SI_V4SF_INT)
41037         (VOID_PDOUBLE_QI_V8SI_V8DF_INT, VOID_PDOUBLE_QI_V4SI_V4DF_INT)
41038         (VOID_PDOUBLE_QI_V4SI_V2DF_INT, VOID_PFLOAT_QI_V8DI_V8SF_INT)
41039         (VOID_PFLOAT_QI_V4DI_V4SF_INT, VOID_PFLOAT_QI_V2DI_V4SF_INT)
41040         (VOID_PDOUBLE_QI_V8DI_V8DF_INT, VOID_PDOUBLE_QI_V4DI_V4DF_INT)
41041         (VOID_PDOUBLE_QI_V2DI_V2DF_INT, VOID_PINT_HI_V16SI_V16SI_INT)
41042         (VOID_PINT_QI_V8SI_V8SI_INT, VOID_PINT_QI_V4SI_V4SI_INT)
41043         (VOID_PLONGLONG_QI_V8SI_V8DI_INT, VOID_PLONGLONG_QI_V4SI_V4DI_INT)
41044         (VOID_PLONGLONG_QI_V4SI_V2DI_INT, VOID_PINT_QI_V8DI_V8SI_INT)
41045         (VOID_PINT_QI_V4DI_V4SI_INT, VOID_PINT_QI_V2DI_V4SI_INT)
41046         (VOID_PLONGLONG_QI_V8DI_V8DI_INT, VOID_QI_V8SI_PCINT64_INT_INT)
41047         (VOID_PLONGLONG_QI_V4DI_V4DI_INT, VOID_PLONGLONG_QI_V2DI_V2DI_INT)
41048         (VOID_HI_V16SI_PCINT_INT_INT, VOID_QI_V8DI_PCINT64_INT_INT)
41049         (VOID_QI_V8DI_PCINT_INT_INT): Remove.
41050         (V16SF_V16SF_PCVOID_V16SI_HI_INT,  V8DF_V8DF_PCVOID_V8SI_QI_INT)
41051         (V8SF_V8SF_PCVOID_V8DI_QI_INT, V8DF_V8DF_PCVOID_V8DI_QI_INT)
41052         (V16SI_V16SI_PCVOID_V16SI_HI_INT, V8DI_V8DI_PCVOID_V8SI_QI_INT)
41053         (V8SI_V8SI_PCVOID_V8DI_QI_INT, V8DI_V8DI_PCVOID_V8DI_QI_INT)
41054         (VOID_PVOID_HI_V16SI_V16SF_INT, VOID_PVOID_QI_V8SI_V8DF_INT)
41055         (VOID_PVOID_QI_V8DI_V8SF_INT, VOID_PVOID_QI_V8DI_V8DF_INT)
41056         (VOID_PVOID_HI_V16SI_V16SI_INT, VOID_PVOID_QI_V8SI_V8DI_INT)
41057         (VOID_PVOID_QI_V8DI_V8SI_INT, VOID_PVOID_QI_V8DI_V8DI_INT)
41058         (V2DF_V2DF_PCVOID_V4SI_QI_INT, V4DF_V4DF_PCVOID_V4SI_QI_INT)
41059         (V2DF_V2DF_PCVOID_V2DI_QI_INT, V4DF_V4DF_PCVOID_V4DI_QI_INT
41060         (V4SF_V4SF_PCVOID_V4SI_QI_INT, V8SF_V8SF_PCVOID_V8SI_QI_INT)
41061         (V4SF_V4SF_PCVOID_V2DI_QI_INT, V4SF_V4SF_PCVOID_V4DI_QI_INT)
41062         (V2DI_V2DI_PCVOID_V4SI_QI_INT, V4DI_V4DI_PCVOID_V4SI_QI_INT)
41063         (V2DI_V2DI_PCVOID_V2DI_QI_INT, V4DI_V4DI_PCVOID_V4DI_QI_INT)
41064         (V4SI_V4SI_PCVOID_V4SI_QI_INT, V8SI_V8SI_PCVOID_V8SI_QI_INT)
41065         (V4SI_V4SI_PCVOID_V2DI_QI_INT, V4SI_V4SI_PCVOID_V4DI_QI_INT)
41066         (VOID_PVOID_QI_V8SI_V8SF_INT, VOID_PVOID_QI_V4SI_V4SF_INT)
41067         (VOID_PVOID_QI_V4SI_V4DF_INT, VOID_PVOID_QI_V4SI_V2DF_INT)
41068         (VOID_PVOID_QI_V4DI_V4SF_INT, VOID_PVOID_QI_V2DI_V4SF_INT)
41069         (VOID_PVOID_QI_V4DI_V4DF_INT, VOID_PVOID_QI_V2DI_V2DF_INT)
41070         (VOID_PVOID_QI_V8SI_V8SI_INT, VOID_PVOID_QI_V4SI_V4SI_INT)
41071         (VOID_PVOID_QI_V4SI_V4DI_INT, VOID_PVOID_QI_V4SI_V2DI_INT)
41072         (VOID_PVOID_QI_V4DI_V4SI_INT, VOID_PVOID_QI_V2DI_V4SI_INT)
41073         (VOID_PVOID_QI_V4DI_V4DI_INT, VOID_PVOID_QI_V2DI_V2DI_INT)
41074         (VOID_QI_V8SI_PCVOID_INT_INT, VOID_HI_V16SI_PCVOID_INT_INT)
41075         (VOID_QI_V8DI_PCVOID_INT_INT): Add.
41076         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Adjust builtin
41077         definitions accordingly.
41079 2017-01-17  Kito Cheng  <kito.cheng@gmail.com>
41080             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
41082         PR target/79079
41083         * internal-fn.c (expand_mul_overflow): Use convert_modes instead of
41084         gen_lowpart.
41086 2017-01-17  Vladimir Makarov  <vmakarov@redhat.com>
41088         PR target/79058
41089         * ira-conflicts.c (ira_build_conflicts): Update total conflict
41090         hard regs for inner regno.
41092 2017-01-17  Martin Liska  <mliska@suse.cz>
41094         PR ipa/71207
41095         * ipa-polymorphic-call.c (contains_type_p): Fix wrong
41096         assumption and add comment.
41098 2017-01-17  Nathan Sidwell  <nathan@acm.org>
41100         * ipa-visibility.c (localize_node): New function, broken out of ...
41101         (function_and_variable_visibility): ... here. Call it.
41103 2017-01-17  Jan Hubicka  <hubicka@ucw.cz>
41105         PR middle-end/77445
41106         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges):
41107         correctly set frequency of oudgoing edge.
41108         (duplicate_thread_path): Fix profile updating.
41110 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
41112         PR other/79046
41113         * configure.ac: Add GCC_BASE_VER.
41114         * Makefile.in (version): Use @get_gcc_base_ver@ instead of cat to get
41115         version from BASE-VER file.
41116         (CFLAGS-gcc.o): Add -DBASEVER=$(BASEVER_s).
41117         (gcc.o): Depend on $(BASEVER).
41118         * common.opt (dumpfullversion): New option.
41119         * gcc.c (driver_handle_option): Handle OPT_dumpfullversion.
41120         * doc/invoke.texi: Document -dumpfullversion.
41121         * doc/install.texi: Document --with-gcc-major-version-only.
41122         * configure: Regenerated.
41124 2017-01-17  Richard Biener  <rguenther@suse.de>
41126         PR tree-optimization/71433
41127         * tree-vrp.c (register_new_assert_for): Merge same asserts
41128         on all incoming edges.
41129         (process_assert_insertions_for): Handle insertions at the
41130         beginning of BBs.
41132 2017-01-17  Gerald Pfeifer  <gerald@pfeifer.com>
41134         * config/i386/cygwin.h (LIBGCJ_SONAME): Remove.
41135         * config/i386/mingw32.h (LIBGCJ_SONAME): Remove.
41137 2017-01-17  Kaz Kojima  <kkojima@gcc.gnu.org>
41139         PR target/78633
41140         * config/sh/sh.md (cmpeqsi_t+1): Call copy_rtx to avoid invalid
41141         RTL sharing.
41143 2017-01-17  Alan Modra  <amodra@gmail.com>
41145         PR target/79066
41146         * config/rs6000/rs6000.md (elf_high, elf_low): Disable when pic.
41147         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack): Don't allow
41148         symbolic stack limit when pic.
41150 2017-01-16  Martin Sebor  <msebor@redhat.com>
41152         PR tree-optimization/78608
41153         * gimple-ssa-sprintf.c (tree_digits): Avoid negating TYPE_MIN.
41155 2017-01-16  Jeff Law  <law@redhat.com>
41157         Revert:
41158         2016-12-02  Tadek Kijkowski  <tkijkowski@gmail.com>
41159         * Makefile.in (PREPROCESSOR_DEFINES): Add a level of indirection
41160         for several include directories that may be relative to sysroot.
41161         * config/i386/x-mingw32 (gplus_includedir): Define.
41162         (gplus_tool_includedir, gplus_backward_include_dir): Likewise.
41163         (native_system_includedir): Likewise.
41164         * config/i386/mingw32.h (STANDARD_STARTFILE_PREFIX_1): Do not
41165         override if TARGET_SYSTEM_ROOT is defined.
41166         (NATIVE_SYSTEM_HEADER_DIR): Likewise.
41168         PR tree-optimization/79090
41169         PR tree-optimization/33562
41170         PR tree-optimization/61912
41171         PR tree-optimization/77485
41172         * tree-ssa-dse.c (compute_trims): Accept STMT argument.  Dump STMT
41173         and computed trims into the dump file.
41175 2017-01-17  Uros Bizjak  <ubizjak@gmail.com>
41177         * config/i386/i386.h (LIMIT_RELOAD_CLASS): Remove.
41179 2017-01-16  Jakub Jelinek  <jakub@redhat.com>
41181         PR c/79089
41182         * gimplify.c (gimplify_init_constructor): If want_value and
41183         object == lhs, unshare lhs to avoid invalid tree sharing.  Formatting
41184         fix.
41186         PR target/79080
41187         * loop-doloop.c (doloop_modify): Call unshare_all_rtl_in_chain on
41188         sequence.  Formatting fixes.
41189         (doloop_optimize): Formatting fixes.
41191         PR driver/49726
41192         * gcc.c (debug_level_greater_than_spec_func): New function.
41193         (static_spec_functions): Add debug-level-gt spec function.
41194         (ASM_DEBUG_SPEC, cpp_options): Use %:debug-level-gt(0) instead of
41195         !g0.
41196         * config/darwin.h (DSYMUTIL_SPEC, ASM_DEBUG_SPEC): Likewise.
41197         * config/darwin9.h (DSYMUTIL_SPEC, ASM_DEBUG_SPEC): Likewise.
41198         * common.opt (g, gcoff, gdwarf, gdwarf-, ggdb, gno-pubnames,
41199         gpubnames, ggnu-pubnames, gno-record-gcc-switches,
41200         grecord-gcc-switches, gno-strict-dwarf, gstrict-dwarf, gstabs,
41201         gstabs+, gtoggle, gvms, gxcoff, gxcoff+): Add Driver flag.
41203 2017-01-16  Uros Bizjak  <ubizjak@gmail.com>
41205         * config/i386/i386.h (HARD_REGNO_CALLER_SAVE_MODE): Apply HImode and
41206         QImode fixups to general and mask registers only.
41208 2017-01-16  Carl Love  <cel@us.ibm.com>
41210         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Add support
41211         for built-in functions
41212         vector signed char vec_nabs (vector signed char)
41213         vector signed short vec_nabs (vector signed short)
41214         vector signed int vec_nabs (vector signed int)
41215         vector signed long long vec_nabs (vector signed long long)
41216         vector float vec_nabs (vector float)
41217         vector double vec_nabs (vector double)
41218         * config/rs6000/rs6000-builtin.def: Add definitions for NABS functions
41219         and NABS overload.
41220         * config/rs6000/altivec.md: New define_expand nabs<mode>2 types
41221         * config/rs6000/altivec.h: New define for vec_nabs built-in function.
41222         * doc/extend.texi: Update the documentation file for the new built-in
41223         functions.
41225 2017-01-16  Martin Sebor  <msebor@redhat.com>
41227         * gimple-ssa-sprintf.c (format_directive): Correct a typo in a warning
41228         message.
41230 2017-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
41232         * config/rs6000/rs6000.c (rtx_is_swappable_p): Change
41233         UNSPEC_VSX__XXSPLTD to require special splat handling.
41235 2017-01-16  David Malcolm  <dmalcolm@redhat.com>
41237         PR bootstrap/78616
41238         * system.h: Poison strndup.
41240 2017-01-16  Alan Modra  <amodra@gmail.com>
41242         PR target/79098
41243         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Don't
41244         use a switch.
41246 2017-01-16  Georg-Johann Lay  <avr@gjlay.de>
41248         * config/avr/avr.h (BRANCH_COST) [reload_completed]: Increase by 4.
41250 2017-01-15  Uros Bizjak  <ubizjak@gmail.com>
41252         * config/i386/i386.c (ix86_legitimate_combined_insn): Do not
41253         call recog here.  Assert that INSN_CODE (insn) is non-negative.
41255 2017-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
41257         PR target/72749
41258         * cfgrtl.c (rtl_split_edge): Also patch jump insns that jump to the
41259         fallthrough.
41260         * haifa-sched.c (dump_insn_stream): Don't crash if there is a label
41261         in the currently scheduled RTL fragment.
41263 2017-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
41265         PR rtl-optimization/78751
41266         * ifcvt.c (find_cond_trap): If we generated a non-existing insn,
41267         give up.
41269 2017-01-14  Jeff Law  <law@redhat.com>
41271         PR tree-optimization/79090
41272         * tree-ssa-dse.c (valid_ao_ref_for_dse): Reject zero length and
41273         variable length stores.
41274         (compute_trims): Delete dead assignment to *trim_tail.
41275         (dse_dom_walker::dse_optimize_stmt): Optimize mem* calls with
41276         zero length.
41278 2017-01-14  Bernd Schmidt  <bschmidt@redhat.com>
41280         PR rtl-optimization/78626
41281         PR rtl-optimization/78727
41282         * cprop.c (one_cprop_pass): Collect unconditional traps in the middle
41283         of a block, and split such blocks after everything else is finished.
41285 2017-01-14  Alan Modra  <amodra@gmail.com>
41287         PR target/72749
41288         * combine.c (recog_for_combine_1): Set INSN_CODE before calling
41289         target legitimate_combined_insn.
41290         * config/rs6000/rs6000.c (TARGET_LEGITIMATE_COMBINED_INSN): Define.
41291         (rs6000_legitimate_combined_insn): New function.
41292         * config/rs6000/rs6000.md (UNSPEC_DOLOOP): Delete, and remove
41293         all uses.
41294         (ctr<mode>_internal3): Rename from *ctr<mode>_internal5.
41295         (ctr<mode>_internal4): Rename from *ctr<mode>_internal6.
41296         (ctr<mode>_internal1, ctr<mode>_internal2): Remove '*' from name.
41298 2017-01-14  Gerald Pfeifer  <gerald@pfeifer.com>
41300         * doc/frontends.texi (G++ and GCC): Remove references to Java.
41302 2017-01-13  Jeff Law  <law@redhat.com>
41304         PR tree-optimization/33562
41305         PR tree-optimization/61912
41306         PR tree-optimization/77485
41307         * tree-ssa-dse.c (delete_dead_call): Accept gsi rather than
41308         a statement.
41309         (delete_dead_assignment): Likewise.
41310         (dse_dom_walker::dse_optimize_stmt): Pass in the gsi rather than
41311         statement to delete_dead_call and delete_dead_assignment.
41313 2017-01-13  David Malcolm  <dmalcolm@redhat.com>
41315         PR c/78304
41316         * substring-locations.c (format_warning_va): Strengthen case 1 so
41317         that both endpoints of the substring must be within the format
41318         range for just the substring to be printed.
41320 2017-01-13  Uros Bizjak  <ubizjak@gmail.com>
41322         * config/i386/i386.opt (msgx): Use ix86_isa_flags2 variable.
41323         * config/i386/i386.c (ix86_target_string): Add missing options
41324         to isa_opts and reorder options by implied ISAs.  Rename isa_opts2 to
41325         isa2_opts, ix86_flag_opts to flag2_opts, ix86_target_other to
41326         flags_other and ix86_target_other to flags2_other.  Display unknown
41327         isa2 options.
41328         (ix86_valid_target_attribute_inner_p): Add missing options and
41329         reorder options by implied ISAs, as in ix86_target_string.
41331 2017-01-13  Richard Sandiford  <richard.sandiford@arm.com>
41333         * hash-table.h (hash_table::too_empty_p): New function.
41334         (hash_table::expand): Use it.
41335         (hash_table::traverse): Likewise.
41336         (hash_table::empty_slot): Use sizeof (value_type) instead of
41337         sizeof (PTR) to convert bytes to elements.  Shrink the table
41338         if the current size is excessive for the current number of
41339         elements.
41341 2017-01-13  Richard Sandiford  <richard.sandiford@arm.com>
41343         * ira-costs.c (record_reg_classes): Break from the inner loop
41344         early once alt_fail is known to be true.  Update outer loop
41345         handling accordingly.
41347 2017-01-13  Jeff Law  <law@redhat.com>
41349         * tree-ssa-dse.c (decrement_count): New function.
41350         (increment_start_addr, maybe_trim_memstar_call): Likewise.
41351         (dse_dom_walker::optimize_stmt): Call maybe_trim_memstar_call directly
41352         when we know the partially dead statement is a mem* function.
41354         PR tree-optimization/61912
41355         PR tree-optimization/77485
41356         * tree-ssa-dse.c: Include expr.h.
41357         (maybe_trim_constructor_store): New function.
41358         (maybe_trim_partially_dead_store): Call maybe_trim_constructor_store.
41360         PR tree-optimization/33562
41361         PR tree-optimization/61912
41362         PR tree-optimization/77485
41363         * doc/invoke.texi: Document new dse-max-object-size param.
41364         * params.def (PARM_DSE_MAX_OBJECT_SIZE): New PARAM.
41365         * tree-ssa-dse.c: Include params.h.
41366         (dse_store_status): New enum.
41367         (initialize_ao_ref_for_dse): New, partially extracted from
41368         dse_optimize_stmt.
41369         (valid_ao_ref_for_dse, normalize_ref): New.
41370         (setup_live_bytes_from_ref, compute_trims): Likewise.
41371         (clear_bytes_written_by, maybe_trim_complex_store): Likewise.
41372         (maybe_trim_partially_dead_store): Likewise.
41373         (maybe_trim_complex_store): Likewise.
41374         (dse_classify_store): Renamed from dse_possibly_dead_store_p.
41375         Track what bytes live from the original store.  Return tri-state
41376         for dead, partially dead or live.
41377         (dse_dom_walker): Add constructor, destructor and new private members.
41378         (delete_dead_call, delete_dead_assignment): New extracted from
41379         dse_optimize_stmt.
41380         (dse_optimize_stmt): Make a member of dse_dom_walker.
41381         Use initialize_ao_ref_for_dse.
41383         PR tree-optimization/33562
41384         PR tree-optimization/61912
41385         PR tree-optimization/77485
41386         * sbitmap.h (bitmap_count_bits): Prototype.
41387         (bitmap_clear_range, bitmap_set_range): Likewise.
41388         * sbitmap.c (bitmap_clear_range): New function.
41389         (bitmap_set_range, sbitmap_popcount, bitmap_count_bits): Likewise.
41391 2017-01-13  Martin Liska  <mliska@suse.cz>
41393         PR ipa/79043
41394         * function.c (set_cfun): Add new argument force.
41395         * function.h (set_cfun): Likewise.
41396         * ipa-inline-transform.c (inline_call): Use the function when
41397         strict alising from is dropped for function we inline to.
41399 2017-01-13  Richard Biener  <rguenther@suse.de>
41401         * tree-pretty-print.c (dump_generic_node): Fix inverted condition
41402         for dumping GIMPLE INTEGER_CSTs.
41404 2017-01-13  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
41406         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define __STDC_VERSION__
41407         to 201112L since C++17.
41409 2017-01-13  Maxim Ostapenko  <m.ostapenko@samsung.com>
41411         PR sanitizer/78887
41412         * asan.c (asan_needs_odr_indicator_p): Don't emit ODR indicators
41413         if -fsanitize=kernel-address is present.
41415 2017-01-13  Richard Biener  <rguenther@suse.de>
41417         * tree-pretty-print.c (dump_generic_node): Dump INTEGER_CSTs
41418         as _Literal ( type ) number in case usual suffixes do not
41419         preserve all information.
41421 2017-01-13  Richard Biener  <rguenther@suse.de>
41423         PR tree-optimization/77283
41424         * gimple-ssa-split-paths.c: Include gimple-ssa.h, tree-phinodes.h
41425         and ssa-iterators.h.
41426         (is_feasible_trace): Implement a cost model based on joiner
41427         PHI node uses.
41429 2017-01-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
41431         PR target/79004
41432         * config/rs6000/rs6000.md (FP_ISA3): Do not optimize converting
41433         char or short to __float128/_Float128 directly.
41435 2017-01-12  Martin Sebor  <msebor@redhat.com>
41437         to -Wformat-overflow.
41438         * gimple-ssa-sprintf.c (pass_sprintf_length::gate): Adjust.
41439         (min_bytes_remaining): Same.
41440         (get_string_length): Same.
41441         (format_string): Same.
41442         (format_directive): Same.
41443         (add_bytes): Same.
41444         (pass_sprintf_length::handle_gimple_call): Same.
41446 2017-01-12  Jakub Jelinek  <jakub@redhat.com>
41448         * gimple-ssa-sprintf.c (try_substitute_return_value): Remove
41449         info.nowrite calls with no lhs that can't throw.  Return bool
41450         whether gsi_remove has been called or not.
41451         (pass_sprintf_length::handle_gimple_call): Return bool whether
41452         try_substitute_return_value called gsi_remove.  Formatting fix.
41453         (pass_sprintf_length::execute): Don't use gsi_remove if
41454         handle_gimple_call returned true.
41456         PR bootstrap/79069
41457         * cfgrtl.c (rtl_tidy_fallthru_edge): For any_uncondjump_p that can't
41458         be removed due to side-effects, don't remove following barrier nor
41459         turn the successor edge into fallthru edge.
41461 2017-01-12  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
41463         PR target/79044
41464         * config/rs6000/rs6000.c (insn_is_swappable_p): Mark
41465         element-reversing loads and stores as not swappable.
41467 2017-01-12  Nathan Sidwell  <nathan@acm.org>
41468             Nicolai Stange  <nicstange@gmail.com>
41470         * combine.c (try_combine): Don't ignore result of overlap checking
41471         loop.  Combine overlap & asm check into single loop.
41473 2017-01-12  Richard Biener  <rguenther@suse.de>
41475         * tree-pretty-print.c (dump_generic_node): Provide -gimple
41476         variant for MEM_REF.  Sanitize INTEGER_CST for -gimple.
41478 2017-01-12  Richard Biener  <rguenther@suse.de>
41480         * tree.c (initialize_tree_contains_struct): Make TS_OPTIMIZATION
41481         and TS_TARGET_OPTION directly derive from TS_BASE.
41482         * tree-core.h (tree_optimization_option): Derive from tree_base.
41483         (tree_target_option): Likewise.
41485 2017-01-11  Uros Bizjak  <ubizjak@gmail.com>
41487         * config/i386/i386.c (memory_address_length): Increase len
41488         only when rip_relative_addr_p returns false.
41490 2017-01-11  Julia Koval  <julia.koval@intel.com>
41492         * common/config/i386/i386-common.c (OPTION_MASK_ISA_SGX_UNSET): New.
41493         (OPTION_MASK_ISA_SGX_SET): New.
41494         (ix86_handle_option): Handle OPT_msgx.
41495         * config.gcc: Added sgxintrin.h.
41496         * config/i386/driver-i386.c (host_detect_local_cpu): Detect sgx.
41497         * config/i386/i386-c.c (ix86_target_macros_internal): Define __SGX__.
41498         * config/i386/i386.c (ix86_target_string): Add -msgx.
41499         (PTA_SGX): New.
41500         (ix86_option_override_internal): Handle new options.
41501         (ix86_valid_target_attribute_inner_p): Add sgx.
41502         * config/i386/i386.h (TARGET_SGX, TARGET_SGX_P): New.
41503         * config/i386/i386.opt: Add msgx.
41504         * config/i386/sgxintrin.h: New file.
41505         * config/i386/x86intrin.h: Add sgxintrin.h.
41507 2017-01-11  Jakub Jelinek  <jakub@redhat.com>
41509         PR c++/71537
41510         * fold-const.c (maybe_nonzero_address): Return 1 for function
41511         local objects.
41512         (tree_single_nonzero_warnv_p): Don't handle function local objects
41513         here.
41515         PR c++/72813
41516         * gcc.c (default_compilers): Don't add -o %g.s for -S -save-temps
41517         of c-header.
41519 2017-01-11  David Malcolm  <dmalcolm@redhat.com>
41521         PR driver/78877
41522         * opts.c: Include "spellcheck.h"
41523         (struct string_fragment): New struct.
41524         (struct edit_distance_traits<const string_fragment &>): New
41525         struct.
41526         (get_closest_sanitizer_option): New function.
41527         (parse_sanitizer_options): Offer suggestions for unrecognized arguments.
41529 2017-01-11  Jakub Jelinek  <jakub@redhat.com>
41531         * dwarf2out.c (DWARF_COMPILE_UNIT_HEADER_SIZE): For DWARF5 decrease
41532         by 12.
41533         (DWARF_COMDAT_TYPE_UNIT_HEADER_SIZE): Always
41534         DWARF_COMPILE_UNIT_HEADER_SIZE plus 12.
41535         (DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE): Define.
41536         (calc_base_type_die_sizes): Use DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE
41537         for initial die_offset if dwarf_split_debug_info.
41538         (output_comp_unit): Use DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE for
41539         initial next_die_offset if dwo_id is non-NULL.  Don't emit padding
41540         fields.
41541         (output_skeleton_debug_sections): Formatting fix.  Use
41542         DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE instead of
41543         DWARF_COMPILE_UNIT_HEADER_SIZE.  Don't emit padding.
41545 2017-01-11  Wilco Dijkstra  <wdijkstr@arm.com>
41547         * config/arm/cortex-a53.md: Add bypasses for
41548         cortex_a53_r2f_cvt.
41549         (cortex_a53_r2f): Only use for transfers.
41550         (cortex_a53_f2r): Likewise.
41551         (cortex_a53_r2f_cvt): Add reservation for conversions.
41552         (cortex_a53_f2r_cvt): Likewise.
41554 2017-01-11  Tamar Christina  <tamar.christina@arm.com>
41556         * config/arm/arm_neon.h: Add __artificial__ and gnu_inline
41557         to all inlined functions, change static to extern.
41559 2017-01-11  Christophe Lyon  <christophe.lyon@linaro.org>
41561         PR target/78253
41562         * config/arm/arm.c (legitimize_pic_address): Handle reference to
41563         weak symbol.
41564         (arm_assemble_integer): Likewise.
41566 2017-01-11  Richard Earnshaw  <rearnsha@arm.com>
41568         * config.gcc: Use new awk script to check CPU, FPU and architecture
41569         parameters for --with-... options.
41570         * config/arm/parsecpu.awk: New file
41571         * config/arm/arm-cpus.in: New file.
41572         * config/arm/arm-opts.h: Include arm-cpu.h instead of processing .def
41573         files.
41574         * config/arm/arm.c: Include arm-cpu-data.h instead of processing .def
41575         files.
41576         * config/arm/t-arm: Update dependency rules.
41577         * common/config/arm/arm-common.c: Include arm-cpu-cdata.h instead
41578         of processing .def files.
41579         * config/arm/genopt.sh: Deleted.
41580         * config/arm/gentune.sh: Deleted.
41581         * config/arm/arm-cores.def: Deleted.
41582         * config/arm/arm-arches.def: Deleted.
41583         * config/arm/arm-fpus.def: Deleted.
41584         * config/arm/arm-tune.md: Regenerated.
41585         * config/arm/arm-tables.opt: Regenerated.
41586         * config/arm/arm-cpu.h: New generated file.
41587         * config/arm/arm-cpu-data.h: New generated file.
41588         * config/arm/arm-cpu-cdata.h: New generated file.
41590 2017-01-11  Maxim Ostapenko  <m.ostapenko@samsung.com>
41592         PR lto/79042
41593         * lto-cgraph.c (lto_output_varpool_node): Pack dynamically_initialized
41594         bit.
41595         (input_varpool_node): Unpack dynamically_initialized bit.
41597 2017-01-11  Eric Botcazou  <ebotcazou@adacore.com>
41599         PR rtl-optimization/79032
41600         * lra-constraints.c (simplify_operand_subreg): In the MEM case, test
41601         the alignment of the adjusted memory reference against that of MODE,
41602         instead of the alignment of the original memory reference.
41604 2017-01-11  Martin Jambor  <mjambor@suse.cz>
41606         * hsa.c (hsa_callable_function_p): Revert addition of DECL_ARTIFICIAL
41607         test.
41608         * ipa-hsa.c (process_hsa_functions): Only duplicate non-artificial
41609         decorated functions.
41611 2017-01-11  Richard Biener  <rguenther@suse.de>
41613         * tree-vrp.c (evrp_dom_walker::before_dom_children): Also
41614         set range/nonnull info for PHI results.  Do not set it on
41615         stmts marked for removal.
41617 2017-01-10  Eric Botcazou  <ebotcazou@adacore.com>
41619         * expr.c (store_field): In the bitfield case, fetch the return value
41620         from the registers before applying a single big-endian adjustment.
41621         Always do a final load for a BLKmode value not larger than a word.
41623 2017-01-10  David Malcolm  <dmalcolm@redhat.com>
41625         PR c++/77949
41626         * input.c (selftest::test_accessing_ordinary_linemaps): Verify
41627         that we correctly handle column numbers greater than
41628         LINE_MAP_MAX_COLUMN_NUMBER.
41630 2017-01-10  Martin Sebor  <msebor@redhat.com>
41632         PR middle-end/78245
41633         * gimple-ssa-sprintf.c (get_destination_size): Call
41634         {init,fini}object_sizes.
41635         * tree-object-size.c (addr_object_size): Adjust.
41636         (pass_through_call): Adjust.
41637         (pass_object_sizes::execute): Adjust.
41638         * tree-object-size.h (fini_object_sizes): Declare.
41640 2017-01-10  Martin Sebor  <msebor@redhat.com>
41642         PR tree-optimization/78775
41643         * builtins.c (get_size_range): Move...
41644         * calls.c: ...to here.
41645         (alloc_max_size): Accept zero argument.
41646         (operand_signed_p): Remove.
41647         (maybe_warn_alloc_args_overflow): Call get_size_range.
41648         * calls.h (get_size_range): Declare.
41650 2017-01-10  Joe Seymour  <joe.s@somniumtech.com>
41652         * config/msp430/driver-msp430.c (msp430_mcu_data): Sync with data
41653         from TI's devices.csv file as of September 2016.
41654         * config/msp430/msp430.c (msp430_mcu_data): Likewise.
41656 2017-01-10  Sandra Loosemore  <sandra@codesourcery.com>
41658         * doc/extend.texi: Tweak formatting to fix overfull hbox warnings.
41659         * doc/invoke.texi: Likewise.
41660         * doc/md.texi: Likewise.
41661         * doc/objc.texi: Likewise.
41663 2017-01-10  Joshua Conner  <joshconner@google.com>
41665         * config/arm/fuchsia-elf.h: New file.
41666         * config/fuchsia.h: New file.
41667         * config.gcc (*-*-fuchsia*): Set native_system_header_dir.
41668         (aarch64*-*-fuchsia*, arm*-*-fuchsia*, x86_64-*-fuchsia*): Add to
41669         targets.
41670         * config.host: (aarch64*-*-fuchsia*, arm*-*-fuchsia*): Add to hosts.
41672 2016-01-10  Richard Biener  <rguenther@suse.de>
41674         PR tree-optimization/79034
41675         * tree-call-cdce.c (shrink_wrap_one_built_in_call_with_conds):
41676         Propagate out degenerate PHIs in the joiner.
41678 2017-01-10  Martin Liska  <mliska@suse.cz>
41680         * ipa-icf.c (sort_sem_items_by_decl_uid): New function.
41681         (sort_congruence_classes_by_decl_uid): Likewise.
41682         (sort_congruence_class_groups_by_decl_uid): Likewise.
41683         (sem_item_optimizer::merge_classes): Sort class, groups in these
41684         classes and members in the groups by DECL_UID of declarations.
41685         This would make merge operations stable.
41687 2017-01-10  Martin Liska  <mliska@suse.cz>
41689         * ipa-icf.c (sem_item_optimizer::sem_item_optimizer): Remove
41690         usage of m_classes_vec.
41691         (sem_item_optimizer::~sem_item_optimizer):  Likewise.
41692         (sem_item_optimizer::get_group_by_hash): Likewise.
41693         (sem_item_optimizer::subdivide_classes_by_equality): Likewise.
41694         (sem_item_optimizer::subdivide_classes_by_sensitive_refs): Likewise.
41695         (sem_item_optimizer::verify_classes): Likewise.
41696         (sem_item_optimizer::process_cong_reduction): Likewise.
41697         (sem_item_optimizer::dump_cong_classes): Likewise.
41698         (sem_item_optimizer::merge_classes): Likewise.
41699         * ipa-icf.h (congruence_class_hash): Rename from
41700         congruence_class_group_hash.  Remove declaration of m_classes_vec.
41702 2017-01-10  Andrew Senkevich  <andrew.senkevich@intel.com>
41704         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VPOPCNTDQ_SET,
41705         OPTION_MASK_ISA_AVX512VPOPCNTDQ_UNSET): New.
41706         * config.gcc: Add avx512vpopcntdqintrin.h.
41707         * config/i386/avx512vpopcntdqintrin.h: New.
41708         * config/i386/cpuid.h (bit_AVX512VPOPCNTDQ): New.
41709         * config/i386/i386-builtin-types.def: Add new types.
41710         * config/i386/i386-builtin.def (__builtin_ia32_vpopcountd_v16si,
41711         __builtin_ia32_vpopcountd_v16si_mask, __builtin_ia32_vpopcountq_v8di,
41712         __builtin_ia32_vpopcountq_v8di_mask): New.
41713         * config/i386/i386-c.c (ix86_target_macros_internal): Define
41714         __AVX512VPOPCNTDQ__.
41715         * config/i386/i386.c (ix86_target_string): Add -mavx512vpopcntdq.
41716         (PTA_AVX512VPOPCNTDQ): Define.
41717         * config/i386/i386.h (TARGET_AVX512VPOPCNTDQ,
41718         TARGET_AVX512VPOPCNTDQ_P): Define.
41719         * config/i386/i386.opt: Add mavx512vpopcntdq.
41720         * config/i386/immintrin.h: Include avx512vpopcntdqintrin.h.
41721         * config/i386/sse.md (define_insn "vpopcount<mode><mask_name>"): New.
41723 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
41725         PR middle-end/77484
41726         * predict.def (PRED_CALL): Set to 67.
41728 2017-01-09  Eric Botcazou  <ebotcazou@adacore.com>
41730         * expr.c (store_field): In the bitfield case, if the value comes from
41731         a function call and is of an aggregate type returned in registers, do
41732         not modify the field mode; extract the value in all cases if the mode
41733         is BLKmode and the size is not larger than a word.
41735 2017-01-09  Dominique d'Humieres  <dominiq@lps.ens.fr>
41737         PR target/71017
41738         * config/i386/cpuid.h: Fix undefined behavior.
41740 2017-01-04  Jeff Law  <law@redhat.com>
41742         PR tree-optimization/79007
41743         PR tree-optimization/67955
41744         * tree-ssa-alias.c (same_addr_size_stores_p): Only need to be
41745         conservative for pt.null when flag_non_call_exceptions is on.
41747 2017-01-09  Jakub Jelinek  <jakub@redhat.com>
41749         PR translation/79019
41750         PR translation/79020
41751         * params.def (PARAM_INLINE_MIN_SPEEDUP,
41752         PARAM_IPA_CP_SINGLE_CALL_PENALTY,
41753         PARAM_USE_AFTER_SCOPE_DIRECT_EMISSION_THRESHOLD): Fix typos
41754         in descriptions.
41755         * config/avr/avr.opt (maccumulate-args): Likewise.
41756         * config/msp430/msp430.opt (mwarn-mcu): Likewise.
41757         * common.opt (freport-bug): Likewise.
41758         * cif-code.def (CIF_FINAL_ERROR): Likewise.
41759         * doc/invoke.texi (ipa-cp-single-call-penalty): Likewise.
41760         * config/s390/s390.c (s390_invalid_binary_op): Fix spelling in
41761         translatable string.
41762         * config/i386/i386.c (function_value_32): Likewise.
41763         * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Likewise.
41764         * config/msp430/msp430.c (msp430_option_override, msp430_attr):
41765         Likewise.
41766         * config/msp430/driver-msp430.c (msp430_select_hwmult_lib): Likewise.
41767         * common/config/msp430/msp430-common.c (msp430_handle_option):
41768         Likewise.
41769         * symtab.c (symtab_node::verify_base): Likewise.
41770         * opts.c (set_debug_level): Likewise.
41771         * tree.c (verify_type_variant): Likewise.  Fix typo in comment.
41772         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Add
41773         missing whitespace to translatable strings.
41774         * config/avr/avr.md (bswapsi2): Fix typo in comment.
41775         * config/sh/superh.h: Likewise.
41776         * config/i386/xopintrin.h: Likewise.
41777         * config/i386/znver1.md: Likewise.
41778         * config/rs6000/rs6000.c (struct rs6000_opt_mask): Likewise.
41779         * ipa-inline-analysis.c (compute_inline_parameters): Likewise.
41780         * double-int.h (struct double_int): Likewise.
41781         * double-int.c (div_and_round_double): Likewise.
41782         * wide-int.cc: Likewise.
41783         * tree-ssa.c (non_rewritable_mem_ref_base): Likewise.
41784         * tree-ssa-sccvn.c (vn_reference_lookup_3): Likewise.
41785         * cfgcleanup.c (crossjumps_occured): Renamed to ...
41786         (crossjumps_occurred): ... this.
41787         (try_crossjump_bb, try_head_merge_bb, try_optimize_cfg, cleanup_cfg):
41788         Adjust all uses.
41790         PR tree-optimization/78899
41791         * tree-if-conv.c (version_loop_for_if_conversion): Instead of
41792         returning bool return struct loop *, NULL for failure and the new
41793         loop on success.
41794         (versionable_outer_loop_p): Don't version outer loop if it has
41795         dont_vectorized bit set.
41796         (tree_if_conversion): When versioning outer loop, ensure
41797         tree_if_conversion is performed also on the inner loop of the
41798         non-vectorizable outer loop copy.
41799         * tree-vectorizer.c (set_uid_loop_bbs): Formatting fix.  Fold
41800         LOOP_VECTORIZED in inner loop of the scalar outer loop and
41801         prevent vectorization of it.
41802         (vectorize_loops): For outer + inner LOOP_VECTORIZED, ensure
41803         the outer loop vectorization of the non-scalar version is attempted
41804         before vectorization of the inner loop in scalar version.  If
41805         outer LOOP_VECTORIZED guarded loop is not vectorized, prevent
41806         vectorization of its inner loop.
41807         * tree-vect-loop-manip.c (rename_variables_in_bb): If outer_loop
41808         has 2 inner loops, rename also on edges from bb whose single pred
41809         is outer_loop->header.  Fix typo in function comment.
41811 2017-01-09  Martin Sebor  <msebor@redhat.com>
41813         PR bootstrap/79033
41814         * asan.c (asan_emit_stack_protection): Increase local buffer size
41815         to avoid snprintf truncation warning.
41817 2017-01-09  Andrew Pinski  <apinski@cavium.com>
41819         * config/aarch64/aarch64-cores.def: Add thunderx2t99.  Change vulcan
41820         to reference thunderx2t99 for the tuning structure
41821         * config/aarch64/aarch64-cost-tables.h (vulcan_extra_costs):
41822         Rename to ...
41823         (thunderx2t99_extra_costs): This.
41824         * config/aarch64/aarch64-tune.md: Regenerate.
41825         * config/aarch64/aarch64.c (vulcan_addrcost_table): Rename to ...
41826         (vulcan_addrcost_table): This.
41827         (vulcan_regmove_cost): Rename to ...
41828         (thunderx2t99_regmove_cost): This.
41829         (vulcan_vector_cost): Rename to ...
41830         (thunderx2t99_vector_cost): this.
41831         (vulcan_branch_cost): Rename to ...
41832         (thunderx2t99_branch_cost): This.
41833         (vulcan_tunings): Rename to ...
41834         (thunderx2t99_tunings): This and s/vulcan/thunderx2t99 .
41835         * doc/invoke.texi (AARCH64/mtune): Add thunderx2t99.
41837 2017-01-09  Martin Jambor  <mjambor@suse.cz>
41839         PR ipa/78365
41840         PR ipa/78599
41841         * ipa-prop.h (ipa_jump_func): Swap positions of vr_known and m_vr.
41842         * ipa-cp.c (ipa_vr_operation_and_type_effects): New function.
41843         (propagate_vr_accross_jump_function): Use the above function for all
41844         value range computations for pass-through jump functions and type
41845         converasion from explicit value range values.
41846         (ipcp_propagate_stage): Do not attempt to deduce types of formal
41847         parameters from TYPE_ARG_TYPES.
41848         * ipa-prop.c (ipa_write_jump_function): Remove trailing whitespace.
41849         (ipa_write_node_info): Stream type of the actual argument.
41850         (ipa_read_node_info): Likewise. Also remove trailing whitespace.
41852 2017-01-09  Martin Liska  <mliska@suse.cz>
41854         PR pch/78970
41855         * gcc.c (driver_handle_option): Handle OPT_E and set have_E.
41856         (lookup_compiler): Do not show error message with have_E.
41858 2017-01-09  Jakub Jelinek  <jakub@redhat.com>
41860         PR tree-optimization/78938
41861         * tree-vect-stmts.c (vectorizable_condition): For non-masked COND_EXPR
41862         where comp_vectype is VECTOR_BOOLEAN_TYPE_P, use
41863         BIT_{NOT,XOR,AND,IOR}_EXPR on the comparison operands instead of
41864         {EQ,NE,GE,GT,LE,LT}_EXPR directly inside of VEC_COND_EXPR.  Formatting
41865         fixes.
41867 2017-01-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
41869         * tree-ssa-address.c (gen_addr_rtx): Don't handle index if it
41870         is const0_rtx.
41872 2017-01-09  Richard Biener  <rguenther@suse.de>
41874         PR tree-optimization/78997
41875         * tree-vect-slp.c (vect_mask_constant_operand_p): Handle SSA
41876         name condition properly.
41878 2017-01-09  Richard Biener  <rguenther@suse.de>
41880         PR debug/79000
41881         * dwarf2out.c (is_cxx): New overload with context.
41882         (is_naming_typedef_decl): Use it.
41884 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41886         * invoke.texi (Option Summary): Correct spacing in option lists
41887         and add line breaks to fix over-long lines.
41889 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41891         PR middle-end/17660
41893         * extend.texi (Common Variable Attributes): Add xref to GCC
41894         Internals manual to explain mode attribute keywords.
41896 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41898         PR other/16519
41899         * doc/invoke.texi (Option Summary): Move -pthread to Linker Options
41900         and Preprocessor Options.
41901         (Options for Linking): Document -pthread here....
41902         (RS/6000 and PowerPC Options): ...not here.
41903         (Solaris 2 Options): ...or here.
41904         * doc/cppopts.texi: Document -pthread.
41906 2017-01-08  Martin Sebor  <msebor@redhat.com>
41908         PR middle-end/77708
41909         * doc/invoke.texi (Warning Options): Document -Wformat-truncation.
41910         * gimple-ssa-sprintf.c (call_info::reval_used, call_info::warnopt):
41911         New member functions.
41912         (format_directive): Used them.
41913         (add_bytes): Same.
41914         (pass_sprintf_length::handle_gimple_call): Same.
41915         * graphite-sese-to-poly.c (tree_int_to_gmp): Increase buffer size
41916         to avoid truncation for any argument.
41917         (extract_affine_mul): Same.
41918         * tree.c (get_file_function_name): Same.
41920 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
41922         PR middle-end/77484
41923         * predict.def (PRED_INDIR_CALL): Set to 86.
41925 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41927         PR preprocessor/54124
41928         * doc/cppopts.texi: Reformat -d subtable to list the full name
41929         of the options.  Add cross-reference to the docs for the general
41930         compiler -d options.
41931         * doc/invoke.texi (Developer Options): Add cross-reference to the
41932         preprocessor-specific -d option documentation.
41934 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41936         PR preprocessor/13498
41937         * doc/cpp.texi (Search Path): Rewrite to remove obsolete and
41938         redudant material, and reflect new command-line options.
41939         (System Headers): Likewise.
41941 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41943         * doc/cppdiropts.texi: Merge documentation of -I, -iquote,
41944         -isystem, and -idirafter.  Copy-edit.
41945         * doc/cppopts.texi: Copy-edit.  Remove contradiction about
41946         default for -ftrack-macro-expansion.  Delete obsolete and
41947         badly-formatted implementation details about -fdebug-cpp output.
41948         * doc/cppwarnopts.texi: Copy-edit.
41950 2017-01-07  David Malcolm  <dmalcolm@redhat.com>
41952         PR c++/72803
41953         * input.c (selftest::test_accessing_ordinary_linemaps): Verify
41954         that the transition from a max line width >= 1<<10 to narrower
41955         lines works correctly.
41957 2017-01-07  Alexandre Oliva  <aoliva@redhat.com>
41959         * doc/options.texi (PerFunction): New.
41960         * opt-functions.awk (switch_flags): Map both Optimization and
41961         PerFunction to CL_OPTIMIZATION.
41962         * opth-gen.awk: Test for PerFunction flag along with
41963         Optimization.
41964         * optc-save-gen.awk: Likewise.  Introduce var_opt_hash and set
41965         it only when the latter is present.  Skip those that don't in
41966         the hash function generator.
41967         * common.opt (fvar-tracking): Mark as PerFunction instead of
41968         Optimization.
41969         (fvar-tracking-assignments): Likewise.
41970         (fvar-tracking-assignments-toggle): Likewise.
41971         (fvar-tracking-uninit): Likewise.
41973 2017-01-07  Jakub Jelinek  <jakub@redhat.com>
41975         PR translation/79018
41976         * params.def (PARAM_MAX_STORES_TO_MERGE): Add missing space between
41977         the and store.
41979 2017-01-06  Mikael Pettersson  <mikpelinux@gmail.com>
41981         PR target/57583
41982         * config/m68k/m68k.opt (LONG_JUMP_TABLE_OFFSETS): New option.
41983         * config/m68k/linux.h (ASM_RETURN_CASE_JUMP): Handle
41984         TARGET_LONG_JUMP_TABLE_OFFSETS.
41985         * config/m68k/m68kelf.h (ASM_RETURN_CASE_JUMP): Likewise.
41986         * config/m68k/netbsd-elf.h (ASM_RETURN_CASE_JUMP): Likewise.
41987         * config/m68k/m68k.h (CASE_VECTOR_MODE): Likewise.
41988         (ASM_OUTPUT_ADDR_DIFF_ELF): Likewise.
41989         * config/m68k/m68k.md (tablejump expander): Likewise.
41990         (*tablejump_pcrel_hi): Renamed from unnamed insn, reject
41991         TARGET_LONG_JUMP_TABLE_OFFSETS.
41992         (*tablejump_pcrel_si): New insn, handle TARGET_LONG_JUMP_TABLE_OFFSETS.
41993         * doc/invoke.texi (M68K options): Add -mlong-jump-table-offsets.
41995 2017-01-06  Edgar E. Iglesias <edgar.iglesias@xilinx.com>
41996             David Holsgrove <david.holsgrove@xilinx.com>
41998         * common/config/microblaze/microblaze-common.c
41999         (TARGET_EXCEPT_UNWIND_INFO): Remove.
42000         * config/microblaze/microblaze-protos.h (microblaze_eh_return):
42001         New prototype.
42002         * config/microblaze/microblaze.c (microblaze_must_save_register)
42003         (microblaze_expand_epilogue, microblaze_return_addr): Handle
42004         calls_eh_return.
42005         (microblaze_eh_return): New function.
42006         * config/microblaze/microblaze.h (RETURN_ADDR_OFFSET)
42007         (EH_RETURN_DATA_REGNO, MB_EH_STACKADJ_REGNUM)
42008         (EH_RETURN_STACKADJ_RTX, ASM_PREFERRED_EH_DATA_FORMAT): New macros.
42009         * config/microblaze/microblaze.md (eh_return): New pattern.
42011 2017-01-06  Jakub Jelinek  <jakub@redhat.com>
42013         * system.h (GCC_DIAGNOSTIC_PUSH_IGNORED, GCC_DIAGNOSTIC_POP,
42014         GCC_DIAGNOSTIC_STRINGIFY): Define.
42016         * read-rtl.c (rtx_reader::read_rtx_code): Avoid -Wsign-compare warning.
42018 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
42020         * config/arm/arm.md (<mcrr>): New.
42021         (<mrrc>): New.
42022         * config/arm/arm.c (arm_arch5te): New.
42023         (arm_option_override): Set arm_arch5te.
42024         (arm_coproc_builtin_available): Add support for mcrr, mcrr2, mrrc
42025         and mrrc2.
42026         * config/arm/arm-builtins.c (MCRR_QUALIFIERS): Define to...
42027         (arm_mcrr_qualifiers): ... this. New.
42028         (MRRC_QUALIFIERS): Define to...
42029         (arm_mrrc_qualifiers): ... this. New.
42030         * config/arm/arm_acle.h (__arm_mcrr, __arm_mcrr2, __arm_mrrc,
42031         __arm_mrrc2): New.
42032         * config/arm/arm_acle_builtins.def (mcrr, mcrr2, mrrc, mrrc2): New.
42033         * config/arm/iterators.md (MCRRI, mcrr, MCRR): New.
42034         (MRRCI, mrrc, MRRC): New.
42035         * config/arm/unspecs.md (VUNSPEC_MCRR, VUNSPEC_MCRR2, VUNSPEC_MRRC,
42036         VUNSPEC_MRRC2): New.
42038 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
42040         * config/arm/arm.md (<mcr>): New.
42041         (<mrc>): New.
42042         * config/arm/arm.c (arm_coproc_builtin_available): Add
42043         support for mcr, mrc, mcr2 and mrc2.
42044         * config/arm/arm-builtins.c (MCR_QUALIFIERS): Define to...
42045         (arm_mcr_qualifiers): ... this. New.
42046         (MRC_QUALIFIERS): Define to ...
42047         (arm_mrc_qualifiers): ... this. New.
42048         (MCR_QUALIFIERS): Define to ...
42049         (arm_mcr_qualifiers): ... this. New.
42050         * config/arm/arm_acle.h (__arm_mcr, __arm_mrc, __arm_mcr2,
42051         __arm_mrc2): New.
42052         * config/arm/arm_acle_builtins.def (mcr, mcr2, mrc, mrc2): New.
42053         * config/arm/iterators.md (MCRI, mcr, MCR, MRCI, mrc, MRC): New.
42054         * config/arm/unspecs.md (VUNSPEC_MCR, VUNSPEC_MCR2, VUNSPEC_MRC,
42055         VUNSPEC_MRC2): New.
42057 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
42059         * config/arm/arm.md (*ldc): New.
42060         (*stc): New.
42061         (<ldc>): New.
42062         (<stc>): New.
42063         * config/arm/arm.c (arm_coproc_builtin_available): Add
42064         support for ldc,ldcl,stc,stcl,ldc2,ldc2l,stc2 and stc2l.
42065         (arm_coproc_ldc_stc_legitimate_address): New.
42066         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
42067         'qualifier_const_pointer'.
42068         (LDC_QUALIFIERS): Define to...
42069         (arm_ldc_qualifiers): ... this. New.
42070         (STC_QUALIFIERS): Define to...
42071         (arm_stc_qualifiers): ... this. New.
42072         * config/arm/arm-protos.h
42073         (arm_coproc_ldc_stc_legitimate_address): New.
42074         * config/arm/arm_acle.h (__arm_ldc, __arm_ldcl, __arm_stc,
42075         __arm_stcl, __arm_ldc2, __arm_ldc2l, __arm_stc2, __arm_stc2l): New.
42076         * config/arm/arm_acle_builtins.def (ldc, ldc2, ldcl, ldc2l, stc,
42077         stc2, stcl, stc2l): New.
42078         * config/arm/constraints.md (Uz): New.
42079         * config/arm/iterators.md (LDCI, STCI, ldc, stc, LDC STC): New.
42080         * config/arm/unspecs.md (VUNSPEC_LDC, VUNSPEC_LDC2, VUNSPEC_LDCL,
42081         VUNSPEC_LDC2L, VUNSPEC_STC, VUNSPEC_STC2, VUNSPEC_STCL,
42082         VUNSPEC_STC2L): New.
42084 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
42086         * config/arm/arm.md (<cdp>): New.
42087         * config/arm/arm.c (neon_const_bounds): Rename this ...
42088         (arm_const_bounds): ... this.
42089         (arm_coproc_builtin_available): New.
42090         * config/arm/arm-builtins.c (SIMD_MAX_BUILTIN_ARGS): Increase.
42091         (arm_type_qualifiers): Add 'qualifier_unsigned_immediate'.
42092         (CDP_QUALIFIERS): Define to...
42093         (arm_cdp_qualifiers): ... this. New.
42094         (void_UP): Define.
42095         (arm_expand_builtin_args): Add case for 6 arguments.
42096         * config/arm/arm-protos.h (neon_const_bounds): Rename this ...
42097         (arm_const_bounds): ... this.
42098         (arm_coproc_builtin_available): New.
42099         * config/arm/arm_acle.h (__arm_cdp): New.
42100         (__arm_cdp2): New.
42101         * config/arm/arm_acle_builtins.def (cdp): New.
42102         (cdp2): New.
42103         * config/arm/iterators.md (CDPI,CDP,cdp): New.
42104         * config/arm/neon.md: Rename all 'neon_const_bounds' to
42105         'arm_const_bounds'.
42106         * config/arm/types.md (coproc): New.
42107         * config/arm/unspecs.md (VUNSPEC_CDP, VUNSPEC_CDP2): New.
42108         * doc/extend.texi (ACLE): Add a mention of Coprocessor intrinsics.
42109         * doc/sourcebuild.texi (arm_coproc1_ok, arm_coproc2_ok,
42110         arm_coproc3_ok, arm_coproc4_ok): Document new effective targets.
42112 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
42114         * config/arm/arm-builtins.c (arm_unsigned_binop_qualifiers): New.
42115         (UBINOP_QUALIFIERS): New.
42116         (si_UP): Define.
42117         (acle_builtin_data): New. Change comment.
42118         (arm_builtins): Remove ARM_BUILTIN_CRC32B, ARM_BUILTIN_CRC32H,
42119         ARM_BUILTIN_CRC32W, ARM_BUILTIN_CRC32CB, ARM_BUILTIN_CRC32CH,
42120         ARM_BUILTIN_CRC32CW. Add ARM_BUILTIN_ACLE_BASE and include
42121         arm_acle_builtins.def.
42122         (ARM_BUILTIN_ACLE_PATTERN_START): Define.
42123         (arm_init_acle_builtins): New.
42124         (CRC32_BUILTIN): Remove.
42125         (bdesc_2arg): Remove entries for crc32b, crc32h, crc32w,
42126         crc32cb, crc32ch and crc32cw.
42127         (arm_init_crc32_builtins): Remove.
42128         (arm_init_builtins): Use arm_init_acle_builtins rather
42129         than arm_init_crc32_builtins.
42130         (arm_expand_acle_builtin): New.
42131         (arm_expand_builtin): Use 'arm_expand_acle_builtin'.
42132         * config/arm/arm_acle_builtins.def: New.
42134 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
42136         * config/arm/arm-builtins.c (neon_builtin_datum): Rename to ..
42137         (arm_builtin_datum): ... this.
42138         (arm_init_neon_builtin): Rename to ...
42139         (arm_init_builtin): ... this. Add a new parameters PREFIX
42140         and USE_SIG_IN_NAME.
42141         (arm_init_neon_builtins): Replace 'arm_init_neon_builtin' with
42142         'arm_init_builtin'. Replace type 'neon_builtin_datum' with
42143         'arm_builtin_datum'.
42144         (arm_init_vfp_builtins): Likewise.
42145         (builtin_arg): Rename enum's replacing 'NEON_ARG' with
42146         'ARG_BUILTIN' and add a 'ARG_BUILTIN_NEON_MEMORY.
42147         (arm_expand_neon_args): Rename to ...
42148         (arm_expand_builtin_args): ... this. Rename builtin_arg
42149         enum values and differentiate between ARG_BUILTIN_MEMORY
42150         and ARG_BUILTIN_NEON_MEMORY.
42151         (arm_expand_neon_builtin_1): Rename to ...
42152         (arm_expand_builtin_1): ... this. Rename builtin_arg enum
42153         values, arm_expand_builtin_args and add bool parameter NEON.
42154         (arm_expand_neon_builtin): Use arm_expand_builtin_1.
42155         (arm_expand_vfp_builtin): Likewise.
42156         (NEON_MAX_BUILTIN_ARGS): Remove, it was unused.
42158 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
42160         PR middle-end/77484
42161         * predict.def (PRED_POLYMORPHIC_CALL): Set to 59.
42162         * predict.c (tree_estimate_probability_bb): Reverse direction of
42163         polymorphic call predictor.
42165 2017-01-06  David Malcolm  <dmalcolm@redhat.com>
42167         * passes.c (execute_one_pass): Split out pass-skipping logic into...
42168         (determine_pass_name_match): ...this new function and...
42169         (should_skip_pass_p): ...this new function.
42171 2017-01-06  Nathan Sidwell  <nathan@acm.org>
42173         * ipa-visibility.c (function_and_variable_visibility): Reformat
42174         comments and long lines.  Remove extrneous if.
42175         * symtab.c (symtab_node::make_decl_local): Fix code format.
42176         (symtab_node::set_section_for_node): Fix comment typo.
42178 2017-01-06  Martin Liska  <mliska@suse.cz>
42180         PR bootstrap/79003
42181         * lra-constraints.c: Rename invariant to lra_invariant.
42182         * predict.c (set_even_probabilities): Initialize e to NULL.
42184 2017-01-05  Martin Sebor  <msebor@redhat.com>
42186         PR tree-optimization/78910
42187         * gimple-ssa-sprintf.c (tree_digits): Add an argument.
42188         (format_integer): Correct off-by-one error in the handling
42189         of precision with negative numbers in signed conversions..
42191 2017-01-05  Eric Botcazou  <ebotcazou@adacore.com>
42193         * doc/invoke.texi (C Dialect Options): Adjust -fsso-struct entry.
42195 2017-01-05  Jakub Jelinek  <jakub@redhat.com>
42197         PR tree-optimization/71016
42198         * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Pass cond_stmt to
42199         factor_out_conditional_conversion.  Formatting fix.
42200         (factor_out_conditional_conversion): Add cond_stmt argument.
42201         If arg1 is INTEGER_CST, punt if new_arg0 is not any operand of
42202         cond_stmt and if arg0_def_stmt is not the only stmt in its bb.
42203         Formatting fix.
42205 2017-01-05  David Malcolm  <dmalcolm@redhat.com>
42207         * Makefile.in (OBJS): Add read-md.o, read-rtl.o,
42208         read-rtl-function.o, and selftest-rtl.o.
42209         * config/aarch64/aarch64.c: Include selftest.h and selftest-rtl.h.
42210         (selftest::aarch64_test_loading_full_dump): New function.
42211         (selftest::aarch64_run_selftests): New function.
42212         (TARGET_RUN_TARGET_SELFTESTS): Wire it up to
42213         selftest::aarch64_run_selftests.
42214         * config/i386/i386.c
42215         (selftest::ix86_test_loading_dump_fragment_1): New function.
42216         (selftest::ix86_test_loading_call_insn): New function.
42217         (selftest::ix86_test_loading_full_dump): New function.
42218         (selftest::ix86_test_loading_unspec): New function.
42219         (selftest::ix86_run_selftests): Call the new functions.
42220         * emit-rtl.c (maybe_set_max_label_num): New function.
42221         * emit-rtl.h (maybe_set_max_label_num): New decl.
42222         * function.c (instantiate_decls): Guard call to
42223         instantiate_decls_1 with if (DECL_INITIAL (fndecl)).
42224         * function-tests.c (selftest::verify_three_block_rtl_cfg): Remove
42225         "static".
42226         * gensupport.c (gen_reader::gen_reader): Pass "false"
42227         for new "compact" param of rtx_reader.
42228         * print-rtl.c (rtx_writer::print_rtx_operand): Print "(nil)"
42229         rather than an empty string for NULL strings.
42230         * read-md.c: Potentially include config.h rather than bconfig.h.
42231         Wrap include of errors.h with #ifdef GENERATOR_FILE.
42232         (have_error): New global, copied from errors.c.
42233         (md_reader::read_name): Rename to...
42234         (md_reader::read_name_1): ...this, adding "out_loc" param,
42235         and converting "missing name or number" to returning false, rather
42236         than failing.
42237         (md_reader::read_name): Reimplement in terms of read_name_1.
42238         (md_reader::read_name_or_nil): New function.
42239         (md_reader::read_string): Handle "(nil)" by returning NULL.
42240         (md_reader::md_reader): Add new param "compact".
42241         (md_reader::read_md_files): Wrap with #ifdef GENERATOR_FILE.
42242         (md_reader::read_file): New method.
42243         * read-md.h (md_reader::md_reader): Add new param "compact".
42244         (md_reader::read_file): New method.
42245         (md_reader::is_compact): New accessor.
42246         (md_reader::read_name): Convert return type from void to file_location.
42247         (md_reader::read_name_or_nil): New decl.
42248         (md_reader::read_name_1): New decl.
42249         (md_reader::m_compact): New field.
42250         (noop_reader::noop_reader): Pass "false" for new "compact" param
42251         of rtx_reader.
42252         (rtx_reader::rtx_reader): Add new "compact" param.
42253         (rtx_reader::read_rtx_operand): Make virtual and convert return
42254         type from void to rtx.
42255         (rtx_reader::read_until): New decl.
42256         (rtx_reader::handle_any_trailing_information): New virtual function.
42257         (rtx_reader::postprocess): New virtual function.
42258         (rtx_reader::finalize_string): New virtual function.
42259         (rtx_reader::m_in_call_function_usage): New field.
42260         (rtx_reader::m_reuse_rtx_by_id): New field.
42261         * read-rtl-function.c: New file.
42262         * selftest-rtl.c (selftest::assert_rtx_ptr_eq_at): New function.
42263         * selftest-rtl.h (ASSERT_RTX_PTR_EQ): New macro.
42264         (selftest::verify_three_block_rtl_cfg): New decl.
42265         * read-rtl-function.h: New file.
42266         * read-rtl.c: Potentially include config.h rather than bconfig.h.
42267         For host, include function.h, memmodel.h, and emit-rtl.h.
42268         (one_time_initialization): New function.
42269         (struct compact_insn_name): New struct.
42270         (compact_insn_names): New array.
42271         (find_code): Handle insn codes in compact dumps.
42272         (apply_subst_iterator): Wrap with #ifdef GENERATOR_FILE.
42273         (bind_subst_iter_and_attr): Likewise.
42274         (add_condition_to_string): Likewise.
42275         (add_condition_to_rtx): Likewise.
42276         (apply_attribute_uses): Likewise.
42277         (add_current_iterators): Likewise.
42278         (apply_iterators): Likewise.
42279         (initialize_iterators): Guard usage of apply_subst_iterator with
42280         #ifdef GENERATOR_FILE.
42281         (read_conditions): Wrap with #ifdef GENERATOR_FILE.
42282         (md_reader::read_mapping): Likewise.
42283         (add_define_attr_for_define_subst): Likewise.
42284         (add_define_subst_attr): Likewise.
42285         (read_subst_mapping): Likewise.
42286         (check_code_iterator): Likewise.
42287         (rtx_reader::read_rtx): Likewise.  Move one-time initialization
42288         logic to...
42289         (one_time_initialization): New function.
42290         (rtx_reader::read_until): New method.
42291         (read_flags): New function.
42292         (parse_reg_note_name): New function.
42293         (rtx_reader::read_rtx_code): Initialize "iterator" to NULL.
42294         Handle reuse_rtx ids.
42295         Wrap iterator lookup within #ifdef GENERATOR_FILE.
42296         Add parsing support for RTL dumps, mirroring the special-cases in
42297         print_rtx, by calling read_flags, reading REG_NOTE names, INSN_UID
42298         values, and calling handle_any_trailing_information.
42299         (rtx_reader::read_rtx_operand): Convert return type from void
42300         to rtx, returning return_rtx.  Handle case 'e'.  Call
42301         finalize_string on XSTR and XTMPL fields.
42302         (rtx_reader::read_nested_rtx):  Handle dumps in which trailing
42303          "(nil)" values were omitted.  Call the postprocess vfunc on the
42304         return_rtx.
42305         (rtx_reader::rtx_reader): Add new "compact" param and pass to base
42306         class ctor.  Initialize m_in_call_function_usage.  Call
42307         one_time_initialization.
42308         * rtl-tests.c (selftest::test_uncond_jump): Call
42309         set_new_first_and_last_insn.
42310         * rtl.h (read_rtx): Wrap decl with #ifdef GENERATOR_FILE.
42311         * selftest-rtl.c: New file.
42312         * selftest-rtl.h (class selftest::rtl_dump_test): New class.
42313         (selftest::get_insn_by_uid): New decl.
42314         * selftest-run-tests.c (selftest::run_tests): Call
42315         read_rtl_function_c_tests.
42316         * selftest.h  (selftest::read_rtl_function_c_tests): New decl.
42317         * tree-dfa.c (ssa_default_def): Return NULL_TREE for rtl function
42318         dumps.
42320 2017-01-05  Uros Bizjak  <ubizjak@gmail.com>
42322         * config/i386/i386.md (*testqi_ext_3): No need to handle memory
42323         operands in a special way.  Assert that pos+len <= mode precision.
42325 2017-01-05  Jakub Jelinek  <jakub@redhat.com>
42327         * common.opt (fvect-cost-model): Remove RejectNegative flag, use
42328         3 argument Alias with unlimited for the negative form.
42329         (fno-vect-cost-model): Removed.
42331 2017-01-05  Martin Liska  <mliska@suse.cz>
42333         * hsa-gen.c (gen_hsa_divmod): New function.
42334         (gen_hsa_insn_for_internal_fn_call): Use the function for IFN_DIVMOD.
42336 2017-01-05  Martin Liska  <mliska@suse.cz>
42338         PR pch/78970
42339         * gcc.c (lookup_compiler): Reject '-' filename for a precompiled
42340         header.
42342 2017-01-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
42344         * config/s390/s390.c (s390_expand_setmem): Unroll the loop for
42345         small constant length operands.
42347 2017-01-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
42349         * config/s390/s390.c (s390_expand_setmem): Avoid overlapping bytes
42350         between loop iterations.
42352 2017-01-05  Martin Liska  <mliska@suse.cz>
42354         PR sanitizer/78815
42355         * gimplify.c (gimplify_decl_expr): Compare to
42356         asan_poisoned_variables instread of checking flags.
42357         (gimplify_target_expr): Likewise.
42358         (gimplify_expr): Likewise.
42359         (gimplify_function_tree): Conditionally initialize
42360         asan_poisoned_variables.
42362 2017-01-04  Jeff Law  <law@redhat.com>
42364         PR tree-optimizatin/78812
42365         * rtl.h (contains_mem_rtx_p): Prototype.
42366         * ifcvt.c (containts_mem_rtx_p): Move from here to...
42367         * rtlanal.c (contains_mem_rtx_p): Here and remove static linkage.
42368         * gcse.c (prune_expressions): Use contains_mem_rtx_p to discover
42369         and prune MEMs that are not at the toplevel of a SET_SRC rtx.  Look
42370         through ZERO_EXTEND and SIGN_EXTEND when trying to avoid pruning MEMs.
42372 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42374         * input.c (assert_char_at_range): Default-initialize actual_range.
42376 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42378         * df-scan.c (df_ref_create_structure): Make regno unsigned,
42379         to match the caller.
42381 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42383         * cfgexpand.c (expand_gimple_basic_block): Disregard debug
42384         insns after final jump in test to emit dummy move.
42386 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42388         * gimple-iterator.h (gsi_one_nondebug_before_end_p): New.
42389         * tree-eh.c (cleanup_empty_eh): Skip more debug stmts.
42391 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42393         * multiple_target.c (create_dispatcher_calls): Init e_next.
42394         * tree-ssa-loop-split.c (split_loop): Init border.
42395         * tree-vect-loop.c (vect_determine_vectorization_factor): Init
42396         scalar_type.
42398 2017-01-04  Michael Meissner  <meissner@linux.vnet.ibm.com>
42400         PR target/71977
42401         PR target/70568
42402         PR target/78823
42403         * config/rs6000/predicates.md (sf_subreg_operand): New predicate.
42404         (altivec_register_operand): Do not return true if the operand
42405         contains a SUBREG mixing SImode and SFmode.
42406         (vsx_register_operand): Likewise.
42407         (vsx_reg_sfsubreg_ok): New predicate.
42408         (vfloat_operand): Do not return true if the operand contains a
42409         SUBREG mixing SImode and SFmode.
42410         (vint_operand): Likewise.
42411         (vlogical_operand): Likewise.
42412         (gpc_reg_operand): Likewise.
42413         (int_reg_operand): Likewise.
42414         * config/rs6000/rs6000-protos.h (valid_sf_si_move): Add declaration.
42415         * config/rs6000/rs6000.c (valid_sf_si_move): New function to
42416         determine if a MOVSI or MOVSF operation contains SUBREGs that mix
42417         SImode and SFmode.
42418         (rs6000_emit_move_si_sf_subreg): New helper function.
42419         (rs6000_emit_move): Call rs6000_emit_move_si_sf_subreg to possbily
42420         fixup SUBREGs involving SImode and SFmode.
42421         * config/rs6000/vsx.md (SFBOOL_*): New constants that are operand
42422         numbers for the new peephole2 optimization.
42423         (peephole2 for SFmode unions): New peephole2 to optimize cases in
42424         the GLIBC math library that do AND/IOR/XOR operations on single
42425         precision floating point.
42426         * config/rs6000/rs6000.h (TARGET_NO_SF_SUBREG): New internal
42427         target macros to say whether we need to avoid SUBREGs mixing
42428         SImode and SFmode.
42429         (TARGET_ALLOW_SF_SUBREG): Likewise.
42430         * config/rs6000/rs6000.md (UNSPEC_SF_FROM_SI): New unspecs.
42431         (UNSPEC_SI_FROM_SF): Likewise.
42432         (iorxor): Change spacing.
42433         (and_ior_xor): New iterator for AND, IOR, and XOR.
42434         (movsi_from_sf): New insns for SImode/SFmode SUBREG support.
42435         (movdi_from_sf_zero_ext): Likewise.
42436         (mov<mode>_hardfloat, FMOVE32 iterator): Use register_operand
42437         instead of gpc_reg_operand.  Add SImode/SFmode SUBREG support.
42438         (movsf_from_si): New insn for SImode/SFmode SUBREG support.
42439         (fma<mode>4): Use gpc_reg_operand instead of register_operand.
42440         (fms<mode>4): Likewise.
42441         (fnma<mode>4): Likewise.
42442         (fnms<mode>4): Likewise.
42443         (nfma<mode>4): Likewise.
42444         (nfms<mode>4): Likewise.
42446 2017-01-04  Marek Polacek  <polacek@redhat.com>
42448         PR c++/64767
42449         * doc/invoke.texi: Document -Wpointer-compare.
42451 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42453         * optc-gen.awk: Emit #error for -W*/-f*/-m* Enum without
42454         RejectNegative.
42456         * dwarf2out.c (output_loc_list): Don't throw away 64K+ location
42457         descriptions for -gdwarf-5 and emit them as uleb128 instead of
42458         2-byte data.
42460 2017-01-04  Kelvin Nilsen  <kelvin@gcc.gnu.org>
42462         PR target/78056
42463         * doc/sourcebuild.texi (PowerPC-specific attributes): Add
42464         documentation of the powerpc_popcntb_ok attribute.
42465         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
42466         code to issue warning messages if a requested CPU configuration is
42467         not supported by the binary (assembler and loader) toolchain.
42468         (spe_init_builtins): Add two assertions to prevent ICE if attempt is
42469         made to define a built-in function that has been disabled.
42470         (paired_init_builtins): Add assertion to prevent ICE if attempt is
42471         made to define a built-in function that has been disabled.
42472         (altivec_init_builtins): Add comment explaining why definition
42473         of the DST built-in functions is not preceded by an assertion
42474         check.  Add assertions to prevent ICE if attempts are made to
42475         define an altivec predicate or an abs* built-in function that has
42476         been disabled.
42477         (htm_init_builtins): Add comment explaining why definition of the
42478         htm built-in functions is not preceded by an assertion check.
42480 2017-01-04  Jeff Law  <law@redhat.com>
42482         PR tree-optimizatin/67955
42483         * tree-ssa-alias.c (same_addr_size_stores_p): Check offsets first.
42484         Allow any SSA_VAR_P as the base objects.  Use integer_zerop.  Verify
42485         the points-to solution does not include pt_null.  Use DECL_PT_UID
42486         unconditionally.
42488 2017-01-04  Uros Bizjak  <ubizjak@gmail.com>
42490         * config/i386/i386.md (HI/SImode test with imm to QImode splitters):
42491         Use gen_int_mode instead of gen_lopwart for const_int operands.
42493 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42495         PR tree-optimization/71563
42496         * match.pd: Simplify X << Y into X if Y is known to be 0 or
42497         out of range value - has low bits known to be zero.
42499 2017-01-04  Alan Modra  <amodra@gmail.com>
42501         * Makefile.in (aclocal_deps): Update and order as per aclocal.m4.
42502         * configure: Regenerate.
42503         * config.in: Regenerate.
42505 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42507         PR bootstrap/77569
42508         * input.c (ebcdic_execution_charset::on_error): Don't use strstr for
42509         a substring of the message, but strcmp with the whole message.  Ifdef
42510         ENABLE_NLS, translate the message first using dgettext.
42512 2017-01-03  Jeff Law  <law@redhat.com>
42514         PR tree-optimizatin/78856
42515         * tree-ssa-threadupdate.c: Include tree-vectorizer.h.
42516         (mark_threaded_blocks): Remove code to truncate thread paths that
42517         cross multiple loop headers.  Instead invalidate the cached loop
42518         iteration information and handle case of a thread path walking
42519         into an irreducible region.
42521 2017-01-03  Michael Meissner  <meissner@linux.vnet.ibm.com>
42523         PR target/78900
42524         * config/rs6000/rs6000.c (rs6000_split_signbit): Change some
42525         assertions.  Add support for doing the signbit if the IEEE 128-bit
42526         floating point value is in a GPR.
42527         * config/rs6000/rs6000.md (Fsignbit): Delete.
42528         (signbit<mode>2_dm): Delete using <Fsignbit> and just use "wa".
42529         Update the length attribute if the value is in a GPR.
42530         (signbit<mode>2_dm_<su>ext): Add combiner pattern to eliminate
42531         the sign or zero extension instruction, since the value is always 0/1.
42532         (signbit<mode>2_dm2): Delete using <Fsignbit>.
42534         PR target/78953
42535         * config/rs6000/vsx.md (vsx_extract_<mode>_store_p9): If we are
42536         extracting SImode to a GPR register so that we can generate a
42537         store, limit the vector to be in a traditional Altivec register
42538         for the vextuwrx instruction.
42540 2017-01-03  Ian Lance Taylor  <iant@google.com>
42542         * godump.c (go_format_type): Treat ENUMERAL_TYPE like INTEGER_TYPE.
42544 2017-01-03  Martin Sebor  <msebor@redhat.com>
42546         PR tree-optimization/78696
42547         * gimple-ssa-sprintf.c (format_floating): Correct handling of
42548         precision.  Use MPFR for %f for greater fidelity.  Correct handling
42549         of %g.
42550         (pass_sprintf_length::compute_format_length): Set width and precision
42551         specified by asrerisk to void_node for vararg functions.
42552         (try_substitute_return_value): Adjust dump output.
42554 2017-01-03  David Edelsohn  <dje.gcc@gmail.com>
42556         * doc/invoke.texi (RS6000 options): LRA is enabled by default.
42558 2017-01-03  Eric Botcazou  <ebotcazou@adacore.com>
42560         * doc/invoke.texi (SPARC options): Document -mlra as the default.
42561         * config/sparc/sparc.c (sparc_option_override): Force LRA unless
42562         -mlra/-mno-lra was passed to the compiler.
42564 2017-01-03  James Cowgill  <James.Cowgill@imgtec.com>
42566         PR rtl-optimization/65618
42567         * emit-rtl.c (try_split): Move initialization of "before" and
42568         "after" to just before the call to emit_insn_after_setloc.
42570 2017-01-03  Gerald Pfeifer  <gerald@pfeifer.com>
42572         * doc/md.texi (Standard Names): Remove reference to Java frontend.
42574 2017-01-03  Pierre-Marie de Rodat  <derodat@adacore.com>
42576         * dwarf2out.c (gen_enumeration_type_die): When
42577         -gno-strict-dwarf, add a DW_AT_encoding attribute.
42579 2017-01-03  Jakub Jelinek  <jakub@redhat.com>
42581         PR tree-optimization/78965
42582         * gimple-ssa-sprintf.c (pass_sprintf_length::compute_format_length):
42583         Change first argument from const call_info & to call_info &.  For %n
42584         set info.nowrite to false.
42586         PR middle-end/78901
42587         * gimple-ssa-sprintf.c (try_substitute_return_value): Don't change
42588         possibly throwing calls.
42590         * genmatch.c (dt_node::gen_kids_1): If generic_exprs include SSA_NAME
42591         and exprs_len || fns_len, emit the code for SSA_NAME next to the exprs
42592         and fns handling, rather than in a separate case SSA_NAME.
42594 2017-01-02  Jeff Law  <law@redhat.com>
42596         * config/darwin-driver.c (darwin_driver_init): Const-correctness
42597         fixes for first_period and second_period variables.
42599 2017-01-02  Uros Bizjak  <ubizjak@gmail.com>
42601         PR target/78967
42602         * config/i386/i386.md (UNSPEC_NOREX_MEM): New unspec.
42603         (*insvqi_1): New insn pattern.
42604         (*insvqi_1_mem_rex64): Ditto.
42605         (*insvqi_2): Ditto.
42606         (*insvqi_3): Rename from *insvqi.
42608         (*extzvqi_mem_rex64): Add UNSPEC_NOREX_MEM tag.
42610 2017-01-02  Gerald Pfeifer  <gerald@pfeifer.com>
42612         * doc/cfg.texi (Edges): Remove reference to Java.
42613         (Maintaining the CFG): Ditto.
42615 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
42617         PR middle-end/77674
42618         * symtab.c (symtab_node::binds_to_current_def_p): Fix handling of
42619         transparent aliases.
42621 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
42623         PR middle-end/77484
42624         * predict.def (PRED_CALL): Update hitrate.
42625         (PRED_INDIR_CALL, PRED_POLYMORPHIC_CALL): New predictors.
42626         * predict.c (tree_estimate_probability_bb): Split CALL predictor
42627         into direct/indirect/polymorphic variants.
42629 2017-01-01  Jakub Jelinek  <jakub@redhat.com>
42631         Update copyright years.
42633         * gcc.c (process_command): Update copyright notice dates.
42634         * gcov-dump.c (print_version): Ditto.
42635         * gcov.c (print_version): Ditto.
42636         * gcov-tool.c (print_version): Ditto.
42637         * gengtype.c (create_file): Ditto.
42638         * doc/cpp.texi: Bump @copying's copyright year.
42639         * doc/cppinternals.texi: Ditto.
42640         * doc/gcc.texi: Ditto.
42641         * doc/gccint.texi: Ditto.
42642         * doc/gcov.texi: Ditto.
42643         * doc/install.texi: Ditto.
42644         * doc/invoke.texi: Ditto.
42646 Copyright (C) 2017 Free Software Foundation, Inc.
42648 Copying and distribution of this file, with or without modification,
42649 are permitted in any medium without royalty provided the copyright
42650 notice and this notice are preserved.