poly_int: expand_expr_real_2
[official-gcc.git] / gcc / ChangeLog
blob6cef62278cea152228c00d155177a6f35e06de29
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         * expr.c (expand_expr_real_2): When handling conversions involving
6         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
7         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
8         as a poly_uint64 too.
10 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
11             Alan Hayward  <alan.hayward@arm.com>
12             David Sherwood  <david.sherwood@arm.com>
14         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
16 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
17             Alan Hayward  <alan.hayward@arm.com>
18             David Sherwood  <david.sherwood@arm.com>
20         * combine.c (can_change_dest_mode): Handle polynomial
21         REGMODE_NATURAL_SIZE.
22         * expmed.c (store_bit_field_1): Likewise.
23         * expr.c (store_constructor): Likewise.
24         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
25         and polynomial REGMODE_NATURAL_SIZE.
26         (gen_lowpart_common): Likewise.
27         * reginfo.c (record_subregs_of_mode): Likewise.
28         * rtlanal.c (read_modify_subreg_p): Likewise.
30 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31             Alan Hayward  <alan.hayward@arm.com>
32             David Sherwood  <david.sherwood@arm.com>
34         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
35         numbers of elements.
37 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
38             Alan Hayward  <alan.hayward@arm.com>
39             David Sherwood  <david.sherwood@arm.com>
41         * match.pd: Cope with polynomial numbers of vector elements.
43 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
44             Alan Hayward  <alan.hayward@arm.com>
45             David Sherwood  <david.sherwood@arm.com>
47         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
48         in a POINTER_PLUS_EXPR.
50 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
51             Alan Hayward  <alan.hayward@arm.com>
52             David Sherwood  <david.sherwood@arm.com>
54         * omp-simd-clone.c (simd_clone_subparts): New function.
55         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
56         (ipa_simd_modify_function_body): Likewise.
58 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
59             Alan Hayward  <alan.hayward@arm.com>
60             David Sherwood  <david.sherwood@arm.com>
62         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
63         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
64         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
65         (expand_vector_condition, vector_element): Likewise.
66         (subparts_gt): New function.
67         (get_compute_type): Use subparts_gt.
68         (count_type_subparts): Delete.
69         (expand_vector_operations_1): Use subparts_gt instead of
70         count_type_subparts.
72 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
73             Alan Hayward  <alan.hayward@arm.com>
74             David Sherwood  <david.sherwood@arm.com>
76         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
77         (vect_compile_time_alias): ...this new function.  Do the calculation
78         on poly_ints rather than trees.
79         (vect_prune_runtime_alias_test_list): Update call accordingly.
81 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
82             Alan Hayward  <alan.hayward@arm.com>
83             David Sherwood  <david.sherwood@arm.com>
85         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
86         numbers of units.
87         (vect_schedule_slp_instance): 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         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
94         constant and extern definitions for variable-length vectors.
95         (vect_get_constant_vectors): Note that the number of units
96         is known to be constant.
98 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
99             Alan Hayward  <alan.hayward@arm.com>
100             David Sherwood  <david.sherwood@arm.com>
102         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
103         of units as polynomial.  Choose between WIDE and NARROW based
104         on multiple_p.
106 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
107             Alan Hayward  <alan.hayward@arm.com>
108             David Sherwood  <david.sherwood@arm.com>
110         * tree-vect-stmts.c (simd_clone_subparts): New function.
111         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
113 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
114             Alan Hayward  <alan.hayward@arm.com>
115             David Sherwood  <david.sherwood@arm.com>
117         * tree-vect-stmts.c (vectorizable_call): Treat the number of
118         vectors as polynomial.  Use build_index_vector for
119         IFN_GOMP_SIMD_LANE.
121 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
122             Alan Hayward  <alan.hayward@arm.com>
123             David Sherwood  <david.sherwood@arm.com>
125         * tree-vect-stmts.c (get_load_store_type): Treat the number of
126         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
127         for variable-length vectors.
128         (vectorizable_mask_load_store): Treat the number of units as
129         polynomial, asserting that it is constant if the condition has
130         already been enforced.
131         (vectorizable_store, vectorizable_load): Likewise.
133 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
134             Alan Hayward  <alan.hayward@arm.com>
135             David Sherwood  <david.sherwood@arm.com>
137         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
138         of units as polynomial.  Punt if we can't tell at compile time
139         which vector contains the final result.
141 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
142             Alan Hayward  <alan.hayward@arm.com>
143             David Sherwood  <david.sherwood@arm.com>
145         * tree-vect-loop.c (vectorizable_induction): Treat the number
146         of units as polynomial.  Punt on SLP inductions.  Use an integer
147         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
148         cast of such a series for variable-length floating-point
149         reductions.
151 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
152             Alan Hayward  <alan.hayward@arm.com>
153             David Sherwood  <david.sherwood@arm.com>
155         * tree.h (build_index_vector): Declare.
156         * tree.c (build_index_vector): New function.
157         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
158         of units as polynomial, forcibly converting it to a constant if
159         vectorizable_reduction has already enforced the condition.
160         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
161         to create a {1,2,3,...} vector.
162         (vectorizable_reduction): Treat the number of units as polynomial.
163         Choose vectype_in based on the largest scalar element size rather
164         than the smallest number of units.  Enforce the restrictions
165         relied on above.
167 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
168             Alan Hayward  <alan.hayward@arm.com>
169             David Sherwood  <david.sherwood@arm.com>
171         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
172         number of units as polynomial.
174 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
175             Alan Hayward  <alan.hayward@arm.com>
176             David Sherwood  <david.sherwood@arm.com>
178         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
179         * target.def (autovectorize_vector_sizes): Return the vector sizes
180         by pointer, using vector_sizes rather than a bitmask.
181         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
182         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
183         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
184         Likewise.
185         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
186         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
187         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
188         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
189         * omp-general.c (omp_max_vf): Likewise.
190         * omp-low.c (omp_clause_aligned_alignment): Likewise.
191         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
192         * tree-vect-loop.c (vect_analyze_loop): Likewise.
193         * tree-vect-slp.c (vect_slp_bb): Likewise.
194         * doc/tm.texi: Regenerate.
195         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
196         to a poly_uint64.
197         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
198         the vector size as a poly_uint64 rather than an unsigned int.
199         (current_vector_size): Change from an unsigned int to a poly_uint64.
200         (get_vectype_for_scalar_type): Update accordingly.
201         * tree.h (build_truth_vector_type): Take the size and number of
202         units as a poly_uint64 rather than an unsigned int.
203         (build_vector_type): Add a temporary overload that takes
204         the number of units as a poly_uint64 rather than an unsigned int.
205         * tree.c (make_vector_type): Likewise.
206         (build_truth_vector_type): Take the number of units as a poly_uint64
207         rather than an unsigned int.
209 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
210             Alan Hayward  <alan.hayward@arm.com>
211             David Sherwood  <david.sherwood@arm.com>
213         * target.def (get_mask_mode): Take the number of units and length
214         as poly_uint64s rather than unsigned ints.
215         * targhooks.h (default_get_mask_mode): Update accordingly.
216         * targhooks.c (default_get_mask_mode): Likewise.
217         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
218         * doc/tm.texi: Regenerate.
220 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
221             Alan Hayward  <alan.hayward@arm.com>
222             David Sherwood  <david.sherwood@arm.com>
224         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
225         * omp-general.c (omp_max_vf): Likewise.
226         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
227         (expand_omp_simd): Handle polynomial safelen.
228         * omp-low.c (omplow_simd_context): Add a default constructor.
229         (omplow_simd_context::max_vf): Change from int to poly_uint64.
230         (lower_rec_simd_input_clauses): Update accordingly.
231         (lower_rec_input_clauses): Likewise.
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         * tree-vectorizer.h (vect_nunits_for_cost): New function.
238         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
239         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
240         (vect_analyze_slp_cost): Likewise.
241         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
242         (vect_model_load_cost): Likewise.
244 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
245             Alan Hayward  <alan.hayward@arm.com>
246             David Sherwood  <david.sherwood@arm.com>
248         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
249         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
250         from an unsigned int * to a poly_uint64_pod *.
251         (calculate_unrolling_factor): New function.
252         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
254 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
255             Alan Hayward  <alan.hayward@arm.com>
256             David Sherwood  <david.sherwood@arm.com>
258         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
259         from an unsigned int to a poly_uint64.
260         (_loop_vec_info::slp_unrolling_factor): Likewise.
261         (_loop_vec_info::vectorization_factor): Change from an int
262         to a poly_uint64.
263         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
264         (vect_get_num_vectors): New function.
265         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
266         (vect_get_num_copies): Use vect_get_num_vectors.
267         (vect_analyze_data_ref_dependences): Change max_vf from an int *
268         to an unsigned int *.
269         (vect_analyze_data_refs): Change min_vf from an int * to a
270         poly_uint64 *.
271         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
272         than an unsigned HOST_WIDE_INT.
273         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
274         (vect_analyze_data_ref_dependence): Change max_vf from an int *
275         to an unsigned int *.
276         (vect_analyze_data_ref_dependences): Likewise.
277         (vect_compute_data_ref_alignment): Handle polynomial vf.
278         (vect_enhance_data_refs_alignment): Likewise.
279         (vect_prune_runtime_alias_test_list): Likewise.
280         (vect_shift_permute_load_chain): Likewise.
281         (vect_supportable_dr_alignment): Likewise.
282         (dependence_distance_ge_vf): Take the vectorization factor as a
283         poly_uint64 rather than an unsigned HOST_WIDE_INT.
284         (vect_analyze_data_refs): Change min_vf from an int * to a
285         poly_uint64 *.
286         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
287         vfm1 as a poly_uint64 rather than an int.  Make the same change
288         for the returned bound_scalar.
289         (vect_gen_vector_loop_niters): Handle polynomial vf.
290         (vect_do_peeling): Likewise.  Update call to
291         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
292         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
293         be constant.
294         * tree-vect-loop.c (vect_determine_vectorization_factor)
295         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
296         (vect_get_known_peeling_cost): Likewise.
297         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
298         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
299         (vect_transform_loop): Likewise.  Use the lowest possible VF when
300         updating the upper bounds of the loop.
301         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
302         rather than an int.
303         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
304         polynomial unroll factors.
305         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
306         (vect_make_slp_decision): Likewise.
307         (vect_supported_load_permutation_p): Likewise, and polynomial
308         vf too.
309         (vect_analyze_slp_cost): Handle polynomial vf.
310         (vect_slp_analyze_node_operations): Likewise.
311         (vect_slp_analyze_bb_1): Likewise.
312         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
313         than an unsigned HOST_WIDE_INT.
314         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
315         (vectorizable_load): Handle polynomial vf.
316         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
317         a poly_uint64.
318         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
320 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
321             Alan Hayward  <alan.hayward@arm.com>
322             David Sherwood  <david.sherwood@arm.com>
324         * match.pd: Handle bit operations involving three constants
325         and try to fold one pair.
327 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
329         * tree-vect-loop-manip.c: Include gimple-fold.h.
330         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
331         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
332         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
333         Add a path that uses a step of VF instead of 1, but disable it
334         for now.
335         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
336         and niters_no_overflow parameters.  Update calls to
337         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
338         Create a new SSA name if the latter choses to use a ste other
339         than zero, and return it via niters_vector_mult_vf_var.
340         * tree-vect-loop.c (vect_transform_loop): Update calls to
341         vect_do_peeling, vect_gen_vector_loop_niters and
342         slpeel_make_loop_iterate_ntimes.
343         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
344         (vect_gen_vector_loop_niters): Update declarations after above changes.
346 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
348         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
349         128-bit round to integer instructions.
350         (ceil<mode>2): Likewise.
351         (btrunc<mode>2): Likewise.
352         (round<mode>2): Likewise.
354 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
356         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
357         unaligned VSX load/store on P8/P9.
358         (expand_block_clear): Allow the use of unaligned VSX
359         load/store on P8/P9.
361 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
363         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
364         New function.
365         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
366         swap associated with both a load and a store.
368 2018-01-02  Andrew Waterman  <andrew@sifive.com>
370         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
371         * config/riscv/riscv.md (clear_cache): Use it.
373 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
375         * web.c: Remove out-of-date comment.
377 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
379         * expr.c (fixup_args_size_notes): Check that any existing
380         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
381         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
382         (emit_single_push_insn): ...here.
384 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
386         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
387         (const_vector_encoded_nelts): New function.
388         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
389         (const_vector_int_elt, const_vector_elt): Declare.
390         * emit-rtl.c (const_vector_int_elt_1): New function.
391         (const_vector_elt): Likewise.
392         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
393         of CONST_VECTOR_ELT.
395 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
397         * expr.c: Include rtx-vector-builder.h.
398         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
399         directly on the tree encoding.
400         (const_vector_from_tree): Likewise.
401         * optabs.c: Include rtx-vector-builder.h.
402         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
403         sequence of "u" values.
404         * vec-perm-indices.c: Include rtx-vector-builder.h.
405         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
406         directly on the vec_perm_indices encoding.
408 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
410         * doc/rtl.texi (const_vector): Describe new encoding scheme.
411         * Makefile.in (OBJS): Add rtx-vector-builder.o.
412         * rtx-vector-builder.h: New file.
413         * rtx-vector-builder.c: Likewise.
414         * rtl.h (rtx_def::u2): Add a const_vector field.
415         (CONST_VECTOR_NPATTERNS): New macro.
416         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
417         (CONST_VECTOR_DUPLICATE_P): Likewise.
418         (CONST_VECTOR_STEPPED_P): Likewise.
419         (CONST_VECTOR_ENCODED_ELT): Likewise.
420         (const_vec_duplicate_p): Check for a duplicated vector encoding.
421         (unwrap_const_vec_duplicate): Likewise.
422         (const_vec_series_p): Check for a non-duplicated vector encoding.
423         Say that the function only returns true for integer vectors.
424         * emit-rtl.c: Include rtx-vector-builder.h.
425         (gen_const_vec_duplicate_1): Delete.
426         (gen_const_vector): Call gen_const_vec_duplicate instead of
427         gen_const_vec_duplicate_1.
428         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
429         (gen_const_vec_duplicate): Use rtx_vector_builder.
430         (gen_const_vec_series): Likewise.
431         (gen_rtx_CONST_VECTOR): Likewise.
432         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
433         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
434         Build a new vector rather than modifying a CONST_VECTOR in-place.
435         (handle_special_swappables): Update call accordingly.
436         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
437         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
438         Build a new vector rather than modifying a CONST_VECTOR in-place.
439         (handle_special_swappables): Update call accordingly.
441 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
443         * simplify-rtx.c (simplify_const_binary_operation): Use
444         CONST_VECTOR_ELT instead of XVECEXP.
446 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
448         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
449         the selector elements to be different from the data elements
450         if the selector is a VECTOR_CST.
451         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
452         ssizetype for the selector.
454 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
456         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
457         before testing each element individually.
458         * tree-vect-generic.c (lower_vec_perm): Likewise.
460 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
462         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
463         * selftest-run-tests.c (selftest::run_tests): Call it.
464         * vector-builder.h (vector_builder::operator ==): New function.
465         (vector_builder::operator !=): Likewise.
466         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
467         (vec_perm_indices::all_from_input_p): New function.
468         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
469         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
470         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
471         instead of reading the VECTOR_CST directly.  Detect whether both
472         vector inputs are the same before constructing the vec_perm_indices,
473         and update the number of inputs argument accordingly.  Use the
474         utility functions added above.  Only construct sel2 if we need to.
476 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
478         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
479         the broadcast of the low byte.
480         (expand_mult_highpart): Use an explicit encoding for the permutes.
481         * optabs-query.c (can_mult_highpart_p): Likewise.
482         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
483         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
484         (vectorizable_bswap): Likewise.
485         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
486         explicit encoding for the power-of-2 permutes.
487         (vect_permute_store_chain): Likewise.
488         (vect_grouped_load_supported): Likewise.
489         (vect_permute_load_chain): Likewise.
491 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
493         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
494         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
495         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
496         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
497         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
498         (vect_gen_perm_mask_any): Likewise.
500 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
502         * int-vector-builder.h: New file.
503         * vec-perm-indices.h: Include int-vector-builder.h.
504         (vec_perm_indices): Redefine as an int_vector_builder.
505         (auto_vec_perm_indices): Delete.
506         (vec_perm_builder): Redefine as a stand-alone class.
507         (vec_perm_indices::vec_perm_indices): New function.
508         (vec_perm_indices::clamp): Likewise.
509         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
510         (vec_perm_indices::new_vector): New function.
511         (vec_perm_indices::new_expanded_vector): Update for new
512         vec_perm_indices class.
513         (vec_perm_indices::rotate_inputs): New function.
514         (vec_perm_indices::all_in_range_p): Operate directly on the
515         encoded form, without computing elided elements.
516         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
517         encoding.  Update for new vec_perm_indices class.
518         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
519         the given vec_perm_builder.
520         (expand_vec_perm_var): Update vec_perm_builder constructor.
521         (expand_mult_highpart): Use vec_perm_builder instead of
522         auto_vec_perm_indices.
523         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
524         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
525         or double series encoding as appropriate.
526         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
527         vec_perm_indices instead of auto_vec_perm_indices.
528         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
529         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
530         (vect_permute_store_chain): Likewise.
531         (vect_grouped_load_supported): Likewise.
532         (vect_permute_load_chain): Likewise.
533         (vect_shift_permute_load_chain): Likewise.
534         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
535         (vect_transform_slp_perm_load): Likewise.
536         (vect_schedule_slp_instance): Likewise.
537         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
538         (vectorizable_mask_load_store): Likewise.
539         (vectorizable_bswap): Likewise.
540         (vectorizable_store): Likewise.
541         (vectorizable_load): Likewise.
542         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
543         vec_perm_indices instead of auto_vec_perm_indices.  Use
544         tree_to_vec_perm_builder to read the vector from a tree.
545         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
546         vec_perm_builder instead of a vec_perm_indices.
547         (have_whole_vector_shift): Use vec_perm_builder and
548         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
549         truncation to calc_vec_perm_mask_for_shift.
550         (vect_create_epilog_for_reduction): Likewise.
551         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
552         from auto_vec_perm_indices to vec_perm_indices.
553         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
554         instead of changing individual elements.
555         (aarch64_vectorize_vec_perm_const): Use new_vector to install
556         the vector in d.perm.
557         * config/arm/arm.c (expand_vec_perm_d::perm): Change
558         from auto_vec_perm_indices to vec_perm_indices.
559         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
560         instead of changing individual elements.
561         (arm_vectorize_vec_perm_const): Use new_vector to install
562         the vector in d.perm.
563         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
564         Update vec_perm_builder constructor.
565         (rs6000_expand_interleave): Likewise.
566         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
567         (rs6000_expand_interleave): Likewise.
569 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
571         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
572         to qimode could truncate the indices.
573         * optabs.c (expand_vec_perm_var): Likewise.
575 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
577         * Makefile.in (OBJS): Add vec-perm-indices.o.
578         * vec-perm-indices.h: New file.
579         * vec-perm-indices.c: Likewise.
580         * target.h (vec_perm_indices): Replace with a forward class
581         declaration.
582         (auto_vec_perm_indices): Move to vec-perm-indices.h.
583         * optabs.h: Include vec-perm-indices.h.
584         (expand_vec_perm): Delete.
585         (selector_fits_mode_p, expand_vec_perm_var): Declare.
586         (expand_vec_perm_const): Declare.
587         * target.def (vec_perm_const_ok): Replace with...
588         (vec_perm_const): ...this new hook.
589         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
590         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
591         * doc/tm.texi: Regenerate.
592         * optabs.def (vec_perm_const): Delete.
593         * doc/md.texi (vec_perm_const): Likewise.
594         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
595         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
596         expand_vec_perm for constant permutation vectors.  Assert that
597         the mode of variable permutation vectors is the integer equivalent
598         of the mode that is being permuted.
599         * optabs-query.h (selector_fits_mode_p): Declare.
600         * optabs-query.c: Include vec-perm-indices.h.
601         (selector_fits_mode_p): New function.
602         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
603         is defined, instead of checking whether the vec_perm_const_optab
604         exists.  Use targetm.vectorize.vec_perm_const instead of
605         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
606         fit in the vector mode before using a variable permute.
607         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
608         vec_perm_indices instead of an rtx.
609         (expand_vec_perm): Replace with...
610         (expand_vec_perm_const): ...this new function.  Take the selector
611         as a vec_perm_indices rather than an rtx.  Also take the mode of
612         the selector.  Update call to shift_amt_for_vec_perm_mask.
613         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
614         Use vec_perm_indices::new_expanded_vector to expand the original
615         selector into bytes.  Check whether the indices fit in the vector
616         mode before using a variable permute.
617         (expand_vec_perm_var): Make global.
618         (expand_mult_highpart): Use expand_vec_perm_const.
619         * fold-const.c: Includes vec-perm-indices.h.
620         * tree-ssa-forwprop.c: Likewise.
621         * tree-vect-data-refs.c: Likewise.
622         * tree-vect-generic.c: Likewise.
623         * tree-vect-loop.c: Likewise.
624         * tree-vect-slp.c: Likewise.
625         * tree-vect-stmts.c: Likewise.
626         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
627         Delete.
628         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
629         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
630         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
631         (aarch64_vectorize_vec_perm_const): ...this new function.
632         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
633         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
634         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
635         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
636         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
637         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
638         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
639         into...
640         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
641         check for NEON modes.
642         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
643         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
644         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
645         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
646         into...
647         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
648         the old VEC_PERM_CONST conditions.
649         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
650         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
651         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
652         (ia64_vectorize_vec_perm_const_ok): Merge into...
653         (ia64_vectorize_vec_perm_const): ...this new function.
654         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
655         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
656         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
657         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
658         * config/mips/mips.c (mips_expand_vec_perm_const)
659         (mips_vectorize_vec_perm_const_ok): Merge into...
660         (mips_vectorize_vec_perm_const): ...this new function.
661         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
662         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
663         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
664         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
665         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
666         (rs6000_expand_vec_perm_const): Delete.
667         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
668         Delete.
669         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
670         (altivec_expand_vec_perm_const_le): Take each operand individually.
671         Operate on constant selectors rather than rtxes.
672         (altivec_expand_vec_perm_const): Likewise.  Update call to
673         altivec_expand_vec_perm_const_le.
674         (rs6000_expand_vec_perm_const): Delete.
675         (rs6000_vectorize_vec_perm_const_ok): Delete.
676         (rs6000_vectorize_vec_perm_const): New function.
677         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
678         an element count and rtx array.
679         (rs6000_expand_extract_even): Update call accordingly.
680         (rs6000_expand_interleave): Likewise.
681         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
682         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
683         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
684         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
685         (rs6000_expand_vec_perm_const): Delete.
686         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
687         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
688         (altivec_expand_vec_perm_const_le): Take each operand individually.
689         Operate on constant selectors rather than rtxes.
690         (altivec_expand_vec_perm_const): Likewise.  Update call to
691         altivec_expand_vec_perm_const_le.
692         (rs6000_expand_vec_perm_const): Delete.
693         (rs6000_vectorize_vec_perm_const_ok): Delete.
694         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
695         reference to the SPE evmerge intructions.
696         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
697         an element count and rtx array.
698         (rs6000_expand_extract_even): Update call accordingly.
699         (rs6000_expand_interleave): Likewise.
700         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
701         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
702         new function.
703         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
705 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
707         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
708         vector mode and that that mode matches the mode of the data
709         being permuted.
710         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
711         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
712         directly using expand_vec_perm_1 when forcing selectors into
713         registers.
714         (expand_vec_perm_var): New function, split out from expand_vec_perm.
716 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
718         * optabs-query.h (can_vec_perm_p): Delete.
719         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
720         * optabs-query.c (can_vec_perm_p): Split into...
721         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
722         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
723         particular selector is valid.
724         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
725         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
726         (vect_grouped_load_supported): Likewise.
727         (vect_shift_permute_load_chain): Likewise.
728         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
729         (vect_transform_slp_perm_load): Likewise.
730         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
731         (vectorizable_bswap): Likewise.
732         (vect_gen_perm_mask_checked): Likewise.
733         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
734         implementations of variable permutation vectors into account
735         when deciding which selector to use.
736         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
737         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
738         with a false third argument.
739         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
740         to test whether the constant selector is valid and can_vec_perm_var_p
741         to test whether a variable selector is valid.
743 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
745         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
746         * optabs-query.c (can_vec_perm_p): Likewise.
747         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
748         instead of vec_perm_indices.
749         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
750         (vect_gen_perm_mask_checked): Likewise,
751         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
752         (vect_gen_perm_mask_checked): Likewise,
754 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
756         * optabs-query.h (qimode_for_vec_perm): Declare.
757         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
758         (qimode_for_vec_perm): ...this new function.
759         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
761 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
763         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
764         does not have a conditional at the top.
766 2018-01-02  Richard Biener  <rguenther@suse.de>
768         * ipa-inline.c (big_speedup_p): Fix expression.
770 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
772         PR target/81616
773         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
774         for generic 4->6.
776 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
778         PR target/81616
779         Generic tuning.
780         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
781         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
782         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
783         cond_taken_branch_cost 3->4.
785 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
787         PR tree-optimization/83581
788         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
789         TODO_cleanup_cfg if any changes have been made.
791         PR middle-end/83608
792         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
793         convert_modes if target mode has the right side, but different mode
794         class.
796         PR middle-end/83609
797         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
798         last argument when extracting from CONCAT.  If either from_real or
799         from_imag is NULL, use expansion through memory.  If result is not
800         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
801         the parts directly to inner mode, if even that fails, use expansion
802         through memory.
804         PR middle-end/83623
805         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
806         check for bswap in mode rather than HImode and use that in expand_unop
807         too.
809 2017-12-31  Jakub Jelinek  <jakub@redhat.com>
811         PR target/83536
812         * config/i386/i386.c (ix86_attribute_table): Remove excess
813         initializer for "shared" attribute.
815 2017-12-30  Tom de Vries  <tom@codesourcery.com>
817         PR libgomp/83046
818         * omp-expand.c (expand_omp_target): If in_lto_p, mark offload_funcs with
819         DECL_PRESERVE_P.
820         * lto-streamer-out.c (prune_offload_funcs): New function.  Remove
821         offload_funcs entries that no longer have a corresponding cgraph_node.
822         Mark the remaining ones as DECL_PRESERVE_P.
823         (output_lto): Call prune_offload_funcs.
825 2017-12-30  Jakub Jelinek  <jakub@redhat.com>
827         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
828         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>,
829         vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
830         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
831         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
832         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1, vpdpbusd_<mode>,
833         vpdpbusd_<mode>_mask, vpdpbusd_<mode>_maskz, vpdpbusd_<mode>_maskz_1,
834         vpdpbusds_<mode>, vpdpbusds_<mode>_mask, vpdpbusds_<mode>_maskz,
835         vpdpbusds_<mode>_maskz_1, vpdpwssd_<mode>, vpdpwssd_<mode>_mask,
836         vpdpwssd_<mode>_maskz, vpdpwssd_<mode>_maskz_1, vpdpwssds_<mode>,
837         vpdpwssds_<mode>_mask, vpdpwssds_<mode>_maskz,
838         vpdpwssds_<mode>_maskz_1, vaesdec_<mode>, vaesdeclast_<mode>,
839         vaesenc_<mode>, vpclmulqdq_<mode>,
840         avx512vl_vpshufbitqmb<mode><mask_scalar_merge_name>): Formatting fixes.
842 2017-12-28  Michael Meissner  <meissner@linux.vnet.ibm.com>
844         * builtins.def: (_Float<N> and _Float<N>X BUILT_IN_CEIL): Add
845         _Float<N> and _Float<N>X variants for rounding built-in
846         functions.
847         (_Float<N> and _Float<N>X BUILT_IN_FLOOR): Likewise.
848         (_Float<N> and _Float<N>X BUILT_IN_NEARBYINT): Likewise.
849         (_Float<N> and _Float<N>X BUILT_IN_RINT): Likewise.
850         (_Float<N> and _Float<N>X BUILT_IN_ROUND): Likewise.
851         (_Float<N> and _Float<N>X BUILT_IN_TRUNC): Likewise.
852         * builtins.c (mathfn_built_in_2): Likewise.
853         * internal-fn.def (CEIL): Likewise.
854         (FLOOR): Likewise.
855         (NEARBYINT): Likewise.
856         (RINT): Likewise.
857         (ROUND): Likewise.
858         (TRUNC): Likewise.
859         * convert.c (convert_to_integer_1): Likewise.
860         * fold-const.c (tree_call_nonnegative_warnv_p): Likewise.
861         (integer_valued_real_call_p): Likewise.
862         * fold-const-call.c (fold_const_call_ss): Likewise.
863         * gencfn-macros.c (print_case_cfn): Change CFN and operator
864         printers to take a const char * suffix instead of a bool.
865         (print_define_operator_list): Likewise.
866         (fltall_suffixes): New list of suffixes, that include the
867         traditional suffixes as well as all of the _Float<N> and
868         _Float<N>X suffixes.
869         (main): For _Float<N> and _Float<N>X functions, emit both
870         <name>_FN and <name>_ALL variants.  The <macro>_FN variant only
871         has the _Float<N> and _Float<N>X case names or operators.  The
872         <name>_ALL variant has both the traditional and the
873         _Float<N>/_Float<N>X case names or operators.
874         * match.pd (COPYSIGN optimizations): Provide optimizations for
875         _Float<N> and _Float<N>X types where possible.
876         (MIN/MAX optimizations): Likewise.
877         (sqrt optimizations): Likewise.
878         (rounding optimizations): Likewise.
880 2017-12-28  Richard Sandiford  <richard.sandiford@linaro.org>
882         * config/powerpcspe/powerpcspe.md (ieee_128bit_negative_zero): Use
883         gen_int_mode rather than GEN_INT.
884         * config/rs6000/rs6000.md (ieee_128bit_negative_zero): Likewise.
886 2017-12-28  Richard Sandiford  <richard.sandiford@linaro.org>
888         * emit-rtl.c (gen_const_vec_series): Use valid_for_const_vector_p
889         instead of CONSTANT_P.
890         (gen_vec_series): Likewise.
891         * simplify-rtx.c (simplify_binary_operation_1): Likewise.
893 2017-12-28  Andreas Schwab  <schwab@linux-m68k.org>
895         * config/m68k/m68k.md (ashrdi3_const1, lshrdi3_const1): Add
896         CC_STATUS_INIT.
898 2017-12-27  Kugan Vivekanandarajah  <kuganv@linaro.org>
900         * config/aarch64/aarch64-simd.md (aarch64_ld1x2<VQ:mode>): New.
901         (aarch64_ld1x2<VDC:mode>): Likewise.
902         (aarch64_simd_ld1<mode>_x2): Likewise.
903         (aarch64_simd_ld1<mode>_x2): Likewise.
904         * config/aarch64/arm_neon.h (vld1_u8_x2): New.
905         (vld1_s8_x2): Likewise.
906         (vld1_u16_x2): Likewise.
907         (vld1_s16_x2): Likewise.
908         (vld1_u32_x2): Likewise.
909         (vld1_s32_x2): Likewise.
910         (vld1_u64_x2): Likewise.
911         (vld1_s64_x2): Likewise.
912         (vld1_f16_x2): Likewise.
913         (vld1_f32_x2): Likewise.
914         (vld1_f64_x2): Likewise.
915         (vld1_p8_x2): Likewise.
916         (vld1_p16_x2): Likewise.
917         (vld1_p64_x2): Likewise.
918         (vld1q_u8_x2): Likewise.
919         (vld1q_s8_x2): Likewise.
920         (vld1q_u16_x2): Likewise.
921         (vld1q_s16_x2): Likewise.
922         (vld1q_u32_x2): Likewise.
923         (vld1q_s32_x2): Likewise.
924         (vld1q_u64_x2): Likewise.
925         (vld1q_s64_x2): Likewise.
926         (vld1q_f16_x2): Likewise.
927         (vld1q_f32_x2): Likewise.
928         (vld1q_f64_x2): Likewise.
929         (vld1q_p8_x2): Likewise.
930         (vld1q_p16_x2): Likewise.
931         (vld1q_p64_x2): Likewise.
933 2017-12-27  Martin Liska  <mliska@suse.cz>
935         PR tree-optimization/83552
936         * tree-ssa-strlen.c (fold_strstr_to_strncmp): Assign result
937         of get_string_lenth to a SSA_NAME if not a GIMPLE value.
939 2017-12-27  Tom de Vries  <tom@codesourcery.com>
941         * config/nvptx/nvptx.c (nvptx_option_override): Disable
942         -gstatement-frontiers.
944 2017-12-26  Alexander Monakov  <amonakov@ispras.ru>
946         PR rtl-optimization/83513
947         * sel-sched.c (sel_rank_for_schedule): Order by non-zero usefulness
948         before priority comparison.
950 2017-12-25  Jakub Jelinek  <jakub@redhat.com>
952         PR target/83488
953         * config/i386/i386.opt (-mavx512vpopcntdq, -mmavx512bitalg): Move from
954         ix86_isa_flags2 to ix86_isa_flags.
955         * config/i386/i386-c.c (ix86_target_macros_internal): Test
956         OPTION_MASK_ISA_AVX512BITALG and OPTION_MASK_ISA_AVX512VPOPCNTDQ in
957         isa_flags rather than isa_flags2.
958         * config/i386/i386.c (ix86_target_string): Move -mavx512vpopcntdq
959         and -mavx512bitalg from isa2_opts to isa_opts.
960         (ix86_option_override_internal): Test OPTION_MASK_ISA_AVX512VPOPCNTDQ
961         in x_ix86_isa_flags_explicit rather than x_ix86_isa_flags2_explicit
962         and set it in x_ix86_isa_flags rather than x_ix86_isa_flags2.
963         Formatting fixes.
964         (def_builtin): Treat OPTION_MASK_ISA_AVX512BW or
965         OPTION_MASK_ISA_AVX512F ored with another option similarly to
966         OPTION_MASK_ISA_AVX512VL.  Even for OPTION_MASK_ISA_AVX512VL don't
967         clear it if mask is just OPTION_MASK_ISA_AVX512VL itself.
968         (ix86_expand_builtin): Don't handle OPTION_MASK_ISA_GFNI and
969         OPTION_MASK_ISA_VPCLMULQDQ specially, instead handle
970         OPTION_MASK_ISA_AVX512BW and OPTION_MASK_ISA_AVX512F that way.
971         * config/i386/i386-builtin.def: Move AVX512VPOPCNTDQ and AVX512BITALG
972         builtins from bdesc_args2 to bdesc_args section.
973         (__builtin_ia32_compressstoreuqi512_mask,
974         __builtin_ia32_compressstoreuhi512_mask,
975         __builtin_ia32_compressstoreuqi256_mask,
976         __builtin_ia32_expandloadqi512_mask,
977         __builtin_ia32_expandloadqi512_maskz,
978         __builtin_ia32_expandloadhi512_mask,
979         __builtin_ia32_expandloadhi512_maskz,
980         __builtin_ia32_compressqi512_mask, __builtin_ia32_compresshi512_mask,
981         __builtin_ia32_compressqi256_mask, __builtin_ia32_expandqi512_mask,
982         __builtin_ia32_expandqi512_maskz, __builtin_ia32_expandhi512_mask,
983         __builtin_ia32_expandhi512_maskz, __builtin_ia32_expandqi256_mask,
984         __builtin_ia32_expandqi256_maskz, __builtin_ia32_vpshrd_v32hi_mask,
985         __builtin_ia32_vpshld_v32hi_mask, __builtin_ia32_vpshrdv_v32hi_mask,
986         __builtin_ia32_vpshrdv_v32hi_maskz, __builtin_ia32_vpshldv_v32hi_mask,
987         __builtin_ia32_vpshldv_v32hi_maskz,
988         __builtin_ia32_vpopcountb_v64qi_mask,
989         __builtin_ia32_vpopcountw_v32hi_mask,
990         __builtin_ia32_vpshufbitqmb512_mask,
991         __builtin_ia32_vpshufbitqmb256_mask): Add
992         " | OPTION_MASK_ISA_AVX512BW".
993         (__builtin_ia32_expandloadqi256_mask,
994         __builtin_ia32_expandloadqi256_maskz,
995         __builtin_ia32_vpopcountb_v32qi_mask): Add
996         " | OPTION_MASK_ISA_AVX512VL | OPTION_MASK_ISA_AVX512BW".
997         (__builtin_ia32_expandloadhi256_mask,
998         __builtin_ia32_expandloadhi256_maskz,
999         __builtin_ia32_expandloadqi128_mask,
1000         __builtin_ia32_expandloadqi128_maskz,
1001         __builtin_ia32_expandloadhi128_mask,
1002         __builtin_ia32_expandloadhi128_maskz,
1003         __builtin_ia32_vpshrd_v16hi, __builtin_ia32_vpshrd_v16hi_mask,
1004         __builtin_ia32_vpshrd_v8hi, __builtin_ia32_vpshrd_v8hi_mask,
1005         __builtin_ia32_vpshrd_v8si, __builtin_ia32_vpshrd_v8si_mask,
1006         __builtin_ia32_vpshrd_v4si, __builtin_ia32_vpshrd_v4si_mask,
1007         __builtin_ia32_vpshrd_v4di, __builtin_ia32_vpshrd_v4di_mask,
1008         __builtin_ia32_vpshrd_v2di, __builtin_ia32_vpshrd_v2di_mask,
1009         __builtin_ia32_vpshld_v16hi, __builtin_ia32_vpshld_v16hi_mask,
1010         __builtin_ia32_vpshld_v8hi, __builtin_ia32_vpshld_v8hi_mask,
1011         __builtin_ia32_vpshld_v8si, __builtin_ia32_vpshld_v8si_mask,
1012         __builtin_ia32_vpshld_v4si, __builtin_ia32_vpshld_v4si_mask,
1013         __builtin_ia32_vpshld_v4di, __builtin_ia32_vpshld_v4di_mask,
1014         __builtin_ia32_vpshld_v2di, __builtin_ia32_vpshld_v2di_mask,
1015         __builtin_ia32_vpshrdv_v16hi, __builtin_ia32_vpshrdv_v16hi_mask,
1016         __builtin_ia32_vpshrdv_v16hi_maskz, __builtin_ia32_vpshrdv_v8hi,
1017         __builtin_ia32_vpshrdv_v8hi_mask, __builtin_ia32_vpshrdv_v8hi_maskz,
1018         __builtin_ia32_vpshrdv_v8si, __builtin_ia32_vpshrdv_v8si_mask,
1019         __builtin_ia32_vpshrdv_v8si_maskz, __builtin_ia32_vpshrdv_v4si,
1020         __builtin_ia32_vpshrdv_v4si_mask, __builtin_ia32_vpshrdv_v4si_maskz,
1021         __builtin_ia32_vpshrdv_v4di, __builtin_ia32_vpshrdv_v4di_mask,
1022         __builtin_ia32_vpshrdv_v4di_maskz, __builtin_ia32_vpshrdv_v2di,
1023         __builtin_ia32_vpshrdv_v2di_mask, __builtin_ia32_vpshrdv_v2di_maskz,
1024         __builtin_ia32_vpshldv_v16hi, __builtin_ia32_vpshldv_v16hi_mask,
1025         __builtin_ia32_vpshldv_v16hi_maskz, __builtin_ia32_vpshldv_v8hi,
1026         __builtin_ia32_vpshldv_v8hi_mask, __builtin_ia32_vpshldv_v8hi_maskz,
1027         __builtin_ia32_vpshldv_v8si, __builtin_ia32_vpshldv_v8si_mask,
1028         __builtin_ia32_vpshldv_v8si_maskz, __builtin_ia32_vpshldv_v4si,
1029         __builtin_ia32_vpshldv_v4si_mask, __builtin_ia32_vpshldv_v4si_maskz,
1030         __builtin_ia32_vpshldv_v4di, __builtin_ia32_vpshldv_v4di_mask,
1031         __builtin_ia32_vpshldv_v4di_maskz, __builtin_ia32_vpshldv_v2di,
1032         __builtin_ia32_vpshldv_v2di_mask, __builtin_ia32_vpshldv_v2di_maskz,
1033         __builtin_ia32_vpopcountb_v32qi, __builtin_ia32_vpopcountb_v16qi,
1034         __builtin_ia32_vpopcountb_v16qi_mask, __builtin_ia32_vpopcountw_v16hi,
1035         __builtin_ia32_vpopcountw_v16hi_mask, __builtin_ia32_vpopcountw_v8hi,
1036         __builtin_ia32_vpopcountw_v8hi_mask): Add
1037         " | OPTION_MASK_ISA_AVX512VL".
1038         * config/i386/avx512vbmi2intrin.h (_mm512_shrdi_epi16,
1039         _mm512_shrdi_epi32, _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
1040         _mm512_shrdi_epi64, _mm512_mask_shrdi_epi64, _mm512_maskz_shrdi_epi64,
1041         _mm512_shldi_epi16, _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
1042         _mm512_maskz_shldi_epi32, _mm512_shldi_epi64, _mm512_mask_shldi_epi64,
1043         _mm512_maskz_shldi_epi64, _mm512_shrdv_epi16, _mm512_shrdv_epi32,
1044         _mm512_mask_shrdv_epi32, _mm512_maskz_shrdv_epi32, _mm512_shrdv_epi64,
1045         _mm512_mask_shrdv_epi64, _mm512_maskz_shrdv_epi64, _mm512_shldv_epi16,
1046         _mm512_shldv_epi32, _mm512_mask_shldv_epi32, _mm512_maskz_shldv_epi32,
1047         _mm512_shldv_epi64, _mm512_mask_shldv_epi64,
1048         _mm512_maskz_shldv_epi64): Don't require avx512bw for these intrinsics.
1049         * config/i386/avx512bitalgintrin.h (_mm_bitshuffle_epi64_mask,
1050         _mm_mask_bitshuffle_epi64_mask): Likewise.
1051         * common/config/i386/i386-common.c
1052         (OPTION_MASK_ISA_AVX512VPOPCNTDQ_SET,
1053         OPTION_MASK_ISA_AVX512BITALG_SET): Or in OPTION_MASK_ISA_AVX512F_SET.
1054         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
1055         OPTION_MASK_ISA_AVX512VPOPCNTDQ_UNSET and
1056         OPTION_MASK_ISA_AVX512BITALG_UNSET.
1057         (OPTION_MASK_ISA2_AVX512F_UNSET,
1058         OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Define.
1059         (ix86_handle_option): For -mno-general-regs-only, clear from
1060         ix86_isa_flags2 OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET rather than
1061         just OPTION_MASK_ISA_MPX.  For -mno-sse{,2,3,4,4.1,4.2,avx,avx2} and
1062         -mno-ssse3 clear OPTION_MASK_ISA2_AVX512F_UNSET bits from
1063         ix86_isa_flags2.  For -mno-avx512f likewise, instead of masking
1064         individually listed ISAs.  For -m{,no-}avx512{vpopcntdq,bitalg} adjust
1065         for moving from ix86_isa_flags2 to ix86_isa_flags.
1067 2017-12-23  Jakub Jelinek  <jakub@redhat.com>
1069         PR c++/83553
1070         * fold-const.c (struct contains_label_data): New type.
1071         (contains_label_1): Return non-NULL even for CASE_LABEL_EXPR, unless
1072         inside of a SWITCH_BODY seen during the walk.
1073         (contains_label_p): Use walk_tree instead of
1074         walk_tree_without_duplicates, prepare data for contains_label_1 and
1075         provide own pset.
1077 2017-12-22  Jakub Jelinek  <jakub@redhat.com>
1079         PR debug/83547
1080         * tree-iterator.c (alloc_stmt_list): Start with cleared
1081         TREE_SIDE_EFFECTS regardless whether a new STATEMENT_LIST is allocated
1082         or old one reused.
1084         PR target/83488
1085         * config/i386/avx512vnniintrin.h: Don't check for __AVX512F__ nor
1086         enable avx512f explicitly in #pragma GCC target.
1087         * config/i386/i386-builtin.def (__builtin_ia32_vpdpbusd_v8si,
1088         __builtin_ia32_vpdpbusd_v8si_mask, __builtin_ia32_vpdpbusd_v8si_maskz,
1089         __builtin_ia32_vpdpbusd_v4si, __builtin_ia32_vpdpbusd_v4si_mask,
1090         __builtin_ia32_vpdpbusd_v4si_maskz, __builtin_ia32_vpdpbusds_v8si,
1091         __builtin_ia32_vpdpbusds_v8si_mask,
1092         __builtin_ia32_vpdpbusds_v8si_maskz, __builtin_ia32_vpdpbusds_v4si,
1093         __builtin_ia32_vpdpbusds_v4si_mask,
1094         __builtin_ia32_vpdpbusds_v4si_maskz, __builtin_ia32_vpdpwssd_v8si,
1095         __builtin_ia32_vpdpwssd_v8si_mask, __builtin_ia32_vpdpwssd_v8si_maskz,
1096         __builtin_ia32_vpdpwssd_v4si, __builtin_ia32_vpdpwssd_v4si_mask,
1097         __builtin_ia32_vpdpwssd_v4si_maskz, __builtin_ia32_vpdpwssds_v8si,
1098         __builtin_ia32_vpdpwssds_v8si_mask,
1099         __builtin_ia32_vpdpwssds_v8si_maskz, __builtin_ia32_vpdpwssds_v4si,
1100         __builtin_ia32_vpdpwssds_v4si_mask,
1101         __builtin_ia32_vpdpwssds_v4si_maskz): Use
1102         OPTION_MASK_ISA_AVX512VNNI | OPTION_MASK_ISA_AVX512VL instead of
1103         just OPTION_MASK_ISA_AVX512VNNI.
1105 2017-12-22  Martin Jambor  <mjambor@suse.cz>
1107         PR lto/82027
1108         * lto-cgraph.c (output_cgraph_opt_summary_p): Also check former
1109         clones.
1111 2017-12-22  Julia Koval  <julia.koval@intel.com>
1112             Sebastian Peryt  <sebastian.peryt@intel.com>
1114         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512BITALG_SET,
1115         OPTION_MASK_ISA_AVX512BITALG_UNSET): New.
1116         (ix86_handle_option): Handle -mavx512bitalg, fix 4VNNIW formatting.
1117         * config.gcc: Add avx512vpopcntdqvlintrin.h and avx512bitalgintrin.h.
1118         * config/i386/avx512bitalgintrin.h (_mm512_popcnt_epi8, _mm512_popcnt_epi16,
1119         _mm512_mask_popcnt_epi8, _mm512_maskz_popcnt_epi8, _mm512_mask_popcnt_epi16,
1120         _mm512_maskz_popcnt_epi16, _mm512_bitshuffle_epi64_mask, _mm256_popcnt_epi8,
1121         _mm512_mask_bitshuffle_epi64_mask, _mm256_mask_popcnt_epi8, _mm_popcnt_epi8,
1122         _mm256_maskz_popcnt_epi8, _mm_bitshuffle_epi64_mask, _mm256_popcnt_epi16,
1123         _mm_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
1124         _mm256_mask_bitshuffle_epi64_mask, _mm_popcnt_epi16, _mm_maskz_popcnt_epi8,
1125         _mm256_mask_popcnt_epi16, _mm256_maskz_popcnt_epi16, _mm_mask_popcnt_epi8,
1126         _mm_mask_popcnt_epi16, _mm_maskz_popcnt_epi16): New intrinsics.
1127         * config/i386/avx512vpopcntdqvlintrin.h (_mm_popcnt_epi32, _mm_popcnt_epi64,
1128         _mm_mask_popcnt_epi32, _mm_maskz_popcnt_epi32, _mm256_popcnt_epi32,
1129         _mm256_mask_popcnt_epi32, _mm256_maskz_popcnt_epi32, _mm_mask_popcnt_epi64,
1130         _mm_maskz_popcnt_epi64, _mm256_popcnt_epi64, _mm256_mask_popcnt_epi64,
1131         _mm256_maskz_popcnt_epi64): New intrinsics.
1132         * config/i386/cpuid.h (bit_AVX512BITALG): New bit.
1133         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mavx512bitalg.
1134         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI, V64QI_FTYPE_V64QI,
1135         V4DI_FTYPE_V4DI, UHI_FTYPE_V2DI_V2DI_UHI, USI_FTYPE_V4DI_V4DI_USI,
1136         V4SI_FTYPE_V4SI_V4SI_UHI, V8SI_FTYPE_V8SI_V8SI_UHI): New types.
1137         * config/i386/i386-builtin.def (__builtin_ia32_vpopcountq_v4di,
1138         __builtin_ia32_vpopcountq_v4di_mask, __builtin_ia32_vpopcountq_v2di,
1139         __builtin_ia32_vpopcountq_v2di_mask, __builtin_ia32_vpopcountd_v4si,
1140         __builtin_ia32_vpopcountd_v4si_mask, __builtin_ia32_vpopcountd_v8si,
1141         __builtin_ia32_vpopcountd_v8si_mask, __builtin_ia32_vpopcountb_v64qi,
1142         __builtin_ia32_vpopcountb_v64qi_mask, __builtin_ia32_vpopcountb_v32qi,
1143         __builtin_ia32_vpopcountb_v32qi_mask, __builtin_ia32_vpopcountb_v16qi,
1144         __builtin_ia32_vpopcountb_v16qi_mask, __builtin_ia32_vpopcountw_v32hi,
1145         __builtin_ia32_vpopcountw_v32hi_mask, __builtin_ia32_vpopcountw_v16hi,
1146         __builtin_ia32_vpopcountw_v16hi_mask, __builtin_ia32_vpopcountw_v8hi,
1147         __builtin_ia32_vpopcountw_v8hi_mask, __builtin_ia32_vpshufbitqmb128_mask,
1148         __builtin_ia32_vpshufbitqmb256_mask,
1149         __builtin_ia32_vpshufbitqmb512_mask): New builtins.
1150         * config/i386/i386-c.c (__AVX512BITALG__): New.
1151         * config/i386/i386.c (isa2_opts): Add -mavx512bitalg.
1152         (ix86_valid_target_attribute_inner_p): Ditto.
1153         (ix86_expand_args_builtin): Handle new types.
1154         * config/i386/i386.h (TARGET_AVX512BITALG, TARGET_AVX512BITALG_P): New.
1155         * config/i386/i386.opt: Add -mavx512bitalg.
1156         * config/i386/immintrin.h: Add avx512vpopcntdqvlintrin.h and
1157         avx512bitalgintrin.h.
1158         * config/i386/sse.md (VI48_AVX512VLBW): New iterator.
1159         (vpopcount<mode><mask_name>): Add more types.
1160         (avx512vl_vpshufbitqmb<mode><mask_scalar_merge_name>): New.
1161         * doc/invoke.texi: Add -mavx512bitalg and -mavx512vpopcntdq.
1163 2017-12-22  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
1165         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VNNI_SET):
1166         Or in OPTION_MASK_ISA_AVX512F_SET.
1167         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
1168         OPTION_MASK_ISA_AVX512VNNI_UNSET.
1169         (ix86_handle_option): Adjust for
1170         OPTION_MASK_ISA_AVX512VNNI_*SET being in ix86_isa_flags.
1171         * config/i386/i386-builtin.def: Move VNNI builtins from ARGS2
1172         section to ARGS.
1173         * config/i386/i386-c.c: Check for OPTION_MASK_ISA_AVX512VNNI in
1174         isa_flag instead of isa_flag2.
1175         * config/i386/i386.c (ix86_target_string): Move -mavx512vnni from
1176         isa_opts2 to isa_opts.
1177         * config/i386/i386.opt (mavx512vnni): Move from ix86_isa_flags2
1178         to ix86_isa_flags.
1180 2017-12-22  Mike Stump  <mikestump@comcast.net>
1181             Eric Botcazou  <ebotcazou@adacore.com>
1183         * doc/extend.texi (Loop-Specific Pragmas): Document pragma GCC unroll.
1185 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1187         PR middle-end/83487
1188         * config/i386/i386.c (ix86_function_arg_boundary): Return
1189         PARM_BOUNDARY for TYPE_EMPTY_P types.
1191         PR c/83448
1192         * gimple-ssa-sprintf.c (maybe_warn): Don't call set_caret_index
1193         if navail is >= dir.len.
1195 2017-12-21  Steve Ellcey  <sellcey@cavium.com>
1197         * config/aarch64/t-aarch64-linux (MULTILIB_OSDIRNAMES): Fix
1198         triplet for ilp32.
1200 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1202         PR rtl-optimization/80747
1203         PR rtl-optimization/83512
1204         * cfgrtl.c (force_nonfallthru_and_redirect): When splitting
1205         succ edge from ENTRY, copy partition from e->dest to the newly
1206         created bb.
1207         * bb-reorder.c (reorder_basic_blocks_simple): If last_tail is
1208         ENTRY, use BB_PARTITION of its successor block as current_partition.
1209         Don't copy partition when splitting succ edge from ENTRY.
1211         PR tree-optimization/83523
1212         * tree-ssa-math-opts.c (is_widening_mult_p): Return false if
1213         for INTEGER_TYPE TYPE_OVERFLOW_TRAPS.
1214         (convert_mult_to_fma): Likewise.
1216         PR tree-optimization/83521
1217         * tree-ssa-phiopt.c (factor_out_conditional_conversion): Use
1218         gimple_build_assign without code on result of
1219         fold_build1 (VIEW_CONVERT_EXPR, ...), as it might not create
1220         a VIEW_CONVERT_EXPR.
1222 2017-12-21  Andrew Pinski  <apinski@cavium.com>
1223             Steve Ellcey  <sellcey@cavium.com>
1225         * config/aarch64/t-aarch64-linux (MULTILIB_OSDIRNAMES): Handle
1226         multi-arch for ilp32.
1228 2017-12-21  Uros Bizjak  <ubizjak@gmail.com>
1230         PR target/83467
1231         * config/i386/i386.md (*ashl<mode>3_mask): Add operand
1232         constraints to operand 2.
1233         (*ashl<mode>3_mask_1): Ditto.
1234         (*<shift_insn><mode>3_mask): Ditto.
1235         (*<shift_insn><mode>3_mask_1): Ditto.
1236         (*<rotate_insn><mode>3_mask): Ditto.
1237         (*<rotate_insn><mode>3_mask_1): Ditto.
1239 2017-12-21  Alexandre Oliva  <aoliva@redhat.com>
1241         * reorg.c (make_return_insns): Reemit each insn with its own location.
1243 2017-12-21  James Greenhalgh  <james.greenhalgh@arm.com>
1245         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
1246         generation for cases where splatting a value is not useful.
1247         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
1248         across a vec_duplicate and a paradoxical subreg forming a vector
1249         mode to a vec_concat.
1251 2017-12-21  James Greenhalgh  <james.greenhalgh@arm.com>
1253         * combine.c (simplify_set): Do not transform subregs to zero_extends
1254         if the destination is not a scalar int mode.
1256 2017-12-21  Eric Botcazou  <ebotcazou@adacore.com>
1258         PR c++/82872
1259         * convert.c (convert_to_integer_1) <POINTER_TYPE>: Do not return the
1260         shared zero if the input has overflowed.
1262 2017-12-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1264         * config/arm/driver-arm.c (arm_cpu_table): Specify dotprod
1265         support for Cortex-A55 and Cortex-A75.
1267 2017-12-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1269         * common/config/arm/arm-common.c (compare_opt_names): Add function
1270         comment.  Use strcmp instead of manual loop.
1272 2017-12-21  Martin Liska  <mliska@suse.cz>
1274         PR gcov-profile/83509
1275         * gcov-dump.c (dump_gcov_file): Do not read info about
1276         support_unexecuted_blocks for gcda files.
1278 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1280         PR rtl-optimization/82973
1281         * emit-rtl.h (valid_for_const_vec_duplicate_p): Rename to ...
1282         (valid_for_const_vector_p): ... this.
1283         * emit-rtl.c (valid_for_const_vec_duplicate_p): Rename to ...
1284         (valid_for_const_vector_p): ... this.  Adjust function comment.
1285         (gen_vec_duplicate): Adjust caller.
1286         * optabs.c (expand_vector_broadcast): Likewise.
1287         * simplify-rtx.c (simplify_const_unary_operation): Don't optimize into
1288         CONST_VECTOR if some element isn't simplified valid_for_const_vector_p
1289         constant.
1290         (simplify_const_binary_operation): Likewise.  Use CONST_FIXED_P macro
1291         instead of GET_CODE == CONST_FIXED.
1292         (simplify_subreg): Use CONST_FIXED_P macro instead of
1293         GET_CODE == CONST_FIXED.
1295         PR target/83488
1296         * config/i386/i386.c (ix86_target_string): Move -mavx512vbmi2 and
1297         -mshstk entries from isa_opts2 to isa_opts and -mhle, -mmovbe,
1298         -mclzero and -mmwaitx entries from isa_opts to isa_opts2.
1299         (ix86_option_override_internal): Adjust for
1300         OPTION_MASK_ISA_{HLE,MOVBE,CLZERO,MWAITX} moving to ix86_isa_flags2
1301         and OPTION_MASK_ISA_SHSTK moving to ix86_isa_flags.
1302         (BDESC_VERIFYS): Remove SPECIAL_ARGS2 related checks.
1303         (ix86_init_mmx_sse_builtins): Remove bdesc_special_args2 handling.
1304         Use def_builtin2 instead of def_builtin for OPTION_MASK_ISA_MWAITX
1305         and OPTION_MASK_ISA_CLZERO builtins.  Use def_builtin instead of
1306         def_builtin2 for CET builtins.
1307         (ix86_expand_builtin): Remove bdesc_special_args2 handling.  Fix
1308         up formatting in IX86_BUILTIN_RDPID code.
1309         * config/i386/i386-builtin.def: Move VBMI2 builtins from SPECIAL_ARGS2
1310         section to SPECIAL_ARGS and from ARGS2 section to ARGS.
1311         * config/i386/i386.opt (mavx512vbmi2, mshstk): Move from
1312         ix86_isa_flags2 to ix86_isa_flags.
1313         (mhle, mmovbe, mclzero, mmwaitx): Move from ix86_isa_flags to
1314         ix86_isa_flags2.
1315         * config/i386/i386-c.c (ix86_target_macros_internal): Check for
1316         OPTION_MASK_ISA_{CLZERO,MWAITX} in isa_flag2 instead of isa_flag.
1317         Check for OPTION_MASK_ISA_{SHSTK,AVX512VBMI2} in isa_flag instead
1318         of isa_flag2.
1319         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VBMI2_SET):
1320         Or in OPTION_MASK_ISA_AVX512F_SET.
1321         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
1322         OPTION_MASK_ISA_AVX512VBMI2_UNSET.
1323         (ix86_handle_option): Adjust for
1324         OPTION_MASK_ISA_{SHSTK,AVX512VBMI2}_*SET being in ix86_isa_flags
1325         and OPTION_MASK_ISA_{MOVBE,MWAITX,CLZERO}_*SET in ix86_isa_flags2.
1327 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1328             Alan Hayward  <alan.hayward@arm.com>
1329             David Sherwood  <david.sherwood@arm.com>
1331         * tree-data-ref.h (prune_runtime_alias_test_list): Take the
1332         factor as a poly_uint64 rather than an unsigned HOST_WIDE_INT.
1333         * tree-data-ref.c (prune_runtime_alias_test_list): Likewise.
1334         Track polynomial offsets.
1336 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1337             Alan Hayward  <alan.hayward@arm.com>
1338             David Sherwood  <david.sherwood@arm.com>
1340         * tree-vect-data-refs.c (vect_compute_data_ref_alignment):
1341         Treat drb->init as a poly_int.  Fail if its misalignment wrt
1342         vector_alignment isn't known.
1344 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1345             Alan Hayward  <alan.hayward@arm.com>
1346             David Sherwood  <david.sherwood@arm.com>
1348         * tree-vectorizer.h (_loop_vec_info): Add a versioning_threshold
1349         field.
1350         (LOOP_VINFO_VERSIONING_THRESHOLD): New macro
1351         (vect_loop_versioning): Take the loop versioning threshold as a
1352         separate parameter.
1353         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
1354         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
1355         versioning_threshold.
1356         (vect_analyze_loop_2): Compute the loop versioning threshold
1357         whenever loop versioning is needed, and store it in the new
1358         field rather than combining it with the cost model threshold.
1359         (vect_transform_loop): Update call to vect_loop_versioning.
1360         Try to combine the loop versioning and cost thresholds here.
1362 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1363             Alan Hayward  <alan.hayward@arm.com>
1364             David Sherwood  <david.sherwood@arm.com>
1366         * tree-ssa-loop-ivopts.h (strip_offset): Return the offset as
1367         poly_uint64_pod rather than an unsigned HOST_WIDE_INT.
1368         * tree-loop-distribution.c (classify_builtin_st): Update accordingly.
1369         * tree-ssa-loop-ivopts.c (iv_use::addr_offset): Change from
1370         an unsigned HOST_WIDE_INT to a poly_uint64_pod.
1371         (group_compare_offset): Update accordingly.
1372         (split_small_address_groups_p): Likewise.
1373         (record_use): Take addr_offset as a poly_uint64 rather than
1374         an unsigned HOST_WIDE_INT.
1375         (strip_offset): Return the offset as a poly_uint64 rather than
1376         an unsigned HOST_WIDE_INT.
1377         (record_group_use, split_address_groups): Track polynomial offsets.
1378         (add_iv_candidate_for_use): Likewise.
1379         (addr_offset_valid_p): Take the offset as a poly_int64 rather
1380         than a HOST_WIDE_INT.
1381         (strip_offset_1): Return the offset as a poly_int64 rather than
1382         a HOST_WIDE_INT.
1384 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1385             Alan Hayward  <alan.hayward@arm.com>
1386             David Sherwood  <david.sherwood@arm.com>
1388         * tree.h (get_binfo_at_offset): Take the offset as a poly_int64
1389         rather than a HOST_WIDE_INT.
1390         * tree.c (get_binfo_at_offset): Likewise.
1392 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1393             Alan Hayward  <alan.hayward@arm.com>
1394             David Sherwood  <david.sherwood@arm.com>
1396         * ipa-prop.h (build_ref_for_offset): Take the offset as a poly_int64
1397         rather than a HOST_WIDE_INT.
1398         * tree-sra.c (build_ref_for_offset): Likewise.
1400 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1401             Alan Hayward  <alan.hayward@arm.com>
1402             David Sherwood  <david.sherwood@arm.com>
1404         * fold-const.h (mem_ref_offset): Return a poly_offset_int rather
1405         than an offset_int.
1406         * tree.c (mem_ref_offset): Likewise.
1407         (build_simple_mem_ref_loc): Treat MEM_REF offsets as poly_ints.
1408         * builtins.c (get_object_alignment_2): Likewise.
1409         * expr.c (get_inner_reference, expand_expr_real_1): Likewise.
1410         * gimple-fold.c (get_base_constructor): Likewise.
1411         * gimple-ssa-strength-reduction.c (restructure_reference): Likewise.
1412         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
1413         Likewise.
1414         * ipa-polymorphic-call.c
1415         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Likewise.
1416         * ipa-prop.c (compute_complex_assign_jump_func): Likewise.
1417         (get_ancestor_addr_info): Likewise.
1418         * ipa-param-manipulation.c (ipa_get_adjustment_candidate): Likewise.
1419         * match.pd: Likewise.
1420         * tree-data-ref.c (dr_analyze_innermost): Likewise.
1421         * tree-dfa.c (get_addr_base_and_unit_offset_1): Likewise.
1422         * tree-eh.c (tree_could_trap_p): Likewise.
1423         * tree-object-size.c (addr_object_size): Likewise.
1424         * tree-ssa-address.c (copy_ref_info): Likewise.
1425         * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Likewise.
1426         (indirect_refs_may_alias_p): Likewise.
1427         * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Likewise.
1428         * tree-ssa.c (maybe_rewrite_mem_ref_base): Likewise.
1429         (non_rewritable_mem_ref_base): Likewise.
1430         * tree-vect-data-refs.c (vect_check_gather_scatter): Likewise.
1431         * tree-vrp.c (vrp_prop::check_array_ref): Likewise.
1432         * varasm.c (decode_addr_const): Likewise.
1434 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1435             Alan Hayward  <alan.hayward@arm.com>
1436             David Sherwood  <david.sherwood@arm.com>
1438         * gimple-ssa-stor-merging.c (find_bswap_or_nop_load): Track polynomial
1439         offsets for MEM_REFs.
1441 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1442             Alan Hayward  <alan.hayward@arm.com>
1443             David Sherwood  <david.sherwood@arm.com>
1445         * tree-ssanames.h (adjust_ptr_info_misalignment): Take the increment
1446         as a poly_uint64 rather than an unsigned int.
1447         * tree-ssanames.c (adjust_ptr_info_misalignment): Likewise.
1449 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1450             Alan Hayward  <alan.hayward@arm.com>
1451             David Sherwood  <david.sherwood@arm.com>
1453         * varasm.c (addr_const::offset): Change from HOST_WIDE_INT
1454         to poly_int64.
1455         (decode_addr_const): Update accordingly.
1457 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1458             Alan Hayward  <alan.hayward@arm.com>
1459             David Sherwood  <david.sherwood@arm.com>
1461         * tree.h (bit_field_size, bit_field_offset): New functions.
1462         * hsa-gen.c (gen_hsa_addr): Use them.
1463         * tree-ssa-forwprop.c (simplify_bitfield_ref): Likewise.
1464         (simplify_vector_constructor): Likewise.
1465         * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Likewise.
1466         * tree-cfg.c (verify_expr): Require the sizes and offsets of a
1467         BIT_FIELD_REF to be poly_uint64s rather than uhwis.
1468         * fold-const.c (fold_ternary_loc): Protect tree_to_uhwi with
1469         tree_fits_uhwi_p.
1471 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1472             Alan Hayward  <alan.hayward@arm.com>
1473             David Sherwood  <david.sherwood@arm.com>
1475         * expr.h (emit_group_load, emit_group_load_into_temps)
1476         (emit_group_store): Take the size as a poly_int64 rather than an int.
1477         * expr.c (emit_group_load_1, emit_group_load): Likewise.
1478         (emit_group_load_into_temp, emit_group_store): Likewise.
1480 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1481             Alan Hayward  <alan.hayward@arm.com>
1482             David Sherwood  <david.sherwood@arm.com>
1484         * ira-int.h (ira_spilled_reg_stack_slot::width): Change from
1485         an unsigned int to a poly_uint64.
1486         * ira.h (ira_reuse_stack_slot, ira_mark_new_stack_slot): Take the
1487         sizes as poly_uint64s rather than unsigned ints.
1488         * ira-color.c (ira_reuse_stack_slot, ira_mark_new_stack_slot):
1489         Likewise.
1491 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1492             Alan Hayward  <alan.hayward@arm.com>
1493             David Sherwood  <david.sherwood@arm.com>
1495         * lra-constraints.c (emit_inc): Change inc_amount from an int
1496         to a poly_int64.
1498 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1499             Alan Hayward  <alan.hayward@arm.com>
1500             David Sherwood  <david.sherwood@arm.com>
1502         * cfgexpand.c (stack_var::size): Change from a HOST_WIDE_INT
1503         to a poly_uint64.
1504         (add_stack_var, stack_var_cmp, partition_stack_vars)
1505         (dump_stack_var_partition): Update accordingly.
1506         (alloc_stack_frame_space): Take the size as a poly_int64 rather
1507         than a HOST_WIDE_INT.
1508         (expand_stack_vars, expand_one_stack_var_1): Handle polynomial sizes.
1509         (defer_stack_allocation, estimated_stack_frame_size): Likewise.
1510         (account_stack_vars, expand_one_var): Likewise.  Return a poly_uint64
1511         rather than a HOST_WIDE_INT.
1513 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1514             Alan Hayward  <alan.hayward@arm.com>
1515             David Sherwood  <david.sherwood@arm.com>
1517         * target.def (return_pops_args): Treat both the input and output
1518         sizes as poly_int64s rather than HOST_WIDE_INTS.
1519         * targhooks.h (default_return_pops_args): Update accordingly.
1520         * targhooks.c (default_return_pops_args): Likewise.
1521         * doc/tm.texi: Regenerate.
1522         * emit-rtl.h (incoming_args): Change pops_args, size and
1523         outgoing_args_size from int to poly_int64_pod.
1524         * function.h (expr_status): Change x_pending_stack_adjust and
1525         x_stack_pointer_delta from int to poly_int64.
1526         (args_size::constant): Change from HOST_WIDE_INT to poly_int64.
1527         (ARGS_SIZE_RTX): Update accordingly.
1528         * calls.c (highest_outgoing_arg_in_use): Change from int to
1529         unsigned int.
1530         (stack_usage_watermark, stored_args_watermark): New variables.
1531         (stack_region_maybe_used_p, mark_stack_region_used): New functions.
1532         (emit_call_1): Change the stack_size and rounded_stack_size
1533         parameters from HOST_WIDE_INT to poly_int64.  Track n_popped
1534         as a poly_int64.
1535         (save_fixed_argument_area): Check stack_usage_watermark.
1536         (initialize_argument_information): Change old_pending_adj from
1537         a HOST_WIDE_INT * to a poly_int64_pod *.
1538         (compute_argument_block_size): Return the size as a poly_int64
1539         rather than an int.
1540         (finalize_must_preallocate): Track polynomial argument sizes.
1541         (compute_argument_addresses): Likewise.
1542         (internal_arg_pointer_based_exp): Track polynomial offsets.
1543         (mem_overlaps_already_clobbered_arg_p): Rename to...
1544         (mem_might_overlap_already_clobbered_arg_p): ...this and take the
1545         size as a poly_uint64 rather than an unsigned HOST_WIDE_INT.
1546         Check stored_args_used_watermark.
1547         (load_register_parameters): Update accordingly.
1548         (check_sibcall_argument_overlap_1): Likewise.
1549         (combine_pending_stack_adjustment_and_call): Take the unadjusted
1550         args size as a poly_int64 rather than an int.  Return a bool
1551         indicating whether the optimization was possible and return
1552         the new adjustment by reference.
1553         (check_sibcall_argument_overlap): Track polynomail argument sizes.
1554         Update stored_args_watermark.
1555         (can_implement_as_sibling_call_p): Handle polynomial argument sizes.
1556         (expand_call): Likewise.  Maintain stack_usage_watermark and
1557         stored_args_watermark.  Update calls to
1558         combine_pending_stack_adjustment_and_call.
1559         (emit_library_call_value_1): Handle polynomial argument sizes.
1560         Call stack_region_maybe_used_p and mark_stack_region_used.
1561         Maintain stack_usage_watermark.
1562         (store_one_arg): Likewise.  Update call to
1563         mem_overlaps_already_clobbered_arg_p.
1564         * config/arm/arm.c (arm_output_function_prologue): Add a cast to
1565         HOST_WIDE_INT.
1566         * config/avr/avr.c (avr_outgoing_args_size): Likewise.
1567         * config/microblaze/microblaze.c (microblaze_function_prologue):
1568         Likewise.
1569         * config/cr16/cr16.c (cr16_return_pops_args): Update for new
1570         TARGET_RETURN_POPS_ARGS interface.
1571         (cr16_compute_frame, cr16_initial_elimination_offset): Add casts
1572         to HOST_WIDE_INT.
1573         * config/ft32/ft32.c (ft32_compute_frame): Likewise.
1574         * config/i386/i386.c (ix86_return_pops_args): Update for new
1575         TARGET_RETURN_POPS_ARGS interface.
1576         (ix86_expand_split_stack_prologue): Add a cast to HOST_WIDE_INT.
1577         * config/moxie/moxie.c (moxie_compute_frame): Likewise.
1578         * config/m68k/m68k.c (m68k_return_pops_args): Update for new
1579         TARGET_RETURN_POPS_ARGS interface.
1580         * config/vax/vax.c (vax_return_pops_args): Likewise.
1581         * config/pa/pa.h (STACK_POINTER_OFFSET): Add a cast to poly_int64.
1582         (EXIT_IGNORE_STACK): Update reference to crtl->outgoing_args_size.
1583         * config/arm/arm.h (CALLER_INTERWORKING_SLOT_SIZE): Likewise.
1584         * config/powerpcspe/aix.h (STACK_DYNAMIC_OFFSET): Likewise.
1585         * config/powerpcspe/darwin.h (STACK_DYNAMIC_OFFSET): Likewise.
1586         * config/powerpcspe/powerpcspe.h (STACK_DYNAMIC_OFFSET): Likewise.
1587         * config/rs6000/aix.h (STACK_DYNAMIC_OFFSET): Likewise.
1588         * config/rs6000/darwin.h (STACK_DYNAMIC_OFFSET): Likewise.
1589         * config/rs6000/rs6000.h (STACK_DYNAMIC_OFFSET): Likewise.
1590         * dojump.h (saved_pending_stack_adjust): Change x_pending_stack_adjust
1591         and x_stack_pointer_delta from int to poly_int64.
1592         * dojump.c (do_pending_stack_adjust): Update accordingly.
1593         * explow.c (allocate_dynamic_stack_space): Handle polynomial
1594         stack_pointer_deltas.
1595         * function.c (STACK_DYNAMIC_OFFSET): Add a cast to poly_int64.
1596         (pad_to_arg_alignment): Track polynomial offsets.
1597         (assign_parm_find_stack_rtl): Likewise.
1598         (assign_parms, locate_and_pad_parm): Handle polynomial argument sizes.
1599         * toplev.c (output_stack_usage): Update reference to
1600         current_function_pushed_stack_size.
1602 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1603             Alan Hayward  <alan.hayward@arm.com>
1604             David Sherwood  <david.sherwood@arm.com>
1606         * function.c (in_arg_offset, var_offset, dynamic_offset)
1607         (out_arg_offset, cfa_offset): Change from int to poly_int64.
1608         (instantiate_new_reg): Return the new offset as a poly_int64_pod
1609         rather than a HOST_WIDE_INT.
1610         (instantiate_virtual_regs_in_rtx): Track polynomial offsets.
1611         (instantiate_virtual_regs_in_insn): Likewise.
1613 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1614             Alan Hayward  <alan.hayward@arm.com>
1615             David Sherwood  <david.sherwood@arm.com>
1617         * rtl.h (get_args_size, add_args_size_note): New functions.
1618         (find_args_size_adjust): Return a poly_int64 rather than a
1619         HOST_WIDE_INT.
1620         (fixup_args_size_notes): Likewise.  Make the same change to the
1621         end_args_size parameter.
1622         * rtlanal.c (get_args_size, add_args_size_note): New functions.
1623         * builtins.c (expand_builtin_trap): Use add_args_size_note.
1624         * calls.c (emit_call_1): Likewise.
1625         * explow.c (adjust_stack_1): Likewise.
1626         * cfgcleanup.c (old_insns_match_p): Update use of
1627         find_args_size_adjust.
1628         * combine.c (distribute_notes): Track polynomial arg sizes.
1629         * dwarf2cfi.c (dw_trace_info): Change beg_true_args_size,
1630         end_true_args_size, beg_delay_args_size and end_delay_args_size
1631         from HOST_WIDE_INT to poly_int64.
1632         (add_cfi_args_size): Take the args_size as a poly_int64 rather
1633         than a HOST_WIDE_INT.
1634         (notice_args_size, notice_eh_throw, maybe_record_trace_start)
1635         (maybe_record_trace_start_abnormal, scan_trace, connect_traces): Track
1636         polynomial arg sizes.
1637         * emit-rtl.c (try_split): Use get_args_size.
1638         * recog.c (peep2_attempt): Likewise.
1639         * reload1.c (reload_as_needed): Likewise.
1640         * expr.c (find_args_size_adjust): Return the adjustment as a
1641         poly_int64 rather than a HOST_WIDE_INT.
1642         (fixup_args_size_notes): Change end_args_size from a HOST_WIDE_INT
1643         to a poly_int64 and change the return type in the same way.
1644         (emit_single_push_insn): Track polynomial arg sizes.
1646 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1647             Alan Hayward  <alan.hayward@arm.com>
1648             David Sherwood  <david.sherwood@arm.com>
1650         * expr.h (push_block, emit_push_insn): Change the "extra" parameter
1651         from HOST_WIDE_INT to poly_int64.
1652         * expr.c (push_block, emit_push_insn): Likewise.
1654 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1655             Alan Hayward  <alan.hayward@arm.com>
1656             David Sherwood  <david.sherwood@arm.com>
1658         * function.h (frame_space): Change start and length from HOST_WIDE_INT
1659         to poly_int64.
1660         (get_frame_size): Return the size as a poly_int64 rather than a
1661         HOST_WIDE_INT.
1662         (frame_offset_overflow): Take the offset as a poly_int64 rather
1663         than a HOST_WIDE_INT.
1664         (assign_stack_local_1, assign_stack_local, assign_stack_temp_for_type)
1665         (assign_stack_temp): Likewise for the size.
1666         * function.c (get_frame_size): Return a poly_int64 rather than
1667         a HOST_WIDE_INT.
1668         (frame_offset_overflow): Take the offset as a poly_int64 rather
1669         than a HOST_WIDE_INT.
1670         (try_fit_stack_local): Take the start, length and size as poly_int64s
1671         rather than HOST_WIDE_INTs.  Return the offset as a poly_int64_pod
1672         rather than a HOST_WIDE_INT.
1673         (add_frame_space): Take the start and end as poly_int64s rather than
1674         HOST_WIDE_INTs.
1675         (assign_stack_local_1, assign_stack_local, assign_stack_temp_for_type)
1676         (assign_stack_temp): Likewise for the size.
1677         (temp_slot): Change size, base_offset and full_size from HOST_WIDE_INT
1678         to poly_int64.
1679         (find_temp_slot_from_address): Handle polynomial offsets.
1680         (combine_temp_slots): Likewise.
1681         * emit-rtl.h (rtl_data::x_frame_offset): Change from HOST_WIDE_INT
1682         to poly_int64.
1683         * cfgexpand.c (alloc_stack_frame_space): Return the offset as a
1684         poly_int64 rather than a HOST_WIDE_INT.
1685         (expand_one_stack_var_at): Take the offset as a poly_int64 rather
1686         than a HOST_WIDE_INT.
1687         (expand_stack_vars, expand_one_stack_var_1, expand_used_vars): Handle
1688         polynomial frame offsets.
1689         * config/m32r/m32r-protos.h (m32r_compute_frame_size): Take the size
1690         as a poly_int64 rather than an int.
1691         * config/m32r/m32r.c (m32r_compute_frame_size): Likewise.
1692         * config/v850/v850-protos.h (compute_frame_size): Likewise.
1693         * config/v850/v850.c (compute_frame_size): Likewise.
1694         * config/xtensa/xtensa-protos.h (compute_frame_size): Likewise.
1695         * config/xtensa/xtensa.c (compute_frame_size): Likewise.
1696         * config/pa/pa-protos.h (pa_compute_frame_size): Likewise.
1697         * config/pa/pa.c (pa_compute_frame_size): Likewise.
1698         * explow.h (get_dynamic_stack_base): Take the offset as a poly_int64
1699         rather than a HOST_WIDE_INT.
1700         * explow.c (get_dynamic_stack_base): Likewise.
1701         * final.c (final_start_function): Use the constant lower bound
1702         of the frame size for -Wframe-larger-than.
1703         * ira.c (do_reload): Adjust for new get_frame_size return type.
1704         * lra.c (lra): Likewise.
1705         * reload1.c (reload): Likewise.
1706         * config/avr/avr.c (avr_asm_function_end_prologue): Likewise.
1707         * config/pa/pa.h (EXIT_IGNORE_STACK): Likewise.
1708         * rtlanal.c (get_initial_register_offset): Return the offset as
1709         a poly_int64 rather than a HOST_WIDE_INT.
1711 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1712             Alan Hayward  <alan.hayward@arm.com>
1713             David Sherwood  <david.sherwood@arm.com>
1715         * reload1.c (elim_table): Change initial_offset, offset and
1716         previous_offset from HOST_WIDE_INT to poly_int64_pod.
1717         (offsets_at): Change the target array's element type from
1718         HOST_WIDE_INT to poly_int64_pod.
1719         (set_label_offsets, eliminate_regs_1, eliminate_regs_in_insn)
1720         (elimination_costs_in_insn, update_eliminable_offsets)
1721         (verify_initial_elim_offsets, set_offsets_for_label)
1722         (init_eliminable_invariants): Update after above changes.
1724 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1725             Alan Hayward  <alan.hayward@arm.com>
1726             David Sherwood  <david.sherwood@arm.com>
1728         * reload.h (reload::inc): Change from an int to a poly_int64_pod.
1729         * reload.c (combine_reloads, debug_reload_to_stream): Likewise.
1730         (decomposition): Change start and end from HOST_WIDE_INT
1731         to poly_int64_pod.
1732         (decompose, immune_p): Update accordingly.
1733         (find_inc_amount): Return a poly_int64 rather than an int.
1734         * reload1.c (inc_for_reload): Take the inc_amount as a poly_int64
1735         rather than an int.
1737 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1738             Alan Hayward  <alan.hayward@arm.com>
1739             David Sherwood  <david.sherwood@arm.com>
1741         * tree.h (get_inner_reference): Return the bitsize and bitpos
1742         as poly_int64_pods rather than HOST_WIDE_INT.
1743         * fold-const.h (ptr_difference_const): Return the pointer difference
1744         as a poly_int64_pod rather than a HOST_WIDE_INT.
1745         * expr.c (get_inner_reference): Return the bitsize and bitpos
1746         as poly_int64_pods rather than HOST_WIDE_INT.
1747         (expand_expr_addr_expr_1, expand_expr_real_1): Track polynomial
1748         offsets and sizes.
1749         * fold-const.c (make_bit_field_ref): Take the bitpos as a poly_int64
1750         rather than a HOST_WIDE_INT.  Update call to get_inner_reference.
1751         (optimize_bit_field_compare): Update call to get_inner_reference.
1752         (decode_field_reference): Likewise.
1753         (fold_unary_loc): Track polynomial offsets and sizes.
1754         (split_address_to_core_and_offset): Return the bitpos as a
1755         poly_int64_pod rather than a HOST_WIDE_INT.
1756         (ptr_difference_const): Likewise for the pointer difference.
1757         * asan.c (instrument_derefs): Track polynomial offsets and sizes.
1758         * config/mips/mips.c (r10k_safe_mem_expr_p): Likewise.
1759         * dbxout.c (dbxout_expand_expr): Likewise.
1760         * dwarf2out.c (loc_list_for_address_of_addr_expr_of_indirect_ref)
1761         (loc_list_from_tree_1, fortran_common): Likewise.
1762         * gimple-laddress.c (pass_laddress::execute): Likewise.
1763         * gimple-ssa-store-merging.c (find_bswap_or_nop_load): Likewise.
1764         * gimplify.c (gimplify_scan_omp_clauses): Likewise.
1765         * simplify-rtx.c (delegitimize_mem_from_attrs): Likewise.
1766         * tree-affine.c (tree_to_aff_combination): Likewise.
1767         (get_inner_reference_aff): Likewise.
1768         * tree-data-ref.c (split_constant_offset_1): Likewise.
1769         (dr_analyze_innermost): Likewise.
1770         * tree-scalar-evolution.c (interpret_rhs_expr): Likewise.
1771         * tree-sra.c (ipa_sra_check_caller): Likewise.
1772         * tree-vect-data-refs.c (vect_check_gather_scatter): Likewise.
1773         * ubsan.c (maybe_instrument_pointer_overflow): Likewise.
1774         (instrument_bool_enum_load, instrument_object_size): Likewise.
1775         * gimple-ssa-strength-reduction.c (slsr_process_ref): Update call
1776         to get_inner_reference.
1777         * hsa-gen.c (gen_hsa_addr): Likewise.
1778         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Likewise.
1779         * tsan.c (instrument_expr): Likewise.
1780         * match.pd: Update call to ptr_difference_const.
1782 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1783             Alan Hayward  <alan.hayward@arm.com>
1784             David Sherwood  <david.sherwood@arm.com>
1786         * fold-const.c (fold_comparison): Track sizes and offsets as
1787         poly_int64s rather than HOST_WIDE_INTs when folding address
1788         comparisons.
1790 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1791             Alan Hayward  <alan.hayward@arm.com>
1792             David Sherwood  <david.sherwood@arm.com>
1794         * expr.h (get_bit_range): Return the bitstart and bitend as
1795         poly_uint64s rather than unsigned HOST_WIDE_INTs.  Return the bitpos
1796         as a poly_int64 rather than a HOST_WIDE_INT.
1797         * expr.c (get_bit_range): Likewise.
1798         (expand_assignment): Update call accordingly.
1799         * fold-const.c (optimize_bit_field_compare): Likewise.
1801 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1802             Alan Hayward  <alan.hayward@arm.com>
1803             David Sherwood  <david.sherwood@arm.com>
1805         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): New enum.
1806         (aarch64_legitimate_address_p): Use it instead of an rtx code,
1807         as an optional final parameter.
1808         * config/aarch64/aarch64.c (aarch64_classify_address): Likewise.
1809         (aarch64_legitimate_address_p): Likewise.
1810         (aarch64_print_address_internal): Take an aarch64_addr_query_type
1811         instead of an rtx code.
1812         (aarch64_address_valid_for_prefetch_p): Update calls accordingly.
1813         (aarch64_legitimate_address_hook_p): Likewise.
1814         (aarch64_print_ldpstp_address): Likewise.
1815         (aarch64_print_operand_address): Likewise.
1816         (aarch64_address_cost): Likewise.
1817         * config/aarch64/constraints.md (Uml, Umq, Ump, Utq): Likewise.
1818         * config/aarch64/predicates.md (aarch64_mem_pair_operand): Likewise.
1819         (aarch64_mem_pair_lanes_operand): Likewise.
1821 2017-12-20  Richard Biener  <rguenther@suse.de>
1823         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Call
1824         update_stmt_if_modified.
1826 2017-12-20  Wilco Dijkstra  <wdijkstr@arm.com>
1828         PR tree-optimization/83491
1829         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Check for SSA_NAME
1830         before walking uses.  Improve coding style and comments.
1832 2017-12-20  Tom de Vries  <tom@codesourcery.com>
1834         * gimple-fold.c (fold_internal_goacc_dim): Simplify.
1836 2017-12-20  Jakub Jelinek  <jakub@redhat.com>
1838         PR ipa/83506
1839         * ipa-fnsummary.c (pass_data_ipa_free_fn_summary): Use 0 for
1840         todo_flags_finish.
1841         (pass_ipa_free_fn_summary): Add small_p private data member,
1842         initialize to false in the ctor.
1843         (pass_ipa_free_fn_summary::clone,
1844         pass_ipa_free_fn_summary::set_pass_param,
1845         pass_ipa_free_fn_summary::gate): New methods.
1846         (pass_ipa_free_fn_summary::execute): Return TODO_remove_functions
1847         | TODO_dump_symtab if small_p.
1848         * passes.def: Add true parm for the existing pass_ipa_free_fn_summary
1849         entry and add another instance of the pass with false parm after
1850         ipa-pure-const.
1851         * ipa-pure-const.c (pass_ipa_pure_const): Don't call
1852         ipa_free_fn_summary here.
1854 2017-12-20  Paolo Carlini  <paolo.carlini@oracle.com>
1856         * gimplify.c (gimplify_return_expr): Remove dead error_mark_node check.
1858 2017-12-20  Martin Sebor  <msebor@redhat.com>
1860         PR testsuite/83131
1861         * builtins.c (expand_builtin_strlen): Use get_callee_fndecl.
1862         (expand_builtin_strcmp): Call maybe_warn_nonstring_arg.
1863         (expand_builtin_strncmp): Same.
1865 2017-12-20  Alexandre Oliva  <aoliva@redhat.com>
1867         PR bootstrap/83396
1868         * cfgexpand.c (label_rtx_for_bb): Revert SFN changes that
1869         allowed debug stmts before labels.
1870         (expand_gimple_basic_block): Likewise.
1871         * gimple-iterator.c (gimple_find_edge_insert_loc): Likewise.
1872         * gimple-iterator.h (gsi_after_labels): Likewise.
1873         * tree-cfgcleanup (remove_forwarder_block): Likewise, but
1874         rename reused variable, and simplify using gsi_move_before.
1875         * tree-ssa-tail-merge.c (find_duplicate): Likewise.
1876         * tree-cfg.c (make_edges, cleanup_dead_labels): Likewise.
1877         (gimple_can_merge_blocks_p, verify_gimple_in_cfg): Likewise.
1878         (gimple_verify_flow_info, gimple_block_label): Likewise.
1879         (make_blocks): Move debug markers after adjacent labels.
1880         * cfgrtl.c (skip_insns_after_block): Revert SFN changes that
1881         allowed debug insns outside blocks.
1882         * df-scan.c (df_insn_delete): Likewise.
1883         * lra-constraints.c (update_ebb_live_info): Likewise.
1884         * var-tracking.c (get_first_insn, vt_emit_notes): Likewise.
1885         (vt_initialize, delete_vta_debug_insns): Likewise.
1886         (reemit_marker_as_note): Drop BB parm.  Adjust callers.
1888 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1889             Alan Hayward  <alan.hayward@arm.com>
1890             David Sherwood  <david.sherwood@arm.com>
1892         * poly-int-types.h (round_down_to_byte_boundary): New macro.
1893         (round_up_to_byte_boundary): Likewise.
1894         * expr.h (get_bit_range): Add temporary shim.
1895         * gimple-ssa-store-merging.c (store_operand_info): Change the
1896         bitsize, bitpos, bitregion_start and bitregion_end fields from
1897         unsigned HOST_WIDE_INT to poly_uint64.
1898         (merged_store_group): Likewise load_align_base.
1899         (compatible_load_p, compatible_load_p): Update accordingly.
1900         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
1901         (split_group, imm_store_chain_info::output_merged_store): Likewise.
1902         (mem_valid_for_store_merging): Return the bitsize, bitpos,
1903         bitregion_start and bitregion_end as poly_uint64s rather than
1904         unsigned HOST_WIDE_INTs.  Track polynomial offsets internally.
1905         (handled_load): Take the bitsize, bitpos,
1906         bitregion_start and bitregion_end as poly_uint64s rather than
1907         unsigned HOST_WIDE_INTs.
1908         (pass_store_merging::process_store): Update call to
1909         mem_valid_for_store_merging.
1911 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1912             Alan Hayward  <alan.hayward@arm.com>
1913             David Sherwood  <david.sherwood@arm.com>
1915         * builtins.c (get_object_alignment_2): Track polynomial offsets
1916         and sizes.  Update the alignment handling.
1918 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1919             Alan Hayward  <alan.hayward@arm.com>
1920             David Sherwood  <david.sherwood@arm.com>
1922         * tree.h (get_inner_reference): Add a version that returns the
1923         offset and size as poly_int64_pods rather than HOST_WIDE_INTs.
1924         * cfgexpand.c (expand_debug_expr): Track polynomial offsets.  Simply
1925         the case in which bitpos is not associated with the first byte.
1927 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1928             Alan Hayward  <alan.hayward@arm.com>
1929             David Sherwood  <david.sherwood@arm.com>
1931         * tree-affine.h (get_inner_reference_aff): Return the size as a
1932         poly_widest_int.
1933         * tree-affine.c (get_inner_reference_aff): Likewise.
1934         * tree-data-ref.c (dr_may_alias_p): Update accordingly.
1935         * tree-ssa-loop-im.c (mem_refs_may_alias_p): Likewise.
1937 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1938             Alan Hayward  <alan.hayward@arm.com>
1939             David Sherwood  <david.sherwood@arm.com>
1941         * fold-const.c (pointer_may_wrap_p): Take the offset as a
1942         HOST_WIDE_INT rather than a poly_int64.
1944 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1945             Alan Hayward  <alan.hayward@arm.com>
1946             David Sherwood  <david.sherwood@arm.com>
1948         * gimple-ssa-store-merging.c (symbolic_number::bytepos): Change from
1949         HOST_WIDE_INT to poly_int64_pod.
1950         (perform_symbolic_merge): Update accordingly.
1951         (bswap_replace): Likewise.
1953 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1954             Alan Hayward  <alan.hayward@arm.com>
1955             David Sherwood  <david.sherwood@arm.com>
1957         * tree-affine.h (aff_tree::offset): Change from widest_int
1958         to poly_widest_int.
1959         (wide_int_ext_for_comb): Delete.
1960         (aff_combination_const, aff_comb_cannot_overlap_p): Take the
1961         constants as poly_widest_int rather than widest_int.
1962         (aff_combination_constant_multiple_p): Return the multiplier
1963         as a poly_widest_int.
1964         (aff_combination_zero_p, aff_combination_singleton_var_p): Handle
1965         polynomial offsets.
1966         * tree-affine.c (wide_int_ext_for_comb): Make original widest_int
1967         version static and add an overload for poly_widest_int.
1968         (aff_combination_const, aff_combination_add_cst)
1969         (wide_int_constant_multiple_p, aff_comb_cannot_overlap_p): Take
1970         the constants as poly_widest_int rather than widest_int.
1971         (tree_to_aff_combination): Generalize INTEGER_CST case to
1972         poly_int_tree_p.
1973         (aff_combination_to_tree): Track offsets as poly_widest_ints.
1974         (aff_combination_add_product, aff_combination_mult): Handle
1975         polynomial offsets.
1976         (aff_combination_constant_multiple_p): Return the multiplier
1977         as a poly_widest_int.
1978         * tree-predcom.c (determine_offset): Return the offset as a
1979         poly_widest_int.
1980         (split_data_refs_to_components, suitable_component_p): Update
1981         accordingly.
1982         (valid_initializer_p): Update call to
1983         aff_combination_constant_multiple_p.
1984         * tree-ssa-address.c (addr_to_parts): Handle polynomial offsets.
1985         * tree-ssa-loop-ivopts.c (get_address_cost_ainc): Take the step
1986         as a poly_int64 rather than a HOST_WIDE_INT.
1987         (get_address_cost): Handle polynomial offsets.
1988         (iv_elimination_compare_lt): Likewise.
1989         (rewrite_use_nonlinear_expr): Likewise.
1991 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1992             Alan Hayward  <alan.hayward@arm.com>
1993             David Sherwood  <david.sherwood@arm.com>
1995         * tree-dfa.h (get_addr_base_and_unit_offset_1): Return the offset
1996         as a poly_int64_pod rather than a HOST_WIDE_INT.
1997         (get_addr_base_and_unit_offset): Likewise.
1998         * tree-dfa.c (get_addr_base_and_unit_offset_1): Likewise.
1999         (get_addr_base_and_unit_offset): Likewise.
2000         * doc/match-and-simplify.texi: Change off from HOST_WIDE_INT
2001         to poly_int64 in example.
2002         * fold-const.c (fold_binary_loc): Update call to
2003         get_addr_base_and_unit_offset.
2004         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
2005         (maybe_canonicalize_mem_ref_addr): Likewise.
2006         (gimple_fold_stmt_to_constant_1): Likewise.
2007         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
2008         Likewise.
2009         * ipa-param-manipulation.c (ipa_modify_call_arguments): Likewise.
2010         * match.pd: Likewise.
2011         * omp-low.c (lower_omp_target): Likewise.
2012         * tree-sra.c (build_ref_for_offset): Likewise.
2013         (build_debug_ref_for_model): Likewise.
2014         * tree-ssa-address.c (maybe_fold_tmr): Likewise.
2015         * tree-ssa-alias.c (ao_ref_init_from_ptr_and_size): Likewise.
2016         * tree-ssa-ccp.c (optimize_memcpy): Likewise.
2017         * tree-ssa-forwprop.c (forward_propagate_addr_expr_1): Likewise.
2018         (constant_pointer_difference): Likewise.
2019         * tree-ssa-loop-niter.c (expand_simple_operations): Likewise.
2020         * tree-ssa-phiopt.c (jump_function_from_stmt): Likewise.
2021         * tree-ssa-pre.c (create_component_ref_by_pieces_1): Likewise.
2022         * tree-ssa-sccvn.c (vn_reference_fold_indirect): Likewise.
2023         (vn_reference_maybe_forwprop_address, vn_reference_lookup_3): Likewise.
2024         (set_ssa_val_to): Likewise.
2025         * tree-ssa-strlen.c (get_addr_stridx, addr_stridxptr)
2026         (maybe_diag_stxncpy_trunc): Likewise.
2027         * tree-vrp.c (vrp_prop::check_array_ref): Likewise.
2028         * tree.c (build_simple_mem_ref_loc): Likewise.
2029         (array_at_struct_end_p): Likewise.
2031 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2032             Alan Hayward  <alan.hayward@arm.com>
2033             David Sherwood  <david.sherwood@arm.com>
2035         * tree-dfa.h (get_ref_base_and_extent): Return the base, size and
2036         max_size as poly_int64_pods rather than HOST_WIDE_INTs.
2037         (get_ref_base_and_extent_hwi): Declare.
2038         * tree-dfa.c (get_ref_base_and_extent): Return the base, size and
2039         max_size as poly_int64_pods rather than HOST_WIDE_INTs.
2040         (get_ref_base_and_extent_hwi): New function.
2041         * cfgexpand.c (expand_debug_expr): Update call to
2042         get_ref_base_and_extent.
2043         * dwarf2out.c (add_var_loc_to_decl): Likewise.
2044         * gimple-fold.c (get_base_constructor): Return the offset as a
2045         poly_int64_pod rather than a HOST_WIDE_INT.
2046         (fold_const_aggregate_ref_1): Track polynomial sizes and offsets.
2047         * ipa-polymorphic-call.c
2048         (ipa_polymorphic_call_context::set_by_invariant)
2049         (extr_type_from_vtbl_ptr_store): Track polynomial offsets.
2050         (ipa_polymorphic_call_context::ipa_polymorphic_call_context)
2051         (check_stmt_for_type_change): Use get_ref_base_and_extent_hwi
2052         rather than get_ref_base_and_extent.
2053         (ipa_polymorphic_call_context::get_dynamic_type): Likewise.
2054         * ipa-prop.c (ipa_load_from_parm_agg, compute_complex_assign_jump_func)
2055         (get_ancestor_addr_info, determine_locally_known_aggregate_parts):
2056         Likewise.
2057         * ipa-param-manipulation.c (ipa_get_adjustment_candidate): Update
2058         call to get_ref_base_and_extent.
2059         * tree-sra.c (create_access, get_access_for_expr): Likewise.
2060         * tree-ssa-alias.c (ao_ref_base, aliasing_component_refs_p)
2061         (stmt_kills_ref_p): Likewise.
2062         * tree-ssa-dce.c (mark_aliased_reaching_defs_necessary_1): Likewise.
2063         * tree-ssa-scopedtables.c (avail_expr_hash, equal_mem_array_ref_p):
2064         Likewise.
2065         * tree-ssa-sccvn.c (vn_reference_lookup_3): Likewise.
2066         Use get_ref_base_and_extent_hwi rather than get_ref_base_and_extent
2067         when calling native_encode_expr.
2068         * tree-ssa-structalias.c (get_constraint_for_component_ref): Update
2069         call to get_ref_base_and_extent.
2070         (do_structure_copy): Use get_ref_base_and_extent_hwi rather than
2071         get_ref_base_and_extent.
2072         * var-tracking.c (track_expr_p): Likewise.
2074 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2075             Alan Hayward  <alan.hayward@arm.com>
2076             David Sherwood  <david.sherwood@arm.com>
2078         * ipa-param-manipulation.h (ipa_parm_adjustment::offset): Change from
2079         HOST_WIDE_INT to poly_int64_pod.
2080         * ipa-param-manipulation.c (ipa_modify_call_arguments): Track
2081         polynomail parameter offsets.
2083 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2084             Alan Hayward  <alan.hayward@arm.com>
2085             David Sherwood  <david.sherwood@arm.com>
2087         * gengtype.c (main): Handle poly_int64_pod.
2088         * dwarf2out.h (dw_cfi_oprnd_cfa_loc): New dw_cfi_oprnd_type.
2089         (dw_cfi_oprnd::dw_cfi_cfa_loc): New field.
2090         (dw_cfa_location::offset, dw_cfa_location::base_offset): Change
2091         from HOST_WIDE_INT to poly_int64_pod.
2092         * dwarf2cfi.c (queued_reg_save::cfa_offset): Likewise.
2093         (copy_cfa): New function.
2094         (lookup_cfa_1): Use the cached dw_cfi_cfa_loc, if it exists.
2095         (cfi_oprnd_equal_p): Handle dw_cfi_oprnd_cfa_loc.
2096         (cfa_equal_p, dwarf2out_frame_debug_adjust_cfa)
2097         (dwarf2out_frame_debug_cfa_offset, dwarf2out_frame_debug_expr)
2098         (initial_return_save): Treat offsets as poly_ints.
2099         (def_cfa_0): Likewise.  Cache the CFA in dw_cfi_cfa_loc if either
2100         offset is nonconstant.
2101         (reg_save): Take the offset as a poly_int64.  Fall back to
2102         DW_CFA_expression for nonconstant offsets.
2103         (queue_reg_save): Take the offset as a poly_int64.
2104         * dwarf2out.c (dw_cfi_oprnd2_desc): Handle DW_CFA_def_cfa_expression.
2106 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2107             Alan Hayward  <alan.hayward@arm.com>
2108             David Sherwood  <david.sherwood@arm.com>
2110         * rtl.h (operand_subword, operand_subword_force): Take the offset
2111         as a poly_uint64 an unsigned int.
2112         * emit-rtl.c (operand_subword, operand_subword_force): Likewise.
2114 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2115             Alan Hayward  <alan.hayward@arm.com>
2116             David Sherwood  <david.sherwood@arm.com>
2118         * doc/rtl.texi: Update documentation of SUBREG_BYTE.  Document the
2119         'p' format code.  Use INT_LIST rather than SUBREG as the example of
2120         a code with an XINT and an XEXP.  Remove the implication that
2121         accessing an rtx field using XINT is expected to work.
2122         * rtl.def (SUBREG): Change format from "ei" to "ep".
2123         * rtl.h (rtunion::rt_subreg): New field.
2124         (XCSUBREG): New macro.
2125         (SUBREG_BYTE): Use it.
2126         (subreg_shape): Change offset from an unsigned int to a poly_uint16.
2127         Update constructor accordingly.
2128         (subreg_shape::operator ==): Update accordingly.
2129         (subreg_shape::unique_id): Return an unsigned HOST_WIDE_INT rather
2130         than an unsigned int.
2131         (subreg_lsb, subreg_lowpart_offset, subreg_highpart_offset): Return
2132         a poly_uint64 rather than an unsigned int.
2133         (subreg_lsb_1): Likewise.  Take the offset as a poly_uint64 rather
2134         than an unsigned int.
2135         (subreg_size_offset_from_lsb, subreg_size_lowpart_offset)
2136         (subreg_size_highpart_offset): Return a poly_uint64 rather than
2137         an unsigned int.  Take the sizes as poly_uint64s.
2138         (subreg_offset_from_lsb): Return a poly_uint64 rather than
2139         an unsigned int.  Take the shift as a poly_uint64 rather than
2140         an unsigned int.
2141         (subreg_regno_offset, subreg_offset_representable_p): Take the offset
2142         as a poly_uint64 rather than an unsigned int.
2143         (simplify_subreg_regno): Likewise.
2144         (byte_lowpart_offset): Return the memory offset as a poly_int64
2145         rather than an int.
2146         (subreg_memory_offset): Likewise.  Take the subreg offset as a
2147         poly_uint64 rather than an unsigned int.
2148         (simplify_subreg, simplify_gen_subreg, subreg_get_info)
2149         (gen_rtx_SUBREG, validate_subreg): Take the subreg offset as a
2150         poly_uint64 rather than an unsigned int.
2151         * rtl.c (rtx_format): Describe 'p' in comment.
2152         (copy_rtx, rtx_equal_p_cb, rtx_equal_p): Handle 'p'.
2153         * emit-rtl.c (validate_subreg, gen_rtx_SUBREG): Take the subreg
2154         offset as a poly_uint64 rather than an unsigned int.
2155         (byte_lowpart_offset): Return the memory offset as a poly_int64
2156         rather than an int.
2157         (subreg_memory_offset): Likewise.  Take the subreg offset as a
2158         poly_uint64 rather than an unsigned int.
2159         (subreg_size_lowpart_offset, subreg_size_highpart_offset): Take the
2160         mode sizes as poly_uint64s rather than unsigned ints.  Return a
2161         poly_uint64 rather than an unsigned int.
2162         (subreg_lowpart_p): Treat subreg offsets as poly_ints.
2163         (copy_insn_1): Handle 'p'.
2164         * rtlanal.c (set_noop_p): Treat subregs offsets as poly_uint64s.
2165         (subreg_lsb_1): Take the subreg offset as a poly_uint64 rather than
2166         an unsigned int.  Return the shift in the same way.
2167         (subreg_lsb): Return the shift as a poly_uint64 rather than an
2168         unsigned int.
2169         (subreg_size_offset_from_lsb): Take the sizes and shift as
2170         poly_uint64s rather than unsigned ints.  Return the offset as
2171         a poly_uint64.
2172         (subreg_get_info, subreg_regno_offset, subreg_offset_representable_p)
2173         (simplify_subreg_regno): Take the offset as a poly_uint64 rather than
2174         an unsigned int.
2175         * rtlhash.c (add_rtx): Handle 'p'.
2176         * genemit.c (gen_exp): Likewise.
2177         * gengenrtl.c (type_from_format, gendef): Likewise.
2178         * gensupport.c (subst_pattern_match, get_alternatives_number)
2179         (collect_insn_data, alter_predicate_for_insn, alter_constraints)
2180         (subst_dup): Likewise.
2181         * gengtype.c (adjust_field_rtx_def): Likewise.
2182         * genrecog.c (find_operand, find_matching_operand, validate_pattern)
2183         (match_pattern_2): Likewise.
2184         (rtx_test::SUBREG_FIELD): New rtx_test::kind_enum.
2185         (rtx_test::subreg_field): New function.
2186         (operator ==, safe_to_hoist_p, transition_parameter_type)
2187         (print_nonbool_test, print_test): Handle SUBREG_FIELD.
2188         * genattrtab.c (attr_rtx_1): Say that 'p' is deliberately not handled.
2189         * genpeep.c (match_rtx): Likewise.
2190         * print-rtl.c (print_poly_int): Include if GENERATOR_FILE too.
2191         (rtx_writer::print_rtx_operand): Handle 'p'.
2192         (print_value): Handle SUBREG.
2193         * read-rtl.c (apply_int_iterator): Likewise.
2194         (rtx_reader::read_rtx_operand): Handle 'p'.
2195         * alias.c (rtx_equal_for_memref_p): Likewise.
2196         * cselib.c (rtx_equal_for_cselib_1, cselib_hash_rtx): Likewise.
2197         * caller-save.c (replace_reg_with_saved_mem): Treat subreg offsets
2198         as poly_ints.
2199         * calls.c (expand_call): Likewise.
2200         * combine.c (combine_simplify_rtx, expand_field_assignment): Likewise.
2201         (make_extraction, gen_lowpart_for_combine): Likewise.
2202         * loop-invariant.c (hash_invariant_expr_1, invariant_expr_equal_p):
2203         Likewise.
2204         * cse.c (remove_invalid_subreg_refs): Take the offset as a poly_uint64
2205         rather than an unsigned int.  Treat subreg offsets as poly_ints.
2206         (exp_equiv_p): Handle 'p'.
2207         (hash_rtx_cb): Likewise.  Treat subreg offsets as poly_ints.
2208         (equiv_constant, cse_insn): Treat subreg offsets as poly_ints.
2209         * dse.c (find_shift_sequence): Likewise.
2210         * dwarf2out.c (rtl_for_decl_location): Likewise.
2211         * expmed.c (extract_low_bits): Likewise.
2212         * expr.c (emit_group_store, undefined_operand_subword_p): Likewise.
2213         (expand_expr_real_2): Likewise.
2214         * final.c (alter_subreg): Likewise.
2215         (leaf_renumber_regs_insn): Handle 'p'.
2216         * function.c (assign_parm_find_stack_rtl, assign_parm_setup_stack):
2217         Treat subreg offsets as poly_ints.
2218         * fwprop.c (forward_propagate_and_simplify): Likewise.
2219         * ifcvt.c (noce_emit_move_insn, noce_emit_cmove): Likewise.
2220         * ira.c (get_subreg_tracking_sizes): Likewise.
2221         * ira-conflicts.c (go_through_subreg): Likewise.
2222         * ira-lives.c (process_single_reg_class_operands): Likewise.
2223         * jump.c (rtx_renumbered_equal_p): Likewise.  Handle 'p'.
2224         * lower-subreg.c (simplify_subreg_concatn): Take the subreg offset
2225         as a poly_uint64 rather than an unsigned int.
2226         (simplify_gen_subreg_concatn, resolve_simple_move): Treat
2227         subreg offsets as poly_ints.
2228         * lra-constraints.c (operands_match_p): Handle 'p'.
2229         (match_reload, curr_insn_transform): Treat subreg offsets as poly_ints.
2230         * lra-spills.c (assign_mem_slot): Likewise.
2231         * postreload.c (move2add_valid_value_p): Likewise.
2232         * recog.c (general_operand, indirect_operand): Likewise.
2233         * regcprop.c (copy_value, maybe_mode_change): Likewise.
2234         (copyprop_hardreg_forward_1): Likewise.
2235         * reginfo.c (simplifiable_subregs_hasher::hash, simplifiable_subregs)
2236         (record_subregs_of_mode): Likewise.
2237         * rtlhooks.c (gen_lowpart_general, gen_lowpart_if_possible): Likewise.
2238         * reload.c (operands_match_p): Handle 'p'.
2239         (find_reloads_subreg_address): Treat subreg offsets as poly_ints.
2240         * reload1.c (alter_reg, choose_reload_regs): Likewise.
2241         (compute_reload_subreg_offset): Likewise, and return an poly_int64.
2242         * simplify-rtx.c (simplify_truncation, simplify_binary_operation_1):
2243         (test_vector_ops_duplicate): Treat subreg offsets as poly_ints.
2244         (simplify_const_poly_int_tests<N>::run): Likewise.
2245         (simplify_subreg, simplify_gen_subreg): Take the subreg offset as
2246         a poly_uint64 rather than an unsigned int.
2247         * valtrack.c (debug_lowpart_subreg): Likewise.
2248         * var-tracking.c (var_lowpart): Likewise.
2249         (loc_cmp): Handle 'p'.
2251 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2252             Alan Hayward  <alan.hayward@arm.com>
2253             David Sherwood  <david.sherwood@arm.com>
2255         * ira.c (get_subreg_tracking_sizes): New function.
2256         (init_live_subregs): Take an integer size rather than a register.
2257         (build_insn_chain): Use get_subreg_tracking_sizes.  Update calls
2258         to init_live_subregs.
2260 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2261             Alan Hayward  <alan.hayward@arm.com>
2262             David Sherwood  <david.sherwood@arm.com>
2264         * expr.c (store_constructor_field): Change bitsize from a
2265         unsigned HOST_WIDE_INT to a poly_uint64 and bitpos from a
2266         HOST_WIDE_INT to a poly_int64.
2267         (store_constructor): Change size from a HOST_WIDE_INT to
2268         a poly_int64.
2269         (store_field): Likewise bitsize and bitpos.
2271 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2272             Alan Hayward  <alan.hayward@arm.com>
2273             David Sherwood  <david.sherwood@arm.com>
2275         * expmed.h (store_bit_field): Change bitregion_start and
2276         bitregion_end from unsigned HOST_WIDE_INT to poly_uint64.
2277         * expmed.c (adjust_bit_field_mem_for_reg, strict_volatile_bitfield_p)
2278         (store_bit_field_1, store_integral_bit_field, store_bit_field)
2279         (store_fixed_bit_field, store_split_bit_field): Likewise.
2280         * expr.c (store_constructor_field, store_field): Likewise.
2281         (optimize_bitfield_assignment_op): Likewise.  Make the same change
2282         to bitsize and bitpos.
2283         * machmode.h (bit_field_mode_iterator): Change m_bitregion_start
2284         and m_bitregion_end from HOST_WIDE_INT to poly_int64.  Make the
2285         same change in the constructor arguments.
2286         (get_best_mode): Change bitregion_start and bitregion_end from
2287         unsigned HOST_WIDE_INT to poly_uint64.
2288         * stor-layout.c (bit_field_mode_iterator::bit_field_mode_iterator):
2289         Change bitregion_start and bitregion_end from HOST_WIDE_INT to
2290         poly_int64.
2291         (bit_field_mode_iterator::next_mode): Update for new types
2292         of m_bitregion_start and m_bitregion_end.
2293         (get_best_mode): Change bitregion_start and bitregion_end from
2294         unsigned HOST_WIDE_INT to poly_uint64.
2296 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2297             Alan Hayward  <alan.hayward@arm.com>
2298             David Sherwood  <david.sherwood@arm.com>
2300         * rtl.h (simplify_gen_subreg): Add a temporary overload that
2301         accepts poly_uint64 offsets.
2302         * expmed.h (extract_bit_field): Take bitsize and bitnum as
2303         poly_uint64s rather than unsigned HOST_WIDE_INTs.
2304         * expmed.c (lowpart_bit_field_p): Likewise.
2305         (extract_bit_field_as_subreg): New function, split out from...
2306         (extract_bit_field_1): ...here.  Take bitsize and bitnum as
2307         poly_uint64s rather than unsigned HOST_WIDE_INTs.  For vector
2308         extractions, check that BITSIZE matches the size of the extracted
2309         value and that BITNUM is an exact multiple of that size.
2310         If all else fails, try forcing the value into memory if
2311         BITNUM is variable, and adjusting the address so that the
2312         offset is constant.  Split the part that can only handle constant
2313         bitsize and bitnum out into...
2314         (extract_integral_bit_field): ...this new function.
2315         (extract_bit_field): Take bitsize and bitnum as poly_uint64s
2316         rather than unsigned HOST_WIDE_INTs.
2318 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2319             Alan Hayward  <alan.hayward@arm.com>
2320             David Sherwood  <david.sherwood@arm.com>
2322         * expmed.h (store_bit_field): Take bitsize and bitnum as
2323         poly_uint64s rather than unsigned HOST_WIDE_INTs.
2324         * expmed.c (simple_mem_bitfield_p): Likewise.  Add a parameter
2325         that returns the byte size.
2326         (store_bit_field_1): Take bitsize and bitnum as
2327         poly_uint64s rather than unsigned HOST_WIDE_INTs.  Update call
2328         to simple_mem_bitfield_p.  Split the part that can only handle
2329         constant bitsize and bitnum out into...
2330         (store_integral_bit_field): ...this new function.
2331         (store_bit_field): Take bitsize and bitnum as poly_uint64s rather
2332         than unsigned HOST_WIDE_INTs.
2333         (extract_bit_field_1): Update call to simple_mem_bitfield_p.
2335 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2336             Alan Hayward  <alan.hayward@arm.com>
2337             David Sherwood  <david.sherwood@arm.com>
2339         * lra-int.h (lra_reg): Change offset from int to poly_int64.
2340         (lra_insn_recog_data): Change sp_offset from HOST_WIDE_INT
2341         to poly_int64.
2342         (lra_eliminate_regs_1, eliminate_regs_in_insn): Change
2343         update_sp_offset from a HOST_WIDE_INT to a poly_int64.
2344         (lra_update_reg_val_offset, lra_reg_val_equal_p): Take the
2345         offset as a poly_int64 rather than an int.
2346         * lra-assigns.c (find_hard_regno_for_1): Handle poly_int64 offsets.
2347         (setup_live_pseudos_and_spill_after_risky_transforms): Likewise.
2348         * lra-constraints.c (equiv_address_substitution): Track offsets
2349         as poly_int64s.
2350         (emit_inc): Check poly_int_rtx_p instead of CONST_INT_P.
2351         (curr_insn_transform): Handle the new form of sp_offset.
2352         * lra-eliminations.c (lra_elim_table): Change previous_offset
2353         and offset from HOST_WIDE_INT to poly_int64.
2354         (print_elim_table, update_reg_eliminate): Update accordingly.
2355         (self_elim_offsets): Change from HOST_WIDE_INT to poly_int64_pod.
2356         (get_elimination): Update accordingly.
2357         (form_sum): Check poly_int_rtx_p instead of CONST_INT_P.
2358         (lra_eliminate_regs_1, eliminate_regs_in_insn): Change
2359         update_sp_offset from a HOST_WIDE_INT to a poly_int64.  Handle
2360         poly_int64 offsets generally.
2361         (curr_sp_change): Change from HOST_WIDE_INT to poly_int64.
2362         (mark_not_eliminable, init_elimination): Update accordingly.
2363         (remove_reg_equal_offset_note): Return a bool and pass the new
2364         offset back by pointer as a poly_int64.
2365         * lra-remat.c (change_sp_offset): Take sp_offset as a poly_int64
2366         rather than a HOST_WIDE_INT.
2367         (do_remat): Track offsets poly_int64s.
2368         * lra.c (lra_update_insn_recog_data, setup_sp_offset): Likewise.
2370 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2371             Alan Hayward  <alan.hayward@arm.com>
2372             David Sherwood  <david.sherwood@arm.com>
2374         * rtl.h (mem_attrs): Add a default constructor.  Change size and
2375         offset from HOST_WIDE_INT to poly_int64.
2376         * emit-rtl.h (set_mem_offset, set_mem_size, adjust_address_1)
2377         (adjust_automodify_address_1, set_mem_attributes_minus_bitpos)
2378         (widen_memory_access): Take the sizes and offsets as poly_int64s
2379         rather than HOST_WIDE_INTs.
2380         * alias.c (ao_ref_from_mem): Handle the new form of MEM_OFFSET.
2381         (offset_overlap_p): Take poly_int64s rather than HOST_WIDE_INTs
2382         and ints.
2383         (adjust_offset_for_component_ref): Change the offset from a
2384         HOST_WIDE_INT to a poly_int64.
2385         (nonoverlapping_memrefs_p): Track polynomial offsets and sizes.
2386         * cfgcleanup.c (merge_memattrs): Update after mem_attrs changes.
2387         * dce.c (find_call_stack_args): Likewise.
2388         * dse.c (record_store): Likewise.
2389         * dwarf2out.c (tls_mem_loc_descriptor, dw_sra_loc_expr): Likewise.
2390         * print-rtl.c (rtx_writer::print_rtx): Likewise.
2391         * read-rtl-function.c (test_loading_mem): Likewise.
2392         * rtlanal.c (may_trap_p_1): Likewise.
2393         * simplify-rtx.c (delegitimize_mem_from_attrs): Likewise.
2394         * var-tracking.c (int_mem_offset, track_expr_p): Likewise.
2395         * emit-rtl.c (mem_attrs_eq_p, get_mem_align_offset): Likewise.
2396         (mem_attrs::mem_attrs): New function.
2397         (set_mem_attributes_minus_bitpos): Change bitpos from a
2398         HOST_WIDE_INT to poly_int64.
2399         (set_mem_alias_set, set_mem_addr_space, set_mem_align, set_mem_expr)
2400         (clear_mem_offset, clear_mem_size, change_address)
2401         (get_spill_slot_decl, set_mem_attrs_for_spill): Directly
2402         initialize mem_attrs.
2403         (set_mem_offset, set_mem_size, adjust_address_1)
2404         (adjust_automodify_address_1, offset_address, widen_memory_access):
2405         Likewise.  Take poly_int64s rather than HOST_WIDE_INT.
2407 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2408             Alan Hayward  <alan.hayward@arm.com>
2409             David Sherwood  <david.sherwood@arm.com>
2411         * rtlanal.c (rtx_addr_can_trap_p_1): Take the offset and size
2412         as poly_int64s rather than HOST_WIDE_INTs.  Use a size of -1
2413         rather than 0 to represent an unknown size.  Assert that the size
2414         is known when the mode isn't BLKmode.
2415         (may_trap_p_1): Use -1 for unknown sizes.
2416         (rtx_addr_can_trap_p): Likewise.  Pass BLKmode rather than VOIDmode.
2418 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2419             Alan Hayward  <alan.hayward@arm.com>
2420             David Sherwood  <david.sherwood@arm.com>
2422         * dse.c (store_info): Change offset and width from HOST_WIDE_INT
2423         to poly_int64.  Update commentary for positions_needed.large.
2424         (read_info_type): Change offset and width from HOST_WIDE_INT
2425         to poly_int64.
2426         (set_usage_bits): Likewise.
2427         (canon_address): Return the offset as a poly_int64 rather than
2428         a HOST_WIDE_INT.  Use strip_offset_and_add.
2429         (set_all_positions_unneeded, any_positions_needed_p): Use
2430         positions_needed.large to track stores with non-constant widths.
2431         (all_positions_needed_p): Likewise.  Take the offset and width
2432         as poly_int64s rather than ints.  Assert that rhs is nonnull.
2433         (record_store): Cope with non-constant offsets and widths.
2434         Nullify the rhs of an earlier store if we can't tell which bytes
2435         of it are needed.
2436         (find_shift_sequence): Take the access_size and shift as poly_int64s
2437         rather than ints.
2438         (get_stored_val): Take the read_offset and read_width as poly_int64s
2439         rather than HOST_WIDE_INTs.
2440         (check_mem_read_rtx, scan_stores, scan_reads, dse_step5): Handle
2441         non-constant offsets and widths.
2443 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2444             Alan Hayward  <alan.hayward@arm.com>
2445             David Sherwood  <david.sherwood@arm.com>
2447         * inchash.h (inchash::hash::add_poly_int): New function.
2448         * tree-ssa-alias.h (ao_ref::offset, ao_ref::size, ao_ref::max_size):
2449         Use poly_int64 rather than HOST_WIDE_INT.
2450         (ao_ref::max_size_known_p): New function.
2451         * tree-ssa-sccvn.h (vn_reference_op_struct::off): Use poly_int64_pod
2452         rather than HOST_WIDE_INT.
2453         * tree-ssa-alias.c (ao_ref_base): Apply get_ref_base_and_extent
2454         to temporaries until its interface is adjusted to match.
2455         (ao_ref_init_from_ptr_and_size): Handle polynomial offsets and sizes.
2456         (aliasing_component_refs_p, decl_refs_may_alias_p)
2457         (indirect_ref_may_alias_decl_p, indirect_refs_may_alias_p): Take
2458         the offsets and max_sizes as poly_int64s instead of HOST_WIDE_INTs.
2459         (refs_may_alias_p_1, stmt_kills_ref_p): Adjust for changes to
2460         ao_ref fields.
2461         * alias.c (ao_ref_from_mem): Likewise.
2462         * tree-ssa-dce.c (mark_aliased_reaching_defs_necessary_1): Likewise.
2463         * tree-ssa-dse.c (valid_ao_ref_for_dse, normalize_ref)
2464         (clear_bytes_written_by, setup_live_bytes_from_ref, compute_trims)
2465         (maybe_trim_complex_store, maybe_trim_constructor_store)
2466         (live_bytes_read, dse_classify_store): Likewise.
2467         * tree-ssa-sccvn.c (vn_reference_compute_hash, vn_reference_eq):
2468         (copy_reference_ops_from_ref, ao_ref_init_from_vn_reference)
2469         (fully_constant_vn_reference_p, valueize_refs_1): Likewise.
2470         (vn_reference_lookup_3): Likewise.
2471         * tree-ssa-uninit.c (warn_uninitialized_vars): Likewise.
2473 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2474             Alan Hayward  <alan.hayward@arm.com>
2475             David Sherwood  <david.sherwood@arm.com>
2477         * tree-ssa-alias.c (indirect_ref_may_alias_decl_p)
2478         (indirect_refs_may_alias_p): Use ranges_may_overlap_p
2479         instead of ranges_overlap_p.
2481 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2482             Alan Hayward  <alan.hayward@arm.com>
2483             David Sherwood  <david.sherwood@arm.com>
2485         * tree-ssa-alias.c (same_addr_size_stores_p): Take the offsets and
2486         sizes as poly_int64s rather than HOST_WIDE_INTs.
2488 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2489             Alan Hayward  <alan.hayward@arm.com>
2490             David Sherwood  <david.sherwood@arm.com>
2492         * gimple-fold.h (fold_ctor_reference): Take the offset and size
2493         as poly_uint64 rather than unsigned HOST_WIDE_INT.
2494         * gimple-fold.c (fold_ctor_reference): Likewise.
2496 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2497             Alan Hayward  <alan.hayward@arm.com>
2498             David Sherwood  <david.sherwood@arm.com>
2500         * target.def (dwarf_poly_indeterminate_value): New hook.
2501         * targhooks.h (default_dwarf_poly_indeterminate_value): Declare.
2502         * targhooks.c (default_dwarf_poly_indeterminate_value): New function.
2503         * doc/tm.texi.in (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Document.
2504         * doc/tm.texi: Regenerate.
2505         * dwarf2out.h (build_cfa_loc, build_cfa_aligned_loc): Take the
2506         offset as a poly_int64.
2507         * dwarf2out.c (new_reg_loc_descr): Move later in file.  Take the
2508         offset as a poly_int64.
2509         (loc_descr_plus_const, loc_list_plus_const, build_cfa_aligned_loc):
2510         Take the offset as a poly_int64.
2511         (build_cfa_loc): Likewise.  Use loc_descr_plus_const.
2512         (frame_pointer_fb_offset): Change to a poly_int64.
2513         (int_loc_descriptor): Take the offset as a poly_int64.  Use
2514         targetm.dwarf_poly_indeterminate_value for polynomial offsets.
2515         (based_loc_descr): Take the offset as a poly_int64.
2516         Use strip_offset_and_add to handle (plus X (const)).
2517         Use new_reg_loc_descr instead of an open-coded version of the
2518         previous implementation.
2519         (mem_loc_descriptor): Handle CONST_POLY_INT.
2520         (compute_frame_pointer_to_fb_displacement): Take the offset as a
2521         poly_int64.  Use strip_offset_and_add to handle (plus X (const)).
2523 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2524             Alan Hayward  <alan.hayward@arm.com>
2525             David Sherwood  <david.sherwood@arm.com>
2527         * rtl.h (reg_attrs::offset): Change from HOST_WIDE_INT to poly_int64.
2528         (gen_rtx_REG_offset): Take the offset as a poly_int64.
2529         * inchash.h (inchash::hash::add_poly_hwi): New function.
2530         * gengtype.c (main): Register poly_int64.
2531         * emit-rtl.c (reg_attr_hasher::hash): Use inchash.  Treat the
2532         offset as a poly_int.
2533         (reg_attr_hasher::equal): Use must_eq to compare offsets.
2534         (get_reg_attrs, update_reg_offset, gen_rtx_REG_offset): Take the
2535         offset as a poly_int64.
2536         (set_reg_attrs_from_value): Treat the offset as a poly_int64.
2537         * print-rtl.c (print_poly_int): New function.
2538         (rtx_writer::print_rtx_operand_code_r): Treat REG_OFFSET as
2539         a poly_int.
2540         * var-tracking.c (track_offset_p, get_tracked_reg_offset): New
2541         functions.
2542         (var_reg_set, var_reg_delete_and_set, var_reg_delete): Use them.
2543         (same_variable_part_p, track_loc_p): Take the offset as a poly_int64.
2544         (vt_get_decl_and_offset): Return the offset as a poly_int64.
2545         Enforce track_offset_p for parts of a PARALLEL.
2546         (vt_add_function_parameter): Use const_offset for the final
2547         offset to track.  Use get_tracked_reg_offset for the parts
2548         of a PARALLEL.
2550 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2551             Alan Hayward  <alan.hayward@arm.com>
2552             David Sherwood  <david.sherwood@arm.com>
2554         * target.def (truly_noop_truncation): Take poly_uint64s instead of
2555         unsigned ints.  Change default to hook_bool_puint64_puint64_true.
2556         * doc/tm.texi: Regenerate.
2557         * hooks.h (hook_bool_uint_uint_true): Delete.
2558         (hook_bool_puint64_puint64_true): Declare.
2559         * hooks.c (hook_bool_uint_uint_true): Delete.
2560         (hook_bool_puint64_puint64_true): New function.
2561         * config/mips/mips.c (mips_truly_noop_truncation): Take poly_uint64s
2562         instead of unsigned ints.
2563         * config/spu/spu.c (spu_truly_noop_truncation): Likewise.
2564         * config/tilegx/tilegx.c (tilegx_truly_noop_truncation): Likewise.
2566 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2567             Alan Hayward  <alan.hayward@arm.com>
2568             David Sherwood  <david.sherwood@arm.com>
2570         * optabs.h (expand_operand): Add an int_value field.
2571         (create_expand_operand): Add an int_value parameter and use it
2572         to initialize the new expand_operand field.
2573         (create_integer_operand): Replace with a declaration of a function
2574         that accepts poly_int64s.  Move the implementation to...
2575         * optabs.c (create_integer_operand): ...here.
2576         (maybe_legitimize_operand): For EXPAND_INTEGER, check whether
2577         the mode preserves the value of int_value, instead of calling
2578         const_int_operand on the rtx.  Use gen_int_mode to generate
2579         the new rtx.
2581 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2582             Alan Hayward  <alan.hayward@arm.com>
2583             David Sherwood  <david.sherwood@arm.com>
2585         * dumpfile.h (dump_dec): Declare.
2586         * dumpfile.c (dump_dec): New function.
2587         * pretty-print.h (pp_wide_integer): Turn into a function and
2588         declare a poly_int version.
2589         * pretty-print.c (pp_wide_integer): New function for poly_ints.
2591 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2592             Alan Hayward  <alan.hayward@arm.com>
2593             David Sherwood  <david.sherwood@arm.com>
2595         * doc/generic.texi (POLY_INT_CST): Document.
2596         * tree.def (POLY_INT_CST): New tree code.
2597         * treestruct.def (TS_POLY_INT_CST): New tree layout.
2598         * tree-core.h (tree_poly_int_cst): New struct.
2599         (tree_node): Add a poly_int_cst field.
2600         * tree.h (POLY_INT_CST_P, POLY_INT_CST_COEFF): New macros.
2601         (wide_int_to_tree, force_fit_type): Take a poly_wide_int_ref
2602         instead of a wide_int_ref.
2603         (build_int_cst, build_int_cst_type): Take a poly_int64 instead
2604         of a HOST_WIDE_INT.
2605         (build_int_cstu, build_array_type_nelts): Take a poly_uint64
2606         instead of an unsigned HOST_WIDE_INT.
2607         (build_poly_int_cst, tree_fits_poly_int64_p, tree_fits_poly_uint64_p)
2608         (ptrdiff_tree_p): Declare.
2609         (tree_to_poly_int64, tree_to_poly_uint64): Likewise.  Provide
2610         extern inline implementations if the target doesn't use POLY_INT_CST.
2611         (poly_int_tree_p): New function.
2612         (wi::unextended_tree): New class.
2613         (wi::int_traits <unextended_tree>): New override.
2614         (wi::extended_tree): Add a default constructor.
2615         (wi::extended_tree::get_tree): New function.
2616         (wi::widest_extended_tree, wi::offset_extended_tree): New typedefs.
2617         (wi::tree_to_widest_ref, wi::tree_to_offset_ref): Use them.
2618         (wi::tree_to_poly_widest_ref, wi::tree_to_poly_offset_ref)
2619         (wi::tree_to_poly_wide_ref): New typedefs.
2620         (wi::ints_for): Provide overloads for extended_tree and
2621         unextended_tree.
2622         (poly_int_cst_value, wi::to_poly_widest, wi::to_poly_offset)
2623         (wi::to_wide): New functions.
2624         (wi::fits_to_boolean_p, wi::fits_to_tree_p): Handle poly_ints.
2625         * tree.c (poly_int_cst_hasher): New struct.
2626         (poly_int_cst_hash_table): New variable.
2627         (tree_node_structure_for_code, tree_code_size, simple_cst_equal)
2628         (valid_constant_size_p, add_expr, drop_tree_overflow): Handle
2629         POLY_INT_CST.
2630         (initialize_tree_contains_struct): Handle TS_POLY_INT_CST.
2631         (init_ttree): Initialize poly_int_cst_hash_table.
2632         (build_int_cst, build_int_cst_type, build_invariant_address): Take
2633         a poly_int64 instead of a HOST_WIDE_INT.
2634         (build_int_cstu, build_array_type_nelts): Take a poly_uint64
2635         instead of an unsigned HOST_WIDE_INT.
2636         (wide_int_to_tree): Rename to...
2637         (wide_int_to_tree_1): ...this.
2638         (build_new_poly_int_cst, build_poly_int_cst): New functions.
2639         (force_fit_type): Take a poly_wide_int_ref instead of a wide_int_ref.
2640         (wide_int_to_tree): New function that takes a poly_wide_int_ref.
2641         (ptrdiff_tree_p, tree_to_poly_int64, tree_to_poly_uint64)
2642         (tree_fits_poly_int64_p, tree_fits_poly_uint64_p): New functions.
2643         * lto-streamer-out.c (DFS::DFS_write_tree_body, hash_tree): Handle
2644         TS_POLY_INT_CST.
2645         * tree-streamer-in.c (lto_input_ts_poly_tree_pointers): Likewise.
2646         (streamer_read_tree_body): Likewise.
2647         * tree-streamer-out.c (write_ts_poly_tree_pointers): Likewise.
2648         (streamer_write_tree_body): Likewise.
2649         * tree-streamer.c (streamer_check_handled_ts_structures): Likewise.
2650         * asan.c (asan_protect_global): Require the size to be an INTEGER_CST.
2651         * cfgexpand.c (expand_debug_expr): Handle POLY_INT_CST.
2652         * expr.c (expand_expr_real_1, const_vector_from_tree): Likewise.
2653         * gimple-expr.h (is_gimple_constant): Likewise.
2654         * gimplify.c (maybe_with_size_expr): Likewise.
2655         * print-tree.c (print_node): Likewise.
2656         * tree-data-ref.c (data_ref_compare_tree): Likewise.
2657         * tree-pretty-print.c (dump_generic_node): Likewise.
2658         * tree-ssa-address.c (addr_for_mem_ref): Likewise.
2659         * tree-vect-data-refs.c (dr_group_sort_cmp): Likewise.
2660         * tree-vrp.c (compare_values_warnv): Likewise.
2661         * tree-ssa-loop-ivopts.c (determine_base_object, constant_multiple_of)
2662         (get_loop_invariant_expr, add_candidate_1, get_computation_aff_1)
2663         (force_expr_to_var_cost): Likewise.
2664         * tree-ssa-loop.c (for_each_index): Likewise.
2665         * fold-const.h (build_invariant_address, size_int_kind): Take a
2666         poly_int64 instead of a HOST_WIDE_INT.
2667         * fold-const.c (fold_negate_expr_1, const_binop, const_unop)
2668         (fold_convert_const, multiple_of_p, fold_negate_const): Handle
2669         POLY_INT_CST.
2670         (size_binop_loc): Likewise.  Allow int_const_binop_1 to fail.
2671         (int_const_binop_2): New function, split out from...
2672         (int_const_binop_1): ...here.  Handle POLY_INT_CST.
2673         (size_int_kind): Take a poly_int64 instead of a HOST_WIDE_INT.
2674         * expmed.c (make_tree): Handle CONST_POLY_INT_P.
2675         * gimple-ssa-strength-reduction.c (slsr_process_add)
2676         (slsr_process_mul): Check for INTEGER_CSTs before using them
2677         as candidates.
2678         * stor-layout.c (bits_from_bytes): New function.
2679         (bit_from_pos): Use it.
2680         (layout_type): Likewise.  For vectors, multiply the TYPE_SIZE_UNIT
2681         by BITS_PER_UNIT to get the TYPE_SIZE.
2682         * tree-cfg.c (verify_expr, verify_types_in_gimple_reference): Allow
2683         MEM_REF and TARGET_MEM_REF offsets to be a POLY_INT_CST.
2685 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2686             Alan Hayward  <alan.hayward@arm.com>
2687             David Sherwood  <david.sherwood@arm.com>
2689         * doc/rtl.texi (const_poly_int): Document.  Also document the
2690         rtl sharing behavior.
2691         * gengenrtl.c (excluded_rtx): Return true for CONST_POLY_INT.
2692         * rtl.h (const_poly_int_def): New struct.
2693         (rtx_def::u): Add a cpi field.
2694         (CASE_CONST_UNIQUE, CASE_CONST_ANY): Add CONST_POLY_INT.
2695         (CONST_POLY_INT_P, CONST_POLY_INT_COEFFS): New macros.
2696         (wi::rtx_to_poly_wide_ref): New typedef
2697         (const_poly_int_value, wi::to_poly_wide, rtx_to_poly_int64)
2698         (poly_int_rtx_p): New functions.
2699         (trunc_int_for_mode): Declare a poly_int64 version.
2700         (plus_constant): Take a poly_int64 instead of a HOST_WIDE_INT.
2701         (immed_wide_int_const): Take a poly_wide_int_ref rather than
2702         a wide_int_ref.
2703         (strip_offset): Declare.
2704         (strip_offset_and_add): New function.
2705         * rtl.def (CONST_POLY_INT): New rtx code.
2706         * rtl.c (rtx_size): Handle CONST_POLY_INT.
2707         (shared_const_p): Use poly_int_rtx_p.
2708         * emit-rtl.h (gen_int_mode): Take a poly_int64 instead of a
2709         HOST_WIDE_INT.
2710         (gen_int_shift_amount): Likewise.
2711         * emit-rtl.c (const_poly_int_hasher): New class.
2712         (const_poly_int_htab): New variable.
2713         (init_emit_once): Initialize it when NUM_POLY_INT_COEFFS > 1.
2714         (const_poly_int_hasher::hash): New function.
2715         (const_poly_int_hasher::equal): Likewise.
2716         (gen_int_mode): Take a poly_int64 instead of a HOST_WIDE_INT.
2717         (immed_wide_int_const): Rename to...
2718         (immed_wide_int_const_1): ...this and make static.
2719         (immed_wide_int_const): New function, taking a poly_wide_int_ref
2720         instead of a wide_int_ref.
2721         (gen_int_shift_amount): Take a poly_int64 instead of a HOST_WIDE_INT.
2722         (gen_lowpart_common): Handle CONST_POLY_INT.
2723         * cse.c (hash_rtx_cb, equiv_constant): Likewise.
2724         * cselib.c (cselib_hash_rtx): Likewise.
2725         * dwarf2out.c (const_ok_for_output_1): Likewise.
2726         * expr.c (convert_modes): Likewise.
2727         * print-rtl.c (rtx_writer::print_rtx, print_value): Likewise.
2728         * rtlhash.c (add_rtx): Likewise.
2729         * explow.c (trunc_int_for_mode): Add a poly_int64 version.
2730         (plus_constant): Take a poly_int64 instead of a HOST_WIDE_INT.
2731         Handle existing CONST_POLY_INT rtxes.
2732         * expmed.h (expand_shift): Take a poly_int64 instead of a
2733         HOST_WIDE_INT.
2734         * expmed.c (expand_shift): Likewise.
2735         * rtlanal.c (strip_offset): New function.
2736         (commutative_operand_precedence): Give CONST_POLY_INT the same
2737         precedence as CONST_DOUBLE and put CONST_WIDE_INT between that
2738         and CONST_INT.
2739         * rtl-tests.c (const_poly_int_tests): New struct.
2740         (rtl_tests_c_tests): Use it.
2741         * simplify-rtx.c (simplify_const_unary_operation): Handle
2742         CONST_POLY_INT.
2743         (simplify_const_binary_operation): Likewise.
2744         (simplify_binary_operation_1): Fold additions of symbolic constants
2745         and CONST_POLY_INTs.
2746         (simplify_subreg): Handle extensions and truncations of
2747         CONST_POLY_INTs.
2748         (simplify_const_poly_int_tests): New struct.
2749         (simplify_rtx_c_tests): Use it.
2750         * wide-int.h (storage_ref): Add default constructor.
2751         (wide_int_ref_storage): Likewise.
2752         (trailing_wide_ints): Use GTY((user)).
2753         (trailing_wide_ints::operator[]): Add a const version.
2754         (trailing_wide_ints::get_precision): New function.
2755         (trailing_wide_ints::extra_size): Likewise.
2757 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2758             Alan Hayward  <alan.hayward@arm.com>
2759             David Sherwood  <david.sherwood@arm.com>
2761         * emit-rtl.h (gen_int_shift_amount): Declare.
2762         * emit-rtl.c (gen_int_shift_amount): New function.
2763         * asan.c (asan_emit_stack_protection): Use gen_int_shift_amount
2764         instead of GEN_INT.
2765         * calls.c (shift_return_value): Likewise.
2766         * cse.c (fold_rtx): Likewise.
2767         * dse.c (find_shift_sequence): Likewise.
2768         * expmed.c (init_expmed_one_mode, store_bit_field_1, expand_shift_1)
2769         (expand_shift, expand_smod_pow2): Likewise.
2770         * lower-subreg.c (shift_cost): Likewise.
2771         * optabs.c (expand_superword_shift, expand_doubleword_mult)
2772         (expand_unop, expand_binop, shift_amt_for_vec_perm_mask)
2773         (expand_vec_perm_var): Likewise.
2774         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
2775         (simplify_binary_operation_1): Likewise.
2776         * combine.c (try_combine, find_split_point, force_int_to_mode)
2777         (simplify_shift_const_1, simplify_shift_const): Likewise.
2778         (change_zero_ext): Likewise.  Use simplify_gen_binary.
2780 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2782         * poly-int.h (multiple_p): Fix handling of two non-poly_ints.
2784 2017-12-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2786         * doc/invoke.texi (ARM Options): Document accepted extension options
2787         for -march=armv8.3-a.
2789 2017-12-20  Richard Earnshaw  <rearnsha@arm.com>
2791         PR target/83105
2792         * config.gcc (arm*-*-linux*): When configured with --with-float=hard
2793         or --with-float=softfp, set the default CPU to arm10e.
2795 2017-12-20  Eric Botcazou  <ebotcazou@adacore.com>
2797         * config/visium/constraints.md (J, K, L): Use IN_RANGE macro.
2798         * config/visium/predicates.md (const_shift_operand): Likewise.
2799         * config/visium/visium.c (visium_legitimize_address): Fix oversight.
2800         (visium_legitimize_reload_address): Likewise.
2802 2017-12-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2804         PR target/82975
2805         * config/arm/arm.h (TEST_REGNO): Adjust comment as expected in r255830.
2807 2017-12-20  Jakub Jelinek  <jakub@redhat.com>
2809         PR c++/83490
2810         * calls.c (compute_argument_addresses): Ignore TYPE_EMPTY_P arguments.
2812 2017-12-20  Julia Koval  <julia.koval@intel.com>
2814         * common/config/i386/i386-common.c (OPTION_MASK_ISA_VPCLMULQDQ_SET,
2815         OPTION_MASK_ISA_VPCLMULQDQ_UNSET): New.
2816         (ix86_handle_option): Handle -mvpclmulqdq, move cx6 to flags2.
2817         * config.gcc: Include vpclmulqdqintrin.h.
2818         * config/i386/cpuid.h: Handle bit_VPCLMULQDQ.
2819         * config/i386/driver-i386.c (host_detect_local_cpu): Handle
2820         -mvpclmulqdq.
2821         * config/i386/i386-builtin.def (__builtin_ia32_vpclmulqdq_v2di,
2822         __builtin_ia32_vpclmulqdq_v4di, __builtin_ia32_vpclmulqdq_v8di): New.
2823         * config/i386/i386-c.c (__VPCLMULQDQ__): New.
2824         * config/i386/i386.c (isa2_opts): Add -mcx16.
2825         (isa_opts): Add -mpclmulqdq, remove -mcx16.
2826         (ix86_option_override_internal): Move mcx16 to flags2.
2827         (ix86_valid_target_attribute_inner_p): Add vpclmulqdq.
2828         (ix86_expand_builtin): Handle OPTION_MASK_ISA_VPCLMULQDQ.
2829         * config/i386/i386.h (TARGET_VPCLMULQDQ, TARGET_VPCLMULQDQ_P): New.
2830         * config/i386/i386.opt: Add mvpclmulqdq, move mcx16 to flags2.
2831         * config/i386/immintrin.h: Include vpclmulqdqintrin.h.
2832         * config/i386/sse.md (vpclmulqdq_<mode>): New pattern.
2833         * config/i386/vpclmulqdqintrin.h (_mm512_clmulepi64_epi128,
2834         _mm_clmulepi64_epi128, _mm256_clmulepi64_epi128): New intrinsics.
2835         * doc/invoke.texi: Add -mvpclmulqdq.
2837 2017-12-20  Tom de Vries  <tom@codesourcery.com>
2839         PR middle-end/83423
2840         * config/i386/i386.c (ix86_static_chain): Move
2841         DECL_STATIC_CHAIN test ...
2842         * calls.c (rtx_for_static_chain): ... here.  New function.
2843         * calls.h (rtx_for_static_chain): Declare.
2844         * builtins.c (expand_builtin_setjmp_receiver): Use rtx_for_static_chain
2845         instead of targetm.calls.static_chain.
2846         * df-scan.c (df_get_entry_block_def_set): Same.
2848 2017-12-19  Tom de Vries  <tom@codesourcery.com>
2850         PR tree-optimization/83493
2851         * graphite-isl-ast-to-gimple.c (translate_isl_ast_node_for): Unshare ub
2852         and lb.
2854 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2856         * gimple-ssa-sprintf.c (format_directive): Use inform_n instead of
2857         inform with hardcoded english plural handling.
2859 2017-12-18  Jeff Law  <law@redhat.com>
2861         PR tree-optimization/83477
2862         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis): For
2863         a non-virtual PHI, always push a new range.
2865 2017-12-19  Martin Sebor  <msebor@redhat.com>
2867         PR middle-end/77608
2868         * builtins.c (compute_objsize): Handle non-constant offsets.
2870 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2872         PR tree-optimization/83444
2873         * tree-ssa-strlen.c (strlen_check_and_optimize_stmt): For the
2874         character load case, if get_stridx on MEM_REF's operand doesn't
2875         look usable, retry with get_addr_stridx.
2877 2017-12-19  Alexandre Oliva  <aoliva@redhat.com>
2879         PR debug/83422
2880         * var-tracking.c (vt_debug_insns_local): Do not drop markers.
2881         (variable_tracking_main_1): Keep markers even when VTA fails.
2883         PR bootstrap/83396
2884         * cfgexpand.c (expand_gimple_basic_block): Expand label first,
2885         even if there are markers before it.
2886         * cfgrtl.c (rtl_verify_bb_layout): Reject DEBUG_INSNs outside BBs.
2888 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2890         * read-rtl.c (parse_reg_note_name): Replace Yoda conditions with
2891         typical order conditions.
2892         * sel-sched.c (extract_new_fences_from): Likewise.
2893         * config/visium/constraints.md (J, K, L): Likewise.
2894         * config/visium/predicates.md (const_shift_operand): Likewise.
2895         * config/visium/visium.c (visium_legitimize_address,
2896         visium_legitimize_reload_address): Likewise.
2897         * config/m68k/m68k.c (output_reg_adjust, emit_reg_adjust): Likewise.
2898         * config/arm/arm.c (arm_block_move_unaligned_straight): Likewise.
2899         * config/avr/constraints.md (Y01, Ym1, Y02, Ym2): Likewise.
2900         * config/avr/avr-log.c (avr_vdump, avr_log_set_avr_log,
2901         SET_DUMP_DETAIL): Likewise.
2902         * config/avr/predicates.md (const_8_16_24_operand): Likewise.
2903         * config/avr/avr.c (STR_PREFIX_P, avr_popcount_each_byte,
2904         avr_is_casesi_sequence, avr_casei_sequence_check_operands,
2905         avr_set_core_architecture, avr_set_current_function,
2906         avr_legitimize_reload_address, avr_asm_len, avr_print_operand,
2907         output_movqi, output_movsisf, avr_out_plus, avr_out_bitop,
2908         avr_out_fract, avr_adjust_insn_length, avr_encode_section_info,
2909         avr_2word_insn_p, output_reload_in_const, avr_has_nibble_0xf,
2910         avr_map_decompose, avr_fold_builtin): Likewise.
2911         * config/avr/driver-avr.c (avr_devicespecs_file): Likewise.
2912         * config/avr/gen-avr-mmcu-specs.c (str_prefix_p, print_mcu): Likewise.
2913         * config/i386/i386.c (ix86_parse_stringop_strategy_string): Likewise.
2914         * config/m32c/m32c-pragma.c (m32c_pragma_memregs): Likewise.
2915         * config/m32c/m32c.c (m32c_conditional_register_usage,
2916         m32c_address_cost): Likewise.
2917         * config/m32c/predicates.md (shiftcount_operand,
2918         longshiftcount_operand): Likewise.
2919         * config/iq2000/iq2000.c (iq2000_expand_prologue): Likewise.
2920         * config/nios2/nios2.c (nios2_handle_custom_fpu_insn_option,
2921         can_use_cdx_ldstw): Likewise.
2922         * config/nios2/nios2.h (CDX_REG_P): Likewise.
2923         * config/cr16/cr16.h (RETURN_ADDR_RTX, REGNO_MODE_OK_FOR_BASE_P):
2924         Likewise.
2925         * config/cr16/cr16.md (*mov<mode>_double): Likewise.
2926         * config/cr16/cr16.c (cr16_create_dwarf_for_multi_push): Likewise.
2927         * config/h8300/h8300.c (h8300_rtx_costs, get_shift_alg): Likewise.
2928         * config/vax/constraints.md (U06, U08, U16, CN6, S08, S16): Likewise.
2929         * config/vax/vax.c (adjacent_operands_p): Likewise.
2930         * config/ft32/constraints.md (L, b, KA): Likewise.
2931         * config/ft32/ft32.c (ft32_load_immediate, ft32_expand_prologue):
2932         Likewise.
2933         * cfgexpand.c (expand_stack_alignment): Likewise.
2934         * gcse.c (insert_expr_in_table): Likewise.
2935         * print-rtl.c (rtx_writer::print_rtx_operand_codes_E_and_V): Likewise.
2936         * cgraphunit.c (cgraph_node::expand): Likewise.
2937         * ira-build.c (setup_min_max_allocno_live_range_point): Likewise.
2938         * emit-rtl.c (add_insn): Likewise.
2939         * input.c (dump_location_info): Likewise.
2940         * passes.c (NEXT_PASS): Likewise.
2941         * read-rtl-function.c (parse_note_insn_name,
2942         function_reader::read_rtx_operand_r, function_reader::parse_mem_expr):
2943         Likewise.
2944         * sched-rgn.c (sched_rgn_init): Likewise.
2945         * diagnostic-show-locus.c (layout::show_ruler): Likewise.
2946         * combine.c (find_split_point, simplify_if_then_else, force_to_mode,
2947         if_then_else_cond, simplify_shift_const_1, simplify_comparison):
2948         Likewise.
2949         * explow.c (eliminate_constant_term): Likewise.
2950         * final.c (leaf_renumber_regs_insn): Likewise.
2951         * cfgrtl.c (print_rtl_with_bb): Likewise.
2952         * genhooks.c (emit_init_macros): Likewise.
2953         * poly-int.h (maybe_ne, maybe_le, maybe_lt): Likewise.
2954         * tree-data-ref.c (conflict_fn): Likewise.
2955         * selftest.c (assert_streq): Likewise.
2956         * expr.c (store_constructor_field, expand_expr_real_1): Likewise.
2957         * fold-const.c (fold_range_test, extract_muldiv_1, fold_truth_andor,
2958         fold_binary_loc, multiple_of_p): Likewise.
2959         * reload.c (push_reload, find_equiv_reg): Likewise.
2960         * et-forest.c (et_nca, et_below): Likewise.
2961         * dbxout.c (dbxout_symbol_location): Likewise.
2962         * reorg.c (relax_delay_slots): Likewise.
2963         * dojump.c (do_compare_rtx_and_jump): Likewise.
2964         * gengtype-parse.c (type): Likewise.
2965         * simplify-rtx.c (simplify_gen_ternary, simplify_gen_relational,
2966         simplify_const_relational_operation): Likewise.
2967         * reload1.c (do_output_reload): Likewise.
2968         * dumpfile.c (get_dump_file_info_by_switch): Likewise.
2969         * gengtype.c (type_for_name): Likewise.
2970         * gimple-ssa-sprintf.c (format_directive): Likewise.
2972 2017-12-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2974         PR target/82975
2975         * config/arm/arm.h (TEST_REGNO): Check reg_renumber is set before
2976         accessing it.  Adjust comment.
2978 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2980         PR middle-end/81914
2981         * predict.c (zero_one_minusone): New function.
2982         (apply_return_prediction): Avoid return prediction for functions
2983         returning only -1, 0 and 1 values, unless they only return -1 and 0
2984         or 0 and 1.
2986 2017-12-19  Claudiu Zissulescu  <claziss@synopsys.com>
2988         * config/arc/arc.c (legitimate_scaled_address_p): Clean
2989         fall-through warning.
2990         (arc_compute_frame_size): Remove unused variables.
2991         (arc_print_operand): Fix fprintif format.
2992         (arc_can_follow_jump): Clean fall-through warning.
2994 2017-12-19  Marek Polacek  <polacek@redhat.com>
2996         PR c++/83489
2997         * config/i386/i386.c (init_cumulative_args): Don't check TYPE_EMPTY_P
2998         on an error node.
3000 2017-12-19  Claudiu Zissulescu  <claziss@synopsys.com>
3002         * config/arc/arc.c (overriderregs): New variable.
3003         (arc_override_options): Track fixed/call saved/call options.
3004         (arc_conditional_register_usage): Check against overrideregs
3005         variable whenever we change register properties.
3007 2017-12-19  Nathan Sidwell  <nathan@acm.org>
3009         * opts.c (finish_options): Don't prefix dump_base_name if it
3010         already contains directories.
3012 2017-12-19  Martin Liska  <mliska@suse.cz>
3014         PR rtl-optimization/82675
3015         * loop-unroll.c (unroll_loop_constant_iterations): Allocate one
3016         more element in sbitmap.
3018 2017-12-19  Paolo Carlini  <paolo.carlini@oracle.com>
3020         * gimplify.c (gimplify_expr): Use error_operand_p.
3022 2017-12-19  Sebastian Huber  <sebastian.huber@embedded-brains.de>
3024         PR target/83387
3025         * config/rs6000/t-rtems (MULTILIB_REQUIRED): Remove 64-bit soft-float
3026         multilib.
3028 2017-12-19  Daniel Cederman  <cederman@gaisler.com>
3030         * config/sparc/sparc.c (sparc_do_work_around_errata): Make sure
3031         the jump is to a label.
3033 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
3035         PR tree-optimization/83444
3036         * tree-ssa-strlen.c (strlen_check_and_optimize_stmt): Optimize
3037         character loads.
3039         PR ipa/82801
3040         PR ipa/83346
3041         * ipa-inline.c (flatten_remove_node_hook): New function.
3042         (ipa_inline): Keep only nodes with flatten attribute at the end of
3043         the array in the order from ipa_reverse_postorder, only walk that
3044         portion of array for flattening, if there is more than one such
3045         node, temporarily register a removal hook and ignore removed nodes.
3047         PR tree-optimization/80631
3048         * tree-vect-loop.c (vect_create_epilog_for_reduction): Compare
3049         induc_code against MAX_EXPR or MIN_EXPR instead of reduc_fn against
3050         IFN_REDUC_MAX or IFN_REDUC_MIN.
3052 2017-12-18  Jeff Law  <law@redhat.com>
3054         * tree-ssa-dom.c (record_equivalences_from_phis): Do not
3055         record symbolic equivalences from backedges in the CFG.
3057         Revert
3058         2017-11-19  Jeff Law  <law@redhat.com>
3060         * tree-ssa-dom.c (record_equivalences_from_phis): Fix handling
3061         of degenerates resulting from ignoring an edge.
3063 2017-12-18  Martin Sebor  <msebor@redhat.com>
3065         PR middle-end/83373
3066         PR tree-optimization/78450
3067         * tree-ssa-strlen.c (maybe_set_strlen_range): New function.
3068         (handle_builtin_strlen): Call it.
3070 2017-12-18  Segher Boessenkool  <segher@kernel.crashing.org>
3072         PR rtl-optimization/83424
3073         * rtlanal.c (dead_or_set_regno_p): Handle CLOBBER just like SET.
3075 2017-12-18  Marek Polacek  <polacek@redhat.com>
3077         PR middle-end/83463
3078         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
3079         Check if TYPE is INTEGRAL_TYPE_P before accessing its min/max
3080         values.
3082 2017-12-18  Claudiu Zissulescu  <claziss@synopsys.com>
3084         * config/arc/arc.md (maddsidi4, maddsidi4_split): Update pattern.
3085         (umaddsidi4, umaddsidi_split): Likewise.
3087 2017-12-18  Claudiu Zissulescu  <claziss@synopsys.com>
3089         * config/arc/arc.c (arc_legitimate_constant_p): Always check all
3090         constants.
3092 2017-12-18  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
3094         PR target/83420
3095         * config/s390/s390.c (s390_option_override): Avoid strncpy.
3097 2017-12-18  Richard Biener  <rguenther@suse.de>
3099         PR tree-optimization/81877
3100         * tree-ssa-loop-im.c (ref_indep_loop_p): Remove safelen parameters.
3101         (outermost_indep_loop): Adjust.
3102         (ref_indep_loop_p_1): Likewise.  Remove safelen handling again.
3103         (can_sm_ref_p): Adjust.
3105 2017-12-18  Richard Biener  <rguenther@suse.de>
3107         PR middle-end/77291
3108         * tree.c (array_at_struct_end_p): Return true if the underlying
3109         object has space for at least one element in excess of what
3110         the array domain specifies.
3112 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
3114         * doc/extend.texi (x86 Function Attributes): Reformat nocf_check
3115         example to avoid overfull hbox.
3116         * doc/invoke.texi (Option Summary): Add missing @gol.
3117         (C++ Dialect Options): Reformat -Wnoexcept-type example to avoid
3118         overfull hbox.
3120 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
3121             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3123         * doc/invoke.texi (Option Summary): Add -mverbose-cost-dump
3124         to AArch64 and ARM lists, plus missing -mflip-thumb for ARM.
3125         (AArch64 Options): Document -mverbose-cost-dump.
3126         (ARM Options): Likewise, plus -mflip-thumb.
3128 2017-12-17  Martin Sebor  <msebor@redhat.com>
3130         PR bootstrap/83446
3131         * gimple-ssa-warn-restrict.c
3132         (builtin_memref::offset_out_of_bounds): Correct the handling of
3133         anti-ranges.
3135 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
3137         * doc/invoke.texi (ARC Options): Add missing -mlra entry.
3139 2017-12-17  John David Anglin  <danglin@gcc.gnu.org>
3141         * config/pa/pa.c (pa_som_asm_init_sections): Fix comment.
3143 2017-12-17  Markus Trippelsdorf  <markus@trippelsdorf.de>
3145         * x86-tune-costs.h (skylake_cost, core_cost): Decrease r64 multiply
3146         latencies.
3148 2017-12-16  Sandra Loosemore  <sandra@codesourcery.com>
3150         * doc/invoke.texi: Fix some typos.
3152 2017-12-16  Martin Sebor  <msebor@redhat.com>
3154         PR tree-optimization/78918
3155         * Makefile.in (OBJS): Add gimple-ssa-warn-restrict.o.
3156         * builtins.c (check_sizes): Rename...
3157         (check_access): ...to this.  Rename function arguments for clarity.
3158         (check_memop_sizes): Adjust names.
3159         (expand_builtin_memchr, expand_builtin_memcpy): Same.
3160         (expand_builtin_memmove, expand_builtin_mempcpy): Same.
3161         (expand_builtin_strcat, expand_builtin_stpncpy): Same.
3162         (check_strncat_sizes, expand_builtin_strncat): Same.
3163         (expand_builtin_strncpy, expand_builtin_memset): Same.
3164         (expand_builtin_bzero, expand_builtin_memcmp): Same.
3165         (expand_builtin_memory_chk, maybe_emit_chk_warning): Same.
3166         (maybe_emit_sprintf_chk_warning): Same.
3167         (expand_builtin_strcpy): Adjust.
3168         (expand_builtin_stpcpy): Same.
3169         (expand_builtin_with_bounds): Detect out-of-bounds accesses
3170         in pointer-checking forms of memcpy, memmove, and mempcpy.
3171         (gcall_to_tree_minimal, max_object_size): Define new functions.
3172         * builtins.h (max_object_size): Declare.
3173         * calls.c (alloc_max_size): Call max_object_size instead of
3174         hardcoding ssizetype limit.
3175         (get_size_range): Handle new argument.
3176         * calls.h (get_size_range): Add a new argument.
3177         * cfgexpand.c (expand_call_stmt): Propagate no-warning bit.
3178         * doc/invoke.texi (-Wrestrict): Adjust, add example.
3179         * gimple-fold.c (gimple_fold_builtin_memory_op): Detect overlapping
3180         operations.
3181         (gimple_fold_builtin_memory_chk): Same.
3182         (gimple_fold_builtin_stxcpy_chk): New function.
3183         * gimple-ssa-warn-restrict.c: New source.
3184         * gimple-ssa-warn-restrict.h: New header.
3185         * gimple.c (gimple_build_call_from_tree): Propagate location.
3186         * passes.def (pass_warn_restrict): Add new pass.
3187         * tree-pass.h (make_pass_warn_restrict): Declare.
3188         * tree-ssa-strlen.c (handle_builtin_strcpy): Detect overlapping
3189         operations.
3190         (handle_builtin_strcat): Same.
3191         (strlen_optimize_stmt): Rename...
3192         (strlen_check_and_optimize_stmt): ...to this.  Handle strncat,
3193         stpncpy, strncpy, and their checking forms.
3195 2017-12-16  Jan Hubicka  <hubicka@ucw.cz>
3197         PR rtl-optimization/82849
3198         * modulo-sched.c (sms_schedule): Use get_estimated_loop_iterations_int
3199         and get_max_loop_iterations_int.
3201 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3202             Alan Hayward  <alan.hayward@arm.com>
3203             David Sherwood  <david.sherwood@arm.com>
3205         * machmode.h (mode_for_size, int_mode_for_size, float_mode_for_size)
3206         (smallest_mode_for_size, smallest_int_mode_for_size): Take the mode
3207         size as a poly_uint64.
3208         (mode_for_vector, mode_for_int_vector): Take the number of vector
3209         elements as a poly_uint64.
3210         * stor-layout.c (mode_for_size, smallest_mode_for_size): Take the mode
3211         size as a poly_uint64.
3212         (mode_for_vector, mode_for_int_vector): Take the number of vector
3213         elements as a poly_uint64.
3215 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3216             Alan Hayward  <alan.hayward@arm.com>
3217             David Sherwood  <david.sherwood@arm.com>
3219         * machmode.h (MACRO_MODE): New macro.
3220         * addresses.h (base_reg_class, ok_for_base_p_1): Use it.
3221         * targhooks.c (default_libcall_value, default_secondary_reload)
3222         (default_memory_move_cost, default_register_move_cost)
3223         (default_class_max_nregs): Likewise.
3225 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3226             Alan Hayward  <alan.hayward@arm.com>
3227             David Sherwood  <david.sherwood@arm.com>
3229         * doc/sourcebuild.texi: Document IN_TARGET_CODE.
3230         * genattrtab.c (write_header): Define IN_TARGET_CODE to 1 in the
3231         target C file.
3232         * genautomata.c (main): Likewise.
3233         * genconditions.c (write_header): Likewise.
3234         * genemit.c (main): Likewise.
3235         * genextract.c (print_header): Likewise.
3236         * genopinit.c (main): Likewise.
3237         * genoutput.c (output_prologue): Likewise.
3238         * genpeep.c (main): Likewise.
3239         * genpreds.c (write_insn_preds_c): Likewise.
3240         * genrecog.c (writer_header): Likewise.
3241         * config/aarch64/aarch64-builtins.c (IN_TARGET_CODE): Define.
3242         * config/aarch64/aarch64-c.c (IN_TARGET_CODE): Likewise.
3243         * config/aarch64/aarch64.c (IN_TARGET_CODE): Likewise.
3244         * config/aarch64/cortex-a57-fma-steering.c (IN_TARGET_CODE): Likewise.
3245         * config/aarch64/driver-aarch64.c (IN_TARGET_CODE): Likewise.
3246         * config/alpha/alpha.c (IN_TARGET_CODE): Likewise.
3247         * config/alpha/driver-alpha.c (IN_TARGET_CODE): Likewise.
3248         * config/arc/arc-c.c (IN_TARGET_CODE): Likewise.
3249         * config/arc/arc.c (IN_TARGET_CODE): Likewise.
3250         * config/arc/driver-arc.c (IN_TARGET_CODE): Likewise.
3251         * config/arm/aarch-common.c (IN_TARGET_CODE): Likewise.
3252         * config/arm/arm-builtins.c (IN_TARGET_CODE): Likewise.
3253         * config/arm/arm-c.c (IN_TARGET_CODE): Likewise.
3254         * config/arm/arm.c (IN_TARGET_CODE): Likewise.
3255         * config/arm/driver-arm.c (IN_TARGET_CODE): Likewise.
3256         * config/avr/avr-c.c (IN_TARGET_CODE): Likewise.
3257         * config/avr/avr-devices.c (IN_TARGET_CODE): Likewise.
3258         * config/avr/avr-log.c (IN_TARGET_CODE): Likewise.
3259         * config/avr/avr.c (IN_TARGET_CODE): Likewise.
3260         * config/avr/driver-avr.c (IN_TARGET_CODE): Likewise.
3261         * config/avr/gen-avr-mmcu-specs.c (IN_TARGET_CODE): Likewise.
3262         * config/bfin/bfin.c (IN_TARGET_CODE): Likewise.
3263         * config/c6x/c6x.c (IN_TARGET_CODE): Likewise.
3264         * config/cr16/cr16.c (IN_TARGET_CODE): Likewise.
3265         * config/cris/cris.c (IN_TARGET_CODE): Likewise.
3266         * config/darwin.c (IN_TARGET_CODE): Likewise.
3267         * config/epiphany/epiphany.c (IN_TARGET_CODE): Likewise.
3268         * config/epiphany/mode-switch-use.c (IN_TARGET_CODE): Likewise.
3269         * config/epiphany/resolve-sw-modes.c (IN_TARGET_CODE): Likewise.
3270         * config/fr30/fr30.c (IN_TARGET_CODE): Likewise.
3271         * config/frv/frv.c (IN_TARGET_CODE): Likewise.
3272         * config/ft32/ft32.c (IN_TARGET_CODE): Likewise.
3273         * config/h8300/h8300.c (IN_TARGET_CODE): Likewise.
3274         * config/i386/djgpp.c (IN_TARGET_CODE): Likewise.
3275         * config/i386/driver-i386.c (IN_TARGET_CODE): Likewise.
3276         * config/i386/driver-mingw32.c (IN_TARGET_CODE): Likewise.
3277         * config/i386/host-cygwin.c (IN_TARGET_CODE): Likewise.
3278         * config/i386/host-i386-darwin.c (IN_TARGET_CODE): Likewise.
3279         * config/i386/host-mingw32.c (IN_TARGET_CODE): Likewise.
3280         * config/i386/i386-c.c (IN_TARGET_CODE): Likewise.
3281         * config/i386/i386.c (IN_TARGET_CODE): Likewise.
3282         * config/i386/intelmic-mkoffload.c (IN_TARGET_CODE): Likewise.
3283         * config/i386/msformat-c.c (IN_TARGET_CODE): Likewise.
3284         * config/i386/winnt-cxx.c (IN_TARGET_CODE): Likewise.
3285         * config/i386/winnt-stubs.c (IN_TARGET_CODE): Likewise.
3286         * config/i386/winnt.c (IN_TARGET_CODE): Likewise.
3287         * config/i386/x86-tune-sched-atom.c (IN_TARGET_CODE): Likewise.
3288         * config/i386/x86-tune-sched-bd.c (IN_TARGET_CODE): Likewise.
3289         * config/i386/x86-tune-sched-core.c (IN_TARGET_CODE): Likewise.
3290         * config/i386/x86-tune-sched.c (IN_TARGET_CODE): Likewise.
3291         * config/ia64/ia64-c.c (IN_TARGET_CODE): Likewise.
3292         * config/ia64/ia64.c (IN_TARGET_CODE): Likewise.
3293         * config/iq2000/iq2000.c (IN_TARGET_CODE): Likewise.
3294         * config/lm32/lm32.c (IN_TARGET_CODE): Likewise.
3295         * config/m32c/m32c-pragma.c (IN_TARGET_CODE): Likewise.
3296         * config/m32c/m32c.c (IN_TARGET_CODE): Likewise.
3297         * config/m32r/m32r.c (IN_TARGET_CODE): Likewise.
3298         * config/m68k/m68k.c (IN_TARGET_CODE): Likewise.
3299         * config/mcore/mcore.c (IN_TARGET_CODE): Likewise.
3300         * config/microblaze/microblaze-c.c (IN_TARGET_CODE): Likewise.
3301         * config/microblaze/microblaze.c (IN_TARGET_CODE): Likewise.
3302         * config/mips/driver-native.c (IN_TARGET_CODE): Likewise.
3303         * config/mips/frame-header-opt.c (IN_TARGET_CODE): Likewise.
3304         * config/mips/mips.c (IN_TARGET_CODE): Likewise.
3305         * config/mmix/mmix.c (IN_TARGET_CODE): Likewise.
3306         * config/mn10300/mn10300.c (IN_TARGET_CODE): Likewise.
3307         * config/moxie/moxie.c (IN_TARGET_CODE): Likewise.
3308         * config/msp430/driver-msp430.c (IN_TARGET_CODE): Likewise.
3309         * config/msp430/msp430-c.c (IN_TARGET_CODE): Likewise.
3310         * config/msp430/msp430.c (IN_TARGET_CODE): Likewise.
3311         * config/nds32/nds32-cost.c (IN_TARGET_CODE): Likewise.
3312         * config/nds32/nds32-fp-as-gp.c (IN_TARGET_CODE): Likewise.
3313         * config/nds32/nds32-intrinsic.c (IN_TARGET_CODE): Likewise.
3314         * config/nds32/nds32-isr.c (IN_TARGET_CODE): Likewise.
3315         * config/nds32/nds32-md-auxiliary.c (IN_TARGET_CODE): Likewise.
3316         * config/nds32/nds32-memory-manipulation.c (IN_TARGET_CODE): Likewise.
3317         * config/nds32/nds32-pipelines-auxiliary.c (IN_TARGET_CODE): Likewise.
3318         * config/nds32/nds32-predicates.c (IN_TARGET_CODE): Likewise.
3319         * config/nds32/nds32.c (IN_TARGET_CODE): Likewise.
3320         * config/nios2/nios2.c (IN_TARGET_CODE): Likewise.
3321         * config/nvptx/mkoffload.c (IN_TARGET_CODE): Likewise.
3322         * config/nvptx/nvptx.c (IN_TARGET_CODE): Likewise.
3323         * config/pa/pa.c (IN_TARGET_CODE): Likewise.
3324         * config/pdp11/pdp11.c (IN_TARGET_CODE): Likewise.
3325         * config/powerpcspe/driver-powerpcspe.c (IN_TARGET_CODE): Likewise.
3326         * config/powerpcspe/host-darwin.c (IN_TARGET_CODE): Likewise.
3327         * config/powerpcspe/host-ppc64-darwin.c (IN_TARGET_CODE): Likewise.
3328         * config/powerpcspe/powerpcspe-c.c (IN_TARGET_CODE): Likewise.
3329         * config/powerpcspe/powerpcspe-linux.c (IN_TARGET_CODE): Likewise.
3330         * config/powerpcspe/powerpcspe.c (IN_TARGET_CODE): Likewise.
3331         * config/riscv/riscv-builtins.c (IN_TARGET_CODE): Likewise.
3332         * config/riscv/riscv-c.c (IN_TARGET_CODE): Likewise.
3333         * config/riscv/riscv.c (IN_TARGET_CODE): Likewise.
3334         * config/rl78/rl78-c.c (IN_TARGET_CODE): Likewise.
3335         * config/rl78/rl78.c (IN_TARGET_CODE): Likewise.
3336         * config/rs6000/driver-rs6000.c (IN_TARGET_CODE): Likewise.
3337         * config/rs6000/host-darwin.c (IN_TARGET_CODE): Likewise.
3338         * config/rs6000/host-ppc64-darwin.c (IN_TARGET_CODE): Likewise.
3339         * config/rs6000/rs6000-c.c (IN_TARGET_CODE): Likewise.
3340         * config/rs6000/rs6000-linux.c (IN_TARGET_CODE): Likewise.
3341         * config/rs6000/rs6000-p8swap.c (IN_TARGET_CODE): Likewise.
3342         * config/rs6000/rs6000-string.c (IN_TARGET_CODE): Likewise.
3343         * config/rs6000/rs6000.c (IN_TARGET_CODE): Likewise.
3344         * config/rx/rx.c (IN_TARGET_CODE): Likewise.
3345         * config/s390/driver-native.c (IN_TARGET_CODE): Likewise.
3346         * config/s390/s390-c.c (IN_TARGET_CODE): Likewise.
3347         * config/s390/s390.c (IN_TARGET_CODE): Likewise.
3348         * config/sh/sh-c.c (IN_TARGET_CODE): Likewise.
3349         * config/sh/sh-mem.cc (IN_TARGET_CODE): Likewise.
3350         * config/sh/sh.c (IN_TARGET_CODE): Likewise.
3351         * config/sh/sh_optimize_sett_clrt.cc (IN_TARGET_CODE): Likewise.
3352         * config/sh/sh_treg_combine.cc (IN_TARGET_CODE): Likewise.
3353         * config/sparc/driver-sparc.c (IN_TARGET_CODE): Likewise.
3354         * config/sparc/sparc-c.c (IN_TARGET_CODE): Likewise.
3355         * config/sparc/sparc.c (IN_TARGET_CODE): Likewise.
3356         * config/spu/spu-c.c (IN_TARGET_CODE): Likewise.
3357         * config/spu/spu.c (IN_TARGET_CODE): Likewise.
3358         * config/stormy16/stormy16.c (IN_TARGET_CODE): Likewise.
3359         * config/tilegx/mul-tables.c (IN_TARGET_CODE): Likewise.
3360         * config/tilegx/tilegx-c.c (IN_TARGET_CODE): Likewise.
3361         * config/tilegx/tilegx.c (IN_TARGET_CODE): Likewise.
3362         * config/tilepro/mul-tables.c (IN_TARGET_CODE): Likewise.
3363         * config/tilepro/tilepro-c.c (IN_TARGET_CODE): Likewise.
3364         * config/tilepro/tilepro.c (IN_TARGET_CODE): Likewise.
3365         * config/v850/v850-c.c (IN_TARGET_CODE): Likewise.
3366         * config/v850/v850.c (IN_TARGET_CODE): Likewise.
3367         * config/vax/vax.c (IN_TARGET_CODE): Likewise.
3368         * config/visium/visium.c (IN_TARGET_CODE): Likewise.
3369         * config/vms/vms-c.c (IN_TARGET_CODE): Likewise.
3370         * config/vms/vms-f.c (IN_TARGET_CODE): Likewise.
3371         * config/vms/vms.c (IN_TARGET_CODE): Likewise.
3372         * config/xtensa/xtensa.c (IN_TARGET_CODE): Likewise.
3374 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3375             Alan Hayward  <alan.hayward@arm.com>
3376             David Sherwood  <david.sherwood@arm.com>
3378         * tree-ssa-sccvn.c (vn_reference_lookup_3): Avoid repeated
3379         checks for MEM_REF.
3381 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3382             Alan Hayward  <alan.hayward@arm.com>
3383             David Sherwood  <david.sherwood@arm.com>
3385         * doc/generic.texi (VEC_SERIES_EXPR): Document.
3386         * doc/md.texi (vec_series@var{m}): Document.
3387         * tree.def (VEC_SERIES_EXPR): New tree code.
3388         * tree.h (build_vec_series): Declare.
3389         * tree.c (build_vec_series): New function.
3390         * cfgexpand.c (expand_debug_expr): Handle VEC_SERIES_EXPR.
3391         * tree-pretty-print.c (dump_generic_node): Likewise.
3392         * gimple-pretty-print.c (dump_binary_rhs): Likewise.
3393         * tree-inline.c (estimate_operator_cost): Likewise.
3394         * expr.c (expand_expr_real_2): Likewise.
3395         * optabs-tree.c (optab_for_tree_code): Likewise.
3396         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
3397         * fold-const.c (const_binop): Fold VEC_SERIES_EXPRs of constants.
3398         * expmed.c (make_tree): Handle VEC_SERIES.
3399         * optabs.def (vec_series_optab): New optab.
3400         * optabs.h (expand_vec_series_expr): Declare.
3401         * optabs.c (expand_vec_series_expr): New function.
3402         * tree-vect-generic.c (expand_vector_operations_1): Check that
3403         the operands also have vector type.
3405 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3406             Alan Hayward  <alan.hawyard@arm.com>
3407             David Sherwood  <david.sherwood@arm.com>
3409         * doc/generic.texi (VEC_DUPLICATE_EXPR): Document.
3410         (VEC_COND_EXPR): Add missing @tindex.
3411         * doc/md.texi (vec_duplicate@var{m}): Document.
3412         * tree.def (VEC_DUPLICATE_EXPR): New tree codes.
3413         * tree.c (build_vector_from_val): Add stubbed-out handling of
3414         variable-length vectors, using VEC_DUPLICATE_EXPR.
3415         (uniform_vector_p): Handle VEC_DUPLICATE_EXPR.
3416         * cfgexpand.c (expand_debug_expr): Likewise.
3417         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
3418         * tree-inline.c (estimate_operator_cost): Likewise.
3419         * tree-pretty-print.c (dump_generic_node): Likewise.
3420         * tree-vect-generic.c (ssa_uniform_vector_p): Likewise.
3421         * fold-const.c (const_unop): Fold VEC_DUPLICATE_EXPRs of a constant.
3422         (test_vec_duplicate_folding): New function.
3423         (fold_const_c_tests): Call it.
3424         * optabs.def (vec_duplicate_optab): New optab.
3425         * optabs-tree.c (optab_for_tree_code): Handle VEC_DUPLICATE_EXPR.
3426         * optabs.h (expand_vector_broadcast): Declare.
3427         * optabs.c (expand_vector_broadcast): Make non-static.  Try using
3428         vec_duplicate_optab.
3429         * expr.c (store_constructor): Try using vec_duplicate_optab for
3430         uniform vectors.
3431         (expand_expr_real_2): Handle VEC_DUPLICATE_EXPR.
3433 2017-12-15  Markus Trippelsdorf  <markus@trippelsdorf.de>
3435         PR target/83358
3436         * config/i386/x86-tune-costs.h (skylake_cost, core_cost): Increase
3437         div/mod latencies a bit.
3439 2017-12-15  Jeff Law  <law@redhat.com>
3441         PR tree-optimization/36550
3442         * tree-ssa-threadupdate.c (count_stmts_and_phis_in_block): New.
3443         (mark_threaded_blocks): Rewrite code to avoid block copying when
3444         optimizing for size.  Don't pessimize blocks which will be
3445         copied, but all the statements will be dead.
3447 2017-12-15  Alexandre Oliva  <aoliva@redhat.com>
3449         PR tree-optimization/81165
3450         * tree-ssa-threadupdate.c (uses_in_bb): New.
3451         (estimate_threading_killed_stmts): New.
3452         * tree-ssa-threadupdate.h (estimate_threading_killed_stmts): Prototype.
3453         * tree-ssa-threadedge.c
3454         (record_temporary_equivalences_from_stmts_at_dest): Expand limit
3455         when its hit.
3457 2017-12-15  Jeff Law  <law@redhat.com>
3459         PR tree-optimization/83410
3460         * tree-ssa-threadupdate.c (thread_block_1): Avoid certain jump
3461         threads when parallelizing loops.
3463 2017-12-15  Jakub Jelinek  <jakub@redhat.com>
3465         * tree-core.h (struct attribute_spec): Swap affects_type_identity and
3466         handler fields.
3467         * config/alpha/alpha.c (vms_attribute_table): Swap
3468         affects_type_identity and handler fields, adjust comments.
3469         * config/mips/mips.c (mips_attribute_table): Likewise.
3470         * config/visium/visium.c (visium_attribute_table): Likewise.
3471         * config/epiphany/epiphany.c (epiphany_attribute_table): Likewise.
3472         * config/microblaze/microblaze.c (microblaze_attribute_table):
3473         Likewise.
3474         * config/spu/spu.c (spu_attribute_table): Likewise.
3475         * config/mcore/mcore.c (mcore_attribute_table): Likewise.
3476         * config/arc/arc.c (arc_attribute_table): Likewise.
3477         * config/m68k/m68k.c (m68k_attribute_table): Likewise.
3478         * config/v850/v850.c (v850_handle_interrupt_attribute,
3479         v850_handle_data_area_attribute): Formatting fixes.
3480         (v850_attribute_table): Swap affects_type_identity and handler
3481         fields, adjust comments.
3482         * config/m32r/m32r.c (m32r_attribute_table): Likewise.
3483         * config/arm/arm.c (arm_attribute_table): Likewise.
3484         * config/avr/avr.c (avr_attribute_table): Likewise.
3485         * config/s390/s390.c (s390_attribute_table): Likewise.
3486         * config/sh/sh.c (sh_attribute_table): Likewise.
3487         * config/i386/i386.c (ix86_handle_cconv_attribute,
3488         ix86_handle_callee_pop_aggregate_return): Formatting fixes.
3489         (ix86_attribute_table): Swap affects_type_identity and handler
3490         fields, adjust comments.
3491         * config/i386/cygming.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
3492         * config/sparc/sparc.c (sparc_attribute_table): Likewise.
3493         * config/m32c/m32c.c (m32c_attribute_table): Likewise.
3494         * config/sol2.h (SOLARIS_ATTRIBUTE_TABLE): Likewise.
3495         * config/ia64/ia64.c (ia64_attribute_table): Likewise.
3496         * config/msp430/msp430.c (msp430_attribute_table): Likewise.
3497         * config/rx/rx.c (rx_attribute_table): Likewise.
3498         * config/cr16/cr16.c (cr16_attribute_table): Likewise.
3499         * config/h8300/h8300.c (h8300_attribute_table): Likewise.
3500         * config/nvptx/nvptx.c (nvptx_attribute_table): Likewise.
3501         * config/powerpcspe/powerpcspe.c (rs6000_attribute_table): Likewise.
3502         * config/darwin.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
3503         * config/stormy16/stormy16.c (xstormy16_attribute_table): Likewise.
3504         * config/bfin/bfin.c (bfin_attribute_table): Likewise.
3505         * config/rs6000/rs6000.c (rs6000_attribute_table): Likewise.
3506         * config/rl78/rl78.c (rl78_attribute_table): Likewise.
3507         * config/nds32/nds32.c (nds32_attribute_table): Likewise.
3508         * doc/plugins.texi (user_attr): Likewise.  Add NULL for
3509         exclude.
3510         * attribs.c (empty_attribute_table): Swap affects_type_identity and
3511         handler fields.
3512         (register_scoped_attributes, decl_attributes): Formatting fixes.
3514         PR tree-optimization/83269
3515         * fold-const.c (fold_binary_loc): Perform (-A) - B -> (-B) - A
3516         subtraction in arg0's type if type is signed and arg0 is unsigned.
3517         Formatting fix.
3519         PR sanitizer/81281
3520         * match.pd ((T)(P + A) - (T)P -> (T) A): Use @@0 instead of @0 and
3521         convert? on @0 instead of convert.  Check type of @1, not @0.
3522         ((T)P - (T)(P + A) -> -(T) A): Use @@0 instead of @0 and
3523         convert? on @0 instead of convert.  Check type of @1, not @0.
3524         ((T)(P + A) - (T)(P + B) -> (T)A - (T)B): Use @@0 instead of @0,
3525         only optimize if either both @1 and @2 types are narrower
3526         precision, or both are wider or equal precision, and in the former
3527         case only if both have undefined overflow.
3529 2017-12-15  Richard Biener  <rguenther@suse.de>
3531         PR lto/83388
3532         * internal-fn.def (IFN_NOP): Add.
3533         * internal-fn.c (expand_NOP): Do nothing.
3534         * lto-streamer-in.c (input_function): Instead of removing
3535         sanitizer calls replace them with IFN_NOP calls.
3537 2017-12-15  Richard Sandiford  <richard.sandiford@linaro.org>
3538             Alan Hayward  <alan.hayward@arm.com>
3539             David Sherwood  <david.sherwood@arm.com>
3541         * dse.c (store_info, read_info_type): Replace begin and end with
3542         offset and width.
3543         (print_range): New function.
3544         (set_all_positions_unneeded, any_positions_needed_p)
3545         (check_mem_read_rtx, scan_stores, scan_reads, dse_step5): Update
3546         accordingly.
3547         (record_store): Likewise.  Optimize the case in which all positions
3548         are unneeded.
3549         (get_stored_val): Replace read_begin and read_end with read_offset
3550         and read_width.
3551         (replace_read): Update call accordingly.
3553 2017-12-15  Bin Cheng  <bin.cheng@arm.com>
3555         * gimple-loop-interchange.cc (STMT_COST_RATIO): New macro.
3556         (loop_cand::m_num_stmts, loop_cand::m_const_init_reduc): New members.
3557         (loop_cand::loop_cand): Initialize above members.
3558         (loop_cand::supported_operations): Delete.
3559         (loop_cand::can_interchange_p): Inline above function.
3560         (loop_cand::classify_simple_reduction): Record number of constant
3561         initialized simple reductions.
3562         (should_interchange_loops): New parameters.  Check stmt cost of loops
3563         to be interchange.
3564         (tree_loop_interchange::interchange): Prepare stmt cost of outer loop.
3565         Update call to should_interchange_loops.
3566         (should_interchange_loop_nest): Update call to
3567         should_interchange_loops.
3569 2017-12-15  Eric Botcazou  <ebotcazou@adacore.com>
3571         PR target/66488
3572         * ggc-page.c (HOST_BITS_PER_PTR): Do not define here...
3573         * hwint.h (HOST_BITS_PER_PTR): ...but here instead.
3574         * config/i386/xm-mingw32.h (HOST_BITS_PER_PTR): Delete.
3576 2017-12-15  Julia Koval  <julia.koval@intel.com>
3578         * config/i386/i386-builtin.def (__builtin_ia32_vaesenclast_v16qi,
3579         __builtin_ia32_vaesenclast_v32qi, __builtin_ia32_vaesenclast_v64qi):
3580         New.
3581         * config/i386/sse.md (vaesenclast_<mode>): New pattern.
3582         * config/i386/vaesintrin.h (_mm256_aesenclast_epi128,
3583         _mm512_aesenclast_epi128, _mm_aesenclast_epi128): New intrinsics.
3585 2017-12-15  Julia Koval  <julia.koval@intel.com>
3587         * config/i386/i386-builtin.def (__builtin_ia32_vaesenc_v16qi,
3588         __builtin_ia32_vaesenc_v32qi, __builtin_ia32_vaesenc_v64qi): New.
3589         * config/i386/sse.md (vaesenc_<mode>): New pattern.
3590         * config/i386/vaesintrin.h (_mm256_aesenc_epi128, _mm512_aesenc_epi128,
3591         _mm_aesenc_epi128): New intrinsics.
3593 2017-12-15  Julia Koval  <julia.koval@intel.com>
3595         * config/i386/i386-builtin.def (__builtin_ia32_vaesdeclast_v16qi,
3596         __builtin_ia32_vaesdeclast_v32qi, __builtin_ia32_vaesdeclast_v64qi):
3597         New.
3598         * config/i386/sse.md (vaesdeclast_<mode>): New pattern.
3599         * config/i386/vaesintrin.h (_mm256_aesdeclast_epi128,
3600         _mm512_aesdeclast_epi128, _mm_aesdeclast_epi128): New intrinsics.
3602 2017-12-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3604         * gimple-ssa-strength-reduction.c (analyze_increments):
3605         Distinguish replacement costs for constant strides from those for
3606         unknown strides.
3608 2017-12-14  Jakub Jelinek  <jakub@redhat.com>
3610         * var-tracking.c (variable_tracking_main_1): Formatting fix.
3612 2017-12-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3614         * doc/invoke.texi: Document -Wcast-function-type.
3615         * recog.h (stored_funcptr): Change signature.
3616         * tree-dump.c (dump_node): Avoid warning.
3617         * typed-splay-tree.h (typed_splay_tree): Avoid warning.
3619 2017-12-14  Qing Zhao  <qing.zhao@oracle.com>
3621         PR middle_end/79538
3622         * gimple-fold.c (get_range_strlen): Add the handling of non-member 
3623         array.
3625 2017-12-14  David Malcolm  <dmalcolm@redhat.com>
3627         PR tree-optimization/83312
3628         * domwalk.h (dom_walker::dom_walker): Fix typo in comment.
3629         * tree-cfg.c (find_taken_edge): Update to handle NULL_TREE for
3630         "val" param, and to cope with arbitrary basic blocks.
3631         (find_taken_edge_cond_expr): Add "cond_stmt" param and use it to
3632         handle NULL_TREE for "val", dropping "bb" param.
3633         (find_taken_edge_switch_expr): Make "switch_stmt" param const and
3634         drop "bb" param.  Handle NULL_TREE for "val".
3635         (find_case_label_for_value): Make "switch_stmt" param const.
3636         * tree-vrp.c (class check_array_bounds_dom_walker): New subclass
3637         of dom_walker.
3638         (vrp_prop::check_all_array_refs): Reimplement as...
3639         (check_array_bounds_dom_walker::before_dom_children): ...this new
3640         vfunc.  Replace linear search through BB block list, excluding
3641         those with non-executable in-edges via dominator walk.
3643 2017-12-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3645         * config/arm/arm.opt (mverbose-cost-dump): New option.
3646         * config/arm/arm.c (arm_rtx_costs): Use it.
3648 2017-12-14  Andreas Schwab  <schwab@linux-m68k.org>
3650         PR bootstrap/83396
3651         * reload1.c (emit_input_reload_insns): Skip debug markers.
3653 2017-12-14  Alexandre Oliva  <aoliva@redhat.com>
3655         * config/i386/i386.c (rest_of_insert_endbranch): Use call loc
3656         for its nop_endbr.
3658         PR bootstrap/83396
3659         * config/arc/arc.c (hwloop_optimize): Skip debug insns.
3660         * config/sh/sh-protos.h (sh_find_set_of_reg): Adjust.
3661         * config/sh/sh.c: Skip debug insns besides notes.
3662         * config/sh/sh.md: Likewise.
3663         * config/sh/sh_treg_combine.cc: Likewise.
3664         * config/sh/sync.md: Likewise.
3666 2017-12-14  Tom de Vries  <tom@codesourcery.com>
3668         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
3669         Add item for weak_undefined.
3671 2017-12-14  Richard Biener  <rguenther@suse.de>
3673         PR tree-optimization/67842
3674         * sese.h (bb_in_region): Remove #if 0'ed code.
3676 2017-12-14  Richard Biener  <rguenther@suse.de>
3678         PR tree-optimization/83326
3679         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Add
3680         may_be_zero parameter and handle it by not marking the first
3681         peeled copy as not exiting the loop.
3682         (try_peel_loop): Likewise.
3683         (canonicalize_loop_induction_variables): Use number_of_iterations_exit
3684         to handle the case of constant or zero iterations and perform
3685         loop header copying on-the-fly.
3687 2017-12-14  Richard Biener  <rguenther@suse.de>
3689         PR tree-optimization/83418
3690         * vr-values.c (vr_values::extract_range_for_var_from_comparison_expr):
3691         Instead of asserting we don't get unfolded comparisons deal with them.
3693 2017-12-14  Jakub Jelinek  <jakub@redhat.com>
3695         PR bootstrap/83396
3696         * var-tracking.c (vt_initialize): Ignore non-DEBUG_INSNs outside of
3697         basic blocks.  Assert debug bind insns don't appear outside of bbs,
3698         don't reset them.  Assert insns without BLOCK_FOR_INSN are outside of
3699         bb.  Simplify.
3701         PR tree-optimization/83198
3702         * gimple-ssa-sprintf.c (format_floating): Set type solely based on
3703         dir.modifier, regardless of TREE_TYPE (arg).  Assume non-REAL_CST
3704         value if arg is a REAL_CST with incompatible type.
3706 2017-12-14  Sudakshina Das  <sudi.das@arm.com>
3707             Bin Cheng  <bin.cheng@arm.com>
3709         PR target/81228
3710         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Move LTGT
3711         to CCFPEmode.
3712         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Add
3713         LTGT.
3715 2017-12-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3717         * config/aarch64/aarch64-cores.def (cortex-a55, cortex-a75,
3718         cortex-a75.cortex-a55): Specify AARCH64_FL_F16 in the arch features.
3720 2017-12-14  Richard Sandiford  <richard.sandiford@linaro.org>
3721             Alan Hayward  <alan.hayward@arm.com>
3722             David Sherwood  <david.sherwood@arm.com>
3724         * poly-int.h: New file.
3725         * poly-int-types.h: Likewise.
3726         * coretypes.h: Include them.
3727         (POLY_INT_CONVERSION): Define.
3728         * target.def (estimated_poly_value): New hook.
3729         * doc/tm.texi.in (TARGET_ESTIMATED_POLY_VALUE): New hook.
3730         * doc/tm.texi: Regenerate.
3731         * doc/poly-int.texi: New file.
3732         * doc/gccint.texi: Include it.
3733         * doc/rtl.texi: Describe restrictions on subreg modes.
3734         * Makefile.in (TEXI_GCCINT_FILES): Add poly-int.texi.
3735         * genmodes.c (NUM_POLY_INT_COEFFS): Provide a default definition.
3736         (emit_insn_modes_h): Emit a definition of NUM_POLY_INT_COEFFS.
3737         * targhooks.h (default_estimated_poly_value): Declare.
3738         * targhooks.c (default_estimated_poly_value): New function.
3739         * target.h (estimated_poly_value): Likewise.
3740         * wide-int.h (WI_UNARY_RESULT): Use wi::binary_traits.
3741         (wi::unary_traits): Delete.
3742         (wi::binary_traits::signed_shift_result_type): Define for
3743         offset_int << HOST_WIDE_INT, etc.
3744         (generic_wide_int::operator <<=): Define for all types and use
3745         wi::lshift instead of <<.
3746         (wi::hwi_with_prec): Add a default constructor.
3747         (wi::ints_for): New class.
3748         (operator <<): Define for all wide-int types.
3749         (operator /): New function.
3750         (operator %): Likewise.
3751         * selftest.h (ASSERT_KNOWN_EQ, ASSERT_KNOWN_EQ_AT, ASSERT_MAYBE_NE)
3752         (ASSERT_MAYBE_NE_AT): New macros.
3754 2017-12-13  Eric Botcazou  <ebotcazou@adacore.com>
3755             Dominik Vogt  <vogt@linux.vnet.ibm.com>
3757         PR middle-end/78468
3758         * emit-rtl.c (init_emit): Remove ??? comment.
3759         * explow.c (get_dynamic_stack_size): Take known alignment of stack
3760         pointer + STACK_DYNAMIC_OFFSET into account in lieu of STACK_BOUNDARY.
3761         * config/sparc/sparc.h (INIT_EXPANDERS): In 32-bit mode, lower the
3762         alignment of 3 virtual registers to BITS_PER_WORD.
3764         * config/sparc/sparc.c (sparc_compute_frame_size): Simplify.
3766 2017-12-13  Peter Bergner  <bergner@vnet.ibm.com>
3768         * config/rs6000/ppc-auxv.h (PPC_FEATURE2_HTM_NO_SUSPEND): New define.
3769         * config/rs6000/rs6000.c (cpu_supports_info): Use it.
3771 2017-12-13  Alexandre Oliva  <aoliva@redhat.com>
3773         PR bootstrap/83396
3774         * reload1.c (eliminate_regs_in_insn): Skip debug markers.
3776 2017-12-13  Jakub Jelinek  <jakub@redhat.com>
3778         * tree-cfg.c (verify_gimple_in_cfg): Verify no non-label stmts
3779         with the exception of debug begin stmt markers appear before
3780         labels.
3782         PR bootstrap/83396
3783         * final.c (rest_of_handle_final): Call variable_tracking_main only
3784         if !flag_var_tracking.
3786 2017-12-13  Alexandre Oliva  <aoliva@redhat.com>
3787             Jakub Jelinek  <jakub@redhat.com>
3789         PR bootstrap/83396
3790         PR debug/83391
3791         * tree-cfgcleanup.c (remove_forwarder_block): Keep after
3792         labels debug stmts that can only appear after labels.
3794 2017-12-13  Alexander Monakov  <amonakov@ispras.ru>
3796         PR rtl-optimization/82398
3797         * sel-sched.c (sel_rank_for_schedule): Fix check for zero
3798         EXPR_USEFULNESS in priority comparison.
3800 2017-12-13  Segher Boessenkool  <segher@kernel.crashing.org>
3802         PR rtl-optimization/83393
3803         * combine.c (move_deaths): If reg_stat points to a too new insn in
3804         last_death, do not use it: find the proper insn instead.
3806 2017-12-12  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
3808         PR target/82190
3809         * config/rs6000/rs6000-string.c (expand_block_compare,
3810         expand_strn_compare): Fix set_mem_size() calls.
3812 2017-12-12  Jeff Law  <law@redhat.com>
3814         PR tree-optimization/83298
3815         PR tree-optimization/83362
3816         PR tree-optimization/83383
3817         * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Make
3818         push_value_range a public interface.  Add new argument to
3819         record_ranges_from_stmt.
3820         * gimple-ssa-evrp-analyze.c
3821         (evrp_range_analyzer::record_ranges_from_stmt): Add new argument.
3822         Update comments.  Handle recording temporary equivalences.
3823         * tree-ssa-dom.c (dom_opt_opt_walker::before_dom_children): Add
3824         new argument to call to evrp_range_analyzer::record_ranges_from_stmt.
3825         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Likewise.
3826         * tree-ssa-threadedge.c: Include alloc-pool.h, vr-values.h and
3827         gimple-ssa-evrp-analyze.h.
3828         (record_temporary_equivalences_from_phis): Add new argument.  When
3829         the PHI arg is an SSA_NAME, set the result's range to the range
3830         of the PHI arg.
3831         (record_temporary_equivalences_from_stmts_at_dest): Record ranges
3832         from statements too.
3833         (thread_through_normal_block): Accept new argument, evrp_range_analyzer.
3834         Pass it down to children as needed.
3835         (thread_outgoing_edges): Likewise.
3836         (thread_across_edge): Likewise.   Push/pop range state as needed.
3837         * tree-ssa-threadedge.h (thread_outgoing_edges): Update prototype.
3839 2017-12-12  Julia Koval  <julia.koval@intel.com>
3841         * config/i386/i386.c (PTA_SKYLAKE_AVX512): Add PTA_CLWB.
3842         (PTA_CANNONLAKE): Remove PTA_CLWB.
3844 2017-12-12  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3846         PR target/83332
3847         * config/rs6000/vector.md (vcondv2dfv2di): New define_expand.
3848         (vcondv2div2df): Likewise.
3849         (vconduv2dfv2di): Likewise.
3851 2017-12-12  Jakub Jelinek  <jakub@redhat.com>
3853         * builtins.def (BUILT_IN_NEXTAFTER, BUILT_IN_NEXTAFTERF,
3854         BUILT_IN_NEXTAFTERL, BUILT_IN_NEXTTOWARD, BUILT_IN_NEXTTOWARDF,
3855         BUILT_IN_NEXTTOWARDL): Use ATTR_MATHFN_ERRNO instead of
3856         ATTR_MATHFN_FPROUNDING_ERRNO.
3858 2017-12-12  Richard Biener  <rguenther@suse.de>
3860         PR tree-optimization/83385
3861         * graphite-scop-detection.c (get_order, order): Remove.
3862         (bb_to_rpo): New global.
3863         (cmp_pbbs): Adjust.
3864         (build_scops): Sort pbbs in RPO order.
3866 2017-12-12  James Greenhalgh  <james.greenhalgh@arm.com>
3868         * combine.c (simplify_set): Do not transform subregs to zero_extends
3869         if the destination mode is a vector mode.
3871 2017-12-12  Jakub Jelinek  <jakub@redhat.com>
3873         PR tree-optimization/83359
3874         * tree-cfg.h (fold_loop_internal_call): Declare.
3875         * tree-vectorizer.c (fold_loop_internal_call): Moved to ...
3876         * tree-cfg.c (fold_loop_internal_call): ... here.  No longer static.
3877         (find_loop_dist_alias): New function.
3878         (move_sese_region_to_fn): If any dloop->orig_loop_num value is
3879         updated, also adjust any corresponding LOOP_DIST_ALIAS internal
3880         calls.
3882         PR tree-optimization/80631
3883         * tree-vect-loop.c (get_initial_def_for_reduction): Fix comment typo.
3884         (vect_create_epilog_for_reduction): Add INDUC_VAL and INDUC_CODE
3885         arguments, for INTEGER_INDUC_COND_REDUCTION use INDUC_VAL instead of
3886         hardcoding zero as the value if COND_EXPR is never true.  For
3887         INTEGER_INDUC_COND_REDUCTION don't emit the final COND_EXPR if
3888         INDUC_VAL is equal to INITIAL_DEF, and use INDUC_CODE instead of
3889         hardcoding MAX_EXPR as the reduction operation.
3890         (is_nonwrapping_integer_induction): Allow negative step.
3891         (vectorizable_reduction): Compute INDUC_VAL and INDUC_CODE for
3892         vect_create_epilog_for_reduction, if no value is suitable, don't
3893         use INTEGER_INDUC_COND_REDUCTION for now.  Formatting fixes.
3895 2017-12-12  Richard Biener  <rguenther@suse.de>
3897         PR tree-optimization/81889
3898         * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness): Use
3899         range info from the non-wrapping IV instead of just the range
3900         of the type.
3902 2017-12-12  Julia Koval  <julia.koval@intel.com>
3904         * config.gcc: Add vaesintrin.h.
3905         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI_V64QI):
3906         New type.
3907         * config/i386/i386-builtin.def (__builtin_ia32_vaesdec_v16qi,
3908         __builtin_ia32_vaesdec_v32qi, __builtin_ia32_vaesdec_v64qi):
3909         New builtins.
3910         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
3911         * config/i386/immintrin.h: Include vaesintrin.h.
3912         * config/i386/sse.md (vaesdec_<mode>): New pattern.
3913         * config/i386/vaesintrin.h (_mm256_aesdec_epi128, _mm512_aesdec_epi128,
3914         _mm_aesdec_epi128): New intrinsics.
3916 2017-12-12  Julia Koval  <julia.koval@intel.com>
3918         * common/config/i386/i386-common.c (OPTION_MASK_ISA_VAES_SET,
3919         OPTION_MASK_ISA_VAES_UNSET): New.
3920         (ix86_handle_option): Handle -mvaes.
3921         * config/i386/cpuid.h: Define bit_VAES.
3922         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mvaes.
3923         * config/i386/i386-c.c (__VAES__): New.
3924         * config/i386/i386.c (ix86_target_string): Add -mvaes.
3925         (ix86_valid_target_attribute_inner_p): Ditto.
3926         * config/i386/i386.h (TARGET_VAES, TARGET_VAES_P): New.
3927         * config/i386/i386.opt: Add -mvaes.
3928         * doc/invoke.texi: Ditto.
3930 2017-12-12  Alexandre Oliva  <aoliva@redhat.com>
3932         * debug.h (gcc_debug_hooks): Add inline_entry.
3933         * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Likewise.
3934         * debug.c (do_nothing_debug_hooks): Likewise.
3935         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
3936         * dwarf2out.c (dwarf2_debug_hooks): Likewise.
3937         (dwarf2_lineno_debug_hooks): Likewise.
3939         * common.opt (gstatement-frontiers): New, setting
3940         debug_nonbind_markers_p.
3941         * rtl.h (MAY_HAVE_DEBUG_MARKER_INSNS): Activate.
3942         * toplev.c (process_options): Autodetect value for debug statement
3943         frontiers option.
3944         * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): Activate.
3945         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers): New.
3947         * cfgexpand.c (expand_gimple_basic_block): Handle begin stmt
3948         markers.  Integrate source bind into debug stmt expand loop.
3949         (pass_expand::execute): Check debug marker limit.  Avoid deep
3950         TER and expand debug locations for debug bind insns only.
3951         * cse.c (insn_live_p): Keep nonbind markers and debug bindings
3952         followed by them.
3953         * df-scan.c (df_insn_delete): Accept out-of-block debug insn.
3954         * final.c (reemit_insn_block_notes): Take current block from
3955         nonbind markers.  Declare note where it's first set.
3956         (final_scan_insn): Handle begin stmt notes.  Emit is_stmt according to
3957         begin stmt markers if enabled.
3958         (notice_source_line): Handle nonbind markers.  Fail if their
3959         location is unknown or that of builtins.
3960         (rest_of_handle_final): Convert begin stmt markers to notes if
3961         var-tracking didn't run.
3962         (rest_of_clean_state): Skip begin stmt markers.
3963         * gimple-pretty-print.c (dump_gimple_debug): Handle begin stmt
3964         markers.
3965         * function.c (allocate_struct_function): Set begin_stmt_markers.
3966         * function.h (struct function): Add debug_marker_count counter
3967         and debug_nonbind_markers flag.
3968         * gimple-iterator.c (gsi_remove): Adjust debug_marker_count.
3969         * gimple-low.c (lower_function_body): Adjust
3970         debug_nonbind_markers.
3971         (lower_stmt): Drop or skip gimple debug stmts.
3972         (lower_try_catch): Skip debug stmts.
3973         * gimple.c (gimple_build_debug_begin_stmt): New.
3974         (gimple_copy): Increment debug_marker_count if copying one.
3975         * gimple.h (gimple_build_debug_begin_stmt): Declare.
3976         * gimplify.c (rexpr_location): New.
3977         (rexpr_has_location): New.
3978         (warn_switch_unreachable_r): Handle gimple debug stmts.
3979         (shortcut_cond_r): Call expr_location.
3980         (find_goto): New.
3981         (find_goto_label): New.
3982         (shortcut_cond_expr): Call expr_has_location, expr_location, and
3983         find_goto_label.
3984         (gimplify_cond_expr): Call find_goto_label, expr_has_location, and
3985         expr_location.
3986         (gimplify_expr): Handle begin stmt markers.  Reject debug expr decls.
3987         * langhooks-def.h (LANG_HOOKS_EMITS_BEGIN_STMT): New.  Add to...
3988         (LANG_HOOKS_INITIALIZER): ... this.
3989         * langhooks.h (struct lang_hooks): Add emits_begin_stmt.
3990         * lra-contraints.c (inherit_reload_reg): Tolerate between-blocks
3991         debug insns.
3992         (update_ebb_live_info): Skip debug insn markers.
3993         * lra.c (debug_insn_static_data): Rename to...
3994         (debug_bind_static_data): ... this.
3995         (debug_marker_static_data): New.
3996         (lra_set_insn_recog_data): Select one of the above depending
3997         on debug insn kind.
3998         (lra_update_isn_regno_info): Don't assume debug insns have
3999         freqs.
4000         (push_insns): Skip debug insns.
4001         * lto-streamer-in.c (input_function): Drop debug stmts
4002         depending on active options.  Adjust debug_nonbind_markers.
4003         * params.def (PARAM_MAX_DEBUG_MARKER_COUNT): New.
4004         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
4005         begin stmt marker notes.
4006         (print_insn): Likewise.
4007         * recog.c (extract_insn): Recognize rtl for debug markers.
4008         * rtl.def (DEBUG_MARKER): New.
4009         * tree-inline.c: Include params.h.
4010         (remap_gimple_stmt): Handle nonbind markers.
4011         (maybe_move_debug_stmts_to_successors): Likewise.
4012         (copy_debug_stmt): Likewise.
4013         * tree-iterator.c (append_to_statement_list_1): Append begin stmt
4014         markers regardless of no side effects.
4015         (tsi_link_before): Don't update container's side effects when adding
4016         a begin stmt marker.
4017         (tsi_link_after): Likewise.
4018         (expr_first): Skip begin stmt markers.
4019         (expr_last): Likewise.
4020         * tree-pretty-print (dump_generic_node): Handle begin stmt markers.
4021         * tree-ssa-threadedge.c (propagate_threaded_block_debug_info):
4022         Disregard nonbind markers.
4023         * tree.c (make_node_stat): Don't set side effects for begin stmt
4024         markers.
4025         (build1_stat): Likewise.
4026         * tree.def (DEBUG_BEGIN_STMT): New.
4027         * tree.h (GOTO_DESTINATION): Require a GOTO_EXPR.
4028         * var-tracking.c (delete_debug_insns): Renamed to...
4029         (delete_vta_debug_insns): ... this.
4030         (reemit_marker_as_note): New.
4031         (vt_initialize): Reemit markers.
4032         (delete_vta_debug_insns): Likewise.
4033         (vt_debug_insns_local): Reemit or delete markers.
4034         (variable_tracking_main_1): Likewise.
4035         * doc/generic.texi (DEBUG_BEGIN_STMT): Document.
4036         * doc/gimple.texi (gimple_debug_begin_stmt_p): New.
4037         (gimple_debug_nonbind_marker_p): New.
4038         (gimple_build_debug_bind): Adjust.
4039         (gimple_build_debug_begin_stmt): New.
4040         * doc/invoke.texi (max-debug-marker-count): New param.
4041         * doc/rtl.texi (debug_implicit_ptr, entry_value): New.
4042         (debug_parameter_ref, debug_marker): New.
4043         (NOTE_INSN_BEGIN_STMT): New.
4044         (DEBUG_INSN): Describe begin stmt markers.
4046         * cfgbuild.c (find_bb_boundaries): Don't purge dead edges if,
4047         without debug insns, we wouldn't, but clean up debug insns
4048         after a control flow insn nevertheless.
4050         * cfgcleanup.c (delete_unreachable_blocks): Use alternate
4051         block removal order if MAY_HAVE_DEBUG_BIND_INSNS.
4052         * cfgexpand.c (label_rtx_for_bb): Skip debug insns.
4053         * cfgrtl.c (try_redirect_by_replacing_jump): Skip debug insns.
4054         (rtl_tidy_fallthru_edge): Likewise.
4055         (rtl_verify_fallthru): Likewise.
4056         (rtl_verify_bb_layout): Likewise.
4057         (skip_insns_after_block): Likewise.
4058         (duplicate_insn_chain): Use DEBUG_BIND_INSN_P.
4059         * dwarf2out.c: Include print-rtl.h.
4060         (dwarf2out_next_real_insn): New.
4061         (dwarf2out_var_location): Call it.  Disregard begin stmt markers.
4062         Dump debug binds in asm comments.
4063         * gimple-iterator.c (gimple_find_edge_insert_loc): Skip debug stmts.
4064         * gimple-iterator.h (gsi_start_bb_nondebug): Remove; adjust
4065         callers to use gsi_start_nondebug_bb instead.
4066         (gsi_after_labels): Skip gimple debug stmts.
4067         (gsi_start_nondebug): New.
4068         * gimple-loop-interchange.c (find_deps_in_bb_for_stmt): Adjust.
4069         (proper_loop_form_for_interchange): Adjust.
4070         * gimple-low.c (gimple_seq_may_fallthru): Take last nondebug stmt.
4071         * gimple.h (gimple_seq_last_nondebug_stmt): New.
4072         * gimplify.c (last_stmt_in_scope): Skip debug stmts.
4073         (collect_fallthrough_labels): Likewise.
4074         (should_warn_for_implicit_fallthrough): Likewise.
4075         (warn_implicit_fallthrough_r): Likewise.
4076         (expand_FALLTHROUGH_r): Likewise.
4077         * graphite-isl-ast-to-gimple.c (gsi_insert_earliest): Adjust.
4078         (graphite_copy_stmts_from_block): Skip nonbind markers.
4079         * haifa-sched.c (sched_extend_bb): Skip debug insns.
4080         * ipa-icf-gimple.c (func_checker::compare_bb): Adjust.
4081         * jump.c (clean_barriers): Skip debug insns.
4082         * omp-expand.c (expand_parallel_call): Skip debug insns.
4083         (expand_task_call): Likewise.
4084         (remove_exit_barrier): Likewise.
4085         (expand_omp_taskreg): Likewise.
4086         (expand_omp_for_init_counts): Likewise.
4087         (expand_omp_for_generic): Likewise.
4088         (expand_omp_for_static_nochunk): Likewise.
4089         (expand_omp_for_static_chunk): Likewise.
4090         (expand_omp_simd): Likewise.
4091         (expand_omp_taskloop_for_outer): Likewise.
4092         (expand_omp_taskloop_for_inner): Likewise.
4093         (expand_oacc_for): Likewise.
4094         (expand_omp_sections): Likewise.
4095         (expand_omp_single): Likewise.
4096         (expand_omp_synch): Likewise.
4097         (expand_omp_atomic_load): Likewise.
4098         (expand_omp_atomic_store): Likewise.
4099         (expand_omp_atomic_fetch_op): Likewise.
4100         (expand_omp_atomic_pipeline): Likewise.
4101         (expand_omp_atomic_mutex): Likewise.
4102         (expand_omp_target): Likewise.
4103         (grid_expand_omp_for_loop): Likewise.
4104         (grid_expand_target_grid_body): Likewise.
4105         (build_omp_regions_1): Likewise.
4106         * omp-low.c (check_combined_parallel): Skip debug stmts.
4107         * postreload.c (fixup_debug_insns): Skip nonbind debug insns.
4108         * regcprop.c (find_oldest_value_reg): Ensure REGNO is not a pseudo.
4109         * sese.c (sese_trivially_empty_bb_p): Call is_gimple_debug in
4110         test.
4111         * tree-cfg.c (make_blobs_1): Skip debug stmts.
4112         (make_edges): Likewise.
4113         (cleanup_dead_labels): Likewise.
4114         (gimple_can_merge_blocks_p): Likewise.
4115         (stmt_starts_bb_p): Likewise.
4116         (gimple_block_label): Likewise.
4117         (gimple_redirect_edge_and_branch): Likewise.
4118         * tree-cfgcleanup.c (remove_forwarder_block): Rearrange skipping
4119         of debug stmts.
4120         (execute_cleanup_cfg_post_optimizing): Dump enumerated decls with
4121         TDF_SLIM.
4122         * tree-pretty-print (print_declaration): Omit initializer in slim
4123         dumps.
4124         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Mark begin stmt
4125         markers.
4126         (eliminate_unnecessary_stmts): Stabilize block removal order.
4127         * tree-ssa-tail-merge.c (find_duplicate): Skip debug stmts.
4128         * var-tracking.c (get_first_insn): New.
4129         (vt_emit_notes): Call it.
4130         (vt_initialize): Walk any insns before the first BB.
4131         (delete_debug_insns): Likewise.
4133         * gimple.h (enum gimple_debug_subcode): Add
4134         GIMPLE_DEBUG_BEGIN_STMT.
4135         (gimple_debug_begin_stmt_p): New.
4136         (gimple_debug_nonbind_marker_p): New.
4137         * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): New.
4138         (MAY_HAVE_DEBUG_BIND_STMTS): Renamed from....
4139         (MAY_HAVE_DEBUG_STMTS): ... this.  Check both.
4140         * insn-notes.def (BEGIN_STMT): New.
4141         * rtl.h (MAY_HAVE_DEBUG_MARKER_INSNS): New.
4142         (MAY_HAVE_DEBUG_BIND_INSNS): Renamed from....
4143         (MAY_HAVE_DEBUG_INSNS): ... this.  Check both.
4144         (NOTE_MARKER_LOCATION, NOTE_MARKER_P): New.
4145         (DEBUG_BIND_INSN_P, DEBUG_MARKER_INSN_P): New.
4146         (INSN_DEBUG_MARKER_KIND): New.
4147         (GEN_RTX_DEBUG_MARKER_BEGIN_STMT_PAT): New.
4148         (INSN_VAR_LOCATION): Check for VAR_LOCATION.
4149         (INSN_VAR_LOCATION_PTR): New.
4150         * cfgexpand.c (expand_debug_locations): Handle debug bind insns
4151         only.
4152         (expand_gimple_basic_block): Likewise.  Emit debug temps for TER
4153         deps only if debug bind insns are enabled.
4154         (pass_expand::execute): Avoid deep TER and expand
4155         debug locations for debug bind insns only.
4156         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Narrow
4157         debug stmts special handling down to debug bind stmts.
4158         * combine.c (try_combine): Narrow debug insns special handling
4159         down to debug bind insns.
4160         * cse.c (delete_trivially_dead_insns): Handle debug bindings.
4161         Narrow debug insns preexisting special handling down to debug
4162         bind insns.
4163         * dce.c (rest_of_handle_ud_dce): Narrow debug insns special
4164         handling down to debug bind insns.
4165         * function.c (instantiate_virtual_regs): Skip debug markers,
4166         adjust handling of debug binds.
4167         * gimple-ssa-backprop.c (backprop::prepare_change): Try debug
4168         temp insertion iff MAY_HAVE_DEBUG_BIND_STMTS.
4169         * haifa-sched.c (schedule_insn): Narrow special handling of debug
4170         insns to debug bind insns.
4171         * ipa-param-manipulation.c (ipa_modify_call_arguments): Narrow
4172         special handling of debug stmts to debug bind stmts.
4173         * ipa-split.c (split_function): Likewise.
4174         * ira.c (combine_and_move_insns): Adjust debug bind insns only.
4175         * loop-unroll.c (apply_opt_in_copies): Adjust tests on bind
4176         debug insns.
4177         * reg-stack.c (convert_regs_1): Use DEBUG_BIND_INSN_P.
4178         * regrename.c (build_def_use): Likewise.
4179         * regcprop.c (copyprop_hardreg_forward_1): Likewise.
4180         (pass_cprop_hardreg): Narrow special casing of debug insns to
4181         debug bind insns.
4182         * regstat.c (regstat_init_n_sets_and_refs): Likewise.
4183         * reload1.c (reload): Likewise.
4184         * sese.c (sese_insert_phis_for_liveouts): Narrow special
4185         casing of debug stmts to debug bind stmts.
4186         * shrink-wrap.c (move_insn_for_shrink_wrap): Likewise.
4187         * ssa-iterators.h (num_imm_uses): Likewise.
4188         * tree-cfg.c (gimple_merge_blocks): Narrow special casing of
4189         debug stmts to debug bind stmts.
4190         * tree-inline.c (tree_function_versioning): Narrow special casing
4191         of debug stmts to debug bind stmts.
4192         * tree-loop-distribution.c (generate_loops_for_partition):
4193         Narrow special casing of debug stmts to debug bind stmts.
4194         * tree-sra.c (analyze_access_subtree): Narrow special casing
4195         of debug stmts to debug bind stmts.
4196         * tree-ssa-dce.c (remove_dead_stmt): Narrow special casing of debug
4197         stmts to debug bind stmts.
4198         * tree-ssa-loop-ivopt.c (remove_unused_ivs): Narrow special
4199         casing of debug stmts to debug bind stmts.
4200         * tree-ssa-reassoc.c (reassoc_remove_stmt): Likewise.
4201         * tree-ssa-tail-merge.c (tail_merge_optimize): Narrow special
4202         casing of debug stmts to debug bind stmts.
4203         * tree-ssa-threadedge.c (propagate_threaded_block_debug_info):
4204         Likewise.
4205         * tree-ssa.c (flush_pending_stmts): Narrow special casing of
4206         debug stmts to debug bind stmts.
4207         (gimple_replace_ssa_lhs): Likewise.
4208         (insert_debug_temp_for_var_def): Likewise.
4209         (insert_debug_temps_for_defs): Likewise.
4210         (reset_debug_uses): Likewise.
4211         * tree-ssanames.c (release_ssa_name_fn): Likewise.
4212         * tree-vect-loop-manip.c (adjust_debug_stmts_now): Likewise.
4213         (adjust_debug_stmts): Likewise.
4214         (adjust_phi_and_debug_stmts): Likewise.
4215         (vect_do_peeling): Likewise.
4216         * tree-vect-loop.c (vect_transform_loop): Likewise.
4217         * valtrack.c (propagate_for_debug): Use BIND_DEBUG_INSN_P.
4218         * var-tracking.c (adjust_mems): Narrow special casing of debug
4219         insns to debug bind insns.
4220         (dv_onepart_p, dataflow_set_clar_at_call, use_type): Likewise.
4221         (compute_bb_dataflow, vt_find_locations): Likewise.
4222         (vt_expand_loc, emit_notes_for_changes): Likewise.
4223         (vt_init_cfa_base): Likewise.
4224         (vt_emit_notes): Likewise.
4225         (vt_initialize): Likewise.
4226         (vt_finalize): Likewise.
4228         * emit-rtl.c (next_nondebug_insn, prev_nondebug_insn): Reorder.
4229         (next_nonnote_nondebug_insn, prev_nonnote_nondebug_insn): Reorder.
4230         (next_nonnote_nondebug_insn_bb): New.
4231         (prev_nonnote_nondebug_insn_bb): New.
4232         (prev_nonnote_insn_bb, next_nonnote_insn_bb): Remove.
4233         * rtl.h (prev_nonnote_insn_bb, next_nonnote_insn_bb): Remove decls.
4234         (prev_nonnote_nondebug_insn_bb): Declare.
4235         (next_nonnote_nondebug_insn_bb): Declare.
4236         * cfgbuild.c (find_bb_boundaries): Adjust to skip debug insns.
4237         * cfgrtl.c (get_last_bb_insn): Likewise.
4238         * lra.c (push_insns): Likewise.
4240 2017-12-11  David Malcolm  <dmalcolm@redhat.com>
4242         PR c/82050
4243         * selftest-run-tests.c (selftest::run_tests): Move start/finish code
4244         to...
4245         * selftest.c (selftest::test_runner::test_runner): New ctor.
4246         (selftest::test_runner::~test_runner): New dtor.
4247         * selftest.h (class selftest::test_runner): New class.
4249 2017-12-11  Carl Love  <cel@us.ibm.com>
4251         * config/rs6000/altivec.h (vec_extract_fp32_from_shorth,
4252         vec_extract_fp32_from_shortl]): Add #defines.
4253         * config/rs6000/rs6000-builtin.def (VSLDOI_2DI): Add macro expansion.
4254         * config/rs6000/rs6000-c.c (ALTIVEC_BUILTIN_VEC_UNPACKH,
4255         ALTIVEC_BUILTIN_VEC_UNPACKL, ALTIVEC_BUILTIN_VEC_AND,
4256         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SRL,
4257         ALTIVEC_BUILTIN_VEC_SRO, ALTIVEC_BUILTIN_VEC_SLD,
4258         ALTIVEC_BUILTIN_VEC_SLL): Add expansions.
4259         * doc/extend.texi: Add documentation for the added builtins.
4261 2017-12-11  Alexandre Oliva  <aoliva@redhat.com>
4263         PR rtl-optimization/80693
4264         PR rtl-optimization/81019
4265         PR rtl-optimization/81020
4266         * combine.c (distribute_notes): Reset any REG_UNUSED REGs that
4267         are not mentioned in i3.  Place the REG_UNUSED note on i2,
4268         possibly modified to REG_DEAD, if it did not originate in i3.
4270 2017-12-11  Jakub Jelinek  <jakub@redhat.com>
4272         * recog.c (store_data_bypass_p_1): New function.
4273         (store_data_bypass_p): Handle USE in a PARALLEL like CLOBBER.  Use
4274         store_data_bypass_p_1 to avoid code duplication.  Formatting fixes.
4276 2017-12-11  Segher Boessenkool  <segher@kernel.crashing.org>
4278         PR rtl-optimization/83361
4279         * ifcvt.c (if_convert): Call fixup_partitions.
4281 2017-12-11  Will Schmidt  <will_schmidt@vnet.ibm.com>
4283         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
4284         early folding of splat_u{8,16,32}.
4286 2017-12-11  Jakub Jelinek  <jakub@redhat.com>
4288         * config/aarch64/aarch64.c (aarch64_print_operand): Don't start
4289         output_operand_lossage first argument with capital letter.
4290         (aarch64_override_options): Don't start error and sorry first argument
4291         with capital letter.
4293 2017-12-11  Andi Kleen  <ak@linux.intel.com>
4295         PR gcov-profile/83355
4296         * auto-profile.c (string_table::get_index_by_decl): Don't
4297         recurse when abstract origin points to itself.
4299 2017-12-11  Bin Cheng  <bin.cheng@arm.com>
4301         PR tree-optimization/83320
4302         * gimple-loop-interchange.cc (free_data_refs_with_aux): Use delete.
4303         (prune_datarefs_not_in_loop): Ditto.
4305 2017-12-10  Gerald Pfeifer  <gerald@pfeifer.com>
4307         * doc/install.texi (Specific): Tweak link to mkssoftware.com.
4309 2017-12-10  Jakub Jelinek  <jakub@redhat.com>
4311         PR tree-optimization/83337
4312         * gimple-loop-interchange.cc (compute_access_stride): Handle
4313         bitfield DRs properly.
4315 2017-12-09  Jakub Jelinek  <jakub@redhat.com>
4317         PR tree-optimization/83338
4318         * tree-vect-stmts.c (vectorizable_operation): Handle POINTER_DIFF_EXPR
4319         vectorization as MINUS_EXPR with a subsequent VIEW_CONVERT_EXPR from
4320         vector of unsigned integers to vector of signed integers.
4322 2017-12-08  Vladimir Makarov  <vmakarov@redhat.com>
4324         PR rtl-optimization/83317
4325         * lra-constraints.c (process_address_1): Add insn code check.
4327 2017-12-08  Michael Matz  <matz@suse.de>
4329         Fix PR tree-optimization/83323
4330         * gimple-loop-jam (unroll_jam_possible_p): Correct test for
4331         head-controlled loops and loop BBs.
4332         * common.opt (funroll-and-jam): Remove, instead ...
4333         (floop-unroll-and-jam): ... reuse this option.
4334         * opts.c (default_options_table): Use OPT_floop_unroll_and_jam.
4335         * doc/invoke.texi (-funroll-and-jam): Move docu to ...
4336         (-floop-unroll-and-jam): ... this option.
4338 2017-12-08  Jakub Jelinek  <jakub@redhat.com>
4340         * ipa-polymorphic-call.c (noncall_stmt_may_be_vtbl_ptr_store): Fix
4341         a comment typo, get_base_ref_and_offset -> get_ref_base_and_extent.
4342         * ipa-prop.c (stmt_may_be_vtbl_ptr_store): Likewise.
4344 2017-12-08  Richard Biener  <rguenther@suse.de>
4346         PR middle-end/81782
4347         * tree-ssa-uninit.c (warn_uninitialized_vars): Properly
4348         handle accesses outside of zero-sized vars.
4350 2017-12-08  Martin Jambor  <mjambor@suse.cz>
4352         PR tree-optimization/83141
4353         * tree-sra.c (contains_vce_or_bfcref_p): Move up in the file, also
4354         test for MEM_REFs implicitely changing types with padding.  Remove
4355         inline keyword.
4356         (build_accesses_from_assign): Added contains_vce_or_bfcref_p checks.
4358 2017-12-08  Jakub Jelinek  <jakub@redhat.com>
4360         * config/arc/arc.c (arc_attribute_table): Add exclusions to
4361         the comment.
4362         * config/avr/avr.c (avr_attribute_table): Likewise.
4363         * config/msp430/msp430.c (msp430_attribute_table): Likewise.
4364         * config/rl78/rl78.c (rl78_attribute_table): Likewise.
4365         * config/nds32/nds32.c (nds32_attribute_table): Likewise.
4366         * config/darwin.h (SUBTARGET_ATTRIBUTE_TABLE): Initialize new member
4367         of struct attribute_spec.
4368         * config/i386/cygming.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
4370 2017-12-08  Ulrich Weigand  <uweigand@de.ibm.com>
4372         PR target/82960
4373         * config/spu/spu.c (pad_bb): Only check INSN_CODE when INSN_P is true.
4375 2017-12-08  Jan Hubicka  <hubicka@ucw.cz>
4377         PR middle-end/83609
4378         * profile-count.c (profile_count::from_gcov_type): Move from
4379         profile-count.h; handle overflow.
4380         * profile-count.h (profile_count::from_gcov_type): Move offline.
4382 2017-12-08  Segher Boessenkool  <segher@kernel.crashing.org>
4384         PR rtl-optimization/83304
4385         * combine.c (move_deaths): If we do not know where a register died,
4386         search for it.
4388 2017-12-08  Richard Biener  <rguenther@suse.de>
4390         * gimple-loop-interchange.cc (tree_loop_interchange::interchange):
4391         Provide -fopt-info-loop feedback when we interchange in a nest.
4393 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4395         * config/arm/driver-arm.c (arm_cpu_table): Use fp-capable product names
4396         for armv6 ARM CPU IDs.
4398 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4400         * common/config/arm/arm-common.c: Include <algorithm>.
4401         (INCLUDE_VECTOR): Define.
4402         (compare_opt_names): New function.
4403         (arm_rewrite_selected_arch): Only strip out extensions that can be
4404         expressed through -mfpu.  Sort the remaining extensions
4405         alphabetically.
4407 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4409         * config/arm/arm.h (arm_asm_auto_mfpu): Declare.
4410         (ASM_CPU_SPEC_FUNCTIONS): Add new rule asm_auto_mfpu.
4411         (ASM_CPU_SPEC): Use it if -mfpu is set to auto.
4412         * common/config/arm/arm-common.c (arm_asm_auto_mfpu): New function.
4414 2017-06-08  Tristan Gingold  <gindold@adacore.com>
4416         PR ada/81470
4417         * dwarf2out.c (dwarf2out_do_cfi_startproc): Only emit
4418         .cfi_personality or .cfi_lsda if the eh data format is dwarf2.
4420 2017-12-08  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4422         * config/sol2.h (SOLARIS_ATTRIBUTE_TABLE): Initialize new member
4423         of struct attribute_spec.
4425 2017-12-08  Julia Koval  <julia.koval@intel.com>
4427         * config/i386/avx512vnniintrin.h (_mm512_dpwssds_epi32,
4428         _mm512_mask_dpwssds_epi32, _mm512_maskz_dpwssds_epi32): New intrinsics.
4429         * config/i386/avx512vnnivlintrin.h (_mm256_dpwssds_epi32,
4430         _mm256_mask_dpwssds_epi32, _mm256_maskz_dpwssds_epi32,
4431         _mm_dpwssds_epi32, _mm_mask_dpwssds_epi32,
4432         _mm_maskz_dpwssds_epi32): Ditto.
4434 2017-12-08  Richard Biener  <rguenther@suse.de>
4436         PR tree-optimization/81303
4437         * tree-vect-stmts.c (vect_is_simple_cond): For invariant
4438         conditions try to create a comparison vector type matching
4439         the data vector type.
4440         (vectorizable_condition): Adjust.
4441         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
4442         Leave invariant conditions alone in case we can vectorize those.
4444 2017-12-08  Julia Koval  <julia.koval@intel.com>
4446         * config/i386/avx512vnniintrin.h (_mm512_dpwssd_epi32,
4447         _mm512_mask_dpwssd_epi32, _mm512_maskz_dpwssd_epi32): New intrinsics.
4448         * config/i386/avx512vnnivlintrin.h (_mm256_dpwssd_epi32,
4449         _mm256_mask_dpwssd_epi32, _mm256_maskz_dpwssd_epi32, _mm_dpwssd_epi32,
4450         _mm_mask_dpwssd_epi32, _mm_maskz_dpwssd_epi32): Ditto.
4452 2017-12-08  Julia Koval  <julia.koval@intel.com>
4454         * config/i386/avx512vnniintrin.h (_mm512_dpbusds_epi32,
4455         _mm512_mask_dpbusds_epi32, _mm512_maskz_dpbusds_epi32): New.
4456         * config/i386/avx512vnnivlintrin.h (_mm256_dpbusds_epi32,
4457         _mm256_mask_dpbusds_epi32, _mm256_maskz_dpbusds_epi32,
4458         _mm_dpbusds_epi32, _mm_mask_dpbusds_epi32,
4459         _mm_maskz_dpbusds_epi32): New intrinsics.
4461 2017-12-07  Sandra Loosemore  <sandra@codesourcery.com>
4463         * config/nios2/nios2.md (ld<bhw_uns>io): Add splitter for memory
4464         operand.
4465         (ld<bh>io_signed): Likewise.
4466         (st<bhw>io): Likewise.
4467         * config/nios2/predicates.md (ldstio_memory_operand): Allow
4468         SMALL_INT12 constant integer operand.
4470 2017-12-07  Sandra Loosemore  <sandra@codesourcery.com>
4472         * config/nios2/nios2.c (nios2_symbolic_constant_allowed):
4473         Rename to...
4474         (nios2_large_constant_allowed): ...this.  Adjust uses.
4475         (nios2_plus_symbolic_constant_p): Rename to...
4476         (nios2_plus_large_constant_p): ...this.  Adjust uses.
4477         (nios2_legitimate_address_p): Correct CONST_INT handling.
4478         (nios2_symbolic_memory_operand_p): Rename to...
4479         (nios2_large_constant_memory_operand_p): ...this.  Adjust uses.
4480         (nios2_large_constant_p): Check for large constant integers too.
4481         (nios2_split_large_constant): Handle constant integers.
4482         (nios2_split_symbolic_memory_operand): Rename to...
4483         (nios2_split_large_constant_memory_operand): ...this.  Adjust uses.
4484         (nios2_legitimize_constant_address): Handle constant integers.
4485         (r0rel_constant_p): Handle small constant integers.
4486         (nios2_print_operand_address): Handle r0-relative integer addresses.
4487         * config/nios2/nios2-protos.h: Adjust for renamed functions.
4488         * config/nios2/nios2.md: Adjust for renamed functions.
4490 2017-12-07  Andrew Waterman  <andrew@sifive.com>
4492         * config/riscv/riscv.c (TARGET_ASM_SELECT_SECTION): New define.
4493         (TARGET_HAVE_SRODATA_SECTION): New define.
4494         (riscv_select_section): New function.
4496 2017-12-08  Joseph Myers  <joseph@codesourcery.com>
4497             Alexander Monakov  <amonakov@ispras.ru>
4498             Jakub Jelinek  <jakub@redhat.com>
4500         PR target/81906
4501         * config/i386/i386.c (ix86_expand_rint): Handle flag_rounding_math.
4503 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4505         * config/aarch64/aarch64.c (aarch64_print_address_internal): Return
4506         a bool success value.  Don't call output_operand_lossage here.
4507         (aarch64_print_ldpstp_address): Return a bool success value.
4508         (aarch64_print_operand_address): Call output_addr_const if
4509         aarch64_print_address_internal fails.
4510         (aarch64_print_operand): Don't assert that the mode is 16 bytes for
4511         'y'; call output_operand_lossage instead.  Call output_operand_lossage
4512         if aarch64_print_ldpstp_address fails.
4514 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4516         * tree-vector-builder.h
4517         (tree_vector_builder::binary_encoded_nelts): Declare.
4518         * tree-vector-builder.c
4519         (tree_vector_builder::binary_encoded_nelts): New function.
4520         * fold-const.c (negate_expr_p): Likewise.
4521         (operand_equal_p, fold_checksum_tree): Likewise.
4522         * tree-loop-distribution.c (const_with_all_bytes_same): Likewise.
4523         * tree.c (integer_zerop, integer_onep, integer_all_onesp, real_zerop)
4524         (real_onep, real_minus_onep, add_expr, initializer_zerop): Likewise.
4525         (uniform_vector_p): Likewise.
4526         * varasm.c (const_hash_1, compare_constant): Likewise.
4527         * tree-ssa-ccp.c: Include tree-vector-builder.h.
4528         (valid_lattice_transition): Operate directly on the VECTOR_CST
4529         encoding.
4530         * ipa-icf.c: Include tree-vector-builder.h.
4531         (sem_variable::equals): Operate directly on the VECTOR_CST encoding.
4532         * print-tree.c (print_node): Print encoding of VECTOR_CSTs.
4534 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4536         * tree.c (build_vector): Delete.
4537         * tree.h (build_vector): Make static and move into the self-testing
4538         block.
4540 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4542         * vector-builder.h (vector_builder::derived): New const overload.
4543         (vector_builder::elt): New function.
4544         * tree-vector-builder.h (tree_vector_builder::type): New function.
4545         (tree_vector_builder::apply_step): Declare.
4546         * tree-vector-builder.c (tree_vector_builder::apply_step): New
4547         function.
4548         * gimple-fold.h (tree_vector_builder): Declare.
4549         (gimple_build_vector): Take a tree_vector_builder instead of a
4550         type and vector of elements.
4551         * gimple-fold.c (gimple_build_vector): Likewise.
4552         * tree-vect-loop.c (get_initial_def_for_reduction): Update call
4553         accordingly.
4554         (get_initial_defs_for_reduction): Likewise.
4555         (vectorizable_induction): Likewise.
4557 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4559         * tree-vector-builder.h
4560         (tree_vector_builder::new_binary_operation): Declare.
4561         * tree-vector-builder.c
4562         (tree_vector_builder::new_binary_operation): New function.
4563         * fold-const.c (fold_relational_const): Use it.
4564         (const_binop): Likewise.  Check that both input vectors have
4565         the same number of elements, thus excluding things like WIDEN_SUM.
4566         Check whether it is possible to operate directly on the encodings
4567         of stepped inputs.
4569 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4571         * fold-const.c (fold_negate_expr_1): Use tree_vector_builder and
4572         new_unary_operation, operating only on the encoded elements.
4573         (const_unop): Likewise.
4574         (exact_inverse): Likewise.
4575         (distributes_over_addition_p): New function.
4576         (const_binop): Use tree_vector_builder and new_unary_operation
4577         for combinations of VECTOR_CST and INTEGER_CST.  Operate only
4578         on the encoded elements unless the encoding is strided and the
4579         operation does not distribute over addition.
4580         (fold_convert_const):  Use tree_vector_builder and
4581         new_unary_operation.  Operate only on the encoded elements
4582         for truncating integer conversions, or for non-stepped encodings.
4584 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4586         * config/sparc/sparc.c: Include tree-vector-builder.h.
4587         (sparc_fold_builtin): Use tree_vector_builder instead of build_vector.
4588         * expmed.c: Include tree-vector-builder.h.
4589         (make_tree): Use tree_vector_builder instead of build_vector.
4590         * fold-const.c: Include tree-vector-builder.h.
4591         (const_binop): Use tree_vector_builder instead of build_vector.
4592         (const_unop): Likewise.
4593         (native_interpret_vector): Likewise.
4594         (fold_vec_perm): Likewise.
4595         (fold_ternary_loc): Likewise.
4596         * gimple-fold.c: Include tree-vector-builder.h.
4597         (gimple_fold_stmt_to_constant_1): Use tree_vector_builder instead
4598         of build_vector.
4599         * tree-ssa-forwprop.c: Include tree-vector-builder.h.
4600         (simplify_vector_constructor): Use tree_vector_builder instead
4601         of build_vector.
4602         * tree-vect-generic.c: Include tree-vector-builder.h.
4603         (add_rshift): Use tree_vector_builder instead of build_vector.
4604         (expand_vector_divmod): Likewise.
4605         (optimize_vector_constructor): Likewise.
4606         * tree-vect-loop.c: Include tree-vector-builder.h.
4607         (vect_create_epilog_for_reduction): Use tree_vector_builder instead
4608         of build_vector.  Explicitly use a stepped encoding for
4609         { 1, 2, 3, ... }.
4610         * tree-vect-slp.c: Include tree-vector-builder.h.
4611         (vect_get_constant_vectors): Use tree_vector_builder instead
4612         of build_vector.
4613         (vect_transform_slp_perm_load): Likewise.
4614         (vect_schedule_slp_instance): Likewise.
4615         * tree-vect-stmts.c: Include tree-vector-builder.h.
4616         (vectorizable_bswap): Use tree_vector_builder instead of build_vector.
4617         (vect_gen_perm_mask_any): Likewise.
4618         (vectorizable_call): Likewise.  Explicitly use a stepped encoding.
4619         * tree.c: (build_vector_from_ctor): Use tree_vector_builder instead
4620         of build_vector.
4621         (build_vector_from_val): Likewise.  Explicitly use a duplicate
4622         encoding.
4624 2017-12-07  Richard Sandiford  <richard.sandiford@arm.com>
4626         * doc/generic.texi (VECTOR_CST): Describe new representation of
4627         vector constants.
4628         * vector-builder.h: New file.
4629         * tree-vector-builder.h: Likewise.
4630         * tree-vector-builder.c: Likewise.
4631         * Makefile.in (OBJS): Add tree-vector-builder.o.
4632         * tree.def (VECTOR_CST): Update comment to refer to generic.texi.
4633         * tree-core.h (tree_base): Add a vector_cst field to the u union.
4634         (tree_vector): Change the number of elements to
4635         vector_cst_encoded_nelts.
4636         * tree.h (VECTOR_CST_NELTS): Redefine using TYPE_VECTOR_SUBPARTS.
4637         (VECTOR_CST_ELTS): Delete.
4638         (VECTOR_CST_ELT): Redefine using vector_cst_elt.
4639         (VECTOR_CST_LOG2_NPATTERNS, VECTOR_CST_NPATTERNS): New macros.
4640         (VECTOR_CST_NELTS_PER_PATTERN, VECTOR_CST_DUPLICATE_P): Likewise.
4641         (VECTOR_CST_STEPPED_P, VECTOR_CST_ENCODED_ELTS): Likewise.
4642         (VECTOR_CST_ENCODED_ELT): Likewise.
4643         (vector_cst_encoded_nelts): New function.
4644         (make_vector): Take the values of VECTOR_CST_LOG2_NPATTERNS and
4645         VECTOR_CST_NELTS_PER_PATTERN as arguments.
4646         (vector_cst_int_elt, vector_cst_elt): Declare.
4647         * tree.c: Include tree-vector-builder.h.
4648         (tree_code_size): Abort if passed VECTOR_CST.
4649         (tree_size): Update for new VECTOR_CST layout.
4650         (make_vector): Take the values of VECTOR_CST_LOG2_NPATTERNS and
4651         VECTOR_CST_NELTS_PER_PATTERN as arguments.
4652         (build_vector): Use tree_vector_builder.
4653         (vector_cst_int_elt, vector_cst_elt): New functions.
4654         (drop_tree_overflow): For VECTOR_CST, drop the TREE_OVERFLOW from the
4655         encoded elements and then create the vector in the canonical form.
4656         (check_vector_cst, check_vector_cst_duplicate, check_vector_cst_fill)
4657         (check_vector_cst_stepped, test_vector_cst_patterns): New functions.
4658         (tree_c_tests): Call test_vector_cst_patterns.
4659         * lto-streamer-out.c (DFS::DFS_write_tree_body): Handle the new
4660         VECTOR_CST fields.
4661         (hash_tree): Likewise.
4662         * tree-streamer-out.c (write_ts_vector_tree_pointers): Likewise.
4663         (streamer_write_tree_header): Likewise.
4664         * tree-streamer-in.c (lto_input_ts_vector_tree_pointers): Likewise.
4665         (streamer_alloc_tree): Likewise.  Update call to make_vector.
4666         * fold-const.c (fold_ternary_loc): Avoid using VECTOR_CST_ELTS.
4668 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4670         * selftest.h (ASSERT_TRUE_AT, ASSERT_FALSE_AT, ASSERT_EQ_AT)
4671         (ASSERT_NE, ASSERT_PRED1): Add underscores to local variable names
4672         * selftest-rtl.h (ASSERT_RTX_EQ, ASSERT_RTX_PTR_EQ): Likewise.
4674 2017-12-07  Bin Cheng  <bin.cheng@arm.com>
4675             Richard Biener  <rguenther@suse.de>
4677         PR tree-optimization/81303
4678         * Makefile.in (gimple-loop-interchange.o): New object file.
4679         * common.opt (floop-interchange): Reuse the option from graphite.
4680         * doc/invoke.texi (-floop-interchange): Ditto.  New document for
4681         -floop-interchange and mention it for -O3.
4682         * opts.c (default_options_table): Enable -floop-interchange at -O3.
4683         * gimple-loop-interchange.cc: New file.
4684         * params.def (PARAM_LOOP_INTERCHANGE_MAX_NUM_STMTS): New parameter.
4685         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): New parameter.
4686         * passes.def (pass_linterchange): New pass.
4687         * timevar.def (TV_LINTERCHANGE): New time var.
4688         * tree-pass.h (make_pass_linterchange): New declaration.
4689         * tree-ssa-loop-ivcanon.c (create_canonical_iv): Change to external
4690         interchange.  Record IV before/after increment in new parameters.
4691         * tree-ssa-loop-ivopts.h (create_canonical_iv): New declaration.
4692         * tree-vect-loop.c (vect_is_simple_reduction): Factor out reduction
4693         path check into...
4694         (check_reduction_path): ...New function here.
4695         * tree-vectorizer.h (check_reduction_path): New declaration.
4697 2017-12-07  Vladimir Makarov  <vmakarov@redhat.com>
4699         PR target/83252
4700         PR rtl-optimization/80818
4701         * lra.c (add_regs_to_insn_regno_info): Make a hard reg in CLOBBER
4702         always early clobbered.
4703         * lra-lives.c (process_bb_lives): Check input hard regs for early
4704         clobbered non-operand hard reg.
4706 2017-12-07  Jakub Jelinek  <jakub@redhat.com>
4708         PR middle-end/83164
4709         * tree-cfg.c (verify_gimple_assign_binary): Don't require
4710         types_compatible_p, just that TYPE_MODE is the same.
4712 2017-12-07  Martin Sebor  <msebor@redhat.com>
4714         PR c/81544
4715         * attribs.c (empty_attribute_table): Initialize new member of
4716         struct attribute_spec.
4717         (decl_attributes): Add argument.  Handle mutually exclusive
4718         combinations of attributes.
4719         (selftests::test_attribute_exclusions): New function.
4720         (selftests::attribute_c_tests): Ditto.
4721         * attribs.h (decl_attributes): Add default argument.
4722         * selftest.h (attribute_c_tests): Declare.
4723         * selftest-run-tests.c (selftest::run_tests): Call attribute_c_tests.
4724         * tree-core.h (attribute_spec::exclusions, exclude): New type and
4725         member.
4726         * doc/extend.texi (Common Function Attributes): Update const and pure.
4727         * config/alpha/alpha.c (vms_attribute_table): Initialize new member
4728         of struct attribute_spec.
4729         * config/arc/arc.c (arc_attribute_table): Same.
4730         * config/arm/arm.c (arm_attribute_table): Same.
4731         * config/avr/avr.c ( avr_attribute_table): Same.
4732         * config/bfin/bfin.c (bfin_attribute_table): Same.
4733         * config/cr16/cr16.c (cr16_attribute_table): Same.
4734         * config/epiphany/epiphany.c (epiphany_attribute_table): Same.
4735         * config/h8300/h8300.c (h8300_attribute_table): Same.
4736         * config/i386/i386.c (ix86_attribute_table): Same.
4737         * config/ia64/ia64.c (ia64_attribute_table): Same.
4738         * config/m32c/m32c.c (m32c_attribute_table): Same.
4739         * config/m32r/m32r.c (m32r_attribute_table): Same.
4740         * config/m68k/m68k.c (m68k_attribute_table): Same.
4741         * config/mcore/mcore.c (mcore_attribute_table): Same.
4742         * config/microblaze/microblaze.c (microblaze_attribute_table): Same.
4743         * config/mips/mips.c (mips_attribute_table): Same.
4744         * config/msp430/msp430.c (msp430_attribute_table): Same.
4745         * config/nds32/nds32.c (nds32_attribute_table): Same.
4746         * config/nvptx/nvptx.c (nvptx_attribute_table): Same.
4747         * config/powerpcspe/powerpcspe.c (rs6000_attribute_table): Same.
4748         * config/rl78/rl78.c (rl78__attribute_table): Same.
4749         * config/rs6000/rs6000.c (rs6000_attribute_table): Same.
4750         * onfig/rx/rx.c (rx_attribute_table): Same.
4751         * config/s390/s390.c (s390_handle_vectorbool_attribute): Same.
4752         * config/sh/sh.c (sh_attribute_table): Same.
4753         * config/sparc/sparc.c (sparc_attribute_table): Same.
4754         * config/spu/spu.c (spu_attribute_table): Same.
4755         * config/stormy16/stormy16.c (xstormy16_attribute_table): Same.
4756         * config/v850/v850.c (v850_attribute_table): Same.
4757         * config/visium/visium.c (visium_attribute_table): Same.
4759 2017-12-07  Tamar Christina  <tamar.christina@arm.com>
4761         PR target/82641
4762         * config/arm/arm.c (INCLUDE_STRING): Define.
4763         (arm_last_printed_arch_string, arm_last_printed_fpu_string): New.
4764         (arm_declare_function_name): Conservatively emit .arch, .arch_extensions
4765         and .fpu.
4767 2017-12-07  Michael Matz  <matz@suse.de>
4769         Add unroll and jam pass
4771         * gimple-loop-jam.c: New file.
4772         * Makefile.in (OBJS): Add gimple-loop-jam.o.
4773         * common.opt (funroll-and-jam): New option.
4774         * opts.c (default_options_table): Add unroll-and-jam at -O3.
4775         * params.def (PARAM_UNROLL_JAM_MIN_PERCENT): New param.
4776         (PARAM_UNROLL_JAM_MAX_UNROLL): Ditto.
4777         * passes.def: Add pass_loop_jam.
4778         * timevar.def (TV_LOOP_JAM): Add.
4779         * tree-pass.h (make_pass_loop_jam): Declare.
4780         * cfgloop.c (flow_loop_tree_node_add): Add AT argument.
4781         * cfgloop.h (flow_loop_tree_node_add): Adjust declaration.
4782         * cfgloopmanip.c (duplicate_loop): Add AT argument, adjust call
4783         to flow_loop_tree_node_add.
4784         (duplicate_subloops, copy_loops_to): Append to sibling list.
4785         * cfgloopmanip.h: (duplicate_loop): Adjust declaration.
4786         * doc/invoke.texi (-funroll-and-jam): Document new option.
4787         (unroll-jam-min-percent, unroll-jam-max-unroll): Document new params.
4789 2017-12-07  Richard Biener  <rguenther@suse.de>
4791         PR tree-optimization/83296
4792         PR tree-optimization/67769
4793         * tree-ssa-phiopt.c (conditional_replacement): Do not reset
4794         flow sensitive info in an unrelated BB.
4795         (value_replacement): Use reset_flow_sensitive_info.
4796         (minmax_replacement): Reset flow sensitive info on the def
4797         we move.  Do not reset flow sensitive info in the whole BB
4798         we move the stmt to.
4799         (abs_replacement): Likewise.
4801 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4803         PR target/43871
4804         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
4805         rs6000_cpu to the given -mcpu=, or to the default processor.
4807 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4809         * config/rs6000/rs6000.h (rs6000_cpu_attr): Delete.
4810         * config/rs6000/rs6000.c (rs6000_variable_issue_1): Use rs6000_tune
4811         instead of rs6000_cpu_attr.
4812         (rs6000_adjust_cost): Ditto.
4813         (is_microcoded_insn): Ditto.
4814         (rs6000_adjust_priority): Ditto.
4815         (rs6000_issue_rate): Ditto.
4816         (rs6000_use_sched_lookahead): Ditto.
4817         (rs6000_use_sched_lookahead_guard): Ditto.
4818         (rs6000_sched_reorder): Ditto.
4819         (force_new_group): Ditto.
4820         * config/rs6000/rs6000.md (cpu attribute): Ditto.
4821         (group_ending_nop): Ditto.
4823 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4825         * config/rs6000/rs6000.opt (rs6000_tune): New variable.
4826         * config/rs6000/rs6000.c (rs6000_option_override_internal): Also set
4827         rs6000_tune.  Use rs6000_tune instead of rs6000_cpu where appropriate.
4828         (rs6000_loop_align): Use rs6000_tune instead of rs6000_cpu where
4829         appropriate.
4830         (rs6000_reassociation_width): Ditto.
4831         (rs6000_emit_epilogue): Ditto.
4832         (rs6000_adjust_cost): Ditto.
4833         (is_microcoded_insn): Ditto.
4834         (is_cracked_insn): Ditto.
4835         (rs6000_adjust_priority): Ditto.
4836         (rs6000_sched_reorder): Ditto.
4837         (rs6000_sched_reorder2): Ditto.
4838         (insn_must_be_first_in_group): Ditto.
4839         (insn_must_be_last_in_group): Ditto.
4840         (rs6000_register_move_cost): Ditto.
4841         * config/rs6000/rs6000.h (rs6000_cpu_attr): Use rs6000_tune instead of
4842         rs6000_cpu.
4844 2017-12-07  Julia Koval  <julia.koval@intel.com>
4846         * config.gcc: Add vaesintrin.h.
4847         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI_V64QI):
4848         New type.
4849         * config/i386/i386-builtin.def (__builtin_ia32_vaesdec_v16qi,
4850         __builtin_ia32_vaesdec_v32qi, __builtin_ia32_vaesdec_v64qi):
4851         New builtins.
4852         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
4853         * config/i386/immintrin.h: Include vaesintrin.h.
4854         * config/i386/sse.md (vaesdec_<mode>): New pattern.
4855         * config/i386/vaesintrin.h (_mm256_aesdec_epi128, _mm512_aesdec_epi128,
4856         _mm_aesdec_epi128): New intrinsics.
4858 2017-12-06  David Malcolm  <dmalcolm@redhat.com>
4860         * Makefile.in (C_COMMON_OBJS): Add c-family/c-spellcheck.o.
4861         * spellcheck-tree.c (find_closest_macro_cpp_cb): Move to
4862         c-family/c-spellcheck.cc.
4863         (best_macro_match::best_macro_match): Likewise.
4864         * spellcheck-tree.h
4865         (struct edit_distance_traits<cpp_hashnode *>): Move to
4866         c-family/c-spellcheck.h.
4867         (class best_macro_match): Likewise.
4869 2017-12-06  Jakub Jelinek  <jakub@redhat.com>
4871         PR tree-optimization/83293
4872         * gimple-ssa-strength-reduction.c (insert_initializers): Use
4873         GSI_NEW_STMT instead of GSI_SAME_STMT in gsi_insert_after that
4874         might insert into empty bb.
4876         PR sanitizer/81281
4877         * match.pd ((T)(P + A) - (T)P -> (T) A): Split into separate
4878         simplify for plus with :c added, and pointer_plus without that.
4879         ((T)P - (T)(P + A) -> -(T) A): Likewise.  If type is integral
4880         with undefined overflow and the conversion is not widening,
4881         perform negation in utype and only convert to type afterwards.
4882         ((T)(P + A) - (T)(P + B) -> (T)A - (T)B): Split into separate
4883         simplify for plus with :c added, and pointer_plus without that.
4884         If type is integral with undefined overflow and the conversion is
4885         not widening, perform minus in utype and only convert to type
4886         afterwards.  Move the last pointer_diff_expr simplify into the
4887         two outermost ifs.
4889 2017-12-06  Martin Sebor  <msebor@redhat.com>
4891         PR tree-optimization/82646
4892         * builtins.c (maybe_emit_chk_warning): Use size as the bound for
4893         strncpy, not maxlen.
4895 2017-12-06  Martin Sebor  <msebor@redhat.com>
4897         * doc/invoke.texi (-Wstringop-truncation): Mention attribute
4898         nonstring.
4900         PR tree-optimization/83075
4901         * tree-ssa-strlen.c (handle_builtin_stxncpy): Avoid assuming
4902         strncat/strncpy don't change length of source string.
4904 2017-12-06  Eric Botcazou  <ebotcazou@adacore.com>
4906         Revert
4907         2017-11-29  Martin Aberg  <maberg@gaisler.com>
4909         * config/sparc/sparc.md (divdf3_fix): Add NOP and adjust length
4910         to prevent b2bst errata sequence.
4911         (sqrtdf2_fix): Likewise.
4913 2017-12-06  Jakub Jelinek  <jakub@redhat.com>
4915         PR tree-optimization/81945
4916         * cfgloop.h (FOR_EACH_LOOP_FN): Use FN instead of hardcoding fn.
4917         * tree-cfg.c (move_sese_region_to_fn): If any of the loops moved
4918         to dest_cfun has orig_loop_num set, either remap it to the new
4919         loop number if the loop got moved too, or clear it.
4921 2017-12-05  Steve Ellcey  <sellcey@cavium.com>
4923         * config/aarch64/thunderx2-t99.md (thunderx2t99_branch): Add trap
4924         to reservation.
4925         (thunderx2t99_nothing): New insn reservation.
4926         (thunderx2t99_mrs): New insn reservation.
4927         (thunderx2t99_multiple): New insn reservation.
4928         (thunderx2t99_alu_basi): Add bfx to reservation.
4929         (thunderx2t99_fp_cmp): Add fccmps and fccmpd to reservation.
4931 2017-12-05  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
4933         PR target/82248
4934         * config/arm/arm.md (probe_stack) : Use the 'o' constraint.
4936 2017-12-05  Bin Cheng  <bin.cheng@arm.com>
4938         * tree-ssa-dce.c (simple_dce_from_worklist): Move and rename from
4939         tree-ssa-pre.c::remove_dead_inserted_code.
4940         * tree-ssa-dce.h: New file.
4941         * tree-ssa-pre.c (tree-ssa-dce.h): Include new header file.
4942         (remove_dead_inserted_code): Move and rename to function
4943         tree-ssa-dce.c::simple_dce_from_worklist.
4944         (pass_pre::execute): Update use.
4946 2017-12-05  Richard Biener  <rguenther@suse.de>
4948         PR tree-optimization/83277
4949         * graphite-isl-ast-to-gimple.c (should_copy_to_new_region): Make sure
4950         to code-gen liveout vars.
4952 2017-12-05  Richard Sandiford  <richard.sandiford@linaro.org>
4954         * config/aarch64/aarch64-simd.md (aarch64_simd_bsldi_internal)
4955         (aarch64_simd_bsldi_alt): Check REG_P before GP_REGNUM_P.
4956         (aarch64_cm<optab>di, aarch64_cmtstdi): Add leading "&&" to
4957         split condition.
4959 2017-12-05  Max Filippov  <jcmvbkbc@gmail.com>
4961         * config/xtensa/xtensa.c (xtensa_asan_shadow_offset): New
4962         function.
4963         (TARGET_ASAN_SHADOW_OFFSET): New macro definition.
4964         * config/xtensa/xtensa.h (FRAME_GROWS_DOWNWARD): Set to 1 if
4965         ASAN is enabled.
4967 2017-12-05  Richard Biener  <rguenther@suse.de>
4969         * timevar.def (TV_TREE_RECIP, TV_TREE_SINCOS, TV_TREE_WIDEN_MUL): Add.
4970         * tree-ssa-math-opts.c (pass_data_cse_reciprocal): Use TV_TREE_RECIP.
4971         (pass_data_cse_sincos): Use TV_TREE_SINCOS.
4972         (pass_data_optimize_widening_mul): Use TV_TREE_WIDEN_MUL.
4974 2017-12-05  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4976         * dbxout.c (dbxout_block): Grow buf to 30 bytes.
4978 2017-12-05  Martin Liska  <mliska@suse.cz>
4979             Jakub Jelinek  <jakub@redhat.com>
4981         * doc/invoke.texi: Document the options.
4982         * flag-types.h (enum sanitize_code): Add
4983         SANITIZE_POINTER_COMPARE and SANITIZE_POINTER_SUBTRACT.
4984         * ipa-inline.c (sanitize_attrs_match_for_inline_p): Add handling
4985         of SANITIZE_POINTER_COMPARE and SANITIZE_POINTER_SUBTRACT.
4986         * opts.c: Define new sanitizer options.
4987         * sanitizer.def (BUILT_IN_ASAN_POINTER_COMPARE): Likewise.
4988         (BUILT_IN_ASAN_POINTER_SUBTRACT): Likewise.
4990 2017-12-05  Julia Koval  <julia.koval@intel.com>
4992         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VNNI_SET,
4993         OPTION_MASK_ISA_AVX512VNNI_UNSET): New.
4994         (ix86_handle_option): Handle -mavx512vnni.
4995         * config/i386/cpuid.h (bit_AVX512VNNI): New bit.
4996         * config/i386/driver-i386.c (host_detect_local_cpu): Handle new bit.
4997         * config/i386/i386-c (__AVX512VNNI__): New.
4998         * config/i386/i386.c (ix86_target_string): Handle new option.
4999         (ix86_valid_target_attribute_inner_p): Handle new option.
5000         * config/i386/i386.h (TARGET_AVX512VNNI, TARGET_AVX512VNNI_P): New.
5001         * config/i386/i386.opt (mavx512vnni): New option.
5003 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
5005         PR target/81616
5006         * athlon.md: Disable for generic.
5007         * haswell.md: Enable for generic.
5008         * i386.c (ix86_sched_init_global): Add core hooks for generic.
5009         * x86-tune-sched.c (ix86_issue_rate): Increase issue rate for generic
5010         to 4.
5011         (ix86_adjust_cost): Move generic to haswell path.
5013 2017-12-04  Eric Botcazou  <ebotcazou@adacore.com>
5015         * config/sparc/sparc.c (sparc_do_work_around_errata): Use mem_ref
5016         instead of MEM_P in a couple more places.  Fix formatting issues.
5018 2017-12-04  Jim Wilson  <jimw@sifive.com>
5020         * config/riscv/riscv.c (riscv_for_each_saved_reg): Use GP_REG_LAST
5021         instead of GP_REG_LAST-1.
5022         (riscv_adjust_libcall_cfi_prologue): Likewise.
5023         (riscv_adjust_libcall_cri_epilogue): Likewise.
5024         * config/riscv/riscv.h (CALL_USED_REGISTERS): Change a6 to t6 in
5025         comment.
5027 2017-12-04  Luis Machado  <luis.machado@linaro.org>
5029         * ipa-pure-const.c (check_decl): Add missing newline.
5030         (state_from_flags): Likewise.
5032 2017-12-04  Jeff Law  <law@redhat.com>
5034         PR tree-optimizatin/78496
5035         * gimple-ssa-evrp-analyze.h
5036         (evrp_range_analyzer::get_vr_values): Simplify.
5037         * gimple-ssa-evrp-analyze.c: Corresponding changes.
5038         * tree-ssa-dom.c: Include alloc-pool.h, tree-vrp.h, vr-values.h
5039         and gimple-ssa-evrp-analyze.h.
5040         (dom_opt_dom_walker class): Add evrp_range_analyzer member.
5041         (simplify_stmt_for_jump_threading): Copy a blob of code from
5042         tree-vrp.c to use ranges to simplify statements.
5043         (dom_opt_dom_walker::before_dom_children): Call
5044         evrp_range_analyzer::{enter,record_ranges_from_stmt} methods.
5045         (dom_opt_dom_walker::after_dom_children): Similarly for
5046         evrp_range_analyzer::leave.
5047         (dom_opt_dom_walker::optimize_stmt): Use EVRP ranges to optimize
5048         conditionals.
5050         * gimple-ssa-evrp-analyze.c
5051         (evrp_range_analyzer::extract_range_from_stmt):  Always use
5052         vr_values::update_value_range so preexisting range info is
5053         medged with new range info, even if the new range is VR_VARYING.
5055 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
5057         * combine.c: Adjust comment.
5058         (use_crosses_set_p): Delete.
5059         (can_combine_p): Use modified_between_p instead of use_crosses_set_p.
5060         (try_combine): Ditto.
5062 2017-12-04  Richard Biener  <rguenther@suse.de>
5064         PR tree-optimization/83255
5065         * graphite-isl-ast-to-gimple.c (translate_isl_ast_node_for):
5066         Re-add zero-iteration check.
5068 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
5070         PR rtl-optimization/83245
5071         * lra.c (collect_non_operand_hard_regs): Treat clobbers of non-operand
5072         hard registers as earlyclobber, also if not in an asm.
5074 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
5076         PR bootstrap/83265
5077         Revert
5078         2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5080         PR target/43871
5081         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
5082         rs6000_cpu based on cpu_index, not tune_index.
5084 2017-12-04  Richard Biener  <rguenther@suse.de>
5086         PR tree-optimization/83238
5087         * graphite-scop-detection.c (scop_detection::merge_sese): Make
5088         code match comment, rejecting invalid SESE regions.
5090 2017-12-03  John David Anglin  <danglin@gcc.gnu.org>
5092         * config/pa/pa.c (pa_legitimate_address_p): For scaled indexing,
5093         require base operand is a REG_POINTER prior to reload on targets
5094         with non-equivalent space registers.
5096 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
5098         * ipa-cp.c (ipcp_lattice<valtype>::print): Update dumping.
5099         (update_specialized_profile): Fix updating of counts.
5100         (perhaps_add_new_callers): Likewise.
5102 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
5104         PR target/81616
5105         * x86-tune.def: Remove obsolette FIXMEs.
5106         (X86_TUNE_PARTIAL_FLAG_REG_STALL): Disable for generic
5107         (X86_TUNE_FUSE_CMP_AND_BRANCH_32, X86_TUNE_FUSE_CMP_AND_BRANCH_64,
5108         X86_TUNE_FUSE_CMP_AND_BRANCH_SOFLAGS, X86_TUNE_FUSE_ALU_AND_BRANCH):
5109         Enable for generic.
5110         (X86_TUNE_PAD_RETURNS): Disable for generic.
5112 2017-12-02  Jakub Jelinek  <jakub@redhat.com>
5114         PR tree-optimization/83170
5115         PR tree-optimization/83241
5116         * gimple-ssa-store-merging.c
5117         (imm_store_chain_info::try_coalesce_bswap): Update vuse field from
5118         gimple_vuse (ins_stmt) in case it has changed.
5119         (imm_store_chain_info::output_merged_store): Likewise.
5121         * tree-chkp.c (chkp_compute_bounds_for_assignment): Handle
5122         POINTER_DIFF_EXPR.
5124         PR c++/81212
5125         * tree-cfg.c (pass_warn_function_return::execute): Handle
5126         __builtin_ubsan_handle_missing_return like __builtin_unreachable
5127         with BUILTINS_LOCATION.
5129         PR target/78643
5130         PR target/80583
5131         * expr.c (get_inner_reference): If DECL_MODE of a non-bitfield
5132         is BLKmode for vector field with vector raw mode, use TYPE_MODE
5133         instead of DECL_MODE.
5135         * config/i386/i386-protos.h (standard_sse_constant_opcode): Change
5136         last argument to rtx pointer.
5137         * config/i386/i386.c (standard_sse_constant_opcode): Replace X argument
5138         with OPERANDS.  For AVX+ 128-bit VEX encoded instructions over 256-bit
5139         or 512-bit.  If setting EXT_REX_SSE_REG_P, use EVEX encoded insn
5140         depending on the chosen ISAs.
5141         * config/i386/i386.md (*movxi_internal_avx512f, *movoi_internal_avx,
5142         *movti_internal, *movdi_internal, *movsi_internal, *movtf_internal,
5143         *movdf_internal, *movsf_internal): Adjust standard_sse_constant_opcode
5144         callers.
5145         * config/i386/sse.md (mov<mode>_internal): Likewise.
5146         * config/i386/mmx.md (*mov<mode>_internal): Likewise.
5148 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5150         * doc/invoke.texi (-dp): Say that instruction cost is printed as well.
5152 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5154         * config/rs6000/rs6000-protos.h (emit_fusion_addis): Remove last two
5155         parameters from prototype.
5156         * config/rs6000/rs6000.c (emit_fusion_addis): Remove last two
5157         parameters.  Don't print a comment.
5158         (emit_fusion_gpr_load): Adjust.
5159         (emit_fusion_load_store): Adjust.
5160         * config/rs6000/rs6000.md (*fusion_p9_<mode>_constant): Adjust.
5161         * config/rs6000/vsx.md (two peepholes): Print the "vector load fusion"
5162         comment on the second line.
5164 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5166         PR target/43871
5167         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
5168         rs6000_cpu based on cpu_index, not tune_index.
5170 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5172         * final.c (output_asm_name): Print insn_cost.  Shorten output.  Print
5173         which_alternative instead of which_alternative + 1.
5174         (output_asm_insn): Print an extra tab if the template is short.
5176 2017-12-01  Jim Wilson  <jimw@sifive.com>
5178         * common.opt (use_gnu_debug_info_extensions): Delete DWARF_DEBUG from
5179         comment.
5180         * config/vx-common.h (DWARF_DEBUGGING_INFO): Delete undef.
5181         * doc/tm.texi.in (PREFERRED_DEBUGGING_TYPE): Delete DWARF_DEBUG
5182         reference.
5183         * doc/tm.texi: Regenerate.
5185 2017-12-01  Michael Meissner  <meissner@linux.vnet.ibm.com>
5187         PR target/81959
5188         * config/rs6000/rs6000.c (rs6000_address_for_fpconvert): Check for
5189         whether we can allocate pseudos before trying to fix an address.
5190         * config/rs6000/rs6000.md (float_<mode>si2_hw): Make sure the
5191         memory address is indexed or indirect.
5192         (floatuns_<mode>si2_hw2): Likewise.
5194 2017-12-01  Jason Merrill  <jason@redhat.com>
5196         * Makefile.in (TAGS): Add c-family/*.cc.
5198 2017-12-01  Wilco Dijkstra  <wdijkstr@arm.com>
5200         * config/aarch64/aarch64.md (call_insn): Use %c rather than %a.
5201         (call_value_insn): Likewise.
5202         (sibcall_insn): Likewise.
5203         (sibcall_value_insn): Likewise.
5204         (movsi_aarch64): Likewise.
5205         (movdi_aarch64): Likewise.
5206         (add_losym_): Likewise.
5207         (ldr_got_small_): Likewise.
5208         (ldr_got_small_sidi): Likewise.
5209         (ldr_got_small_28k_): Likewise.
5210         (ldr_got_small_28k_sidi): Likewise.
5211         * config/aarch64/aarch64.c (aarch64_print_address_internal):
5212         Move output_addr_const to symbolic case. Add error check.
5214 2017-12-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
5216         * config/s390/predicates.md (plus16_Q_operand): New predicate.
5217         * config/s390/s390.md: Disable MVC merging peephole if it would
5218         disable operand forwarding.
5219         (new peephole2): Split MVCs if it would turn them into up to 2
5220         forwardable MVCs.
5222 2017-12-01  Richard Biener  <rguenther@suse.de>
5224         PR tree-optimization/83232
5225         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Fix
5226         detection of same access. Instead of breaking the group here
5227         do not consider the duplicate.  Add comment explaining real fix.
5229 2017-12-01  Jonathan Wakely  <jwakely@redhat.com>
5231         * doc/md.texi (Insn Splitting): Fix "central flowgraph" typo.
5233 2017-12-01  Sudakshina Das  <sudi.das@arm.com>
5235         * config/arm/vfp.md (*movhf_vfp_fp16): Add conds attribute.
5237 2017-12-01  Jakub Jelinek  <jakub@redhat.com>
5239         * function.h (struct function): Remove cilk_frame_decl,
5240         is_cilk_function and calls_cilk_spawn fields.
5241         * tree-inline.h (struct copy_body_data): Remove remap_var_for_cilk
5242         field.
5243         * omp-simd-clone.c (simd_clone_clauses_extract): Don't clear
5244         cilk_elemental field.
5245         * cgraph.h (struct cgraph_simd_clone): Remove cilk_elemental field.
5246         * target.def: Adjust comment.
5247         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
5248         Don't test cilk_elemental.
5250         PR tree-optimization/83233
5251         * gimple-ssa-store-merging.c (nop_stats, bswap_stats): Use
5252         bswap_stat name for the struct.
5254         PR c/79153
5255         * tree.h (SWITCH_BREAK_LABEL_P): Define.
5256         * gimplify.c (collect_fallthrough_labels): Handle GIMPLE_BIND
5257         starting with a GIMPLE_SWITCH and ending with GIMPLE_LABEL with
5258         SWITCH_BREAK_LABEL_P set on the label.
5259         (gimplify_switch_expr): Set SWITCH_BREAK_LABEL_P on the label
5260         added for default case if it was missing and not all cases covered.
5261         Wrap GIMPLE_SWITCH and the switch_body_seq into a GIMPLE_BIND if
5262         switch_body_seq ends with a GIMPLE_LABEL with SWITCH_BREAK_LABEL_P
5263         set on the label.
5264         * tree-chrec.c (evolution_function_is_univariate_p): Add return true;
5265         to avoid -Wimplicit-fallthrough warning.
5266         * config/i386/i386.c (ix86_expand_special_args_builtin): Add
5267         FALLTHRU comment to avoid -Wimplicit-fallthrough warning.
5269         PR tree-optimization/83221
5270         * tree-ssa-reassoc.c (sort_by_operand_rank): Shift bb_rank
5271         down by 16.
5272         (init_reassoc): Formatting fix.
5274         PR sanitizer/81275
5275         * tree-cfg.c (group_case_labels_stmt): Don't optimize away
5276         C++ FE implicitly added __builtin_unreachable () until -Wreturn-type
5277         is diagnosed.
5279         PR sanitizer/83219
5280         * tree-cfg.c: Include asan.h.
5281         (gimple_seq_unreachable_p): Return false for -fsanitize=unreachable.
5283 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5285         * config/i386/i386.md: Fix AVX512 register width in AVX512 instruction.
5287 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5289         * config/i386/i386.c (standard_sse_constant_opcode): Fix wrong form for
5290         vpcmpeqd instruction.
5292 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5294         * config/i386/i386.c (standard_sse_constant_opcode): Fix
5295         registers type for 128bit mode.
5297 2017-11-30  Jakub Jelinek  <jakub@redhat.com>
5299         * spellcheck-tree.c (test_find_closest_identifier): Use ; instead
5300         of ;;.
5301         * gengtype-state.c (read_state_pair): Likewise.
5302         * gimple-fold.c (gimple_fold_builtin_string_compare): Likewise.
5303         * sel-sched-dump.c (dump_insn_rtx_1): Likewise.
5304         * ipa-cp.c (intersect_aggregates_with_edge): Likewise.
5305         * ifcvt.c (noce_try_store_flag_constants): Likewise.
5306         * tree-ssa-ccp.c (ccp_finalize): Likewise.
5307         * omp-grid.c (grid_process_kernel_body_copy): Likewise.
5308         * builtins.c (fold_builtin_3): Likewise.
5309         * graphite-scop-detection.c
5310         (scop_detection::stmt_has_simple_data_refs_p): Likewise.
5311         * hsa-gen.c (hsa_function_representation::hsa_function_representation):
5312         Likewise.
5314 2017-12-01  Maxim Ostapenko  <m.ostapenko@samsung.com>
5316         PR sanitizer/81697
5317         * asan.c (asan_protect_global): Add new ignore_decl_rtl_set_p
5318         parameter. Return true if ignore_decl_rtl_set_p is true and other
5319         conditions are satisfied.
5320         * asan.h (asan_protect_global): Add new parameter.
5321         * varasm.c (categorize_decl_for_section): Pass true as second parameter
5322         to asan_protect_global calls.
5324 2017-11-30  Jim Wilson  <jimw@sifive.com>
5326         * doc/invoke.texi (RISC-V Options): Delete nonexistent -mmemcpy and
5327         -mno-memcpy options.  For -mplt, -mfdiv, -mdiv, -msave-restore, and
5328         -mstrict-align, add info on default value.  Delete redundant lines for
5329         -mabi.  Add missing -mexplicit-relocs docs.
5331 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5333         * config/arc/arc.md (trap): New pattern.
5335 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5337         * config/arc/arc.c (hwloop_optimize): Prevent the last ZOL
5338         instruction to end into a delay slot.
5339         * config/arc/arc.md (cond_delay_insn): Check if the instruction
5340         can be placed into a delay slot against reg_note.
5342 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5344         * config/arc/arc.c (hwloop_optimize): Update hw-loop's end/start
5345         labels number of usages.
5347 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5349         * config/arc/arc.c (arc_cannot_substitue_mem_equiv_p): New
5350         function.
5351         (TARGET_CANNOT_SUBSTITUTE_MEM_EQUIV_P): Define.
5353 2017-11-30  Jakub Jelinek  <jakub@redhat.com>
5355         PR target/83210
5356         * internal-fn.c (expand_mul_overflow): Optimize unsigned
5357         multiplication by power of 2 constant into two shifts + comparison.
5359 2017-11-30  Jan Hubicka  <hubicka@ucw.cz>
5361         PR target/81616
5362         * config/i386/x86-tune-costs.h (generic_cost): Revise for modern CPUs.
5364 2017-11-30  Richard Biener  <rguenther@suse.de>
5366         PR tree-optimization/83202
5367         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Add
5368         allow_peel argument and guard peeling.
5369         (canonicalize_loop_induction_variables): Likewise.
5370         (canonicalize_induction_variables): Pass false.
5371         (tree_unroll_loops_completely_1): Pass unroll_outer to disallow
5372         peeling from cunrolli.
5374 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5376         * combine.c (try_combine): Print a message to dump file whenever
5377         I0, I1, or I2 cannot be combined into I3.
5379 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5381         PR rtl-optimization/83156
5382         PR rtl-optimization/82621
5383         * combine.c (try_combine): Don't split an I2 if one of the dests is
5384         set again before I3.  Allow unused dests.
5386 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5388         * config/rs6000/rs6000.md (*add<mode>3_carry_in_internal2): New.
5390 2017-11-29  Vladimir Makarov  <vmakarov@redhat.com>
5392         PR rtl-optimization/80818
5393         * lra.c (collect_non_operand_hard_regs): New arg insn.  Pass it
5394         recursively.  Use insn code for clobber.
5395         (lra_set_insn_recog_data): Pass the new arg to
5396         collect_non_operand_hard_regs.
5397         (add_regs_to_insn_regno_info): Pass insn instead of uid.  Use insn
5398         code for clobber.
5399         (lra_update_insn_regno_info): Pass insn to
5400         add_regs_to_insn_regno_info.
5402 2017-11-29  Jim Wilson  <jimw@sifive.com>
5403             Andrew Waterman  <andrew@sifive.com>
5405         * config/riscv/riscv.c (SINGLE_SHIFT_COST): New.
5406         (riscv_rtx_costs): Case ZERO_EXTRACT, match new pattern, and return
5407         SINGLE_SHIFT_COST.  Case LT and ZERO_EXTEND, likewise.  Case ASHIFT,
5408         use SINGLE_SHIFT_COST.
5409         * config/riscv/riscv.md (lshrsi3_zero_extend_1): New.
5410         (lshrsi3_zero_extend_2, lshrsi3_zero_extend_3): New.
5412 2017-11-29  Julia Koval  <julia.koval@intel.com>
5414         * config/i386/avx512vbmi2intrin.h (_mm512_shldv_epi16,
5415         _mm512_mask_shldv_epi16, _mm512_maskz_shldv_epi16, _mm512_shldv_epi32,
5416         _mm512_mask_shldv_epi32, _mm512_maskz_shldv_epi32, _mm512_shldv_epi64,
5417         _mm512_mask_shldv_epi64, _mm512_maskz_shldv_epi64): New intrinsics.
5418         * config/i386/avx512vbmi2vlintrin.h (_mm256_shldv_epi16,
5419         _mm256_mask_shldv_epi16, _mm256_maskz_shldv_epi16, _mm256_shldv_epi32,
5420         _mm256_mask_shldv_epi32, _mm256_maskz_shldv_epi32, _mm256_shldv_epi64,
5421         _mm256_mask_shldv_epi64, _mm256_maskz_shldv_epi64, _mm_shldv_epi16,
5422         _mm_mask_shldv_epi16, _mm_maskz_shldv_epi16, _mm_shldv_epi32,
5423         _mm_mask_shldv_epi32, _mm_maskz_shldv_epi32, _mm_shldv_epi64,
5424         _mm_mask_shldv_epi64, _mm_maskz_shldv_epi64): Ditto.
5425         * config/i386/i386-builtin.def (__builtin_ia32_vpshldv_v32hi,
5426         __builtin_ia32_vpshldv_v32hi_mask, __builtin_ia32_vpshldv_v32hi_maskz,
5427         __builtin_ia32_vpshldv_v16hi, __builtin_ia32_vpshldv_v16hi_mask,
5428         __builtin_ia32_vpshldv_v16hi_maskz, __builtin_ia32_vpshldv_v8hi,
5429         __builtin_ia32_vpshldv_v8hi_mask, __builtin_ia32_vpshldv_v8hi_maskz,
5430         __builtin_ia32_vpshldv_v16si, __builtin_ia32_vpshldv_v16si_mask,
5431         __builtin_ia32_vpshldv_v16si_maskz, __builtin_ia32_vpshldv_v8si,
5432         __builtin_ia32_vpshldv_v8si_mask, __builtin_ia32_vpshldv_v8si_maskz,
5433         __builtin_ia32_vpshldv_v4si, __builtin_ia32_vpshldv_v4si_mask,
5434         __builtin_ia32_vpshldv_v4si_maskz, __builtin_ia32_vpshldv_v8di,
5435         __builtin_ia32_vpshldv_v8di_mask, __builtin_ia32_vpshldv_v8di_maskz,
5436         __builtin_ia32_vpshldv_v4di, __builtin_ia32_vpshldv_v4di_mask,
5437         __builtin_ia32_vpshldv_v4di_maskz, __builtin_ia32_vpshldv_v2di,
5438         __builtin_ia32_vpshldv_v2di_mask,
5439         __builtin_ia32_vpshldv_v2di_maskz): New builtins.
5440         * config/i386/sse.md (vpshldv_<mode>, vpshldv_<mode>_mask,
5441         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): New patterns.
5443 2017-11-29  Julia Koval  <julia.koval@intel.com>
5445         * config/i386/avx512vbmi2intrin.h (_mm512_shrdv_epi16,
5446         _mm512_mask_shrdv_epi16, _mm512_maskz_shrdv_epi16, _mm512_shrdv_epi32,
5447         _mm512_mask_shrdv_epi32, _mm512_maskz_shrdv_epi32, _mm512_shrdv_epi64,
5448         _mm512_mask_shrdv_epi64, _mm512_maskz_shrdv_epi64): New intrinsics.
5449         * config/i386/avx512vbmi2vlintrin.h (_mm256_shrdv_epi16,
5450         _mm256_mask_shrdv_epi16, _mm256_maskz_shrdv_epi16, _mm256_shrdv_epi32,
5451         _mm256_mask_shrdv_epi32, _mm256_maskz_shrdv_epi32, _mm256_shrdv_epi64,
5452         _mm256_mask_shrdv_epi64, _mm256_maskz_shrdv_epi64, _mm_shrdv_epi16,
5453         _mm_mask_shrdv_epi16, _mm_maskz_shrdv_epi16, _mm_shrdv_epi32,
5454         _mm_mask_shrdv_epi32, _mm_maskz_shrdv_epi32, _mm_shrdv_epi64,
5455         _mm_mask_shrdv_epi64, _mm_maskz_shrdv_epi64): Ditto.
5456         * config/i386/i386-builtin-types.def (V32HI_FTYPE_V32HI_V32HI_V32HI,
5457         V32HI_FTYPE_V32HI_V32HI_V32HI_INT, V16HI_FTYPE_V16HI_V16HI_V16HI_INT,
5458         V8HI_FTYPE_V8HI_V8HI_V8HI_INT, V8SI_FTYPE_V8SI_V8SI_V8SI_INT,
5459         V4SI_FTYPE_V4SI_V4SI_V4SI_INT, V8DI_FTYPE_V8DI_V8DI_V8DI,
5460         V8DI_FTYPE_V8DI_V8DI_V8DI_INT, V4DI_FTYPE_V4DI_V4DI_V4DI_INT,
5461         V16SI_FTYPE_V16SI_V16SI_V16SI, V16SI_FTYPE_V16SI_V16SI_V16SI_INT,
5462         V2DI_FTYPE_V2DI_V2DI_V2DI_INT): New types.
5463         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
5464         * config/i386/sse.md (vpshrdv_<mode>, vpshrdv_<mode>_mask,
5465         vpshrdv_<mode>_maskz, vpshrdv_<mode>_maskz_1): New pattern.
5467 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5469         * config/sparc/sparc.c (sparc_do_work_around_errata): Treat the
5470         movsi_pic_gotdata_op instruction as a load for the UT699 errata
5471         workaround.
5473 2017-11-29  Martin Aberg  <maberg@gaisler.com>
5475         * config/sparc/sparc.md (divdf3_fix): Add NOP and adjust length
5476         to prevent b2bst errata sequence.
5477         (sqrtdf2_fix): Likewise.
5479 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5481         * config/sparc/sparc.c (fpop_reg_depend_p): New function.
5482         (div_sqrt_insn_p): New function.
5483         (sparc_do_work_around_errata): Insert NOP instructions to
5484         prevent sequences that could trigger the TN-0013 errata for
5485         certain LEON3 processors.
5486         (pass_work_around_errata::gate): Also test sparc_fix_lost_divsqrt.
5487         (sparc_option_override): Set sparc_fix_lost_divsqrt appropriately.
5488         * config/sparc/sparc.md (fix_lost_divsqrt): New attribute.
5489         (in_branch_delay): Prevent div and sqrt in delay slot if
5490         fix_lost_divsqrt.
5491         * config/sparc/sparc.opt (sparc_fix_lost_divsqrt): New variable.
5493 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5495         * config/sparc/sparc.c (atomic_insn_p): New function.
5496         (sparc_do_work_around_errata): Insert NOP instructions to
5497         prevent sequences that could trigger the TN-0010 errata for
5498         UT700.
5499         * config/sparc/sync.md (atomic_compare_and_swap_leon3_1): Make
5500         instruction referable in atomic_insns_p.
5502 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5504         * config/sparc/sync.md (swapsi): 16-byte align if sparc_fix_gr712rc.
5505         (atomic_compare_and_swap_leon3_1): Likewise.
5506         (ldstub): Likewise.
5508 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5510         * config/sparc/sparc.c (fpop_insn_p): New function.
5511         (sparc_do_work_around_errata): Insert NOP instructions to
5512         prevent sequences that could trigger the TN-0012 errata for
5513         GR712RC.
5514         (pass_work_around_errata::gate): Also test sparc_fix_gr712rc.
5515         * config/sparc/sparc.md (fix_gr712rc): New attribute.
5516         (in_branch_annul_delay): Prevent floating-point instructions
5517         in delay slot of annulled integer branch.
5519 2017-11-29  Richard Biener  <rguenther@suse.de>
5521         PR tree-optimization/83202
5522         * tree-vect-slp.c (scalar_stmts_set_t): New typedef.
5523         (bst_fail): Use it.
5524         (vect_analyze_slp_cost_1): Add visited set, do not account SLP
5525         nodes vectorized to the same stmts multiple times.
5526         (vect_analyze_slp_cost): Allocate a visited set and pass it down.
5527         (vect_analyze_slp_instance): Adjust.
5528         (scalar_stmts_to_slp_tree_map_t): New typedef.
5529         (vect_schedule_slp_instance): Add a map recording the SLP node
5530         representing the vectorized stmts for a set of scalar stmts.
5531         Avoid code-generating redundancies.
5532         (vect_schedule_slp): Allocate map and pass it down.
5534 2017-11-29  Nathan Sidwell  <nathan@acm.org>
5536         PR c++/83187
5537         * tree.c (build_complex_type): Fix canonicalization.  Only fill in
5538         type if it is new.
5540 2017-11-29  Wilco Dijkstra  <wdijkstr@arm.com>
5542         * config/aarch64/aarch64.c (aarch64_print_operand): Add new
5543         cases for printing LDP/STP memory addresses.
5544         (aarch64_print_address_internal): Renamed from
5545         aarch64_print_operand_address, added parameter, add Pmode check.
5546         (aarch64_print_ldpstp_address): New function for LDP/STP addresses.
5547         (aarch64_print_operand_address): Indirect to
5548         aarch64_print_address_internal.
5549         * config/aarch64/aarch64-simd.md (store_pair_lanes): Use new
5550         'y' operand output specifier.
5552 2017-11-29  Jakub Jelinek  <jakub@redhat.com>
5554         PR middle-end/83185
5555         * tree.c (build_simple_mem_ref_loc): Handle
5556         get_addr_base_and_unit_offset returning a MEM_REF.
5558         PR middle-end/80929
5559         * rtlanal.c (seq_cost): For non-single_set insns try to use insn_cost.
5561         PR target/80819
5562         * config/i386/sse.md (vec_concatv2di): Remove * from (=Yr,0,*rm)
5563         alternative.
5565 2017-11-28  David Malcolm  <dmalcolm@redhat.com>
5567         * diagnostic-show-locus.c (layout::print_trailing_fixits): Handle
5568         m_x_offset.
5569         (layout::move_to_column): Likewise.
5571 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5573         PR sanitizer/81275
5574         * tree.c (block_may_fallthru): Return false if SWITCH_ALL_CASES_P
5575         is set on SWITCH_EXPR and !block_may_fallthru (SWITCH_BODY ()).
5577 2017-11-28  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
5578             Martin Jambor  <mjambor@suse.cz>
5580         PR ipa/82808
5581         * tree.h (expr_type_first_operand_type_p): Declare
5582         * tree.c (expr_type_first_operand_type_p): New function.
5583         * ipa-prop.h (ipa_get_type): Allow i to be out of bounds.
5584         (ipa_value_from_jfunc): Adjust declaration.
5585         * ipa-cp.c (ipa_get_jf_pass_through_result): New parameter RES_TYPE.
5586         Use it as result type for arithmetics, unless it is NULL in which case
5587         be more conservative.
5588         (ipa_value_from_jfunc): New parameter PARM_TYPE, pass it to
5589         ipa_get_jf_pass_through_result.
5590         (propagate_vals_across_pass_through): Likewise.
5591         (propagate_scalar_across_jump_function): New parameter PARM_TYPE, pass
5592         is to propagate_vals_across_pass_through.
5593         (propagate_constants_across_call): Pass PARM_TYPE to
5594         propagate_scalar_across_jump_function.
5595         (find_more_scalar_values_for_callers_subset): Pass parameter type to
5596         ipa_value_from_jfunc.
5597         (cgraph_edge_brings_all_scalars_for_node): Likewise.
5598         * ipa-fnsummary.c (evaluate_properties_for_edge): Renamed parms_info
5599         to caller_parms_info, pass parameter type to ipa_value_from_jfunc.
5600         * ipa-prop.c (try_make_edge_direct_simple_call): New parameter
5601         target_type, pass it to ipa_value_from_jfunc.
5602         (update_indirect_edges_after_inlining): Pass parameter type to
5603         try_make_edge_direct_simple_call.
5605 2017-11-28  Jeff Law  <law@redhat.com>
5607         * gimple-ssa-evrp-analyze.c
5608         (evrp_range_analyzer::record_ranges_from_phis): Only use SCEV to
5609         refine ranges if scev_initialized_p returns true.
5610         * vr-values.c (vr_values::extract_range_from_phi_node): Likewise.
5612 2017-11-28  Julia Koval  <julia.koval@intel.com>
5614         * config/i386/avx512vbmi2intrin.h (_mm512_shrdi_epi16,
5615         _mm512_mask_shrdi_epi16, _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
5616         _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32, _mm512_shrdi_epi64,
5617         _mm512_mask_shrdi_epi64, _mm512_maskz_shrdi_epi64): New intrinsics.
5618         * config/i386/avx512vbmi2vlintrin.h (_mm256_shrdi_epi16,
5619         _mm256_mask_shrdi_epi16, _mm256_maskz_shrdi_epi16,
5620         _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32, _mm256_shrdi_epi32,
5621         _mm256_mask_shrdi_epi64, _mm256_maskz_shrdi_epi64, _mm256_shrdi_epi64,
5622         _mm_mask_shrdi_epi16, _mm_maskz_shrdi_epi16, _mm_shrdi_epi16,
5623         _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32, _mm_shrdi_epi32,
5624         _mm_mask_shrdi_epi64, _mm_maskz_shrdi_epi64, _mm_shrdi_epi64): Ditto.
5625         * config/i386/i386-builtin.def (__builtin_ia32_vpshrd_v32hi,
5626         __builtin_ia32_vpshrd_v32hi_mask, __builtin_ia32_vpshrd_v16hi,
5627         __builtin_ia32_vpshrd_v16hi_mask, __builtin_ia32_vpshrd_v8hi,
5628         __builtin_ia32_vpshrd_v8hi_mask, __builtin_ia32_vpshrd_v16si,
5629         __builtin_ia32_vpshrd_v16si_mask, __builtin_ia32_vpshrd_v8si,
5630         __builtin_ia32_vpshrd_v8si_mask, __builtin_ia32_vpshrd_v4si,
5631         __builtin_ia32_vpshrd_v4si_mask, __builtin_ia32_vpshrd_v8di,
5632         __builtin_ia32_vpshrd_v8di_mask, __builtin_ia32_vpshrd_v4di,
5633         __builtin_ia32_vpshrd_v4di_mask, __builtin_ia32_vpshrd_v2di,
5634         __builtin_ia32_vpshrd_v2di_mask): New builtins.
5635         * config/i386/sse.md (vpshrd_<mode><mask_name>): New pattern.
5637 2017-11-28  Julia Koval  <julia.koval@intel.com>
5639         * config/i386/avx512vbmi2intrin.h (_mm512_shldi_epi16,
5640         _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16, _mm512_shldi_epi32,
5641         _mm512_mask_shldi_epi32, _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
5642         _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): New intrinsics.
5643         * config/i386/avx512vbmi2vlintrin.h (_mm256_shldi_epi16,
5644         _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
5645         _mm256_mask_shldi_epi32, _mm256_maskz_shldi_epi32, _mm256_shldi_epi32,
5646         _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64, _mm256_shldi_epi64,
5647         _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16, _mm_shldi_epi16,
5648         _mm_mask_shldi_epi32, _mm_maskz_shldi_epi32, _mm_shldi_epi32,
5649         _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64, _mm_shldi_epi64): Ditto.
5650         * config/i386/i386-builtin-types.def (V32HI_FTYPE_V32HI_V32HI_INT,
5651         V32HI_FTYPE_V32HI_V32HI_INT_V32HI_INT, V16SI_FTYPE_V16SI_V16SI_INT,
5652         V16SI_FTYPE_V16SI_V16SI_INT_V16SI_INT,
5653         V8DI_FTYPE_V8DI_V8DI_INT_V8DI_INT, V8SI_FTYPE_V8SI_V8SI_INT_V8SI_INT,
5654         V16HI_FTYPE_V16HI_V16HI_INT_V16HI_INT,
5655         V4DI_FTYPE_V4DI_V4DI_INT_V4DI_INT,
5656         V8HI_FTYPE_V8HI_V8HI_INT_V8HI_INT,
5657         V4SI_FTYPE_V4SI_V4SI_INT_V4SI_INT,
5658         V2DI_FTYPE_V2DI_V2DI_INT_V2DI_INT): New types.
5659         * config/i386/i386-builtin.def (__builtin_ia32_vpshld_v32hi,
5660         __builtin_ia32_vpshld_v32hi_mask, __builtin_ia32_vpshld_v16hi,
5661         __builtin_ia32_vpshld_v16hi_mask, __builtin_ia32_vpshld_v8hi,
5662         __builtin_ia32_vpshld_v8hi_mask, __builtin_ia32_vpshld_v16si,
5663         __builtin_ia32_vpshld_v16si_mask, __builtin_ia32_vpshld_v8si,
5664         __builtin_ia32_vpshld_v8si_mask, __builtin_ia32_vpshld_v4si,
5665         __builtin_ia32_vpshld_v4si_mask, __builtin_ia32_vpshld_v8di,
5666         __builtin_ia32_vpshld_v8di_mask, __builtin_ia32_vpshld_v4di,
5667         __builtin_ia32_vpshld_v4di_mask, __builtin_ia32_vpshld_v2di,
5668         __builtin_ia32_vpshld_v2di_mask): New builtins.
5669         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
5670         * config/i386/sse.md (vpshld_<mode><mask_name>): New pattern.
5672 2017-11-28  Richard Biener  <rguenther@suse.de>
5674         PR tree-optimization/80776
5675         * gimple-ssa-evrp-analyze.h (evrp_range_analyzer::set_ssa_range_info):
5676         Declare.
5677         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
5678         New function.
5679         (evrp_range_analyzer::record_ranges_from_incoming_edges):
5680         If the incoming edge is an effective fallthru because the other
5681         edge only reaches a __builtin_unreachable () then record ranges
5682         derived from the controlling condition in SSA info.
5683         (evrp_range_analyzer::record_ranges_from_phis): Use set_ssa_range_info.
5684         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5686 2017-11-28  Olivier Hainque  <hainque@adacore.com>
5688         * Makefile.in (SELFTEST_FLAGS): Use nul instead of /dev/null
5689         on mingw build hosts.
5691 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5693         PR debug/81307
5694         * dbxout.c (dbx_block_with_cold_children): Fix function comment.
5696 2017-11-28  Richard Biener  <rguenther@suse.de>
5698         PR middle-end/83141
5699         * gimple-fold.c (gimple_fold_builtin_memory_op): For aggregate
5700         copies generated from memcpy use a character array as reference
5701         type.
5703 2017-11-28  Julia Koval  <julia.koval@intel.com>
5704             Sebastian Peryt  <sebastian.peryt@intel.com>
5706         * Makefile.in (cilkplus.def, cilk-builtins.def, c-family/cilk.o,
5707         c-family/c-cilkplus.o, c-family/array-notation-common.o,
5708         cilk-common.o, cilk.h, cilk-common.c): Remove.
5709         * builtin-types.def
5710         (BT_FN_INT_PTR_PTR_PTR_FTYPE_BT_INT_BT_PTR_BT_PTR_BT_PTR): Remove.
5711         * builtins.c (is_builtin_name): Remove cilkplus condition.
5712         (BUILT_IN_CILK_DETACH, BUILT_IN_CILK_POP_FRAME): Remove.
5713         * builtins.def (DEF_CILK_BUILTIN_STUB, DEF_CILKPLUS_BUILTIN,
5714         cilk-builtins.def, cilkplus.def): Remove.
5715         * cif-code.def (CILK_SPAWN): Remove.
5716         * cilk-builtins.def: Delete.
5717         * cilk-common.c: Ditto.
5718         * cilk.h: Ditto.
5719         * cilkplus.def: Ditto.
5720         * config/darwin.h (fcilkplus): Delete.
5721         * cppbuiltin.c: Ditto.
5722         * doc/extend.texi: Remove cilkplus doc.
5723         * doc/generic.texi: Ditto.
5724         * doc/invoke.texi: Ditto.
5725         * doc/passes.texi: Ditto.
5726         * gcc.c (fcilkplus): Remove.
5727         * gengtype.c (cilk.h): Remove.
5728         * gimple-pretty-print.c (dump_gimple_omp_for): Remove cilkplus
5729         support.
5730         * gimple.h (GF_OMP_FOR_KIND_CILKFOR, GF_OMP_FOR_KIND_CILKSIMD):
5731         Remove.
5732         * gimplify.c (gimplify_return_expr, maybe_fold_stmt,
5733         gimplify_call_expr,
5734         is_gimple_stmt, gimplify_modify_expr, gimplify_scan_omp_clauses,
5735         gimplify_adjust_omp_clauses, gimplify_omp_for, gimplify_expr): Remove
5736         cilkplus conditions.
5737         * ipa-fnsummary.c (ipa_dump_fn_summary, compute_fn_summary,
5738         inline_read_section): Ditto.
5739         * ipa-inline-analysis.c (cilk.h): Remove.
5740         * ira.c (ira_setup_eliminable_regset): Remove cilkplus support.
5741         * lto-wrapper.c (merge_and_complain, append_compiler_options,
5742         append_linker_options): Remove condition for fcilkplus.
5743         * lto/lto-lang.c (cilk.h): Remove.
5744         (lto_init): Remove condition for fcilkplus.
5745         * omp-expand.c (expand_cilk_for_call): Delete.
5746         (expand_omp_taskreg, expand_omp_for_static_chunk,
5747         expand_omp_for): Remove cilkplus
5748         conditions.
5749         (expand_cilk_for): Delete.
5750         * omp-general.c (omp_extract_for_data): Remove cilkplus support.
5751         * omp-low.c (scan_sharing_clauses, create_omp_child_function,
5752         execute_lower_omp, diagnose_sb_0): Ditto.
5753         * omp-simd-clone.c (simd_clone_clauses_extract): Ditto.
5754         * tree-core.h (OMP_CLAUSE__CILK_FOR_COUNT_): Delete.
5755         * tree-nested.c: Ditto.
5756         * tree-pretty-print.c (dump_omp_clause): Remove cilkplus support.
5757         (dump_generic_node): Ditto.
5758         * tree.c (OMP_CLAUSE__CILK_FOR_COUNT_): Delete.
5759         * tree.def (cilk_simd, cilk_for, cilk_spawn_stmt, cilk_sync_stmt):
5760         Delete.
5761         * tree.h (CILK_SPAWN_FN, EXPR_CILK_SPAWN): Delete.
5763 2017-11-28  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
5765         * config/aarch64/aarch64.md (div<mode>3): Change check to TARGET_FLOAT.
5766         * config/aarch64/aarch64.c (aarch64_emit_approx_div): Add early exit
5767         for vector mode and !TARGET_SIMD.
5769 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5771         * tree.def (SWITCH_EXPR): Change from 3 operand to 2 operand tree.
5772         Adjust comment.
5773         * tree.h (SWITCH_LABELS): Remove.
5774         * gimplify.c (gimplify_switch_expr): Don't test SWITCH_LABELS,
5775         assert SWITCH_BODY is non-NULL.
5776         * tree-pretty-print.c (dump_generic_node): Remove SWITCH_LABELS
5777         handling.
5778         * tree.c (block_may_fallthru): Always return true; for SWITCH_EXPR.
5780         PR tree-optimization/80788
5781         * match.pd (X +- C1 CMP C2 -> X CMP C2 -+ C1): If res
5782         has TREE_OVERFLOW set, call drop_tree_overflow.
5784 2017-11-28  Richard Biener  <rguenther@suse.de>
5786         PR tree-optimization/83158
5787         * tree-vrp.c (intersect_ranges): Prefer ~[0, 0] in a few more cases.
5789 2017-11-28  Segher Boessenkool  <segher@kernel.crashing.org>
5791         PR 81288/target
5792         * config/rs6000/rs6000.c (rs6000_rtx_costs): Do not handle
5793         TARGET_ISEL && !TARGET_MFCRF differently.  Simplify code.
5795 2017-11-27  Segher Boessenkool  <segher@kernel.crashing.org>
5797         * config/rs6000/rs6000.md (<code><GPR:mode><GPR2:mode>2_isel): Change
5798         LT/GT/LTU/GTU to LE/GE/LEU/GEU where possible.
5800 2017-11-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
5802         PR middle_end/82333
5803         * varasm.c (compare_constant): Take the mode of the constants into
5804         account when comparing floating point constants.
5806 2017-11-27  Gerald Pfeifer  <gerald@pfeifer.com>
5808         * hash-set.h (DEFINE_DEBUG_HASH_SET): Remove static qualifier
5809         from explicit instantiation of debug_helper.
5810         * vec.h (DEFINE_DEBUG_VEC): Ditto.
5812 2017-11-27  Richard Biener  <rguenther@suse.de>
5814         * gimple-fold.c (gimple_fold_builtin_memory_op): Remove dead code,
5815         refactor a bit.
5817 2017-11-27  Richard Biener  <rguenther@suse.de>
5819         * tree.c (wide_int_to_tree): Free discarded INTEGER_CST.
5820         (type_hash_canon): Also clear int_cst_hash_table entry for
5821         TYPE_MIN/MAX_VALUE.
5822         (build_nonstandard_integer_type): Hash all TYPE_MAX_VALUEs.
5824 2017-11-27  Tamar Christina  <tamar.christina@arm.com>
5826         * doc/extend.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5827         * doc/invoke.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5828         * doc/sourcebuild.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5830 2017-11-27  Markus Trippelsdorf  <markus@trippelsdorf.de>
5832         * hash-map.h (gt_cleare_cache): Avoid UB.
5834 2017-11-27  Eric Botcazou  <ebotcazou@adacore.com>
5836         * cfgloop.h (struct loop): Document usage of USHRT_MAX for unroll.
5837         * loop-unroll.c (decide_unroll_constant_iterations): Implement it.
5838         (decide_unroll_runtime_iterations): Likewise.
5839         (decide_unroll_stupid): Likewise.
5841 2017-11-27  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
5843         PR target/83109
5844         * config/i386/i386.md: Add a loop with incssp.
5846 2017-11-27  Martin Jambor  <mjambor@suse.cz>
5848         PR tree-optimization/81248
5849         * tree-sra.c (splice_param_accesses): Remove size check.
5850         (decide_one_param_reduction): Fix size check.
5851         * gimple-pretty-print.c (dump_profile): Silence warning.
5852         * params.def (PARAM_IPA_SRA_PTR_GROWTH_FACTOR): Adjust description.
5854 2017-11-27  Jakub Jelinek  <jakub@redhat.com>
5856         PR debug/81307
5857         * dbxout.c (lastlineno): New variable.
5858         (dbx_debug_hooks): Use dbxout_switch_text_section as
5859         switch_text_section debug hook.
5860         (dbxout_function_end): Switch to current_function_section
5861         rather than function_section.  If crtl->has_bb_partition,
5862         output just one N_FUN, depending on in_cold_section_p.
5863         (dbxout_source_line): Remember last lineno in lastlineno.
5864         (dbxout_switch_text_section): New function.
5865         (dbxout_function_decl): Adjust dbxout_block caller.
5866         (dbx_block_with_cold_children): New function.
5867         (dbxout_block): Return true if any LBRAC/RBRAC have been
5868         emitted.  Use dbx_block_with_cold_children at depth == 0
5869         in second partition.  Add PARENT_BLOCKNUM argument, pass
5870         it optionally adjusted to children.  Output LBRAC/RBRAC
5871         around recursive call only if the block is in the current
5872         partition, if not and anything was output, emit empty
5873         range LBRAC/RBRAC.
5874         * final.c (final_scan_insn): Compute cold_function_name
5875         before calling switch_text_section debug hook.  Call
5876         that hook even if dwarf2out_do_frame if not emitting
5877         dwarf debug info.
5879         PR target/83100
5880         * varasm.c (bss_initializer_p): Return true for DECL_COMMON
5881         TREE_READONLY decls.
5883 2017-11-27  Markus Trippelsdorf  <markus@trippelsdorf.de>
5885         PR rtl-optimization/82488
5886         * expr.c (fixup_args_size_notes): Avoid signed integer overflow.
5888 2017-11-26  Julia Koval  <julia.koval@intel.com>
5890         * config/i386/i386.c (processor_target_table): Add skylake_cost for
5891         skylake-avx512.
5892         * config/i386/x86-tune-costs.h (skylake_memcpy, skylake_memset,
5893         skylake_cost): New.
5895 2017-11-26  Julia Koval  <julia.koval@intel.com>
5897         * config/i386/driver-i386.c (host_detect_local_cpu):
5898         Detect skylake-avx512.
5900 2017-11-26  Julia Koval  <julia.koval@intel.com>
5902         * config.gcc: Add -march=cannonlake.
5903         * config/i386/driver-i386.c (host_detect_local_cpu): Detect cannonlake.
5904         * config/i386/i386-c.c (ix86_target_macros_internal): Handle cannonlake.
5905         * config/i386/i386.c (processor_costs): Add m_CANNONLAKE.
5906         (PTA_CANNONLAKE): New.
5907         (processor_target_table): Add cannonlake.
5908         (ix86_option_override_internal): Ditto.
5909         (fold_builtin_cpu): Ditto.
5910         (get_builtin_code_for_version): Handle cannonlake.
5911         (M_INTEL_COREI7_CANNONLAKE): New.
5912         * config/i386/i386.h (TARGET_CANNONLAKE, PROCESSOR_CANNONLAKE): New.
5913         * doc/invoke.texi: Add -march=cannonlake.
5915 2017-11-14  Boris Kolpackov  <boris@codesynthesis.com>
5917         * plugin.c (add_new_plugin): Use platform-specific library extensions.
5918         (try_init_one_plugin): Alternative implementation for MinGW.
5919         * Makefile.in (plugin_implib): New.
5920         (gengtype-lex.c): Fix broken AIX workaround.
5921         * configure: Regenerate.
5922         * doc/plugins.texi: Document support for MinGW.
5924 2017-11-25  Jakub Jelinek  <jakub@redhat.com>
5926         PR rtl-optimization/81553
5927         * combine.c (simplify_if_then_else): In (if_then_else COND (OP Z C1) Z)
5928         to (OP Z (mult COND (C1 * STORE_FLAG_VALUE))) optimization, if OP
5929         is a shift where C1 has different mode than the whole shift, use C1's
5930         mode for MULT rather than the shift's mode.
5932         PR target/82848
5933         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Don't fold
5934         builtins not enabled in the currently selected ISA.
5936 2017-11-24  Jackson Woodruff  <jackson.woodruff@arm.com>
5938         PR tree-optimization/71026
5939         * tree-ssa-math-opts (is_division_by_square, is_square_of): New.
5940         (insert_reciprocals): Change to insert reciprocals before a division
5941         by a square and to insert the square of a reciprocal.
5942         (execute_cse_reciprocals_1): Change to consider division by a square.
5943         (register_division_in): Add importance parameter.
5945 2017-11-24  Richard Biener  <rguenther@suse.de>
5947         PR tree-optimization/82402
5948         * tree-vect-loop-manip.c (create_lcssa_for_virtual_phi): Properly
5949         set SSA_NAME_OCCURS_IN_ABNORMAL_PHI.
5951 2017-11-24  Marc Glisse  <marc.glisse@inria.fr>
5953         * match.pd (0-ptr): New transformation.
5955 2017-11-24  Jan Hubicka  <hubicka@ucw.cz>
5957         PR bootstrap/83015
5958         * ipa-inline.c (inline_small_functions): Set current badnes correctly
5959         when skipping checking.
5961 2017-11-24  Richard Biener  <rguenther@suse.de>
5963         PR tree-optimization/83128
5964         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Handle STRING_CSTs.
5965         (vn_reference_lookup_3): Likewise.
5967 2017-11-24  Jakub Jelinek  <jakub@redhat.com>
5969         PR sanitizer/83014
5970         * ubsan.c (ubsan_type_descriptor): Use pp_unsigned_wide_integer
5971         instead of pp_printf with HOST_WIDE_INT_PRINT_DEC.  Avoid calling
5972         tree_to_uhwi twice.
5974         * tree-object-size.c (pass_through_call): Use gimple_call_return_flags
5975         ERF_RETURN*ARG* for builtins other than BUILT_IN_ASSUME_ALIGNED,
5976         check for the latter with gimple_call_builtin_p.  Do not handle
5977         BUILT_IN_STPNCPY_CHK which is not a pass through call.
5979 2017-11-24  Christophe Lyon  <christophe.lyon@linaro.org>
5981         * config/arm/arm_neon.h: Fix pragma GCC push_options before
5982         vdot_u32.
5984 2017-11-23  Julia Koval  <julia.koval@intel.com>
5986         * config/i386/avx512vbmi2intrin.h (_mm512_mask_expand_epi8,
5987         _mm512_maskz_expand_epi8, _mm512_mask_expandloadu_epi8,
5988         _mm512_maskz_expandloadu_epi8, _mm512_mask_expand_epi16,
5989         _mm512_maskz_expand_epi16, _mm512_mask_expandloadu_epi16,
5990         _mm512_maskz_expandloadu_epi16): New intrinsics.
5991         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_expand_epi8,
5992         _mm_maskz_expand_epi8, _mm_mask_expandloadu_epi8,
5993         _mm_maskz_expandloadu_epi8, _mm_mask_expand_epi16,
5994         _mm_maskz_expand_epi16, _mm_mask_expandloadu_epi16,
5995         _mm_maskz_expandloadu_epi16, _mm256_mask_expand_epi16,
5996         _mm256_maskz_expand_epi16, _mm256_mask_expandloadu_epi16,
5997         _mm256_maskz_expandloadu_epi16, _mm256_mask_expand_epi8,
5998         _mm256_maskz_expand_epi8, _mm256_mask_expandloadu_epi8,
5999         _mm256_maskz_expandloadu_epi8): New intrinsics.
6000         * config/i386/i386-builtin-types.def (V64QI_FTYPE_PCV64QI_V64QI_UDI,
6001         V32HI_FTYPE_PCV32HI_V32HI_USI, V32QI_FTYPE_PCV32QI_V32QI_USI,
6002         V16HI_FTYPE_PCV16HI_V16HI_UHI, V16QI_FTYPE_PCV16QI_V16QI_UHI,
6003         V8HI_FTYPE_PCV8HI_V8HI_UQI): New types.
6004         * config/i386/i386.c (ix86_expand_special_args_builtin): Use new types.
6005         * config/i386/sse.md (VI248_VLBW): New iterator.
6006         (expand<mode>_mask, expand<mode>_maskz): New patterns.
6008 2017-11-23  Julia Koval  <julia.koval@intel.com>
6010         * config.gcc (avx512vbmi2intrin.h, avx512vbmi2vlintrin): New headers.
6011         * config/i386/avx512vbmi2intrin.h (_mm512_mask_compress_epi8,
6012         _mm512_maskz_compress_epi8, _mm512_mask_compressstoreu_epi8,
6013         _mm512_mask_compress_epi16, _mm512_maskz_compress_epi16,
6014         _mm512_mask_compressstoreu_epi16): New.
6015         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_compress_epi8,
6016         _mm_maskz_compress_epi8, _mm256_mask_compressstoreu_epi16,
6017         _mm_mask_compress_epi16, _mm_maskz_compress_epi16,
6018         _mm256_mask_compress_epi16, _mm256_maskz_compress_epi16,
6019         _mm_mask_compressstoreu_epi8, _mm_mask_compressstoreu_epi16,
6020         _mm256_mask_compress_epi8, _mm256_maskz_compress_epi8,
6021         _mm256_mask_compressstoreu_epi8): New.
6022         * config/i386/i386-builtin-types.def (VOID_FTYPE_PV64QI_V64QI_UDI,
6023         VOID_FTYPE_PV32HI_V32HI_USI, VOID_FTYPE_PV32QI_V32QI_USI,
6024         VOID_FTYPE_PV16QI_V16QI_UHI, VOID_FTYPE_PV16HI_V16HI_UHI,
6025         VOID_FTYPE_PV8HI_V8HI_UQI): New types.
6026         * config/i386/i386-builtin.def (__builtin_ia32_compressqi512_mask,
6027         __builtin_ia32_compresshi512_mask, __builtin_ia32_compressqi256_mask,
6028         __builtin_ia32_compressqi128_mask, __builtin_ia32_compresshi256_mask,
6029         __builtin_ia32_compresshi128_mask,
6030         __builtin_ia32_compressstoreuqi512_mask,
6031         __builtin_ia32_compressstoreuhi512_mask,
6032         __builtin_ia32_compressstoreuqi256_mask,
6033         __builtin_ia32_compressstoreuqi128_mask,
6034         __builtin_ia32_compressstoreuhi256_mask,
6035         __builtin_ia32_compressstoreuhi128_mask): New builtins.
6036         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Create special args
6037         array for flags2.
6038         (ix86_expand_special_args_builtin): Handle new types.
6039         (s4fma_expand): Handle new builtin array.
6040         * config/i386/immintrin.h: Include new headers.
6041         * config/i386/sse.md (VI12_AVX512VLBW): New iterator.
6042         (compress<mode>_mask, compressstore<mode>_mask): New patterns.
6044 2017-11-23  Jakub Jelinek  <jakub@redhat.com>
6046         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Formatting
6047         fixes.  Declare temp and g variables at the top in order to avoid
6048         {} in most of the cases.
6050 2017-11-23  Marc Glisse  <marc.glisse@inria.fr>
6052         * match.pd (ptr-0): New transformation.
6054 2017-11-23  Charles Baylis  <charles.baylis@linaro.org>
6056         * config/arm/arm-protos.h (enum arm_addr_mode_op): New.
6057         (struct addr_mode_cost_table): New.
6058         (struct tune_params): Add field addr_mode_costs.
6059         * config/arm/arm.c (generic_addr_mode_costs): New.
6060         (arm_slowmul_tune): Initialise addr_mode_costs field.
6061         (arm_fastmul_tune): Likewise.
6062         (arm_strongarm_tune): Likewise.
6063         (arm_xscale_tune): Likewise.
6064         (arm_9e_tune): Likewise.
6065         (arm_marvell_pj4_tune): Likewise.
6066         (arm_v6t2_tune): Likewise.
6067         (arm_cortex_tune): Likewise.
6068         (arm_cortex_a8_tune): Likewise.
6069         (arm_cortex_a7_tune): Likewise.
6070         (arm_cortex_a15_tune): Likewise.
6071         (arm_cortex_a35_tune): Likewise.
6072         (arm_cortex_a53_tune): Likewise.
6073         (arm_cortex_a57_tune): Likewise.
6074         (arm_exynosm1_tune): Likewise.
6075         (arm_xgene1_tune): Likewise.
6076         (arm_cortex_a5_tune): Likewise.
6077         (arm_cortex_a9_tune): Likewise.
6078         (arm_cortex_a12_tune): Likewise.
6079         (arm_cortex_a73_tune): Likewise.
6080         (arm_v7m_tune): Likewise.
6081         (arm_cortex_m7_tune): Likewise.
6082         (arm_v6m_tune): Likewise.
6083         (arm_fa726te_tune): Likewise.
6084         (arm_mem_costs): Use table lookup to calculate cost of addressing mode.
6086 2017-11-23  Charles Baylis  <charles.baylis@linaro.org>
6088         * config/arm/arm.c (arm_mem_costs): New function.
6089         (arm_rtx_costs_internal): Use arm_mem_costs.
6091 2017-11-23  Mark Wielaard  <mark@klomp.org>
6093         * dwarf2out.c (init_sections_and_labels): Use generation to create
6094         unique ranges_section_label and ranges_base_label. Return generation.
6095         (output_rnglists): Add generation argument. Use generation to create
6096         unique ranges labels.
6097         (dwarf2out_finish): Get generation from init_sections_and_labels
6098         and pass generation to output_rnglists.
6100 2017-11-23  Mike Stump  <mikestump@comcast.net>
6101             Eric Botcazou  <ebotcazou@adacore.com>
6103         * doc/generic.texi (ANNOTATE_EXPR): Document 3rd operand.
6104         * cfgloop.h (struct loop): Add unroll field.
6105         * function.h (struct function): Add has_unroll bitfield.
6106         * gimplify.c (gimple_boolify) <ANNOTATE_EXPR>: Deal with unroll kind.
6107         (gimplify_expr) <ANNOTATE_EXPR>: Propagate 3rd operand.
6108         * loop-init.c (pass_loop2::gate): Return true if cfun->has_unroll.
6109         (pass_rtl_unroll_loops::gate): Likewise.
6110         * loop-unroll.c (decide_unrolling): Tweak note message.  Skip loops
6111         for which loop->unroll==1.
6112         (decide_unroll_constant_iterations): Use note for consistency and
6113         take loop->unroll into account.  Return early if loop->unroll is set.
6114         Fix thinko in existing test.
6115         (decide_unroll_runtime_iterations): Use note for consistency and
6116         take loop->unroll into account.
6117         (decide_unroll_stupid): Likewise.
6118         * lto-streamer-in.c (input_cfg): Read loop->unroll.
6119         * lto-streamer-out.c (output_cfg): Write loop->unroll.
6120         * tree-cfg.c (replace_loop_annotate_in_block) <annot_expr_unroll_kind>:
6121         New case.
6122         (replace_loop_annotate) <annot_expr_unroll_kind>: Likewise.
6123         (print_loop): Print loop->unroll if set.
6124         * tree-core.h (enum annot_expr_kind): Add annot_expr_unroll_kind.
6125         * tree-inline.c (copy_loops): Copy unroll and set cfun->has_unroll.
6126         * tree-pretty-print.c (dump_generic_node) <annot_expr_unroll_kind>:
6127         New case.
6128         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Bail out if
6129         loop->unroll is set and smaller than the trip count.  Otherwise bypass
6130         entirely the heuristics if loop->unroll is set.  Remove dead note.
6131         Fix off-by-one bug in other note.
6132         (try_peel_loop): Bail out if loop->unroll is set.  Fix formatting.
6133         (tree_unroll_loops_completely_1): Force unrolling if loop->unroll
6134         is greater than 1.
6135         (tree_unroll_loops_completely): Make static.
6136         (pass_complete_unroll::execute): Use correct type for variable.
6137         (pass_complete_unrolli::execute): Fix formatting.
6138         * tree.def (ANNOTATE_EXPR): Add 3rd operand.
6140 2017-11-23  Sergey Shalnov  <Sergey.Shalnov@intel.com>
6142         * config/i386/i386.h (TARGET_PREFER_AVX256): Also
6143         enable when TARGET_PREFER_AVX128 is set.
6145 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
6147         * ipa-profile.c (ipa_propagate_frequency_1): Use count instead of
6148         frequency.
6149         * cgraph.c (cgraph_edge::dump_edge_flags): Dump sreal frequencies.
6150         (cgraph_edge::maybe_hot_p): Use sreal frequencies.
6152 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
6154         * ipa-fnsummary.c (record_modified_bb_info): Use sreal
6155         frequencies. Fix estimation of aggregate parameters.
6157 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
6159         * cgraphclones.c (cgraph_node::create_clone): Fix updating of profile
6160         when inlining.
6162 2017-11-23  Tom de Vries  <tom@codesourcery.com>
6164         * config/sh/sh.h (ASM_OUTPUT_ADDR_VEC_ELT): Wrap in "do {} while (0)".
6166 2017-11-23  Tom de Vries  <tom@codesourcery.com>
6168         * config/ft32/ft32.h (ASM_OUTPUT_ADDR_VEC_ELT): Remove semicolon after
6169         macro.
6171 2017-11-23  Oleg Endo  <olegendo@gcc.gnu.org>
6173         PR target/83111
6174         * config/sh/sh.md (udivsi3, divsi3, sibcall_value_pcrel,
6175         sibcall_value_pcrel_fdpic): Use local variable instead of
6176         operands[3].
6177         (calli_tbr_rel): Add missing operand 2.
6178         (call_valuei_tbr_rel): Add missing operand 3.
6180 2017-11-23  Jakub Jelinek  <jakub@redhat.com>
6182         PR middle-end/82253
6183         * expr.c (expand_assignment): For CONCAT to_rtx, complex type from and
6184         bitpos/bitsize covering the whole destination, use store_expr only if
6185         the complex mode is the same.  Otherwise, use expand_normal and if
6186         it returns CONCAT, subreg each part separately instead of trying to
6187         subreg the whole result.
6189 2017-11-23  Richard Biener  <rguenther@suse.de>
6191         PR tree-optimization/23094
6192         * tree-ssa-sccvn.c (vuse_ssa_val): Handle VN_TOP when we
6193         come here from walking over backedges in the first iteration.
6194         (vn_reference_lookup_3): Skip clobbers that store the same value.
6196 2017-11-23  Richard Biener  <rguenther@suse.de>
6198         PR tree-optimization/81403
6199         * tree-ssa-pre.c (get_representative_for): Add parameter specifying
6200         a block we need a leader relative to.
6201         (phi_translate_1): For nary processing require a leader from
6202         get_representative_for given we run expression simplification
6203         using match-and-simplify.  Remove previous fix.
6205 2017-11-22  Jeff Law  <law@redhat.com>
6207         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::try_find_new_range):
6208         Use new method allocate_value_range rather than accessing the
6209         vrp_value_range_pool data member directly.
6210         * tree-vrp.c (simplify_stmt_for_jump_threading): Tweak slightly
6211         to use extract_range_from_stmt method to avoid need for
6212         extract_range_from_assignment method.
6213         (vrp_prop::vrp_finalize): Use set_lattice_propagation_complete
6214         method rather than setting values_propgated data member directly.
6215         * vr-values.h (class vr_values): Privatize vrp_value_range_pool,
6216         and values propagated data members and extract_range_from_assignment
6217         method.  Reorder private data members to conform to standards.
6218         Add new methods set_lattice_propagation_complete and
6219         allocate_value_range.
6221 2017-11-22  Eric Botcazou  <ebotcazou@adacore.com>
6223         PR rtl-optimization/83030
6224         * doc/rtl.texi (Flags in an RTL Expression): Alphabetize, add entry
6225         for CROSSING_JUMP_P and mention usage of 'jump' for JUMP_INSNs.
6226         (Insns): Delete entry for REG_CROSSING_JUMP in register notes.
6227         * bb-reorder.c (update_crossing_jump_flags): Do not test whether the
6228         CROSSING_JUMP_P flag is already set before setting it.
6229         * cfgrtl.c (fixup_partition_crossing): Likewise.
6230         * reorg.c (relax_delay_slots): Do not consider a CROSSING_JUMP_P
6231         insn as useless.
6233 2017-11-22  Jakub Jelinek  <jakub@redhat.com>
6235         * simplify-rtx.c (simplify_binary_operation_1) <case VEC_SERIES>:
6236         Handle the case where both arguments are using gen_const_vec_series.
6238 2017-11-22  David Malcolm  <dmalcolm@redhat.com>
6240         PR c++/62170
6241         * pretty-print.c (pp_format): Move quoting implementation to
6242         pp_begin_quote and pp_end_quote.  Update pp_format_decoder call
6243         to pass address of "quote" local.
6244         (pp_begin_quote): New function.
6245         (pp_end_quote): New function.
6246         * pretty-print.h (printer_fn): Convert penultimate param from bool
6247         to bool *.
6248         (pp_begin_quote): New decl.
6249         (pp_end_quote): New decl.
6250         * tree-diagnostic.c (default_tree_printer): Convert penultimate
6251         param from bool to bool *.
6252         * tree-diagnostic.h (default_tree_printer): Likewise.
6254 2017-11-22  Jeff Law  <law@redhat.com>
6256         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::evrp_range_analyzer)
6257         Initialize vr_values.
6258         (evrp_range_analyzer::try_find_new_range): Call methods attached to
6259         vr_values via vr_values class instance rather than delegators.
6260         (evrp_range_analyzer::record_ranges_from_phis): Likewise.
6261         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6262         (evrp_range_analyzer::push_value_range): Likewise.
6263         (evrp_range_analyzer::pop_value_range): Likewise.
6264         * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
6265         most delegators.  Those remaining are exposed as public interfaces.
6266         Make vr_values a pointer and private.
6267         (evrp_range_analyzer::~evrp_range_analyzer): Delete the attached
6268         vr_values.
6269         (evrp_range_analyzer::get_vr_value): New method.
6270         * gimple-ssa-evrp.c (class evrp_folder): Use DISABLE_COPY_AND_ASSIGN.
6271         (evrp_folder::evrp_folder): New ctor to initialize vr_values.
6272         (class evrp_dom_walker): Attach evrp_folder class, initialize
6273         it in the ctor.  Remove temporary delegators.
6274         (evrp_dom_walker::before_dom_children): Call methods in attached
6275         evrp_range_analyzer class via class instance pointer.  Use
6276         free value_range_constant_singleton to remove need for
6277         op_with_constant_singleton_value delegator method.  Do not
6278         create a vrp_prop class instance for every call!  Narrow
6279         scope of a couple variables.
6280         (evrp_dom_walker::cleanup): Call methods in attached
6281         evrp_range_analyzer class via class instance pointer.
6282         * vr-values.h (class vr_values): Privatize many methods and
6283         data members.
6285 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6287         * tree.def (REDUC_MAX_EXPR, REDUC_MIN_EXPR, REDUC_PLUS_EXPR): Delete.
6288         * cfgexpand.c (expand_debug_expr): Remove handling for them.
6289         * expr.c (expand_expr_real_2): Likewise.
6290         * fold-const.c (const_unop): Likewise.
6291         * optabs-tree.c (optab_for_tree_code): Likewise.
6292         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
6293         * tree-inline.c (estimate_operator_cost): Likewise.
6294         * tree-pretty-print.c (dump_generic_node): Likewise.
6295         (op_code_prio): Likewise.
6296         (op_symbol_code): Likewise.
6297         * internal-fn.def (DEF_INTERNAL_SIGNED_OPTAB_FN): Define.
6298         (IFN_REDUC_PLUS, IFN_REDUC_MAX, IFN_REDUC_MIN): New internal functions.
6299         * internal-fn.c (direct_internal_fn_optab): New function.
6300         (direct_internal_fn_array, direct_internal_fn_supported_p
6301         (internal_fn_expanders): Handle DEF_INTERNAL_SIGNED_OPTAB_FN.
6302         * fold-const-call.c (fold_const_reduction): New function.
6303         (fold_const_call): Handle CFN_REDUC_PLUS, CFN_REDUC_MAX and
6304         CFN_REDUC_MIN.
6305         * tree-vect-loop.c: Include internal-fn.h.
6306         (reduction_code_for_scalar_code): Rename to...
6307         (reduction_fn_for_scalar_code): ...this and return an internal
6308         function.
6309         (vect_model_reduction_cost): Take an internal_fn rather than
6310         a tree_code.
6311         (vect_create_epilog_for_reduction): Likewise.  Build calls rather
6312         than assignments.
6313         (vectorizable_reduction): Use internal functions rather than tree
6314         codes for the reduction operation.  Update calls to the functions
6315         above.
6316         * config/aarch64/aarch64-builtins.c (aarch64_gimple_fold_builtin):
6317         Use calls to internal functions rather than REDUC tree codes.
6318         * config/aarch64/aarch64-simd.md: Update comment accordingly.
6320 2017-11-22  Olivier Hainque  <hainque@adacore.com>
6322         * config/vxworks.c (vxworks_override_options): Pick default
6323         dwarf version from macro value, VXWORKS_DWARF_VERSION_DEFAULT.
6324         * config/vxworks.h: Define VXWORKS_DWARF_VERSION_DEFAULT and
6325         DWARF_GNAT_ENCODINGS_DEFAULT.
6326         * config/vxworksae.h: Likewise.
6328 2017-11-22  Marc Glisse  <marc.glisse@inria.fr>
6330         PR tree-optimization/83104
6331         * vr-values.c (simplify_stmt_using_ranges): Check integral argument,
6332         not result.
6334 2017-11-22  Marek Polacek  <polacek@redhat.com>
6335             H.J. Lu  <hongjiu.lu@intel.com>
6336             Jason Merrill  <jason@redhat.com>
6338         PR c++/60336
6339         PR middle-end/67239
6340         PR target/68355
6341         * calls.c (initialize_argument_information): Call
6342         warn_parameter_passing_abi target hook.
6343         (store_one_arg): Use 0 for empty record size.  Don't push 0 size
6344         argument onto stack.
6345         (must_pass_in_stack_var_size_or_pad): Return false for empty types.
6346         * common.opt: Update -fabi-version description.
6347         * config/i386/i386.c (init_cumulative_args): Set cum->warn_empty.
6348         (ix86_gimplify_va_arg): Call arg_int_size_in_bytes instead of
6349         int_size_in_bytes.
6350         (ix86_is_empty_record): New function.
6351         (ix86_warn_parameter_passing_abi): New function.
6352         (TARGET_EMPTY_RECORD_P): Redefine.
6353         (TARGET_WARN_PARAMETER_PASSING_ABI): Redefine.
6354         * config/i386/i386.h (CUMULATIVE_ARGS): Add warn_empty.
6355         * doc/tm.texi: Regenerated.
6356         * doc/tm.texi.in (TARGET_EMPTY_RECORD_P,
6357         TARGET_WARN_PARAMETER_PASSING_ABI): Add.
6358         * dwarf2out.c (get_ultimate_context): Move to tree.c.
6359         * explow.c (hard_function_value): Call arg_int_size_in_bytes
6360         instead of int_size_in_bytes.
6361         * expr.c (copy_blkmode_to_reg): Likewise.
6362         * function.c (aggregate_value_p): Return 0 for empty types.
6363         (assign_parm_find_entry_rtl): Call warn_parameter_passing_abi
6364         target hook.
6365         (locate_and_pad_parm): Call arg size_in_bytes instead
6366         size_in_bytes.
6367         * lto-streamer-out.c (hash_tree): Hash TYPE_EMPTY_P and DECL_PADDING_P.
6368         * stor-layout.c (finalize_type_size): Set TYPE_EMPTY_P.
6369         * target.def (empty_record_p, warn_parameter_passing_abi): New target
6370         hooks.
6371         * targhooks.c (hook_void_CUMULATIVE_ARGS_tree): New hook.
6372         (std_gimplify_va_arg_expr): Skip empty records.  Call
6373         arg_size_in_bytes instead size_in_bytes.
6374         * targhooks.h (hook_void_CUMULATIVE_ARGS_tree): Declare.
6375         * tree-core.h (tree_type_common): Add empty_flag.
6376         (tree_decl_common): Update comments.
6377         * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Stream
6378         DECL_PADDING_P.
6379         (unpack_ts_type_common_value_fields): Stream TYPE_EMPTY_P.
6380         * tree-streamer-out.c (pack_ts_decl_common_value_fields): Stream
6381         DECL_PADDING_P.
6382         (pack_ts_type_common_value_fields): Stream TYPE_EMPTY_P.
6383         * tree.c (default_is_empty_type): New function.
6384         (default_is_empty_record): New function.
6385         (arg_int_size_in_bytes): New function.
6386         (arg_size_in_bytes): New function.
6387         (get_ultimate_context): New function.
6388         * tree.h: Define TYPE_EMPTY_P, DECL_PADDING_P and
6389         TRANSLATION_UNIT_WARN_EMPTY_P.
6390         (default_is_empty_record, arg_int_size_in_bytes,
6391         arg_size_in_bytes, get_ultimate_context): Declare.
6393 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6395         * config/arm/arm.c (cmse_clear_registers): New function.
6396         (cmse_nonsecure_call_clear_caller_saved): Replace register clearing
6397         code by call to cmse_clear_registers.
6398         (cmse_nonsecure_entry_clear_before_return): Likewise.
6400 2017-11-22  Tamar Christina  <tamar.christina@arm.com>
6402         * config/arm/arm_neon.h (vdot_u32, vdotq_u32)
6403         (vdot_s32, vdotq_s32): New.
6404         (vdot_lane_u32, vdotq_lane_u32): New.
6405         (vdot_lane_s32, vdotq_lane_s32): New.
6408 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6410         PR middle-end/82547
6411         * wide-int.cc (wi::add_large, wi::sub_large): Fix overflow detection
6412         for unsigned values with fewer HWIs than the precision.
6413         (test_overflow): New function.
6414         (wide_int_cc_tests): Call it.
6416 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6418         * emit-rtl.c (init_derived_machine_modes): Make sure ptr_mode
6419         has the same mode class as Pmode.
6421 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6423         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Get rid of
6424         padding_bits_to_clear_ptr.
6425         (cmse_nonsecure_entry_clear_before_return): Likewise.
6427 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6429         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Use
6430         auto_sbitap instead of integer bitfield to control register needing
6431         clearing.
6433 2017-11-22  Jakub Jelinek  <jakub@redhat.com>
6435         PR tree-optimization/83044
6436         * tree-vrp.c (vrp_prop::check_array_ref): If eltsize is not
6437         INTEGER_CST or is 0, clear up_bound{,_p1} and later ignore tests
6438         that need the upper bound.  Subtract offset from
6439         get_addr_base_and_unit_offset only if positive and subtract it
6440         before division by eltsize rather than after it.
6442         PR debug/83084
6443         * valtrack.c (propagate_for_debug_subst, propagate_for_debug): Reset
6444         debug insns if they would contain UNSPEC_VOLATILE or volatile asm.
6445         (dead_debug_insert_temp): Likewise, but also ignore even non-volatile
6446         asm.
6448         PR middle-end/82875
6449         * optabs.c (expand_doubleword_mult, expand_binop): Before calling
6450         expand_binop with *mul_widen_optab, make sure at least one of the
6451         operands doesn't have VOIDmode.
6453         PR debug/83034
6454         * dwarf2out.c (mem_loc_descriptor): Handle VEC_SERIES.
6456         PR rtl-optimization/82044
6457         PR tree-optimization/82042
6458         * dse.c (record_store): Check for overflow.
6459         (check_mem_read_rtx): Properly check for overflow if width == -1, call
6460         add_wild_read instead of clear_rhs_from_active_local_stores on
6461         overflow and log it into dump_file.
6463 2017-11-22  Richard Biener  <rguenther@suse.de>
6465         * gimple-iterator.c (gimple_find_edge_insert_loc): Ignore
6466         fake edges to exit when looking for a place to insert.
6467         * tree-ssa-pre.c (clear_expression_ids): Inline into callers
6468         and remove.
6469         (insert_into_preds_of_block): Commit edge insertion immediately,
6470         assert that doesn't require new BBs.
6471         (fini_pre): Release expressions.
6472         (pass_pre::execute): Shuffle things around a bit, if the fn
6473         is too large do not compute AVAIL either as this is really the
6474         quadratic bit.
6476 2017-11-22  Richard Biener  <rguenther@suse.de>
6478         PR tree-optimization/83089
6479         * tree-if-conv.c (pass_if_conversion::execute): If anything
6480         changed reset SCEV and free the number of iteration estimates.
6482 2017-11-21  Martin Sebor  <msebor@redhat.com>
6484         PR tree-optimization/82945
6485         * calls.h (warn_nonstring_bound): Remove unused function.
6487 2017-11-21  Martin Sebor  <msebor@redhat.com>
6489         PR tree-optimization/82945
6490         * builtins.c (expand_builtin_strlen): Call maybe_warn_nonstring_arg.
6491         * calls.h (maybe_warn_nonstring_arg): Declare new function.
6492         * calls.c (get_attr_nonstring_decl, maybe_warn_nonstring_arg): New
6493         functions.
6494         (initialize_argument_information): Call maybe_warn_nonstring_arg.
6495         * calls.h (get_attr_nonstring_decl): Declare new function.
6496         * doc/extend.texi (attribute nonstring): Update.
6497         * gimple-fold.c (gimple_fold_builtin_strncpy): Call
6498         get_attr_nonstring_decl and handle it.
6499         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Same.  Improve
6500         detection of nul-termination.
6501         (strlen_to_stridx): Change to a pointer.
6502         (handle_builtin_strlen, handle_builtin_stxncpy): Adjust.
6503         (pass_strlen::execute): Same.
6505 2017-11-21  Sergey Shalnov  <Sergey.Shalnov@intel.com>
6507         * config/i386/i386-opts.h (enum prefer_vector_width): Added new enum
6508         for the new option -mprefer-vector-width=[none|128|256|512].
6509         * config/i386/i386.c (ix86_target_string): remove old style options
6510         -mprefer-avx256 and make -mprefer-avx128 as alias.
6511         (ix86_option_override_internal):  Apply defaults for the
6512         -mprefer-vector-width=[128|256] option.
6513         * config/i386/i386.h (TARGET_PREFER_AVX128, TARGET_PREFER_AVX256):
6514         Implement macros to work with -mprefer-vector-width=.
6515         * config/i386/i386.opt: Implemented option
6516         -mprefer-vector-width=[none|128|256|512].
6517         * doc/invoke.texi: Documentation for
6518         -mprefer-vector-width=[none|128|256|512].
6520 2017-11-21  Pat Haugen  <pthaugen@us.ibm.com>
6522         * config/rs6000/ppc-asm.h (f50, vs50): Fix values.
6524 2017-11-21  Jonathan Wakely  <jwakely@redhat.com>
6526         * doc/invoke.texi (-Wmaybe-uninitialized): Rephrase for clarity.
6528 2017-11-21  Marc Glisse  <marc.glisse@inria.fr>
6530         * doc/generic.texi: Document POINTER_DIFF_EXPR, update
6531         POINTER_PLUS_EXPR.
6532         * cfgexpand.c (expand_debug_expr): Handle POINTER_DIFF_EXPR.
6533         * expr.c (expand_expr_real_2): Likewise.
6534         * fold-const.c (const_binop, fold_addr_of_array_ref_difference,
6535         fold_binary_loc): Likewise.
6536         * match.pd (X-X, P+(Q-P), &D-P, (P+N)-P, P-(P+N), (P+M)-(P+N),
6537         P-Q==0, -(A-B), X-Z<Y-Z, (X-Z)-(Y-Z), Z-X<Z-Y, (Z-X)-(Z-Y),
6538         (A-B)+(C-A)): New transformations for POINTER_DIFF_EXPR, based on
6539         MINUS_EXPR transformations.
6540         * optabs-tree.c (optab_for_tree_code): Handle POINTER_DIFF_EXPR.
6541         * tree-cfg.c (verify_expr, verify_gimple_assign_binary): Likewise.
6542         * tree-inline.c (estimate_operator_cost): Likewise.
6543         * tree-pretty-print.c (dump_generic_node, op_code_prio,
6544         op_symbol_code): Likewise.
6545         * tree-vect-stmts.c (vectorizable_operation): Likewise.
6546         * vr-values.c (extract_range_from_binary_expr): Likewise.
6547         * varasm.c (initializer_constant_valid_p_1): Likewise.
6548         * tree.def: New tree code POINTER_DIFF_EXPR.
6550 2017-11-21  Uros Bizjak  <ubizjak@gmail.com>
6552         * config/i386/i386.md (*bswap<mode>2_movbe): Add
6553         integer suffix to movbe mnemonic.
6554         (*bswaphi2_movbe): Ditto.
6555         (bswaphi_lowpart): Merge with *bswaphi_lowpart_1.
6557 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6559         PR c++/83045
6560         * tree-cfg.c (pass_warn_function_return::execute): Formatting fix.
6561         Also warn if seen __builtin_unreachable () call with BUILTINS_LOCATION.
6562         Use LOCATION_LOCUS when comparing against UNKNOWN_LOCATION.
6564 2017-11-21  Martin Liska  <mliska@suse.cz>
6566         * tree-inline.c (expand_call_inline): Remove not needed
6567         xstrdup_for_dump.
6569 2017-11-21  James Cowgill  <James.Cowgill@imgtec.com>
6570             Jakub Jelinek  <jakub@redhat.com>
6572         PR target/82880
6573         * config/mips/frame-header-opt.c (mips_register_frame_header_opt):
6574         Remove static keyword from f variable.
6576 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6578         PR tree-optimization/83086
6579         * gimple-ssa-store-merging.c
6580         (imm_store_chain_info::try_coalesce_bswap): Test this_n.base_addr
6581         rather than n.base_addr.
6583 2017-11-21  Martin Liska  <mliska@suse.cz>
6585         PR rtl-optimization/82044
6586         PR tree-optimization/82042
6587         * dse.c (check_mem_read_rtx): Check for overflow.
6589 2017-11-21  Aldy Hernandez  <aldyh@redhat.com>
6591         * tree-ssa-threadbackward.c (find_jump_threads_backwards): Fix
6592         typo in comment.
6594 2017-11-21  Claudiu Zissulescu  <claziss@synopsys.com>
6596         * cfgrtl.c (force_nonfallthru_and_redirect): Don't split a call
6597         and its corresponding call arg location note.
6599 2017-11-21  Claudiu Zissulescu  <claziss@synopsys.com>
6600             Andrew Burgess  <andrew.burgess@embecosm.com>
6602         * config/arc/arc-protos.h (arc_compute_frame_size): Delete
6603         declaration.
6604         (arc_return_slot_offset): Likewise.
6605         (arc_eh_return_address_location): New declaration.
6606         * config/arc/arc.c (TARGET_BUILTIN_SETJMP_FRAME_VALUE): Define.
6607         (MUST_SAVE_REGISTER): Add exception handler case.
6608         (MUST_SAVE_RETURN_ADDR): Likewise.
6609         (arc_frame_pointer_required): Likewise.
6610         (arc_frame_pointer_needed): New function.
6611         (arc_compute_frame_size): Changed.
6612         (arc_expand_prologue): Likewise.
6613         (arc_expand_epilogue): Likewise.
6614         (arc_initial_elimination_offset): Likewise.
6615         (arc_return_slot_offset): Delete.
6616         (arc_eh_return_address_location): New function.
6617         (arc_builtin_setjmp_frame_value): Likewise.
6618         * config/arc/arc.h (EH_RETURN_DATA_REGNO): Use 2 registers.
6619         (EH_RETURN_STACKADJ_RTX): Define.
6620         (EH_RETURN_HANDLER_RTX): Likewise.
6621         * config/arc/arc.md (eh_return): Delete.
6623 2017-11-21  Aldy Hernandez  <aldyh@redhat.com>
6625         * print-tree.h (debug_vec_tree): Remove prototype.
6626         * gdbinit.in (pvt): Remove macro.
6628 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6630         PR tree-optimization/83047
6631         * gimple-ssa-store-merging.c
6632         (imm_store_chain_info::output_merged_store): If the loads with the
6633         same vuse are in different basic blocks, for load_gsi pick a load
6634         location that is dominated by the other loads.
6636         PR c++/83059
6637         * config/i386/i386.c (ix86_memmodel_check): Start
6638         -Winvalid-memory-model diagnostics with lowercase letter.
6640         PR debug/82718
6641         * dwarf2out.c (dw_loc_list): If crtl->has_bb_partition, temporarily
6642         set in_cold_section_p to the partition containing loc_list->first.
6643         When seeing loc_list->last_before_switch node, update secname and
6644         perform range_across_switch second partition handling only after that.
6646         PR debug/82933
6647         * run-rtl-passes.c: Include debug.h.
6648         (run_rtl_passes): Call debug_hooks->assembly_start.
6649         * dwarf2out.c (dwarf2out_assembly_start): Return early if invoked
6650         multiple times.
6652         PR target/82981
6653         * internal-fn.c (expand_mul_overflow): Use OPTAB_WIDEN instead of
6654         OPTAB_DIRECT in calls to expand_simple_binop.
6656 2017-11-20  David Malcolm  <dmalcolm@redhat.com>
6658         PR c/81404
6659         * Makefile.in (C_COMMON_OBJS): Add c-family/known-headers.o.
6661 2017-11-20  David Malcolm  <dmalcolm@redhat.com>
6663         PR c++/72786
6664         * spellcheck.h (best_match::blithely_get_best_candidate): New
6665         accessor.
6667 2017-11-20  Jakub Jelinek  <jakub@redhat.com>
6669         * config/i386/i386.c (parse_mtune_ctrl_str): Start diagnostics
6670         with lower case letter.
6672 2017-11-20  Uros Bizjak  <ubizjak@gmail.com>
6674         * config/i386/i386.md (bswaphi2): New expander.
6675         (*bswaphi2_movbe): New insn pattern.
6676         (bswaphi -> rorhi pepehole2): New peephole pattern.
6678 2017-11-20  Jeff Law  <law@redhat.com>
6680         * Makefile.in (OBJS): Add gimple-ssa-evrp-analyze.o.
6681         * gimple-ssa-evrp-analyze.c: New file pulled from gimple-ssa-evrp.c.
6682         * gimple-ssa-evrp-analyze.h: New file pulled from gimple-ssa-evrp.c.
6683         * gimple-ssa-evrp.c: Remove bits moved into new files.  Include
6684         gimple-ssa-evrp-analyze.h.
6686         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Do not
6687         set BB_VISITED here.
6688         (evrp_range_analyzer::enter): Set BB_VISITED here instead.
6690 2017-11-20  Bin Cheng  <bin.cheng@arm.com>
6692         * tree-predcom.c: Add general comment on Store-Store chains.
6693         (split_data_refs_to_components): Postpone clearing eliminate_store_p
6694         flag in component.
6695         (get_chain_last_ref_at): Rename into...
6696         (get_chain_last_write_at): ...this.
6697         (get_chain_last_write_before_load): New function.
6698         (add_ref_to_chain): Promote type of chain from CT_STORE_LOAD to
6699         CT_STORE_STORE when write reference is added.
6700         (determine_roots_comp): Support load ref in CT_STORE_STORE chains.
6701         (is_inv_store_elimination_chain): Update get_chain_last_write_at call.
6702         (initialize_root_vars_store_elim_1): Ditto.
6703         (initialize_root_vars_store_elim_2): Ditto.  Replace rhs once default
6704         definition is created.
6705         (execute_pred_commoning_chain): Support load ref in CT_STORE_STORE
6706         chain by replacing it with dominant stored value.
6708 2017-11-20  Bin Cheng  <bin.cheng@arm.com>
6710         * tree-predcom.c (add_ref_to_chain): Remove check on distance.
6712 2017-11-20  Marc Glisse  <marc.glisse@inria.fr>
6714         * vr-values.c (extract_range_from_binary_expr): Use a full range
6715         for VR_VARYING.
6717 2017-11-20  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6719         * config/arm/arm.md (R4_REGNUM): Define constant.
6720         (nonsecure_call_internal): Remove r4 clobber.
6721         (nonsecure_call_value_internal): Likewise.
6722         * config/arm/thumb1.md (nonsecure_call_reg_thumb1_v5): Remove second
6723         clobber and resequence match_operands.
6724         (nonsecure_call_value_reg_thumb1_v5): Likewise.
6725         * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Likewise.
6726         (nonsecure_call_value_reg_thumb2): Likewise.
6728 2017-11-20  Jakub Jelinek  <jakub@redhat.com>
6730         PR tree-optimization/78821
6731         * gimple-ssa-store-merging.c (find_bswap_or_nop_load): Give up
6732         if base is TARGET_MEM_REF.  If base is not MEM_REF, set base_addr
6733         to the address of the base rather than the base itself.
6734         (find_bswap_or_nop_1): Just use pointer comparison for vuse check.
6735         (find_bswap_or_nop_finalize): New function.
6736         (find_bswap_or_nop): Use it.
6737         (bswap_replace): Return a tree rather than bool, change first
6738         argument from gimple * to gimple_stmt_iterator, allow inserting
6739         into an empty sequence, allow ins_stmt to be NULL - then emit
6740         all stmts into gsi.  Fix up MEM_REF address gimplification.
6741         (pass_optimize_bswap::execute): Adjust bswap_replace caller.
6742         (struct store_immediate_info): Add N and INS_STMT non-static
6743         data members.
6744         (store_immediate_info::store_immediate_info): Initialize them
6745         from newly added ctor args.
6746         (merged_store_group::apply_stores): Formatting fixes.  Sort by
6747         bitpos at the end.
6748         (stmts_may_clobber_ref_p): For stores call also
6749         refs_anti_dependent_p.
6750         (gather_bswap_load_refs): New function.
6751         (imm_store_chain_info::try_coalesce_bswap): New method.
6752         (imm_store_chain_info::coalesce_immediate_stores): Use it.
6753         (split_group): Handle LROTATE_EXPR and NOP_EXPR rhs_code specially.
6754         (imm_store_chain_info::output_merged_store): Fail if number of
6755         new estimated stmts is bigger or equal than old.  Handle LROTATE_EXPR
6756         and NOP_EXPR rhs_code.
6757         (pass_store_merging::process_store): Compute n and ins_stmt, if
6758         ins_stmt is non-NULL and the store rhs is otherwise invalid, use
6759         LROTATE_EXPR rhs_code.  Pass n and ins_stmt to store_immediate_info
6760         ctor.
6761         (pass_store_merging::execute): Calculate dominators.
6763         * tree-ssa-math-opts.c (nop_stats, bswap_stats, struct symbolic_number,
6764         BITS_PER_MARKER, MARKER_MASK, MARKER_BYTE_UNKNOWN, HEAD_MARKER, CMPNOP,
6765         CMPXCHG, do_shift_rotate, verify_symbolic_number_p,
6766         init_symbolic_number, find_bswap_or_nop_load, perform_symbolic_merge,
6767         find_bswap_or_nop_1, find_bswap_or_nop, pass_data_optimize_bswap,
6768         class pass_optimize_bswap, bswap_replace,
6769         pass_optimize_bswap::execute): Moved to ...
6770         * gimple-ssa-store-merging.c: ... this file.
6771         Include optabs-tree.h.
6772         (nop_stats, bswap_stats, do_shift_rotate, verify_symbolic_number_p,
6773         init_symbolic_number, find_bswap_or_nop_load, perform_symbolic_merge,
6774         find_bswap_or_nop_1, find_bswap_or_nop, bswap_replace): Put into
6775         anonymous namespace, remove static keywords.
6776         (pass_optimize_bswap::gate): Test BITS_PER_UNIT == 8 here...
6777         (pass_optimize_bswap::execute): ... rather than here.  Formatting fix.
6779 2017-11-20  Jan Hubicka  <hubicka@ucw.cz>
6781         PR bootstrap/83062
6782         * ipa-inline.c (can_inline_edge_p): Fix typo in previous patch.
6784 2017-11-20  Aldy Hernandez  <aldyh@redhat.com>
6786         * vec.h (debug_helper): New function.
6787         (DEFINE_DEBUG_VEC): New macro.
6788         * hash-set.h (debug_helper): New function.
6789         (DEFINE_DEBUG_HASH_SET): New macro.
6790         * cfg.c (debug_slim (edge)): New function.
6791         Call DEFINE_DEBUG_VEC for edges.
6792         Call DEFINE_DEBUG_HASH_SET for edges.
6793         * cfghooks.c (debug_slim (basic_block)): New function.
6794         Call DEFINE_DEBUG_VEC for basic blocks.
6795         Call DEFINE_DEBUG_HASH_SET for basic blocks.
6796         * print-tree.c (debug_slim): New function to handle trees.
6797         Call DEFINE_DEBUG_VEC for trees.
6798         Call DEFINE_DEBUG_HASH_SET for trees.
6799         (debug (vec<tree, va_gc>) &): Remove.
6800         (debug (<vec<tree, va_gc>) *): Remove.
6801         * print-rtl.c (debug_slim): New function to handle const_rtx.
6802         Call DEFINE_DEBUG_VEC for rtx_def.
6803         Call DEFINE_DEBUG_VEC for rtx_insn.
6804         Call DEFINE_DEBUG_HASH_SET for rtx_def.
6805         Call DEFINE_DEBUG_HASH_SET for rtx_insn.
6806         * sel-sched-dump.c (debug (vec<rtx_insn *> &): Remove.
6807         (debug (vec<rtx_insn *> *ptr): Remove.
6808         (debug_insn_vector): Remove.
6809         * stor-layout.c (debug_rli): Call debug() instead of debug_vec_tree.
6811 2017-11-20  Tom de Vries  <tom@codesourcery.com>
6813         PR rtl-optimization/82020
6814         * simplify-rtx.c (simplify_ternary_operation): Fix comparison mode of
6815         IF_THEN_ELSE condition.
6817 2017-11-19  Jeff Law  <law@redhat.com>
6819         * tree-ssa-dom.c (record_equivalences_from_phis): Fix handling
6820         of degenerates resulting from ignoring an edge.
6822 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6824         PR ipa/81360
6825         * ipa-inline.c (can_inline_edge_p): Also check that caller is optimized
6827 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6829         PR ipa/83001
6830         * profile-count.c (profile_count::to_sreal_scale): Fix return value
6831         for uninitialied counts.
6833 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6835         PR ipa/60243
6836         * tree-inline.c (estimate_num_insns): Set to 1 at least.
6838 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6840         PR target/82713
6841         * i386.c (ix86_builtin_vectorization_cost): Be ready for insane types.
6843 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6845         * config/arc/arc.h (FUNCTION_PROFILER): Remove semicolon after
6846         "do while (0)".
6848 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6850         * config/phoenix.h (TARGET_OS_CPP_BUILTINS): Remove semicolon after
6851         "do {} while (0)".
6853 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6855         * config/visium/visium.h (ASM_OUTPUT_CASE_END): Remove semicolon after
6856         macro body.
6858 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6860         * config/ft32/ft32.h (REGISTER_TARGET_PRAGMAS): Remove semicolon after
6861         "do {} while (0)".
6862         * config/spu/spu.h (REGISTER_TARGET_PRAGMAS): Same.
6864 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6866         * config/mcore/mcore-elf.h (MCORE_EXPORT_NAME): Remove semicolon after
6867         "do {} while (0)".
6868         * config/mcore/mcore.h (ASM_OUTPUT_ALIGNED_COMMON): After missing
6869         semicolon after MCORE_EXPORT_NAME call.
6871 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6873         PR target/82961
6874         * vmsdbgout.c (vmsdbgout_early_finish): New function.
6875         (vmsdbg_debug_hooks): Set early_finish field to vmsdbgout_early_finish.
6877 2017-11-18  Jan Hubicka  <hubicka@ucw.cz>
6879         * cgraphclones.c (cgraph_edge::clone): Rename gcov_count to prof_count.
6880         (cgraph_edge::clone): Cleanup updating of profile.
6881         * ipa-cp.c (update_profiling_info): Likewise.
6882         * ipa-inline-transform.c (inline_transform): Likewise.
6883         * ipa-inline.c (inline_small_functions): Add missing space to dump.
6884         * ipa-split.c (execute_split_functions): Do not split when function
6885         is cold.
6886         * predict.c (estimate_bb_frequencies): Cleanup updating of profile.
6887         * profile-count.c (profile_count::dump): Add global0.
6888         (profile_count::to_cgraph_frequency): Do not ICE when entry is
6889         undefined.
6890         (profile_count::to_sreal_scale): Likewise.
6891         (profile_count::adjust_for_ipa_scaling): Fix typo in comment.
6892         (profile_count::combine_with_ipa_count): New function.
6893         * profile-count.h (profile_guessed_global0adjusted): New.
6894         (profile_count::adjusted_zero): New.
6895         (profile_count::global0adjusted): New.
6896         (profile_count::combine_with_ipa_count): New.
6897         * tree-inline.c (copy_edges_for_bb): Add NUM/DEN arugment;
6898         correct profile of return block of split functions.
6899         (copy_cfg_body): Remove unused profile_count.
6900         (copy_body): Likewise.
6901         (expand_call_inline): Update.
6902         (tree_function_versioning): Update.
6904 2017-11-18  Aldy Hernandez  <aldyh@redhat.com>
6906         * hash-set.h (hash_set::empty): New.
6907         * tree-ssa-threadbackward.h: Delete.
6908         * tree-ssa-threadbackward.c (class thread_jumps): New.
6909         Move max_threaded_paths into class.
6910         (fsm_find_thread_path): Remove arguments that are now in class.
6911         (profitable_jump_thread_path): Rename to...
6912         (thread_jumps::profitable_jump_thread_path): ...this.
6913         (convert_and_register_jump_thread_path): Rename to...
6914         (thread_jumps::convert_and_register_current_path): ...this.
6915         (check_subpath_and_update_thread_path): Rename to...
6916         (thread_jumps::check_subpath_and_update_thread_path): ...this.
6917         (register_jump_thread_path_if_profitable): Rename to...
6918         (thread_jumps::register_jump_thread_path_if_profitable): ...this.
6919         (handle_phi): Rename to...
6920         (thread_jumps::handle_phi): ...this.
6921         (handle_assignment): Rename to...
6922         (thread_jumps::handle_assignment): ...this.
6923         (fsm_find_control_statement_thread_paths): Rename to...
6924         (thread_jumps::fsm_find_control_statement_thread_paths): ...this.
6925         (find_jump_threads_backwards): Rename to...
6926         (thread_jumps::find_jump_threads_backwards): ...this.
6927         Initialize path local data.
6928         (pass_thread_jumps::execute): Call find_jump_threads_backwards
6929         from within thread_jumps class.
6930         (pass_early_thread_jumps::execute): Same.
6932 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6934         * cgraph.c (cgraph_node::dump): Do IPA sanity checking on IPA counts.
6936 2017-11-17  Steve Ellcey  <sellcey@cavium.com>
6938         * config/aarch64/aarch64-simd.md (fnma<mode>4): Move neg operator
6939         to canonical location.
6941 2017-11-17  Steve Ellcey  <sellcey@cavium.com>
6943         PR target/81356
6944         * config/aarch64/aarch64.c (aarch64_use_by_pieces_infrastructure_p):
6945         Remove.
6946         (TARGET_USE_BY_PIECES_INFRASTRUCTURE_P): Remove define.
6948 2017-11-17  H.J. Lu  <hongjiu.lu@intel.com>
6950         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Use
6951         rcrt1.o%s/grcrt1.o%s for -static-pie.
6953 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6955         * i386.c (ix86_multiplication_cost, ix86_division_cost,
6956         ix86_shift_rotate_cost): Break out from ...
6957         (ix86_rtx_costs): ... here.
6958         (ix86_add_stmt_cost): Use rtx cost machinery to compute cost of
6959         vector operations.
6961 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6963         * predict.c (determine_unlikely_bbs): Set cgraph node count to 0
6964         when entry block was promoted unlikely.
6965         (estimate_bb_frequencies): Increase frequency scale.
6966         * profile-count.h (profile_count): Export precision info.
6968 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6970         * tree-tailcall.c (eliminate_tail_call): Be more careful about not
6971         disturbin profile of entry block.
6973 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6975         * ipa-fnsummary.c (estimate_node_size_and_time): Be more tolerant for
6976         roundoff errors.
6978 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6980         * ipa-cp.c (update_profiling_info): Handle conversion to local profile.
6981         * tree-cfg.c (execute_fixup_cfg): Do fixup same way as inliner does.
6983 2017-11-17  Jeff Law  <law@redhat.com>
6985         * gimple-ssa-evrp.c (class evrp_range_analyzer): New class extracted
6986         from evrp_dom_walker class.  Various methods moved into new class.
6987         (evrp_range_analyzer::evrp_range_analyzer): Constructor for new class.
6988         (evrp_range_analyzer::enter): New method.
6989         (evrp_range_analyzer::leave): New method.
6990         (evrp_dom_walker): Remove delegators no longer needed by this class.
6991         Replace vr_values data member with evrp_range_analyzer
6993         * gimple-ssa-evrp.c (evrp_dom_walker::record_ranges_from_phis): New
6994         method extracted from evrp_dom_walker::before_dom_children.
6995         (evrp_dom_walker::record_ranges_from_stmt): Likewise.
6996         (evrp_dom_walker::record_ranges_from_incoming_edge): Likewise.
6998         * gimple-ssa-evrp.c (evrp_dom_walker): Add cleanup method.
6999         Add private copy constructor and move assignment operators.
7000         Privatize methods and class data where trivially possible.
7001         (evrp_dom_walker::cleanup): New function, extracted from
7002         execute_early_vrp.  Simplify access to class data.
7004         * vr-values.h (get_output_for_vrp): Prototype.
7005         * vr-values.c (get_output_for_vrp): New function extracted from
7006         vrp_visit_assignment_or_call and extract_range_from_stmt.
7007         (vrp_visit_assignment_or_call): Use get_output_for_vrp.  Simplify.
7009 2017-11-17  Luis Machado  <luis.machado@linaro.org>
7011         * config/aarch64/aarch64.c
7012         (qdf24xx_prefetch_tune) <default_opt_level>: Set to -1.
7013         (qdf24xx_tunings) <autoprefetcher_model>: Set to
7014         tune_params::AUTOPREFETCHER_WEAK.
7016 2017-11-17  Tamar Christina  <tamar.christina@arm.com>
7018         PR target/82641
7019         * config/arm/arm.c (arm_valid_target_attribute_rec):
7020         Parse "arch=" and "+<ext>".
7021         (arm_valid_target_attribute_tree): Re-init global options.
7022         (arm_option_override): Make non-static.
7023         (arm_options_perform_arch_sanity_checks): Make errors fatal.
7024         * gcc/config/arm/arm-c.c (__ARM_FEATURE_CMSE): Support undef.
7025         (__ARM_FEATURE_CRC32): Support undef.
7026         * config/arm/arm_acle.h (__ARM_FEATURE_CRC32): Replace with pragma.
7027         * doc/extend.texi (ARM Function Attributes): Add pragma and target.
7029 2017-11-17  David Malcolm  <dmalcolm@redhat.com>
7031         * gdbinit.in (break-on-diagnostic): New command.
7033 2017-11-17  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
7035         * config/i386/i386.c (ix86_expand_epilogue): Change simple
7036         return to indirect jump for EH return if control-flow
7037         protection is enabled. Change explicit 'false' argument in
7038         pro_epilogue_adjust_stack with a value of
7039         flag_cf_protection.
7040         * config/i386/i386.md (simple_return_indirect_internal):
7041         Remove SImode restriction to support 64-bit.
7043 2017-11-17  Segher Boessenkool  <segher@kernel.crashing.org>
7045         * combine.c (added_notes_insn): New.
7046         (try_combine): Handle added_notes_insn like added_links_insn.
7047         Rewrite return value code.
7048         (distribute_notes): Set added_notes_insn to the earliest insn we added
7049         a note to.
7051 2017-11-17  Segher Boessenkool  <segher@kernel.crashing.org>
7053         PR rtl-optimization/82621
7054         * combine.c (try_combine): Do not split PARALLELs of two SETs if the
7055         dest of one of those SETs is unused.
7057 2017-11-17  Richard Biener  <rguenther@suse.de>
7059         PR fortran/83017
7060         * tree-core.h (enum annot_expr_kind): Add annot_expr_parallel_kind.
7061         * tree-pretty-print.c (dump_generic_node): Handle
7062         annot_expr_parallel_kind.
7063         * tree-cfg.c (replace_loop_annotate_in_block): Likewise.
7064         * gimplify.c (gimple_boolify): Likewise.
7066 2017-11-17  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
7068         * config.gcc (extra_headers): Add cet.h for x86 targets.
7069         * config/i386/cet.h: New file.
7070         * doc/install.texi: Add --enable-cet/--disable-cet.
7072 2017-11-17  Richard Biener  <rguenther@suse.de>
7074         PR tree-optimization/83017
7075         * tree-parloops.c (MIN_PER_THREAD): Use --param parloops-min-per-thread.
7076         (gen_parallel_loop): Properly count iterations.
7077         (parallelize_loops): Handle loop->can_be_parallel independent
7078         of flag_loop_parallelize_all.  Make static profitability test match
7079         the runtime one.
7080         * params.def (PARAM_PARLOOPS_MIN_PER_THREAD): New.
7081         * invoke.texi (parloops-min-per-thread): Document.
7083 2017-11-17  Vineet Gupta  <vgupta@synopsys.com>
7085         * config/arc/linux.h: GLIBC_DYNAMIC_LINKER update per glibc
7086         upstreaming review comments.
7088 2017-11-17  Tamar Christina  <tamar.christina@arm.com>
7090         * expr.c (copy_blkmode_to_reg): Fix bitsize for targets
7091         with fast unaligned access.
7092         * doc/sourcebuild.texi (word_mode_no_slow_unalign): New.
7094 2017-11-17  Thomas Preud'homme  <thomas.preudhomme@arm.com>
7096         * config/arm/arm.c (cmse_nonsecure_entry_clear_before_return): Allocate
7097         to_clear_arg_regs_bitmap to the same size as to_clear_bitmap.
7099 2017-11-17  Richard Biener  <rguenther@suse.de>
7101         * tree-ssa-pre.c (phi_translate_1): Remove redundant constant
7102         folding of references.
7104 2017-11-17  Qing Zhao  <qing.zhao@oracle.com>
7106         PR middle-end/78809
7107         * gimple-fold.c (gimple_fold_builtin_string_compare): Add handling
7108         of replacing call to strncmp with corresponding call to strcmp when
7109         meeting conditions.
7111 2017-11-17  Sergey Shalnov  <Sergey.Shalnov@intel.com>
7113         * config/i386/x86-tune.def (X86_TUNE_AVX256_OPTIMAL): Add tuning
7114         option prefer-avx256 for skylake-avx512 configuration.
7115         * config/i386/i386.c (ix86_option_override_internal): Ditto.
7116         (get_builtin_code_for_version): Ditto.
7118 2017-11-17  Chung-Ju Wu  <jasonwucj@gmail.com>
7119             Monk Chiang  <sh.chiang04@gmail.com>
7121         * config/nds32/nds32.h (FIRST_PSEUDO_REGISTER): Modify.
7122         (FIXED_REGISTERS): Reserve more register numbers.
7123         (CALL_USED_REGISTERS): Likewise.
7124         (REG_ALLOC_ORDER): Likewise.
7125         (REG_CLASS_CONTENTS): Likewise.
7126         (REGISTER_NAMES): Likewise.
7128 2017-11-17  Chung-Ju Wu  <jasonwucj@gmail.com>
7129             Kito Cheng  <kito.cheng@gmail.com>
7131         * config/nds32/nds32-modes.def: Add vector mode V4QI V2HI V8QI V4HI
7132         V2SI.
7133         * config/nds32/iterators.md: Add vector mode iterators and attributes.
7135 2017-11-16  Steven Munroe  <munroesj@gcc.gnu.org>
7137         * config/rs6000/mmintrin.h (_mm_add_pi32[_ARCH_PWR]): Correct
7138         parameter list for vec_splats.
7140 2017-11-16  Joseph Myers  <joseph@codesourcery.com>
7142         * doc/invoke.texi (-std=c17): Refer to 2018 expected publication
7143         date of C17.
7144         (-std=c18, -std=iso9899:2018, -std=gnu18): Document option aliases.
7146 2017-11-16  Thomas Koenig  <tkoenig@gcc.gnu.org>
7148         PR bootstrap/82856
7149         * doc/install.texi: Document incompatibility of Perl >=5.6.26
7150         with the required version of automake 1.11.6.
7152 2017-11-16  Pat Haugen  <pthaugen@us.ibm.com>
7154         * rs6000/power9.md (power9fpdiv): New automaton and cpu_unit defined
7155         for it.
7156         (DU_C2_3_power9): Correct reservation combinations.
7157         (FP_DIV_power9, VEC_DIV_power9): New.
7158         (power9-alu): Split out rotate/shift...
7159         (power9-rot): ...to here, correct dispatch resource.
7160         (power9-cracked-alu, power9-mul, power9-mul-compare): Correct dispatch
7161         resource.
7162         (power9-fp): Correct latency.
7163         (power9-sdiv): Add div/sqrt resource.
7164         (power9-ddiv): Correct latency, add div/sqrt resource.
7165         (power9-sqrt, power9-dsqrt): Add div/sqrt resource.
7166         (power9-vecfdiv, power9-vecdiv): Correct latency, add div/sqrt
7167         resource.
7168         (power9-qpdiv, power9-qpmul): Adjust resource usage.
7170 2017-11-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
7172         * config/rs6000/rs6000.c (rs6000_expand_builtin): Do not do the
7173         switch statement mapping KF built-ins to TF built-ins if we don't
7174         have the proper ISA 3.0 assembler support.
7176 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7178         * tree-emutls.c (lower_emutls_data): Remove unused bb_freq.
7179         (lower_emutls_function_body): Do not compute it.
7181 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7183         * ipa-split.c (split_bb_info): Turn time to sreal.
7184         (split_point): Likewise.
7185         (dump_split_point): Likewise.
7186         (fine_split_points): Likewise.
7187         (execute_split_functions): Only zero split_bbs; turn time to sreals.
7189 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7191         * ipa-fnsummary.c (analyze_function_body): Accumulate time consistently
7192         in sreal.
7194 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7196         * predict.c (combine_predictions_for_bb): Preserve zero predicted
7197         edges.
7198         (expensive_function_p): Remove useless assert.
7199         (determine_unlikely_bbs): Propagate also forward; determine cold blocks
7201 2017-11-16  Martin Sebor  <msebor@redhat.com>
7203         PR tree-optimization/82588
7204         PR tree-optimization/82583
7205         * tree-vrp.c (check_array_ref): Handle flexible array members,
7206         string literals, and inner indices.
7207         (search_for_addr_array): Add detail to diagnostics.
7209 2017-11-16  Nathan Sidwell  <nathan@acm.org>
7211         PR c++/82836
7212         PR c++/82737
7213         * tree.h (COPY_DECL_RTL): Rename parms for clarity.
7214         (SET_DECL_ASSEMBLER_NAME): Forward to
7215         overwrite_decl_assembler_name.
7216         (COPY_DECL_ASSEMBLER_NAME): Rename parms for clarity.
7217         (overwrite_decl_assembler_name): Declare.
7218         * tree.c (overwrite_decl_assembler_name): New.
7219         * langhooks-def.h (lhd_overwrite_decl_assembler_name): Declare.
7220         (LANG_HOOKS_OVERWRITE_DECL_ASSEMBLER_NAME): Provide default.
7221         (LANG_HOOKS_INITIALIZER): Add it.
7222         * langhooks.h (struct lang_hooks): Add overwrite_decl_assembler_name.
7223         * langhooks.c (lhd_set_decl_assembler_name): Use
7224         SET_DECL_ASSEMBLER_NAME.
7225         (lhd_overwrite_decl_assembler_name): Default implementation.
7227 2017-11-16  Wilco Dijkstra  <wdijkstr@arm.com>
7228             Jackson Woodruff  <jackson.woodruff@arm.com>
7230         PR tree-optimization/71026
7231         * match.pd: Canonicalize constant multiplies in division.
7233 2017-11-16  Wilco Dijkstra  <wdijkstr@arm.com>
7235         * opts.c (default_options_table): Add OPT_fomit_frame_pointer entry.
7236         * common/config/alpha/alpha-common.c (TARGET_OPTION_OPTIMIZATION_TABLE):
7237         Remove OPT_fomit_frame_pointer entry.
7238         * common/config/arc/arc-common.c: Likewise.
7239         * common/config/arm/arm-common.c: Likewise.
7240         * common/config/avr/avr-common.c: Likewise.
7241         * common/config/c6x/c6x-common.c: Likewise.
7242         * common/config/cr16/cr16-common.c: Likewise.
7243         * common/config/cris/cris-common.c: Likewise.
7244         * common/config/epiphany/epiphany-common.c: Likewise.
7245         * common/config/fr30/fr30-common.c: Likewise.
7246         * common/config/frv/frv-common.c: Likewise.
7247         * common/config/ia64/ia64-common.c: Likewise.
7248         * common/config/iq2000/iq2000-common.c: Likewise.
7249         * common/config/lm32/lm32-common.c: Likewise.
7250         * common/config/m32r/m32r-common.c: Likewise.
7251         * common/config/mcore/mcore-common.c: Likewise.
7252         * common/config/microblaze/microblaze-common.c: Likewise.
7253         * common/config/mips/mips-common.c: Likewise.
7254         * common/config/mmix/mmix-common.c: Likewise.
7255         * common/config/mn10300/mn10300-common.c: Likewise.
7256         * common/config/nios2/nios2-common.c: Likewise.
7257         * common/config/pa/pa-common.c: Likewise.
7258         * common/config/pdp11/pdp11-common.c: Likewise.
7259         * common/config/powerpcspe/powerpcspe-common.c: Likewise.
7260         * common/config/riscv/riscv-common.c: Likewise.
7261         * common/config/rs6000/rs6000-common.c: Likewise.
7262         * common/config/rx/rx-common.c: Likewise.
7263         * common/config/s390/s390-common.c: Likewise.
7264         * common/config/sh/sh-common.c: Likewise.
7265         * common/config/sparc/sparc-common.c: Likewise.
7266         * common/config/tilegx/tilegx-common.c: Likewise.
7267         * common/config/tilepro/tilepro-common.c: Likewise.
7268         * common/config/v850/v850-common.c: Likewise.
7269         * common/config/visium/visium-common.c: Likewise.
7270         * common/config/xstormy16/xstormy16-common.c: Likewise.
7271         * common/config/xtensa/xtensa-common.c: Likewise.
7272         * invoke.texi (-fomit-frame-pointer): Update documentation.
7274 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7276         * tree-cfg.c (gimple_find_sub_bbs): Do not compute freq.
7278 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7280         * cfg.c (scale_bbs_frequencies_int,
7281         cale_bbs_frequencies_gcov_type): Remove.
7282         * cfg.h (scale_bbs_frequencies_int,
7283         cale_bbs_frequencies_gcov_type): Remove.
7285 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7287         * tree-ssa-loop-manip.c
7288         (scale_dominated_blocks_in_loop): Update to profile counts.
7289         (tree_transform_and_unroll_loop): Likewise.
7291 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7293         * tree-vect-loop-manip.c (vect_do_peeling): Do not use
7294         scale_bbs_frequencies_int.
7296 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7298         * final.c (compute_alignments): Use counts rather than frequencies.
7300 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7302         * cfgloopanal.c: Include sreal.h
7303         (average_num_loop_insns): Use counts and sreal for accounting.
7305 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7307         * cfgloopmanip.c (duplicate_loop_to_header_edge): Cleanup profile
7308         manipulation.
7310 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7311             Kito Cheng  <kito.cheng@gmail.com>
7313         * config/nds32/constraints.md: Provide more constraints.
7314         * config/nds32/nds32.h (enum reg_class, REG_CLASS_NAMES,
7315         REG_CLASS_CONTENTS): Define R5_REG, R8_REG, and FRAME_POINTER_REG to
7316         support constraints usage.
7318 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7320         * config/nds32/constants.md (UNSPEC_VOLATILE_FUNC_RETURN): Remove.
7322 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7323             Kito Cheng  <kito.cheng@gmail.com>
7325         * config/nds32/nds32.opt: Add mext-perf, mext-perf2, mext-string.
7326         * config/nds32/nds32.opt: Refine the layout.
7327         * config/nds32/nds32.c (TARGET_EXT_PERF, TARGET_EXT_PERF2,
7328         TARGET_EXT_STRING): Support new options.
7329         * config/nds32/nds32.h: Likewise.
7330         * config/nds32/nds32.md: Likewise.
7331         * config/nds32/nds32-predicates.c: Likewise.
7332         * config/nds32/constraints.md: Likewise.
7333         * common/config/nds32/nds32-common.c: Likewise.
7335 2017-11-16  Julia Koval  <julia.koval@intel.com>
7337         PR target/82983
7338         * config/i386/gfniintrin.h: Add sse check.
7339         * config/i386/i386.c (ix86_expand_builtin): Fix gfni check.
7341 2017-11-16  Julia Koval  <julia.koval@intel.com>
7343         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VBMI2_SET,
7344         OPTION_MASK_ISA_AVX512VBMI2_UNSET): New.
7345         (ix86_handle_option): Handle -mavx512vbmi2.
7346         * config/i386/cpuid.h: Add bit_AVX512VBMI2.
7347         * config/i386/driver-i386.c (host_detect_local_cpu): Handle new bit.
7348         * config/i386/i386-c.c (__AVX512VBMI2__): New.
7349         * config/i386/i386.c (ix86_target_string): Handle -mavx512vbmi2.
7350         (ix86_valid_target_attribute_inner_p): Ditto.
7351         * config/i386/i386.h (TARGET_AVX512VBMI2, TARGET_AVX512VBMI2_P): New.
7352         * config/i386/i386.opt (mavx512vbmi2): New option.
7353         * doc/invoke.texi: Add new option.
7355 2017-11-16  Julia Koval  <julia.koval@intel.com>
7357         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8, _mm256_gf2p8mul_epi8,
7358         _mm_mask_gf2p8mul_epi8, _mm_maskz_gf2p8mul_epi8,
7359         _mm256_mask_gf2p8mul_epi8, _mm256_maskz_gf2p8mul_epi8,
7360         _mm512_mask_gf2p8mul_epi8, _mm512_maskz_gf2p8mul_epi8,
7361         _mm512_gf2p8mul_epi8): New intrinsics.
7362         * config/i386/i386-builtin-types.def
7363         (V64QI_FTYPE_V64QI_V64QI): New type.
7364         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8mulb_v64qi,
7365         __builtin_ia32_vgf2p8mulb_v64qi_mask, __builtin_ia32_vgf2p8mulb_v32qi,
7366         __builtin_ia32_vgf2p8mulb_v32qi_mask, __builtin_ia32_vgf2p8mulb_v16qi,
7367         __builtin_ia32_vgf2p8mulb_v16qi_mask): New builtins.
7368         * config/i386/sse.md (vgf2p8mulb_*): New pattern.
7369         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
7371 2017-11-15  Uros Bizjak  <ubizjak@gmail.com>
7373         * config/i386/i386.c (x86_print_call_or_nop): Emit 5 byte nop
7374         explicitly as a stream of bytes.
7376 2017-11-15  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
7378         * config/rs6000/altivec.h (vec_xst_be): New #define.
7379         * config/rs6000/altivec.md (altivec_vperm_<mode>_direct): Rename
7380         and externalize from *altivec_vperm_<mode>_internal.
7381         * config/rs6000/rs6000-builtin.def (XL_BE_V16QI): Remove macro
7382         instantiation.
7383         (XL_BE_V8HI): Likewise.
7384         (XL_BE_V4SI): Likewise.
7385         (XL_BE_V4SI): Likewise.
7386         (XL_BE_V2DI): Likewise.
7387         (XL_BE_V4SF): Likewise.
7388         (XL_BE_V2DF): Likewise.
7389         (XST_BE): Add BU_VSX_OVERLOAD_X macro instantiation.
7390         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Correct
7391         all array entries with these keys: VSX_BUILTIN_VEC_XL,
7392         VSX_BUILTIN_VEC_XL_BE, VSX_BUILTIN_VEC_XST.  Add entries for key
7393         VSX_BUILTIN_VEC_XST_BE.
7394         * config/rs6000/rs6000.c (altivec_expand_xl_be_builtin): Remove.
7395         (altivec_expand_builtin): Remove handling for VSX_BUILTIN_XL_BE_*
7396         built-ins.
7397         (altivec_init_builtins): Replace conditional calls to def_builtin
7398         for __builtin_vsx_ld_elemrev_{v8hi,v16qi} and
7399         __builtin_vsx_st_elemrev_{v8hi,v16qi} based on TARGET_P9_VECTOR
7400         with unconditional calls.  Remove calls to def_builtin for
7401         __builtin_vsx_le_be_<mode>.  Add a call to def_builtin for
7402         __builtin_vec_xst_be.
7403         * config/rs6000/vsx.md (vsx_ld_elemrev_v8hi): Convert define_insn
7404         to define_expand, and add alternate RTL generation for P8.
7405         (*vsx_ld_elemrev_v8hi_internal): New define_insn based on
7406         vsx_ld_elemrev_v8hi.
7407         (vsx_ld_elemrev_v16qi): Convert define_insn to define_expand, and
7408         add alternate RTL generation for P8.
7409         (*vsx_ld_elemrev_v16qi_internal): New define_insn based on
7410         vsx_ld_elemrev_v16qi.
7411         (vsx_st_elemrev_v8hi): Convert define_insn
7412         to define_expand, and add alternate RTL generation for P8.
7413         (*vsx_st_elemrev_v8hi_internal): New define_insn based on
7414         vsx_st_elemrev_v8hi.
7415         (vsx_st_elemrev_v16qi): Convert define_insn to define_expand, and
7416         add alternate RTL generation for P8.
7417         (*vsx_st_elemrev_v16qi_internal): New define_insn based on
7418         vsx_st_elemrev_v16qi.
7420 2017-11-15  H.J. Lu  <hongjiu.lu@intel.com>
7422         PR target/82990
7423         * config/i386/i386.c (pass_insert_vzeroupper::gate): Remove
7424         TARGET_AVX512ER check.
7425         (ix86_option_override_internal): Set MASK_VZEROUPPER if
7426         neither -mzeroupper nor -mno-zeroupper is used and
7427         TARGET_EMIT_VZEROUPPER is set.
7428         * config/i386/i386.h (TARGET_EMIT_VZEROUPPER): New.
7429         * config/i386/x86-tune.def: Add X86_TUNE_EMIT_VZEROUPPER.
7431 2017-11-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
7433         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
7434         folding of vector compares.
7435         (fold_build_vec_cmp): New helper function.
7436         (fold_compare_helper): New helper function.
7437         (builtin_function_type): Add compare builtins to the list of functions
7438         having unsigned arguments.  Cosmetic updates to comment indentation.
7439         * config/rs6000/vsx.md (vcmpneb, vcmpneh, vcmpnew): Update to specify
7440         the not+eq combination.
7442 2017-11-15  Bin Cheng  <bin.cheng@arm.com>
7444         PR tree-optimization/82726
7445         PR tree-optimization/70754
7446         * tree-predcom.c (order_drefs_by_pos): New function.
7447         (combine_chains): Move code setting has_max_use_after to...
7448         (try_combine_chains): ...here.  New parameter.  Sort combined chains
7449         according to position information.
7450         (tree_predictive_commoning_loop): Update call to above function.
7451         (update_pos_for_combined_chains, pcom_stmt_dominates_stmt_p): New.
7453 2017-11-15  Bin Cheng  <bin.cheng@arm.com>
7455         PR tree-optimization/82726
7456         Revert
7457         2017-01-23  Bin Cheng  <bin.cheng@arm.com>
7459         PR tree-optimization/70754
7460         * tree-predcom.c (stmt_combining_refs): New parameter INSERT_BEFORE.
7461         (reassociate_to_the_same_stmt): New parameter INSERT_BEFORE.  Insert
7462         combined stmt before it if not NULL.
7463         (combine_chains): Process refs reversely and compute dominance point
7464         for root ref.
7466         Revert
7467         2017-02-23  Bin Cheng  <bin.cheng@arm.com>
7469         PR tree-optimization/79663
7470         * tree-predcom.c (combine_chains): Process refs in reverse order
7471         only for ZERO length chains, and add explaining comment.
7473 2017-11-15  Tamar Christina  <tamar.christina@arm.com>
7475         * config/arm/arm-cpus.in (armv8_3, ARMv8_3a, armv8.3-a): New
7476         * config/arm/arm-tables.opt (armv8.3-a): Regenerated.
7477         * doc/invoke.texi (ARM Options): Add armv8.3-a.
7479 2017-11-15  Tamar Christina  <tamar.christina@arm.com>
7481         * config/arm/arm.h (TARGET_DOTPROD): Add arm_arch8_2.
7483 2017-11-15  Martin Liska  <mliska@suse.cz>
7485         * tree-cfg.c (pass_warn_function_return::execute):
7486         Compare warn_return_type for greater than zero.
7488 2017-11-15  Sebastian Peryt  <sebastian.peryt@intel.com>
7490         PR target/82941
7491         PR target/82942
7492         * config/i386/i386.c (pass_insert_vzeroupper): Modify gate condition
7493         to return true on Xeon and not on Xeon Phi.
7494         (ix86_check_avx256_register): Changed to ...
7495         (ix86_check_avx_upper_register): ... this. Add extra check for
7496         VALID_AVX512F_REG_OR_XI_MODE.
7497         (ix86_avx_u128_mode_needed): Changed
7498         ix86_check_avx256_register to ix86_check_avx_upper_register.
7499         (ix86_check_avx256_stores): Changed to ...
7500         (ix86_check_avx_upper_stores): ... this. Changed
7501         ix86_check_avx256_register to ix86_check_avx_upper_register.
7502         (ix86_avx_u128_mode_after): Changed
7503         avx_reg256_found to avx_upper_reg_found. Changed
7504         ix86_check_avx256_stores to ix86_check_avx_upper_stores.
7505         (ix86_avx_u128_mode_entry): Changed
7506         ix86_check_avx256_register to ix86_check_avx_upper_register.
7507         (ix86_avx_u128_mode_exit): Ditto.
7508         * config/i386/i386.h: (host_detect_local_cpu): New define.
7510 2017-11-15  Dominik Infuehr  <dominik.infuehr@theobroma-systems.com>
7512         * config/arm/xgene1.md (xgene1): Split into automatons
7513         xgene1_main, xgene1_decoder, xgene1_div, xgene1_simd.
7514         (xgene1_f_load): Adjust reservations and/or types.
7515         (xgene1_f_store): Likewise.
7516         (xgene1_load_pair): Likewise.
7517         (xgene1_store_pair): Likewise.
7518         (xgene1_fp_load1): Likewise.
7519         (xgene1_load1): Likewise.
7520         (xgene1_store1): Likewise.
7521         (xgene1_move): Likewise.
7522         (xgene1_alu): Likewise.
7523         (xgene1_simd): Likewise.
7524         (xgene1_bfm): Likewise.
7525         (xgene1_neon_load1): Likewise.
7526         (xgene1_neon_store1): Likewise.
7527         (xgene1_neon_logic): Likewise.
7528         (xgene1_neon_st1): Likewise.
7529         (xgene1_neon_ld1r): Likewise.
7530         (xgene1_alu_cond): Added.
7531         (xgene1_shift_reg): Likwise.
7532         (xgene1_bfx): Likewise.
7533         (xgene1_mul): Split into xgene1_mul32, xgene1_mul64.
7535 2017-11-15  Jakub Jelinek  <jakub@redhat.com>
7537         PR target/82981
7538         * internal-fn.c: Include gimple-ssa.h, tree-phinodes.h and
7539         ssa-iterators.h.
7540         (can_widen_mult_without_libcall): New function.
7541         (expand_mul_overflow): If only checking unsigned mul overflow,
7542         not result, and can do efficiently MULT_HIGHPART_EXPR, emit that.
7543         Don't use WIDEN_MULT_EXPR if it would involve a libcall, unless
7544         no other way works.  Add MULT_HIGHPART_EXPR + MULT_EXPR support.
7545         (expand_DIVMOD): Formatting fix.
7546         * expmed.h (expand_mult): Add NO_LIBCALL argument.
7547         * expmed.c (expand_mult): Likewise.  Use OPTAB_WIDEN rather
7548         than OPTAB_LIB_WIDEN if NO_LIBCALL is true, and allow it to fail.
7550         PR tree-optimization/82977
7551         * tree-ssa-strlen.c (strlen_optimize_stmt): Pass a reference to a copy
7552         constructed temporary to strlen_to_stridx.put.
7554 2017-11-15  Martin Liska  <mliska@suse.cz>
7556         * configure.ac: Remove -fkeep-inline-functions from coverage_flags.
7557         * configure: Regenerate.
7559 2017-11-15  Martin Liska  <mliska@suse.cz>
7561         PR target/82927
7562         * config/sh/sh-mem.cc: Use proper probability for
7563         REG_BR_PROB_NOTE.
7565 2017-11-14  Jeff Law  <law@redhat.com>
7567         * explow.c (anti_adjust_stack_and_probe_stack_clash): Avoid probing
7568         the red zone for stack_clash_protection_final_dynamic_probe targets
7569         when the total dynamic stack size is zero bytes.
7571         * tree-ssa-threadupdate.c (thread_through_all_blocks): Thread
7572         blocks is post order.
7574 2017-11-15  Alexandre Oliva  <aoliva@redhat.com>
7576         * dumpfile.h (TDF_COMPARE_DEBUG): New.
7577         * final.c (rest_of_clean_state): Set it for the
7578         -fcompare-debug dump.
7579         * tree-pretty-print.c (dump_generic_node): Omit OBJ_TYPE_REF
7580         class when TDF_COMPARE_DEBUG is set.
7582         * dwarf2out.c (gen_producer_string): Discard
7583         OPT_fcompare_debug.
7585 2017-11-15  Joseph Myers  <joseph@codesourcery.com>
7587         PR c/81156
7588         * doc/extend.texi (Other Builtins): Document __builtin_tgmath.
7589         * ginclude/tgmath.h (__tg_cplx, __tg_ldbl, __tg_dbl, __tg_choose)
7590         (__tg_choose_2, __tg_choose_3, __TGMATH_REAL_1_2)
7591         (__TGMATH_REAL_2_3): Remove macros.
7592         (__TGMATH_CPLX, __TGMATH_CPLX_2, __TGMATH_REAL, __TGMATH_REAL_2)
7593         (__TGMATH_REAL_3, __TGMATH_CPLX_ONLY): Define using
7594         __builtin_tgmath.
7595         (frexp, ldexp, nexttoward, scalbn, scalbln): Define using
7596         __TGMATH_REAL_2.
7597         (remquo): Define using __TGMATH_REAL_3.
7599 2017-11-14  Jeff Law  <law@redhat.com>
7601         * vr-values.c: New file with contents extracted from tree-vrp.c.
7602         * Makefile.in (OBJS): Add vr-values.o
7603         * tree-vrp.h (set_value_range_to_nonnull): Prototype.
7604         (set_value_range, set_and_canonicalize_value_range): Likewise.
7605         (vrp_bitmap_equal_p, range_is_nonnull): Likewise.
7606         (value_range_constant_singleton, symbolic_range_p): Likewise.
7607         (compare_values, compare_values_warnv, vrp_val_is_min): Likewise.
7608         (vrp_val_is_max, copy_value_range, set_value_range_to_value): Likewise.
7609         (extract_range_from_binary_expr_1, vrp_val_min, vrp_val_max): Likewise.
7610         (set_value_range_to_null, range_int_cst_p, opreand_less_p): Likewise.
7611         (find_case_label_range, find_case_label_index): Likewise.
7612         (zero_nonzero_bits_from_vr, overflow_comparison_p): Likewise.
7613         (range_int_cst_singleton_p, value_inside_range): Likewise.
7614         (get_single_symbol): Likewise.
7615         (switch_update): Move structure definition here.
7616         (to_remove_edges, to_update_switch_stmts): Provide externs.
7617         * tree-vrp.c: Move all methods for vr-values class to vr-values.c
7618         (vrp_val_max, vrp_val_min, vrp_val_is_max): Make externally visible.
7619         (vrp_val_is_min, set_value_range): Likewise.
7620         (set_and_canonicalize_value_range, copy_value_range): Likewise.
7621         (set_value_range_to_value, set_value_range_to_nonnull): Likewise.
7622         (set_value_range_to_null, vrp_bitmap_equal_p): Likewise.
7623         (range_is_nonnull, range_int_cst_p): Likewwise.
7624         (range_int_cst_singleton_p, symbolic_range_p): Likewise.
7625         (get_single_symbol, operand_less_p): Likewise
7626         (compare_values_warnv, compare_values): Likewise.
7627         (value_inside_range, value_range_constant_singleton): Likewise.
7628         (zero_nonzero_bitgs_from_vr): Likewise.
7629         (extract_range_from_binary_expr_1): Likewise.
7630         (overflow_comparison_p): Likewise.
7631         (to_remove_edges, to_update_switch_stmts): Likewise.
7632         (find_case_label-index, find_case_label_range): Likewise.
7633         (switch_update, set_value_range_to_nonnegative): Remove.
7634         (set_value_range_to_truthvalue): Likewise.
7635         (symbolic_range_based_on_p, gimple_assign_nonzero_p): Likewise.
7636         (gimple_stmt_nonzero_p, compare_ranges): Likewise.
7637         (compare_range_with_value, vrp_valueize, vrp_valueize_1): Likewise.
7638         (find_case_label_ranges, test_for_singularity): Likewise.
7639         (range_fits_type_p, simplify_conversion_using_ranges): LIkewise.
7640         (x_vr_values): Move to its remaining use site.
7642 2017-11-10  Jeff Law  <law@redhat.com>
7644         * vr-values.h (VR_INITIALIZER): Move #define here.
7645         * gimple-ssa-evrp.c: New file with contents extracted from tree-vrp.c
7646         * Makefile.in (OBJS): Add tree-evrp.o
7647         * tree-vrp.h (assert_info): Move structure definition here.
7648         (set_value_range_to_varying): Prototype.
7649         (vrp_operand_equal_p, range_includes_zero_p): Likewise.
7650         (infer_value_range, register_edge_assert_for): Likewise.
7651         (stmt_interesting_for_vrp): Likewise.
7652         * tree-vrp.c: Move all methods for evrp class into gimple-ssa-evrp.c.
7653         (set_value_range_to_varying): No longer static.
7654         (vrp_operand_equal_p, range_includes_zero_p): Likewise.
7655         (infer_value_range, register_edge_assert_for): Likewise.
7657 2017-11-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
7659         * config/rs6000/rs6000.md (bswapdi2): On 32-bit ISA 3.0, don't
7660         generate the XXBRD instruction.
7662         * config/rs6000/rs6000-c.c (is_float128_p): New helper function.
7663         (rs6000_builtin_type_compatible): Treat _Float128 and long double
7664         as being compatible if -mabi=ieeelongdouble.
7665         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_HW_1): New macros
7666         to setup float128 built-ins with hardware support.
7667         (BU_FLOAT128_HW_2): Likewise.
7668         (BU_FLOAT128_HW_3): Likewise.
7669         (BU_FLOAT128_HW_VSX_1): Likewise.
7670         (BU_FLOAT128_HW_VSX_2): Likewise.
7671         (scalar_extract_expq): Change float128 built-in functions to
7672         accommodate having both KFmode and TFmode functions.  Use the
7673         KFmode variant as the default.
7674         (scalar_extract_sigq): Likewise.
7675         (scalar_test_neg_qp): Likewise.
7676         (scalar_insert_exp_q): Likewise.
7677         (scalar_insert_exp_qp): Likewise.
7678         (scalar_test_data_class_qp): Likewise.
7679         (sqrtf128_round_to_odd): Delete processing the round to odd
7680         built-in functions as special built-in functions, and define them
7681         as float128 built-ins.  Use the KFmode variant as the default.
7682         (truncf128_round_to_odd): Likewise.
7683         (addf128_round_to_odd): Likewise.
7684         (subf128_round_to_odd): Likewise.
7685         (mulf128_round_to_odd): Likewise.
7686         (divf128_round_to_odd): Likewise.
7687         (fmaf128_round_to_odd): Likewise.
7688         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Add
7689         support for KFmode and TFmode xststdcqp calls.
7690         (rs6000_expand_builtin): If long double is IEEE 128-bit floating
7691         point, switch the built-in handlers for the get/set float128
7692         exponent, get float128 mantissa, float128 test built-ins, and the
7693         float128 round to odd built-in functions.  Eliminate creating the
7694         float128 round to odd built-in functions as special built-ins.
7695         (rs6000_init_builtins): Eliminate special creation of the float128
7696         round to odd built-in functions.
7697         * config/rs6000/vsx.md (xsxexpqp_<mode>): Change float128 built-in
7698         function insns to support both TFmode and KFmode variants.
7699         (xsxsigqp_<mode>): Likewise.
7700         (xsiexpqpf_<mode>): Likewise.
7701         (xsiexpqp_<mode>): Likewise.
7702         (xststdcqp_<mode>): Likewise.
7703         (xststdcnegqp_<mode>): Likewise.
7704         (xststdcqp_<mode>): Likewise.
7706 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7708         * tree-ssa-threadupdate.c (compute_path_counts): Remove
7709         unused path_in_freq_ptr parameter.
7710         (ssa_fix_duplicate_block_edges): Do not pass around path_in_freq
7712 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7714         * ipa-inline.c (edge_badness): Dump sreal frequency.
7715         (compute_inlined_call_time): Match natural implementaiton ...
7716         * ipa-fnsummary.c (estimate_edge_size_and_time): ... here; remove
7717         forgotten division by CGRAPH_FREQ_BASE.
7719 2017-11-14  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
7721         * config.gcc (*-*-solaris2*): Enable default_use_cxa_atexit since
7722         Solaris 11.  Update comment.
7723         * configure.ac (gcc_cv_ld_pid): Adapt comment for Solaris 12
7724         renaming.
7725         * config/sol2.h (STARTFILE_SPEC): Likewise.
7726         * configure: Regenerate.
7728 2017-11-14  Carl Love  <cel@us.ibm.com>
7730         * config/rs6000/rs6000.c (swap_endian_selector_for_mode): Remove
7731         le_ and be_ prefixes to swap* variables.  Remove
7732         if (VECTOR_ELT_ORDER_BIG) statement. Remove E_V16QImode case
7733         statements.
7735 2017-11-14  Jason Merrill  <jason@redhat.com>
7737         Support GTY((cache)) on hash_map.
7738         * hash-traits.h (ggc_remove): Add ggc_maybe_mx member function.
7739         (ggc_cache_remove): Override it instead of ggc_mx.
7740         * hash-table.h (gt_ggc_mx): Call it instead of ggc_mx.
7741         (gt_cleare_cache): Call ggc_mx instead of gt_ggc_mx.
7742         * hash-map-traits.h (simple_hashmap_traits): Add maybe_mx member.
7743         (simple_cache_map_traits): Override maybe_mx.
7744         * hash-map.h (hash_entry): Add ggc_maybe_mx and keep_cache_entry.
7745         (hash_map): Friend gt_cleare_cache.
7746         (gt_cleare_cache): New.
7747         * tree.h (tree_cache_traits): New hash_map traits class.
7748         (tree_cache_map): New typedef.
7750 2017-11-14  Richard Biener  <rguenther@suse.de>
7752         * tree-cfgcleanup.c (cleanup_control_expr_graph): Remove first_p
7753         paramter and handling.
7754         (cleanup_control_flow_bb): Likewise.
7755         (cleanup_control_flow_pre): New helper performing a DFS walk
7756         to call cleanup_control_flow_bb in PRE order.
7757         (cleanup_tree_cfg_1): Do the first phase of cleanup_control_flow_bb
7758         via cleanup_control_flow_pre.
7760 2017-11-14  James Greenhalgh  <james.greenhalgh@arm.com>
7762         * config/aarch64/aarch64-simd.md
7763         (aarch64_simd_bsl<mode>_internal): Remove DImode.
7764         (*aarch64_simd_bsl<mode>_alt): Likewise.
7765         (aarch64_simd_bsldi_internal): New.
7766         (aarch64_simd_bsldi_alt): Likewise.
7768 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7770         * tracer.c (better_p): Do not compare frequencies.
7771         * reg-stack.c (better_edge): Likewise.
7772         * shrink-wrap.c (try_shrink_wrapping): Do not convert to gcov counts
7773         and back.
7775 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7777         * auto-profile.c (afdo_annotate_cfg): Use update_max_bb_count.
7778         * cgraphunit.c (cgraph_node::expand_thunk): Use update_max_bb_count.
7779         * ipa-utils.c (ipa_merge_profiles): Use update_max_bb_count.
7780         * lto-streamer-in.c (input_function): Use update_max_bb_count.
7781         * omp-expand.c (expand_omp_taskreg): Use update_max_bb_count.
7782         * predict.c (maybe_hot_frequency_p): Inline to ...
7783         (maybe_hot_count_p): ... here; rewrite to counts.
7784         (counts_to_freqs): Rename to ...
7785         (update_max_bb_count): ... this one.
7786         (expensive_function_p): Use counts.
7787         (estimate_bb_frequencies): Update.
7788         (rebuild_frequencies): Update.
7789         * predict.h (counts_to_freqs): Rename to ...
7790         (update_max_bb_count): ... this one.
7791         * profile.c (compute_branch_probabilities): Add debug info
7792         * tree-inline.c (expand_call_inline): Update debug info.
7793         (optimize_inline_calls): Use update_max_bb_count..
7794         (tree_function_versioning): Use update_max_bb_count..
7795         * value-prof.c (gimple_value_profile_transformations):
7796         Do not use update_max_bb_count.
7798 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7800         * ipa-inline.c (compute_uninlined_call_time, compute_inlined_call_time):
7801         always use frequencies.
7803 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7805         * bb-reorder.c: Remove frequencies from comments.
7806         (better_edge_p): Use profile counts.
7807         (find_traces): Dump profile counts.
7808         (rotate_loop): Use profile counts.
7809         (find_traces_1_round): Likewise.
7810         (connect_better_edge_p): Use counts instead of probabilities for
7811         reverse walk.
7812         (copy_bb_p): Drop early check for non-0 frequency.
7813         (sanitize_hot_paths): Update comments.
7815 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7817         * ipa-split.c (struct split_point): Add count.
7818         (consider_split): Do not compute incoming frequency; compute incoming
7819         count and store it to split_point.
7820         (split_function): Set count of the call to split part correctly.
7822 2017-11-13  Carl Love  <cel@us.ibm.com>
7824         * config/rs6000/altivec.md (altivec_vsumsws_be): Add define_expand.
7826 2017-11-13  Tom Tromey  <tom@tromey.com>
7828         * doc/cpp.texi (Variadic Macros): Document __VA_OPT__.
7830 2017-11-13  Carl Love  <cel@us.ibm.com>
7832         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
7833         Add support for builtins:
7834         unsigned int vec_first_{,miss}_match_{,or_eos}index,
7835         vector {un,}signed {char,int,short},
7836         vector {un,}signed {char,int,short}) arguments.
7837         * config/rs6000/rs6000-builtin.def (VFIRSTMATCHINDEX,
7838         VFIRSTMATCHOREOSINDEX, VFIRSTMISMATCHINDEX, VFIRSTMISMATCHOREOSINDEX):
7839         Add BU_P9V_AV_2 expansions for the builtins.
7840         * config/rs6000/altivec.h (vec_first_match_index,
7841         vec_first_mismatch_index, vec_first_match_or_eos_index,
7842         vec_first_mismatch_or_eos_index): Add #defines for the builtins.
7843         * config/rs6000/rs6000-protos.h (bytes_in_mode): Add
7844         new extern declaration.
7845         * config/rs6000/rs6000.c (bytes_in_mode): Add new function.
7846         * config/rs6000/vsx.md (first_match_index_<mode>,
7847         first_match_or_eos_index_<mode>, first_mismatch_index_<mode>,
7848         first_mismatch_or_eos_index_<mode>): Add define expand.
7849         (vctzlsbb_<mode>): Add mode field to define_insn for vctzlsbb.
7850         * doc/extend.texi: Update the built-in documenation file for the new
7851         built-in functions.
7853 2017-11-13  Michael Meissner  <meissner@linux.vnet.ibm.com>
7855         * match.pd: Convert fminf<N>, fminf<N>x, fmax<N>, and fmax<N>x
7856         into the min/max operations for _Float<N> and _Float<N>X types.
7858 2017-11-13  Eric Botcazou  <ebotcazou@adacore.com>
7860         PR lto/81351
7861         * dwarf2out.c (do_eh_frame): New static variable.
7862         (dwarf2out_begin_prologue): Set it.
7863         (dwarf2out_frame_finish): Test it instead of dwarf2out_do_eh_frame.
7865 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7867         * tree-ssa-coalesce.c (coalesce_cost): Fix formating.
7869         * tree-ssa-sink.c (select_best_block): Do not use frequencies.
7871 2017-11-13  Eric Botcazou  <ebotcazou@adacore.com>
7873         PR lto/81351
7874         * debug.h (dwarf2out_do_eh_frame): Declare.
7875         * dwarf2cfi.c (dwarf2out_do_eh_frame): New predicate.
7876         (dwarf2out_do_frame): Use it.
7877         (dwarf2out_do_cfi_asm): Likewise.
7878         * dwarf2out.c (dwarf2out_frame_finish): Likewise.
7879         (dwarf2out_assembly_start): Likewise.
7880         (dwarf2out_begin_prologue): Fix comment.
7881         * toplev.c (compile_file): Always call dwarf2out_frame_finish
7882         if the target needs either debug or unwind DWARF2 info.
7883         * lto-opts.c (lto_write_options): Do not save -fexceptions,
7884         -fnon-call-exceptions, -ffp-contract, -fmath-errno, -fsigned-zeros,
7885         -ftrapping-math, -ftrapv and -fwrapv.
7887 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7889         * cgraph.c (cgraph_edge::sreal_frequency): New function.
7890         * cgraph.h (cgraph_edge::sreal_frequency): Declare.
7891         * ipa-fnsummary.c (dump_ipa_call_summary): Use sreal_frequency.
7892         (estimate_edge_size_and_time): Likewise.
7893         (ipa_merge_fn_summary_after_inlining): Likewise.
7894         * ipa-inline.c (cgraph_freq_base_rec): Remove.
7895         (compute_uninlined_call_time): Use sreal_frequency.
7896         (compute_inlined_call_time): Likewise.
7897         (ipa_inline): Do not initialize cgraph_freq_base_rec.
7898         * profile-count.c: Include sreal.h.
7899         (profile_count::to_sreal_scale): New.
7900         * profile-count.h: Forward declare sreal.
7901         (profile_count::to_sreal_scale): Declare.
7903 2017-11-13  Nathan Sidwell  <nathan@acm.org>
7905         * diagnostic.c (maybe_line_and_column): New.
7906         (diagnostic_get_location_text): Use it.
7907         (diagnostic_report_current_module): Likewise.
7908         (test_diagnostic_get_location_text): Add tests.
7910 2017-11-13  Luis Machado  <luis.machado@linaro.org>
7912         * doc/md.texi (Specifying processor pipeline description): Fix
7913         incorrect latency for the div instruction example.
7915 2017-11-13  Jakub Jelinek  <jakub@redhat.com>
7917         PR tree-optimization/78821
7918         * gimple-ssa-store-merging.c (compatible_load_p): Don't require
7919         that bit_not_p is the same.
7920         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
7921         (split_group): Count precisely bit_not_p bits in each statement.
7922         (invert_op): New function.
7923         (imm_store_chain_info::output_merged_store): Use invert_op to
7924         emit BIT_XOR_EXPR with a xor_mask instead of BIT_NOT_EXPR if some
7925         but not all orig_stores have BIT_NOT_EXPR in the corresponding spots.
7927 2017-11-13  Martin Liska  <mliska@suse.cz>
7929         * gcov.c (struct coverage_info): Remove typedef of coverage_t.
7930         (struct source_info): Likewise.
7931         (add_branch_counts): Likewise.
7932         (add_line_counts): Likewise.
7933         (function_summary): Likewise.
7934         (output_intermediate_line): Likewise.
7935         (generate_results): Likewise.
7937 2017-11-13  Martin Liska  <mliska@suse.cz>
7939         * gcov.c (struct block_info): Remove typedef for block_t.
7940         (struct line_info): Likewise.
7941         (line_info::has_block): Likewise.
7942         (EXIT_BLOCK): Likewise.
7943         (unblock): Likewise.
7944         (circuit): Likewise.
7945         (get_cycles_count): Likewise.
7946         (process_file): Likewise.
7947         (read_graph_file): Likewise.
7948         (solve_flow_graph): Likewise.
7949         (find_exception_blocks): Likewise.
7950         (add_line_counts): Likewise.
7951         (accumulate_line_info): Likewise.
7952         (output_line_details): Likewise.
7954 2017-11-13  Martin Liska  <mliska@suse.cz>
7956         * gcov.c (struct arc_info): Remove typedef for arc_t.
7957         (struct line_info): Likewise.
7958         (add_branch_counts): Likewise.
7959         (output_branch_count): Likewise.
7960         (function_info::~function_info): Likewise.
7961         (circuit): Likewise.
7962         (output_intermediate_line): Likewise.
7963         (read_graph_file): Likewise.
7964         (solve_flow_graph): Likewise.
7965         (find_exception_blocks): Likewise.
7966         (add_line_counts): Likewise.
7967         (accumulate_line_info): Likewise.
7968         (output_line_details): Likewise.
7969         (output_function_details): Likewise.
7971 2017-11-13  Martin Liska  <mliska@suse.cz>
7973         * gcov.c (struct function_info): Remove typedef for function_t.
7974         (struct source_info): Likewise.
7975         (source_info::get_functions_at_location): Likewise.
7976         (solve_flow_graph): Likewise.
7977         (find_exception_blocks): Likewise.
7978         (add_line_counts): Likewise.
7979         (output_intermediate_file): Likewise.
7980         (process_file): Likewise.
7981         (generate_results): Likewise.
7982         (release_structures): Likewise.
7983         (read_graph_file): Likewise.
7984         (read_count_file): Likewise.
7985         (accumulate_line_counts): Likewise.
7986         (output_lines): Likewise.
7988 2017-11-13  Martin Liska  <mliska@suse.cz>
7990         * gcov.c (function_info::function_info): Remove num_counts
7991         and add vector<gcov_type>.
7992         (function_info::~function_info): Use the vector.
7993         (process_file): Likewise.
7994         (read_graph_file): Likewise.
7995         (read_count_file): Likewise.
7996         (solve_flow_graph): Likewise.
7998 2017-11-13  Martin Liska  <mliska@suse.cz>
8000         * gcov.c (function_info::is_artificial): New function.
8001         (process_file): Erase all artificial early.
8002         (generate_results): Skip as all artificial are already
8003         removed.
8005 2017-11-13  Martin Liska  <mliska@suse.cz>
8007         * gcov.c (read_graph_file): Store to global vector of functions.
8008         (read_count_file): Iterate the vector.
8009         (process_file): Likewise.
8010         (generate_results): Likewise.
8011         (release_structures): Likewise.
8013 2017-11-13  Jakub Jelinek  <jakub@redhat.com>
8015         PR tree-optimization/82954
8016         * gimple-ssa-store-merging.c
8017         (imm_store_chain_info::coalesce_immediate_stores): If
8018         !infof->ops[N].base_addr, split group if info->ops[N].base_addr.
8020 2017-11-13  Richard Sandiford  <richard.sandiford@linaro.org>
8022         * config/aarch64/aarch64-simd.md (aarch64_store_lane0<mode>):
8023         Upddate call to ENDIAN_LANE_N.
8024         (aarch64_<sur>dot_lane<vsi2qi>): Use aarch64_endian_lane_rtx.
8025         (aarch64_<sur>dot_laneq<vsi2qi>): Likewise.
8026         (*aarch64_simd_vec_copy_lane<mode>): Update calls to ENDIAN_LANE_N
8027         and use aarch64_endian_lane_rtx.
8028         (*aarch64_simd_vec_copy_lane_<vswap_width_name><mode>): Likewise.
8030 2017-11-12  Tom de Vries  <tom@codesourcery.com>
8032         * config/riscv/riscv.h (ASM_OUTPUT_LABELREF): Wrap in do {} while (0).
8034 2017-11-12  Tom de Vries  <tom@codesourcery.com>
8036         * config/elfos.h (ASM_OUTPUT_ASCII): Remove semicolon after macro body.
8038 2017-11-12  Tom de Vries  <tom@codesourcery.com>
8040         * config/cr16/cr16.h (ASM_OUTPUT_LABELREF): Remove semicolon after macro
8041         body.
8042         * config/powerpcspe/xcoff.h (ASM_OUTPUT_LABELREF): Same.
8043         * config/rs6000/xcoff.h (ASM_OUTPUT_LABELREF): Same.
8044         * defaults.h (ASM_OUTPUT_LABELREF): Same.
8046 2017-11-11  Martin Sebor  <msebor@redhat.com>
8048         PR c/81117
8049         * doc/extend.texi (attribute nonstring): Remove spurious argument.
8051         PR bootstrap/82948
8052         * prefic.c (translate_name): Replace strncpy with memcpy to
8053         avoid -Wstringop-truncation.
8055 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
8057         * tree-ssa-loop-im.c (execute_sm_if_changed): Do not compute freq_sum.
8059 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
8061         * predict.c (maybe_hot_frequency_p): Do not use cfun.
8063 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
8065         * tree-ssa-tail-merge.c (replace_block_by): Fix and re-enable profile
8066         merging.
8068 2017-11-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
8070         * config/rs6000/rs6000.md (bswaphi2_reg): On ISA 3.0 systems,
8071         enable generating XXBRH if the value is in a vector register.
8072         (bswapsi2_reg): On ISA 3.0 systems, enable generating XXBRW if the
8073         value is in a vector register.
8074         (bswapdi2_reg): On ISA 3.0 systems, always use XXBRD to do
8075         register to register bswap64's instead of doing the GPR sequence
8076         used on previous machines.
8077         (bswapdi2_xxbrd): New insn.
8078         (bswapdi2_reg): Disallow on ISA 3.0.
8079         (register to register bswap64 splitter): Do not split the insn on
8080         ISA 3.0 systems that use XXBRD.
8082 2017-11-10  Martin Sebor  <msebor@redhat.com>
8084         PR c/81117
8085         * config/darwin-c.c (framework_construct_pathname): Replace strncpy
8086         with memcpy.
8087         (find_subframework_file): Same.
8089 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
8091         * auto-profile.c (afdo_indirect_call): Drop frequency.
8092         * cgraph.c (symbol_table::create_edge): Drop frequency argument.
8093         (cgraph_node::create_edge): Drop frequency argument.
8094         (cgraph_node::create_indirect_edge): Drop frequency argument.
8095         (cgraph_edge::make_speculative): Drop frequency arguments.
8096         (cgraph_edge::resolve_speculation): Do not update frequencies
8097         (cgraph_edge::dump_edge_flags): Do not dump frequency.
8098         (cgraph_node::dump): Check consistency in IPA mode.
8099         (cgraph_edge::maybe_hot_p): Use IPA counter.
8100         (cgraph_edge::verify_count_and_frequency): Rename to ...
8101         (cgraph_edge::verify_count): ... this one; drop frequency checking.
8102         (cgraph_node::verify_node): Update.
8103         * cgraph.h (struct cgraph_edge): Drop frequency.
8104         (cgraph_edge::frequency): New function.
8105         * cgraphbuild.c (pass_build_cgraph_edges::execute): Donot pass
8106         frequencies.
8107         (cgraph_edge::rebuild_edges): Likewise.
8108         * cgraphclones.c (cgraph_edge::clone): Scale only counts.
8109         (duplicate_thunk_for_node): Do not pass frequency.
8110         (cgraph_node::create_clone): Scale only counts.
8111         (cgraph_node::create_virtual_clone): Do not pass frequency.
8112         (cgraph_node::create_edge_including_clones): Do not pass frequency.
8113         (cgraph_node::create_version_clone): Do not pass frequency.
8114         * cgraphunit.c (cgraph_node::analyze): Do not pass frequency.
8115         (cgraph_node::expand_thunk): Do not pass frequency.
8116         (cgraph_node::create_wrapper): Do not pass frequency.
8117         * gimple-iterator.c (update_call_edge_frequencies): Do not pass
8118         frequency.
8119         * gimple-streamer-in.c (input_bb): Scale only IPA counts.
8120         * ipa-chkp.c (chkp_produce_thunks): Do not pass frequency.
8121         * ipa-cp.c (ipcp_lattice::print): Use frequency function.
8122         (gather_caller_stats): Use frequency function.
8123         (ipcp_cloning_candidate_p): Use frequency function.
8124         (ipcp_propagate_stage): Use frequency function.
8125         (get_info_about_necessary_edges): Use frequency function.
8126         (update_profiling_info): Update only IPA profile.
8127         (update_specialized_profile): Use frequency functoin.
8128         (perhaps_add_new_callers): Update only IPA profile.
8129         * ipa-devirt.c (ipa_devirt): Use IPA profile.
8130         * ipa-fnsummary.c (redirect_to_unreachable): Do not set frequrency.
8131         (dump_ipa_call_summary): Use frequency function.
8132         (estimate_edge_size_and_time): Use frequency function.
8133         (ipa_merge_fn_summary_after_inlining): Use frequency function.
8134         * ipa-inline-analysis.c (do_estimate_edge_time): Use IPA profile.
8135         * ipa-inline-transform.c (update_noncloned_frequencies): Rename to ..
8136         (update_noncloned_counts): ... ths one; scale counts only.
8137         (clone_inlined_nodes): Do not scale frequency.
8138         (inline_call): Do not pass frequency.
8139         * ipa-inline.c (compute_uninlined_call_time): Use IPA profile.
8140         (compute_inlined_call_time): Use IPA profile.
8141         (want_inline_small_function_p): Use IPA profile.
8142         (want_inline_self_recursive_call_p): Use IPA profile.
8143         (edge_badness): Use IPA profile.
8144         (lookup_recursive_calls): Use IPA profile.
8145         (recursive_inlining): Do not pass frequency.
8146         (resolve_noninline_speculation): Do not update frequency.
8147         (inline_small_functions): Collect max of IPA profile.
8148         (dump_overall_stats): Dump IPA porfile.
8149         (dump_inline_stats): Dump IPA porfile.
8150         (ipa_inline): Collect IPA stats.
8151         * ipa-inline.h (clone_inlined_nodes): Update prototype.
8152         * ipa-profile.c (ipa_propagate_frequency_1): Use frequency function.
8153         (ipa_propagate_frequency): Use frequency function.
8154         (ipa_profile): Cleanup.
8155         * ipa-prop.c (ipa_make_edge_direct_to_target): Do not pass frequency
8156         * ipa-utils.c (ipa_merge_profiles): Merge all profiles.
8157         * lto-cgraph.c (lto_output_edge): Do not stream frequency.
8158         (input_node): Do not stream frequency.
8159         (input_edge): Do not stream frequency.
8160         (merge_profile_summaries): Scale only IPA profiles.
8161         * omp-simd-clone.c (simd_clone_adjust): Do not pass frequency.
8162         * predict.c (drop_profile): Do not recompute frequency.
8163         * trans-mem.c (ipa_tm_insert_irr_call): Do not pass frequency.
8164         (ipa_tm_insert_gettmclone_call): Do not pass frequency.
8165         * tree-cfg.c (execute_fixup_cfg): Drop profile to global0 if needed.
8166         * tree-chkp.c (chkp_copy_bounds_for_assign): Do not pass frequency.
8167         * tree-emutls.c (gen_emutls_addr): Do not pass frequency.
8168         * tree-inline.c (copy_bb): Do not scale frequency.
8169         (expand_call_inline): Do not scale frequency.
8170         (tree_function_versioning): Do not scale frequency.
8171         * ubsan.c (ubsan_create_edge): Do not pass frequency.
8173 2017-11-10  Julia Koval  <julia.koval@intel.com>
8175         * config/i386/gfniintrin.h (_mm_gf2p8affine_epi64_epi8)
8176         (_mm256_gf2p8affine_epi64_epi8, _mm_mask_gf2p8affine_epi64_epi8)
8177         (_mm_maskz_gf2p8affine_epi64_epi8, _mm256_mask_gf2p8affine_epi64_epi8)
8178         (_mm256_maskz_gf2p8affine_epi64_epi8)
8179         (_mm512_mask_gf2p8affine_epi64_epi8, _mm512_gf2p8affine_epi64_epi8)
8180         (_mm512_maskz_gf2p8affine_epi64_epi8): New intrinsics.
8181         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8affineqb_v64qi)
8182         (__builtin_ia32_vgf2p8affineqb_v32qi)
8183         (__builtin_ia32_vgf2p8affineqb_v16qi): New builtins.
8184         * config/i386/sse.md (vgf2p8affineqb_<mode><mask_name>): New pattern.
8186 2017-11-10  Tamar Christina  <tamar.christina@arm.com>
8188         PR target/82641
8189         * config/arm/arm.c
8190         (arm_option_override): Refactor.
8191         (arm_option_reconfigure_globals): New.
8192         (arm_options_perform_arch_sanity_checks): New.
8193         * config/arm/arm-protos.h (arm_option_reconfigure_globals):
8194         New prototype.
8195         (arm_options_perform_arch_sanity_checks): Likewise
8197 2017-11-10  Pat Haugen  <pthaugen@us.ibm.com>
8199         * rs6000/power9.md (power9-qpdiv): Correct DFU pipe usage.
8200         (power9-qpmul): New.
8201         * rs6000/rs6000.md ("type" attr): Add qmul.
8202         (mul<mode>3, fma<mode>4_hw, *fms<mode>4_hw, *nfma<mode>4_hw,
8203         *nfms<mode>4_hw, mul<mode>3_odd, fma<mode>4_odd, *fms<mode>4_odd,
8204         *nfma<mode>4_odd, *nfms<mode>4_odd): Change type to qmul.
8206 2017-11-10  Martin Sebor  <msebor@redhat.com>
8208         PR c/81117
8209         * builtins.c (compute_objsize): Handle arrays that
8210         compute_builtin_object_size likes to fail for.  Make extern.
8211         * builtins.h (compute_objsize): Declare.
8212         (check_strncpy_sizes): New function.
8213         (expand_builtin_strncpy): Call check_strncpy_sizes.
8214         * gimple-fold.c (gimple_fold_builtin_strncpy): Implement
8215         -Wstringop-truncation.
8216         (gimple_fold_builtin_strncat): Same.
8217         * gimple.c (gimple_build_call_from_tree): Set call location.
8218         * tree-ssa-strlen.c (strlen_to_stridx): New global variable.
8219         (maybe_diag_bound_equal_length, is_strlen_related_p): New functions.
8220         (handle_builtin_stxncpy, handle_builtin_strncat): Same.
8221         (handle_builtin_strlen): Use strlen_to_stridx.
8222         (strlen_optimize_stmt): Handle flavors of strncat, strncpy, and
8223         stpncpy.
8224         Use strlen_to_stridx.
8225         (pass_strlen::execute): Release strlen_to_stridx.
8226         * doc/invoke.texi (-Wsizeof-pointer-memaccess): Document enhancement.
8227         (-Wstringop-truncation): Document new option.
8229 2017-11-10  Martin Liska  <mliska@suse.cz>
8231         PR gcov-profile/82702
8232         * gcov.c (main): Handle intermediate files in a different
8233         way.
8234         (get_gcov_intermediate_filename): New function.
8235         (output_gcov_file): Remove support of intermediate files.
8236         (generate_results): Allocate intermediate file.
8237         (release_structures): Clean-up properly fn_end.
8238         (output_intermediate_file): Start iterating with line 1.
8240 2017-11-10  Jakub Jelinek  <jakub@redhat.com>
8242         PR tree-optimization/82929
8243         * gimple-ssa-store-merging.c (struct store_immediate_info): Add
8244         ops_swapped_p non-static data member.
8245         (store_immediate_info::store_immediate_info): Clear it.
8246         (imm_store_chain_info::coalesce_immediate_stores): If swapping
8247         ops set ops_swapped_p.
8248         (count_multiple_uses): Handle ops_swapped_p.
8250 2017-11-10  Martin Liska  <mliska@suse.cz>
8252         * coverage.c (coverage_init): Stream information about
8253         support of has_unexecuted_blocks.
8254         * doc/gcov.texi: Document that.
8255         * gcov-dump.c (dump_gcov_file): Support it in gcov_dump tool.
8256         * gcov.c (read_graph_file): Likewise.
8257         (output_line_beginning): Fix a small issue with
8258         color output.
8260 2017-11-10  Bin Cheng  <bin.cheng@arm.com>
8262         * tree-predcom.c (determine_roots_comp): Avoid memory leak by freeing
8263         reference of trivial component.
8265 2017-11-10  Jakub Jelinek  <jakub@redhat.com>
8267         PR bootstrap/82916
8268         * gimple-ssa-store-merging.c
8269         (pass_store_merging::terminate_all_aliasing_chains): For
8270         gimple_store_p stmts also call refs_output_dependent_p.
8272         PR rtl-optimization/82913
8273         * compare-elim.c (try_merge_compare): Punt if def_insn is not
8274         single set.
8276 2017-11-09  Jeff Law  <law@redhat.com>
8278         * vr-values.h: New file with vr_values class.
8279         * tree-vrp.c: Include vr-values.h
8280         (vrp_value_range_pool, vrp_equiv_obstack, num_vr_values): Move static
8281         data objects into the vr_values class.
8282         (vr_value, values_propagated, vr_phi_edge_counts): Likewise.
8283         (get_value_range): Make it a member function within vr_values class.
8284         (set_defs_to_varying, update_value_range, add_equivalence): Likewise.
8285         (vrp_stmt_computes_nonzero_p, op_with_boolean_value_range_p): Likewise.
8286         (op_with_constant_singleton_value_range): Likewise.
8287         (extract_range_for_var_from_comparison_expr): Likewise.
8288         (extract_range_from_assert, extract_range_from_ssa_name): Likewise.
8289         (extract_range_from_binary_expr): Likewise.
8290         (extract_range_from_unary_expr): Likewise.
8291         (extract_range_from_cond_expr, extrat_range_from_comparison): Likewise.
8292         (check_for_binary_op_overflow, extract_range_basic): Likewise.
8293         (extract_range_from_assignment, adjust_range_with_scev): Likewise.
8294         (dump_all_value_ranges, get_vr_for_comparison): Likewise.
8295         (compare_name_with_value, compare_names): Likewise.
8296         (vrp_evaluate_conditional_warnv_with_ops_using_ranges): Likewise.
8297         (vrp_evaluate_conditional_warnv_with_ops): Likewise.  Remove prototype.
8298         (vrp_evaluate_conditional, vrp_visit_cond_stmt): Likewise.
8299         (vrp_visit_switch_stmt, extract_range_from_stmt): Likewise.
8300         (extract_range_from_phi_node): Likewise.
8301         (simplify_truth_ops_using_ranges): Likewise.
8302         (simplify_div_or_mod_using_ranges): Likewise.
8303         (simplify_min_or_max_using_ranges, simplify_abs_using_ranges): Likewise.
8304         (simplify_bit_ops_using_ranges, simplify_cond_using_ranges_1): Likewise.
8305         (simplify_cond_using_ranges_2, simplify_switch_using_ranges): Likewise.
8306         (simplify_float_conversion_using_ranges): Likewise.
8307         (simplify_internal_call_using_ranges): Likewise.
8308         (two_valued_val_range_p, simplify_stmt_using_ranges): Likewise.
8309         (vrp_visit_assignment_or_call): Likewise.  Smuggle class instance
8310         poitner via x_vr_values for calls into gimple folder.
8311         (vrp_initialize_lattice): Make this the vr_values ctor.
8312         (vrp_free_lattice): Make this the vr_values dtor.
8313         (set_vr_value): New function.
8314         (class vrp_prop): Add vr_values data member.  Add various member
8315         functions as well as member functions that delegate to vr_values.
8316         (check_array_ref): Make a member function within vrp_prop class.
8317         (search_for_addr_array, vrp_initialize): Likewise.
8318         (vrp_finalize): Likewise.  Revamp to avoid direct access to
8319         vr_value, values_propagated, etc.
8320         (check_array_bounds): Extract vrp_prop class instance pointer from
8321         walk info structure.  Use it to call member functions.
8322         (check_all_array_refs): Make a member function within vrp_prop class.
8323         Smuggle class instance pointer via walk info structure.
8324         (x_vr_values): New local static.
8325         (vrp_valueize): Use x_vr_values to get class instance.
8326         (vr_valueize_1): Likewise.
8327         (class vrp_folder): Add vr_values data member.  Add various member
8328         functions as well as member functions that delegate to vr_values.
8329         (fold_predicate_in): Make a mber fucntion within vrp_folder class.
8330         (simplify_stmt_for_jump_threading): Extract smuggled vr_values
8331         class instance from vr_values.  Use it to call member functions.
8332         (vrp_dom_walker): Add vr_values data member.
8333         (vrp_dom_walker::after_dom_children): Smuggle vr_values class
8334         instance via x_vr_values.
8335         (identify_jump_threads): Accept vr_values as argument.  Store
8336         it into the walker structure.
8337         (evrp_dom_walker): Add vr_values class data member.  Add various
8338         delegators.
8339         (evrp_dom_walker::try_find_new_range): Use vr_values data
8340         member to access the memory allocator.
8341         (evrp_dom_walker::before_dom_children): Store vr_values class
8342         instance into the vrp_folder class.
8343         (evrp_dom_walker::push_value_range): Rework to avoid direct
8344         access to num_vr_values and vr_value.
8345         (evrp_dom_walker::pop_value_range): Likewise.
8346         (execute_early_vrp): Remove call to vrp_initialize_lattice.
8347         Use vr_values to get to dump_all_value_ranges member function.
8348         Remove call to vrp_free_lattice.  Call vrp_initialize, vrp_finalize,
8349         and simplify_cond_using_ranges_2 via vrp_prop class instance.
8350         Pass vr_values class instance down to identify_jump_threads.
8351         Remove call to vrp_free_lattice.
8352         (debug_all_value_ranges): Remove.
8354         * tree-vrp.c (vrp_prop): Move class to earlier point in the file.
8355         (vrp_folder): Likewise.
8357         * tree-vrp.c (set_value_range): Do not reference vrp_equiv_obstack.
8358         Get it from the existing bitmap instead.
8359         (vrp_intersect_ranges_1): Likewise.
8361 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8363         * gimple-ssa-store-merging.c (struct store_immediate_info): Add
8364         bit_not_p field.
8365         (store_immediate_info::store_immediate_info): Add bitnotp argument,
8366         set bit_not_p to it.
8367         (imm_store_chain_info::coalesce_immediate_stores): Break group
8368         if bit_not_p is different.
8369         (count_multiple_uses, split_group,
8370         imm_store_chain_info::output_merged_store): Handle info->bit_not_p.
8371         (handled_load): Avoid multiple chained BIT_NOT_EXPRs.
8372         (pass_store_merging::process_store): Handle BIT_{AND,IOR,XOR}_EXPR
8373         result inverted using BIT_NOT_EXPR, compute bit_not_p, pass it
8374         to store_immediate_info ctor.
8376 2017-11-09  Jim Wilson  <jimw@sifive.com>
8378         * collect2.c (OBJECT_FORMAT_COFF): Remove EXTENDED_COFF support.
8379         (scan_prog_file): Likewise.
8381 2017-11-09  Jan Hubicka  <hubicka@ucw.cz>
8383         * bb-reorder.c (max_entry_frequency): Remove.
8384         (find_traces, rotate_loop, mark_bb_visited, connect_better_edge_p,
8385         connect_traces, push_to_next_round_p): Remove prototypes.
8386         (find_traces_1_round): Use counts only.
8387         (push_to_next_round_p): Likewise.
8388         (find_traces): Likewise.
8389         (rotate_loop): Likewise.
8390         (find_traces_1_round): Likewise.
8391         (connect_traces): Likewise.
8392         (edge_order): Likewise.
8394 2017-11-09  Thomas Preud'homme  <thomas.preudhomme@arm.com>
8396         * config/arm/arm.c (output_return_instruction): Add comments to
8397         indicate requirement for cmse_nonsecure_entry return to account
8398         for the size of clearing instruction output here.
8399         (thumb_exit): Likewise.
8400         * config/arm/thumb2.md (thumb2_cmse_entry_return): Fix length for
8401         return in hardfloat mode.
8403 2017-11-09  Segher Boessenkool  <segher@kernel.crashing.org>
8405         * config/rs6000/rs6000.c (machine_function): Add a bool,
8406         "toc_is_wrapped_separately".
8407         (rs6000_option_override_internal): Enable OPTION_MASK_SAVE_TOC_INDIRECT
8408         if it wasn't explicitly set or unset, we are optimizing for speed, and
8409         doing separate shrink-wrapping.
8410         (rs6000_get_separate_components): Enable the TOC component if
8411         saving the TOC register in the prologue.
8412         (rs6000_components_for_bb): Handle the TOC component.
8413         (rs6000_emit_prologue_components): Store the TOC register where needed.
8414         (rs6000_set_handled_components): Mark TOC as handled, if handled.
8415         (rs6000_emit_prologue): Don't save the TOC if that is already done.
8417 2017-11-09  Martin Jambor  <mjambor@suse.cz>
8419         * ipa-param-manipulation.c: New file.
8420         * ipa-param-manipulation.h: Likewise.
8421         * Makefile.in (OBJS): Add ipa-param-manipulation.o.
8422         (PLUGIN_HEADERS): Addded ipa-param-manipulation.h
8423         * ipa-param.h (ipa_parm_op): Moved to ipa-param-manipulation.h.
8424         (ipa_parm_adjustment): Likewise.
8425         (ipa_parm_adjustment_vec): Likewise.
8426         (ipa_get_vector_of_formal_parms): Moved declaration to
8427         ipa-param-manipulation.h.
8428         (ipa_get_vector_of_formal_parm_types): Likewise.
8429         (ipa_modify_formal_parameters): Likewise.
8430         (ipa_modify_call_arguments): Likewise.
8431         (ipa_combine_adjustments): Likewise.
8432         (ipa_dump_param_adjustments): Likewise.
8433         (ipa_modify_expr): Likewise.
8434         (ipa_get_adjustment_candidate): Likewise.
8435         * ipa-prop.c (ipa_get_vector_of_formal_parms): Moved to
8436         ipa-param-manipulation.c.
8437         (ipa_get_vector_of_formal_parm_types): Likewise.
8438         (ipa_modify_formal_parameters): Likewise.
8439         (ipa_modify_call_arguments): Likewise.
8440         (ipa_modify_expr): Likewise.
8441         (get_ssa_base_param): Likewise.
8442         (ipa_get_adjustment_candidate): Likewise.
8443         (index_in_adjustments_multiple_times_p): Likewise.
8444         (ipa_combine_adjustments): Likewise.
8445         (ipa_dump_param_adjustments): Likewise.
8446         * tree-sra.c: Also include ipa-param-manipulation.h
8447         * omp-simd-clone.c: Include ipa-param-manipulation.h instead of
8448         ipa-param.h.
8450 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8451             Alan Hayward  <alan.hayward@arm.com>
8452             David Sherwood  <david.sherwood@arm.com>
8454         * doc/sourcebuild.texi (vect_masked_store): Document.
8456 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8457             Alan Hayward  <alan.hayward@arm.com>
8458             David Sherwood  <david.sherwood@arm.com>
8460         * doc/sourcebuild.texi (vect_align_stack_vars): Document.
8462 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8463             Alan Hayward  <alan.hayward@arm.com>
8464             David Sherwood  <david.sherwood@arm.com>
8466         * doc/sourcebuild.texi (vect_variable_length): Document.
8468 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8469             Alan Hayward  <alan.hayward@arm.com>
8470             David Sherwood  <david.sherwood@arm.com>
8472         * doc/sourcebuild.texi (vect_unaligned_possible): Document.
8474 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8475             Alan Hayward  <alan.hayward@arm.com>
8476             David Sherwood  <david.sherwood@arm.com>
8478         * doc/sourcebuild.texi (vect_element_align_preferred): Document.
8480 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8481             Alan Hayward  <alan.hayward@arm.com>
8482             David Sherwood  <david.sherwood@arm.com>
8484         * doc/sourcebuild.texi (vect_perm_short, vect_perm_byte): Document
8485         previously undocumented selectors.
8486         (vect_perm3_byte, vect_perm3_short, vect_perm3_int): Document.
8488 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8490         * doc/rtl.texi (const_vector): Say that elements can be
8491         const_wide_ints too.
8492         * emit-rtl.h (valid_for_const_vec_duplicate_p): Declare.
8493         * emit-rtl.c (valid_for_const_vec_duplicate_p): New function.
8494         (gen_vec_duplicate): Use it instead of CONSTANT_P.
8495         * optabs.c (expand_vector_broadcast): Likewise.
8497 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8498             Alan Hayward  <alan.hayward@arm.com>
8499             David Sherwood  <david.sherwood@arm.com>
8501         * tree-ssa-loop-ivopts.c (get_address_cost): Try using a
8502         scaled index even if the unscaled address was invalid.
8503         Don't increase the complexity of using a scale in that case.
8505 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8506             Alan Hayward  <alan.hayward@arm.com>
8507             David Sherwood  <david.sherwood@arm.com>
8509         * doc/rtl.texi: Rewrite the subreg rules so that they partition
8510         the inner register into REGMODE_NATURAL_SIZE bytes rather than
8511         UNITS_PER_WORD bytes.
8512         * emit-rtl.c (validate_subreg): Divide subregs into blocks
8513         based on REGMODE_NATURAL_SIZE of the inner mode.
8514         (gen_lowpart_common): Split the SCALAR_FLOAT_MODE_P and
8515         !SCALAR_FLOAT_MODE_P cases.  Use REGMODE_NATURAL_SIZE for the latter.
8516         * expmed.c (lowpart_bit_field_p): Divide the value up into
8517         chunks of REGMODE_NATURAL_SIZE rather than UNITS_PER_WORD.
8518         * expr.c (store_constructor): Use REGMODE_NATURAL_SIZE to test
8519         whether something is likely to occupy more than one register.
8521 2017-11-09  Jan Hubicka  <hubicka@ucw.cz>
8523         PR ipa/82879
8524         * ipa-inline-transform.c (update_noncloned_frequencies): Use
8525         profile_count::adjust_for_ipa_scaling.
8526         * tree-inline.c (copy_bb, copy_cfg_body): Likewise.
8527         * profile-count.c (profile_count::adjust_for_ipa_scaling): New member
8528         function.
8529         * profile-count.h (profile_count::adjust_for_ipa_scaling): Declare.
8531 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8533         * gimple-ssa-store-merging.c (count_multiple_uses): New function.
8534         (split_group): Add total_orig and total_new arguments, estimate the
8535         number of statements related to the store group without store merging
8536         and with store merging.
8537         (imm_store_chain_info::output_merged_store): Adjust split_group
8538         callers, punt if estimated number of statements with store merging
8539         is not smaller than estimated number of statements without it.
8540         Formatting fix.
8541         (handled_load): Remove has_single_use checks.
8542         (pass_store_merging::process_store): Likewise.
8544 2017-11-09  Richard Biener  <rguenther@suse.de>
8546         PR tree-optimization/82902
8547         * tree-ssa-phiprop.c (propagate_with_phi): Test proper type.
8549 2017-11-09  Martin Liska  <mliska@suse.cz>
8551         PR target/82863
8552         * emit-rtl.c (init_emit_regs): Initialize split_branch_probability to
8553         uninitialized.
8555 2017-11-09  Martin Liska  <mliska@suse.cz>
8557         PR tree-optimization/82669
8558         * sbitmap.h (bmp_iter_set_init): Remove non needed check.
8560 2017-11-09  Martin Liska  <mliska@suse.cz>
8562         PR gcov-profile/48463
8563         * coverage.c (coverage_begin_function): Output also end locus
8564         of a function and information whether the function is
8565         artificial.
8566         * gcov-dump.c (tag_function): Parse and print the information.
8567         * gcov.c (INCLUDE_MAP): Add include.
8568         (INCLUDE_SET): Likewise.
8569         (struct line_info): Move earlier in the source file because
8570         of vector<line_info> in function_info structure.
8571         (line_info::line_info): Likewise.
8572         (line_info::has_block): Likewise.
8573         (struct source_info): Add new member index.
8574         (source_info::get_functions_at_location): New function.
8575         (function_info::group_line_p): New function.
8576         (output_intermediate_line): New function.
8577         (output_intermediate_file): Use the mentioned function.
8578         (struct function_start): New.
8579         (struct function_start_pair_hash): Likewise.
8580         (process_file): Add code that identifies group functions.
8581         Assign lines either to global or function scope.
8582         (generate_results): Skip artificial functions.
8583         (find_source): Assign index for each source file.
8584         (read_graph_file): Read new flag artificial and end_line.
8585         (add_line_counts): Assign it either to global of function scope.
8586         (accumulate_line_counts): Isolate core of the function to
8587         accumulate_line_info and call it for both function and global
8588         scope lines.
8589         (accumulate_line_info): New function.
8590         (output_line_beginning): Fix GNU coding style.
8591         (print_source_line): New function.
8592         (output_line_details): Likewise.
8593         (output_function_details): Likewise.
8594         (output_lines): Iterate both source (global) scope and function
8595         scope.
8596         (struct function_line_start_cmp): New class.
8597         * doc/gcov.texi: Reflect changes in documentation.
8599 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8601         PR debug/82837
8602         * dwarf2out.c (const_ok_for_output_1): Reject NEG in addition to NOT.
8603         (mem_loc_descriptor): Handle (const (neg (...))) as (neg (const (...)))
8604         and similarly for not instead of neg.
8606 2017-11-08  Andi Kleen  <ak@linux.intel.com>
8608         * config/i386/i386.opt: Add -mforce-indirect-call.
8609         * config/i386/predicates.md: Check for flag_force_indirect_call.
8610         * doc/invoke.texi: Document -mforce-indirect-call
8612 2017-11-08  Kito Cheng  <kito.cheng@gmail.com>
8614         * config/riscv/riscv-protos.h (riscv_slow_unaligned_access_p):
8615         New extern.
8616         (MOVE_RATIO): Use riscv_slow_unaligned_access_p.
8617         config/riscv/riscv.c (predict.h): New include.
8618         (riscv_slow_unaligned_access_p): No longer static.
8619         (riscv_block_move_straight): Add require.
8620         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): Delete.
8622 2017-11-08  Jakub Jelinek  <jakub@redhat.com>
8624         PR target/82855
8625         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>,
8626         <avx512>_eq<mode>3<mask_scalar_merge_name>_1): Use
8627         nonimmediate_operand predicate for operand 1 instead of
8628         register_operand.
8630 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8632         * config/aarch64/aarch64-simd.md (store_pair_lanes<mode>):
8633         New pattern.
8634         * config/aarch64/constraints.md (Uml): New constraint.
8635         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): New
8636         predicate.
8638 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8640         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
8641         of two vec_duplicates into a vec_concat.
8643 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8645         * simplify-rtx.c (simplify_ternary_operation, VEC_MERGE):
8646         Simplify vec_merge of vec_duplicate and vec_concat.
8647         * config/aarch64/constraints.md (Utq): New constraint.
8648         * config/aarch64/aarch64-simd.md (load_pair_lanes<mode>): New
8649         define_insn.
8651 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8653         * simplify-rtx.c (simplify_ternary_operation, VEC_MERGE):
8654         Simplify vec_merge of vec_duplicate and const_vector.
8655         * config/aarch64/predicates.md (aarch64_simd_or_scalar_imm_zero):
8656         New predicate.
8657         * config/aarch64/aarch64-simd.md (*aarch64_combinez<mode>): Use VDC
8658         mode iterator.  Update predicate on operand 1 to
8659         handle non-const_vec constants.  Delete constraints.
8660         (*aarch64_combinez_be<mode>): Likewise for operand 2.
8662 2017-11-08  Jakub Jelinek  <jakub@redhat.com>
8664         PR tree-optimization/78821
8665         * gimple-ssa-store-merging.c (struct store_operand_info): Add bit_not_p
8666         data member.
8667         (store_operand_info::store_operand_info): Initialize it to false.
8668         (pass_store_merging::terminate_all_aliasing_chains): Rewritten to use
8669         ref_maybe_used_by_stmt_p and stmt_may_clobber_ref_p on lhs of each
8670         store in the group, and if chain_info is non-NULL, to ignore altogether
8671         that chain.
8672         (compatible_load_p): Fail if bit_not_p does not match.
8673         (imm_store_chain_info::output_merged_store): Handle bit_not_p loads.
8674         (handled_load): Fill in bit_not_p.  Handle BIT_NOT_EXPR.
8675         (pass_store_merging::process_store): Adjust
8676         terminate_all_aliasing_chains calls to pass NULL in all current spots,
8677         call terminate_all_aliasing_chains newly when adding a store into
8678         a chain with non-NULL chain_info.
8680 2017-11-08  Wilco Dijkstra  <wdijkstr@arm.com>
8682         * config/aarch64/aarch64.c (aarch64_can_eliminate): Simplify logic.
8684 2017-11-08  Wilco Dijkstra  <wdijkstr@arm.com>
8686         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
8687         Remove.
8688         (aarch64_layout_frame): Initialise emit_frame_chain.
8689         (aarch64_can_eliminate): Remove omit leaf frame pointer code.
8690         (TARGET_FRAME_POINTER_REQUIRED): Remove define.
8692 2017-11-08  Martin Liska  <mliska@suse.cz>
8694         * gimplify.c (expand_FALLTHROUGH_r): Simplify usage
8695         of gimple_call_internal_p.
8697 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8699         * config/mips/mips.h (ASM_OUTPUT_LABELREF): Wrap in "do {} while (0)".
8701 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8703         * config/mips/mips.h (ASM_OUTPUT_CASE_END): Remove semicolon after
8704         "do {} while (0)".
8706 2017-11-08  Martin Liska  <mliska@suse.cz>
8708         PR sanitizer/82792
8709         * gimplify.c (expand_FALLTHROUGH_r): Skip IFN_ASAN_MARK.
8711 2017-11-07  Eric Botcazou  <ebotcazou@adacore.com>
8713         * gimple-pretty-print.c (dump_profile): Return "" instead of NULL.
8715 2017-11-07  Jakub Jelinek  <jakub@redhat.com>
8717         PR target/82855
8718         * config/i386/i386.md (SWI1248_AVX512BWDQ2_64): New mode iterator.
8719         (*cmp<mode>_ccz_1): New insn with $k alternative.
8721         PR target/82855
8722         * config/i386/i386.c (ix86_swap_binary_operands_p): Treat
8723         RTX_COMM_COMPARE as commutative as well.
8724         (ix86_binary_operator_ok): Formatting fix.
8725         * config/i386/sse.md (*mul<mode>3<mask_name><round_name>,
8726         *<code><mode>3<mask_name><round_saeonly_name>,
8727         *<code><mode>3<mask_name>, *<code>tf3, *mul<mode>3<mask_name>,
8728         *<s>mul<mode>3_highpart<mask_name>,
8729         *vec_widen_umult_even_v16si<mask_name>,
8730         *vec_widen_umult_even_v8si<mask_name>,
8731         *vec_widen_umult_even_v4si<mask_name>,
8732         *vec_widen_smult_even_v16si<mask_name>,
8733         *vec_widen_smult_even_v8si<mask_name>, *sse4_1_mulv2siv2di3<mask_name>,
8734         *avx2_pmaddwd, *sse2_pmaddwd, *<sse4_1_avx2>_mul<mode>3<mask_name>,
8735         *avx2_<code><mode>3, *avx512f_<code><mode>3<mask_name>,
8736         *sse4_1_<code><mode>3<mask_name>, *<code>v8hi3,
8737         *sse4_1_<code><mode>3<mask_name>, *<code>v16qi3, *avx2_eq<mode>3,
8738         <avx512>_eq<mode>3<mask_scalar_merge_name>_1, *sse4_1_eqv2di3,
8739         *sse2_eq<mode>3, <mask_codefor><code><mode>3<mask_name>,
8740         *<code><mode>3, *<sse2_avx2>_uavg<mode>3<mask_name>,
8741         *<ssse3_avx2>_pmulhrsw<mode>3<mask_name>, *ssse3_pmulhrswv4hi3): Use
8742         !(MEM_P (operands[1]) && MEM_P (operands[2])) condition instead of
8743         ix86_binary_operator_ok.  Formatting fixes.
8744         (*<plusminus_insn><mode>3<mask_name><round_name>,
8745         *<plusminus_insn><mode>3, *<plusminus_insn><mode>3_m): Formatting
8746         fixes.
8748 2017-11-07  Segher Boessenkool  <segher@kernel.crashing.org>
8750         * config/rs6000/rs6000.md (GPR2): New mode_iterator.
8751         ("cstore<mode>4"): Don't always expand with rs6000_emit_int_cmove for
8752         eq and ne if TARGET_ISEL.
8753         (cmp): New code_iterator.
8754         (UNS, UNSU_, UNSIK): New code_attrs.
8755         (<code><GPR:mode><GPR2:mode>2_isel): New define_insn_and_split.
8756         ("eq<mode>3"): New define_expand, rename the define_insn_and_split
8757         to...
8758         ("eq<mode>3"): ... this.
8759         ("ne<mode>3"): New define_expand, rename the define_insn_and_split
8760         to...
8761         ("ne<mode>3"): ... this.
8763 2017-11-07  Julia Koval  <julia.koval@intel.com>
8765         PR target/82812
8766         * common/config/i386/i386-common.c
8767         (OPTION_MASK_ISA_GENERAL_REGS_ONLY_UNSET): Remove MPX from flag.
8768         (ix86_handle_option): Move MPX to isa_flags2 and GFNI to isa_flags.
8769         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto.
8770         * config/i386/i386.opt: Ditto.
8771         * config/i386/i386.c (ix86_target_string): Ditto.
8772         (ix86_option_override_internal): Ditto.
8773         (ix86_init_mpx_builtins): Move MPX to args2.
8774         (ix86_expand_builtin): Special handling for OPTION_MASK_ISA_GFNI.
8775         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8affineinvqb_v64qi,
8776         __builtin_ia32_vgf2p8affineinvqb_v64qi_mask,
8777         __builtin_ia32_vgf2p8affineinvqb_v32qi,
8778         __builtin_ia32_vgf2p8affineinvqb_v32qi_mask,
8779         __builtin_ia32_vgf2p8affineinvqb_v16qi,
8780         __builtin_ia32_vgf2p8affineinvqb_v16qi_mask): Move to ARGS array.
8782 2017-11-07  Uros Bizjak  <ubizjak@gmail.com>
8784         PR target/80425
8785         * config/i386.i386.md (*zero_extendsidi2): Change (?r,*Yj), (?*Yi,r)
8786         and (*x,m) to ($r,Yj), ($Yi,r) and ($x,m).
8787         (zero-extendsidi peephole2): Remove peephole.
8789 2017-11-07  Eric Botcazou  <ebotcazou@adacore.com>
8791         PR c/53037
8792         * stor-layout.c: Include attribs.h.
8793         (handle_warn_if_not_align): Replace test on TYPE_USER_ALIGN with
8794         explicit lookup of "aligned" attribute.
8796 2017-11-07  Andrew Waterman  <andrew@sifive.com>
8798         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): New prototype.
8799         (riscv_expand_block_move): Likewise.
8800         * config/riscv/riscv.h (MOVE_RATIO): Tune cost to movmemsi
8801         implementation.
8802         (RISCV_MAX_MOVE_BYTES_PER_LOOP_ITER): New define.
8803         (RISCV_MAX_MOVE_BYTES_STRAIGHT): New define.
8804         * config/riscv/riscv.c (riscv_block_move_straight): New function.
8805         (riscv_adjust_block_mem): Likewise.
8806         (riscv_block_move_loop): Likewise.
8807         (riscv_expand_block_move): Likewise.
8808         * config/riscv/riscv.md (movmemsi): New pattern.
8810 2017-11-07  Michael Clark  <michaeljclark@mac.com>
8812         * config/riscv/linux.h (MUSL_ABI_SUFFIX): New define.
8813         (MUSL_DYNAMIC_LINKER): Likewise.
8815 2017-11-07  Richard Sandiford  <richard.sandiford@linaro.org>
8817         * config/aarch64/aarch64.md (ashl<mode>3, ashr<mode>3, lshr<mode>3)
8818         (rotr<mode>3, rotl<mode>3): Use aarch64_reg_or_imm instead of
8819         nonmmory_operand.
8821 2017-11-07  Richard Biener  <rguenther@suse.de>
8823         * match.pd: Fix build.
8825 2017-11-07  Wilco Dijkstra  <wdijkstr@arm.com>
8826             Jackson Woodruff  <jackson.woodruff@arm.com>
8828         PR tree-optimization/71026
8829         * match.pd: Canonicalize negate in division.
8831 2017-11-07  Sudakshina Das  <sudi.das@arm.com>
8833         PR middle-end/80131
8834         * match.pd: Simplify 1 << (C - x) where C = precision (x) - 1.
8836 2017-11-07  Marc Glisse  <marc.glisse@inria.fr>
8838         * match.pd ((a&~b)|(a^b),(a&~b)^~a,(a|b)&~(a^b),a|~(a^b),
8839         (a|b)|(a&^b),(a&b)|~(a^b),~(~a&b),~X^Y): New transformations.
8841 2017-11-07  Marc Glisse  <marc.glisse@inria.fr>
8843         * fold-const.c (negate_expr_p) [PLUS_EXPR, MINUS_EXPR]: Handle
8844         non-scalar integral types.
8845         * match.pd (negate_expr_p): Handle MINUS_EXPR.
8846         (-(A-B), -(~A)): New transformations.
8848 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8850         * config/powerpcspe/aix43.h (SUBTARGET_OVERRIDE_OPTIONS): Remove
8851         semicolon after "do {} while (0)".
8852         * config/powerpcspe/aix51.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8853         * config/powerpcspe/aix52.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8854         * config/powerpcspe/aix53.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8855         * config/powerpcspe/aix61.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8856         * config/powerpcspe/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8858 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8860         * config/rs6000/aix43.h (SUBTARGET_OVERRIDE_OPTIONS): Remove semicolon
8861         after "do {} while (0)".
8862         * config/rs6000/aix51.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8863         * config/rs6000/aix52.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8864         * config/rs6000/aix53.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8865         * config/rs6000/aix61.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8866         * config/rs6000/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8868 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8870         PR other/82784
8871         * config/arm/arm.c (HANDLE_NARROW_SHIFT_ARITH): Remove semicolon after
8872         "while {} do (0)".
8873         (arm_rtx_costs_internal): Add missing semicolon after
8874         HANDLE_NARROW_SHIFT_ARITH call.
8876 2017-11-06  Segher Boessenkool  <segher@kernel.crashing.org>
8878         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
8879         disable isel if it was not set explicitly.
8881 2017-11-06  James Bowman  <james.bowman@ftdichip.com>
8883         * gcc/dwarf2out.c (modified_type_die): Retain ADDR_SPACE qualifiers.
8884         (add_type_attribute) likewise.
8886 2017-11-06  H.J. Lu  <hongjiu.lu@intel.com>
8888         * config/i386/i386.c (ix86_can_use_return_insn_p): Use reference
8889         of struct ix86_frame.
8890         (ix86_initial_elimination_offset): Likewise.
8891         (ix86_expand_split_stack_prologue): Likewise.
8893 2017-11-06  Marc Glisse  <marc.glisse@inria.fr>
8895         * tree-vrp.h (enum value_range_type): Update stale comment.
8897 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8898             Alan Hayward  <alan.hayward@arm.com>
8899             David Sherwood  <david.sherwood@arm.com>
8901         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm)
8902         (aarch64_expand_vec_perm_const): Take the number of units too.
8903         * config/aarch64/aarch64.c (aarch64_expand_vec_perm)
8904         (aarch64_expand_vec_perm_const): Likewise.
8905         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>)
8906         (vec_perm<mode>): Update accordingly.
8908 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8909             Alan Hayward  <alan.hayward@arm.com>
8910             David Sherwood  <david.sherwood@arm.com>
8912         * config/aarch64/aarch64-protos.h (aarch64_simd_vect_par_cnst_half):
8913         Take the number of units too.
8914         * config/aarch64/aarch64.c (aarch64_simd_vect_par_cnst_half): Likewise.
8915         (aarch64_simd_check_vect_par_cnst_half): Update call accordingly,
8916         but check for a vector mode before rather than after the call.
8917         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
8918         (move_hi_quad_<mode>, vec_unpack<su>_hi_<mode>)
8919         (vec_unpack<su>_lo_<mode, vec_widen_<su>mult_lo_<mode>)
8920         (vec_widen_<su>mult_hi_<mode>, vec_unpacks_lo_<mode>)
8921         (vec_unpacks_hi_<mode>, aarch64_saddl2<mode>, aarch64_uaddl2<mode>)
8922         (aarch64_ssubl2<mode>, aarch64_usubl2<mode>, widen_ssum<mode>3)
8923         (widen_usum<mode>3, aarch64_saddw2<mode>, aarch64_uaddw2<mode>)
8924         (aarch64_ssubw2<mode>, aarch64_usubw2<mode>, aarch64_sqdmlal2<mode>)
8925         (aarch64_sqdmlsl2<mode>, aarch64_sqdmlal2_lane<mode>)
8926         (aarch64_sqdmlal2_laneq<mode>, aarch64_sqdmlsl2_lane<mode>)
8927         (aarch64_sqdmlsl2_laneq<mode>, aarch64_sqdmlal2_n<mode>)
8928         (aarch64_sqdmlsl2_n<mode>, aarch64_sqdmull2<mode>)
8929         (aarch64_sqdmull2_lane<mode>, aarch64_sqdmull2_laneq<mode>)
8930         (aarch64_sqdmull2_n<mode>): Update accordingly.
8932 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8933             Alan Hayward  <alan.hayward@arm.com>
8934             David Sherwood  <david.sherwood@arm.com>
8936         * config/aarch64/aarch64-protos.h (aarch64_reverse_mask): Take
8937         the number of units too.
8938         * config/aarch64/aarch64.c (aarch64_reverse_mask): Likewise.
8939         * config/aarch64/aarch64-simd.md (vec_load_lanesoi<mode>)
8940         (vec_store_lanesoi<mode>, vec_load_lanesci<mode>)
8941         (vec_store_lanesci<mode>, vec_load_lanesxi<mode>)
8942         (vec_store_lanesxi<mode>): Update accordingly.
8944 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8945             Alan Hayward  <alan.hayward@arm.com>
8946             David Sherwood  <david.sherwood@arm.com>
8948         * config/aarch64/aarch64-protos.h (aarch64_endian_lane_rtx): Declare.
8949         * config/aarch64/aarch64.c (aarch64_endian_lane_rtx): New function.
8950         * config/aarch64/aarch64.h (ENDIAN_LANE_N): Take the number
8951         of units rather than the mode.
8952         * config/aarch64/iterators.md (nunits): New mode attribute.
8953         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args):
8954         Use aarch64_endian_lane_rtx instead of GEN_INT (ENDIAN_LANE_N ...).
8955         * config/aarch64/aarch64-simd.md (aarch64_dup_lane<mode>)
8956         (aarch64_dup_lane_<vswap_width_name><mode>, *aarch64_mul3_elt<mode>)
8957         (*aarch64_mul3_elt_<vswap_width_name><mode>): Likewise.
8958         (*aarch64_mul3_elt_to_64v2df, *aarch64_mla_elt<mode>): Likewise.
8959         (*aarch64_mla_elt_<vswap_width_name><mode>, *aarch64_mls_elt<mode>)
8960         (*aarch64_mls_elt_<vswap_width_name><mode>, *aarch64_fma4_elt<mode>)
8961         (*aarch64_fma4_elt_<vswap_width_name><mode>):: Likewise.
8962         (*aarch64_fma4_elt_to_64v2df, *aarch64_fnma4_elt<mode>): Likewise.
8963         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
8964         (*aarch64_fnma4_elt_to_64v2df, reduc_plus_scal_<mode>): Likewise.
8965         (reduc_plus_scal_v4sf, reduc_<maxmin_uns>_scal_<mode>): Likewise.
8966         (reduc_<maxmin_uns>_scal_<mode>): Likewise.
8967         (*aarch64_get_lane_extend<GPI:mode><VDQQH:mode>): Likewise.
8968         (*aarch64_get_lane_zero_extendsi<mode>): Likewise.
8969         (aarch64_get_lane<mode>, *aarch64_mulx_elt_<vswap_width_name><mode>)
8970         (*aarch64_mulx_elt<mode>, *aarch64_vgetfmulx<mode>): Likewise.
8971         (aarch64_sq<r>dmulh_lane<mode>, aarch64_sq<r>dmulh_laneq<mode>)
8972         (aarch64_sqrdml<SQRDMLH_AS:rdma_as>h_lane<mode>): Likewise.
8973         (aarch64_sqrdml<SQRDMLH_AS:rdma_as>h_laneq<mode>): Likewise.
8974         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>): Likewise.
8975         (aarch64_sqdml<SBINQOPS:as>l_laneq<mode>): Likewise.
8976         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
8977         (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>_internal): Likewise.
8978         (aarch64_sqdmull_lane<mode>, aarch64_sqdmull_laneq<mode>): Likewise.
8979         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
8980         (aarch64_sqdmull2_laneq<mode>_internal): Likewise.
8981         (aarch64_vec_load_lanesoi_lane<mode>): Likewise.
8982         (aarch64_vec_store_lanesoi_lane<mode>): Likewise.
8983         (aarch64_vec_load_lanesci_lane<mode>): Likewise.
8984         (aarch64_vec_store_lanesci_lane<mode>): Likewise.
8985         (aarch64_vec_load_lanesxi_lane<mode>): Likewise.
8986         (aarch64_vec_store_lanesxi_lane<mode>): Likewise.
8987         (aarch64_simd_vec_set<mode>): Update use of ENDIAN_LANE_N.
8988         (aarch64_simd_vec_setv2di): Likewise.
8990 2017-11-06  Carl Love  <cel@us.ibm.com>
8992         * config/rs6000/rs6000-c.c (P8V_BUILTIN_VEC_REVB): Add power 8
8993         definitions.
8994         (P9V_BUILTIN_VEC_REVB): Remove the power 9 instance definitions.
8995         * config/rs6000/altivec.h (vec_revb): Change the #define from power 9
8996         to power 8.
8997         * config/rs6000/r6000-protos.h (swap_endian_selector_for_mode): Add new
8998         extern declaration.
8999         * config/rs6000/rs6000.c (swap_endian_selector_for_mode): Add function.
9000         * config/rs6000/rs6000-builtin.def (BU_P8V_VSX_1, BU_P8V_OVERLOAD_1):
9001         Add power 8 macro expansions.
9002         (BU_P9V_OVERLOAD_1): Remove power 9 overload expansion.
9003         * config/rs6000/vsx.md (revb_<mode>): Add define_expand to generate
9004         power 8 instructions.  (VSX_XXBR): Add iterator.
9006 2017-11-06  Wilco Dijkstra  <wdijkstr@arm.com>
9008         * config/arm/arm.md (predicable_short_it): Change default to "no",
9009         improve documentation, remove uses that are identical to the default.
9010         (enabled_for_depr_it): Rename to enabled_for_short_it.
9011         * gcc/config/arm/arm-fixed.md (predicable_short_it):
9012         Remove default uses.
9013         * gcc/config/arm/ldmstm.md (predicable_short_it): Likewise.
9014         * gcc/config/arm/sync.md (predicable_short_it): Likewise.
9015         * gcc/config/arm/thumb2.md (predicable_short_it): Likewise.
9016         * gcc/config/arm/vfp.md (predicable_short_it): Likewise.
9018 2017-11-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
9020         PR target/82748
9021         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_1): Delete
9022         float128 helper macros, which are no longer used after deleting
9023         the old 'q' built-in functions, and moving the round to odd
9024         built-in functions to being special built-in functions.
9025         (BU_FLOAT128_2): Likewise.
9026         (BU_FLOAT128_1_HW): Likewise.
9027         (BU_FLOAT128_2_HW): Likewise.
9028         (BU_FLOAT128_3_HW): Likewise.
9029         (FABSQ): Delete old 'q' built-in functions.
9030         (COPYSIGNQ): Likewise.
9031         (SQRTF128_ODD): Move round to odd built-in functions to be
9032         special built-in functions, so that we can handle
9033         -mabi=ieeelongdouble.
9034         (TRUNCF128_ODD): Likewise.
9035         (ADDF128_ODD): Likewise.
9036         (SUBF128_ODD): Likewise.
9037         (MULF128_ODD): Likewise.
9038         (DIVF128_ODD): Likewise.
9039         (FMAF128_ODD): Likewise.
9040         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Map old 'q'
9041         built-in names to 'f128'.
9042         * config/rs6000/rs6000.c (rs6000_fold_builtin): Remove folding the
9043         old 'q' built-in functions, as the machine independent code for
9044         'f128' built-in functions handles this.
9045         (rs6000_expand_builtin): Add expansion for float128 round to odd
9046         functions, keying off on -mabi=ieeelongdouble of whether to use
9047         the KFmode or TFmode variant.
9048         (rs6000_init_builtins): Initialize the _Float128 round to odd
9049         built-in functions.
9050         * doc/extend.texi (PowerPC Built-in Functions): Document the old
9051         _Float128 'q' built-in functions are now mapped into the new
9052         'f128' built-in functions.
9054 2017-11-06  David Edelsohn  <dje.gcc@gmail.com>
9056         * collect2.c (add_lto_object): Compile for OBJECT_COFF.
9057         (scan_prog_file): Don't skip PASS_LTOINFO. Scan for LTO objects.
9059 2017-11-06  David Malcolm  <dmalcolm@redhat.com>
9061         PR jit/82826
9062         * ipa-fnsummary.c (ipa_fnsummary_c_finalize): New function.
9063         * ipa-fnsummary.h (ipa_fnsummary_c_finalize): New decl.
9064         * toplev.c: Include "ipa-fnsummary.h".
9065         (toplev::finalize): Call ipa_fnsummary_c_finalize.
9067 2017-11-06  Jakub Jelinek  <jakub@redhat.com>
9069         PR tree-optimization/82838
9070         * gimple-ssa-store-merging.c
9071         (imm_store_chain_info::output_merged_store): Call force_gimple_operand_1
9072         on a separate gimple_seq which is then appended to seq.
9074 2017-11-06  Jeff Law  <law@redhat.com>
9076         PR target/82788
9077         * config/i386/i386.c (PROBE_INTERVAL): Remove.
9078         (get_probe_interval): New functions.
9079         (ix86_adjust_stack_and_probe_stack_clash): Use get_probe_interval.
9080         (ix86_adjust_stack_and_probe): Likewise.
9081         (output_adjust_stack_and_probe): Likewise.
9082         (ix86_emit_probe_stack_range): Likewise.
9083         (ix86_expand_prologue): Likewise.
9085 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
9087         PR tree-optimization/82816
9088         * tree-ssa-math-opts.c (convert_mult_to_widen): Return false
9089         if the modes of the two types are the same.
9090         (convert_plusminus_to_widen): Likewise.
9092 2017-11-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
9094         * config/rs6000/altivec.md (*p9_vadu<mode>3) Rename to
9095         p9_vadu<mode>3.
9096         (usadv16qi): New define_expand.
9097         (usadv8hi): New define_expand.
9099 2017-11-06  Jan Hubicka  <hubicka@ucw.cz>
9101         PR bootstrap/82832
9102         * ipa-inline-transform.c (update_noncloned_frequencies): Always
9103         scale.
9104         (inline_transform): Likewise.
9105         * predict.c (counts_to_freqs): Remove useless conditional.
9106         * profile-count.h (profile_count::apply_scale): Move sanity check.
9107         * tree-inline.c (copy_bb): Always scale.
9108         (copy_cfg_body): Likewise.
9110 2017-11-06  Christophe Lyon  <christophe.lyon@linaro.org>
9112         PR target/67591
9113         * config/arm/arm.md (*sub_shiftsi): Add predicable_short_it
9114         attribute.
9115         (*cmp_ite0): Add enabled_for_depr_it attribute.
9116         (*cmp_ite1): Likewise.
9118 2017-11-06  Segher Boessenkool  <segher@kernel.crashing.org>
9120         * config/rs6000/rs6000.c (rs6000_insn_cost): Handle TYPE_MFCR and
9121         TYPE_MFCRF.
9123 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
9125         * tree-vrp.c (vrp_int_const_binop): Return true on success and
9126         return the value by pointer.
9127         (extract_range_from_multiplicative_op_1): Update accordingly.
9128         Return as soon as an operation fails.
9130 2017-11-05  Tom de Vries  <tom@codesourcery.com>
9132         PR other/82784
9133         * asan.c (DEF_SANITIZER_BUILTIN_1): Factor out of ...
9134         (DEF_SANITIZER_BUILTIN): ... here.
9135         (initialize_sanitizer_builtins): Use DEF_SANITIZER_BUILTIN_1 instead of
9136         DEF_SANITIZER_BUILTIN in if stmt.  Add missing semicolon.
9138 2017-11-05  Tom de Vries  <tom@codesourcery.com>
9140         PR other/82784
9141         * config/elfos.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove semicolon after
9142         macro body.
9143         (ASM_OUTPUT_CASE_LABEL): Add semicolon after
9144         ASM_OUTPUT_BEFORE_CASE_LABEL call.
9145         * config/arc/arc.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove semicolon
9146         after macro body.
9147         * config/m68k/m68kelf.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
9148         * config/mips/mips.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
9149         * config/v850/v850.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
9151 2017-11-05  Tom de Vries  <tom@codesourcery.com>
9153         PR other/82784
9154         * graphite-scop-detection.c (DEBUG_PRINT): Remove semicolon after
9155         "do {} while (0)".
9157 2017-11-04  Michael Clark  <michaeljclark@mac.com>
9159         * config/riscv/riscv.c (riscv_print_operand): Add a 'i' format.
9160         config/riscv/riscv.md (addsi3): Use 'i' for immediates.
9161         (adddi3): Likewise.
9162         (*addsi3_extended): Likewise.
9163         (*addsi3_extended2): Likewise.
9164         (<optab>si3): Likewise.
9165         (<optab>di3): Likewise.
9166         (<optab><mode>3): Likewise.
9167         (<*optabe>si3_internal): Likewise.
9168         (zero_extendqi<SUPERQI:mode>2): Likewise.
9169         (*add<mode>hi3): Likewise.
9170         (*xor<mode>hi3): Likewise.
9171         (<optab>di3): Likewise.
9172         (*<optab>si3_extend): Likewise.
9173         (*sge<u>_<X:mode><GPR:mode>): Likewise.
9174         (*slt<u>_<X:mode><GPR:mode>): Likewise.
9175         (*sle<u>_<X:mode><GPR:mode>): Likewise.
9177 2017-11-04  Andrew Waterman  <andrew@sifive.com>
9179         * config/riscv/riscv.c (riscv_option_override): Conditionally set
9180         TARGET_STRICT_ALIGN based upon -mtune argument.
9182 2017-11-04  Andrew Waterman  <andrew@sifive.com>
9184         * config/riscv/riscv.h (SLOW_BYTE_ACCESS): Change to 1.
9186 2017-11-04  Daniel Santos  <daniel.santos@pobox.com>
9188         * config/i386/i386.c (choose_basereg): Use optional scratch
9189         register and add assertion.
9190         (x86_emit_outlined_ms2sysv_save): Use scratch register when
9191         needed, and don't allocate stack.
9192         (ix86_expand_prologue): Rearrange where SSE saves/stub call is
9193         emitted, correct wrong allocation with -mcall-ms2sysv-xlogues.
9194         (ix86_emit_outlined_ms2sysv_restore): Fix non-immediate offsets.
9196 2017-11-03  Jeff Law  <law@redhat.com>
9198         * config/i386/i386.c (ix86_emit_restore_reg_using_pop): Prototype.
9199         (ix86_adjust_stack_and_probe_stack_clash): Use a push/pop sequence
9200         to probe at the start of a noreturn function.
9202 2017-11-03  Jakub Jelinek  <jakub@redhat.com>
9204         PR tree-optimization/78821
9205         * gimple-ssa-store-merging.c: Update the file comment.
9206         (MAX_STORE_ALIAS_CHECKS): Define.
9207         (struct store_operand_info): New type.
9208         (store_operand_info::store_operand_info): New constructor.
9209         (struct store_immediate_info): Add rhs_code and ops data members.
9210         (store_immediate_info::store_immediate_info): Add rhscode, op0r
9211         and op1r arguments to the ctor, initialize corresponding data members.
9212         (struct merged_store_group): Add load_align_base and load_align
9213         data members.
9214         (merged_store_group::merged_store_group): Initialize them.
9215         (merged_store_group::do_merge): Update them.
9216         (merged_store_group::apply_stores): Pick the constant for
9217         encode_tree_to_bitpos from one of the two operands, or skip
9218         encode_tree_to_bitpos if neither operand is a constant.
9219         (class pass_store_merging): Add process_store method decl.  Remove
9220         bool argument from terminate_all_aliasing_chains method decl.
9221         (pass_store_merging::terminate_all_aliasing_chains): Remove
9222         var_offset_p argument and corresponding handling.
9223         (stmts_may_clobber_ref_p): New function.
9224         (compatible_load_p): New function.
9225         (imm_store_chain_info::coalesce_immediate_stores): Terminate group
9226         if there is overlap and rhs_code is not INTEGER_CST.  For
9227         non-overlapping stores terminate group if rhs is not mergeable.
9228         (get_alias_type_for_stmts): Change first argument from
9229         auto_vec<gimple *> & to vec<gimple *> &.  Add IS_LOAD, CLIQUEP and
9230         BASEP arguments.  If IS_LOAD is true, look at rhs1 of the stmts
9231         instead of lhs.  Compute *CLIQUEP and *BASEP in addition to the
9232         alias type.
9233         (get_location_for_stmts): Change first argument from
9234         auto_vec<gimple *> & to vec<gimple *> &.
9235         (struct split_store): Remove orig_stmts data member, add orig_stores.
9236         (split_store::split_store): Create orig_stores rather than orig_stmts.
9237         (find_constituent_stmts): Renamed to ...
9238         (find_constituent_stores): ... this.  Change second argument from
9239         vec<gimple *> * to vec<store_immediate_info *> *, push pointers
9240         to info structures rather than the statements.
9241         (split_group): Rename ALLOW_UNALIGNED argument to
9242         ALLOW_UNALIGNED_STORE, add ALLOW_UNALIGNED_LOAD argument and handle
9243         it.  Adjust find_constituent_stores caller.
9244         (imm_store_chain_info::output_merged_store): Handle rhs_code other
9245         than INTEGER_CST, adjust split_group, get_alias_type_for_stmts and
9246         get_location_for_stmts callers.  Set MR_DEPENDENCE_CLIQUE and
9247         MR_DEPENDENCE_BASE on the MEM_REFs if they are the same in all stores.
9248         (mem_valid_for_store_merging): New function.
9249         (handled_load): New function.
9250         (pass_store_merging::process_store): New method.
9251         (pass_store_merging::execute): Use process_store method.  Adjust
9252         terminate_all_aliasing_chains caller.
9254 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9256         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9257         Return true for more constants, symbols and label references.
9258         (aarch64_valid_floating_const): Remove unused function.
9260 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9262         PR target/82786
9263         * config/aarch64/aarch64.c (aarch64_layout_frame):
9264         Undo forcing of LR at bottom of frame.
9266 2017-11-03  Jeff Law  <law@redhat.com>
9268         PR target/82823
9269         * config/i386/i386.c (ix86_expand_prologue): Tighten assert
9270         for int_registers_saved.
9272         * cfganal.c (single_pred_edge_ignoring_loop_edges): New function
9273         extracted from tree-ssa-dom.c.
9274         * cfganal.h (single_pred_edge_ignoring_loop_edges): Prototype.
9275         * tree-ssa-dom.c (single_incoming_edge_ignoring_loop_edges): Remove.
9276         (record_equivalences_from_incoming_edge): Add additional argument
9277         to single_pred_edge_ignoring_loop_edges call.
9278         * tree-ssa-uncprop.c (single_incoming_edge_ignoring_loop_edges): Remove.
9279         (uncprop_dom_walker::before_dom_children): Add additional argument
9280         to single_pred_edge_ignoring_loop_edges call.
9281         * tree-ssa-sccvn.c (sccvn_dom_walker::before_dom_children): Use
9282         single_pred_edge_ignoring_loop_edges rather than open coding.
9283         * tree-vrp.c (evrp_dom_walker::before_dom_children): Similarly.
9285 2017-11-03  Marc Glisse  <marc.glisse@inria.fr>
9287         * match.pd (-(-A)): Rewrite.
9289 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9291         * config/rs6000/rs60000-protos.h (rs6000_emit_sISEL): Delete.
9292         (rs6000_emit_int_cmove): New declaration.
9293         * config/rs6000/rs6000.c (rs6000_emit_int_cmove): Delete declaration.
9294         (rs6000_emit_sISEL): Delete.
9295         (rs6000_emit_int_cmove): Make non-static.
9296         * config/rs6000/rs6000.md (cstore<mode>4): Use rs6000_emit_int_cmove
9297         instead of rs6000_emit_sISEL.
9299 2017-11-03  Jan Hubicka  <hubicka@ucw.cz>
9301         * asan.c (create_cond_insert_point): Maintain profile.
9302         * ipa-utils.c (ipa_merge_profiles): Be sure only IPA profiles are
9303         merged.
9304         * basic-block.h (struct basic_block_def): Remove frequency.
9305         (EDGE_FREQUENCY): Use to_frequency
9306         * bb-reorder.c (push_to_next_round_p): Use only IPA counts for global
9307         heuristics.
9308         (find_traces): Update to use to_frequency.
9309         (find_traces_1_round): Likewise; use only IPA counts.
9310         (bb_to_key): Likewise.
9311         (connect_traces): Use IPA counts only.
9312         (copy_bb_p): Update to use to_frequency.
9313         (fix_up_crossing_landing_pad): Likewise.
9314         (sanitize_hot_paths): Likewise.
9315         * bt-load.c (basic_block_freq): Likewise.
9316         * cfg.c (init_flow): Set count_max to uninitialized.
9317         (check_bb_profile): Remove frequencies; check counts.
9318         (dump_bb_info): Do not dump frequencies.
9319         (update_bb_profile_for_threading): Update counts only.
9320         (scale_bbs_frequencies_int): Likewise.
9321         (MAX_SAFE_MULTIPLIER): Remove.
9322         (scale_bbs_frequencies_gcov_type): Update counts only.
9323         (scale_bbs_frequencies_profile_count): Update counts only.
9324         (scale_bbs_frequencies): Update counts only.
9325         * cfg.h (struct control_flow_graph): Add count-max.
9326         (update_bb_profile_for_threading): Update prototype.
9327         * cfgbuild.c (find_bb_boundaries): Do not update frequencies.
9328         (find_many_sub_basic_blocks): Likewise.
9329         * cfgcleanup.c (try_forward_edges): Likewise.
9330         (try_crossjump_to_edge): Likewise.
9331         * cfgexpand.c (expand_gimple_cond): Likewise.
9332         (expand_gimple_tailcall): Likewise.
9333         (construct_init_block): Likewise.
9334         (construct_exit_block): Likewise.
9335         * cfghooks.c (verify_flow_info): Check consistency of counts.
9336         (dump_bb_for_graph): Do not dump frequencies.
9337         (split_block_1): Do not update frequencies.
9338         (split_edge): Do not update frequencies.
9339         (make_forwarder_block): Do not update frequencies.
9340         (duplicate_block): Do not update frequencies.
9341         (account_profile_record): Do not update frequencies.
9342         * cfgloop.c (find_subloop_latch_edge_by_profile): Use IPA counts
9343         for global heuristics.
9344         * cfgloopanal.c (average_num_loop_insns): Update to use to_frequency.
9345         (expected_loop_iterations_unbounded): Use counts only.
9346         * cfgloopmanip.c (scale_loop_profile): Simplify.
9347         (create_empty_loop_on_edge): Simplify
9348         (loopify): Simplify
9349         (duplicate_loop_to_header_edge): Simplify
9350         * cfgrtl.c (force_nonfallthru_and_redirect): Update profile.
9351         (update_br_prob_note): Take care of removing note when profile
9352         becomes undefined.
9353         (relink_block_chain): Do not dump frequency.
9354         (rtl_account_profile_record): Use to_frequency.
9355         * cgraph.c (symbol_table::create_edge): Convert count to ipa count.
9356         (cgraph_edge::redirect_call_stmt_to_calle): Conver tcount to ipa count.
9357         (cgraph_update_edges_for_call_stmt_node): Likewise.
9358         (cgraph_edge::verify_count_and_frequency): Update.
9359         (cgraph_node::verify_node): Temporarily disable frequency verification.
9360         * cgraphbuild.c (compute_call_stmt_bb_frequency): Use
9361         to_cgraph_frequency.
9362         (cgraph_edge::rebuild_edges): Convert to ipa counts.
9363         * cgraphunit.c (init_lowered_empty_function): Do not initialize
9364         frequencies.
9365         (cgraph_node::expand_thunk): Update profile.
9366         * except.c (dw2_build_landing_pads): Do not update frequency.
9367         * final.c (compute_alignments): Use to_frequency.
9368         (dump_basic_block_info): Do not dump frequency.
9369         * gimple-pretty-print.c (dump_profile): Do not dump frequency.
9370         (dump_gimple_bb_header): Do not dump frequency.
9371         * gimple-ssa-isolate-paths.c (isolate_path): Do not update frequency;
9372         do update count.
9373         * gimple-streamer-in.c (input_bb): Do not stream frequency.
9374         * gimple-streamer-out.c (output_bb): Do not stream frequency.
9375         * haifa-sched.c (sched_pressure_start_bb): Use to_freuqency.
9376         (init_before_recovery): Do not update frequency.
9377         (sched_create_recovery_edges): Do not update frequency.
9378         * hsa-gen.c (convert_switch_statements): Do not update frequency.
9379         * ipa-cp.c (ipcp_propagate_stage): Update search for max_count.
9380         (ipa_cp_c_finalize): Set max_count to uninitialized.
9381         * ipa-fnsummary.c (get_minimal_bb): Use counts.
9382         (param_change_prob): Use counts.
9383         * ipa-profile.c (ipa_profile_generate_summary): Do not summarize
9384         local profiles.
9385         * ipa-split.c (consider_split): Use to_frequency.
9386         (split_function): Use to_frequency.
9387         * ira-build.c (loop_compare_func): Likewise.
9388         (mark_loops_for_removal): Likewise.
9389         (mark_all_loops_for_removal): Likewise.
9390         * loop-doloop.c (doloop_modify): Do not update frequency.
9391         * loop-unroll.c (unroll_loop_runtime_iterations): Do not update
9392         frequency.
9393         * lto-streamer-in.c (input_function): Update count_max.
9394         * omp-expand.c (expand_omp_taskreg): Update count_max.
9395         * omp-simd-clone.c (simd_clone_adjust): Update profile.
9396         * predict.c (maybe_hot_frequency_p): Use to_frequency.
9397         (maybe_hot_count_p): Use ipa counts only.
9398         (maybe_hot_bb_p): Simplify.
9399         (maybe_hot_edge_p): Simplify.
9400         (probably_never_executed): Do not take frequency argument.
9401         (probably_never_executed_bb_p): Do not pass frequency.
9402         (probably_never_executed_edge_p): Likewise.
9403         (combine_predictions_for_bb): Check that profile is nonzero.
9404         (propagate_freq): Do not set frequency.
9405         (drop_profile): Simplify.
9406         (counts_to_freqs): Simplify.
9407         (expensive_function_p): Use to_frequency.
9408         (propagate_unlikely_bbs_forward): Simplify.
9409         (determine_unlikely_bbs): Simplify.
9410         (estimate_bb_frequencies): Add hack to silence graphite issues.
9411         (compute_function_frequency): Use ipa counts.
9412         (pass_profile::execute): Update.
9413         (rebuild_frequencies): Use counts only.
9414         (force_edge_cold): Use counts only.
9415         * profile-count.c (profile_count::dump): Dump new count types.
9416         (profile_count::differs_from_p): Check compatiblity.
9417         (profile_count::to_frequency): New function.
9418         (profile_count::to_cgraph_frequency): New function.
9419         * profile-count.h (struct function): Declare.
9420         (enum profile_quality): Add profile_guessed_local and
9421         profile_guessed_global0.
9422         (class profile_proability): Decrease number of bits to 29;
9423         update from_reg_br_prob_note and to_reg_br_prob_note.
9424         (class profile_count: Update comment; decrease number of bits
9425         to 61. Check compatibility.
9426         (profile_count::compatible_p): New private member function.
9427         (profile_count::ipa_p): New member function.
9428         (profile_count::operator<): Handle global zero correctly.
9429         (profile_count::operator>): Handle global zero correctly.
9430         (profile_count::operator<=): Handle global zero correctly.
9431         (profile_count::operator>=): Handle global zero correctly.
9432         (profile_count::nonzero_p): New member function.
9433         (profile_count::force_nonzero): New member function.
9434         (profile_count::max): New member function.
9435         (profile_count::apply_scale): Handle IPA scalling.
9436         (profile_count::guessed_local): New member function.
9437         (profile_count::global0): New member function.
9438         (profile_count::ipa): New member function.
9439         (profile_count::to_frequency): Declare.
9440         (profile_count::to_cgraph_frequency): Declare.
9441         * profile.c (OVERLAP_BASE): Delete.
9442         (compute_frequency_overlap): Delete.
9443         (compute_branch_probabilities): Do not use compute_frequency_overlap.
9444         * regs.h (REG_FREQ_FROM_BB): Use to_frequency.
9445         * sched-ebb.c (rank): Use counts only.
9446         * shrink-wrap.c (handle_simple_exit): Use counts only.
9447         (try_shrink_wrapping): Use counts only.
9448         (place_prologue_for_one_component): Use counts only.
9449         * tracer.c (find_best_predecessor): Use to_frequency.
9450         (find_trace): Use to_frequency.
9451         (tail_duplicate): Use to_frequency.
9452         * trans-mem.c (expand_transaction): Do not update frequency.
9453         * tree-call-cdce.c: Do not update frequency.
9454         * tree-cfg.c (gimple_find_sub_bbs): Likewise.
9455         (gimple_merge_blocks): Likewise.
9456         (gimple_split_edge): Likewise.
9457         (gimple_duplicate_sese_region): Likewise.
9458         (gimple_duplicate_sese_tail): Likewise.
9459         (move_sese_region_to_fn): Likewise.
9460         (gimple_account_profile_record): Likewise.
9461         (insert_cond_bb): Likewise.
9462         * tree-complex.c (expand_complex_div_wide): Likewise.
9463         * tree-eh.c (lower_resx): Update profile.
9464         * tree-inline.c (copy_bb): Simplify count scaling; do not scale
9465         frequencies.
9466         (initialize_cfun): Do not initialize frequencies
9467         (freqs_to_counts): Delete.
9468         (copy_cfg_body): Ignore count parameter.
9469         (copy_body): Update.
9470         (expand_call_inline): Update count_max.
9471         (optimize_inline_calls): Update count_max.
9472         (tree_function_versioning): Update count_max.
9473         * tree-ssa-coalesce.c (coalesce_cost_bb): Use to_frequency.
9474         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Do not update
9475         frequency.
9476         * tree-ssa-loop-im.c (execute_sm_if_changed): Use counts only.
9477         * tree-ssa-loop-ivcanon.c (unloop_loops): Do not update freuqency.
9478         (try_peel_loop): Likewise.
9479         * tree-ssa-loop-ivopts.c (get_scaled_computation_cost_at): Use
9480         to_frequency.
9481         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): Pass -1.
9482         (tree_transform_and_unroll_loop): Do not use frequencies
9483         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations):
9484         Use reliable prediction only.
9485         * tree-ssa-loop-unswitch.c (hoist_guard): Do not use frequencies.
9486         * tree-ssa-sink.c (select_best_block): Use to_frequency.
9487         * tree-ssa-tail-merge.c (replace_block_by): Temporarily disable
9488         probability scaling.
9489         * tree-ssa-threadupdate.c (create_block_for_threading): Do
9490         not update frequency
9491         (any_remaining_duplicated_blocks): Likewise.
9492         (update_profile): Likewise.
9493         (estimated_freqs_path): Delete.
9494         (freqs_to_counts_path): Delete.
9495         (clear_counts_path): Delete.
9496         (ssa_fix_duplicate_block_edges): Likewise.
9497         (duplicate_thread_path): Likewise.
9498         * tree-switch-conversion.c (gen_inbound_check): Use counts.
9499         * tree-tailcall.c (decrease_profile): Do not update frequency.
9500         (eliminate_tail_call): Likewise.
9501         * tree-vect-loop-manip.c (vect_do_peeling): Likewise.
9502         * tree-vect-loop.c (scale_profile_for_vect_loop): Likewise.
9503         (optimize_mask_stores): Likewise.
9504         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
9505         * ubsan.c (ubsan_expand_null_ifn): Update profile.
9506         (ubsan_expand_ptr_ifn): Update profile.
9507         * value-prof.c (gimple_ic): Simplify.
9508         * value-prof.h (gimple_ic): Update prototype.
9509         * ipa-inline-transform.c (inline_transform): Fix scaling conditoins.
9510         * ipa-inline.c (compute_uninlined_call_time): Be sure that
9511         counts are nonzero.
9512         (want_inline_self_recursive_call_p): Likewise.
9513         (resolve_noninline_speculation): Only cummulate defined counts.
9514         (inline_small_functions): Use nonzero_p.
9515         (ipa_inline): Do not access freed node.
9517 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9519         * config/aarch64/aarch64.c (aarch64_override_options_internal):
9520         Set PARAM_SCHED_PRESSURE_ALGORITHM to SCHED_PRESSURE_MODEL.
9522 2017-11-03  Kito Cheng  <kito.cheng@gmail.com>
9524         * config/riscv/riscv.c (riscv_legitimize_move): Handle
9525         non-legitimate address.
9527 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9529         * config/rs6000/rs6000.md (*lt0_disi): Delete.
9530         (*lt0_<mode>di, *lt0_<mode>si): New.
9532 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9534         * config/rs6000/rs6000.md (move_from_CR_ov_bit): Change condition to
9535         TARGET_PAIRED_FLOAT.
9537 2017-11-03  Siddhesh Poyarekar  <siddhesh.poyarekar@linaro.org>
9538             Jim Wilson  <jim.wilson@linaro.org>
9540         * config/aarch64/aarch64-cores.def (saphira): New CPU.
9541         * config/aarch64/aarch64-tune.md: Regenerated.
9542         * doc/invoke.texi (AArch64 Options/-mtune): Add "saphira".
9543         * gcc/config/aarch64/aarch64.c (saphira_tunings): New tuning table.
9545 2017-11-03  Cupertino Miranda  <cmiranda@synopsys.com>
9547         * config/arc/arc.c (arc_save_restore): Corrected CFA note.
9548         (arc_expand_prologue): Restore blink for millicode.
9549         * config/arc/linux.h (LINK_EH_SPEC): Defined.
9551 2017-11-03  Richard Sandiford  <richard.sandiford@linaro.org>
9553         PR target/82809
9554         * config/i386/i386.c (ix86_vector_duplicate_value): Use
9555         gen_vec_duplicate after forcing the scalar into a register.
9557 2017-11-02  Segher Boessenkool  <segher@kernel.crashing.org>
9559         * combine (try_combine): Print the insns input to try_combine to the
9560         dump file.
9562 2017-11-02  Steve Ellcey  <sellcey@cavium.com>
9564         PR target/79868
9565         * config/aarch64/aarch64-c.c (aarch64_pragma_target_parse):
9566         Remove second argument from aarch64_process_target_attr call.
9567         * config/aarch64/aarch64-protos.h (aarch64_process_target_attr):
9568         Ditto.
9569         * config/aarch64/aarch64.c (aarch64_attribute_info): Change
9570         field type.
9571         (aarch64_handle_attr_arch): Remove second argument.
9572         (aarch64_handle_attr_cpu): Ditto.
9573         (aarch64_handle_attr_tune): Ditto.
9574         (aarch64_handle_attr_isa_flags): Ditto.
9575         (aarch64_process_one_target_attr): Ditto.
9576         (aarch64_process_target_attr): Ditto.
9577         (aarch64_option_valid_attribute_p): Remove second argument.
9578         on aarch64_process_target_attr call.
9580 2017-11-02  David Malcolm  <dmalcolm@redhat.com>
9582         * diagnostic.c: Include "selftest-diagnostic.h".
9583         (selftest::assert_location_text): New function.
9584         (selftest::test_diagnostic_get_location_text): New function.
9585         (selftest::diagnostic_c_tests): Call it.
9587 2017-11-02  David Malcolm  <dmalcolm@redhat.com>
9589         * Makefile.in (OBJS-libcommon): Add selftest-diagnostic.o.
9590         * diagnostic-show-locus.c: Include "selftest-diagnostic.h".
9591         (class selftest::test_diagnostic_context): Move to...
9592         * selftest-diagnostic.c: New file.
9593         * selftest-diagnostic.h: New file.
9595 2017-11-02  James Bowman  <james.bowman@ftdichip.com>
9597         * config/ft32/ft32.c (ft32_addr_space_legitimate_address_p): increase
9598         offset range for FT32B.
9599         * config/ft32/ft32.h: option "mcompress" enables relaxation.
9600         * config/ft32/ft32.md: Add TARGET_NOPM.
9601         * config/ft32/ft32.opt: Add mft32b, mcompress, mnopm.
9602         * gcc/doc/invoke.texi: Add mft32b, mcompress, mnopm.
9604 2017-11-02  Wilco Dijkstra  <wdijkstr@arm.com>
9606         * config/aarch64/aarch64.h (MALLOC_ABI_ALIGNMENT): New define.
9608 2017-11-02  Jeff Law  <law@redhat.com>
9610         * gimple-ssa-sprintf.c (sprintf_dom_walker): Remove
9611         virtual keyword on FINAL OVERRIDE members.
9613         * tree-ssa-propagate.h (ssa_propagation_engine): Group
9614         virtuals together.  Add virtual destructor.
9615         (substitute_and_fold_engine): Similarly.
9617 2017-11-02  Jan Hubicka  <hubicka@ucw.cz>
9619         * x86-tune.def (X86_TUNE_USE_INCDEC): Enable for Haswell+.
9621 2017-11-02  Richard Biener  <rguenther@suse.de>
9623         PR tree-optimization/82795
9624         * tree-if-conv.c (predicate_mem_writes): Remove bogus assert.
9626 2017-11-02  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
9628         * acinclude.m4 (gcc_AC_INITFINI_ARRAY): Don't require
9629         gcc_SUN_LD_VERSION.
9630         (gcc_GAS_CHECK_FEATURE): Remove.
9631         * configure.ac (ld_vers) <*-*-solaris2*>: Move comments from
9632         gcc_AC_INITFINI_ARRAY here.  Update for Solaris 11.4 changes.
9633         * configure: Regenerate.
9635 2017-11-02  Claudiu Zissulescu <claziss@synopsys.com>
9637         * config/arc/arc.c (hwloop_optimize): Account for empty
9638         body loops.
9640 2017-11-02  Richard Biener  <rguenther@suse.de>
9642         PR middle-end/82765
9643         * varasm.c (decode_addr_const): Make offset HOST_WIDE_INT.
9644         Truncate ARRAY_REF index and element size.
9646 2017-11-01  Palmer Dabbelt  <palmer@dabbelt.com>
9648         * doc/invoke.texi (RISC-V Options): Use "@minus{}2 GB", not "-2 GB".
9650 2017-11-01  Jeff Law  <law@redhat.com>
9652         * tree-ssa-ccp.c (ccp_folder): New class derived from
9653         substitute_and_fold_engine.
9654         (ccp_folder::get_value): New member function.
9655         (ccp_folder::fold_stmt): Renamed from ccp_fold_stmt.
9656         (ccp_fold_stmt): Remove prototype.
9657         (ccp_finalize): Call substitute_and_fold from the ccp_class.
9658         * tree-ssa-copy.c (copy_folder): New class derived from
9659         substitute_and_fold_engine.
9660         (copy_folder::get_value): Renamed from get_value.
9661         (fini_copy_prop): Call substitute_and_fold from copy_folder class.
9662         * tree-vrp.c (vrp_folder): New class derived from
9663         substitute_and_fold_engine.
9664         (vrp_folder::fold_stmt): Renamed from vrp_fold_stmt.
9665         (vrp_folder::get_value): New member function.
9666         (vrp_finalize): Call substitute_and_fold from vrp_folder class.
9667         (evrp_dom_walker::before_dom_children): Similarly for replace_uses_in.
9668         * tree-ssa-propagate.h (substitute_and_fold_engine): New class to
9669         provide a class interface to folder/substitute routines.
9670         (ssa_prop_fold_stmt_fn): Remove typedef.
9671         (ssa_prop_get_value_fn): Likewise.
9672         (subsitute_and_fold): Remove prototype.
9673         (replace_uses_in): Likewise.
9674         * tree-ssa-propagate.c (substitute_and_fold_engine::replace_uses_in):
9675         Renamed from replace_uses_in.  Call the virtual member function
9676         (substitute_and_fold_engine::replace_phi_args_in): Similarly.
9677         (substitute_and_fold_dom_walker): Remove initialization of
9678         data member entries for calbacks.  Add substitute_and_fold_engine
9679         member and initialize it.
9680         (substitute_and_fold_dom_walker::before_dom_children0: Use the
9681         member functions for get_value, replace_phi_args_in c
9682         replace_uses_in, and fold_stmt calls.
9683         (substitute_and_fold_engine::substitute_and_fold): Renamed from
9684         substitute_and_fold.  Remove assert.   Update ctor call.
9686         * tree-ssa-propagate.h (ssa_prop_visit_stmt_fn): Remove typedef.
9687         (ssa_prop_visit_phi_fn): Likewise.
9688         (class ssa_propagation_engine): New class to provide an interface
9689         into ssa_propagate.
9690         * tree-ssa-propagate.c (ssa_prop_visit_stmt): Remove file scoped
9691         variable.
9692         (ssa_prop_visit_phi): Likewise.
9693         (ssa_propagation_engine::simulate_stmt): Moved into class.
9694         Call visit_phi/visit_stmt from the class rather than via
9695         file scoped static variables.
9696         (ssa_propagation_engine::simulate_block): Moved into class.
9697         (ssa_propagation_engine::process_ssa_edge_worklist): Similarly.
9698         (ssa_propagation_engine::ssa_propagate): Similarly.  No longer
9699         set file scoped statics for the visit_stmt/visit_phi callbacks.
9700         * tree-complex.c (complex_propagate): New class derived from
9701         ssa_propagation_engine.
9702         (complex_propagate::visit_stmt): Renamed from complex_visit_stmt.
9703         (complex_propagate::visit_phi): Renamed from complex_visit_phi.
9704         (tree_lower_complex): Call ssa_propagate via the complex_propagate
9705         class.
9706         * tree-ssa-ccp.c: (ccp_propagate): New class derived from
9707         ssa_propagation_engine.
9708         (ccp_propagate::visit_phi): Renamed from ccp_visit_phi_node.
9709         (ccp_propagate::visit_stmt): Renamed from ccp_visit_stmt.
9710         (do_ssa_ccp): Call ssa_propagate from the ccp_propagate class.
9711         * tree-ssa-copy.c (copy_prop): New class derived from
9712         ssa_propagation_engine.
9713         (copy_prop::visit_stmt): Renamed from copy_prop_visit_stmt.
9714         (copy_prop::visit_phi): Renamed from copy_prop_visit_phi_node.
9715         (execute_copy_prop): Call ssa_propagate from the copy_prop class.
9716         * tree-vrp.c (vrp_prop): New class derived from ssa_propagation_engine.
9717         (vrp_prop::visit_stmt): Renamed from vrp_visit_stmt.
9718         (vrp_prop::visit_phi): Renamed from vrp_visit_phi_node.
9719         (execute_vrp): Call ssa_propagate from the vrp_prop class.
9721 2017-11-01  Jakub Jelinek  <jakub@redhat.com>
9723         PR rtl-optimization/82778
9724         PR rtl-optimization/82597
9725         * compare-elim.c (struct comparison): Add in_a_setter field.
9726         (find_comparison_dom_walker::before_dom_children): Remove killed
9727         bitmap and df_simulate_find_defs call, instead walk the defs.
9728         Compute last_setter and initialize in_a_setter.  Merge definitions
9729         with first initialization for a few variables.
9730         (try_validate_parallel): Use insn_invalid_p instead of
9731         recog_memoized.  Return insn rather than just the pattern.
9732         (try_merge_compare): Fix up comment.  Don't uselessly test if
9733         in_a is a REG_P.  Use cmp->in_a_setter instead of walking UD
9734         chains.
9735         (execute_compare_elim_after_reload): Remove df_chain_add_problem
9736         call.
9738 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9739             Alan Hayward  <alan.hayward@arm.com>
9740             David Sherwood  <david.sherwood@arm.com>
9742         * config/aarch64/aarch64.c (aarch64_rtx_costs): Use
9743         aarch64_hard_regno_nregs to get the number of registers
9744         in a mode.
9746 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9747             Alan Hayward  <alan.hayward@arm.com>
9748             David Sherwood  <david.sherwood@arm.com>
9750         * config/aarch64/constraints.md (Upl): Rename to...
9751         (Uaa): ...this.
9752         * config/aarch64/aarch64.md
9753         (*zero_extend<SHORT:mode><GPI:mode>2_aarch64, *addsi3_aarch64_uxtw):
9754         Update accordingly.
9756 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9757             Alan Hayward  <alan.hayward@arm.com>
9758             David Sherwood  <david.sherwood@arm.com>
9760         * config/aarch64/aarch64.c (aarch64_add_constant_internal)
9761         (aarch64_add_constant, aarch64_add_sp, aarch64_sub_sp): Move
9762         earlier in file.
9764 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9765             Alan Hayward  <alan.hayward@arm.com>
9766             David Sherwood  <david.sherwood@arm.com>
9768         * config/aarch64/aarch64.c (aarch64_evpc_trn, aarch64_evpc_uzp)
9769         (aarch64_evpc_zip, aarch64_evpc_ext, aarch64_evpc_rev)
9770         (aarch64_evpc_dup): Generate rtl direcly, rather than using
9771         named expanders.
9772         (aarch64_expand_vec_perm_const_1): Explicitly check for permutes
9773         of a single element.
9774         * config/aarch64/iterators.md: Add a comment above the permute
9775         unspecs to say that they are generated directly by
9776         aarch64_expand_vec_perm_const.
9777         * config/aarch64/aarch64-simd.md: Likewise the permute instructions.
9779 2017-11-01  Nathan Sidwell  <nathan@acm.org>
9781         * tree-dump.c (dequeue_and_dump): Use HAS_DECL_ASSEMBLER_NAME_P.
9783 2017-11-01  Palmer Dabbelt  <palmer@dabbelt.com>
9785         * doc/invoke.texi (RISC-V Options): Explicitly name the medlow
9786         and medany code models, and describe what they do.
9788 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9790         Revert accidental duplicate:
9792         * combine.c (can_change_dest_mode): Reject changes in
9793         REGMODE_NATURAL_SIZE.
9795 2017-11-01  Segher Boessenkool  <segher@kernel.crashing.org>
9797         PR rtl-optimization/64682
9798         PR rtl-optimization/69567
9799         PR rtl-optimization/69737
9800         PR rtl-optimization/82683
9801         * combine.c (distribute_notes) <REG_DEAD>: If the new I2 sets the same
9802         register mentioned in the note, drop the note, unless it came from I3,
9803         in which case it should go to I3 again.
9805 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9807         * tree-ssa-dse.c (normalize_ref): Check whether the ranges overlap
9808         and return false if not.
9809         (clear_bytes_written_by, live_bytes_read): Update accordingly.
9811 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9813         * tree-ssa-alias.h (ranges_overlap_p): Return false if either
9814         range is known to be empty.
9816 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9817             Alan Hayward  <alan.hayward@arm.com>
9818             David Sherwood  <david.sherwood@arm.com>
9820         * simplify-rtx.c (simplify_const_unary_operation): Use GET_MODE_NUNITS
9821         and CONST_VECTOR_NUNITS instead of computing the number of units from
9822         the byte sizes of the vector and element.
9823         (simplify_binary_operation_1): Likewise.
9824         (simplify_const_binary_operation): Likewise.
9825         (simplify_ternary_operation): Likewise.
9827 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9828             Alan Hayward  <alan.hayward@arm.com>
9829             David Sherwood  <david.sherwood@arm.com>
9831         * var-tracking.c (INT_MEM_OFFSET): Replace with...
9832         (int_mem_offset): ...this new function.
9833         (var_mem_set, var_mem_delete_and_set, var_mem_delete)
9834         (find_mem_expr_in_1pdv, dataflow_set_preserve_mem_locs)
9835         (same_variable_part_p, use_type, add_stores, vt_get_decl_and_offset):
9836         Update accordingly.
9838 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9839             Alan Hayward  <alan.hayward@arm.com>
9840             David Sherwood  <david.sherwood@arm.com>
9842         * lower-subreg.c (interesting_mode_p): New function.
9843         (compute_costs, find_decomposable_subregs, decompose_register)
9844         (simplify_subreg_concatn, can_decompose_p, resolve_simple_move)
9845         (resolve_clobber, dump_choices): Use it.
9847 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9848             Alan Hayward  <alan.hayward@arm.com>
9849             David Sherwood  <david.sherwood@arm.com>
9851         * rtlhash.c (add_rtx): Use add_hwi for 'w' and add_int for 'i'.
9853 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9854             Alan Hayward  <alan.hayward@arm.com>
9855             David Sherwood  <david.sherwood@arm.com>
9857         * alias.c (find_base_value, find_base_term): Only process integer
9858         truncations.  Check the precision rather than the size.
9860 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9861             Alan Hayward  <alan.hayward@arm.com>
9862             David Sherwood  <david.sherwood@arm.com>
9864         * machmode.h (is_narrower_int_mode): New function
9865         * optabs.c (expand_float, expand_fix): Use it.
9866         * dwarf2out.c (rotate_loc_descriptor): Likewise.
9868 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9869             Alan Hayward  <alan.hayward@arm.com>
9870             David Sherwood  <david.sherwood@arm.com>
9872         * rtl.h (narrower_subreg_mode): New function.
9873         * ira-color.c (update_costs_from_allocno): Use it.
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         * optabs-query.h (convert_optab_p): New function, split out from...
9880         (convert_optab_handler): ...here.
9881         (widening_optab_handler): Delete.
9882         (find_widening_optab_handler): Remove permit_non_widening parameter.
9883         (find_widening_optab_handler_and_mode): Likewise.  Provide an
9884         override that operates on mode class wrappers.
9885         * optabs-query.c (widening_optab_handler): Delete.
9886         (find_widening_optab_handler_and_mode): Remove permit_non_widening
9887         parameter.  Assert that the two modes are the same class and that
9888         the "from" mode is narrower than the "to" mode.  Use
9889         convert_optab_handler instead of widening_optab_handler.
9890         * expmed.c (expmed_mult_highpart_optab): Use convert_optab_handler
9891         instead of widening_optab_handler.
9892         * expr.c (expand_expr_real_2): Update calls to
9893         find_widening_optab_handler.
9894         * optabs.c (expand_widen_pattern_expr): Likewise.
9895         (expand_binop_directly): Take the insn_code as a parameter.
9896         (expand_binop): Only call find_widening_optab_handler for
9897         conversion optabs; use optab_handler otherwise.  Update calls
9898         to find_widening_optab_handler and expand_binop_directly.
9899         Use convert_optab_handler instead of widening_optab_handler.
9900         * tree-ssa-math-opts.c (convert_mult_to_widen): Update calls to
9901         find_widening_optab_handler and use scalar_mode rather than
9902         machine_mode.
9903         (convert_plusminus_to_widen): Likewise.
9905 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9906             Alan Hayward  <alan.hayward@arm.com>
9907             David Sherwood  <david.sherwood@arm.com>
9909         * machmode.h (fixed_size_mode): New class.
9910         * rtl.h (get_pool_mode): Return fixed_size_mode.
9911         * gengtype.c (main): Add fixed_size_mode.
9912         * target.def (get_raw_result_mode): Return a fixed_size_mode.
9913         (get_raw_arg_mode): Likewise.
9914         * doc/tm.texi: Regenerate.
9915         * targhooks.h (default_get_reg_raw_mode): Return a fixed_size_mode.
9916         * targhooks.c (default_get_reg_raw_mode): Likewise.
9917         * config/ia64/ia64.c (ia64_get_reg_raw_mode): Likewise.
9918         * config/mips/mips.c (mips_get_reg_raw_mode): Likewise.
9919         * config/msp430/msp430.c (msp430_get_raw_arg_mode): Likewise.
9920         (msp430_get_raw_result_mode): Likewise.
9921         * config/avr/avr-protos.h (regmask): Use as_a <fixed_side_mode>
9922         * dbxout.c (dbxout_parms): Require fixed-size modes.
9923         * expr.c (copy_blkmode_from_reg, copy_blkmode_to_reg): Likewise.
9924         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Likewise.
9925         * omp-low.c (lower_oacc_reductions): Likewise.
9926         * simplify-rtx.c (simplify_immed_subreg): Take fixed_size_modes.
9927         (simplify_subreg): Update accordingly.
9928         * varasm.c (constant_descriptor_rtx::mode): Change to fixed_size_mode.
9929         (force_const_mem): Update accordingly.  Return NULL_RTX for modes
9930         that aren't fixed-size.
9931         (get_pool_mode): Return a fixed_size_mode.
9932         (output_constant_pool_2): Take a fixed_size_mode.
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         * doc/rtl.texi (vec_series): Document.
9939         (const): Say that the operand can be a vec_series.
9940         * rtl.def (VEC_SERIES): New rtx code.
9941         * rtl.h (const_vec_series_p_1): Declare.
9942         (const_vec_series_p): New function.
9943         * emit-rtl.h (gen_const_vec_series): Declare.
9944         (gen_vec_series): Likewise.
9945         * emit-rtl.c (const_vec_series_p_1, gen_const_vec_series)
9946         (gen_vec_series): Likewise.
9947         * optabs.c (expand_mult_highpart): Use gen_const_vec_series.
9948         * simplify-rtx.c (simplify_unary_operation): Handle negations
9949         of vector series.
9950         (simplify_binary_operation_series): New function.
9951         (simplify_binary_operation_1): Use it.  Handle VEC_SERIES.
9952         (test_vector_ops_series): New function.
9953         (test_vector_ops): Call it.
9954         * config/powerpcspe/altivec.md (altivec_lvsl): Use
9955         gen_const_vec_series.
9956         (altivec_lvsr): Likewise.
9957         * config/rs6000/altivec.md (altivec_lvsl, altivec_lvsr): Likewise.
9959 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9960             Alan Hayward  <alan.hayward@arm.com>
9961             David Sherwood  <david.sherwood@arm.com>
9963         * doc/rtl.texi (const): Update description of address constants.
9964         Say that vector constants are allowed too.
9965         * common.md (E, F): Use CONSTANT_P instead of checking for
9966         CONST_VECTOR.
9967         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9968         checking for CONST_VECTOR.
9969         * expmed.c (make_tree): Use build_vector_from_val for a CONST
9970         VEC_DUPLICATE.
9971         * expr.c (expand_expr_real_2): Check for vector modes instead
9972         of checking for CONST_VECTOR.
9973         * rtl.h (const_vec_p): New function.
9974         (const_vec_duplicate_p): Check for a CONST VEC_DUPLICATE.
9975         (unwrap_const_vec_duplicate): Handle them here too.
9977 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9978             David Malcolm  <dmalcolm@redhat.com>
9979             Alan Hayward  <alan.hayward@arm.com>
9980             David Sherwood  <david.sherwood@arm.com>
9982         * rtl.h (vec_duplicate_p): New function.
9983         * selftest-rtl.c (assert_rtx_eq_at): New function.
9984         * selftest-rtl.h (ASSERT_RTX_EQ): New macro.
9985         (assert_rtx_eq_at): Declare.
9986         * selftest.h (selftest::simplify_rtx_c_tests): Declare.
9987         * selftest-run-tests.c (selftest::run_tests): Call it.
9988         * simplify-rtx.c: Include selftest.h and selftest-rtl.h.
9989         (simplify_unary_operation_1): Recursively handle vector duplicates.
9990         (simplify_binary_operation_1): Likewise.  Handle VEC_SELECTs of
9991         vector duplicates.
9992         (simplify_subreg): Handle subregs of vector duplicates.
9993         (make_test_reg, test_vector_ops_duplicate, test_vector_ops)
9994         (selftest::simplify_rtx_c_tests): New functions.
9996 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9997             Alan Hayward  <alan.hayward@arm.com>
9998             David Sherwood  <david.sherwood@arm.com>
10000         * emit-rtl.h (gen_const_vec_duplicate): Declare.
10001         (gen_vec_duplicate): Likewise.
10002         * emit-rtl.c (gen_const_vec_duplicate_1): New function, split
10003         out from...
10004         (gen_const_vector): ...here.
10005         (gen_const_vec_duplicate, gen_vec_duplicate): New functions.
10006         (gen_rtx_CONST_VECTOR): Use gen_const_vec_duplicate for constants
10007         whose elements are all equal.
10008         * optabs.c (expand_vector_broadcast): Use gen_const_vec_duplicate.
10009         * simplify-rtx.c (simplify_const_unary_operation): Likewise.
10010         (simplify_relational_operation): Likewise.
10011         * config/aarch64/aarch64.c (aarch64_simd_gen_const_vector_dup):
10012         Likewise.
10013         (aarch64_simd_dup_constant): Use gen_vec_duplicate.
10014         (aarch64_expand_vector_init): Likewise.
10015         * config/arm/arm.c (neon_vdup_constant): Likewise.
10016         (neon_expand_vector_init): Likewise.
10017         (arm_expand_vec_perm): Use gen_const_vec_duplicate.
10018         (arm_block_set_unaligned_vect): Likewise.
10019         (arm_block_set_aligned_vect): Likewise.
10020         * config/arm/neon.md (neon_copysignf<mode>): Likewise.
10021         * config/i386/i386.c (ix86_expand_vec_perm): Likewise.
10022         (expand_vec_perm_even_odd_pack): Likewise.
10023         (ix86_vector_duplicate_value): Use gen_vec_duplicate.
10024         * config/i386/sse.md (one_cmpl<mode>2): Use CONSTM1_RTX.
10025         * config/ia64/ia64.c (ia64_expand_vecint_compare): Use
10026         gen_const_vec_duplicate.
10027         * config/ia64/vect.md (addv2sf3, subv2sf3): Use CONST1_RTX.
10028         * config/mips/mips.c (mips_gen_const_int_vector): Use
10029         gen_const_vec_duplicate.
10030         (mips_expand_vector_init): Use CONST0_RTX.
10031         * config/powerpcspe/altivec.md (abs<mode>2, nabs<mode>2): Likewise.
10032         (define_split): Use gen_const_vec_duplicate.
10033         * config/rs6000/altivec.md (abs<mode>2, nabs<mode>2): Use CONST0_RTX.
10034         (define_split): Use gen_const_vec_duplicate.
10035         * config/s390/vx-builtins.md (vec_genmask<mode>): Likewise.
10036         (vec_ctd_s64, vec_ctd_u64, vec_ctsl, vec_ctul): Likewise.
10037         * config/spu/spu.c (spu_const): Likewise.
10039 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
10040             Alan Hayward  <alan.hayward@arm.com>
10041             David Sherwood  <david.sherwood@arm.com>
10043         * combine.c (can_change_dest_mode): Reject changes in
10044         REGMODE_NATURAL_SIZE.
10046 2017-10-31  Sandra Loosemore  <sandra@codesourcery.com>
10048         * configure.ac (--enable-libssp): New.
10049         (gcc_cv_libc_provides_ssp): Check for explicit setting before
10050         trying to determine target-specific default.  Adjust indentation.
10051         * configure: Regenerated.
10052         * doc/install.texi (Configuration): Expand --disable-libssp
10053         documentation.
10055 2017-10-31  Daniel Santos  <daniel.santos@pobox.com>
10057         config/i386/i386.c (ix86_expand_epilogue): Correct stack
10058         calculation.
10060 2017-10-31  Martin Jambor  <mjambor@suse.cz>
10062         PR c++/81702
10063         * gimple-fold.c (gimple_get_virt_method_for_vtable): Remove assert.
10065 2017-10-31  David Malcolm  <dmalcolm@redhat.com>
10067         * auto-profile.c (autofdo_source_profile::read): Use
10068         UNKNOWN_LOCATION rather than 0.
10069         * diagnostic-core.h (warning_at_rich_loc): Rename to...
10070         (warning_at): ...this overload.
10071         (warning_at_rich_loc_n): Rename to...
10072         (warning_n): ...this overload.
10073         (error_at_rich_loc): Rename to...
10074         (error_at): ...this overload.
10075         (pedwarn_at_rich_loc): Rename to...
10076         (pedwarn): ...this overload.
10077         (permerror_at_rich_loc): Rename to...
10078         (permerror): ...this overload.
10079         (inform_at_rich_loc): Rename to...
10080         (inform): ...this overload.
10081         * diagnostic.c: (diagnostic_n_impl): Delete location_t-based decl.
10082         (diagnostic_n_impl_richloc): Rename to...
10083         (diagnostic_n_impl): ...this rich_location *-based decl.
10084         (inform_at_rich_loc): Rename to...
10085         (inform): ...this, and add an assertion.
10086         (inform_n): Update for removal of location_t-based diagnostic_n_impl.
10087         (warning_at_rich_loc): Rename to...
10088         (warning_at): ...this, and add an assertion.
10089         (warning_at_rich_loc_n): Rename to...
10090         (warning_n): ...this, and add an assertion.
10091         (warning_n): Update location_t-based implementation for removal of
10092         location_t-based diagnostic_n_impl.
10093         (pedwarn_at_rich_loc): Rename to...
10094         (pedwarn): ...this, and add an assertion.
10095         (permerror_at_rich_loc): Rename to...
10096         (permerror): ...this, and add an assertion.
10097         (error_n): Update for removal of location_t-based diagnostic_n_impl.
10098         (error_at_rich_loc): Rename to...
10099         (error_at): ...this, and add an assertion.
10100         * gcc.c (do_spec_1): Use UNKNOWN_LOCATION rather than 0.
10101         (driver::do_spec_on_infiles): Likewise.
10102         * substring-locations.c (format_warning_va): Update for renaming
10103         of inform_at_rich_loc.
10105 2017-10-31  Michael Meissner  <meissner@linux.vnet.ibm.com>
10107         * builtins.def (DEF_FLOATN_BUILTIN): Change most _Float<N> and
10108         _Float<N>X built-in functions so that the variant without the
10109         "__builtin_" prefix is only enabled for the GNU C and Objective C
10110         languages when they are in non-strict ANSI/ISO mode.
10111         (DEF_EXT_LIB_FLOATN_NX_BUILTINS): Likewise.
10112         * target.def (floatn_builtin_p): Add a target hook to control
10113         whether _Float<N> and _Float<N>X built-in functions without the
10114         "__builtin_" prefix are enabled, and return true for C and
10115         Objective C in the default hook.  Include langhooks.h in
10116         targhooks.c.
10117         * targhooks.h (default_floatn_builtin_p): Likewise.
10118         * targhooks.c (default_floatn_builtin_p): Likewise.
10119         * doc/tm.texi.in (TARGET_FLOATN_BUILTIN_P): Document the
10120         floatn_builtin_p target hook.
10121         * doc/tm.texi (TARGET_FLOATN_BUILTIN_P): Likewise.
10123 2017-10-31  Matthew Fortune  <matthew.fortune@imgtec.com>
10124             Eric Botcazou  <ebotcazou@adacore.com>
10126         PR rtl-optimization/81803
10127         * lra-constraints.c (curr_insn_transform): Also reload the whole
10128         register for a strict subreg no wider than a word if this is for
10129         a WORD_REGISTER_OPERATIONS target.
10131 2017-10-31  Jason Merrill  <jason@redhat.com>
10133         * gdbinit.in: Skip over inlines from timevar.h.
10135 2017-10-31  Martin Liska  <mliska@suse.cz>
10137         * doc/gcov.texi: Document new option.
10138         * gcov.c (print_usage): Likewise print it.
10139         (process_args): Support the argument.
10140         (format_count): New function.
10141         (format_gcov): Use the function.
10143 2017-10-31  Martin Liska  <mliska@suse.cz>
10145         * gcov.c (struct name_map): do not use typedef.
10146         Define operator== and operator<.
10147         (name_search): Remove.
10148         (name_sort): Remove.
10149         (main): Do not allocate names.
10150         (process_file): Add vertical space.
10151         (generate_results): Use std::find.
10152         (release_structures): Do not release memory.
10153         (find_source): Use std::find.
10155 2017-10-31  Martin Liska  <mliska@suse.cz>
10157         * gcov.c (struct line_info): Remove it's typedef.
10158         (line_info::line_info): Add proper ctor.
10159         (line_info::has_block): Do not use a typedef.
10160         (struct source_info): Do not use typedef.
10161         (circuit): Likewise.
10162         (get_cycles_count): Likewise.
10163         (output_intermediate_file): Iterate via vector iterator.
10164         (add_line_counts): Use std::vector methods.
10165         (accumulate_line_counts): Likewise.
10166         (output_lines): Likewise.
10168 2017-10-31  Martin Liska  <mliska@suse.cz>
10170         * gcov.c (struct source_info): Remove typedef.
10171         (source_info::source_info): Add proper ctor.
10172         (accumulate_line_counts): Use struct, not it's typedef.
10173         (output_gcov_file): Likewise.
10174         (output_lines): Likewise.
10175         (main): Do not allocate an array.
10176         (output_intermediate_file): Use size of vector container.
10177         (process_file): Resize the vector.
10178         (generate_results): Do not preallocate, use newly added vector
10179         lines.
10180         (release_structures): Do not release sources.
10181         (find_source): Use vector methods.
10182         (add_line_counts): Do not use typedef.
10184 2017-10-31  Martin Liska  <mliska@suse.cz>
10186         * doc/gcov.texi: Document that.
10187         * gcov.c (add_line_counts): Mark lines with a non-executed
10188         statement.
10189         (output_line_beginning): Handle such lines.
10190         (output_lines): Pass new argument.
10191         (output_intermediate_file): Print it in intermediate format.
10193 2017-10-31  Martin Liska  <mliska@suse.cz>
10195         * color-macros.h: New file.
10196         * diagnostic-color.c: Factor out color related to macros to
10197         color-macros.h.
10198         * doc/gcov.texi: Document -k option.
10199         * gcov.c (INCLUDE_STRING): Include string.h.
10200         (print_usage): Add -k option.
10201         (process_args): Parse it.
10202         (pad_count_string): New function.
10203         (output_line_beginning): Likewise.
10204         (DEFAULT_LINE_START): New macro.
10205         (output_lines): Support color output.
10207 2017-10-31  Martin Liska  <mliska@suse.cz>
10209         PR gcov-profile/82633
10210         * doc/gcov.texi: Document -fkeep-{static,inline}-functions and
10211         their interaction with GCOV infrastructure.
10212         * configure.ac: Add -fkeep-{inline,static}-functions to
10213         coverage_flags.
10214         * configure: Regenerate.
10216 2017-10-31  Uros Bizjak  <ubizjak@gmail.com>
10218         PR target/82772
10219         * config/alpha/sync.md (fetchop_constr) <and>: Change to "rINM".
10221 2017-10-31  Segher Boessenkool  <segher@kernel.crashing.org>
10223         PR target/82674
10224         * config/rs6000/rs6000.md (allocate_stack): Force update interval
10225         into a register if it does not fit into an immediate offset field.
10227 2017-10-31  Olivier Hainque  <hainque@adacore.com>
10229         * gcc/Makefile.in (FLAGS_TO_PASS): Pass libsubdir as well.
10231 2017-10-31  Julia Koval  <julia.koval@intel.com>
10233         * config.gcc: Add gfniintrin.h.
10234         * config/i386/gfniintrin.h: New.
10235         * config/i386/i386-builtin-types.def
10236         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
10237         __builtin_ia32_vgf2p8affineinvqb_v64qi_mask,
10238         __builtin_ia32_vgf2p8affineinvqb_v32qi,
10239         __builtin_ia32_vgf2p8affineinvqb_v32qi_mask,
10240         __builtin_ia32_vgf2p8affineinvqb_v16qi,
10241         __builtin_ia32_vgf2p8affineinvqb_v16qi_mask): New builtins.
10242         * config/i386/i386-builtin.def (V64QI_FTYPE_V64QI_V64QI_INT_V64QI_UDI,
10243         V32QI_FTYPE_V32QI_V32QI_INT_V32QI_USI,
10244         V16QI_FTYPE_V16QI_V16QI_INT_V16QI_UHI,
10245         V64QI_FTYPE_V64QI_V64QI_INT): New types.
10246         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
10247         * config/i386/immintrin.h: Include gfniintrin.h.
10248         * config/i386/sse.md (vgf2p8affineinvqb_*) New pattern.
10250 2017-10-30  Eric Botcazou  <ebotcazou@adacore.com>
10252         * gcc.c (HAVE_TARGET_EXECUTABLE_SUFFIX): Remove old kludge.
10254 2017-10-30  Wilco Dijkstra  <wdijkstr@arm.com>
10256         * config/arm/arm.md (ashldi3): Remove shift by 1 expansion.
10257         (arm_ashldi3_1bit): Remove pattern.
10258         (ashrdi3): Remove shift by 1 expansion.
10259         (arm_ashrdi3_1bit): Remove pattern.
10260         (lshrdi3): Remove shift by 1 expansion.
10261         (arm_lshrdi3_1bit): Remove pattern.
10262         * config/arm/arm.c (arm_rtx_costs_internal): Slightly increase
10263         cost of ashldi3 by 1.
10264         * config/arm/neon.md (ashldi3_neon): Remove shift by 1 expansion.
10265         (<shift>di3_neon): Likewise.
10267 2017-10-30  Dominik Infuehr  <dominik.infuehr@theobroma-systems.com>
10269         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov): Rename
10270         both identically named patterns to (*aarch64_simd_mov<VD:mode>)
10271         and (*aarch64_simd_mov<VQ:mode>).
10272         (*aarch64_simd_mov<VD:mode>): Change type attribute to match
10273         pattern alternative.
10274         (*aarch64_simd_mov<VQ:mode>): Re-order and change type
10275         attributes to match pattern alternative.
10277 2017-10-30  Steven Munroe  <munroesj@gcc.gnu.org>
10279         * config.gcc (powerpc*-*-*): Add emmintrin.h.
10280         * config/rs6000/emmintrin.h: New file.
10281         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include emmintrin.h.
10283 2017-10-30  Wilco Dijkstra  <wdijkstr@arm.com>
10285         * config/arm/vfp.md (movdi_vfp): Merge changes from movdi_vfp_cortexa8.
10286         * (movdi_vfp_cortexa8): Remove pattern.
10288 2017-10-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10290         * doc/install.texi (Specific, alpha*-*-*): Remove DEC OSF/1
10291         etc. reference.
10292         (Specific, alpha*-dec-osf5.1): Remove.
10293         (Specific, mips-sgi-irix5): Remove.
10294         (Specific, mips-sgi-irix6): Remove.
10296 2017-10-30  Jakub Jelinek  <jakub@redhat.com>
10298         PR middle-end/22141
10299         * gimple-ssa-store-merging.c (merged_store_group::apply_stores): Fix
10300         arguments to clear_bit_region_be.
10302 2017-10-30  Jim Wilson  <wilson@tuliptree.org>
10304         * gimplify.c: Include memmodel.h.
10306 2017-10-30  Martin Jambor  <mjambor@suse.cz>
10308         * omp-grid.c (grid_attempt_target_gridification): Also insert a
10309         condition whether loop should be executed at all.
10311 2017-10-30  Will Schmidt  <will_schmidt@vnet.ibm.com>
10313         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
10314         gimple folding of vec_madd() intrinsics.
10315         * config/rs6000/altivec.md (mulv8hi3): Rename altivec_vmladduhm to
10316         fmav8hi4.  (altivec_vmladduhm): Rename to fmav8hi4.
10317         * config/rs6000/rs6000-builtin.def: Rename vmladduhm to fmav8hi4
10319 2017-10-30  Richard Biener  <rguenther@suse.de>
10321         PR tree-optimization/82762
10322         Revert
10323         2017-10-23  Richard Biener  <rguenther@suse.de>
10325         PR tree-optimization/82129
10326         Revert
10327         2017-08-01  Richard Biener  <rguenther@suse.de>
10329         PR tree-optimization/81181
10330         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
10331         (compute_antic): ... end of iteration here.
10333 2017-10-30  Joseph Myers  <joseph@codesourcery.com>
10335         * doc/invoke.texi (C Dialect Options): Document -std=c17,
10336         -std=iso9899:2017 and -std=gnu17.
10337         * doc/standards.texi (C Language): Document C17 support.
10338         * doc/cpp.texi (Overview): Mention -std=c17.
10339         (Standard Predefined Macros): Document C11 and C17 values of
10340         __STDC_VERSION__.  Do not refer to C99 support as incomplete.
10341         * doc/extend.texi (Inline): Do not list individual options for
10342         standards newer than C99.
10343         * dwarf2out.c (highest_c_language, gen_compile_unit_die): Handle
10344         "GNU C17".
10345         * config/rl78/rl78.c (rl78_option_override): Handle "GNU C17"
10346         language name.
10348 2017-10-30  Maxim Ostapenko  <m.ostapenko@samsung.com>
10350         * asan.c (asan_finish_file): Align asan globals array by shadow
10351         granularity.
10353 2017-10-30  Jakub Jelinek  <jakub@redhat.com>
10355         PR middle-end/22141
10356         * gimple-ssa-store-merging.c: Include rtl.h and expr.h.
10357         (struct store_immediate_info): Add bitregion_start and bitregion_end
10358         fields.
10359         (store_immediate_info::store_immediate_info): Add brs and bre
10360         arguments and initialize bitregion_{start,end} from those.
10361         (struct merged_store_group): Add bitregion_start, bitregion_end,
10362         align_base and mask fields.  Drop unnecessary struct keyword from
10363         struct store_immediate_info.  Add do_merge method.
10364         (clear_bit_region_be): Use memset instead of loop storing zeros.
10365         (merged_store_group::do_merge): New method.
10366         (merged_store_group::merge_into): Use do_merge.  Allow gaps in between
10367         stores as long as the surrounding bitregions have no gaps.
10368         (merged_store_group::merge_overlapping): Use do_merge.
10369         (merged_store_group::apply_stores): Test that bitregion_{start,end}
10370         is byte aligned, rather than requiring that start and width are
10371         byte aligned.  Drop unnecessary struct keyword from
10372         struct store_immediate_info.  Allocate and populate also mask array.
10373         Make start of the arrays relative to bitregion_start rather than
10374         start and size them according to bitregion_{end,start} difference.
10375         (struct imm_store_chain_info): Drop unnecessary struct keyword from
10376         struct store_immediate_info.
10377         (pass_store_merging::gate): Punt if BITS_PER_UNIT or CHAR_BIT is not 8.
10378         (pass_store_merging::terminate_all_aliasing_chains): Drop unnecessary
10379         struct keyword from struct store_immediate_info.
10380         (imm_store_chain_info::coalesce_immediate_stores): Allow gaps in
10381         between stores as long as the surrounding bitregions have no gaps.
10382         Formatting fixes.
10383         (struct split_store): Add orig non-static data member.
10384         (split_store::split_store): Initialize orig to false.
10385         (find_constituent_stmts): Return store_immediate_info *, non-NULL
10386         if there is exactly a single original stmt.  Change stmts argument
10387         to pointer from reference, if NULL, don't push anything to it.  Add
10388         first argument, use it to optimize skipping over orig stmts that
10389         are known to be before bitpos already.  Simplify.
10390         (split_group): Return unsigned int count how many stores are or
10391         would be needed rather than a bool.  Add allow_unaligned argument.
10392         Change split_stores argument from reference to pointer, if NULL,
10393         only do a dry run computing how many stores would be produced.
10394         Rewritten algorithm to use both alignment and misalign if
10395         !allow_unaligned and handle bitfield stores with gaps.
10396         (imm_store_chain_info::output_merged_store): Set start_byte_pos
10397         from bitregion_start instead of start.  Compute allow_unaligned
10398         here, if true, do 2 split_group dry runs to compute which one
10399         produces fewer stores and prefer aligned if equal.  Punt if
10400         new count is bigger or equal than original before emitting any
10401         statements, rather than during that.  Remove no longer needed
10402         new_ssa_names tracking.  Replace num_stmts with
10403         split_stores.length ().  Use 32-bit stack allocated entries
10404         in split_stores auto_vec.  Try to reuse original store lhs/rhs1
10405         if possible.  Handle bitfields with gaps.
10406         (pass_store_merging::execute): Ignore bitsize == 0 stores.
10407         Compute bitregion_{start,end} for the stores and construct
10408         store_immediate_info with that.  Formatting fixes.
10410 2017-10-30  Uros Bizjak  <ubizjak@gmail.com>
10412         PR target/82725
10413         * config/i386/i386.c (legitimate_pic_address_disp_p): Allow
10414         UNSPEC_DTPOFF and UNSPEC_NTPOFF with SImode immediate offset.
10416 2017-10-29  Jim Wilson  <wilson@tuliptree.org>
10418         * gimplify.c: Include tm_p.h.
10420         * common.opt (gcoff): Re-add as ignored option.
10421         (gcoff1, gcoff2, gcoff3): Likewise.
10423         * Makefile.in (OBJS): Delete sdbout.o.
10424         (GTFILES): Delete $(srcdir)/sdbout.c.
10425         * debug.h: Delete sdb_debug_hooks.
10426         * final.c: Delete sdbout.h include.
10427         (final_scan_insn): Delete SDB_DEBUG check.
10428         (rest_of_clean_state): Likewise.
10429         * output.h: Delete sdb_begin_function_line.
10430         * sdbout.c: Delete.
10431         * sdbout.h: Delete.
10432         * toplev.c: Delete sdbout.h include.
10433         (process_options): Delete SDB_DEBUG check.
10434         * tree-core.h (tree_type_common): Delete pointer field of
10435         tree_type_symtab.
10436         * tree.c (copy_node): Clear TYPE_SYMTAB_DIE instead of
10437         TYPE_SYMTAB_POINTER.
10438         * tree.h (TYPE_SYMTAB_POINTER): Delete.
10439         (TYPE_SYMTAB_IS_POINTER): Delete.
10440         (TYPE_SYMTAB_IS_DIE): Renumber.
10441         * xcoffout.c: Refer to former sdbout.c file.
10442         (xcoffout_begin_prologue): Use past tense for sdbout.c reference.
10444         * doc/install.texi (--with-stabs): Delete COFF and ECOFF info.
10445         * doc/invoke.texi (SEEALSO): Delete adb and sdb references.
10446         (Debugging Options): Delete -gcoff.
10447         (-gstabs): Delete SDB reference.
10448         (-gcoff): Delete.
10449         (-gcoff@var{level}): Delete.
10450         * doc/passes.texi (Debugging information output): Delete SDB and
10451         sdbout.c references.
10452         * doc/tm.texi: Regenerate.
10453         * doc/tm.texi.in (DWARF_CIE_DATA_ALIGNMENT): Delete SDB from xref.
10454         (SDB and DWARF): Change node name to DWARF and delete SDB and COFF
10455         references.
10456         (DEBUGGER_AUTO_OFFSET): Delete COFF and SDB references.
10457         (PREFERRED_DEBUGGING_TYPE): Delete SDB_DEBUG and -gcoff references.
10458         (SDB_DEBUGGING_INFO): Delete.
10459         (PUT_SDB_@dots{}, SDB_DELIM, SDB_ALLOW_UNKNOWN_REFERENCES)
10460         SDB_ALLOW_FORWARD_REFERENCES, SDB_OUTPUT_SOURCE_LINE): Delete.
10461         * target.def (output_source_filename): Delete COFF reference.
10463         * common.opt (gcoff): Delete.
10464         (gxcoff+): Update Negative chain.
10465         * defaults.h: Delete all references to SDB_DEBUGGING_INFO and
10466         SDB_DEBUG.
10467         * dwarf2out.c (gen_array_type_die): Change SDB to debuggers.
10468         * flag-types.h (enum debug_info_type): Delete SDB_DEBUG.
10469         * function.c (number_blocks): Delete SDB_DEBUGGING_INFO, SDB_DEBUG,
10470         and SDB references.
10471         (expand_function_start): Change sdb reference to past tense.
10472         (expand_function_end): Change sdb reference to past tense.
10473         * gcc.c (cpp_unique_options): Delete gcoff3 reference.
10474         * opts.c (debug_type_names): Delete coff entry.
10475         (common_handle_option): Delete OPT_gcoff case.
10476         * system.h (SDB_DEBUG, SDB_DEBUGGING_INFO): Poison.
10478         * config/dbxcoff.h (PREFERRED_DEBUGGING_TYPE): Set to DBX_DEBUG.
10479         * config/cris/cris.h: Delete SDB reference in comment.
10480         * config/i386/cygming.h: Don't define SDB_DEBUGGING_INFO.
10481         (ASM_DECLARE_FUNCTION_NAME): Delete SDB reference from comment.
10482         * config/i386/gas.h: Don't define SDB_DEBUGGING_INFO.
10483         * config/i386/i386.c (svr4_dbx_register_map): Change SDB references
10484         to past tense.
10485         (ix86_expand_prologue): Likewise.
10486         * config/i386/winnt.c (i386_pe_start_function): Don't check SDB_DEBUG.
10487         * config/ia64/ia64.h: Likewise.
10488         * config/m68k/m68kelf.h (DBX_REGISTER_NUMBER): Delete SDB reference.
10489         * config/mips/mips.h (SUBTARGET_ASM_DEBUGGING_SPEC): Delete gcoff*
10490         support.
10491         * config/mmix/mmix.h: Likewise.
10492         * config/nds32/nds32.c: Likewise.
10493         * config/stormy/storym16.h: Likewise.
10494         * config/visium/visium.h: Likewise.
10495         * config/vx-common.h (SDB_DEBUGGING_INFO): Delete undef.
10497 2017-10-28  Sandra Loosemore  <sandra@codesourcery.com>
10499         * config/nios2/nios2.h (FRAME_GROWS_DOWNWARD): Define to 1.
10500         * config/nios2/nios2.c (nios2_initial_elimination_offset):  Make
10501         FRAME_POINTER_REGNUM point at high end of local var area.
10503 2017-10-27  Eric Botcazou  <ebotcazou@adacore.com>
10505         * bb-reorder.c (find_traces_1_round): Fix off-by-one index.
10506         Move comment around.  Do not reset best_edge for a copiable
10507         destination if the copy would cause a partition change.
10508         (better_edge_p): Remove redundant check.
10510 2017-10-27  Uros Bizjak  <ubizjak@gmail.com>
10512         * config/i386/i386-protos.h (ix86_fp_compare_mode): Remove prototype.
10514 2017-10-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
10516         * builtins.c (CASE_MATHFN_FLOATN): New helper macro to add cases
10517         for math functions that have _Float<N> and _Float<N>X variants.
10518         (mathfn_built_in_2): Add support for math functions that have
10519         _Float<N> and _Float<N>X variants.
10520         (DEF_INTERNAL_FLT_FLOATN_FN): New helper macro.
10521         (expand_builtin_mathfn_ternary): Add support for fma with
10522         _Float<N> and _Float<N>X variants.
10523         (expand_builtin): Likewise.
10524         (fold_builtin_3): Likewise.
10525         * builtins.def (DEF_EXT_LIB_FLOATN_NX_BUILTINS): New macro to
10526         create math function _Float<N> and _Float<N>X variants as external
10527         library builtins.
10528         (BUILT_IN_COPYSIGN _Float<N> and _Float<N>X variants) Use
10529         DEF_EXT_LIB_FLOATN_NX_BUILTINS to make built-in functions using
10530         the __builtin_ prefix and if not strict ansi, without the prefix.
10531         (BUILT_IN_FABS _Float<N> and _Float<N>X variants): Likewise.
10532         (BUILT_IN_FMA _Float<N> and _Float<N>X variants): Likewise.
10533         (BUILT_IN_FMAX _Float<N> and _Float<N>X variants): Likewise.
10534         (BUILT_IN_FMIN _Float<N> and _Float<N>X variants): Likewise.
10535         (BUILT_IN_NAN _Float<N> and _Float<N>X variants): Likewise.
10536         (BUILT_IN_SQRT _Float<N> and _Float<N>X variants): Likewise.
10537         * builtin-types.def (BT_FN_FLOAT16_FLOAT16_FLOAT16_FLOAT16): New
10538         function signatures for fma _Float<N> and _Float<N>X variants.
10539         (BT_FN_FLOAT32_FLOAT32_FLOAT32_FLOAT32): Likewise.
10540         (BT_FN_FLOAT64_FLOAT64_FLOAT64_FLOAT64): Likewise.
10541         (BT_FN_FLOAT128_FLOAT128_FLOAT128_FLOAT128): Likewise.
10542         (BT_FN_FLOAT32X_FLOAT32X_FLOAT32X_FLOAT32X): Likewise.
10543         (BT_FN_FLOAT64X_FLOAT64X_FLOAT64X_FLOAT64X): Likewise.
10544         (BT_FN_FLOAT128X_FLOAT128X_FLOAT128X_FLOAT128X): Likewise.
10545         * gencfn-macros.c (print_case_cfn): Add support for math functions
10546         that have _Float<N> and _Float<N>X variants.
10547         (print_define_operator_list): Likewise.
10548         (fltfn_suffixes): Likewise.
10549         (main): Likewise.
10550         * internal-fn.def (DEF_INTERNAL_FLT_FLOATN_FN): New helper macro
10551         for math functions that have _Float<N> and _Float<N>X variants.
10552         (SQRT): Add support for sqrt, copysign, fmin and fmax _Float<N>
10553         and _Float<N>X variants.
10554         (COPYSIGN): Likewise.
10555         (FMIN): Likewise.
10556         (FMAX): Likewise.
10557         * fold-const.c (tree_call_nonnegative_warnv_p): Add support for
10558         copysign, fma, fmax, fmin, and sqrt _Float<N> and _Float<N>X
10559         variants.
10560         (integer_valued_read_call_p): Likewise.
10561         * fold-const-call.c (fold_const_call_ss): Likewise.
10562         (fold_const_call_sss): Add support for copysign, fmin, and fmax
10563         _Float<N> and _Float<N>X variants.
10564         (fold_const_call_ssss): Add support for fma _Float<N> and
10565         _Float<N>X variants.
10566         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Add
10567         support for copysign and fma _Float<N> and _Float<N>X variants.
10568         (backprop::process_builtin_call_use): Likewise.
10569         * tree-call-cdce.c (can_test_argument_range); Add support for
10570         sqrt _Float<N> and _Float<N>X variants.
10571         (edom_only_function): Likewise.
10572         (get_no_error_domain): Likewise.
10573         * tree-ssa-math-opts.c (internal_fn_reciprocal): Likewise.
10574         * tree-ssa-reassoc.c (attempt_builtin_copysign): Add support for
10575         copysign _Float<N> and _Float<N>X variants.
10576         * config/rs6000/rs6000-builtin.def (SQRTF128): Delete, this is now
10577         handled by machine independent code.
10578         (FMAF128): Likewise.
10579         * doc/cpp.texi (Common Predefined Macros): Document defining
10580         __FP_FAST_FMAF<N> and __FP_FAST_FMAF<N>X if the backend supports
10581         fma _Float<N> and _Float<N>X variants.
10583 2017-10-27  Uros Bizjak  <ubizjak@gmail.com>
10585         PR target/82692
10586         * config/i386/i386-modes.def (CCFPU): Remove definition.
10587         * config/i386/i386.c (put_condition_mode): Remove CCFPU mode handling.
10588         (ix86_cc_modes_compatible): Ditto.
10589         (ix86_expand_carry_flag_compare): Ditto.
10590         (ix86_expand_int_movcc): Ditto.
10591         (ix86_expand_int_addcc): Ditto.
10592         (ix86_reverse_condition): Ditto.
10593         (ix86_unordered_fp_compare): Rename from ix86_fp_compare_mode.
10594         Return true/false for unordered/ordered fp comparisons.
10595         (ix86_cc_mode): Always return CCFPmode for float mode comparisons.
10596         (ix86_prepare_fp_compare_args): Update for rename.
10597         (ix86_expand_fp_compare): Update for rename.  Generate unordered
10598         compare RTXes wrapped with UNSPEC_NOTRAP unspec.
10599         (ix86_expand_sse_compare_and_jump): Ditto.
10600         * config/i386/predicates.md (fcmov_comparison_operator):
10601         Remove CCFPU mode handling.
10602         (ix86_comparison_operator): Ditto.
10603         (ix86_carry_flag_operator): Ditto.
10604         * config/i386/i386.md (UNSPEC_NOTRAP): New unspec.
10605         (*cmpu<mode>_i387): Wrap compare RTX with UNSPEC_NOTRAP unspec.
10606         (*cmpu<mode>_cc_i387): Ditto.
10607         (FPCMP): Remove mode iterator.
10608         (unord): Remove mode attribute.
10609         (unord_subst): New define_subst transformation
10610         (unord): New define_subst attribute.
10611         (unordered): Ditto.
10612         (*cmpi<unord><MODEF:mode>): Rewrite using unord_subst transformation.
10613         (*cmpi<unord>xf_i387): Ditto.
10614         * config/i386/sse.md (<sse>_<unord>comi<round_saeonly_name>): Merge
10615         from <sse>_comi<round_saeonly_name> and <sse>_ucomi<round_saeonly_name>
10616         using unord_subst transformation.
10617         * config/i386/subst.md (SUBST_A): Remove CCFP and CCFPU modes.
10618         (round_saeonly): Also handle CCFP mode.
10619         * reg-stack.c (subst_stack_regs_pat): Handle UNSPEC_NOTRAP unspec.
10620         Remove UNSPEC_SAHF unspec handling.
10622 2017-10-27  Jan Hubicka  <hubicka@ucw.cz>
10624         * x86-tune.def (X86_TUNE_INTER_UNIT_MOVES_TO_VEC): Disable for Zen.
10626 2017-10-27  Jeff Law  <law@redhat.com>
10628         * gimple-ssa-sprintf.c: Include domwalk.h.
10629         (class sprintf_dom_walker): New class, derived from dom_walker.
10630         (sprintf_dom_walker::before_dom_children): New function.
10631         (struct call_info): Moved into sprintf_dom_walker class
10632         (compute_formath_length, handle_gimple_call): Likewise.
10633         (sprintf_length::execute): Call the dominator walker rather
10634         than walking the statements.
10636         * tree-vrp.c (check_all_array_refs): Do not use wi->info to smuggle
10637         gimple statement locations.
10638         (check_array_bounds): Corresponding changes.  Get the statement's
10639         location directly from wi->stmt.
10641 2017-10-27  Palmer Dabbelt  <palmer@dabbelt.com>
10643         PR target/82717
10644         * doc/invoke.texi (RISC-V) <-mabi>: Correct and improve.
10646 2017-10-27  Jan Hubicka  <hubicka@ucw.cz>
10648         * config/i386/x86-tune.def (X86_TUNE_PARTIAL_REG_DEPENDENCY,
10649         X86_TUNE_MOVX): Disable for Haswell and newer CPUs.
10651 2017-10-27  Jakub Jelinek  <jakub@redhat.com>
10653         PR target/82703
10654         * config/i386/i386-protos.h (maybe_get_pool_constant): Removed.
10655         * config/i386/i386.c (maybe_get_pool_constant): Removed.
10656         (ix86_split_to_parts): Use avoid_constant_pool_reference instead of
10657         maybe_get_pool_constant.
10658         * config/i386/predicates.md (zero_extended_scalar_load_operand):
10659         Likewise.
10661 2017-10-27  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10663         * doc/install.texi (Specific, i?86-*-solaris2.10): Simplify gas
10664         2.26 caveat.  Update gas and gld versions.
10665         (Specific, *-*-solaris2*): Update binutils version.  Remove caveat
10666         reference.
10668 2017-10-27  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
10670         * cgraph.h (set_malloc_flag): Declare.
10671         * cgraph.c (set_malloc_flag_1): New function.
10672         (set_malloc_flag): Likewise.
10673         * ipa-fnsummary.h (ipa_call_summary): Add new field is_return_callee.
10674         * ipa-fnsummary.c (ipa_call_summary::reset): Set is_return_callee to
10675         false.
10676         (read_ipa_call_summary): Add support for reading is_return_callee.
10677         (write_ipa_call_summary): Stream is_return_callee.
10678         * ipa-inline.c (ipa_inline): Remove call to ipa_free_fn_summary.
10679         * ipa-pure-const.c: Add headers ssa.h, alloc-pool.h, symbol-summary.h,
10680         ipa-prop.h, ipa-fnsummary.h.
10681         (pure_const_names): Change to static.
10682         (malloc_state_e): Define.
10683         (malloc_state_names): Define.
10684         (funct_state_d): Add field malloc_state.
10685         (varying_state): Set malloc_state to STATE_MALLOC_BOTTOM.
10686         (check_retval_uses): New function.
10687         (malloc_candidate_p): Likewise.
10688         (analyze_function): Add support for malloc attribute.
10689         (pure_const_write_summary): Stream malloc_state.
10690         (pure_const_read_summary): Add support for reading malloc_state.
10691         (dump_malloc_lattice): New function.
10692         (propagate_malloc): New function.
10693         (warn_function_malloc): New function.
10694         (ipa_pure_const::execute): Call propagate_malloc and
10695         ipa_free_fn_summary.
10696         (pass_local_pure_const::execute): Add support for malloc attribute.
10697         * ssa-iterators.h (RETURN_FROM_IMM_USE_STMT): New macro.
10698         * doc/invoke.texi: Document Wsuggest-attribute=malloc.
10700 2017-10-27  Martin Liska  <mliska@suse.cz>
10702         PR gcov-profile/82457
10703         * doc/invoke.texi: Document that one needs a non-strict ISO mode
10704         for fork-like functions to be properly instrumented.
10706 2017-10-27  Richard Biener  <rguenther@suse.de>
10708         PR middle-end/81659
10709         * tree-eh.c (pass_lower_eh_dispatch::execute): Free dominator
10710         info when we redirected EH.
10712 2017-10-26  Michael Collison  <michael.collison@arm.com>
10714         * config/aarch64/aarch64.md(<optab>_trunc><vf><GPI:mode>2):
10715         New pattern.
10716         (<optab>_trunchf<GPI:mode>2: New pattern.
10717         (<optab>_trunc<vgp><GPI:mode>2: New pattern.
10718         * config/aarch64/iterators.md (wv): New mode attribute.
10719         (vf, VF): New mode attributes.
10720         (vgp, VGP): New mode attributes.
10721         (s): Update attribute with SImode and DImode prefixes.
10723 2017-10-26  Sandra Loosemore  <sandra@codesourcery.com>
10725         * config/nios2/constraints.md ("S"): Match r0rel_constant_p too.
10726         * config/nios2/nios2-protos.h (r0rel_constant_p): Declare.
10727         * config/nios2/nios2.c: (nios2_r0rel_sec_regex): New.
10728         (nios2_option_overide): Initialize it.  Don't allow R0-relative
10729         addressing with PIC.
10730         (nios2_rtx_costs): Handle r0rel_constant_p like gprel_constant_p.
10731         (nios2_symbolic_constant_p): Likewise.
10732         (nios2_legitimate_address_p): Likewise.
10733         (nios2_r0rel_section_name_p): New.
10734         (nios2_symbol_ref_in_r0rel_data_p): New.
10735         (nios2_emit_move_sequence): Handle r0rel_constant_p.
10736         (r0rel_constant_p): New.
10737         (nios2_print_operand_address): Handle r0rel_constant_p.
10738         (nios2_cdx_narrow_form_p): Likewise.
10739         * config/nios2/nios2.opt (mr0rel-sec=): New option.
10740         * doc/invoke.texi (Option Summary): Add -mr0rel-sec.
10741         (Nios II Options): Document -mr0rel-sec.
10743 2017-10-26  Sandra Loosemore  <sandra@codesourcery.com>
10745         * config/nios2/nios2.c: Include xregex.h.
10746         (nios2_gprel_sec_regex): New.
10747         (nios2_option_overide): Initialize it.  Don't allow GP-relative
10748         addressing with PIC.
10749         (nios2_small_section_name_p): Check for regex match.
10750         * config/nios2/nios2.opt (mgprel-sec=): New option.
10751         * doc/invoke.texi (Option Summary): Add -mgprel-sec.
10752         (Nios II Options): Document -mgprel-sec.
10754 2017-10-26  Jim Wilson  <wilson@tuliptree.org>
10756         * doc/invoke.texi (-fdebug-prefix-map): Expand documentation.
10758 2017-10-26  Tom de Vries  <tom@codesourcery.com>
10760         PR tree-optimization/82707
10761         * gimple.c (gimple_copy): Fix unsharing of
10762         GIMPLE_OMP_{SINGLE,TARGET,TEAMS}.
10764 2017-10-26  Olga Makhotina  <olga.makhotina@intel.com>
10766         * config/i386/avx512fintrin.h (_mm512_cmpeq_pd_mask,
10767         _mm512_cmple_pd_mask, _mm512_cmplt_pd_mask,
10768         _mm512_cmpneq_pd_mask, _mm512_cmpnle_pd_mask,
10769         _mm512_cmpnlt_pd_mask, _mm512_cmpord_pd_mask,
10770         _mm512_cmpunord_pd_mask, _mm512_mask_cmpeq_pd_mask,
10771         _mm512_mask_cmple_pd_mask, _mm512_mask_cmplt_pd_mask,
10772         _mm512_mask_cmpneq_pd_mask, _mm512_mask_cmpnle_pd_mask,
10773         _mm512_mask_cmpnlt_pd_mask, _mm512_mask_cmpord_pd_mask,
10774         _mm512_mask_cmpunord_pd_mask, _mm512_cmpeq_ps_mask,
10775         _mm512_cmple_ps_mask, _mm512_cmplt_ps_mask,
10776         _mm512_cmpneq_ps_mask, _mm512_cmpnle_ps_mask,
10777         _mm512_cmpnlt_ps_mask, _mm512_cmpord_ps_mask,
10778         _mm512_cmpunord_ps_mask, _mm512_mask_cmpeq_ps_mask,
10779         _mm512_mask_cmple_ps_mask, _mm512_mask_cmplt_ps_mask,
10780         _mm512_mask_cmpneq_ps_mask, _mm512_mask_cmpnle_ps_mask,
10781         _mm512_mask_cmpnlt_ps_mask, _mm512_mask_cmpord_ps_mask,
10782         _mm512_mask_cmpunord_ps_mask): New intrinsics.
10784 2017-10-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
10786         * config/rs6000/aix.h (TARGET_IEEEQUAD_DEFAULT): Set long double
10787         default to IBM.
10788         * config/rs6000/darwin.h (TARGET_IEEEQUAD_DEFAULT): Likewise.
10789         * config/rs6000/rs6000.opt (-mabi=ieeelongdouble): Move the
10790         warning to rs6000.c.  Remove the Undocumented flag, since it has
10791         been documented.
10792         (-mabi=ibmlongdouble): Likewise.
10793         * config/rs6000/rs6000.c (TARGET_IEEEQUAD_DEFAULT): If it is not
10794         already set, set the default format for long double.
10795         (rs6000_debug_reg_global): Print whether long double is IBM or
10796         IEEE.
10797         (rs6000_option_override_internal): Rework setting long double
10798         format.  Only warn if the user is changing the long double default
10799         and they did not use -Wno-psabi.
10800         * doc/invoke.texi (PowerPC options): Update the documentation for
10801         -mabi=ieeelongdouble and -mabi=ibmlongdouble.
10803 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10804             Alan Hayward  <alan.hayward@arm.com>
10805             David Sherwood  <david.sherwood@arm.com>
10807         * rtl.h (wider_subreg_mode): New function.
10808         * ira.h (ira_sort_regnos_for_alter_reg): Take a machine_mode *
10809         rather than an unsigned int *.
10810         * ira-color.c (regno_max_ref_width): Replace with...
10811         (regno_max_ref_mode): ...this new variable.
10812         (coalesced_pseudo_reg_slot_compare): Update accordingly.
10813         Use wider_subreg_mode.
10814         (ira_sort_regnos_for_alter_reg): Likewise.  Take a machine_mode *
10815         rather than an unsigned int *.
10816         * lra-constraints.c (uses_hard_regs_p): Use wider_subreg_mode.
10817         (process_alt_operands): Likewise.
10818         (invariant_p): Likewise.
10819         * lra-spills.c (assign_mem_slot): Likewise.
10820         (add_pseudo_to_slot): Likewise.
10821         * lra.c (collect_non_operand_hard_regs): Likewise.
10822         (add_regs_to_insn_regno_info): Likewise.
10823         * reload1.c (regno_max_ref_width): Replace with...
10824         (regno_max_ref_mode): ...this new variable.
10825         (reload): Update accordingly.  Update call to
10826         ira_sort_regnos_for_alter_reg.
10827         (alter_reg): Update to use regno_max_ref_mode.  Call wider_subreg_mode.
10828         (init_eliminable_invariants): Update to use regno_max_ref_mode.
10829         (scan_paradoxical_subregs): Likewise.
10831 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10833         * config/aarch64/aarch64.h (EXIT_IGNORE_STACK): Set if alloca is used.
10834         (aarch64_frame): Add emit_frame_chain boolean.
10835         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
10836         Move eh_return case to aarch64_layout_frame.
10837         (aarch64_layout_frame): Initialize emit_frame_chain.
10838         (aarch64_expand_prologue): Use emit_frame_chain.
10840 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10842         * config/aarch64/aarch64.c (aarch64_layout_frame):
10843         Ensure LR is always stored at the bottom of the callee-saves.
10844         Remove rarely used frame layout which saves callee-saves at top of
10845         frame, so the store of LR can be used as a valid probe in all cases.
10847 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10849         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10850         Improve unaligned TImode/TFmode base/offset split.
10852 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10853             Alan Hayward  <alan.hayward@arm.com>
10854             David Sherwood  <david.sherwood@arm.com>
10856         * caller-save.c (mark_referenced_regs):  Use read_modify_subreg_p.
10857         * combine.c (find_single_use_1): Likewise.
10858         (expand_field_assignment): Likewise.
10859         (move_deaths): Likewise.
10860         * lra-constraints.c (simplify_operand_subreg): Likewise.
10861         (curr_insn_transform): Likewise.
10862         * lra.c (collect_non_operand_hard_regs): Likewise.
10863         (add_regs_to_insn_regno_info): Likewise.
10864         * rtlanal.c (reg_referenced_p): Likewise.
10865         (covers_regno_no_parallel_p): Likewise.
10867 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10869         * wide-int-print.cc (print_hex): Loop based on extract_uhwi.
10870         Don't print any bits outside the precision of the value.
10871         * wide-int.cc (test_printing): Add some new tests.
10873 2017-10-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10875         * configure.ac (gcc_cv_as_ix86_xbrace_comment): Check if assembler
10876         supports -xbrace_comment option.
10877         * configure: Regenerate.
10878         * config.in: Regenerate.
10879         * config/i386/sol2.h (ASM_XBRACE_COMMENT_SPEC): Define.
10880         (ASM_CPU_SPEC): Use it.
10882 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10884         * target.def (static_rtx_alignment): New hook.
10885         * targhooks.h (default_static_rtx_alignment): Declare.
10886         * targhooks.c (default_static_rtx_alignment): New function.
10887         * doc/tm.texi.in (TARGET_STATIC_RTX_ALIGNMENT): New hook.
10888         * doc/tm.texi: Regenerate.
10889         * varasm.c (force_const_mem): Use targetm.static_rtx_alignment
10890         instead of targetm.constant_alignment.  Remove call to
10891         set_mem_attributes.
10892         * config/cris/cris.c (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10893         (cris_preferred_mininum_alignment): New function, split out from...
10894         (cris_constant_alignment): ...here.
10895         (cris_static_rtx_alignment): New function.
10896         * config/i386/i386.c (ix86_static_rtx_alignment): New function,
10897         split out from...
10898         (ix86_constant_alignment): ...here.
10899         (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10900         * config/mmix/mmix.c (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10901         (mmix_static_rtx_alignment): New function.
10902         * config/spu/spu.c (spu_static_rtx_alignment): New function.
10903         (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10905 2017-10-26  Tamar Christina  <tamar.christina@arm.com>
10907         PR target/81800
10908         * config/aarch64/aarch64.md (lrint<GPF:mode><GPI:mode>2):
10909         Add flag_trapping_math and flag_fp_int_builtin_inexact.
10911 2017-10-25  Palmer Dabbelt  <palmer@dabbelt.com>
10913         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Define.
10914         * config/riscv/pic.md (local_pic_load): Rename to local_pic_load_s,
10915         mark as a sign-extending load.
10916         (local_pic_load_u): Define.
10918 2017-10-25  Eric Botcazou  <ebotcazou@adacore.com>
10920         PR middle-end/82062
10921         * fold-const.c (operand_equal_for_comparison_p): Also return true
10922         if ARG0 is a simple variant of ARG1 with narrower precision.
10923         (fold_ternary_loc): Always pass unstripped operands to the predicate.
10925 2017-10-25  Jan Hubicka  <hubicka@ucw.cz>
10927         * i386.c (ix86_builtin_vectorization_cost): Compute scatter/gather
10928         cost correctly.
10929         * i386.h (processor_costs): Add gather_static, gather_per_elt,
10930         scatter_static, scatter_per_elt.
10931         * x86-tune-costs.h: Add new cost entries.
10933 2017-10-25  Richard Biener  <rguenther@suse.de>
10935         * tree-ssa-sccvn.h (vn_eliminate): Declare.
10936         * tree-ssa-pre.c (class eliminate_dom_walker, eliminate,
10937         class pass_fre): Move to ...
10938         * tree-ssa-sccvn.c (class eliminate_dom_walker, vn_eliminate,
10939         class pass_fre): ... here and adjust for statistics.
10941 2017-10-25  Jakub Jelinek  <jakub@redhat.com>
10943         PR libstdc++/81706
10944         * attribs.c (attribute_value_equal): Use omp_declare_simd_clauses_equal
10945         for comparison of OMP_CLAUSEs regardless of flag_openmp{,_simd}.
10946         (duplicate_one_attribute, copy_attributes_to_builtin): New functions.
10947         * attribs.h (duplicate_one_attribute, copy_attributes_to_builtin): New
10948         declarations.
10950 2017-10-25  Richard Biener  <rguenther@suse.de>
10952         * tree-ssa-pre.c (need_eh_cleanup, need_ab_cleanup, el_to_remove,
10953         el_to_fixup, el_todo, el_avail, el_avail_stack, eliminate_avail,
10954         eliminate_push_avail, eliminate_insert): Move inside...
10955         (class eliminate_dom_walker): ... this class in preparation
10956         of move.
10957         (fini_eliminate): Remove by merging with ...
10958         (eliminate): ... this function.  Adjust for class changes.
10959         (pass_pre::execute): Remove fini_eliminate call.
10960         (pass_fre::execute): Likewise.
10962 2017-10-24  Jakub Jelinek  <jakub@redhat.com>
10964         PR target/82460
10965         * config/i386/sse.md (UNSPEC_VPERMI2, UNSPEC_VPERMI2_MASK): Remove.
10966         (VPERMI2, VPERMI2I): New mode iterators.
10967         (<avx512>_vpermi2var<mode>3_maskz): Remove 3 define_expand patterns.
10968         (<avx512>_vpermi2var<mode>3<sd_maskz_name>): Remove 3 define_insn
10969         patterns.
10970         (<avx512>_vpermi2var<mode>3_mask): New define_expand using VPERMI2
10971         mode iterator.  Remove 3 old define_insn patterns.
10972         (*<avx512>_vpermi2var<mode>3_mask): 2 new define_insn patterns.
10973         (<avx512>_vpermt2var<mode>3_maskz): Adjust 1 define_expand to use
10974         VPERMI2 mode iterator, remove the other two expanders.
10975         (<avx512>_vpermt2var<mode>3<sd_maskz_name>): Adjust 1 define_insn
10976         to use VPERMI2 mode iterator, add another alternative for vpermi2*
10977         instructions, remove the other two patterns.
10978         (<avx512>_vpermt2var<mode>3_mask): Adjust 1 define_insn to use VPERMI2
10979         mode iterator, remove the other two patterns.
10980         * config/i386/i386.c (ix86_expand_vec_perm_vpermi2): Renamed to ...
10981         (ix86_expand_vec_perm_vpermt2): ... this.  Swap mask and op0
10982         arguments, use gen_*vpermt2* expanders instead of gen_*vpermi2*
10983         and adjust argument order accordingly.
10984         (ix86_expand_vec_perm): Adjust caller.
10985         (expand_vec_perm_1): Likewise.
10986         (expand_vec_perm_vpermi2_vpshub2): Rename to ...
10987         (expand_vec_perm_vpermt2_vpshub2): ... this.
10988         (ix86_expand_vec_perm_const_1): Adjust caller.
10989         (ix86_vectorize_vec_perm_const_ok): Adjust comments.
10991         PR target/82370
10992         * config/i386/sse.md (VIMAX_AVX2): Remove V4TImode.
10993         (VIMAX_AVX2_AVX512BW, VIMAX_AVX512VL): New mode iterators.
10994         (vec_shl_<mode>): Remove unused expander.
10995         (avx512bw_<shift_insn><mode>3): New define_insn.
10996         (<sse2_avx2>_ashl<mode>3, <sse2_avx2>_lshr<mode>3): Replaced by ...
10997         (<sse2_avx2>_<shift_insn><mode>3): ... this.  New define_insn.
10999 2017-10-24  Paolo Carlini  <paolo.carlini@oracle.com>
11001         PR c++/82466
11002         * doc/invoke.texi ([Wbuiltin-declaration-mismatch]): Extend
11003         description.
11005 2017-10-24  Wilco Dijkstra  <wdijkstr@arm.com>
11007         PR rtl-optimization/82396
11008         * gcc/haifa-sched.c (ready_sort_real): Remove qsort workaround.
11009         (autopref_multipass_init): Simplify initialization.
11010         (autopref_rank_data): Simplify sort order.
11011         * gcc/sched-int.h (autopref_multipass_data_): Remove
11012         multi_mem_insn_p, min_offset and max_offset.
11014 2017-10-24  Wilco Dijkstra  <wdijkstr@arm.com>
11016         PR middle-end/60580
11017         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
11018         Check special value of flag_omit_frame_pointer.
11019         (aarch64_can_eliminate): Likewise.
11020         (aarch64_override_options_after_change_1): Simplify handling of
11021         -fomit-frame-pointer and -fomit-leaf-frame-pointer.
11023 2017-10-24  Richard Biener  <rguenther@suse.de>
11025         PR tree-optimization/82697
11026         * tree-ssa-phiopt.c (cond_store_replacement): Use alias-set
11027         zero for conditional load and unconditional store.
11029 2017-10-24  H.J. Lu  <hongjiu.lu@intel.com>
11031         * doc/install.texi: Document bootstrap-cet.
11033 2017-10-24  H.J. Lu  <hongjiu.lu@intel.com>
11035         PR target/82659
11036         * config/i386/i386.c (rest_of_insert_endbranch): Don't insert
11037         ENDBR instruction at function entrance if function is only
11038         called directly.
11040 2017-10-24  Jakub Jelinek  <jakub@redhat.com>
11042         PR target/82628
11043         * config/i386/i386.md (addcarry<mode>, subborrow<mode>): Change
11044         patterns to better describe from which operation the CF is computed.
11045         (addcarry<mode>_0, subborrow<mode>_0): New patterns.
11046         * config/i386/i386.c (ix86_expand_builtin) <case handlecarry>: Pass
11047         one LTU with [DT]Imode and another one with [SD]Imode.  If arg0
11048         is 0, use _0 suffixed expanders instead of emitting a comparison
11049         before it.
11051 2017-10-06  Sergey Shalnov  <Sergey.Shalnov@intel.com>
11053         * config/i386/i386.md(*movsf_internal, *movdf_internal):
11054         Avoid 512-bit AVX modes for TARGET_PREFER_AVX256.
11056 2017-10-24  Eric Botcazou  <ebotcazou@adacore.com>
11058         PR middle-end/82569
11059         * tree-outof-ssa.h (always_initialized_rtx_for_ssa_name_p): Delete.
11060         * expr.c (expand_expr_real_1) <expand_decl_rtl>: Revert latest change.
11061         * loop-iv.c (iv_get_reaching_def): Likewise.
11062         * cfgexpand.c (expand_one_ssa_partition): Initialize the RTX if the
11063         variable is promoted and the partition contains undefined values.
11065 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
11067         * config/nios2/nios2.c (nios2_rtx_costs): Make costs better
11068         reflect reality.
11069         (nios2_address_cost): Define.
11070         (nios2_legitimize_address): Recognize (exp + constant) directly.
11071         (TARGET_ADDRESS_COST): Define.
11073 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
11075         * config/nios2/nios2-protos.h (nios2_large_constant_p): Declare.
11076         (nios2_symbolic_memory_operand_p): Declare.
11077         (nios2_split_large_constant): Declare.
11078         (nios2_split_symbolic_memory_operand): Declare.
11079         * config/nios2/nios2.c: Adjust includes.
11080         (nios2_symbolic_constant_allowed): New.
11081         (nios2_symbolic_constant_p): New.
11082         (nios2_plus_symbolic_constant_p): New.
11083         (nios2_valid_addr_expr_p): Recognize addresses involving
11084         symbolic constants.
11085         (nios2_legitimate_address_p): Likewise, also LO_SUM.
11086         (nios2_symbolic_memory_operand_p): New.
11087         (nios2_large_constant_p): New.
11088         (nios2_split_large_constant): New.
11089         (nios2_split_plus_large_constant): New.
11090         (nios2_split_symbolic_memory_operand): New.
11091         (nios2_legitimize_address): Code refactoring.  Handle addresses
11092         involving symbolic constants.
11093         (nios2_emit_move_sequence): Likewise.
11094         (nios2_print_operand): Improve error output.
11095         (nios2_print_operand_address): Handle LO_SUM.
11096         (nios2_cdx_narrow_form_p): Likewise.
11097         * config/nios2/nios2.md (movqi_internal): Add splitter for memory
11098         operands involving symbolic constants.
11099         (movhi_internal, movsi_internal): Likewise.
11100         (zero_extendhisi2, zero_extendqi<mode>2): Likewise.
11101         (extendhisi2, extendqi<mode>2): Likewise.
11103 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
11105         * tree-pass.h (PROP_rtl_split_insns): Define.
11106         * recog.c (pass_data_split_all_insns): Provide PROP_rtl_split_insns.
11108 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
11110         * config/nios2/nios2.c (TARGET_LRA_P): Don't override.
11112 2017-10-23  Jakub Jelinek  <jakub@redhat.com>
11114         PR debug/82630
11115         * target.def (const_not_ok_for_debug_p): Default to
11116         default_const_not_ok_for_debug_p instead of hook_bool_rtx_false.
11117         * targhooks.h (default_const_not_ok_for_debug_p): New declaration.
11118         * targhooks.c (default_const_not_ok_for_debug_p): New function.
11119         * dwarf2out.c (const_ok_for_output_1): Only reject UNSPECs for
11120         which targetm.const_not_ok_for_debug_p returned true.
11121         * config/arm/arm.c (arm_const_not_ok_for_debug_p): Return true
11122         for UNSPECs.
11123         * config/powerpcspe/powerpcspe.c (rs6000_const_not_ok_for_debug_p):
11124         Likewise.
11125         * config/rs6000/rs6000.c (rs6000_const_not_ok_for_debug_p): Likewise.
11126         * config/i386/i386.c (ix86_delegitimize_address_1): Don't delegitimize
11127         UNSPEC_GOTOFF with addend into addend - _GLOBAL_OFFSET_TABLE_ + symbol
11128         if !base_term_p.
11129         (ix86_const_not_ok_for_debug_p): New function.
11130         (i386_asm_output_addr_const_extra): Handle UNSPEC_GOTOFF.
11131         (TARGET_CONST_NOT_OK_FOR_DEBUG_P): Redefine.
11133 2017-10-23  David Malcolm  <dmalcolm@redhat.com>
11135         PR bootstrap/82610
11136         * system.h: Conditionally include "unique-ptr.h" if
11137         INCLUDE_UNIQUE_PTR is defined.
11138         * unique-ptr-tests.cc: Remove include of "unique-ptr.h" in favor
11139         of defining INCLUDE_UNIQUE_PTR before including "system.h".
11141 2017-10-23  Sebastian Perta  <sebastian.perta@renesas.com>
11143         * config/rl78/rl78.md: New define_expand "subdi3".
11145 2017-10-23  H.J. Lu  <hongjiu.lu@intel.com>
11147         PR target/82673
11148         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Skip
11149         DF_REF_INSN if DF_REF_INSN_INFO is false.
11151 2017-10-23  Jan Hubicka  <hubicka@ucw.cz>
11153         * i386.c (dimode_scalar_chain::compute_convert_gain): Use
11154         xmm_move instead of sse_move.
11155         (sse_store_index): New function.
11156         (ix86_register_move_cost): Be more sensible about mismatch stall;
11157         model AVX moves correctly; make difference between sse->integer and
11158         integer->sse.
11159         (ix86_builtin_vectorization_cost): Model correctly aligned and unaligned
11160         moves; make difference between SSE and AVX.
11161         * i386.h (processor_costs): Remove sse_move; add xmm_move, ymm_move
11162         and zmm_move. Increase size of sse load and store tables;
11163         add unaligned load and store tables; add ssemmx_to_integer.
11164         * x86-tune-costs.h: Update all entries according to real
11165         move latencies from Agner Fog's manual and chip documentation.
11167 2017-10-23  Jakub Jelinek  <jakub@redhat.com>
11169         PR target/82628
11170         * config/i386/predicates.md (x86_64_dwzext_immediate_operand): New.
11171         * config/i386/constraints.md (Wf): New constraint.
11172         * config/i386/i386.md (UNSPEC_SBB): New unspec.
11173         (cmp<dwi>_doubleword): Removed.
11174         (sub<mode>3_carry_ccc, *sub<mode>3_carry_ccc_1): New patterns.
11175         (sub<mode>3_carry_ccgz): Use unspec instead of compare.
11176         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>: Don't
11177         expand with cmp<dwi>_doubleword.  For LTU and GEU use
11178         sub<mode>3_carry_ccc instead of sub<mode>3_carry_ccgz and use CCCmode.
11180         * common.opt (gcolumn-info): Enable by default.
11181         * doc/invoke.texi (gcolumn-info): Document new default.
11183 2017-10-23  Richard Biener  <rguenther@suse.de>
11185         PR tree-optimization/82672
11186         * graphite-isl-ast-to-gimple.c (graphite_copy_stmts_from_block):
11187         Fold the stmt if we propagated into it.
11189 2017-10-23  Richard Biener  <rguenther@suse.de>
11191         * tree-ssa-pre.c (bitmap_remove_from_set): Rename to...
11192         (bitmap_remove_expr_from_set): ... this.  All callers call this
11193         for non-constant values.
11194         (bitmap_set_subtract): Rename to...
11195         (bitmap_set_subtract_expressions): ... this.  Adjust and
11196         optimize.
11197         (bitmap_set_contains_value): Remove superfluous check.
11198         (bitmap_set_replace_value): Inline into single caller ...
11199         (bitmap_value_replace_in_set): ... here and simplify.
11200         (dependent_clean): Merge into ...
11201         (clean): ... this using an overload.  Adjust.
11202         (prune_clobbered_mems): Adjust.
11203         (compute_antic_aux): Likewise.
11204         (compute_partial_antic_aux): Likewise.
11206 2017-10-23  Richard Biener  <rguenther@suse.de>
11208         PR tree-optimization/82129
11209         Revert
11210         2017-08-01  Richard Biener  <rguenther@suse.de>
11212         PR tree-optimization/81181
11213         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
11214         (compute_antic): ... end of iteration here.
11216 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11218         * target.def (starting_frame_offset): New hook.
11219         * doc/tm.texi (STARTING_FRAME_OFFSET): Remove in favor of...
11220         (TARGET_STARTING_FRAME_OFFSET): ...this new hook.
11221         * doc/tm.texi.in: Regenerate.
11222         * hooks.h (hook_hwi_void_0): Declare.
11223         * hooks.c (hook_hwi_void_0): New function.
11224         * doc/rtl.texi: Refer to TARGET_STARTING_FRAME_OFFSET instead of
11225         STARTING_FRAME_OFFSET.
11226         * builtins.c (expand_builtin_setjmp_receiver): Likewise.
11227         * reload1.c (reload): Likewise.
11228         * cfgexpand.c (expand_used_vars): Use targetm.starting_frame_offset
11229         instead of STARTING_FRAME_OFFSET.
11230         * function.c (try_fit_stack_local): Likewise.
11231         (assign_stack_local_1): Likewise
11232         (instantiate_virtual_regs): Likewise.
11233         * rtlanal.c (rtx_addr_can_trap_p_1): Likewise.
11234         * config/avr/avr.md (nonlocal_goto_receiver): Likewise.
11235         * config/aarch64/aarch64.h (STARTING_FRAME_OFFSET): Delete.
11236         * config/alpha/alpha.h (STARTING_FRAME_OFFSET): Likewise.
11237         * config/arc/arc.h (STARTING_FRAME_OFFSET): Likewise.
11238         * config/arm/arm.h (STARTING_FRAME_OFFSET): Likewise.
11239         * config/bfin/bfin.h (STARTING_FRAME_OFFSET): Likewise.
11240         * config/c6x/c6x.h (STARTING_FRAME_OFFSET): Likewise.
11241         * config/cr16/cr16.h (STARTING_FRAME_OFFSET): Likewise.
11242         * config/cris/cris.h (STARTING_FRAME_OFFSET): Likewise.
11243         * config/fr30/fr30.h (STARTING_FRAME_OFFSET): Likewise.
11244         * config/frv/frv.h (STARTING_FRAME_OFFSET): Likewise.
11245         * config/ft32/ft32.h (STARTING_FRAME_OFFSET): Likewise.
11246         * config/h8300/h8300.h (STARTING_FRAME_OFFSET): Likewise.
11247         * config/i386/i386.h (STARTING_FRAME_OFFSET): Likewise.
11248         * config/ia64/ia64.h (STARTING_FRAME_OFFSET): Likewise.
11249         * config/m32c/m32c.h (STARTING_FRAME_OFFSET): Likewise.
11250         * config/m68k/m68k.h (STARTING_FRAME_OFFSET): Likewise.
11251         * config/mcore/mcore.h (STARTING_FRAME_OFFSET): Likewise.
11252         * config/mn10300/mn10300.h (STARTING_FRAME_OFFSET): Likewise.
11253         * config/moxie/moxie.h (STARTING_FRAME_OFFSET): Likewise.
11254         * config/msp430/msp430.h (STARTING_FRAME_OFFSET): Likewise.
11255         * config/nds32/nds32.h (STARTING_FRAME_OFFSET): Likewise.
11256         * config/nios2/nios2.h (STARTING_FRAME_OFFSET): Likewise.
11257         * config/nvptx/nvptx.h (STARTING_FRAME_OFFSET): Likewise.
11258         * config/pdp11/pdp11.h (STARTING_FRAME_OFFSET): Likewise.
11259         * config/riscv/riscv.h (STARTING_FRAME_OFFSET): Likewise.
11260         * config/rl78/rl78.h (STARTING_FRAME_OFFSET): Likewise.
11261         * config/rx/rx.h (STARTING_FRAME_OFFSET): Likewise.
11262         * config/s390/s390.h (STARTING_FRAME_OFFSET): Likewise.
11263         * config/sh/sh.h (STARTING_FRAME_OFFSET): Likewise.
11264         * config/sparc/sparc.c (sparc_compute_frame_size): Likewise.
11265         * config/sparc/sparc.h (STARTING_FRAME_OFFSET): Likewise.
11266         * config/spu/spu.h (STARTING_FRAME_OFFSET): Likewise.
11267         * config/stormy16/stormy16.h (STARTING_FRAME_OFFSET): Likewise.
11268         * config/tilegx/tilegx.h (STARTING_FRAME_OFFSET): Likewise.
11269         * config/tilepro/tilepro.h (STARTING_FRAME_OFFSET): Likewise.
11270         * config/v850/v850.h (STARTING_FRAME_OFFSET): Likewise.
11271         * config/visium/visium.h (STARTING_FRAME_OFFSET): Likewise.
11272         * config/avr/avr.h (STARTING_FRAME_OFFSET): Likewise.
11273         * config/avr/avr-protos.h (avr_starting_frame_offset): Likewise.
11274         * config/avr/avr.c (avr_starting_frame_offset): Make static and
11275         return a HOST_WIDE_INT.
11276         (avr_builtin_setjmp_frame_value): Use it instead of
11277         STARTING_FRAME_OFFSET.
11278         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11279         * config/epiphany/epiphany.h (STARTING_FRAME_OFFSET): Delete.
11280         * config/epiphany/epiphany.c (epiphany_starting_frame_offset):
11281         New function.
11282         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11283         * config/iq2000/iq2000.h (STARTING_FRAME_OFFSET): Delete.
11284         * config/iq2000/iq2000.c (iq2000_starting_frame_offset): New function.
11285         (TARGET_CONSTANT_ALIGNMENT): Redefine.
11286         * config/lm32/lm32.h (STARTING_FRAME_OFFSET): Delete.
11287         * config/lm32/lm32.c (lm32_starting_frame_offset): New function.
11288         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11289         * config/m32r/m32r.h (STARTING_FRAME_OFFSET): Delete.
11290         * config/m32r/m32r.c (m32r_starting_frame_offset): New function.
11291         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11292         * config/microblaze/microblaze.h (STARTING_FRAME_OFFSET): Delete.
11293         * config/microblaze/microblaze.c (microblaze_starting_frame_offset):
11294         New function.
11295         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11296         * config/mips/mips.h (STARTING_FRAME_OFFSET): Delete.
11297         * config/mips/mips.c (mips_compute_frame_info): Refer to
11298         TARGET_STARTING_FRAME_OFFSET instead of STARTING_FRAME_OFFSET.
11299         (mips_starting_frame_offset): New function.
11300         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11301         * config/mmix/mmix.h (STARTING_FRAME_OFFSET): Delete.
11302         * config/mmix/mmix-protos.h (mmix_starting_frame_offset): Delete.
11303         * config/mmix/mmix.c (mmix_starting_frame_offset): Make static
11304         and return a HOST_WIDE_INT.
11305         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11306         (mmix_initial_elimination_offset): Refer to
11307         TARGET_STARTING_FRAME_OFFSET instead of STARTING_FRAME_OFFSET.
11308         * config/pa/pa.h (STARTING_FRAME_OFFSET): Delete.
11309         * config/pa/pa.c (pa_starting_frame_offset): New function.
11310         (pa_compute_frame_size): Use it instead of STARTING_FRAME_OFFSET.
11311         (pa_expand_prologue): Likewise.
11312         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11313         * config/powerpcspe/aix.h (STARTING_FRAME_OFFSET): Split out
11314         !FRAME_GROWS_DOWNWARD handling to...
11315         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11316         * config/powerpcspe/darwin.h (STARTING_FRAME_OFFSET): Split out
11317         !FRAME_GROWS_DOWNWARD handling to...
11318         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11319         * config/powerpcspe/powerpcspe.h (STARTING_FRAME_OFFSET): Split out
11320         !FRAME_GROWS_DOWNWARD handling to...
11321         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11322         * config/powerpcspe/powerpcspe.c (TARGET_STARTING_FRAME_OFFSET):
11323         Redefine.
11324         (rs6000_starting_frame_offset): New function.
11325         * config/rs6000/aix.h (STARTING_FRAME_OFFSET): Split out
11326         !FRAME_GROWS_DOWNWARD handling to...
11327         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11328         * config/rs6000/darwin.h (STARTING_FRAME_OFFSET): Split out
11329         !FRAME_GROWS_DOWNWARD handling to...
11330         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11331         * config/rs6000/rs6000.h (STARTING_FRAME_OFFSET): Split out
11332         !FRAME_GROWS_DOWNWARD handling to...
11333         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11334         * config/rs6000/rs6000.c (TARGET_STARTING_FRAME_OFFSET): Refine.
11335         (rs6000_starting_frame_offset): New function.
11336         * config/vax/elf.h (STARTING_FRAME_OFFSET): Delete.
11337         * config/vax/vax.h (STARTING_FRAME_OFFSET): Delete.
11338         * config/vax/vax.c (vax_starting_frame_offset): New function.
11339         (vax_expand_prologue): Use it instead of STARTING_FRAME_OFFSET.
11340         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11341         * config/xtensa/xtensa.h (STARTING_FRAME_OFFSET): Delete.
11342         * config/xtensa/xtensa.c (xtensa_starting_frame_offset): New function.
11343         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11344         * system.h (STARTING_FRAME_OFFSET): Poison.
11346 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11348         * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
11349         SCALAR_TYPE_MODE instead of TYPE_MODE.
11351 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11352             Alan Hayward  <alan.hayward@arm.com>
11353             David Sherwood  <david.sherwood@arm.com>
11355         * dwarf2out.c (loc_list_from_tree_1): Use SCALAR_INT_TYPE_MODE
11357 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11358             Alan Hayward  <alan.hayward@arm.com>
11359             David Sherwood  <david.sherwood@arm.com>
11361         * expmed.c (expand_shift_1): Use scalar_mode for scalar_mode.
11363 2017-10-23  Richard Biener  <rguenther@suse.de>
11365         PR tree-optimization/82129
11366         * tree-ssa-pre.c (bitmap_set_and): Remove.
11367         (compute_antic_aux): Compute ANTIC_OUT intersection in a way
11368         canonicalizing expressions in the set to those with lowest
11369         ID rather than taking that from the first edge.
11371 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11373         * combine.c (rtx_equal_for_field_assignment_p): Use
11374         byte_lowpart_offset.
11376 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11377             Alan Hayward  <alan.hayward@arm.com>
11378             David Sherwood  <david.sherwood@arm.com>
11380         * internal-fn.c (expand_direct_optab_fn): Don't assign directly
11381         to a SUBREG_PROMOTED_VAR.
11383 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11384             Alan Hayward  <alan.hayward@arm.com>
11385             David Sherwood  <david.sherwood@arm.com>
11387         * cfgexpand.c (expand_debug_expr): Use GET_MODE_UNIT_PRECISION.
11388         (expand_debug_source_expr): Likewise.
11389         * combine.c (combine_simplify_rtx): Likewise.
11390         * cse.c (fold_rtx): Likewise.
11391         * optabs.c (expand_float): Likewise.
11392         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11393         (simplify_binary_operation_1): Likewise.
11395 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11396             Alan Hayward  <alan.hayward@arm.com>
11397             David Sherwood  <david.sherwood@arm.com>
11399         * combine.c (simplify_comparison): Use HWI_COMPUTABLE_MODE_P.
11400         (record_promoted_value): Likewise.
11401         * expr.c (expand_expr_real_2): Likewise.
11402         * ree.c (update_reg_equal_equiv_notes): Likewise.
11403         (combine_set_extension): Likewise.
11404         * rtlanal.c (low_bitmask_len): Likewise.
11405         * simplify-rtx.c (neg_const_int): Likewise.
11406         (simplify_binary_operation_1): Likewise.
11408 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11409             Alan Hayward  <alan.hayward@arm.com>
11410             David Sherwood  <david.sherwood@arm.com>
11412         * lra-spills.c (assign_mem_slot): Use subreg_size_lowpart_offset.
11413         * regcprop.c (maybe_mode_change): Likewise.
11414         * reload1.c (alter_reg): Likewise.
11416 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11418         * inchash.h (inchash::hash::add_wide_int): New function.
11419         * lto-streamer-out.c (hash_tree): Use it.
11421 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11423         * inchash.h (inchash::hash::add_wide_int): Rename to...
11424         (inchash::hash::add_hwi): ...this.
11425         * ipa-devirt.c (hash_odr_vtable): Update accordingly.
11426         (polymorphic_call_target_hasher::hash): Likewise.
11427         * ipa-icf.c (sem_function::get_hash, sem_function::init): Likewise.
11428         (sem_item::add_expr, sem_item::add_type, sem_variable::get_hash)
11429         (sem_item_optimizer::update_hash_by_addr_refs): Likewise.
11430         * lto-streamer-out.c (hash_tree): Likewise.
11431         * optc-save-gen.awk: Likewise.
11432         * tree.c (add_expr): Likewise.
11434 2017-10-22  Uros Bizjak  <ubizjak@gmail.com>
11436         PR target/52451
11437         * config/i386/i386.c (ix86_fp_compare_mode): Return CCFPmode
11438         for ordered inequality comparisons even with TARGET_IEEE_FP.
11440 2017-10-22  Uros Bizjak  <ubizjak@gmail.com>
11442         PR target/82628
11443         * config/i386/i386.md (cmp<dwi>_doubleword): New pattern.
11444         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>:
11445         Expand with cmp<dwi>_doubleword.
11447 2017-10-21  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11449         * extend.texi: Add x86 specific to 'nocf_check' attribute.
11450         List CET intrinsics.
11451         * invoke.texi: Add -mcet, -mibt, -mshstk options.  Add x86
11452         specific to -fcf-protection option.
11454 2017-10-21  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11456         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET): New.
11457         (OPTION_MASK_ISA_SHSTK_SET): Likewise.
11458         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
11459         (OPTION_MASK_ISA_SHSTK_UNSET): Likewise.
11460         (ix86_handle_option): Add -mibt, -mshstk, -mcet handling.
11461         * config.gcc (extra_headers): Add cetintrin.h for x86 targets.
11462         (extra_objs): Add cet.o for Linux/x86 targets.
11463         (tmake_file): Add i386/t-cet for Linux/x86 targets.
11464         * config/i386/cet.c: New file.
11465         * config/i386/cetintrin.h: Likewise.
11466         * config/i386/t-cet: Likewise.
11467         * config/i386/cpuid.h (bit_SHSTK): New.
11468         (bit_IBT): Likewise.
11469         * config/i386/driver-i386.c (host_detect_local_cpu): Detect and
11470         pass IBT and SHSTK bits.
11471         * config/i386/i386-builtin-types.def
11472         (VOID_FTYPE_UNSIGNED_PVOID): New.
11473         (VOID_FTYPE_UINT64_PVOID): Likewise.
11474         * config/i386/i386-builtin.def: Add CET intrinsics.
11475         * config/i386/i386-c.c (ix86_target_macros_internal): Add
11476         OPTION_MASK_ISA_IBT, OPTION_MASK_ISA_SHSTK handling.
11477         * config/i386/i386-passes.def: Add pass_insert_endbranch pass.
11478         * config/i386/i386-protos.h (make_pass_insert_endbranch): New
11479         prototype.
11480         * config/i386/i386.c (rest_of_insert_endbranch): New.
11481         (pass_data_insert_endbranch): Likewise.
11482         (pass_insert_endbranch): Likewise.
11483         (make_pass_insert_endbranch): Likewise.
11484         (ix86_notrack_prefixed_insn_p): Likewise.
11485         (ix86_target_string): Add -mibt, -mshstk flags.
11486         (ix86_option_override_internal): Add flag_cf_protection
11487         processing.
11488         (ix86_valid_target_attribute_inner_p): Set OPT_mibt, OPT_mshstk.
11489         (ix86_print_operand): Add 'notrack' prefix output.
11490         (ix86_init_mmx_sse_builtins): Add CET intrinsics.
11491         (ix86_expand_builtin): Expand CET intrinsics.
11492         (x86_output_mi_thunk): Add 'endbranch' instruction.
11493         * config/i386/i386.h (TARGET_IBT): New.
11494         (TARGET_IBT_P): Likewise.
11495         (TARGET_SHSTK): Likewise.
11496         (TARGET_SHSTK_P): Likewise.
11497         * config/i386/i386.md (unspecv): Add UNSPECV_NOP_RDSSP,
11498         UNSPECV_INCSSP, UNSPECV_SAVEPREVSSP, UNSPECV_RSTORSSP,
11499         UNSPECV_WRSS, UNSPECV_WRUSS, UNSPECV_SETSSBSY, UNSPECV_CLRSSBSY.
11500         (builtin_setjmp_setup): New pattern.
11501         (builtin_longjmp): Likewise.
11502         (rdssp<mode>): Likewise.
11503         (incssp<mode>): Likewise.
11504         (saveprevssp): Likewise.
11505         (rstorssp): Likewise.
11506         (wrss<mode>): Likewise.
11507         (wruss<mode>): Likewise.
11508         (setssbsy): Likewise.
11509         (clrssbsy): Likewise.
11510         (nop_endbr): Likewise.
11511         * config/i386/i386.opt: Add -mcet, -mibt, -mshstk and -mcet-switch
11512         options.
11513         * config/i386/immintrin.h: Include <cetintrin.h>.
11514         * config/i386/linux-common.h
11515         (file_end_indicate_exec_stack_and_cet): New prototype.
11516         (TARGET_ASM_FILE_END): New.
11518 2017-10-20  Jan Hubicka  <hubicka@ucw.cz>
11520         * i386.c (ix86_builtin_vectorization_cost): Use existing rtx_cost
11521         latencies instead of having separate table; make difference between
11522         integer and float costs.
11523         * i386.h (processor_costs): Remove scalar_stmt_cost,
11524         scalar_load_cost, scalar_store_cost, vec_stmt_cost, vec_to_scalar_cost,
11525         scalar_to_vec_cost, vec_align_load_cost, vec_unalign_load_cost,
11526         vec_store_cost.
11527         * x86-tune-costs.h: Remove entries which has been removed in
11528         procesor_costs from all tables; make cond_taken_branch_cost
11529         and cond_not_taken_branch_cost COST_N_INSNS based.
11531 2017-10-20  Jan Hubicka  <hubicka@ucw.cz>
11533         * x86-tune-costs.h (intel_cost, generic_cost): Fix move costs.
11535 2017-10-20  Jakub Jelinek  <jakub@redhat.com>
11537         * config/i386/i386.md (isa): Remove fma_avx512f.
11538         * config/i386/sse.md (<avx512>_fmadd_<mode>_mask<round_name>,
11539         <avx512>_fmadd_<mode>_mask3<round_name>,
11540         <avx512>_fmsub_<mode>_mask<round_name>,
11541         <avx512>_fmsub_<mode>_mask3<round_name>,
11542         <avx512>_fnmadd_<mode>_mask<round_name>,
11543         <avx512>_fnmadd_<mode>_mask3<round_name>,
11544         <avx512>_fnmsub_<mode>_mask<round_name>,
11545         <avx512>_fnmsub_<mode>_mask3<round_name>,
11546         <avx512>_fmaddsub_<mode>_mask<round_name>,
11547         <avx512>_fmaddsub_<mode>_mask3<round_name>,
11548         <avx512>_fmsubadd_<mode>_mask<round_name>,
11549         <avx512>_fmsubadd_<mode>_mask3<round_name>): Remove isa attribute.
11550         (*vec_widen_umult_even_v16si<mask_name>,
11551         *vec_widen_smult_even_v16si<mask_name>): Likewise.
11552         (<mask_codefor>avx512bw_dbpsadbw<mode><mask_name>): Likewise.
11554 2017-10-20  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11556         * extend.texi: Add 'nocf_check' documentation.
11557         * gimple.texi: Add second parameter to
11558         gimple_build_call_from_tree.
11559         * invoke.texi: Add -fcf-protection documentation.
11560         * rtl.texi: Add REG_CALL_NOTRACK documenation.
11562 2017-10-20  Richard Biener  <rguenther@suse.de>
11564         PR tree-optimization/82473
11565         * tree-vect-loop.c (vectorizable_reduction): Properly get at
11566         the largest input type.
11568 2017-10-20  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11570         * c-attribs.c (handle_nocf_check_attribute): New function.
11571         (c_common_attribute_table): Add 'nocf_check' handling.
11572         * gimple-parser.c: Add second argument NULL to
11573         gimple_build_call_from_tree.
11574         * attrib.c (comp_type_attributes): Check nocf_check attribute.
11575         * cfgexpand.c (expand_call_stmt): Set REG_CALL_NOCF_CHECK for
11576         call insn.
11577         * combine.c (distribute_notes): Add REG_CALL_NOCF_CHECK handling.
11578         * common.opt: Add fcf-protection flag.
11579         * emit-rtl.c (try_split): Add REG_CALL_NOCF_CHECK handling.
11580         * flag-types.h: Add enum cf_protection_level.
11581         * gimple.c (gimple_build_call_from_tree): Add second parameter.
11582         Add 'nocf_check' attribute propagation to gimple call.
11583         * gimple.h (gf_mask): Add GF_CALL_NOCF_CHECK.
11584         (gimple_build_call_from_tree): Update prototype.
11585         (gimple_call_nocf_check_p): New function.
11586         (gimple_call_set_nocf_check): Likewise.
11587         * gimplify.c: Add second argument to gimple_build_call_from_tree.
11588         * ipa-icf.c: Add nocf_check attribute in statement hash.
11589         * recog.c (peep2_attempt): Add REG_CALL_NOCF_CHECK handling.
11590         * reg-notes.def: Add REG_NOTE (CALL_NOCF_CHECK).
11591         * toplev.c (process_options): Add flag_cf_protection handling.
11593 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11595         * x86-tune-costs.h (core_cost): Fix div, move and sqrt latencies.
11597 2017-10-20  Richard Biener  <rguenther@suse.de>
11599         PR tree-optimization/82603
11600         * tree-if-conv.c (predicate_mem_writes): Make sure to only
11601         remove false predicated stores.
11603 2017-10-20  Richard Biener  <rguenther@suse.de>
11605         * graphite-isl-ast-to-gimple.c
11606         (translate_isl_ast_to_gimple::graphite_copy_stmts_from_block):
11607         Remove return value and simplify, dump copied stmt after lhs
11608         adjustment.
11609         (translate_isl_ast_to_gimple::translate_isl_ast_node_user):
11610         Reduce dump verbosity.
11611         (gsi_insert_earliest): Likewise.
11612         (translate_isl_ast_to_gimple::copy_bb_and_scalar_dependences): Adjust.
11613         * graphite.c (print_global_statistics): Adjust dumping.
11614         (print_graphite_scop_statistics): Likewise.
11615         (print_graphite_statistics): Do not dump loops here.
11616         (graphite_transform_loops): But here.
11618 2017-10-20  Nicolas Roche  <roche@adacore.com>
11620         * configure.ac (ACX_PROG_GNAT): Append "libgnat" to include search dir.
11621         * configure: Regenerate.
11623 2017-10-20  Jakub Jelinek  <jakub@redhat.com>
11625         PR target/82158
11626         * tree-cfg.c (pass_warn_function_return::execute): In noreturn
11627         functions when optimizing replace GIMPLE_RETURN stmts with
11628         calls to __builtin_unreachable ().
11630         PR sanitizer/82595
11631         * config/gnu-user.h (LIBTSAN_EARLY_SPEC): Add libtsan_preinit.o
11632         for -fsanitize=thread link of executables.
11633         (LIBLSAN_EARLY_SPEC): Add liblsan_preinit.o for -fsanitize=leak
11634         link of executables.
11636         PR target/82370
11637         * config/i386/sse.md (VI248_AVX2, VI248_AVX512BW, VI248_AVX512BW_2):
11638         New mode iterators.
11639         (<shift_insn><mode>3<mask_name>): Change the last of the 3
11640         define_insns for logical vector shifts to use VI248_AVX512BW
11641         iterator instead of VI48_AVX512, remove <mask_mode512bit_condition>
11642         condition, useless isa and prefix attributes.  Change the first
11643         2 of these define_insns to ...
11644         (<mask_codefor><shift_insn><mode>3<mask_name>): ... this, new
11645         define_insn for avx512vl.
11646         (<shift_insn><mode>3): ... and this, new define_insn without
11647         masking for non-avx512vl.
11649         PR target/82370
11650         * config/i386/sse.md (*andnot<mode>3,
11651         <mask_codefor><code><mode>3<mask_name>, *<code><mode>3): Split
11652         (=v,v,vm) alternative into (=x,x,xm) and (=v,v,vm), for 128-bit
11653         and 256-bit vectors, the (=x,x,xm) alternative and when mask is
11654         not applied use empty suffix even for TARGET_AVX512VL.
11655         * config/i386/subst.md (mask_prefix3, mask_prefix4): When mask
11656         is applied, supply evex,evex or evex,evex,evex instead of just
11657         evex.
11659 2017-10-20  Julia Koval  <julia.koval@intel.com>
11661         * common/config/i386/i386-common.c (OPTION_MASK_ISA_GFNI_SET,
11662         (OPTION_MASK_ISA_GFNI_UNSET): New.
11663         (ix86_handle_option): Handle OPT_mgfni.
11664         * config/i386/cpuid.h (bit_GFNI): New.
11665         * config/i386/driver-i386.c (host_detect_local_cpu): Detect gfni.
11666         * config/i386/i386-c.c (ix86_target_macros_internal): Define __GFNI__.
11667         * config/i386/i386.c (ix86_target_string): Add -mgfni.
11668         (ix86_valid_target_attribute_inner_p): Add OPT_mgfni.
11669         * config/i386/i386.h (TARGET_GFNI, TARGET_GFNI_P): New.
11670         * config/i386/i386.opt: Add mgfni.
11672 2017-10-20  Orlando Arias  <oarias@knights.ucf.edu>
11674         * config/msp430/msp430.c (msp430_option_override): Disable
11675         -fdelete-null-pointer-checks.
11676         * doc/invoke.text (-fdelete-null-pointer-checks): Document that.
11678 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11680         * x86-tune-costs.h (generic_cost, core_cost): Correct costs
11681         of x87 and SSE instructions.
11683 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11685         * asan.c (create_cond_insert_point): Do not update edge count.
11686         * auto-profile.c (afdo_propagate_edge): Update for edge count removal.
11687         (afdo_propagate_circuit): Likewise.
11688         (afdo_calculate_branch_prob): Likewise.
11689         (afdo_annotate_cfg): Likewise.
11690         * basic-block.h (struct edge_def): Remove count.
11691         (edge_def::count): New accessor.
11692         * bb-reorder.c (rotate_loop): Update.
11693         (find_traces_1_round): Update.
11694         (connect_traces): Update.
11695         (sanitize_hot_paths): Update.
11696         * cfg.c (unchecked_make_edge): Update.
11697         (make_single_succ_edge): Update.
11698         (check_bb_profile): Update.
11699         (dump_edge_info): Update.
11700         (update_bb_profile_for_threading): Update.
11701         (scale_bbs_frequencies_int): Update.
11702         (scale_bbs_frequencies_gcov_type): Update.
11703         (scale_bbs_frequencies_profile_count): Update.
11704         (scale_bbs_frequencies): Update.
11705         * cfganal.c (connect_infinite_loops_to_exit): Update.
11706         * cfgbuild.c (compute_outgoing_frequencies): Update.
11707         (find_many_sub_basic_blocks): Update.
11708         * cfgcleanup.c (try_forward_edges): Update.
11709         (try_crossjump_to_edge): Update
11710         * cfgexpand.c (expand_gimple_cond): Update
11711         (expand_gimple_tailcall): Update
11712         (construct_exit_block): Update
11713         * cfghooks.c (verify_flow_info): Update
11714         (redirect_edge_succ_nodup): Update
11715         (split_edge): Update
11716         (make_forwarder_block): Update
11717         (duplicate_block): Update
11718         (account_profile_record): Update
11719         * cfgloop.c (find_subloop_latch_edge_by_profile): Update.
11720         * cfgloopanal.c (expected_loop_iterations_unbounded): Update.
11721         * cfgloopmanip.c (scale_loop_profile): Update.
11722         (loopify): Update.
11723         (lv_adjust_loop_entry_edge): Update.
11724         * cfgrtl.c (try_redirect_by_replacing_jump): Update.
11725         (force_nonfallthru_and_redirect): Update.
11726         (purge_dead_edges): Update.
11727         (rtl_flow_call_edges_add): Update.
11728         * cgraphunit.c (init_lowered_empty_function): Update.
11729         (cgraph_node::expand_thunk): Update.
11730         * gimple-pretty-print.c (dump_probability): Update.
11731         (dump_edge_probability): Update.
11732         * gimple-ssa-isolate-paths.c (isolate_path): Update.
11733         * haifa-sched.c (sched_create_recovery_edges): Update.
11734         * hsa-gen.c (convert_switch_statements): Update.
11735         * ifcvt.c (dead_or_predicable): Update.
11736         * ipa-inline-transform.c (inline_transform): Update.
11737         * ipa-split.c (split_function): Update.
11738         * ipa-utils.c (ipa_merge_profiles): Update.
11739         * loop-doloop.c (add_test): Update.
11740         * loop-unroll.c (unroll_loop_runtime_iterations): Update.
11741         * lto-streamer-in.c (input_cfg): Update.
11742         (input_function): Update.
11743         * lto-streamer-out.c (output_cfg): Update.
11744         * modulo-sched.c (sms_schedule): Update.
11745         * postreload-gcse.c (eliminate_partially_redundant_load): Update.
11746         * predict.c (maybe_hot_edge_p): Update.
11747         (unlikely_executed_edge_p): Update.
11748         (probably_never_executed_edge_p): Update.
11749         (dump_prediction): Update.
11750         (drop_profile): Update.
11751         (propagate_unlikely_bbs_forward): Update.
11752         (determine_unlikely_bbs): Update.
11753         (force_edge_cold): Update.
11754         * profile.c (compute_branch_probabilities): Update.
11755         * reg-stack.c (better_edge): Update.
11756         * shrink-wrap.c (handle_simple_exit): Update.
11757         * tracer.c (better_p): Update.
11758         * trans-mem.c (expand_transaction): Update.
11759         (split_bb_make_tm_edge): Update.
11760         * tree-call-cdce.c: Update.
11761         * tree-cfg.c (gimple_find_sub_bbs): Update.
11762         (gimple_split_edge): Update.
11763         (gimple_duplicate_sese_region): Update.
11764         (gimple_duplicate_sese_tail): Update.
11765         (gimple_flow_call_edges_add): Update.
11766         (insert_cond_bb): Update.
11767         (execute_fixup_cfg): Update.
11768         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update.
11769         * tree-complex.c (expand_complex_div_wide): Update.
11770         * tree-eh.c (lower_resx): Update.
11771         (unsplit_eh): Update.
11772         (cleanup_empty_eh_move_lp): Update.
11773         * tree-inline.c (copy_edges_for_bb): Update.
11774         (freqs_to_counts): Update.
11775         (copy_cfg_body): Update.
11776         * tree-ssa-dce.c (remove_dead_stmt): Update.
11777         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
11778         * tree-ssa-loop-im.c (execute_sm_if_changed): Update.
11779         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update.
11780         (unloop_loops): Update.
11781         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
11782         * tree-ssa-loop-split.c (connect_loops): Update.
11783         (split_loop): Update.
11784         * tree-ssa-loop-unswitch.c (hoist_guard): Update.
11785         * tree-ssa-phionlycprop.c (propagate_rhs_into_lhs): Update.
11786         * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Update.
11787         * tree-ssa-reassoc.c (branch_fixup): Update.
11788         * tree-ssa-tail-merge.c (replace_block_by): Update.
11789         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges): Update.
11790         (compute_path_counts): Update.
11791         (update_profile): Update.
11792         (recompute_probabilities): Update.
11793         (update_joiner_offpath_counts): Update.
11794         (estimated_freqs_path): Update.
11795         (freqs_to_counts_path): Update.
11796         (clear_counts_path): Update.
11797         (ssa_fix_duplicate_block_edges): Update.
11798         (duplicate_thread_path): Update.
11799         * tree-switch-conversion.c (hoist_edge_and_branch_if_true): Update.
11800         (case_bit_test_cmp): Update.
11801         (collect_switch_conv_info): Update.
11802         (gen_inbound_check): Update.
11803         (do_jump_if_equal): Update.
11804         (emit_cmp_and_jump_insns): Update.
11805         * tree-tailcall.c (decrease_profile): Update.
11806         (eliminate_tail_call): Update.
11807         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
11808         (vect_do_peeling): Update.
11809         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
11810         * ubsan.c (ubsan_expand_null_ifn): Update.
11811         (ubsan_expand_ptr_ifn): Update.
11812         * value-prof.c (gimple_divmod_fixed_value): Update.
11813         (gimple_mod_pow2): Update.
11814         (gimple_mod_subtract): Update.
11815         (gimple_ic): Update.
11816         (gimple_stringop_fixed_value): Update.
11818 2017-10-19  Uros Bizjak  <ubizjak@gmail.com>
11820         PR target/82618
11821         * config/i386/i386.md (sub to cmp): New peephole2 pattern.
11823 2017-10-19  Alexander Monakov  <amonakov@ispras.ru>
11825         PR rtl-optimization/82395
11826         * ira-color.c (allocno_priority_compare_func): Fix comparison step
11827         based on non_spilled_static_chain_regno_p.
11829 2017-10-19  Uros Bizjak  <ubizjak@gmail.com>
11831         * config/i386/i386.c (output_387_binary_op): Rewrite SSE part.
11832         (ix86_emit_mode_set): Rewrite insn mnemonic construction.
11833         (ix86_prepare_fp_compare_args): Redefine is_sse as bool.
11835 2017-10-19  Martin Sebor  <msebor@redhat.com>
11837         PR tree-optimization/82596
11838         * tree.c (array_at_struct_end_p): Handle STRING_CST.
11840 2017-10-19  Eric Botcazou  <ebotcazou@adacore.com>
11842         * asan.c (handle_builtin_alloca): Deal with all alloca variants.
11843         (get_mem_refs_of_builtin_call): Likewise.
11844         * builtins.c (expand_builtin_apply): Adjust call to
11845         allocate_dynamic_stack_space.
11846         (expand_builtin_alloca): For __builtin_alloca_with_align_and_max, pass
11847         the third argument to allocate_dynamic_stack_space, otherwise -1.
11848         (expand_builtin): Deal with all alloca variants.
11849         (is_inexpensive_builtin): Likewise.
11850         * builtins.def (BUILT_IN_ALLOCA_WITH_ALIGN_AND_MAX): New.
11851         * calls.c (special_function_p): Deal with all alloca variants.
11852         (initialize_argument_information): Adjust call to
11853         allocate_dynamic_stack_space.
11854         (expand_call): Likewise.
11855         * cfgexpand.c (expand_call_stmt): Deal with all alloca variants.
11856         * doc/extend.texi (Built-ins): Add __builtin_alloca_with_align_and_max
11857         * explow.c (allocate_dynamic_stack_space): Add MAX_SIZE parameter and
11858         use it for the stack usage computation.
11859         * explow.h (allocate_dynamic_stack_space): Adjust prototype.
11860         * function.c (gimplify_parameters): Call build_alloca_call_expr.
11861         * gimple-ssa-warn-alloca.c (alloca_call_type): Simplify control flow.
11862         Take into account 3rd argument of __builtin_alloca_with_align_and_max.
11863         (in_loop_p): Remove first argument and useless check.
11864         (pass_walloca::execute): Remove useless test and adjust call to above.
11865         * gimple.c (gimple_build_call_from_tree): Deal with all alloc variants
11866         * gimplify.c (gimplify_vla_decl): Call build_alloca_call_expr.
11867         (gimplify_call_expr): Deal with all alloca variants.
11868         * hsa-gen.c (gen_hsa_alloca): Likewise.
11869         (gen_hsa_insns_for_call): Likewise.
11870         * ipa-pure-const.c (special_builtin_state): Likewise.
11871         * tree-chkp.c (chkp_build_returned_bound): Likewise.
11872         * tree-object-size.c (alloc_object_size): Likewise.
11873         * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
11874         (call_may_clobber_ref_p_1): Likewise.
11875         * tree-ssa-ccp.c (evaluate_stmt): Likewise.
11876         (ccp_fold_stmt): Likewise.
11877         (optimize_stack_restore): Likewise.
11878         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
11879         (mark_all_reaching_defs_necessary_1): Likewise.
11880         (propagate_necessity): Likewise.
11881         (eliminate_unnecessary_stmts): Likewise.
11882         * tree.c (build_common_builtin_nodes): Build
11883         BUILT_IN_ALLOCA_WITH_ALIGN_AND_MAX.
11884         (build_alloca_call_expr): New function.
11885         * tree.h (ALLOCA_FUNCTION_CODE_P): New macro.
11886         (CASE_BUILT_IN_ALLOCA): Likewise.
11887         (build_alloca_call_expr): Declare.
11888         * varasm.c (incorporeal_function_p): Deal with all alloca variants.
11890 2017-10-19  Eric Botcazou  <ebotcazou@adacore.com>
11892         PR debug/82509
11893         * dwarf2out.c (new_die_raw): New static inline function.
11894         (new_die): Use it to create the DIE.
11895         (add_AT_external_die_ref): Likewise.
11896         (clone_die): Likewise.
11897         (clone_as_declaration): Likewise.
11898         (dwarf2out_vms_debug_main_pointer): Likewise.
11899         (base_type_die): Likewise.  Remove early return for corner cases.
11900         Do not call add_pubtype on the DIE here.
11901         (is_base_type): Remove ERROR_MARK and return 0 for VOID_TYPE.
11902         (modified_type_die): Adjust the lookup for reverse order DIEs.  Skip
11903         typedefs for base types with DW_AT_endianity.  Make sure a DIE with
11904         native order exists for base types, attach the DIE manually and call
11905         add_pubtype on it.  Do not equate a reverse order DIE to the type.
11907 2017-10-19  Richard Earnshaw  <rearnsha@arm.com>
11909         * config/arm/arm.c (align_ok_ldrd_strd): New function.
11910         (mem_ok_for_ldrd_strd): New parameter align.  Extract the alignment of
11911         the mem into it.
11912         (gen_operands_ldrd_strd): Validate the alignment of the accesses.
11914 2017-10-19  Jakub Jelinek  <jakub@redhat.com>
11916         * flag-types.h (enum sanitize_code): Add SANITIZE_BUILTIN.  Or
11917         SANITIZE_BUILTIN into SANITIZE_UNDEFINED.
11918         * sanitizer.def (BUILT_IN_UBSAN_HANDLE_INVALID_BUILTIN,
11919         BUILT_IN_UBSAN_HANDLE_INVALID_BUILTIN_ABORT): New builtins.
11920         * opts.c (sanitizer_opts): Add builtin.
11921         * ubsan.c (instrument_builtin): New function.
11922         (pass_ubsan::execute): Call it.
11923         (pass_ubsan::gate): Enable even for SANITIZE_BUILTIN.
11924         * doc/invoke.texi: Document -fsanitize=builtin.
11926         * ubsan.c (ubsan_expand_null_ifn): Use _v1 suffixed type mismatch
11927         builtins, store max (log2 (align), 0) into uchar field instead of
11928         align into uptr field.
11929         (ubsan_expand_objsize_ifn): Use _v1 suffixed type mismatch builtins,
11930         store uchar 0 field instead of uptr 0 field.
11931         (instrument_nonnull_return): Use _v1 suffixed nonnull return builtin,
11932         instead of passing one address of struct with 2 locations pass
11933         two addresses of structs with 1 location each.
11934         * sanitizer.def (BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH,
11935         BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_ABORT,
11936         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN,
11937         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_ABORT): Removed.
11938         (BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_V1,
11939         BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_V1_ABORT,
11940         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_V1,
11941         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_V1_ABORT): New builtins.
11943 2017-10-19  Martin Liska  <mliska@suse.cz>
11945         PR driver/81829
11946         * file-find.c (remove_prefix): Remove.
11947         * file-find.h (remove_prefix): Likewise.
11948         * gcc-ar.c: Remove smartness of lookup.
11950 2017-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
11952         * config/rs6000/rs6000.md (*call_indirect_aix<mode>,
11953         *call_value_indirect_aix<mode>, *call_indirect_elfv2<mode>,
11954         *call_value_indirect_elfv2<mode>): Add correct mode to the unspec.
11956 2017-10-19  Jakub Jelinek  <jakub@redhat.com>
11958         PR target/82580
11959         * config/i386/i386.md (setcc + movzbl to xor + setcc): New peephole2.
11960         (setcc + and to xor + setcc): New peephole2.
11962 2017-10-19  Tom de Vries  <tom@codesourcery.com>
11964         * doc/sourcebuild.texi (Test Directives, Variants of
11965         dg-require-support): Add dg-require-stack-size.
11967 2017-10-19  Martin Liska  <mliska@suse.cz>
11969         PR sanitizer/82517
11970         * gimplify.c (gimplify_decl_expr): Do not instrument variables
11971         that have a large alignment.
11972         (gimplify_target_expr): Likewise.
11974 2017-10-18  Segher Boessenkool  <segher@kernel.crashing.org>
11976         PR rtl-optimization/82602
11977         * ira.c (rtx_moveable_p): Return false for volatile asm.
11979 2017-10-18  Uros Bizjak  <ubizjak@gmail.com>
11981         PR target/82580
11982         * config/i386/i386-modes.def (CCGZ): New CC mode.
11983         * config/i386/i386.md (sub<mode>3_carry_ccgz): New insn pattern.
11984         * config/i386/predicates.md (ix86_comparison_operator):
11985         Handle CCGZmode.
11986         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>:
11987         Emulate LE, LEU, GT, GTU, LT, LTU, GE and GEU double-word comparisons
11988         with double-word subtraction.
11989         (put_condition_code): Handle CCGZmode.
11991 2017-10-18  Aldy Hernandez  <aldyh@redhat.com>
11993         * wide-int.cc (debug (const wide_int &)): New.
11994         (debug (const wide_int *)): New.
11995         (debug (const widest_int &)): New.
11996         (debug (const widest_int *)): New.
11998 2017-10-18  Vladimir Makarov  <vmakarov@redhat.com>
12000         PR middle-end/82556
12001         * lra-constraints.c (curr_insn_transform): Use non-input operand
12002         instead of output one for matched reload.
12004 2017-10-18  Bin Cheng  <bin.cheng@arm.com>
12006         * tree-loop-distribution.c (INCLUDE_ALGORITHM): New header file.
12007         (tree-ssa-loop-ivopts.h): New header file.
12008         (struct builtin_info): New fields.
12009         (classify_builtin_1): Compute and record base and offset parts for
12010         memset builtin partition by calling strip_offset.
12011         (offset_cmp, fuse_memset_builtins): New functions.
12012         (finalize_partitions): Fuse adjacent memset partitions by calling
12013         above function.
12014         * tree-ssa-loop-ivopts.c (strip_offset): Delete static declaration.
12015         Expose the interface.
12016         * tree-ssa-loop-ivopts.h (strip_offset): New declaration.
12018 2017-10-18  Bin Cheng  <bin.cheng@arm.com>
12020         PR tree-optimization/82574
12021         * tree-loop-distribution.c (find_single_drs): New parameter.  Check
12022         that data reference must be executed exactly once per iteration
12023         against the outermost loop in nest.
12024         (classify_partition): Update call to above function.
12026 2017-10-18  Richard Biener  <rguenther@suse.de>
12028         PR tree-optimization/82591
12029         * graphite.c (graphite_transform_loops): Move code gen message
12030         printing ...
12031         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
12032         Here.  Handle scop_to_isl_ast failing.
12033         (scop_to_isl_ast): Limit the number of ISL operations.
12035 2017-10-18  Richard Biener  <rguenther@suse.de>
12037         * graphite-isl-ast-to-gimple.c
12038         (translate_isl_ast_to_gimple::set_rename): Simplify.
12039         (translate_isl_ast_to_gimple::set_rename_for_each_def): Inline...
12040         (graphite_copy_stmts_from_block): ... here.
12041         (copy_bb_and_scalar_dependences): Simplify.
12042         (add_parameters_to_ivs_params): Canonicalize.
12043         (generate_entry_out_of_ssa_copies): Simplify.
12044         * graphite-sese-to-poly.c (extract_affine_name): Simplify
12045         by passing in ISL dimension.
12046         (parameter_index_in_region_1): Rename to ...
12047         (parameter_index_in_region): ... this.
12048         (extract_affine): Adjust assert, pass down parameter index.
12049         (add_param_constraints): Use range-info when available.
12050         (build_scop_context): Adjust.
12051         * sese.c (new_sese_info): Adjust.
12052         (free_sese_info): Likewise.
12053         * sese.h (bb_map_t, rename_map_t, phi_rename, init_back_edge_pair_t):
12054         Remove unused typedefs.
12055         (struct sese_info_t): Simplify rename_map, remove incomplete_phis.
12057 2017-10-18  Martin Liska  <mliska@suse.cz>
12059         * combine.c (simplify_compare_const): Add gcc_fallthrough.
12061 2017-10-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
12063         * config/s390/s390.c (s390_bb_fallthru_entry_likely): New function.
12064         (s390_sched_init): Do not reset s390_sched_state if we entered the
12065         current basic block via a fallthru edge and all others are unlikely.
12067 2017-10-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
12069         * config/s390/s390.c (NUM_SIDES): New variable.
12070         (LONGRUNNING_THRESHOLD): New variable.
12071         (LATENCY_FACTOR): New variable.
12072         (s390_sched_score): Decrease score for long-running instructions on
12073         wrong side.
12074         (s390_sched_variable_issue): Perform bookkeeping for long-running
12075         instructions.
12077 2017-10-18  Richard Biener  <rguenther@suse.de>
12079         * graphite-isl-ast-to-gimple.c (gcc_expression_from_isl_ast_expr_id):
12080         Simplify with removal of the parameter rename map.
12081         (set_rename): Likewise.
12082         (should_copy_to_new_region): Likewise.
12083         (graphite_copy_stmts_from_block): Likewise.
12084         (copy_bb_and_scalar_dependences): Remove initialization of
12085         unused copied_bb_map.
12086         (copy_def): Remove.
12087         (copy_internal_parameters): Likewise.
12088         (graphite_regenerate_ast_isl): Do not call copy_internal_parameters.
12089         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
12090         Use INTEGRAL_TYPE_P.
12091         (parameter_index_in_region_1): Rename to ...
12092         (assign_parameter_index_in_region): ... this.  Assert we have
12093         a parameter we handle.
12094         (scan_tree_for_params): Adjust.
12095         * sese.h (parameter_rename_map_t): Remove.
12096         (struct sese_info_t): Remove unused parameter_rename_map and
12097         copied_bb_map members.
12098         * sese.c (new_sese_info): Adjust.
12099         (free_sese_info): Likewise.
12101 2017-10-18  Martin Liska  <mliska@suse.cz>
12103         PR sanitizer/82545
12104         * asan.c (asan_expand_poison_ifn): Do not put gimple stmt
12105         on an abnormal edge.
12107 2017-10-18  Sebastian Huber  <sebastian.huber@embedded-brains.de>
12109         * doc/invoke.texi (ffunction-sections and fdata-sections):
12110         Update.
12112 2017-10-17  Eric Botcazou  <ebotcazou@adacore.com>
12114         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Bail out only if
12115         the use statement can throw internally.
12117 2017-10-17  Eric Botcazou  <ebotcazou@adacore.com>
12119         * config/visium/visium.c (visium_select_cc_mode): Return CCmode for
12120         any RTX present on the RHS of a SET.
12121         * compare-elim.c (try_eliminate_compare): Restore comment.
12123 2017-10-17  Jakub Jelinek  <jakub@redhat.com>
12125         * langhooks.h (struct lang_hooks): Document that tree_size langhook
12126         may be also called on tcc_type nodes.
12127         * langhooks.c (lhd_tree_size): Likewise.
12129 2017-10-17  David Malcolm  <dmalcolm@redhat.com>
12131         * gimple-ssa-sprintf.c (fmtwarn): Update for changed signature of
12132         format_warning_at_substring.
12133         (maybe_warn): Convert source_range * param to a location_t.  Pass
12134         UNKNOWN_LOCATION rather than NULL to fmtwarn.
12135         (format_directive): Remove code to extract source_ranges and
12136         source_range * in favor of just a location_t.
12137         (parse_directive): Pass UNKNOWN_LOCATION rather than NULL to
12138         fmtwarn.
12139         * substring-locations.c (format_warning_va): Convert
12140         source_range * param to a location_t.
12141         (format_warning_at_substring): Likewise.
12142         * substring-locations.h (format_warning_va): Likewise.
12143         (format_warning_at_substring): Likewise.
12145 2017-10-17  Jan Hubicka  <hubicka@ucw.cz>
12147         * target.h (enum vect_cost_for_stmt): Add vec_gather_load and
12148         vec_scatter_store
12149         * tree-vect-stmts.c (record_stmt_cost): Make difference between normal
12150         and scatter/gather ops.
12152         * aarch64/aarch64.c (aarch64_builtin_vectorization_cost): Add
12153         vec_gather_load and vec_scatter_store.
12154         * arm/arm.c (arm_builtin_vectorization_cost): Likewise.
12155         * powerpcspe/powerpcspe.c (rs6000_builtin_vectorization_cost): Likewise.
12156         * rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Likewise.
12157         * s390/s390.c (s390_builtin_vectorization_cost): Likewise.
12158         * spu/spu.c (spu_builtin_vectorization_cost): Likewise.
12159         * i386/i386.c (x86_builtin_vectorization_cost): Likewise.
12161 2017-10-17  Uros Bizjak  <ubizjak@gmail.com>
12163         * reg-stack.c (compare_for_stack_reg): Add bool argument.
12164         Detect FTST instruction and handle its register pops.  Only pop
12165         second operand if can_pop_second_op is true.
12166         (subst_stack_regs_pat) <case COMPARE>: Detect FCOMI instruction to
12167         set can_pop_second_op to false in the compare_for_stack_reg call.
12169         * config/i386/i386.md (*cmpi<FPCMP:unord><MODEF:mode>): Only call
12170         output_fp_compare for stack register operands.
12171         * config/i386/i386.c (output_fp_compare): Do not output SSE compare
12172         instructions here.  Do not emit stack register pops here.  Assert
12173         that FCOMPP pops next to top stack register.  Rewrite function.
12175 2017-10-17  Nathan Sidwell  <nathan@acm.org>
12177         PR middle-end/82577
12178         * alias.c (compare_base_decls): Check HAS_DECL_ASSEMBLER_NAME_P,
12179         use DECL_ASSEMBLER_NAME_RAW.
12181         PR middle-end/82546
12182         * tree.c (tree_code_size): Reformat.  Punt to lang hook for unknown
12183         TYPE nodes.
12185 2017-10-17  Qing Zhao <qing.zhao@oracle.com>
12186             Wilco Dijkstra <wilco.dijkstra@arm.com>
12188         * builtins.c (expand_builtin_update_setjmp_buf): Add a
12189         converstion to Pmode from the buf_addr.
12191 2017-10-17  Richard Biener  <rguenther@suse.de>
12193         * graphite-dependences.c (scop_get_reads_and_writes): Change
12194         output parameters to references.
12196 2017-10-17  Jackson Woodruff  <jackson.woodruff@arm.com>
12198         PR 71026/tree-optimization
12199         * fold-const.c (distribute_real_division): Removed.
12200         (fold_binary_loc): Remove calls to distribute_real_divison.
12202 2017-10-17  Richard Biener  <rguenther@suse.de>
12204         * graphite-scop-detection.c
12205         (scop_detection::stmt_has_simple_data_refs_p): Always use
12206         the full nest as region.
12207         (try_generate_gimple_bb): Likewise.
12208         * sese.c (scalar_evolution_in_region): Simplify now that
12209         SCEV can handle instantiation in regions.
12210         * tree-scalar-evolution.c (instantiate_scev_name): Also instantiate
12211         in the non-loop part of a function if requested.
12213 2017-10-17  Richard Biener  <rguenther@suse.de>
12215         PR tree-optimization/82563
12216         * graphite-isl-ast-to-gimple.c (generate_entry_out_of_ssa_copies):
12217         New function.
12218         (graphite_regenerate_ast_isl): Call it.
12219         * graphite-scop-detection.c (build_scops): Remove entry edge split.
12221 2017-10-17  Jakub Jelinek  <jakub@redhat.com>
12223         PR tree-optimization/82549
12224         * fold-const.c (optimize_bit_field_compare, fold_truth_andor_1):
12225         Formatting fixes.  Instead of calling make_bit_field_ref with negative
12226         bitpos return 0.
12228 2017-10-17  Olga Makhotina  <olga.makhotina@intel.com>
12230         * config/i386/avx512dqintrin.h (_mm_mask_reduce_sd,
12231         _mm_maskz_reduce_sd, _mm_mask_reduce_ss,=20
12232         _mm_maskz_reduce_ss): New.
12233         * config/i386/i386-builtin.def (__builtin_ia32_reducesd_mask,
12234         __builtin_ia32_reducess_mask): Ditto..
12235         (__builtin_ia32_reducesd, __builtin_ia32_reducess): Remove.
12236         * config/i386/sse.md (reduces<mode>): Renamed to ...
12237         (reduces<mode><mask_scalar_name>): ... this.
12238         (vreduce<ssescalarmodesuffix>\t{%3, %2, %1, %0|%0, %1, %2, %3}):
12239         Changed to ...
12240         (vreduce<ssescalarmodesuffix>\t{%3, %2, %1, %0<mask_scalar_operand4>|
12241         %0<mask_scalar_operand4>, %1, %2, %3}): ... this.
12243 2017-10-16  David Malcolm  <dmalcolm@redhat.com>
12245         * Makefile.in (OBJS): Add unique-ptr-tests.o.
12246         * selftest-run-tests.c (selftest::run_tests): Call
12247         selftest::unique_ptr_tests_cc_tests.
12248         * selftest.h (selftest::unique_ptr_tests_cc_tests): New decl.
12249         * unique-ptr-tests.cc: New file.
12251 2017-10-16  Vladimir Makarov  <vmakarov@redhat.com>
12253         PR sanitizer/82353
12254         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12255         locations.
12256         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12257         (make_hard_regno_born, make_hard_regno_dead): Update
12258         bb_killed_pseudos and bb_gen_pseudos for fixed regs.
12260 2017-10-16  Jeff Law  <law@redhat.com>
12262         * tree-ssa-dse.c (live_bytes_read): Fix thinko.
12264 2017-10-16  Jan Hubicka  <hubicka@ucw.cz>
12266         * x86-tune-costs.h (znver1_cost): Fix move cost tables.
12268 2017-10-16  Olivier Hainque  <hainque@adacore.com>
12270         * gcc/config.gcc (powerpc*-*-*spe*): Pick 8548 as the default
12271         with_cpu if we were configured for an e500v2 target cpu name.
12273 2017-10-16  Thomas Preud'homme  <thomas.preudhomme@arm.com>
12275         * config/arm/arm-cpus.in (cortex-m33): Add nodsp option.
12276         * doc/invoke.texi: Document +nodsp as a valid extension for
12277         -mcpu=cortex-m33.
12279 2017-10-16  Martin Liska  <mliska@suse.cz>
12281         * sbitmap.c (bitmap_bit_in_range_p_checking): New function.
12282         (test_set_range): Likewise.
12283         (test_range_functions): Rename to ...
12284         (test_bit_in_range): ... this.
12285         (sbitmap_c_tests): Add new test.
12287 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12289         * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
12290         New.
12291         (vdot_lane_u32, vdot_laneq_u32, vdotq_lane_u32, vdotq_laneq_u32): New.
12292         (vdot_lane_s32, vdot_laneq_s32, vdotq_lane_s32, vdotq_laneq_s32): New.
12294 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12296         * config/aarch64/aarch64-builtins.c
12297         (aarch64_types_quadopu_lane_qualifiers): New.
12298         (TYPES_QUADOPU_LANE): New.
12299         * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): New.
12300         (<sur>dot_prod<vsi2qi>, aarch64_<sur>dot_lane<vsi2qi>): New.
12301         (aarch64_<sur>dot_laneq<vsi2qi>): New.
12302         * config/aarch64/aarch64-simd-builtins.def (sdot, udot): New.
12303         (sdot_lane, udot_lane, sdot_laneq, udot_laneq): New.
12304         * config/aarch64/iterators.md (sur): Add UNSPEC_SDOT, UNSPEC_UDOT.
12305         (Vdottype, DOTPROD): New.
12306         (sur): Add SDOT and UDOT.
12308 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12310         * config/aarch64/aarch64.h (AARCH64_FL_DOTPROD): New.
12311         (AARCH64_ISA_DOTPROD, TARGET_DOTPROD): New.
12312         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12313         Add TARGET_DOTPROD.
12314         * config/aarch64/aarch64-option-extensions.def (dotprod): New.
12315         * config/aarch64/aarch64-cores.def (cortex-a55, cortex-a75):
12316         Enable TARGET_DOTPROD.
12317         (cortex-a75.cortex-a55): Likewise.
12318         * doc/invoke.texi (aarch64-feature-modifiers): Document dotprod.
12320 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12322         * config/arm/arm-builtins.c (arm_unsigned_uternop_qualifiers): New.
12323         (UTERNOP_QUALIFIERS, arm_umac_lane_qualifiers, UMAC_LANE_QUALIFIERS):
12324         New.
12325         * config/arm/arm_neon_builtins.def (sdot, udot, sdot_lane, udot_lane):
12326         New.
12327         * config/arm/iterators.md (DOTPROD, VSI2QI, vsi2qi): New.
12328         (UNSPEC_DOT_S, UNSPEC_DOT_U, opsuffix): New.
12329         * config/arm/neon.md (neon_<sup>dot<vsi2qi>): New.
12330         (neon_<sup>dot_lane<vsi2qi>, <sup>dot_prod<vsi2qi>): New.
12331         * config/arm/types.md (neon_dot, neon_dot_q): New.
12332         * config/arm/unspecs.md (sup): Add UNSPEC_DOT_S, UNSPEC_DOT_U.
12334 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12336         * config/arm/arm.h (TARGET_DOTPROD): New.
12337         * config/arm/arm.c (arm_arch_dotprod): New.
12338         (arm_option_reconfigure_globals): Add arm_arch_dotprod.
12339         * config/arm/arm-c.c (__ARM_FEATURE_DOTPROD): New.
12340         * config/arm/arm-cpus.in (armv8.2-a): Enabled +dotprod.
12341         (feature dotprod, group dotprod, ALL_SIMD_INTERNAL): New.
12342         (ALL_FPU_INTERNAL): Use ALL_SIMD_INTERNAL.
12343         * config/arm/t-multilib (v8_2_a_simd_variants): Add dotprod.
12344         * doc/invoke.texi (armv8.2-a): Document dotprod
12346 2017-10-14  Jan Hubicka  <hubicka@ucw.cz>
12348         * i386.c (ix86_vec_cost): New function.
12349         (ix86_rtx_costs): Handle vector operations better.
12350         * i386.h (struct processor_costs): Add sse_op, fmasd, fmass.
12351         * x86-tune-costs.h: Add new costs to all tables.
12353 2017-10-14  Jan Hubicka  <hubicka@ucw.cz>
12355         * i386.c (ix86_rtx_costs): Make difference between x87 and SSE
12356         operations.
12357         * i386.h (struct processor_costs): Add addss, mulss, mulsd, divss,
12358         divsd, sqrtss and sqrtsd
12359         * x86-tune-costs.h: Add new entries to all costs.
12360         (znver1_cost): Fix to match real instruction latencies.
12362 2017-10-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12363             Michael Collison <michael.collison@arm.com>
12365         * compare-elim.c: Include emit-rtl.h.
12366         (can_merge_compare_into_arith): New function.
12367         (try_validate_parallel): Likewise.
12368         (try_merge_compare): Likewise.
12369         (try_eliminate_compare): Call the above when no previous clobber
12370         is available.
12371         (execute_compare_elim_after_reload): Add DF_UD_CHAIN and DF_DU_CHAIN
12372         dataflow problems.
12374 2017-10-14  Jakub Jelinek  <jakub@redhat.com>
12376         PR middle-end/62263
12377         PR middle-end/82498
12378         * tree-ssa-phiopt.c (value_replacement): Comment fix.  Handle
12379         up to 2 preparation statements for ASSIGN in MIDDLE_BB.
12381         PR middle-end/62263
12382         PR middle-end/82498
12383         * tree-ssa-forwprop.c (simplify_rotate): Allow def_arg1[N]
12384         to be any operand_equal_p operands.  For & (B - 1) require
12385         B to be power of 2.  Recognize
12386         (X << (Y & (B - 1))) | (X >> ((-Y) & (B - 1))) and similar patterns.
12388 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12390         PR bootstrap/82553
12391         * optabs.c (expand_memory_blockage): Fix call of
12392         targetm.have_memory_blockage.
12394 2017-10-14  Jakub Jelinek  <jakub@redhat.com>
12396         PR bootstrap/82548
12397         * config.gcc (*-*-solaris2*, i[34567]86-*-cygwin*,
12398         x86_64-*-cygwin*, i[34567]86-*-mingw* | x86_64-*-mingw*): Append
12399         objects to extra_objs instead of overwriting it.
12401 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12403         * config/i386/sync.md (FILD_ATOMIC/FIST_ATOMIC FP load peephole2):
12404         Use any_fp_register_operand as operand[3] predicate.  Simplify
12405         equality test for operands[2] and operands[4] memory location.
12406         (LDX_ATOMIC/STX_ATOMIC FP load peephole2): Ditto.
12407         (FILD_ATOMIC/FIST_ATOMIC FP load peephole2 with mem blockage): New.
12408         (LDX_ATOMIC/LDX_ATOMIC FP load peephole2 with mem blockage): Ditto.
12409         (FILD_ATOMIC/FIST_ATOMIC FP store peephole2): Use
12410         any_fp_register_operand as operand[1] predicate.  Simplify
12411         equality test for operands[0] and operands[3] memory location.
12412         (LDX_ATOMIC/STX_ATOMIC FP store peephole2): Ditto.
12413         (FILD_ATOMIC/FIST_ATOMIC FP store peephole2 with mem blockage): New.
12414         (LDX_ATOMIC/LDX_ATOMIC FP storepeephole2 with mem blockage): Ditto.
12416 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12418         * target-insns.def: Add memory_blockage.
12419         * optabs.c (expand_memory_blockage): New function.
12420         (expand_asm_memory_barrier): Rename ...
12421         (expand_asm_memory_blockage): ... to this.
12422         (expand_mem_thread_fence): Call expand_memory_blockage
12423         instead of expand_asm_memory_barrier.
12424         (expand_mem_singnal_fence): Ditto.
12425         (expand_atomic_load): Ditto.
12426         (expand_atomic_store): Ditto.
12427         * doc/md.texi (Standard Pattern Names For Generation):
12428         Document memory_blockage instruction pattern.
12430 2017-10-13  Sebastian Perta  <sebastian.perta@renesas.com>
12432         * config/rl78/rl78.c (rl78_emit_libcall): New function.
12433         * config/rl78/rl78-protos.h (rl78_emit_libcall): New function.
12434         * config/rl78/rl78.md: New define_expand "adddi3".
12436 2017-10-13  Jan Hubicka  <hubicka@ucw.cz>
12438         * cfghooks.c (verify_flow_info): Disable check that all probabilities
12439         are set correctly.
12441 2017-10-13  Jeff Law  <law@redhat.com>
12443         * tree-ssa-reassoc.c (reassociate_bb): Clarify code slighly.
12445 2017-10-13  Jakub Jelinek  <jakub@redhat.com>
12447         PR target/82274
12448         * internal-fn.c (expand_mul_overflow): If both operands have
12449         the same highpart of -1 or 0 and the topmost bit of lowpart
12450         is different, overflow is if res <= 0 rather than res < 0.
12452 2017-10-13  Pat Haugen  <pthaugen@us.ibm.com>
12454         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Remove
12455         TARGET_P9_VECTOR code for unaligned_load case.
12457 2017-10-13  Jan Hubicka  <hubicka@ucw.cz>
12459         * cfghooks.c (verify_flow_info): Check that edge probabilities are set.
12461 2017-10-13  Nathan Sidwell  <nathan@acm.org>
12463         * tree-core.h (tree_contains_struct): Make bool.
12464         * tree.c (tree_contains_struct): Likewise.
12465         * tree.h (MARK_TS_BASE): Remove do ... while (0) idiom.
12466         (MARK_TS_TYPED, MARK_TS_COMMON, MARK_TS_TYPE_COMMON,
12467         MARK_TS_TYPE_WITH_LANG_SPECIFIC, MARK_TS_DECL_MINIMAL,
12468         MARK_TS_DECL_COMMON, MARK_TS_DECL_WRTL, MARK_TS_DECL_WITH_VIS,
12469         MARK_TS_DECL_NON_COMMON): Likewise, use comma operator.
12471 2017-10-13  Richard Biener  <rguenther@suse.de>
12473         * graphite-isl-ast-to-gimple.c
12474         (translate_isl_ast_to_gimple::get_rename_from_scev): Remove unused
12475         parameters and dominance check.
12476         (translate_isl_ast_to_gimple::graphite_copy_stmts_from_block): Adjust.
12477         (translate_isl_ast_to_gimple::copy_bb_and_scalar_dependences): Likewise.
12478         (translate_isl_ast_to_gimple::graphite_regenerate_ast_isl):
12479         Do not update SSA form here or do intermediate IL verification.
12480         * graphite.c: Include tree-ssa.h and tree-into-ssa.h.
12481         (graphite_initialize): Remove check on the number of loops in
12482         the function and inline into graphite_transform_loops.
12483         (graphite_finalize): Inline into graphite_transform_loops.
12484         (graphite_transform_loops): Perform SSA update and IL verification
12485         here.
12486         * params.def (PARAM_GRAPHITE_MIN_LOOPS_PER_FUNCTION): Remove.
12488 2017-10-13  Richard Biener  <rguenther@suse.de>
12490         * graphite-isl-ast-to-gimple.c (max_mode_int_precision,
12491         graphite_expression_type_precision): Avoid global constructor
12492         by moving ...
12493         (translate_isl_ast_to_gimple::translate_isl_ast_to_gimple): Here.
12494         (translate_isl_ast_to_gimple::graphite_expr_type): Add type member.
12495         (translate_isl_ast_to_gimple::translate_isl_ast_node_for): Use it.
12496         (translate_isl_ast_to_gimple::build_iv_mapping): Likewise.
12497         (translate_isl_ast_to_gimple::graphite_create_new_guard): Likewise.
12498         * graphite-sese-to-poly.c (build_original_schedule): Return nothing.
12500 2017-10-13  H.J. Lu  <hongjiu.lu@intel.com>
12502         PR target/82499
12503         * config/i386/i386.h (ix86_red_zone_size): New.
12504         * config/i386/i386.md (push peephole2s): Replace
12505         "!ix86_using_red_zone ()" with "ix86_red_zone_size == 0".
12507 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12508             Alan Hayward  <alan.hayward@arm.com>
12509             David Sherwood  <david.sherwood@arm.com>
12511         * combine.c (can_change_dest_mode): Reject changes in
12512         REGMODE_NATURAL_SIZE.
12514 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12515             Alan Hayward  <alan.hayward@arm.com>
12516             David Sherwood  <david.sherwood@arm.com>
12518         * cfgexpand.c (expand_debug_expr): Use GET_MODE_UNIT_BITSIZE.
12519         (expand_debug_source_expr): Likewise.
12520         * combine.c (combine_simplify_rtx): Likewise.
12521         * cse.c (fold_rtx): Likewise.
12522         * fwprop.c (canonicalize_address): Likewise.
12523         * targhooks.c (default_shift_truncation_mask): Likewise.
12525 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12526             Alan Hayward  <alan.hayward@arm.com>
12527             David Sherwood  <david.sherwood@arm.com>
12529         * optabs.c (add_equal_note): Use GET_MODE_UNIT_SIZE.
12530         (widened_mode): Likewise.
12531         (expand_unop): Likewise.
12532         * ree.c (transform_ifelse): Likewise.
12533         (merge_def_and_ext): Likewise.
12534         (combine_reaching_defs): Likewise.
12535         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12537 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12538             Alan Hayward  <alan.hayward@arm.com>
12539             David Sherwood  <david.sherwood@arm.com>
12541         * caller-save.c (replace_reg_with_saved_mem): Use byte_lowpart_offset.
12542         * combine.c (gen_lowpart_for_combine): Likewise.
12543         * dwarf2out.c (rtl_for_decl_location): Likewise.
12544         * final.c (alter_subreg): Likewise.
12545         * rtlhooks.c (gen_lowpart_general): Likewise.
12546         (gen_lowpart_if_possible): Likewise.
12548 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12549             Alan Hayward  <alan.hayward@arm.com>
12550             David Sherwood  <david.sherwood@arm.com>
12552         * calls.c (expand_call): Use subreg_lowpart_offset.
12553         * cse.c (cse_insn): Likewise.
12554         * regcprop.c (copy_value): Likewise.
12555         (copyprop_hardreg_forward_1): Likewise.
12557 2017-10-13  Jakub Jelinek  <jakub@redhat.com>
12559         PR target/82524
12560         * config/i386/i386.md (addqi_ext_1, andqi_ext_1,
12561         *andqi_ext_1_cc, *<code>qi_ext_1, *xorqi_ext_1_cc): Change
12562         =Q constraints to +Q and into insn condition add check
12563         that operands[0] and operands[1] are equal.
12564         (*addqi_ext_2, *andqi_ext_2, *<code>qi_ext_2): Change
12565         =Q constraints to +Q and into insn condition add check
12566         that operands[0] is equal to either operands[1] or operands[2].
12568         PR target/82498
12569         * fold-const.c (fold_binary_loc) <bit_rotate>: Code cleanups,
12570         instead of handling MINUS_EXPR twice (once for each argument),
12571         canonicalize operand order and handle just once, use rtype where
12572         possible.  Handle (A << B) | (A >> (-B & (Z - 1))).
12574         PR target/82498
12575         * config/i386/ia32intrin.h (__rold, __rord, __rolq, __rorq): Allow
12576         any values of __C while still being pattern recognizable as a simple
12577         rotate instruction.
12579 2017-10-13  Richard Biener  <rguenther@suse.de>
12581         PR tree-optimization/82451
12582         Revert
12583         2017-10-02  Richard Biener  <rguenther@suse.de>
12585         PR tree-optimization/82355
12586         * graphite-isl-ast-to-gimple.c (build_iv_mapping): Also build
12587         a mapping for the enclosing loop but avoid generating one for
12588         the loop tree root.
12589         (copy_bb_and_scalar_dependences): Remove premature codegen
12590         error on PHIs in blocks duplicated into multiple places.
12591         * graphite-scop-detection.c
12592         (scop_detection::stmt_has_simple_data_refs_p): For a loop not
12593         in the region use it as loop and nest to analyze the DR in.
12594         (try_generate_gimple_bb): Likewise.
12595         * graphite-sese-to-poly.c (extract_affine_chrec): Adjust.
12596         (add_loop_constraints): For blocks in a loop not in the region
12597         create a dimension with a single iteration.
12598         * sese.h (gbb_loop_at_index): Remove assert.
12600         * cfgloop.c (loop_preheader_edge): For the loop tree root
12601         return the single successor of the entry block.
12602         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
12603         Reset the SCEV hashtable and niters.
12604         * graphite-scop-detection.c
12605         (scop_detection::graphite_can_represent_scev): Add SCOP parameter,
12606         assert that we only have POLYNOMIAL_CHREC that vary in loops
12607         contained in the region.
12608         (scop_detection::graphite_can_represent_expr): Adjust.
12609         (scop_detection::stmt_has_simple_data_refs_p): For loops
12610         not in the region set loop to NULL.  The nest is now the
12611         entry edge to the region.
12612         (try_generate_gimple_bb): Likewise.
12613         * sese.c (scalar_evolution_in_region): Adjust for
12614         instantiate_scev change.
12615         * tree-data-ref.h (graphite_find_data_references_in_stmt):
12616         Make nest parameter the edge into the region.
12617         (create_data_ref): Likewise.
12618         * tree-data-ref.c (dr_analyze_indices): Make nest parameter an
12619         entry edge into a region and adjust instantiate_scev calls.
12620         (create_data_ref): Likewise.
12621         (graphite_find_data_references_in_stmt): Likewise.
12622         (find_data_references_in_stmt): Pass the loop preheader edge
12623         from the nest argument.
12624         * tree-scalar-evolution.h (instantiate_scev): Make instantiate_below
12625         parameter the edge into the region.
12626         (instantiate_parameters): Use the loop preheader edge as entry.
12627         * tree-scalar-evolution.c (analyze_scalar_evolution): Handle
12628         NULL loop.
12629         (get_instantiated_value_entry): Make instantiate_below parameter
12630         the edge into the region.
12631         (instantiate_scev_name): Likewise.  Adjust dominance checks,
12632         when we cannot use loop-based instantiation instantiate by
12633         walking use-def chains.
12634         (instantiate_scev_poly): Adjust.
12635         (instantiate_scev_binary): Likewise.
12636         (instantiate_scev_convert): Likewise.
12637         (instantiate_scev_not): Likewise.
12638         (instantiate_array_ref): Remove.
12639         (instantiate_scev_3): Likewise.
12640         (instantiate_scev_2): Likewise.
12641         (instantiate_scev_1): Likewise.
12642         (instantiate_scev_r): Do not blindly handle N-operand trees.
12643         Do not instantiate array-refs.  Handle all constants and invariants.
12644         (instantiate_scev): Make instantiate_below parameter
12645         the edge into the region.
12646         (resolve_mixers): Use the loop preheader edge for the region
12647         parameter to instantiate_scev_r.
12648         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Adjust.
12650 2017-10-13  Richard Biener  <rguenther@suse.de>
12652         PR tree-optimization/82525
12653         * graphite-isl-ast-to-gimple.c
12654         (translate_isl_ast_to_gimple::widest_int_from_isl_expr_int): Split
12655         out from ...
12656         (translate_isl_ast_to_gimple::gcc_expression_from_isl_expr_int): Here.
12657         Fail code generation when we cannot represent the isl integer.
12658         (binary_op_to_tree): Elide modulo operations that are no-ops
12659         in the type we code generate.  Remove now superfluous code
12660         generation errors.
12662 2017-10-13  Richard Biener  <rguenther@suse.de>
12664         * graphite-scop-detection.c (loop_ivs_can_be_represented): Remove.
12665         (scop_detection::harmful_loop_in_region): Remove premature
12666         IV type restriction.
12667         (scop_detection::graphite_can_represent_scev): We can handle
12668         pointer IVs just fine.
12670 2017-10-13  Alan Modra  <amodra@gmail.com>
12672         * doc/extend.texi (Extended Asm <Clobbers>): Rename to
12673         "Clobbers and Scratch Registers".  Add paragraph on
12674         alternative to clobbers for scratch registers and OpenBLAS
12675         example.
12677 2017-10-13  Alan Modra  <amodra@gmail.com>
12679         * doc/extend.texi (Clobbers): Correct vax example.  Delete old
12680         example of a memory input for a string of known length.  Move
12681         commentary out of table.  Add a number of new examples
12682         covering array memory inputs.
12684 2017-10-12  Martin Liska  <mliska@suse.cz>
12686         PR tree-optimization/82493
12687         * sbitmap.c (bitmap_bit_in_range_p): Fix the implementation.
12688         (test_range_functions): New function.
12689         (sbitmap_c_tests): Likewise.
12690         * selftest-run-tests.c (selftest::run_tests): Run new tests.
12691         * selftest.h (sbitmap_c_tests): New function.
12693         * tree-ssa-dse.c (live_bytes_read): Fix thinko.
12695 2017-10-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
12697         * config/rs6000/amo.h: Fix spacing issue.
12699 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12701         PR target/82498
12702         * config/i386/i386.md (*ashl<mode>3_mask_1,
12703         *<shift_insn><mode>3_mask_1, *<rotate_insn><mode>3_mask_1,
12704         *<btsc><mode>_mask_1, *btr<mode>_mask_1): New define_insn_and_split
12705         patterns.
12707 2017-10-12  Jan Hubicka  <hubicka@ucw.cz>
12709         * profile-count.h (safe_scale_64bit): Fix GCC4.x path.
12710         (profile_probability): Set max_probability
12711         to (uint32_t) 1 << (n_bits - 2) and update accessors to avoid overlfows
12712         in temporaries.
12713         * profile-count.c (profile_probability::differs_from_p): Do not
12714         rely on max_probaiblity == 10000
12716 2017-10-12  Jeff Law  <law@redhat.com>
12718         * tree-ssa-dse.c (valid_ao_ref_for_dse): Reject ao_refs with
12719         negative offsets.
12721 2017-10-12  Martin Sebor  <msebor@redhat.com>
12723         PR other/82301
12724         PR c/82435
12725         * cgraphunit.c (maybe_diag_incompatible_alias): New function.
12726         (handle_alias_pairs): Call it.
12727         * common.opt (-Wattribute-alias): New option.
12728         * doc/extend.texi (ifunc attribute): Discuss C++ specifics.
12729         * doc/invoke.texi (-Wattribute-alias): Document.
12731 2017-10-12  Vladimir Makarov  <vmakarov@redhat.com>
12733         Revert
12734         2017-10-11  Vladimir Makarov  <vmakarov@redhat.com>
12735         PR sanitizer/82353
12736         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12737         locations.
12738         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12739         (make_hard_regno_born, make_hard_regno_dead): Update
12740         bb_killed_pseudos and bb_gen_pseudos.
12742 2017-10-12  Jan Hubicka  <hubicka@ucw.cz>
12744         * config/i386/x86-tune-sched.c (ix86_adjust_cost): Fix Zen support.
12746 2017-10-12  Uros Bizjak  <ubizjak@gmail.com>
12748         * config/alpha/alpha.c (alpha_split_conditional_move):
12749         Use std::swap instead of manually swapping.
12750         (alpha_stdarg_optimize_hook): Ditto.
12751         (alpha_canonicalize_comparison): Ditto.
12753 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12755         * tree-loop-distribution.c (struct builtin_info): New struct.
12756         (struct partition): Refactor fields into struct builtin_info.
12757         (partition_free): Free struct builtin_info.
12758         (build_size_arg_loc, build_addr_arg_loc): Delete.
12759         (generate_memset_builtin, generate_memcpy_builtin): Get memory range
12760         information from struct builtin_info.
12761         (find_single_drs): New function refactored from classify_partition.
12762         Also moved builtin validity checks to this function.
12763         (compute_access_range, alloc_builtin): New functions.
12764         (classify_builtin_st, classify_builtin_ldst): New functions.
12765         (classify_partition): Refactor code into functions find_single_drs,
12766         classify_builtin_st and classify_builtin_ldst.
12767         (distribute_loop): Don't do runtime alias check when distributing
12768         loop nest.
12769         (find_seed_stmts_for_distribution): New function.
12770         (pass_loop_distribution::execute): Refactor code finding seed
12771         stmts into above function.  Support distribution for the innermost
12772         two-level loop nest.  Adjust dump information.
12774 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12776         * tree-loop-distribution.c: Adjust the general comment.
12777         (NUM_PARTITION_THRESHOLD): New macro.
12778         (ssa_name_has_uses_outside_loop_p): Support loop nest distribution.
12779         (classify_partition): Skip builtin pattern of loop nest's inner loop.
12780         (merge_dep_scc_partitions): New parameter ignore_alias_p and use it
12781         in call to build_partition_graph.
12782         (finalize_partitions): New parameter.  Make loop distribution more
12783         conservative by fusing more partitions.
12784         (distribute_loop): Don't do runtime alias check in case of loop nest
12785         distribution.
12786         (find_seed_stmts_for_distribution): New function.
12787         (prepare_perfect_loop_nest): New function.
12788         (pass_loop_distribution::execute): Refactor code finding seed stmts
12789         and loop nest into above functions.  Support loop nest distribution.
12790         Adjust dump information accordingly.
12792 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12794         * tree-loop-distribution.c (break_alias_scc_partitions): Add comment
12795         and set PTYPE_SEQUENTIAL for merged partition.
12797 2017-10-12  Richard Biener  <rguenther@suse.de>
12799         PR tree-optimization/69728
12800         Revert
12801         2017-09-19  Richard Biener  <rguenther@suse.de>
12803         PR tree-optimization/69728
12804         * graphite-sese-to-poly.c (schedule_error): New global.
12805         (add_loop_schedule): Handle empty domain by failing the
12806         schedule.
12807         (build_original_schedule): Handle schedule_error.
12809         * graphite-sese-to-poly.c (add_loop_schedule): Handle empty
12810         domain by returning an unchanged schedule.
12812 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12814         * genrecog.c (validate_pattern): For VEC_SELECT verify that
12815         CONST_INT selectors are 0 to GET_MODE_NUNITS (imode) - 1.
12817 2017-10-12  Aldy Hernandez  <aldyh@redhat.com>
12819         * Makefile.in (TAGS): Merge all the *.def files into one pattern.
12820         Handle params.def.
12822 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12824         PR c++/82159
12825         * expr.c (store_field): Don't optimize away bitsize == 0 store
12826         from CALL_EXPR with addressable return type.
12828 2017-10-11  Segher Boessenkool  <segher@kernel.crashing.org>
12830         * config/rs6000/rs6000.h (TARGET_ISEL64): Delete.
12831         * config/rs6000/rs6000.md (sel): Delete mode attribute.
12832         (mov<mode>cc, isel_signed_<mode>, isel_unsigned_<mode>,
12833         *isel_reversed_signed_<mode>, *isel_reversed_unsigned_<mode>): Use
12834         TARGET_ISEL instead of TARGET_ISEL<sel>.
12836 2017-10-11  David Edelsohn  <dje.gcc@gmail.com>
12838         * config/rs6000/rs6000.c
12839         (rs6000_xcoff_asm_output_aligned_decl_common): Test for NULL decl.
12841 2017-10-11  Segher Boessenkool  <segher@kernel.crashing.org>
12843         * config/rs6000/predicates.md (zero_constant, all_ones_constant):
12844         Move up in file.
12845         (reg_or_cint_operand): Fix comment.
12846         (reg_or_zero_operand): New predicate.
12847         * config/rs6000/rs6000-protos.h (output_isel): Delete.
12848         * config/rs6000/rs6000.c (output_isel): Delete.
12849         * config/rs6000/rs6000.md (isel_signed_<mode>): Use reg_or_zero_operand
12850         instead of reg_or_cint_operand.  Output instruction directly (not via
12851         output_isel).
12852         (isel_unsigned_<mode>): Ditto.
12853         (*isel_reversed_signed_<mode>): Use reg_or_zero_operand instead of
12854         gpc_reg_operand.  Add an instruction alternative for this.  Output
12855         instruction directly.
12856         (*isel_reversed_unsigned_<mode>): Ditto.
12858 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12860         * config/i386/i386.c (ix86_canonicalize_comparison): New function.
12861         (TARGET_CANONICALIZE_COMPARISON): Define.
12863 2017-10-11  Qing Zhao  <qing.zhao@oracle.com>
12865         PR target/81422
12866         * config/aarch64/aarch64.c (aarch64_load_symref_appropriately):
12867         Check whether the dest is REG before adding REG_EQUIV note.
12869 2017-10-11  Vladimir Makarov  <vmakarov@redhat.com>
12871         PR sanitizer/82353
12872         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12873         locations.
12874         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12875         (make_hard_regno_born, make_hard_regno_dead): Update
12876         bb_killed_pseudos and bb_gen_pseudos.
12878 2017-10-11  Nathan Sidwell  <nathan@acm.org>
12880         * incpath.h (enum incpath_kind): Name enum, prefix values.
12881         (add_path, add_cpp_dir_path, get_added_cpp_dirs): Use incpath_kind.
12882         * incpath.c (heads, tails): Use INC_MAX.
12883         (add_env_var_paths, add_standard_paths): Use incpath_kind.
12884         (merge_include_chains, split_quote_chain,
12885         register_include_chains): Update incpath_kind names.
12886         (add_cpp_dir_path, add_path, get_added_cpp_dirs): Use incpath_kind.
12887         * config/darwin-c.c (add_system_framework_path): Update incpath_kind
12888         names.
12889         (add_framework_path, darwin_register_objc_includes): Likewise.
12890         * config/vms/vms-c.c (vms_c_register_includes): Likewise.
12892 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12894         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
12895         Do not use float_operator operator predicate.
12896         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
12897         * config/i386/predicates.md (float_operator): Remove predicate.
12899 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12901         * config/i386/i386.md (*jcc<mode>_0_i387): Remove insn pattern.
12902         (*jccxf_i387): Ditto.
12903         (*jcc<mode>_i387): Ditto.
12904         (*jccu<mode>_i387): Ditto.
12905         (*jcc<X87MODEF:mode>_<SWI24:mode>_i387): Ditto.
12906         (*jcc_*_i387 splitters): Remove.
12907         * config/i386/i386-protos.h (ix86_split_fp_branch): Remove prototype.
12908         * config/i386/i386.c (ix86_split_fp_branch): Remove.
12909         * config/i386/predicates.md (ix86_swapped_fp_comparison_operator):
12910         Remove predicate.
12912 2017-10-11  Jan Hubicka  <hubicka@ucw.cz>
12914         * profile-count.h (slow_safe_scale_64bit): New function.
12915         (safe_scale_64bit): New inline.
12916         (profile_count::max_safe_multiplier): Remove; use safe_scale_64bit.
12917         * profile-count.c: Include wide-int.h
12918         (slow_safe_scale_64bit): New.
12920 2017-10-11  Nathan Sidwell  <nathan@acm.org>
12922         * tree.h (DECL_ASSEMBLER_NAME_SET_P): Don't check
12923         HAS_DECL_ASSEMBLER_NAME_P.
12924         * gimple-expr.c (gimple_decl_printable_name: Check
12925         HAS_DECL_ASSEMBLER_NAME_P too.
12926         * ipa-utils.h (type_in_anonymous_namespace_p): Check
12927         DECL_ASSEMBLER_NAME_SET_P of TYPE_NAME.
12928         (odr_type_p): No need to assert TYPE_NAME is a TYPE_DECL.
12929         * passes.c (rest_of_decl_compilation): Check
12930         HAS_DECL_ASSEMBLER_NAME_P too.
12931         * recog.c (verify_changes): Likewise.
12932         * tree-pretty-print.c (dump_decl_name): Likewise.
12933         * tree-ssa-structalias.c (alias_get_name): Likewise.  Reimplement.
12935         * tree.h (DECL_ASSEMBLER_NAME_RAW): New.
12936         (SET_DECL_ASSEMBLER_NAME): Use it.
12937         (DECL_ASSEMBLER_NAME_SET_P): Likewise.
12938         (COPY_DECL_ASSEMBLER_NAME): Likewise.
12939         * tree.c (decl_assembler_name): Use DECL_ASSEMBLER_NAME_RAW.
12941 2017-10-11  Jan Hubicka  <hubicka@ucw.cz>
12943         * config.gcc (i386, x86_64): Add extra objects.
12944         * config/i386/i386-protos.h (ix86_rip_relative_addr_p): Declare.
12945         (ix86_min_insn_size): Declare.
12946         (ix86_issue_rate): Declare.
12947         (ix86_adjust_cost): Declare.
12948         (ia32_multipass_dfa_lookahead): Declare.
12949         (ix86_macro_fusion_p): Declare.
12950         (ix86_macro_fusion_pair_p): Declare.
12951         (ix86_bd_has_dispatch): Declare.
12952         (ix86_bd_do_dispatch): Declare.
12953         (ix86_core2i7_init_hooks): Declare.
12954         (ix86_atom_sched_reorder): Declare.
12955         * config/i386/i386.c Move all CPU cost tables to x86-tune-costs.h.
12956         (COSTS_N_BYTES): Move to x86-tune-costs.h.
12957         (DUMMY_STRINGOP_ALGS):Move to x86-tune-costs.h.
12958         (rip_relative_addr_p): Rename to ...
12959         (ix86_rip_relative_addr_p): ... this one; export.
12960         (memory_address_length): Update.
12961         (ix86_issue_rate): Move to x86-tune-sched.c.
12962         (ix86_flags_dependent): Move to x86-tune-sched.c.
12963         (ix86_agi_dependent): Move to x86-tune-sched.c.
12964         (exact_dependency_1): Move to x86-tune-sched.c.
12965         (exact_store_load_dependency): Move to x86-tune-sched.c.
12966         (ix86_adjust_cost): Move to x86-tune-sched.c.
12967         (ia32_multipass_dfa_lookahead): Move to x86-tune-sched.c.
12968         (ix86_macro_fusion_p): Move to x86-tune-sched.c.
12969         (ix86_macro_fusion_pair_p): Move to x86-tune-sched.c.
12970         (do_reorder_for_imul): Move to x86-tune-sched-atom.c.
12971         (swap_top_of_ready_list): Move to x86-tune-sched-atom.c.
12972         (ix86_sched_reorder): Move to x86-tune-sched-atom.c.
12973         (core2i7_first_cycle_multipass_init): Move to x86-tune-sched-core.c.
12974         (core2i7_dfa_post_advance_cycle): Move to x86-tune-sched-core.c.
12975         (min_insn_size): Rename to ...
12976         (ix86_min_insn_size): ... this one; export.
12977         (core2i7_first_cycle_multipass_begin): Move to x86-tune-sched-core.c.
12978         (core2i7_first_cycle_multipass_issue): Move to x86-tune-sched-core.c.
12979         (core2i7_first_cycle_multipass_backtrack): Move to
12980         x86-tune-sched-core.c.
12981         (core2i7_first_cycle_multipass_end): Move to x86-tune-sched-core.c.
12982         (core2i7_first_cycle_multipass_fini): Move to x86-tune-sched-core.c.
12983         (ix86_sched_init_global): Break up logic to ix86_core2i7_init_hooks.
12984         (ix86_avoid_jump_mispredicts): Update.
12985         (TARGET_SCHED_DISPATCH): Move to ix86-tune-sched-bd.c.
12986         (TARGET_SCHED_DISPATCH_DO): Move to ix86-tune-sched-bd.c.
12987         (TARGET_SCHED_REORDER): Move to ix86-tune-sched-bd.c.
12988         (DISPATCH_WINDOW_SIZE): Move to ix86-tune-sched-bd.c.
12989         (MAX_DISPATCH_WINDOWS): Move to ix86-tune-sched-bd.c.
12990         (MAX_INSN): Move to ix86-tune-sched-bd.c.
12991         (MAX_IMM): Move to ix86-tune-sched-bd.c.
12992         (MAX_IMM_SIZE): Move to ix86-tune-sched-bd.c.
12993         (MAX_IMM_32): Move to ix86-tune-sched-bd.c.
12994         (MAX_IMM_64): Move to ix86-tune-sched-bd.c.
12995         (MAX_LOAD): Move to ix86-tune-sched-bd.c.
12996         (MAX_STORE): Move to ix86-tune-sched-bd.c.
12997         (BIG): Move to ix86-tune-sched-bd.c.
12998         (enum dispatch_group): Move to ix86-tune-sched-bd.c.
12999         (enum insn_path): Move to ix86-tune-sched-bd.c.
13000         (get_mem_group): Move to ix86-tune-sched-bd.c.
13001         (is_cmp): Move to ix86-tune-sched-bd.c.
13002         (dispatch_violation): Move to ix86-tune-sched-bd.c.
13003         (is_branch): Move to ix86-tune-sched-bd.c.
13004         (is_prefetch): Move to ix86-tune-sched-bd.c.
13005         (init_window): Move to ix86-tune-sched-bd.c.
13006         (allocate_window): Move to ix86-tune-sched-bd.c.
13007         (init_dispatch_sched): Move to ix86-tune-sched-bd.c.
13008         (is_end_basic_block): Move to ix86-tune-sched-bd.c.
13009         (process_end_window): Move to ix86-tune-sched-bd.c.
13010         (allocate_next_window): Move to ix86-tune-sched-bd.c.
13011         (find_constant): Move to ix86-tune-sched-bd.c.
13012         (get_num_immediates): Move to ix86-tune-sched-bd.c.
13013         (has_immediate): Move to ix86-tune-sched-bd.c.
13014         (get_insn_path): Move to ix86-tune-sched-bd.c.
13015         (get_insn_group): Move to ix86-tune-sched-bd.c.
13016         (count_num_restricted): Move to ix86-tune-sched-bd.c.
13017         (fits_dispatch_window): Move to ix86-tune-sched-bd.c.
13018         (add_insn_window): Move to ix86-tune-sched-bd.c.
13019         (add_to_dispatch_window): Move to ix86-tune-sched-bd.c.
13020         (debug_dispatch_window_file): Move to ix86-tune-sched-bd.c.
13021         (debug_dispatch_window): Move to ix86-tune-sched-bd.c.
13022         (debug_insn_dispatch_info_file): Move to ix86-tune-sched-bd.c.
13023         (debug_ready_dispatch): Move to ix86-tune-sched-bd.c.
13024         (do_dispatch): Move to ix86-tune-sched-bd.c.
13025         (has_dispatch): Move to ix86-tune-sched-bd.c.
13026         * config/i386/t-i386: Add new object files.
13027         * config/i386/x86-tune-costs.h: New file.
13028         * config/i386/x86-tune-sched-atom.c: New file.
13029         * config/i386/x86-tune-sched-bd.c: New file.
13030         * config/i386/x86-tune-sched-core.c: New file.
13031         * config/i386/x86-tune-sched.c: New file.
13033 2017-10-11  Liu Hao  <lh_mouse@126.com>
13035         * pretty-print.c [_WIN32] (colorize_init): Remove.  Use
13036         the generic version below instead.
13037         (should_colorize): Recognize Windows consoles as terminals
13038         for MinGW targets.
13039         * pretty-print.c [__MINGW32__] (write_all): New function.
13040         [__MINGW32__] (find_esc_head): Likewise.
13041         [__MINGW32__] (find_esc_terminator): Likewise.
13042         [__MINGW32__] (eat_esc_sequence): Likewise.
13043         [__MINGW32__] (mingw_ansi_fputs): New function that handles
13044         ANSI escape codes.
13045         (pp_write_text_to_stream): Use mingw_ansi_fputs instead of fputs
13046         for MinGW targets.
13048 2017-10-11  Richard Biener  <rguenther@suse.de>
13050         * tree-ssa-loop-niter.c (infer_loop_bounds_from_pointer_arith):
13051         Properly call analyze_scalar_evolution with the loop of the stmt.
13053 2017-10-11  Richard Biener  <rguenther@suse.de>
13055         * tree.def (POLYNOMIAL_CHREC): Remove CHREC_VARIABLE tree operand.
13056         * tree-core.h (tree_base): Add chrec_var union member.
13057         * tree.h (CHREC_VAR): Remove.
13058         (CHREC_LEFT, CHREC_RIGHT, CHREC_VARIABLE): Adjust.
13059         * tree-chrec.h (build_polynomial_chrec): Adjust.
13060         * tree-chrec.c (reset_evolution_in_loop): Use build_polynomial_chrec.
13061         * tree-pretty-print.c (dump_generic_node): Use CHREC_VARIABLE.
13063 2017-10-11  Marc Glisse  <marc.glisse@inria.fr>
13065         * fold-const.c (fold_binary_loc) [X +- Y CMP X]: Move ...
13066         * match.pd: ... here.
13067         ((T) X == (T) Y): Relax condition.
13069 2017-10-11  Bin Cheng  <bin.cheng@arm.com>
13071         PR tree-optimization/82472
13072         * tree-loop-distribution.c (sort_partitions_by_post_order): Refine
13073         comment.
13074         (break_alias_scc_partitions): Update postorder number.
13076 2017-10-11  Martin Liska  <mliska@suse.cz>
13078         PR sanitizer/82490
13079         * opts.c (parse_no_sanitize_attribute): Do not use error_value
13080         variable.
13081         * opts.h (parse_no_sanitize_attribute): Remove last argument.
13083 2017-10-11  Martin Liska  <mliska@suse.cz>
13085         * print-rtl.c (print_insn): Move declaration of idbuf
13086         to same scope as name.
13088 2017-10-11  Martin Liska  <mliska@suse.cz>
13090         Revert r253637:
13092         PR sanitizer/82484
13093         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
13094         volatile arguments.
13096 2017-10-11  Martin Liska  <mliska@suse.cz>
13098         PR sanitizer/82484
13099         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
13100         volatile arguments.
13102 2017-10-11  Adhemerval Zanella  <adhemerval.zanella@linaro.org>
13104         * config.gcc (default_gnu_indirect_function): Default to yes for
13105         arm*-*-linux* with glibc.
13107 2017-10-11  Richard Biener  <rguenther@suse.de>
13109         * tree-scalar-evolution.c (get_scalar_evolution): Handle
13110         default-defs and types we do not want to analyze.
13111         (interpret_loop_phi): Replace unreachable code with an assert.
13112         (compute_scalar_evolution_in_loop): Remove and inline ...
13113         (analyze_scalar_evolution_1): ... here, replacing condition with
13114         what makes the intent clearer.  Remove handling of cases
13115         get_scalar_evolution now handles.
13117 2017-10-10  Jim Wilson  <wilson@tuliptree.org>
13119         PR rtl-optimization/81434
13120         * haifa-sched.c (prune_ready_list): Init min_cost_group to 0.  Update
13121         comment for main loop.  In sched_group_found if, also add checks for
13122         pass and min_cost_group.
13124 2017-10-10  Segher Boessenkool  <segher@kernel.crashing.org>
13126         * config/rs6000/rs6000.c (TARGET_INSN_COST): New.
13127         (rs6000_insn_cost): New function.
13128         * config/rs6000/rs6000.md (cost): New attribute.
13130 2017-10-10  Jakub Jelinek  <jakub@redhat.com>
13131             H.J. Lu  <hongjiu.lu@intel.com>
13133         PR target/79565
13134         PR target/82483
13135         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Add
13136         OPTION_MASK_ISA_MMX for __builtin_ia32_maskmovq,
13137         __builtin_ia32_vec_ext_v4hi and __builtin_ia32_vec_set_v4hi.
13138         (ix86_expand_builtin): Treat OPTION_MASK_ISA_MMX similarly
13139         to OPTION_MASK_ISA_AVX512VL - builtins that have both
13140         OPTION_MASK_ISA_MMX and some other bit set require both
13141         mmx and the ISAs without the mmx bit.
13142         * config/i386/i386-builtin.def (__builtin_ia32_cvtps2pi,
13143         __builtin_ia32_cvttps2pi, __builtin_ia32_cvtpi2ps,
13144         __builtin_ia32_pavgb, __builtin_ia32_pavgw, __builtin_ia32_pmulhuw,
13145         __builtin_ia32_pmaxub, __builtin_ia32_pmaxsw, __builtin_ia32_pminub,
13146         __builtin_ia32_pminsw, __builtin_ia32_psadbw, __builtin_ia32_pmovmskb,
13147         __builtin_ia32_pshufw, __builtin_ia32_cvtpd2pi,
13148         __builtin_ia32_cvttpd2pi, __builtin_ia32_cvtpi2pd,
13149         __builtin_ia32_pmuludq, __builtin_ia32_pabsb, __builtin_ia32_pabsw,
13150         __builtin_ia32_pabsd, __builtin_ia32_phaddw, __builtin_ia32_phaddd,
13151         __builtin_ia32_phaddsw, __builtin_ia32_phsubw, __builtin_ia32_phsubd,
13152         __builtin_ia32_phsubsw, __builtin_ia32_pmaddubsw,
13153         __builtin_ia32_pmulhrsw, __builtin_ia32_pshufb, __builtin_ia32_psignb,
13154         __builtin_ia32_psignw, __builtin_ia32_psignd, __builtin_ia32_movntq,
13155         __builtin_ia32_paddq, __builtin_ia32_psubq, __builtin_ia32_palignr):
13156         Add OPTION_MASK_ISA_MMX.
13158 2017-10-10  Andreas Tobler  <andreast@gcc.gnu.org>
13160         * config.gcc (armv7*-*-freebsd*): New target.
13161         (armv6*-*-freebsd*): Remove obsolete TARGET_FREEBSD_ARMv6 define.
13163 2017-10-10  Jan Hubicka  <hubicka@ucw.cz>
13165         * x86-tune.def (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI,
13166         X86_TUNE_ADJUST_UNROLL, X86_TUNE_ONE_IF_CONV_INSN): Move to right
13167         spot in the file.
13169 2017-10-10  Richard Sandiford  <richard.sandiford@linaro.org>
13171         * wide-int.h (wide_int_ref_storage): Make host_dependent_precision
13172         a template parameter.
13173         (WIDE_INT_REF_FOR): Update accordingly.
13174         * tree.h (wi::int_traits <const_tree>): Delete.
13175         (wi::tree_to_widest_ref, wi::tree_to_offset_ref): New typedefs.
13176         (wi::to_widest, wi::to_offset): Use them.  Expand commentary.
13177         (wi::tree_to_wide_ref): New typedef.
13178         (wi::to_wide): New function.
13179         * calls.c (get_size_range): Use wi::to_wide when operating on
13180         trees as wide_ints.
13181         * cgraph.c (cgraph_node::create_thunk): Likewise.
13182         * config/i386/i386.c (ix86_data_alignment): Likewise.
13183         (ix86_local_alignment): Likewise.
13184         * dbxout.c (stabstr_O): Likewise.
13185         * dwarf2out.c (add_scalar_info, gen_enumeration_type_die): Likewise.
13186         * expr.c (const_vector_from_tree): Likewise.
13187         * fold-const-call.c (host_size_t_cst_p, fold_const_call_1): Likewise.
13188         * fold-const.c (may_negate_without_overflow_p, negate_expr_p)
13189         (fold_negate_expr_1, int_const_binop_1, const_binop)
13190         (fold_convert_const_int_from_real, optimize_bit_field_compare)
13191         (all_ones_mask_p, sign_bit_p, unextend, extract_muldiv_1)
13192         (fold_div_compare, fold_single_bit_test, fold_plusminus_mult_expr)
13193         (pointer_may_wrap_p, expr_not_equal_to, fold_binary_loc)
13194         (fold_ternary_loc, multiple_of_p, fold_negate_const, fold_abs_const)
13195         (fold_not_const, round_up_loc): Likewise.
13196         * gimple-fold.c (gimple_fold_indirect_ref): Likewise.
13197         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Likewise.
13198         (alloca_call_type): Likewise.
13199         * gimple.c (preprocess_case_label_vec_for_gimple): Likewise.
13200         * godump.c (go_output_typedef): Likewise.
13201         * graphite-sese-to-poly.c (tree_int_to_gmp): Likewise.
13202         * internal-fn.c (get_min_precision): Likewise.
13203         * ipa-cp.c (ipcp_store_vr_results): Likewise.
13204         * ipa-polymorphic-call.c
13205         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Likewise.
13206         * ipa-prop.c (ipa_print_node_jump_functions_for_edge): Likewise.
13207         (ipa_modify_call_arguments): Likewise.
13208         * match.pd: Likewise.
13209         * omp-low.c (scan_omp_1_op, lower_omp_ordered_clauses): Likewise.
13210         * print-tree.c (print_node_brief, print_node): Likewise.
13211         * stmt.c (expand_case): Likewise.
13212         * stor-layout.c (layout_type): Likewise.
13213         * tree-affine.c (tree_to_aff_combination): Likewise.
13214         * tree-cfg.c (group_case_labels_stmt): Likewise.
13215         * tree-data-ref.c (dr_analyze_indices): Likewise.
13216         (prune_runtime_alias_test_list): Likewise.
13217         * tree-dump.c (dequeue_and_dump): Likewise.
13218         * tree-inline.c (remap_gimple_op_r, copy_tree_body_r): Likewise.
13219         * tree-predcom.c (is_inv_store_elimination_chain): Likewise.
13220         * tree-pretty-print.c (dump_generic_node): Likewise.
13221         * tree-scalar-evolution.c (iv_can_overflow_p): Likewise.
13222         (simple_iv_with_niters): Likewise.
13223         * tree-ssa-address.c (addr_for_mem_ref): Likewise.
13224         * tree-ssa-ccp.c (ccp_finalize, evaluate_stmt): Likewise.
13225         * tree-ssa-loop-ivopts.c (constant_multiple_of): Likewise.
13226         * tree-ssa-loop-niter.c (split_to_var_and_offset)
13227         (refine_value_range_using_guard, number_of_iterations_ne_max)
13228         (number_of_iterations_lt_to_ne, number_of_iterations_lt)
13229         (get_cst_init_from_scev, record_nonwrapping_iv)
13230         (scev_var_range_cant_overflow): Likewise.
13231         * tree-ssa-phiopt.c (minmax_replacement): Likewise.
13232         * tree-ssa-pre.c (compute_avail): Likewise.
13233         * tree-ssa-sccvn.c (vn_reference_fold_indirect): Likewise.
13234         (vn_reference_maybe_forwprop_address, valueized_wider_op): Likewise.
13235         * tree-ssa-structalias.c (get_constraint_for_ptr_offset): Likewise.
13236         * tree-ssa-uninit.c (is_pred_expr_subset_of): Likewise.
13237         * tree-ssanames.c (set_nonzero_bits, get_nonzero_bits): Likewise.
13238         * tree-switch-conversion.c (collect_switch_conv_info, array_value_type)
13239         (dump_case_nodes, try_switch_expansion): Likewise.
13240         * tree-vect-loop-manip.c (vect_gen_vector_loop_niters): Likewise.
13241         (vect_do_peeling): Likewise.
13242         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13243         * tree-vect-stmts.c (vectorizable_load): Likewise.
13244         * tree-vrp.c (compare_values_warnv, vrp_int_const_binop): Likewise.
13245         (zero_nonzero_bits_from_vr, ranges_from_anti_range): Likewise.
13246         (extract_range_from_binary_expr_1, adjust_range_with_scev): Likewise.
13247         (overflow_comparison_p_1, register_edge_assert_for_2): Likewise.
13248         (is_masked_range_test, find_switch_asserts, maybe_set_nonzero_bits)
13249         (vrp_evaluate_conditional_warnv_with_ops, intersect_ranges): Likewise.
13250         (range_fits_type_p, two_valued_val_range_p, vrp_finalize): Likewise.
13251         (evrp_dom_walker::before_dom_children): Likewise.
13252         * tree.c (cache_integer_cst, real_value_from_int_cst, integer_zerop)
13253         (integer_all_onesp, integer_pow2p, integer_nonzerop, tree_log2)
13254         (tree_floor_log2, tree_ctz, mem_ref_offset, tree_int_cst_sign_bit)
13255         (tree_int_cst_sgn, get_unwidened, int_fits_type_p): Likewise.
13256         (get_type_static_bounds, num_ending_zeros, drop_tree_overflow)
13257         (get_range_pos_neg): Likewise.
13258         * ubsan.c (ubsan_expand_ptr_ifn): Likewise.
13259         * config/darwin.c (darwin_mergeable_constant_section): Likewise.
13260         * config/aarch64/aarch64.c (aapcs_vfp_sub_candidate): Likewise.
13261         * config/arm/arm.c (aapcs_vfp_sub_candidate): Likewise.
13262         * config/avr/avr.c (avr_fold_builtin): Likewise.
13263         * config/bfin/bfin.c (bfin_local_alignment): Likewise.
13264         * config/msp430/msp430.c (msp430_attr): Likewise.
13265         * config/nds32/nds32.c (nds32_insert_attributes): Likewise.
13266         * config/powerpcspe/powerpcspe-c.c
13267         (altivec_resolve_overloaded_builtin): Likewise.
13268         * config/powerpcspe/powerpcspe.c (rs6000_aggregate_candidate)
13269         (rs6000_expand_ternop_builtin): Likewise.
13270         * config/rs6000/rs6000-c.c
13271         (altivec_resolve_overloaded_builtin): Likewise.
13272         * config/rs6000/rs6000.c (rs6000_aggregate_candidate): Likewise.
13273         (rs6000_expand_ternop_builtin): Likewise.
13274         * config/s390/s390.c (s390_handle_hotpatch_attribute): Likewise.
13276 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13278         * tree-vect-loop-manip.c (rename_variables_in_bb): Rename PHI nodes
13279         when copying loop nest with only one inner loop.
13281 2017-10-10  Richard Biener  <rguenther@suse.de>
13283         * tree-cfgcleanup.c (cleanup_tree_cfg_noloop): Avoid compacting
13284         blocks if SCEV is active.
13285         * tree-scalar-evolution.c (analyze_scalar_evolution_1): Remove
13286         dead code.
13287         (analyze_scalar_evolution): Handle cached evolutions the obvious way.
13288         (scev_initialize): Assert we are not yet initialized.
13290 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13292         * tree-loop-distribution.c (generate_loops_for_partition): Remove
13293         inner loop's exit stmt by making it always exit the loop, otherwise
13294         we would generate an infinite empty loop.
13296 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13298         * tree-vect-loop-manip.c (slpeel_tree_duplicate_loop_to_edge_cfg): Skip
13299         renaming variables in new preheader if it's deleted.
13301 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13303         * tree-loop-distribution.c (struct partition): Remove unused field
13304         loops of the structure.
13305         (partition_alloc, partition_free): Ditto.
13306         (build_rdg_partition_for_vertex): Ditto.
13308 2017-10-09  Jeff Law  <law@redhat.com>
13310         * targhooks.c (default_stack_clash_protection_final_dynamic_probe): Fix
13311         return type to match prototype and documentation.
13313 2010-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13315         * config/rs6000/rs6000.c (processor_costs): Move to ...
13316         * config/rs6000/rs6000.h: ... here.
13317         (rs6000_cost): Declare.
13319 2017-10-09  Eric Botcazou  <ebotcazou@adacore.com>
13321         * except.c (setjmp_fn): New global variable.
13322         (init_eh): Initialize it if DONT_USE_BUILTIN_SETJMP is defined.
13323         (sjlj_emit_function_enter): Call it instead of BUILTIN_SETJMP
13324         if DONT_USE_BUILTIN_SETJMP is defined.
13326 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13328         * target.def (insn_cost): New hook.
13329         * doc/tm.texi.in (TARGET_INSN_COST): New hook.
13330         * doc/tm.texi: Regenerate.
13331         * rtlanal.c (insn_cost): Use the new hook.
13333 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13335         * combine.c (combine_validate_cost): Compute the new insn_cost,
13336         not just pattern_cost.
13337         (try_combine): Adjust comment.
13339 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13341         * cfgrtl.c (rtl_account_profile_record): Replace insn_rtx_cost with
13342         insn_cost.
13343         * combine.c (uid_insn_cost): Adjust comment.
13344         (combine_validate_cost): Adjust comment.  Use pattern_cost instead
13345         of insn_rtx_cost
13346         (combine_instructions): Use insn_cost instead of insn_rtx_cost.
13347         * dse.c (find_shift_sequence): Ditto.
13348         * ifcvt.c (cheap_bb_rtx_cost_p): Ditto.
13349         (bb_valid_for_noce_process_p): Use pattern_cost.
13350         * rtl.h (insn_rtx_cost): Delete.
13351         (pattern_cost): New prototype.
13352         (insn_cost): New prototype.
13353         * rtlanal.c (insn_rtx_cost): Rename to...
13354         (pattern_cost): ... this.
13355         (insn_cost): New.
13357 2017-10-09  Uros Bizjak  <ubizjak@gmail.com>
13359         * config/i386/i386.md (*jcc_2): Remove insn pattern.
13360         (*jcc<mode>_0_r_i387): Ditto.
13361         (*jccxf_r_i387): Ditto.
13362         (*jcc<mode>_r_i387): Ditto.
13363         (*jccu<mode>_r_i387): Ditto.
13364         (*jcc<X87MODEF:mode>_<SWI24:mode>_r_i387): Ditto.
13365         (*jcc): Rename from *jcc_1.
13367 2017-10-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
13369         * config/rs6000/rs6000-p8swap.c (rs6000_analyze_swaps): Process
13370         deferred rescans after the lvx/stvx recombination pre-pass.
13372 2017-10-09  Michael Meissner  <meissner@linux.vnet.ibm.com>
13374         * config/rs6000/amo.h: New include file to provide ISA 3.0 atomic
13375         memory operation instruction support.
13376         * config.gcc (powerpc*-*-*): Include amo.h as an extra header.
13377         (rs6000-ibm-aix[789]*): Likewise.
13378         * doc/extend.texi (PowerPC Atomic Memory Operation Functions):
13379         Document new functions.
13381 2017-10-09  Richard Biener  <rguenther@suse.de>
13383         PR tree-optimization/82397
13384         * tree-data-ref.c (data_ref_compare_tree): Make sure to return
13385         equality only for semantically equal trees.
13387 2017-10-09  Richard Biener  <rguenther@suse.de>
13389         PR tree-optimization/82449
13390         * sese.c (scev_analyzable_p): Check whether the SCEV is linear.
13391         * tree-chrec.h (evolution_function_is_constant_p): Adjust to
13392         allow constant addresses.
13393         * tree-chrec.c (scev_is_linear_expression): Constant evolutions
13394         are linear.
13396 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13398         * config/s390/s390-builtins.def (vec_nabs, vec_vfi): Fix builtin
13399         flags.
13401 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13403         PR target/82463
13404         * config/s390/vecintrin.h (vec_madd, vec_msub): Fix macro
13405         definitions.
13407 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13409         PR target/82465
13410         * config/s390/s390-builtins.def (vec_sqrt): Fix builtin flags.
13412 2017-10-09  Jakub Jelinek  <jakub@redhat.com>
13414         PR target/82464
13415         * config/s390/s390-builtins.def (s390_vec_xor_flt_a,
13416         s390_vec_xor_flt_b, s390_vec_xor_flt_c): New.
13418 2017-10-09  Richard Sandiford  <richard.sandiford@linaro.org>
13420         * wide-int.h (WI_BINARY_OPERATOR_RESULT): New macro.
13421         (WI_BINARY_PREDICATE_RESULT): Likewise.
13422         (wi::binary_traits::operator_result): New type.
13423         (wi::binary_traits::predicate_result): Likewise.
13424         (generic_wide_int::operator~, unary generic_wide_int::operator-)
13425         (generic_wide_int::operator==, generic_wide_int::operator!=)
13426         (generic_wide_int::operator&, generic_wide_int::and_not)
13427         (generic_wide_int::operator|, generic_wide_int::or_not)
13428         (generic_wide_int::operator^, generic_wide_int::operator+
13429         (binary generic_wide_int::operator-, generic_wide_int::operator*):
13430         Delete.
13431         (operator~, unary operator-, operator==, operator!=, operator&)
13432         (operator|, operator^, operator+, binary operator-, operator*): New
13433         functions.
13434         * expr.c (get_inner_reference): Use wi::bit_and_not.
13435         * fold-const.c (fold_binary_loc): Likewise.
13436         * ipa-prop.c (ipa_compute_jump_functions_for_edge): Likewise.
13437         * tree-ssa-ccp.c (get_value_from_alignment): Likewise.
13438         (bit_value_binop): Likewise.
13439         * tree-ssa-math-opts.c (find_bswap_or_nop_load): Likewise.
13440         * tree-vrp.c (zero_nonzero_bits_from_vr): Likewise.
13441         (extract_range_from_binary_expr_1): Likewise.
13442         (masked_increment): Likewise.
13443         (simplify_bit_ops_using_ranges): Likewise.
13445 2017-10-09  Martin Jambor  <mjambor@suse.cz>
13447         PR hsa/82416
13448         * hsa-common.h (hsa_op_with_type): New method extend_int_to_32bit.
13449         * hsa-gen.c (hsa_extend_inttype_to_32bit): New function.
13450         (hsa_type_for_scalar_tree_type): Use it.  Always force min32int for
13451         COMPLEX types.
13452         (hsa_fixup_mov_insn_type): New function.
13453         (hsa_op_with_type::get_in_type): Use it.
13454         (hsa_build_append_simple_mov): Likewise.  Allow sub-32bit
13455         immediates in an assert.
13456         (hsa_op_with_type::extend_int_to_32bit): New method.
13457         (gen_hsa_insns_for_bitfield): Fixup instruction and intermediary
13458         types.  Convert to dest type if necessary.
13459         (gen_hsa_insns_for_bitfield_load): Fixup load type if necessary.
13460         (reg_for_gimple_ssa): Pass false as min32int to
13461         hsa_type_for_scalar_tree_type.
13462         (gen_hsa_addr): Fixup type when creating addresable temporary.
13463         (gen_hsa_cmp_insn_from_gimple): Extend operands if necessary.
13464         (gen_hsa_unary_operation): Extend operands and convert to dest type if
13465         necessary.  Call hsa_fixup_mov_insn_type.
13466         (gen_hsa_binary_operation): Changed operand types to hsa_op_with_type,
13467         extend operands and convert to dest type if necessary.
13468         (gen_hsa_insns_for_operation_assignment): Extend operands and convert
13469         to dest type if necessary.
13470         (set_output_in_type): Call hsa_fixup_mov_insn_type.  Just ude dest
13471         if conversion nt necessary and size matches.
13472         (gen_hsa_insns_for_load): Call hsa_fixup_mov_insn_type, convert
13473         to dest type if necessary.
13474         (gen_hsa_insns_for_store): Call hsa_fixup_mov_insn_type.
13475         (gen_hsa_insns_for_switch_stmt): Likewise. Also extend operands if
13476         necessary.
13477         (gen_hsa_clrsb): Likewise.
13478         (gen_hsa_ffs): Likewise.
13479         (gen_hsa_divmod): Extend operands and convert to dest type if
13480         necessary.
13481         (gen_hsa_atomic_for_builtin): Change type of op to hsa_op_with_type.
13483 2017-10-08  Segher Boessenkool  <segher@kernel.crashing.org>
13485         * config/rs6000/rs6000.md (conditional branch): Clean up formatting.
13486         Remove empty default arguments.  Use a brace block as output
13487         statement.
13488         (conditional return): Ditto.
13489         (jump): Ditto.
13490         (indirect_jump): Ditto.  Use b%T0 instead of bctr/blr.
13491         (tablejump, tablejumpsi, tablejumpdi, *tablejump<mode>_internal1):
13492         Ditto.
13493         (group_ending_nop): Ditto.
13494         (doloop_end): Ditto.
13495         (ctr<mode>, ctr<mode>_internal1, ctr<mode>_internal2): Ditto.
13496         (splitters for those): Ditto.
13498 2017-10-08  Segher Boessenkool  <segher@kernel.crashing.org>
13500         * config/rs6000/rs6000-string.c (expand_strncmp_align_check): Invert
13501         a conditional jump (and the compare for it) so that pc_rtx is the
13502         last operand.
13503         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Adjust
13504         for the deleted and renamed ctr<mode>_internal[234] patterns.
13505         * config/rs6000/rs6000.md: Delete second conditional branch pattern.
13506         Delete second conditional return pattern.
13507         (ctr<mode>_internal2): Delete this second bdnz pattern.
13508         (ctr<mode>_internal3): Rename to ctr<mode>_internal2.
13509         (ctr<mode>_internal4): Delete this second bdz pattern.
13511 2017-10-08  Eric Botcazou  <ebotcazou@adacore.com>
13513         * tree-outof-ssa.h (ssaexpand): Add partitions_for_undefined_values.
13514         (always_initialized_rtx_for_ssa_name_p): New predicate.
13515         * tree-outof-ssa.c (remove_ssa_form): Initialize new field of SA.
13516         (finish_out_of_ssa): Free new field of SA.
13517         * tree-ssa-coalesce.h (get_undefined_value_partitions): Declare.
13518         * tree-ssa-coalesce.c: Include tree-ssa.h.
13519         (get_parm_default_def_partitions): Remove extern keyword.
13520         (get_undefined_value_partitions): New function.
13521         * expr.c (expand_expr_real_1) <expand_decl_rtl>: For a SSA_NAME, do
13522         not set SUBREG_PROMOTED_VAR_P on the sub-register if it may contain
13523         uninitialized bits.
13524         * loop-iv.c (iv_get_reaching_def): Disqualify all subregs.
13526 2017-10-08  Eric Botcazou  <ebotcazou@adacore.com>
13528         * builtins.def (BUILT_IN_SETJMP): Revert latest change.
13530 2017-10-08  Jan Hubicka  <hubicka@ucw.cz>
13532         * config/i386/i386.c (ix86_expand_set_or_movmem): Disable 512bit loops
13533         for targets that preffer 128bit.
13535 2017-10-08  Jan Hubicka  <hubicka@ucw.cz>
13537         * config/i386/i386.c (has_dispatch): Disable for Ryzen.
13539 2017-10-08  Olivier Hainque  <hainque@adacore.com>
13541         * config/arm/arm.c (arm_set_return_address): Use MEM_VOLATILE_P
13542         on the target mem instead of RTX_FRAME_RELATED_P on the insn to
13543         prevent DSE.
13544         (thumb_set_return_address): Likewise.
13546 2017-10-08  Olivier Hainque  <hainque@adacore.com>
13548         * common/config/arm/arm-common.c (arm_except_unwind_info):
13549         Handle DWARF2_UNWIND_INFO.
13551 2017-10-07  Michael Collison <michael.collison@arm.com>
13553         * config/aarch64/aarch64.md (*aarch64_reg_<optab>_minus<mode>3):
13554         New pattern.
13556 2017-10-07  Eric Botcazou  <ebotcazou@adacore.com>
13558         * builtins.def (BUILT_IN_SETJMP): Declare as library builtin instead
13559         of GCC builtin if DONT_USE_BUILTIN_SETJMP is defined.
13560         * except.c (sjlj_emit_function_enter): If DONT_USE_BUILTIN_SETJMP is
13561         defined, force the creation of a new block for a dispatch label.
13563 2017-10-07  Jan Hubicka  <hubicka@ucw.cz>
13565         * invoke.texi (Wsuggest-attribute=cold): Document.
13566         * common.opt (Wsuggest-attribute=cold): New
13567         * ipa-pure-const.c (warn_function_cold): New function.
13568         * predict.c (compute_function_frequency): Use it.
13569         * predict.h (warn_function_cold): Declare.
13571 2017-10-06  Jan Hubicka  <hubicka@ucw.cz>
13573         * tree-switch-conversion.c (do_jump_if_equal, emit_cmp_and_jump_insns):
13574         Update profile.
13576 2017-10-06  Martin Liska  <mliska@suse.cz>
13578         * sanopt.c (struct sanopt_tree_triplet_hash): Remove inline
13579         keyword for member functions.
13580         (struct sanopt_tree_couple): New struct.
13581         (struct sanopt_tree_couple_hash): New function.
13582         (struct sanopt_ctx): Add new hash_map.
13583         (has_dominating_ubsan_ptr_check): New function.
13584         (record_ubsan_ptr_check_stmt): Likewise.
13585         (maybe_optimize_ubsan_ptr_ifn): Likewise.
13586         (sanopt_optimize_walker): Handle IFN_UBSAN_PTR.
13587         (pass_sanopt::execute): Handle also SANITIZE_POINTER_OVERFLOW.
13589 2017-10-06  Sudakshina Das  <sudi.das@arm.com>
13591         PR target/82440
13592         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Only call
13593         aarch64_simd_valid_immediate on CONST_VECTORs.
13594         (aarch64_reg_or_bic_imm): Likewise.
13596 2017-10-06  Wilco Dijkstra  <wdijkstr@arm.com>
13598         PR rtl-optimization/82396
13599         * haifa-sched.c (ready_sort_real): Disable qsort checking.
13601 2017-10-06  Sebastian Pop  <sebpop@gmail.com>
13603         * graphite-dependences.c (scop_get_reads): Move code to...
13604         (scop_get_must_writes): Move code to...
13605         (scop_get_may_writes): Move code to...
13606         (scop_get_reads_and_writes): ... here.
13607         (scop_get_dependences): Call scop_get_reads_and_writes.
13609 2017-10-06  Jakub Jelinek  <jakub@redhat.com>
13611         PR tree-optimization/82434
13612         * fold-const.h (can_native_encode_type_p,
13613         can_native_encode_string_p): Remove.
13614         * fold-const.c (native_encode_int): Formatting fixes.  If ptr is NULL,
13615         don't encode anything, just return what would be otherwise returned.
13616         (native_encode_fixed, native_encode_complex, native_encode_vector):
13617         Likewise.
13618         (native_encode_string): Likewise.  Inline by hand
13619         can_native_encode_string_p.
13620         (can_native_encode_type_p): Remove.
13621         (can_native_encode_string_p): Remove.
13622         * tree-vect-stmts.c (vectorizable_store): Instead of testing just
13623         STRING_CSTs using can_native_encode_string_p, test all
13624         CONSTANT_CLASS_P values using native_encode_expr with NULL ptr.
13625         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Remove last
13626         argument from native_encode_expr.
13627         (rhs_valid_for_store_merging_p): Use native_encode_expr with NULL ptr.
13628         (pass_store_merging::execute): Don't unnecessarily look for 3 stmts,
13629         but just 2.
13631 2017-10-06  Richard Biener  <rguenther@suse.de>
13633         PR tree-optimization/82397
13634         * tree-vect-data-refs.c (dr_group_sort_cmp): Do not use
13635         operand_equal_p but rely on data_ref_compare_tree for detecting
13636         equalities.
13637         (vect_analyze_data_ref_accesses): Use data_ref_compare_tree
13638         to match up with dr_group_sort_cmp.
13640 2017-10-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13642         PR target/82322
13643         * config/s390/s390-builtins.def (s390_vfi): Define new overloaded
13644         builtin.
13645         * config/s390/s390-builtin-types.def: Regenerate.
13647 2017-10-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13649         PR target/82317
13650         * config/s390/s390-builtin-types.def: Regenerate.
13651         * config/s390/s390-builtins.def (s390_vfmaxdb_4, s390_vfmindb_4):
13652         Change flag from B_VXE to B_VX.
13653         (s390_vec_min_dbl): Remove B_VXE flag.
13655 2017-10-06  Richard Biener  <rguenther@suse.de>
13657         * graphite-isl-ast-to-gimple.c: Include ssa.h and tree-ssa.h.
13658         (translate_isl_ast_to_gimple::translate_pending_phi_nodes,
13659         translate_isl_ast_to_gimple::is_valid_rename,
13660         translate_isl_ast_to_gimple::get_rename,
13661         translate_isl_ast_to_gimple::get_def_bb_for_const,
13662         translate_isl_ast_to_gimple::get_new_name,
13663         translate_isl_ast_to_gimple::collect_all_ssa_names,
13664         translate_isl_ast_to_gimple::copy_loop_phi_args,
13665         translate_isl_ast_to_gimple::collect_all_ssa_names,
13666         translate_isl_ast_to_gimple::copy_loop_phi_args,
13667         translate_isl_ast_to_gimple::copy_loop_phi_nodes,
13668         translate_isl_ast_to_gimple::add_close_phis_to_merge_points,
13669         translate_isl_ast_to_gimple::add_close_phis_to_outer_loops,
13670         translate_isl_ast_to_gimple::copy_loop_close_phi_args,
13671         translate_isl_ast_to_gimple::copy_loop_close_phi_nodes,
13672         translate_isl_ast_to_gimple::copy_cond_phi_args,
13673         translate_isl_ast_to_gimple::copy_cond_phi_nodes,
13674         translate_isl_ast_to_gimple::edge_for_new_close_phis,
13675         translate_isl_ast_to_gimple::add_phi_arg_for_new_expr,
13676         translate_isl_ast_to_gimple::rename_uses,
13677         translate_isl_ast_to_gimple::rename_all_uses): Remove.
13678         (translate_isl_ast_to_gimple::get_rename_from_scev): Simplify.
13679         (set_rename_for_each_def): Likewise.
13680         (graphite_copy_stmts_from_block): Handle debug stmt resetting
13681         here.  Handle rewriting SCEV analyzable uses here.
13682         (copy_bb_and_scalar_dependences): Generate code for PHI
13683         copy-in/outs.
13684         (graphite_regenerate_ast_isl): Adjust.
13685         * graphite-scop-detection.c (trivially_empty_bb_p): Move to sese.[ch].
13686         (add_write, add_read): New functions.
13687         (build_cross_bb_scalars_def): Use it and simplify.
13688         (build_cross_bb_scalars_use): Likewise.
13689         (graphite_find_cross_bb_scalar_vars): Inline into...
13690         (try_generate_gimple_bb): ...here.  Add dependences for PHIs,
13691         simulating out-of-SSA.  Compute liveout and add dependencies.
13692         (build_scops): Force an empty entry block.
13693         * sese.h (sese_info_t::liveout, sese_info_t::debug_liveout): New
13694         members.
13695         (sese_build_liveouts): Declare.
13696         (sese_trivially_empty_bb_p): Likewise.
13697         * sese.c (sese_build_liveouts_bb): Properly handle PHIs,
13698         compute liveout and debug_liveout.
13699         (sese_bad_liveouts_use): Remove.
13700         (sese_reset_debug_liveouts_bb): Likewise.
13701         (sese_reset_debug_liveouts): Rewrite in terms of debug_liveout.
13702         (sese_build_liveouts): Build liveout and debug_liveout and store
13703         it in region.
13704         (new_sese_info): Adjust.
13705         (free_sese_info): Likewise.
13706         (sese_insert_phis_for_liveouts): Reset debug stmts from here,
13707         do not build liveout here.
13708         (move_sese_in_condition): Adjust region entry.
13709         (scev_analyzable_p): Match up with chrec_apply requirements.
13710         (sese_trivially_empty_bb_p): New.
13711         * tree-into-ssa.c (get_reaching_def): Properly support generating
13712         default-defs for incremental rewrite of anonymous names.
13714 2017-10-06  Richard Biener  <rguenther@suse.de>
13716         * graphite-sese-to-poly.c (extract_affine): For casts increasing
13717         precision do not perform modulo reduction.
13719 2017-10-06  Richard Biener  <rguenther@suse.de>
13721         PR tree-optimization/82436
13722         * tree-vect-slp.c (vect_supported_load_permutation_p): More
13723         conservatively choose the vectorization factor when checking
13724         whether we can perform the required load permutation.
13725         (vect_transform_slp_perm_load): Assert when we may not fail.
13727 2017-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
13729         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Correct error
13730         message for incompatible -msdata=* and -mcall-* options.
13732 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13734         * config/i386/i386.c (ia32_multipass_dfa_lookahead): Default to issue
13735         rate for post-reload scheduling.
13737 2017-10-05  Tamar Christina  <tamar.christina@arm.com>
13739         * doc/sourcebuild.texi (vect_sizes_16B_8B, vect_sizes_32B_16B): New.
13741 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13743         * config/i386/i386.c (znver1_cost): Set branch_cost to 3 (instead of 2)
13744         to improve monte carlo in scimark.
13746 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13748         * config/i386/i386.c (ix86_size_cost, i386_cost, i486_cost,
13749         pentium_cost, lakemont_cost, pentiumpro_cost, geode_cost, k6_cost,
13750         athlon_cost, k8_cost, amdfam10_cost, btver1_cost, btver2_cost,
13751         pentium4_cost, nocona_cost): Set reassociation width to 1.
13752         (bdver1_cost, bdver2_cost, bdver3_cost, bdver4_cost): Set reassociation
13753         width to 2 for fp operations and 1 otherwise.
13754         (znver1_cost): Set scalar reassoc width to 4 and vector to 3 and 6
13755         for int and fp.
13756         (atom_cost): Set reassociation width to 2.
13757         (slm_cost, generic_cost): Set fp reassociation width
13758         to 2 and 1 otherwise.
13759         (intel_cost): Set fp reassociation width to 4 and 1 otherwise.
13760         (core_cost): Set fp reassociation width to 4 and vector to 2.
13761         (ix86_reassociation_width): Rewrite using cost table; special case
13762         plus/minus on Zen; honor X86_TUNE_SSE_SPLIT_REGS
13763         and TARGET_AVX128_OPTIMAL.
13764         * config/i386/i386.h (processor_costs): Add
13765         reassoc_int, reassoc_fp, reassoc_vec_int, reassoc_vec_fp.
13766         (TARGET_VECTOR_PARALLEL_EXECUTION, TARGET_REASSOC_INT_TO_PARALLEL,
13767         TARGET_REASSOC_FP_TO_PARALLEL): Remove.
13768         * x86-tune.def (X86_TUNE_REASSOC_INT_TO_PARALLEL): Remove.
13769         (X86_TUNE_REASSOC_FP_TO_PARALLEL): Remove.
13770         (X86_TUNE_VECTOR_PARALLEL_EXECUTION):  Remove.
13772 2017-10-05  Nathan Sidwell  <nathan@acm.org>
13774         * doc/invoke.texi (Wparentheses): Document C++ MVP behaviour.
13776 2017-10-05  Tamar Christina  <tamar.christina@arm.com>
13778         * config/arm/arm.c (arm_test_fpu_data): New.
13779         (arm_run_selftests): Call arm_test_fpu_data.
13781 2017-10-04  Nathan Sidwell  <nathan@acm.org>
13783         * toplev.c (toplev::main): Remove excess parens on pretty_printer
13784         decl.
13785         * caller-save.c (insert_save): Remove excess parens on TO_SAVE parm.
13787 2017-10-04  Sudakshina Das  <sudi.das@arm.com>
13789         * config/aarch64/aarch64-protos.h (enum simd_immediate_check): New
13790         check type for aarch64_simd_valid_immediate.
13791         (aarch64_output_simd_mov_immediate): Update prototype.
13792         (aarch64_simd_valid_immediate): Update prototype.
13793         * config/aarch64/aarch64-simd.md (orr<mode>3): modified pattern to add
13794         support for ORR-immediate.
13795         (and<mode>3): modified pattern to add support for BIC-immediate.
13796         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Function
13797         now checks for valid immediate for BIC and ORR based on new enum
13798         argument.
13799         (aarch64_output_simd_mov_immediate): Function now used to output
13800         BIC/ORR imm as well based on new enum argument.
13801         * config/aarch64/constraints.md (Do): New vector immediate constraint.
13802         (Db) : Likewise.
13803         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): New predicate.
13804         (aarch64_reg_or_bic_imm): Likewise.
13806 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13808         * config/s390/vx-builtins.md ("vec_mergeh<mode>")
13809         ("vec_mergel<mode>"): Change mode iterator to V_128_NOSINGLE.
13811 2017-10-04  Wilco Dijkstra  <wdijkstr@arm.com>
13813         Revert r253399:
13815         PR rtl-optimization/82396
13816         * haifa-sched.c (autopref_multipass_init): Simplify
13817         initialization.
13818         (autopref_rank_data): Simplify sort order.
13819         * sched-int.h (autopref_multipass_data_): Remove
13820         multi_mem_insn_p, min_offset and max_offset.
13822 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13824         * doc/sourcebuild.texi: Document vect_peeling_profitable.
13826 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13828         * doc/sourcebuild.texi: Document vect_intdouble_cvt and
13829         vect_doubleint_cvt.
13831 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13833         * doc/sourcebuild.texi: Document vect_long_mult.
13835 2017-10-04  Richard Sandiford  <richard.sandiford@linaro.org>
13837         PR tree-optimization/82413
13838         * fold-const.c (build_range_check): Use widest_int when comparing
13839         the maximum ETYPE value with HIGH.
13841 2017-10-04  Wilco Dijkstra  <wdijkstr@arm.com>
13843         PR rtl-optimization/82396
13844         * haifa-sched.c (autopref_multipass_init): Simplify
13845         initialization.
13846         (autopref_rank_data): Simplify sort order.
13847         * sched-int.h (autopref_multipass_data_): Remove
13848         multi_mem_insn_p, min_offset and max_offset.
13850 2017-10-04  Jakub Jelinek  <jakub@redhat.com>
13852         PR tree-optimization/82381
13853         * tree-ssa-reassoc.c (sort_by_operand_rank): Check for different
13854         oeN->rank first.  Return 1 or -1 if one op is SSA_NAME and the other
13855         is not.
13857         PR tree-optimization/82374
13858         * omp-low.c (create_omp_child_function): Copy DECL_ATTRIBUTES,
13859         DECL_FUNCTION_SPECIFIC_OPTIMIZATION,
13860         DECL_FUNCTION_SPECIFIC_TARGET and DECL_FUNCTION_VERSIONED from
13861         current_function_decl to the new decl.
13863 2017-10-03  Michael Meissner  <meissner@linux.vnet.ibm.com>
13865         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_2_HW): Define new
13866         helper macro for IEEE float128 hardware built-in functions.
13867         (SQRTF128_ODD): Add built-in functions with the round-to-odd
13868         semantics.
13869         (TRUNCF128_ODD): Likewise.
13870         (ADDF128_ODD): Likewise.
13871         (SUBF128_ODD): Likewise.
13872         (MULF128_ODD): Likewise.
13873         (DIVF128_ODD): Likewise.
13874         (FMAF128_ODD): Likewise.
13875         * config/rs6000/rs6000.md (UNSPEC_ROUND_TO_ODD): Rename to
13876         UNSPEC_TRUNC_ROUND_TO_ODD.
13877         (UNSPEC_TRUNC_ROUND_TO_ODD): Likewise.
13878         (UNSPEC_ADD_ROUND_TO_ODD): New unspec codes for the IEEE 128-bit
13879         floating point round to odd instructions.
13880         (UNSPEC_SUB_ROUND_TO_ODD): Likewise.
13881         (UNSPEC_MUL_ROUND_TO_ODD): Likewise.
13882         (UNSPEC_DIV_ROUND_TO_ODD): Likewise.
13883         (UNSPEC_FMA_ROUND_TO_ODD): Likewise.
13884         (UNSPEC_SQRT_ROUND_TO_ODD): Likewise.
13885         (trunc<mode>sf2_hw): Change the truncate with round to odd
13886         expansion to use UNSPEC_TRUNC_ROUND_TO_ODD.
13887         (add<mode>3_odd): Add insns for IEEE 128-bit floating point round
13888         to odd hardware instructions.
13889         (sub<mode>3_odd): Likewise.
13890         (mul<mode>3_odd): Likewise.
13891         (div<mode>3_odd): Likewise.
13892         (sqrt<mode>2_odd): Likewise.
13893         (fma<mode>4_odd): Likewise.
13894         (fms<mode>4_odd): Likewise.
13895         (nfma<mode>4_odd): Likewise.
13896         (nfms<mode>4_odd): Likewise.
13897         (trunc<mode>df2_odd): Change the truncate with round to odd
13898         expansion to use UNSPEC_TRUNC_ROUND_TO_ODD.  Add a generator
13899         function.
13900         * doc/extend.texi (PowerPC built-in functions): Update documentation
13901         for existing IEEE float128-bit built-in functions.  Add built-in
13902         functions that generate the IEEE 128-bit floating point round to
13903         odd instructions.
13905 2017-10-03  Segher Boessenkool  <segher@kernel.crashing.org>
13907         PR rtl-optimization/77729
13908         * simplify-rtx.c (simplify_binary_operation_1): Delete the (X&C1)|C2
13909         to (X&(C1&~C2))|C2 transformations.
13911 2017-10-03  Martin Jambor  <mjambor@suse.cz>
13913         PR tree-optimization/82363
13914         * tree-sra.c (propagate_subaccesses_across_link): In unrecoverable
13915         mismatch, mark lacc written regardless of racc.
13917 2017-10-03  Jakub Jelinek  <jakub@redhat.com>
13919         PR tree-optimization/82381
13920         * tree-ssa-reassoc.c (sort_by_operand_rank): Don't check
13921         stmt_to_insert nor wheather SSA_NAMEs are default defs.
13922         Return 1 or -1 if one of bba and bbb is NULL. If bb_rank is equal,
13923         fallthrough into reassoc_stmt_dominates_stmt_p.
13925         PR target/82386
13926         * combine.c (combine_instructions): Don't combine in unreachable
13927         basic blocks.
13929 2017-10-02  Peter Bergner  <bergner@vnet.ibm.com>
13931         PR target/80210
13932         * config/rs6000/rs6000.c (rs6000_option_override_internal): Rewrite
13933         function to not use the have_cpu variable.  Do not set cpu_index,
13934         rs6000_cpu_index or rs6000_tune_index if we end up using TARGET_DEFAULT
13935         or the default cpu.
13936         (rs6000_valid_attribute_p): Remove duplicate initializations of
13937         old_optimize and func_optimize.
13938         (rs6000_pragma_target_parse): Call rs6000_activate_target_options ().
13939         (rs6000_activate_target_options): Make global.
13940         * config/rs6000/rs6000-protos.h (rs6000_activate_target_options): Add
13941         prototype.
13943 2017-10-02  Jakub Jelinek  <jakub@redhat.com>
13945         * tree-dfa.c (get_ref_base_and_extent): Set *pmax_size to -1
13946         if *poffset + *pmax_size overflows in HOST_WIDE_INT.
13947         Set *poffset to 0 and *psize and *pmax_size to -1 if
13948         *poffset + *psize overflows in HOST_WIDE_INT.
13950         PR tree-optimization/82387
13951         PR tree-optimization/82388
13952         PR tree-optimization/82389
13953         * tree-ssa-dse.c (dse_classify_store): Test byte_tracking_enabled
13954         instead of live_bytes non-NULL.
13956 2017-10-02  Georg-Johann Lay  <avr@gjlay.de>
13958         PR target/41076
13959         * confg/avr/avr.md (*iorhi3.ashift8-ext.zerox): Add "r,r,0"
13960         alternative.
13962 2017-10-02  Richard Biener  <rguenther@suse.de>
13964         * graphite-isl-ast-to-gimple.c (set_codegen_error): With
13965         -fchecking and --param graphite-allow-codegen-errors=0 ICE.
13966         * params.def (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): New param.
13968 2017-10-02  Richard Sandiford  <richard.sandiford@linaro.org>
13970         * tree.h (wi::int_traits <const_tree>::decompose): Assert that the
13971         requested precision matches the type's.
13972         * calls.c (alloc_max_size): Calculate the new candidate size as
13973         a widest_int and use wi::to_widest when comparing it with the
13974         current candidate size.
13975         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Compare with
13976         zero rather than integer_zero_node.
13977         * match.pd: Check for a no-op conversion before using wi::add
13978         rather than after.  Use tree_to_uhwi when summing small shift
13979         counts into an unsigned int.
13981 2017-10-02  Richard Sandiford  <richard.sandiford@linaro.org>
13982             Alan Hayward  <alan.hayward@arm.com>
13983             David Sherwood  <david.sherwood@arm.com>
13985         PR target/71307
13986         * config/aarch64/aarch64.h (POINTER_AND_FP_REGS): New reg class.
13987         (REG_CLASS_NAMES, REG_CLASS_CONTENTS): Update accordingly.
13988         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
13989         POINTER_AND_FP_REGS.
13991 2017-10-02  Richard Biener  <rguenther@suse.de>
13993         PR tree-optimization/82355
13994         * graphite-isl-ast-to-gimple.c (build_iv_mapping): Also build
13995         a mapping for the enclosing loop but avoid generating one for
13996         the loop tree root.
13997         (copy_bb_and_scalar_dependences): Remove premature codegen
13998         error on PHIs in blocks duplicated into multiple places.
13999         * graphite-scop-detection.c
14000         (scop_detection::stmt_has_simple_data_refs_p): For a loop not
14001         in the region use it as loop and nest to analyze the DR in.
14002         (try_generate_gimple_bb): Likewise.
14003         * graphite-sese-to-poly.c (extract_affine_chrec): Adjust.
14004         (add_loop_constraints): For blocks in a loop not in the region
14005         create a dimension with a single iteration.
14006         * sese.h (gbb_loop_at_index): Remove assert.
14008 2017-10-01  Kevin Buettner  <kevinb@redhat.com>
14010         * omp-expand.c (adjust_context_scope): New function.
14011         (expand_parallel_call): Call adjust_context_scope.
14013 2017-10-01  Jeff Law  <law@redhat.com>
14015         * tree-ssa-dom.c (optimize_stmt): Make this a method within the
14016         dom_opt_dom_walker class with direct access to private members.
14017         Add comments.  Call test_for_singularity.
14018         (dom_opt_dom_walker::before_dom_children): Corresponding changes.
14019         (dom_opt_dom_walker::after_dom_children): Do not lazily initialize
14020         m_dummy_cond anymore.
14021         (class dom_opt_dom_walker): Initialize m_dummy_cond member in the
14022         class ctor.
14023         (pass_dominator:execute): Build the dummy_cond here and pass it
14024         to the dom_opt_dom_walker ctor.
14025         (test_for_singularity): New function.
14027 2017-09-30  Krister Walfridsson  <krister.walfridsson@gmail.com>
14028             Maya Rashish  <coypu@sdf.org>
14030         * config.gcc (*-*-netbsd*): New variable nbsd_tm_file containing
14031         netbsd.h, netbsd-stdint.h, and netbsd-elf.h.
14032         (alpha*-*-netbsd*) Use nbsd_tm_file.
14033         (arm*-*-netbsdelf*) Likewise.
14034         (i[34567]86-*-netbsdelf*) Likewise.
14035         (x86_64-*-netbsd*) Likewise.
14036         (mips*-*-netbsd*) Likewise.
14037         (powerpc-*-netbsd*) Likewise.
14038         (sh*-*-netbsd*) Likewise.
14039         (sparc-*-netbsdelf*) Likewise.
14040         (sparc64-*-netbsd*) Likewise.
14041         (m68k*-*-netbsdelf*) Use nbsd_tm_file and add CHAR_FAST8/SHORT_FAST16
14042         to tm_defines.
14043         (vax-*-netbsdelf*) Likewise.
14044         * config/netbsd-stdint.h (INT_FAST8_TYPE): Check CHAR_FAST8.
14045         (UINT_FAST8_TYPE) Likewise.
14046         (INT_FAST16_TYPE) Check CHAR_FAST16.
14047         (UINT_FAST16_TYPE) Likewise.
14049 2017-09-30  Jakub Jelinek  <jakub@redhat.com>
14051         PR target/82361
14052         * config/i386/i386.md
14053         (TARGET_USE_8BIT_IDIV zext divmodsi4 splitter): New define_split.
14054         (divmodsi4_zext_1, divmodsi4_zext_2, *divmodsi4_zext_1,
14055         *divmodsi4_zext_2): New define_insn_and_split.
14056         (*divmodsi4_noext_zext_1, *divmodsi4_noext_zext_2): New define_insn.
14057         (TARGET_USE_8BIT_IDIV zext udivmodsi4 splitter): New define_split.
14058         (udivmodsi4_zext_1, udivmodsi4_zext_2, *udivmodsi4_zext_1,
14059         *udivmodsi4_zext_2, *udivmodsi4_pow2_zext_1, *udivmodsi4_pow2_zext_2):
14060         New define_insn_and_split.
14061         (*udivmodsi4_noext_zext_1, *udivmodsi4_noext_zext_2): New define_insn.
14062         * config/i386/i386.c (ix86_split_idivmod): Handle operands[0] or
14063         operands[1] having DImode when mode is SImode.
14065         * config/i386/i386.c (ix86_split_idivmod): Use mode instead of
14066         always SImode for DIV and MOD in REG_EQUAL notes.
14068 2017-09-29  Yury Gribov  <tetra2005@gmail.com>
14070         PR middle-end/82319
14071         * match.pd: Fix handling of NaNs in pattern.
14073 2017-09-29  Jeff Law  <law@redhat.com>
14075         * sbitmap.c (bitmap_bit_in_range_p): New function.
14076         * sbitmap.h (bitmap_bit_in_range_p): Prototype.
14077         * tree-ssa-dse.c (live_bytes_read): New function.
14078         (dse_classify_store): Ignore reads of dead bytes.
14080         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Fix
14081         typos and whitespace errors.
14082         * config/i386/predicates.md (address_no_seg_operand): Likewise.
14083         * config/s390/s390.c (s390_emit_prologue): Likewise.
14085 2017-09-29  Vladimir Makarov  <vmakarov@redhat.com>
14087         PR target/81481
14088         * ira-costs.c (scan_one_insn): Don't take into account PIC equiv
14089         with a symbol for LRA.
14091 2017-09-29  Vladimir Makarov  <vmakarov@redhat.com>
14093         PR rtl-optimization/82338
14094         * lra-constraints.c (inherit_in_ebb): Check usage_insns check.
14096 2017-09-29  Alexander Monakov  <amonakov@ispras.ru>
14098         * genmodes.c (calc_wider_mode): Suppress qsort macro.
14099         * system.h [CHECKING_P] (qsort): Redirect to qsort_chk.
14100         (qsort_chk): Declare.
14101         * vec.c [CHECKING_P] (qsort_chk_error): New static function.
14102         (qsort_chk): New function.
14104 2017-09-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
14106         PR tree-optimization/82337
14107         * gimple-ssa-strength-reduction.c (find_phi_def): Don't record a
14108         phi definition if the PHI result appears in an abnormal PHI.
14109         (find_basis_for_base_expr): Don't record a basis if the LHS of the
14110         basis appears in an abnormal PHI.
14112 2017-09-29  Richard Biener  <rguenther@suse.de>
14114         * graphite-isl-ast-to-gimple.c
14115         (translate_isl_ast_to_gimple::set_codegen_error): New function.
14116         (binary_op_to_tree): Use it.
14117         (get_rename_from_scev): Likewise.
14118         (copy_loop_phi_nodes): Likewise.
14119         (copy_bb_and_scalar_dependences): Likewise.
14120         (translate_pending_phi_nodes): Likewise.
14122 2017-09-29  Jakub Jelinek  <jakub@redhat.com>
14124         PR target/82339
14125         * config/i386/i386.md (*movdi_internal peephole2): New -Os peephole
14126         for movabsq $(i32 << shift), r64.
14128 2017-09-28  Uros Bizjak  <ubizjak@gmail.com>
14130         * config/i386/i386.c (ix86_print_operand_address_as): Do not check
14131         index when encoding %esp as %rsp to avoid 0x67 prefix.
14133 2017-09-28  Sergey Shalnov  <Sergey.Shalnov@intel.com>
14135         * config/i386/i386.md (*movsf_internal, *movdf_internal):
14136         Return 256-bit AVX modes for TARGET_PREFER_AVX256.
14138 2017-09-28  Thomas Preud'homme  <thomas.preudhomme@arm.com>
14140         * config/arm/arm.c (arm_option_override): Forbid ARMv8-M Security
14141         Extensions with more than 16 double VFP registers.
14142         (cmse_nonsecure_entry_clear_before_return): Remove second entry of
14143         to_clear_mask and all code related to it.  Replace the remaining
14144         entry by a sbitmap and adapt code accordingly.
14146 2017-09-28  Henry Linjamäki  <henry.linjamaki@parmance.com>
14148         * brig-builtins.def: Change pure attributes to const.
14150 2017-09-28  Joseph Myers  <joseph@codesourcery.com>
14152         * config.gcc (default_gnu_indirect_function): Default to yes for
14153         sparc*-*-linux* with glibc.
14155 2017-09-28  Joseph Myers  <joseph@codesourcery.com>
14157         * config/aarch64/aarch64.c (aarch64_elf_asm_constructor)
14158         (aarch64_elf_asm_destructor): Pass SECTION_NOTYPE to get_section
14159         when creating .init_array and .fini_array sections with priority
14160         specified.
14162 2017-09-27  Christophe Lyon  <christophe.lyon@linaro.org>
14164         PR target/71727
14165         * config/aarch64/aarch64.c
14166         (aarch64_builtin_support_vector_misalignment): Always return false
14167         when misalignment is unknown.
14169 2017-09-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14171         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Revise
14172         this function to return false if the definition used by the swap
14173         instruction is artificial, or if the memory address from which the
14174         constant value is loaded is not represented by a base address held
14175         in a register or if the base address register is a frame or stack
14176         pointer.  Additionally, return false if the base address of the
14177         loaded constant is a SYMBOL_REF but is not considered to be a
14178         constant.
14179         (replace_swapped_load_constant): New function.
14180         (rs6000_analyze_swaps): Add a new pass to replace a swap of a
14181         loaded constant vector with a load of a swapped constant vector.
14183 2017-09-27  Carl Love  <cel@us.ibm.com>
14185         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
14186         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
14187         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
14188         fctiw instruction.
14190 2017-09-27  Alexander Monakov  <amonakov@ispras.ru>
14192         * haifa-sched.c (autopref_rank_for_schedule): Order 'irrelevant' insns
14193         first, always call autopref_rank_data otherwise.
14195 2017-09-27  Richard Biener  <rguenther@suse.de>
14197         * graphite-scop-detection.c (find_scop_parameters): Move
14198         loop bound handling ...
14199         (gather_bbs::before_dom_children): ... here, avoiding the need
14200         to build scop_info->loop_nest.
14201         (record_loop_in_sese): Remove.
14202         * sese.h (sese_info_t::loop_nest): Remove.
14203         * sese.c (new_sese_info): Do not allocate loop_nest.
14204         (free_sese_info): Do not free loop_nest.
14206 2017-09-27  Jakub Jelinek  <jakub@redhat.com>
14208         PR c++/82159
14209         * gimplify.c (gimplify_modify_expr): Don't optimize away zero sized
14210         lhs from calls if the lhs has addressable type.
14212 2017-09-27  Richard Biener  <rguenther@suse.de>
14214         * graphite.h (scop::max_alias_set): New member.
14215         * graphite-scop-detection.c: Remove references to non-existing
14216         --param in comments.
14217         (build_alias_sets): Record the maximum alias set used for drs.
14218         (build_scops): Support zero as unlimited for
14219         --param graphite-max-arrays-per-scop.
14220         * graphite-sese-to-poly.c (add_scalar_version_numbers): Remove
14221         and inline into ...
14222         (build_poly_sr_1): ... here.  Compute alias set based on the
14223         maximum alias set used for drs rather than
14224         PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP
14226 2017-09-27  Richard Biener  <rguenther@suse.de>
14228         * graphite-optimize-isl.c (get_schedule_for_node_st): Allow
14229         --param loop-block-tile-size=0 to disable tiling.
14231 2017-09-27  Richard Biener  <rguenther@suse.de>
14233         * doc/invoke.texi (graphite-max-bbs-per-function): Remove.
14234         (graphite-max-nb-scop-params): Document special value zero.
14235         * domwalk.h (dom_walker::STOP): New symbolical constant.
14236         (dom_walker::dom_walker): Add optional parameter for bb to
14237         RPO mapping.
14238         (dom_walker::~dom_walker): Declare.
14239         (dom_walker::before_dom_children): Document STOP return value.
14240         (dom_walker::m_user_bb_to_rpo): New member.
14241         (dom_walker::m_bb_to_rpo): Likewise.
14242         * domwalk.c (dom_walker::dom_walker): Compute bb to RPO
14243         mapping here if not provided by the user.
14244         (dom_walker::~dom_walker): Free bb to RPO mapping if not
14245         provided by the user.
14246         (dom_walker::STOP): Define.
14247         (dom_walker::walk): Do not compute bb to RPO mapping here.
14248         Support STOP return value from before_dom_children to stop
14249         walking.
14250         * graphite-optimize-isl.c (optimize_isl): If the schedule
14251         is the same still generate code if -fgraphite-identity
14252         or -floop-parallelize-all are given.
14253         * graphite-scop-detection.c: Include cfganal.h.
14254         (gather_bbs::gather_bbs): Get and pass through bb to RPO
14255         mapping.
14256         (gather_bbs::before_dom_children): Return STOP for BBs
14257         not in the region.
14258         (build_scops): Compute bb to RPO mapping and pass it to
14259         the domwalk.  Treat --param graphite-max-nb-scop-params=0
14260         as not limiting the number of params.
14261         * graphite.c (graphite_initialize): Remove limit on the
14262         number of basic-blocks in a function.
14263         * params.def (PARAM_GRAPHITE_MAX_BBS_PER_FUNCTION): Remove.
14264         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Adjust to documented
14265         default value of 10.
14267 2017-09-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
14269         * config/rs6000/vsx.md (peephole for optimizing move SF to GPR):
14270         Adjust code to eliminate needing to do the shift right 32-bits
14271         operation after XSCVDPSPN.
14273 2017-09-26  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
14275         * match.pd ((X / Y) == 0 -> X < Y): New pattern.
14276         ((X / Y) != 0 -> X >= Y): Likewise.
14278 2017-09-26  Carl Love  <cel@us.ibm.com>
14280         * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_XL_LEN_R,
14281         P9V_BUILTIN_VEC_XST_LEN_R): Add support for builtins
14282         vector unsigned char vec_xl_len_r (unsigned char *, size_t);
14283         void vec_xst_len_r (vector unsigned char, unsigned char *, size_t);
14284         * config/rs6000/altivec.h (vec_xl_len_r, vec_xst_len_r): Add defines.
14285         * config/rs6000/rs6000-builtin.def (XL_LEN_R, XST_LEN_R): Add
14286         definitions and overloading.
14287         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case
14288         statement for P9V_BUILTIN_XST_LEN_R.
14289         (altivec_init_builtins): Add def_builtin for P9V_BUILTIN_STXVLL.
14290         * config/rs6000/vsx.md (lxvll, stxvll, xl_len_r, xst_len_r): Add
14291         define_expand and define_insn for the instructions and builtins.
14292         * doc/extend.texi: Update the built-in documentation file for the new
14293         built-in functions.
14294         * config/rs6000/altivec.md (altivec_lvsl_reg, altivec_lvsr_reg): Add
14295         define_insn for the instructions
14297 2017-09-26  Krister Walfridsson  <krister.walfridsson@gmail.com>
14299         PR target/39570
14300         * gcc/config/netbsd-protos.h: New file.
14301         * gcc/config/netbsd.c: New file.
14302         * gcc/config/netbsd.h (SUBTARGET_INIT_BUILTINS): Define.
14303         * gcc/config/t-netbsd: New file.
14304         * gcc/config.gcc (tm_p_file): Add netbsd-protos.h.
14305         (tmake_file) Add t-netbsd.
14306         (extra_objs) Add netbsd.o.
14308 2017-09-26  Janus Weil  <janus@gcc.gnu.org>
14310         PR fortran/82143
14311         PR fortran/82324
14312         * doc/sourcebuild.texi: Document fortran_real_10 and fortran_real_16.
14314 2017-09-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
14316         * config/rs6000/rs6000.md (extendsi<mode>2): Add a splitter to do
14317         sign extension from a vector register to a GPR by doing a 32-bit
14318         direct move and then an EXTSW.
14319         (extendsi<mode>2 splitter): Likewise.
14320         (movsi_from_sf): Adjust code to eliminate doing a 32-bit shift
14321         right or vector extract after doing XSCVDPSPN.  Use
14322         zero_extendsidi2 instead of p8_mfvsrd_4_disf to move the value to
14323         the GPRs.
14324         (movdi_from_sf_zero_ext): Likewise.
14325         (reload_gpr_from_vsxsf): Likewise.
14326         (p8_mfvsrd_4_disf): Delete, no longer used.
14327         (movsi_from_df): Optimize converting a DFmode to a SFmode, and
14328         then needing to move the SFmode to a GPR to use the XSCVDPSP
14329         instruction instead of FRSP and XSCVDPSPN.
14330         * config/rs6000/vsx.md (vsx_xscvspdp_scalar2): Move insn so that
14331         it is adjacent to the other XSCVSPDP insns.
14332         (vsx_xscvdpsp_scalar): Use "ww" constraint instead of "f" to allow
14333         SFmode to be in traditional Altivec registers.
14334         (vsx_xscvdpspn): Eliminate useless alternative constraint.
14335         (vsx_xscvspdpn): Likewise.
14336         (vsx_xscvspdpn_scalar): Likewise.
14338 2017-09-26  Martin Jambor  <mjambor@suse.cz>
14340         * tree-sra.c (compare_access_positions): Put integral types first,
14341         stabilize sorting of integral types, remove conditions putting
14342         non-full-precision integers last.
14343         (sort_and_splice_var_accesses): Disable scalarization if a
14344         non-integert would be represented by a non-full-precision integer.
14346 2017-09-26  Joseph Myers  <joseph@codesourcery.com>
14348         * config/microblaze/linux.h (TARGET_ASM_FILE_END): Likewise.
14349         * config/pa/pa.h (NEED_INDICATE_EXEC_STACK): Likewise.
14350         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Likewise.
14351         * config/pa/pa.c (pa_hpux_file_end): Rename to pa_file_end.
14352         Define unconditionally, with [ASM_OUTPUT_EXTERNAL_REAL]
14353         conditionals inside the function instead of around it.  Call
14354         file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK.
14355         (TARGET_ASM_FILE_END): Define unconditionally to pa_file_end.
14357 2017-09-26  Richard Biener  <rguenther@suse.de>
14359         * graphite-scop-detection.c (scop_detection::build_scop_depth): Rewrite,
14360         fold in ...
14361         (scop_detection::build_scop_breadth): ... this.  Removed.
14362         (scop_detection::loop_is_valid_in_scop): Fold into single caller.
14363         (scop_detection::harmful_stmt_in_bb): Likewise.
14364         (scop_detection::graphite_can_represent_stmt): Likewise.
14365         (scop_detection::loop_body_is_valid_scop): Likewise.  Remove recursion.
14366         (scop_detection::can_represent_loop): Remove recursion, fold in ...
14367         (scop_detection::can_represent_loop_1): ... this.  Removed.
14368         (scop_detection::harmful_loop_in_region): Simplify after inlining
14369         the above and remove more quadraticness.
14370         (build_scops): Adjust.
14371         * tree-data-ref.c (loop_nest_has_data_refs): Remove pointless
14372         quadraticness.
14374 2017-09-26  Jakub Jelinek  <jakub@redhat.com>
14376         PR target/82267
14377         * config/i386/i386.c (ix86_print_operand_address_as): Only test
14378         REGNO (base) == SP_REG if base is a REG.
14380         PR middle-end/35691
14381         * tree-ssa-reassoc.c (update_range_test): Dump r->exp each time
14382         if it is different SSA_NAME.
14383         (optimize_range_tests_cmp_bitwise): New function.
14384         (optimize_range_tests): Call it.
14386 2017-09-26  Richard Biener  <rguenther@suse.de>
14388         PR tree-optimization/82321
14389         * graphite.c (canonicalize_loop_closed_ssa): Properly check
14390         for the def being inside the loop.
14392 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14394         * config/s390/vx-builtins.md ("vmslg"): Add missing operand in
14395         assembler output.
14396         * config/s390/s390-builtins.def: Fix constraint on op4.
14398 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14400         * config/s390/s390.c (s390_expand_vec_compare): Use the new mode
14401         independent expanders.
14402         * config/s390/vector.md ("vec_cmpuneq", "vec_cmpltgt")
14403         ("vec_ordered", "vec_unordered"): New expanders.
14405 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14407         * config/s390/s390.c (s390_preferred_simd_mode): Return V4SFmode
14408         for SFmode.
14410 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14412         * config/s390/vector.md ("vec_unpacks_low_v16qi"): Rename to
14413         vec_unpacks_lo_v16qi.
14414         ("vec_unpacku_low_v16qi"): Rename to vec_unpacku_lo_v16qi.
14416 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14418         * config/s390/vector.md ("vec_unpacks_lo_v4sf")
14419         ("vec_unpacks_hi_v4sf", "vec_unpacks_lo_v2df")
14420         ("vec_unpacks_hi_v2df", "vec_pack_trunc_v2df"): New expanders.
14422 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14424         * config/s390/predicates.md ("const_shift_by_byte_operand"): New
14425         predicate.
14426         * config/s390/vector.md ("*vec_srb<mode>"): Change modes to V_128
14427         and V16QI.
14428         ("*vec_slb<mode>"): New insn pattern.
14429         ("vec_shr_<mode>"): New expander.
14430         * config/s390/vx-builtins.md ("vec_slb<mode>"): Turn into expander
14431         and force the shift count operand to V16QImode.
14432         ("vec_srb<mode>"): Set shift count mode to V16QI.
14434 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14436         * config/s390/vector.md ("vec_widen_umult_lo_<mode>")
14437         ("vec_widen_umult_hi_<mode>", "vec_widen_smult_lo_<mode>")
14438         ("vec_widen_smult_hi_<mode>"): New expander definitions.
14440 2017-09-26  Richard Earnshaw  <rearnsha@arm.com>
14442         PR target/82175
14443         * config/arm/arm.h (DRIVER_SELF_SPECS): Separate sub-rules with commas.
14445 2017-09-26  Richard Biener  <rguenther@suse.de>
14447         PR tree-optimization/82320
14448         * tree-ssa-sccvn.c (set_ssa_val_to): Changing undef to undef
14449         isn't a change.
14451 2017-09-25  Jeff Law  <law@redhat.com>
14453         * config/rs6000/rs6000-protos.h (output_probe_stack_range): Update
14454         prototype for new argument.
14455         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack_1): New function,
14456         mostly extracted from rs6000_emit_allocate_stack.
14457         (rs6000_emit_probe_stack_range_stack_clash): New function.
14458         (rs6000_emit_allocate_stack): Call
14459         rs6000_emit_probe_stack_range_stack_clash as needed.
14460         (rs6000_emit_probe_stack_range): Add additional argument
14461         to call to gen_probe_stack_range{si,di}.
14462         (output_probe_stack_range): New.
14463         (output_probe_stack_range_1): Renamed from output_probe_stack_range.
14464         (output_probe_stack_range_stack_clash): New.
14465         (rs6000_emit_prologue): Emit notes into dump file as requested.
14466         * rs6000.md (allocate_stack): Handle -fstack-clash-protection.
14467         (probe_stack_range<P:mode>): Operand 0 is now early-clobbered.
14468         Add additional operand and pass it to output_probe_stack_range.
14470 2017-09-25  Bin Cheng  <bin.cheng@arm.com>
14472         PR tree-optimization/82163
14473         * tree-ssa-loop-manip.h (verify_loop_closed_ssa): New parameter.
14474         (checking_verify_loop_closed_ssa): New parameter.
14475         * tree-ssa-loop-manip.c (check_loop_closed_ssa_use): Delete.
14476         (check_loop_closed_ssa_stmt): Delete.
14477         (check_loop_closed_ssa_def, check_loop_closed_ssa_bb): New functions.
14478         (verify_loop_closed_ssa): Check loop closed ssa form for LOOP.
14479         (tree_transform_and_unroll_loop): Check loop closed ssa form only for
14480         changed loops.
14482 2017-09-25  Pekka Jaaskelainen <pekka@parmance.com>
14484         * brig-builtins.def: Treat HSAIL barrier builtins as
14485         setjmp/longjump style functions.
14487 2017-09-25  Richard Sandiford  <richard.sandiford@linaro.org>
14489         * target.def (constant_alignment): New hook.
14490         * defaults.h (CONSTANT_ALIGNMENT): Delete.
14491         * doc/tm.texi.in (CONSTANT_ALIGNMENT): Replace with...
14492         (TARGET_CONSTANT_ALIGNMENT): ...this new hook.
14493         * doc/tm.texi: Regenerate.
14494         * targhooks.h (default_constant_alignment): Declare.
14495         (constant_alignment_word_strings): Likewise.
14496         * targhooks.c (default_constant_alignment): New function.
14497         (constant_alignment_word_strings): Likewise.
14498         * builtins.c (get_object_alignment_2): Use targetm.constant_alignment
14499         instead of CONSTANT_ALIGNMENT.
14500         * varasm.c (align_variable, get_variable_align, build_constant_desc)
14501         (force_const_mem): Likewise.
14502         * config/aarch64/aarch64.h (CONSTANT_ALIGNMENT): Delete.
14503         * config/aarch64/aarch64.c (aarch64_constant_alignment): New function.
14504         (aarch64_classify_address): Call it instead of CONSTANT_ALIGNMENT.
14505         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14506         * config/alpha/alpha.h (CONSTANT_ALIGNMENT): Delete commented-out
14507         definition.
14508         * config/arc/arc.h (CONSTANT_ALIGNMENT): Delete.
14509         * config/arc/arc.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14510         constant_alignment_word_strings.
14511         * config/arm/arm.h (CONSTANT_ALIGNMENT_FACTOR): Delete.
14512         (CONSTANT_ALIGNMENT): Likewise.
14513         * config/arm/arm.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14514         (arm_constant_alignment): New function.
14515         * config/bfin/bfin.h (CONSTANT_ALIGNMENT): Delete.
14516         * config/bfin/bfin.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14517         constant_alignment_word_strings.
14518         * config/cr16/cr16.h (CONSTANT_ALIGNMENT): Delete.
14519         * config/cr16/cr16.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14520         constant_alignment_word_strings.
14521         * config/cris/cris.h (CONSTANT_ALIGNMENT): Delete.
14522         * config/cris/cris.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14523         (cris_constant_alignment): New function.
14524         * config/epiphany/epiphany.h (CONSTANT_ALIGNMENT): Delete.
14525         * config/epiphany/epiphany.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14526         (epiphany_constant_alignment): New function.
14527         * config/fr30/fr30.h (CONSTANT_ALIGNMENT): Delete.
14528         * config/fr30/fr30.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14529         constant_alignment_word_strings.
14530         * config/frv/frv.h (CONSTANT_ALIGNMENT): Delete.
14531         * config/frv/frv.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14532         constant_alignment_word_strings.
14533         * config/ft32/ft32.h (CONSTANT_ALIGNMENT): Delete.
14534         * config/ft32/ft32.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14535         constant_alignment_word_strings.
14536         * config/i386/i386.h (CONSTANT_ALIGNMENT): Delete.
14537         * config/i386/i386-protos.h (ix86_constant_alignment): Delete.
14538         * config/i386/i386.c (ix86_constant_alignment): Make static.
14539         Use the same interface as the target hook.
14540         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14541         * config/ia64/ia64.h (CONSTANT_ALIGNMENT): Delete.
14542         * config/ia64/ia64.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14543         constant_alignment_word_strings.
14544         * config/iq2000/iq2000.h (CONSTANT_ALIGNMENT): Delete.
14545         * config/iq2000/iq2000.c (iq2000_constant_alignment): New function.
14546         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14547         * config/lm32/lm32.h (CONSTANT_ALIGNMENT): Delete.
14548         * config/lm32/lm32.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14549         constant_alignment_word_strings.
14550         * config/m32r/m32r.h (CONSTANT_ALIGNMENT): Delete.
14551         * config/m32r/m32r.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14552         constant_alignment_word_strings.
14553         * config/mcore/mcore.h (CONSTANT_ALIGNMENT): Delete.
14554         * config/mcore/mcore.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14555         constant_alignment_word_strings.
14556         * config/microblaze/microblaze.h (CONSTANT_ALIGNMENT): Delete.
14557         * config/microblaze/microblaze.c (microblaze_constant_alignment):
14558         New function.
14559         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14560         * config/mips/mips.h (CONSTANT_ALIGNMENT): Delete.
14561         * config/mips/mips.c (mips_constant_alignment): New function.
14562         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14563         * config/mmix/mmix.h (CONSTANT_ALIGNMENT): Delete.
14564         * config/mmix/mmix-protos.h (mmix_constant_alignment): Delete.
14565         * config/mmix/mmix.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14566         (mmix_constant_alignment): Make static.  Use the same interface
14567         as the target hook.
14568         * config/moxie/moxie.h (CONSTANT_ALIGNMENT): Delete.
14569         * config/moxie/moxie.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14570         constant_alignment_word_strings.
14571         * config/nios2/nios2.h (CONSTANT_ALIGNMENT): Delete.
14572         * config/nios2/nios2.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14573         constant_alignment_word_strings.
14574         * config/pa/pa.h (CONSTANT_ALIGNMENT): Delete.
14575         * config/pa/pa.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14576         constant_alignment_word_strings.
14577         * config/powerpcspe/powerpcspe.h (CONSTANT_ALIGNMENT): Delete.
14578         * config/powerpcspe/powerpcspe.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14579         (rs6000_constant_alignment): New function.
14580         * config/riscv/riscv.h (CONSTANT_ALIGNMENT): Delete.
14581         * config/riscv/riscv.c (riscv_constant_alignment): New function.
14582         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14583         * config/rs6000/rs6000.h (CONSTANT_ALIGNMENT): Delete.
14584         * config/rs6000/rs6000.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14585         (rs6000_constant_alignment): New function.
14586         * config/s390/s390.h (CONSTANT_ALIGNMENT): Delete.
14587         * config/s390/s390.c (s390_constant_alignment): New function.
14588         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14589         * config/sh/sh.h (CONSTANT_ALIGNMENT): Delete.
14590         * config/sh/sh.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14591         constant_alignment_word_strings.
14592         * config/sparc/sparc.h (CONSTANT_ALIGNMENT): Delete.
14593         * config/sparc/sparc.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14594         (sparc_constant_alignment): New function.
14595         * config/spu/spu.h (CONSTANT_ALIGNMENT): Delete.
14596         * config/spu/spu.c (spu_constant_alignment): New function.
14597         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14598         * config/stormy16/stormy16.h (CONSTANT_ALIGNMENT): Delete.
14599         * config/stormy16/stormy16.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14600         constant_alignment_word_strings.
14601         * config/tilegx/tilegx.h (CONSTANT_ALIGNMENT): Delete.
14602         * config/tilegx/tilegx.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14603         constant_alignment_word_strings.
14604         * config/tilepro/tilepro.h (CONSTANT_ALIGNMENT): Delete.
14605         * config/tilepro/tilepro.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14606         constant_alignment_word_strings.
14607         * config/visium/visium.h (CONSTANT_ALIGNMENT): Delete.
14608         * config/visium/visium.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14609         (visium_constant_alignment): New function.
14610         * config/xtensa/xtensa.h (CONSTANT_ALIGNMENT): Delete.
14611         * config/xtensa/xtensa.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14612         (xtensa_constant_alignment): New function.
14613         * system.h (CONSTANT_ALIGNMENT): Poison.
14615 2017-09-25  Will Schmidt  <will_schmidt@vnet.ibm.com>
14617         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
14618         for early folding of vector stores (ALTIVEC_BUILTIN_ST_*).
14619         (rs6000_builtin_valid_without_lhs): New helper function.
14620         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
14621         Remove obsoleted code for handling ALTIVEC_BUILTIN_VEC_ST.
14623 2017-09-25  Richard Sandiford  <richard.sandiford@linaro.org>
14625         * target.h (vec_perm_indices): Use unsigned short rather than
14626         unsigned char.
14627         (auto_vec_perm_indices): Likewise.
14628         * config/aarch64/aarch64.c (aarch64_vectorize_vec_perm_const_ok):
14629         Use unsigned int rather than unsigned char.
14630         * config/arm/arm.c (arm_vectorize_vec_perm_const_ok): Likewise.
14632 2017-09-25  Richard Biener  <rguenther@suse.de>
14634         * cfgloop.h (sort_sibling_loops): Declare.
14635         * cfgloop.c (sort_sibling_loops_cmp): New helper.
14636         (sort_sibling_loops): New function sorting the sibling loop list
14637         in RPO order.
14638         * graphite.c (graphite_transform_loops): Sort sibling loops.
14640 2017-09-25  Richard Sandiford  <richard.sandifird@linaro.org>
14642         * target.def (vec_perm_const_ok): Change sel parameter to
14643         vec_perm_indices.
14644         * optabs-query.c (can_vec_perm_p): Update accordingly.
14645         * doc/tm.texi: Regenerate.
14646         * config/aarch64/aarch64.c (expand_vec_perm_d): Change perm to
14647         auto_vec_perm_indices and remove separate nelt field.
14648         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
14649         (aarch64_evpc_ext, aarch64_evpc_rev, aarch64_evpc_dup)
14650         (aarch64_evpc_tbl, aarch64_expand_vec_perm_const_1)
14651         (aarch64_expand_vec_perm_const): Update accordingly.
14652         (aarch64_vectorize_vec_perm_const_ok): Likewise.  Change sel
14653         to vec_perm_indices.
14654         * config/arm/arm.c (expand_vec_perm_d): Change perm to
14655         auto_vec_perm_indices and remove separate nelt field.
14656         (arm_evpc_neon_vuzp, arm_evpc_neon_vzip, arm_evpc_neon_vrev)
14657         (arm_evpc_neon_vtrn, arm_evpc_neon_vext, arm_evpc_neon_vtbl)
14658         (arm_expand_vec_perm_const_1, arm_expand_vec_perm_const): Update
14659         accordingly.
14660         (arm_vectorize_vec_perm_const_ok): Likewise.  Change sel
14661         to vec_perm_indices.
14662         * config/i386/i386.c (ix86_vectorize_vec_perm_const_ok): Change
14663         sel to vec_perm_indices.
14664         * config/ia64/ia64.c (ia64_vectorize_vec_perm_const_ok): Likewise.
14665         * config/mips/mips.c (mips_vectorize_vec_perm_const_ok): Likewise.
14666         * config/powerpcspe/powerpcspe.c (rs6000_vectorize_vec_perm_const_ok):
14667         Likewise.
14668         * config/rs6000/rs6000.c (rs6000_vectorize_vec_perm_const_ok):
14669         Likewise.
14671 2017-09-25  Pierre-Marie de Rodat  <derodat@adacore.com>
14673         PR debug/82155
14674         * dwarf2out.c (dwarf2out_early_global_decl): Call dwarf2out_decl
14675         on the FUNCTION_DECL function context if it has a DIE that is a
14676         declaration.
14678 2017-09-25  Richard Biener  <rguenther@suse.de>
14680         PR tree-optimization/82285
14681         * tree-vect-patterns.c (vect_recog_bool_pattern): Also handle
14682         enumeral types.
14684 2017-09-25  Tom de Vries  <tom@codesourcery.com>
14686         PR target/80035
14687         PR target/81069
14688         * config/nvptx/nvptx.c (nvptx_output_call_insn): Add exit after call to
14689         noreturn function.
14691 2017-09-25  Richard Biener  <rguenther@suse.de>
14693         * graphite-optimize-isl.c (optimize_isl): Fail and dump if
14694         ISL errors other than isl_error_quota happen.  Dump if the
14695         schedule is the same.
14696         * graphite-sese-to-poly.c (build_poly_scop): Fail on ISL
14697         errors instead of aborting inside ISL.
14699 2017-09-25  Iain Sandoe  <iain@codesourcery.com>
14701         PR target/80556
14702         * config/i386/darwin.h (REAL_LIB_SPEC): New; put libSystem ahead
14703         of libgcc_eh for m64.
14704         * config/i386/darwin64.h: Likewise.
14706 2017-09-25  Richard Biener  <rguenther@suse.de>
14708         PR middle-end/82144
14709         * dwarf2out.c (gen_enumeration_type_die): Do not add alignment
14710         attribute for incomplete types nor twice for complete ones.
14712 2017-09-24  Uros Bizjak  <ubizjak@gmail.com>
14714         PR target/82267
14715         * config/i386/i386.c (ix86_print_operand_address_as): Encode
14716         %esp as %rsp to avoid 0x67 prefix if there is no index or base
14717         register.
14719 2017-09-23  Uros Bizjak  <ubizjak@gmail.com>
14721         PR bootstrap/82306
14722         * config/i386/i386.opt (mprefer-avx256): Use
14723         ix86_target_flags variable.
14724         * config/i386/i386.c (ix86_target_string): Move
14725         -mprefer-avx256 to flag2_opts.
14727 2017-09-22  Jakub Jelinek  <jakub@redhat.com>
14729         PR middle-end/35691
14730         * match.pd: Simplify x == -1 & y == -1 into (x & y) == -1
14731         and x != -1 | y != -1 into (x & y) != -1.
14733 2017-09-22  Steve Ellcey  <sellcey@cavium.com>
14735         * config.gcc: Add new case statement to set
14736         default_gnu_indirect_function.  Remove it from x86_64-*-linux*,
14737         i[34567]86-*, powerpc*-*-linux*spe*, powerpc*-*-linux*, s390-*-linux*,
14738         s390x-*-linux* case statements.   Added aarch64 to the list of
14739         supported architectures.
14741 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14743         PR tree-optimization/82289
14744         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs): Check
14745         STMT_VINFO_RELEVANT_P.
14747 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14748             Alan Hayward  <alan.hayward@arm.com>
14749             David Sherwood  <david.sherwood@arm.com>
14751         * tree-vrp.c (extract_range_from_multiplicative_op_1): Assert
14752         for VR_RANGE only; don't allow VR_ANTI_RANGE.
14753         (extract_range_from_binary_expr_1): Don't call
14754         extract_range_from_multiplicative_op_1 if !range_int_cst_p.
14756 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14757             Alan Hayward  <alan.hayward@arm.com>
14758             David Sherwood  <david.sherwood@arm.com>
14760         * target.def (preferred_vector_alignment): New hook.
14761         * doc/tm.texi.in (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): New
14762         hook.
14763         * doc/tm.texi: Regenerate.
14764         * targhooks.h (default_preferred_vector_alignment): Declare.
14765         * targhooks.c (default_preferred_vector_alignment): New function.
14766         * tree-vectorizer.h (dataref_aux): Add a target_alignment field.
14767         Expand commentary.
14768         (DR_TARGET_ALIGNMENT): New macro.
14769         (aligned_access_p): Update commentary.
14770         (vect_known_alignment_in_bytes): New function.
14771         * tree-vect-data-refs.c (vect_calculate_required_alignment): New
14772         function.
14773         (vect_compute_data_ref_alignment): Set DR_TARGET_ALIGNMENT.
14774         Calculate the misalignment based on the target alignment rather than
14775         the vector size.
14776         (vect_update_misalignment_for_peel): Use DR_TARGET_ALIGMENT
14777         rather than TYPE_ALIGN / BITS_PER_UNIT to update the misalignment.
14778         (vect_enhance_data_refs_alignment): Mask the byte misalignment with
14779         the target alignment, rather than masking the element misalignment
14780         with the number of elements in a vector.  Also use the target
14781         alignment when calculating the maximum number of peels.
14782         (vect_find_same_alignment_drs): Use vect_calculate_required_alignment
14783         instead of TYPE_ALIGN_UNIT.
14784         (vect_duplicate_ssa_name_ptr_info): Remove stmt_info parameter.
14785         Measure DR_MISALIGNMENT relative to DR_TARGET_ALIGNMENT.
14786         (vect_create_addr_base_for_vector_ref): Update call accordingly.
14787         (vect_create_data_ref_ptr): Likewise.
14788         (vect_setup_realignment): Realign by ANDing with
14789         -DR_TARGET_MISALIGNMENT.
14790         * tree-vect-loop-manip.c (vect_gen_prolog_loop_niters): Calculate
14791         the number of peels based on DR_TARGET_ALIGNMENT.
14792         * tree-vect-stmts.c (get_group_load_store_type): Compare the gap
14793         with the guaranteed alignment boundary when deciding whether
14794         overrun is OK.
14795         (vectorizable_mask_load_store): Interpret DR_MISALIGNMENT
14796         relative to DR_TARGET_ALIGNMENT instead of TYPE_ALIGN_UNIT.
14797         (ensure_base_align): Remove stmt_info parameter.  Get the
14798         target base alignment from DR_TARGET_ALIGNMENT.
14799         (vectorizable_store): Update call accordingly.   Interpret
14800         DR_MISALIGNMENT relative to DR_TARGET_ALIGNMENT instead of
14801         TYPE_ALIGN_UNIT.
14802         (vectorizable_load): Likewise.
14804 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14805             Alan Hayward  <alan.hayward@arm.com>
14806             David Sherwood  <david.sherwood@arm.com>
14808         * tree-vectorizer.h (vect_get_scalar_dr_size): New function.
14809         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Use it.
14810         (vect_enhance_data_refs_alignment): Likewise.
14812 2017-09-22  Richard Earnshaw  <richard.earnshaw@arm.com>
14814         * config/arm/parsecpu.awk (fatal): Note that we've encountered an
14815         error.  Only quit immediately if parsing is complete.
14816         (BEGIN): Initialize fatal_err and parse_done.
14817         (begin fpu, end fpu): Check number of arguments.
14818         (begin arch, end arch): Likewise.
14819         (begin cpu, end cpu): Likewise.
14820         (cname, tune for, tune flags, architecture, fpu, option): Likewise.
14821         (optalias): Likewise.
14823 2017-09-22  Richard Earnshaw  <richard.earnshaw@arm.com>
14825         * config.gcc (arm*-*-*): Don't add arm-isa.h to tm_p_file.
14826         * config/arm/arm-isa.h: Delete.  Move definitions to ...
14827         * arm-cpus.in: ... here.  Use new feature and fgroup values.
14828         * config/arm/arm.c (arm_option_override): Use lower case for feature
14829         bit names.
14830         * config/arm/arm.h (TARGET_HARD_FLOAT): Likewise.
14831         (TARGET_VFP3, TARGET_VFP5, TARGET_FMA): Likewise.
14832         * config/arm/parsecpu.awk (END): Add new command 'isa'.
14833         (isa_pfx): Delete.
14834         (print_isa_bits_for): New function.
14835         (gen_isa): New function.
14836         (gen_comm_data): Use print_isa_bits_for.
14837         (define feature): New keyword.
14838         (define fgroup): New keyword.
14839         * config/arm/t-arm (TM_H): Remove.
14840         (GTM_H): Add arm-isa.h.
14841         (arm-isa.h): Add rule to generate file.
14842         * common/config/arm/arm-common.c: (arm_canon_arch_option): Use lower
14843         case for feature bit names.
14845 2017-09-22  Richard Biener  <rguenther@suse.de>
14847         * graphite-isl-ast-to-gimple.c (graphite_verify): Inline into
14848         single caller.
14849         (graphite_regenerate_ast_isl): Do not reset SCEV.  Move debug
14850         print of no dependency loops ...
14851         * graphite.c (graphite_transform_loops): ... here.
14852         (canonicalize_loop_closed_ssa_form): Work from inner to outer
14853         loops.
14854         (same_close_phi_node, remove_duplicate_close_phi,
14855         make_close_phi_nodes_unique, defined_in_loop_p): Fold into ...
14856         (canonicalize_loop_closed_ssa): ... here and simplify.
14857         * graphite-optimize-isl.c: Include tree-vectorizer.h.
14858         (optimize_isl): Use dump_printf_loc to tell when we stopped
14859         optimizing because of an ISL timeout.
14861 2017-09-22  Richard Biener  <rguenther@suse.de>
14863         PR tree-optimization/82291
14864         * tree-if-conv.c (predicate_mem_writes): Make sure to
14865         remove writes in blocks predicated with false.
14867 2017-09-22  Richard Biener  <rguenther@suse.de>
14869         * sese.c: Include cfganal.h.
14870         (if_region_set_false_region): Remove.
14871         (create_if_region_on_edge): Likewise.
14872         (move_sese_in_condition): Re-implement without destroying
14873         dominators.
14875 2017-09-22  Richard Biener  <rguenther@suse.de>
14877         * graphite-isl-ast-to-gimple.c (translate_pending_phi_nodes):
14878         Verify both BBs contain loop PHI nodes before dispatching to
14879         copy_loop_phi_args.
14880         (graphite_regenerate_ast_isl): Do not recompute dominators,
14881         do not verify three times.  Restructure for clarity.
14882         * graphite-scop-detection.c (same_close_phi_node,
14883         remove_duplicate_close_phi, make_close_phi_nodes_unique,
14884         defined_in_loop_p, canonicalize_loop_closed_ssa,
14885         canonicalize_loop_closed_ssa_form): Simplify, remove excess
14886         checking and SSA rewrite, move to ...
14887         * graphite.c: ... here.  Include ssa.h and tree-ssa-loop-manip.h.
14888         (graphite_initialize): Do not pass in ctx, do not reset the
14889         SCEV cache, compute only dominators.
14890         (graphite_transform_loops): Allocate ISL ctx after
14891         graphite_initialize.  Call canonicalize_loop_closed_ssa_form.
14892         Maintain post-dominators only around build_scops.
14893         * sese.c (if_region_set_false_region): Make static.  Free
14894         and recompute dominators.
14895         (move_sese_in_condition): Assert we don't get called with
14896         post-dominators computed.
14897         * sese.h (if_region_set_false_region): Remove.
14899 2017-09-22  Sergey Shalnov  <sergey.shalnov@intel.com>
14901         * config/i386/sse.md ("mov<mode>_internal"): Use <sseinsnmode>
14902         mode attribute for TARGET_AVX512VL.
14904 2017-09-21  Sergey Shalnov  <sergey.shalnov@intel.com>
14906         * config/i386/i386.opt (mprefer-avx256): New option.
14907         * config/i386/i386.c (ix86_target_string): Add -mprefer-avx256
14908         to flag_opts.
14909         (ix86_preferred_simd_mode): Return 256-bit AVX modes
14910         for TARGET_PREFER_AVX256.
14911         * doc/invoke.texi (x86 Options): Document -mprefer-avx256.
14913 2017-09-21  Jeff Law  <law@redhat.com>
14915         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash):
14916         Fix dump output if the only stack space is for pushed registers.
14918 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14920         * config/spu/spu.c (spu_sched_adjust_cost): Update after renaming
14921         of insn_cost.
14923 2017-09-21  Martin Sebor  <msebor@redhat.com>
14925         PR c/81882
14926         * doc/extend.texi (attribute ifunc): Avoid relying on ill-formed
14927         code (in C++) or code that triggers warnings.
14929 2017-09-21  Eric Botcazou  <ebotcazou@adacore.com>
14931         * stor-layout.c (bit_from_pos): Do not distribute the conversion.
14933 2017-09-21  Segher Boessenkool  <segher@kernel.crashing.org>
14935         * haifa-sched.c: Rename insn_cost to insn_sched_cost.
14936         * sched-rgn.c: Ditto.
14937         * sel-sched-ir.c: Ditto.
14939 2017-09-21  Alexander Monakov  <amonakov@ispras.ru>
14941         * toplev.h (set_random_seed): Adjust return type.
14942         * toplev.c (init_local_tick): Move eager initialization of random_seed
14943         to get_random_seed.  Adjust comment.
14944         (init_random_seed): Inline to get_random_seed, delete.
14945         (get_random_seed): Initialize random_seed lazily.
14946         (set_random_seed): Do not return previous value.
14947         (print_switch_value): Do not call get_random_seed.
14949 2017-09-21  Evgeny Kudryashov  <kudryashov@ispras.ru>
14951         * cgraph.c (delete_function_version): New, broken out from...
14952         (cgraph_node::delete_function_version): ...here.  Rename to
14953         cgraph_node::delete_function_version_by_decl.  Update all uses.
14954         (cgraph_node::remove): Call delete_function_version.
14956 2017-09-21  Jakub Jelinek  <jakub@redhat.com>
14958         PR sanitizer/81715
14959         * tree-inline.c (expand_call_inline): Emit clobber stmts for
14960         VAR_DECLs to which addressable non-volatile parameters are mapped
14961         and for id->retvar after the return value assignment.  Clear
14962         id->retval and id->retbnd after inlining.
14964 2017-09-21  Richard Biener  <rguenther@suse.de>
14966         PR tree-optimization/82276
14967         PR tree-optimization/82244
14968         * tree-vrp.c (build_assert_expr_for): Set
14969         SSA_NAME_OCCURS_IN_ABNORMAL_PHI if the variable we assert on
14970         has it set.
14971         (remove_range_assertions): Revert earlier change.
14973 2017-09-21  Wilco Dijkstra  <wdijkstr@arm.com>
14975         PR target/71951
14976         * config/aarch64/aarch64.h (LIBGCC2_UNWIND_ATTRIBUTE): Define.
14978 2017-09-21  Richard Biener  <rguenther@suse.de>
14980         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
14981         Restore valid IL after code generation errors.
14982         * graphite.c (graphite_transform_loops): Diagnose code
14983         generation issues as MSG_MISSED_OPTIMIZATION and continue
14984         with processing SCOPs.
14986 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14987             Alan Hayward  <alan.hayward@arm.com>
14988             David Sherwood  <david.sherwood@arm.com>
14990         * calls.c (compute_argument_addresses): Use simplify_gen_binary
14991         rather than choosing between plus_constant and gen_rtx_<CODE>.
14992         * expr.c (emit_push_insn): Likewise.
14993         (expand_expr_real_2): Likewise.
14995 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14996             Alan Hayward  <alan.hayward@arm.com>
14997             David Sherwood  <david.sherwood@arm.com>
14999         * loop-unroll.c (split_iv): Call copy_rtx on the step.
15001 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
15002             Alan Hayward  <alan.hayward@arm.com>
15003             David Sherwood  <david.sherwood@arm.com>
15005         * tree.c (find_atomic_core_type): Check tree_fits_uhwi_p before
15006         calling tree_to_uhwi.
15008 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
15009             Alan Hayward  <alan.hayward@arm.com>
15010             David Sherwood  <david.sherwood@arm.com>
15012         * tree-ssa-ccp.c (get_value_for_expr): Use a positive test for
15013         INTEGER_CST rather than a negative test for ADDR_EXPR.
15015 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
15016             Alan Hayward  <alan.hayward@arm.com>
15017             David Sherwood  <david.sherwood@arm.com>
15019         * tree-vrp.c (extract_range_from_binary_expr_1): Check
15020         int_cst_rangeN before calling value_range_constant_singleton (&vrN).
15022 2017-09-21  Richard Biener  <rguenther@suse.de>
15024         PR tree-optimization/71351
15025         * graphite-isl-ast-to-gimple.c (translate_isl_ast_to_gimple::
15026         graphite_create_new_loop_guard): Remove, fold remaining parts
15027         into caller ...
15028         (translate_isl_ast_node_for): ... here and simplify.
15030 2017-09-21  Jakub Jelinek  <jakub@redhat.com>
15032         PR target/82260
15033         * config/i386/i386.md (*movqi_internal): Replace (=q,q) alternative
15034         with (=Q,Q), (=R,R) and (=r,r) alternatives, only enable the
15035         latter two for 64-bit, renumber alternatives, for -Os imov (=q,n)
15036         alternative always use QI mode, for -Os imov (=R,R) alternative
15037         always use SI mode, for imov (=Q,Q) or (=r,r) alternatives
15038         ignore -Os.
15040 2017-09-20  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
15041             Jeff Law  <law@redhat.com>
15043         * config/s390/s390.c (MIN_UNROLL_PROBES): Define.
15044         (allocate_stack_space): New function, partially extracted from
15045         s390_emit_prologue.
15046         (s390_emit_prologue): Track offset to most recent stack probe.
15047         Code to allocate space moved into allocate_stack_space.
15048         Dump actions when no stack is allocated.
15049         (s390_prologue_plus_offset): New function.
15050         (s390_emit_stack_probe): Likewise.
15052 2017-09-20  Alexandre Oliva  <aoliva@redhat.com>
15054         * common.opt (Wa, Wl, Wp, g, gz=): Add
15055         RejectNegative.
15056         (gno-column-info): Remove.
15057         (gcolumn-info): Drop RejectNegative.
15058         (gno-): New prefix.
15059         (gno-record-gcc-switches): Remove.
15060         (grecord-gcc-switches): Drop RejectNegative.
15061         (gno-split-dwarf): Remove.
15062         (gsplit-dwarf): Drop RejectNegative.
15063         (gno-strict-dwarf): Remove.
15064         (gstrict-dwarf): Drop RejectNegative.
15065         * config/darwin.opt (gfull, gused): Add RejectNegative.
15066         * dwarf2out.c (gen_producer_string): Drop
15067         gno-record-gcc-switches handler.
15068         * optc-gen.awk: Add g to prefixes with negative forms.
15069         * opts-common.c (remapping_prefix_p): New.
15070         (find_opt): Check it.
15071         (generate_canonical_option): Test g prefix.
15072         (option_map): Add -gno- mapping.
15073         (add_misspelling_candidates): Check remapping_prefix_p.
15075 2017-09-20  Jeff Law  <law@redhat.com>
15077         * config/powerpcspe/powerpcspe.c (rs6000_expand_prologue): Fix
15078         thinko in stack clash protection support.
15080         * explow.c (compute_stack_clash_protection_loop_data): Use
15081         CONST_INT_P instead of explicit test.  Verify object is a
15082         CONST_INT_P before looking at INTVAL.
15083         (anti_adjust_stack_and_probe_stack_clash): Use CONST_INT_P
15084         instead of explicit test.
15086 2017-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
15088         PR target/77687
15089         * config/rs6000/rs6000.md (stack_restore_tie): Store to a scratch
15090         address instead of to r1 and r11.
15092 2017-09-20  Sebastian Peryt  <sebastian.peryt@intel.com>
15094         * config.gcc: Support "knm".
15095         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "knm".
15096         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15097         PROCESSOR_KNM.
15098         * config/i386/i386.c (m_KNM): Define.
15099         (processor_target_table): Add "knm".
15100         (PTA_KNM): Define.
15101         (ix86_option_override_internal): Add "knm".
15102         (ix86_issue_rate): Add PROCESSOR_KNM.
15103         (ix86_adjust_cost): Ditto.
15104         (ia32_multipass_dfa_lookahead): Ditto.
15105         (get_builtin_code_for_version): Handle PROCESSOR_KNM.
15106         (fold_builtin_cpu): Add M_INTEL_KNM.
15107         * config/i386/i386.h (processor_costs): Define TARGET_KNM.
15108         (processor_type): Add PROCESSOR_KNM.
15109         * config/i386/x86-tune.def: Add m_KNM.
15110         * doc/invoke.texi: Add knm as x86 -march=/-mtune= CPU type.
15112 2017-09-20  Richard Biener  <rguenther@suse.de>
15114         PR tree-optimization/80213
15115         * graphite-scop-detection.c (trivially_empty_bb_p): Labels
15116         are allowed in empty BBs as well.
15117         (canonicalize_loop_closed_ssa): Also look for other complex
15118         edges.
15119         (scop_detection::get_sese): Include the loop-closed PHI block
15120         in loop SESEs.
15121         (scop_detection::merge_sese): Remove code adding extra blocks.
15122         (scop_detection::region_has_one_loop): Adjust for get_sese changes.
15123         (build_scops): Assert the final returned scop is invalid.
15125 2017-09-20  Richard Biener  <rguenther@suse.de>
15127         PR tree-optimization/82264
15128         * tree-ssa-sccvn.c (vn_phi_eq): Use safe_dyn_cast to check
15129         for GIMPLE_CONDs.
15130         (vn_phi_lookup): Likewise.
15131         (vn_phi_insert): Likewise.
15133 2017-09-20  Jakub Jelinek  <jakub@redhat.com>
15135         * dwarf2out.c (tree_add_const_value_attribute): For INTEGER_CST
15136         that fits into uhwi or shwi, add DW_AT_const_value regardless
15137         of early_dwarf without going through RTL, using add_AT_unsigned
15138         or add_AT_int.
15140         * dwarf2out.c (DEBUG_LTO_DWO_INFO_SECTION): Reorder defines.
15141         (DEBUG_LTO_ABBREV_SECTION): Likewise.
15142         (DEBUG_LTO_MACINFO_SECTION): Likewise.
15143         (DEBUG_MACRO_SECTION): Likewise.
15144         (DEBUG_LTO_MACRO_SECTION): Likewise.
15145         (DEBUG_STR_DWO_SECTION): Likewise.
15146         (DEBUG_LTO_STR_DWO_SECTION): Likewise.
15147         (DEBUG_LTO_LINE_SECTION): Drop .dwo suffix from the name.
15148         (DEBUG_LTO_DWO_LINE_SECTION): Define.
15149         (DEBUG_LTO_LINE_STR_SECTION): Define.
15150         (init_sections_and_labels): Initialize debug_line_str_section
15151         variable.  Initialize debug_loc_section for -gdwarf-5 to
15152         DEBUG_LOCLISTS_SECTION.  Formatting fixes.
15154 2017-09-20  Richard Biener  <rguenther@suse.de>
15156         * graphite-sese-to-poly.c (extract_affine): Properly handle
15157         POINTER_PLUS_EXPR, BIT_NOT_EXPR and conversion to signed.
15159 2017-09-20  Richard Biener  <rguenther@suse.de>
15161         PR tree-optimization/81373
15162         * graphite-scop-detection.c (build_cross_bb_scalars_def):
15163         Force SESE live-out defs to be handled even if they are
15164         scev_analyzable_p.
15166 2017-09-19  Jeff Law  <law@redhat.com>
15168         * combine-stack-adj.c (combine_stack_adjustments_for_block): Do
15169         nothing for stack adjustments with REG_STACK_CHECK.
15170         * sched-deps.c (parse_add_or_inc): Reject insns with
15171         REG_STACK_CHECK from dependency breaking.
15172         * config/i386/i386.c (pro_epilogue_adjust_stack): Return insn.
15173         (ix86_adjust_satck_and_probe_stack_clash): Add REG_STACK_NOTEs.
15174         * reg-notes.def (STACK_CHECK): New note.
15176         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): New.
15177         (ix86_expand_prologue): Dump stack clash info as needed.
15178         Call ix86_adjust_stack_and_probe_stack_clash as needed.
15180         * function.c (dump_stack_clash_frame_info): New function.
15181         * function.h (dump_stack_clash_frame_info): Prototype.
15182         (enum stack_clash_probes): New enum.
15184         * config/alpha/alpha.c (alpha_expand_prologue): Also check
15185         flag_stack_clash_protection.
15186         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Likewise.
15187         (arm_expand_prologue, thumb1_expand_prologue): Likewise.
15188         (arm_frame_pointer_required): Likewise.
15189         * config/ia64/ia64.c (ia64_compute_frame_size): Likewise.
15190         (ia64_expand_prologue): Likewise.
15191         * config/mips/mips.c (mips_expand_prologue): Likewise.
15192         * config/powerpcspe/powerpcspe.c (rs6000_expand_prologue): Likewise.
15193         * config/sparc/sparc.c (sparc_expand_prologue): Likewise.
15194         (sparc_flat_expand_prologue): Likewise.
15195         * config/spu/spu.c (spu_expand_prologue): Likewise.
15197         * explow.c: Include "params.h".
15198         (anti_adjust_stack_and_probe_stack_clash): New function.
15199         (get_stack_check_protect): Likewise.
15200         (compute_stack_clash_protection_loop_data): Likewise.
15201         (emit_stack_clash_protection_loop_start): Likewise.
15202         (emit_stack_clash_protection_loop_end): Likewise.
15203         (allocate_dynamic_stack_space): Use get_stack_check_protect.
15204         Use anti_adjust_stack_and_probe_stack_clash.
15205         * explow.h (compute_stack_clash_protection_loop_data): Prototype.
15206         (emit_stack_clash_protection_loop_start): Likewise.
15207         (emit_stack_clash_protection_loop_end): Likewise.
15208         * rtl.h (get_stack_check_protect): Prototype.
15209         * target.def (stack_clash_protection_final_dynamic_probe): New hook.
15210         * targhooks.c (default_stack_clash_protection_final_dynamic_probe): New.
15211         * targhooks.h (default_stack_clash_protection_final_dynamic_probe):
15212         Prototype.
15213         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE):
15214         Add @hook.
15215         * doc/tm.texi: Rebuilt.
15216         * config/aarch64/aarch64.c (aarch64_expand_prologue): Use
15217         get_stack_check_protect.
15218         * config/alpha/alpha.c (alpha_expand_prologue): Likewise.
15219         * config/arm/arm.c (arm_expand_prologue): Likewise.
15220         (arm_frame_pointer_required): Likewise.
15221         * config/i386/i386.c (ix86_expand_prologue): Likewise.
15222         * config/ia64/ia64.c (ia64_expand_prologue): Likewise.
15223         * config/mips/mips.c (mips_expand_prologue): Likewise.
15224         * config/powerpcspe/powerpcspe.c (rs6000_emit_prologue): Likewise.
15225         * config/rs6000/rs6000.c (rs6000_emit_prologue): Likewise.
15226         * config/sparc/sparc.c (sparc_expand_prologue): Likewise.
15227         (sparc_flat_expand_prologue): Likewise.
15229         * common.opt (-fstack-clash-protection): New option.
15230         * flag-types.h (enum stack_check_type): Note difference between
15231         -fstack-check= and -fstack-clash-protection.
15232         * params.def (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE): New PARAM.
15233         (PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Likewise.
15234         * toplev.c (process_options): Issue warnings/errors for cases
15235         not handled with -fstack-clash-protection.
15236         * doc/invoke.texi (-fstack-clash-protection): Document new option.
15237         (-fstack-check): Note additional problem with -fstack-check=generic.
15238         Note that -fstack-check is primarily for Ada and refer users
15239         to -fstack-clash-protection for stack-clash-protection.
15240         Document new params for stack clash protection.
15242 2017-09-19  Uros Bizjak  <ubizjak@gmail.com>
15244         * config/i386/i386.c (ix86_split_long_move): Do not handle
15245         address used for LEA in a special way.
15247 2017-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
15249         * simplify-rtx.c (simplify_binary_operation_1): Fix typo in comment.
15251 2017-09-19  Martin Sebor  <msebor@redhat.com>
15253         PR c/81854
15254         * cgraphunit.c (handle_alias_pairs): Reject aliases between functions
15255         of incompatible types.
15257 2017-09-19  Will Schmidt  <will_schmidt@vnet.ibm.com>
15259         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
15260         for early folding of vector loads (ALTIVEC_BUILTIN_LVX_*).
15261         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
15262         Remove obsoleted code for handling ALTIVEC_BUILTIN_VEC_LD.
15264 2017-09-19  Richard Biener  <rguenther@suse.de>
15266         PR tree-optimization/82244
15267         * tree-vrp.c (remove_range_assertions): Do not propagate
15268         a constant to abnormals but replace the assert with a copy.
15270 2017-09-19  Alexander Monakov  <amonakov@ispras.ru>
15272         PR rtl-optimization/57878
15273         PR rtl-optimization/68988
15274         * lra-assigns.c (reload_pseudo_compare_func): Remove fragmentation
15275         avoidance test involving non_reload_pseudos.  Move frequency test
15276         below the general fragmentation avoidance test.
15278 2017-09-19  Richard Biener  <rguenther@suse.de>
15280         PR tree-optimization/69728
15281         * graphite-sese-to-poly.c (schedule_error): New global.
15282         (add_loop_schedule): Handle empty domain by failing the
15283         schedule.
15284         (build_original_schedule): Handle schedule_error.
15286 2017-09-19  Richard Biener  <rguenther@suse.de>
15288         * graphite-scop-detection.c (scop_detection::can_represent_loop):
15289         Do not iterate to sibling loops but only to siblings of inner
15290         loops.
15292 2017-09-18  Andreas Schwab  <schwab@linux-m68k.org>
15294         PR target/81613
15295         * config/m68k/m68k.md (moveq feeding equality comparison): Check
15296         that the registers are different.
15298 2017-09-18  Uros Bizjak  <ubizjak@gmail.com>
15300         * config/i386/i386.c (fold_builtin_cpu): Add M_AMDFAM17H
15301         to processor_model and "amdfam17h" to arch_names_table.
15302         * doc/extend.texi (__builtin_cpu_is): Document amdfam17h CPU name.
15304 2017-09-18  Jakub Jelinek  <jakub@redhat.com>
15306         PR c/82234
15307         * doc/extend.texi: Add @findex entry for __builtin_shuffle.
15309 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15310             Alan Hayward  <alan.hayward@arm.com>
15311             David Sherwood  <david.sherwood@arm.com>
15313         * tree-vectorizer.h (vect_slp_analyze_operations): Replace parameters
15314         with a vec_info *.
15315         * tree-vect-loop.c (vect_analyze_loop_operations): Update call
15316         accordingly.
15317         * tree-vect-slp.c (vect_slp_analyze_node_operations): Add vec_info *
15318         parameter.  Set SLP_TREE_NUMBER_OF_VEC_STMTS here rather than in
15319         vect_schedule_slp_instance.
15320         (vect_slp_analyze_operations): Replace parameters with a vec_info *.
15321         Update call to vect_slp_analyze_node_operations.  Simplify return
15322         value.
15323         (vect_slp_analyze_bb_1): Update call accordingly.
15324         (vect_schedule_slp_instance): Remove vectorization_factor parameter.
15325         Don't calculate SLP_TREE_NUMBER_OF_VEC_STMTS here.
15326         (vect_schedule_slp): Update call accordingly.
15328 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15329             Alan Hayward  <alan.hayward@arm.com>
15330             David Sherwood  <david.sherwood@arm.com>
15332         * tree-vect-data-refs.c (vect_get_smallest_scalar_type): Cope
15333         with types that aren't in fact scalar.
15335 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15337         * tree-vect-slp.c (vect_record_max_nunits): New function,
15338         split out from...
15339         (vect_build_slp_tree_1): ...here.
15340         (vect_build_slp_tree_2): Call it for phis too.
15342 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15344         * tree-vect-stmts.c (vectorizable_mask_load_store): Pass mask_vectype
15345         to vect_get_vec_def_for_operand when getting the mask operand.
15347 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15348             Alan Hayward  <alan.hayward@arm.com>
15349             David Sherwood  <david.sherwood@arm.com>
15351         * tree-vect-loop.c (vectorizable_live_operation): Fix type of
15352         bitstart.
15354 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15355             Alan Hayward  <alan.hayward@arm.com>
15356             David Sherwood  <david.sherwood@arm.com>
15358         * tree-vect-loop.c (vectorizable_live_operation): Fix element size
15359         calculation for vector booleans.
15361 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15362             Alan Hayward  <alan.hayward@arm.com>
15363             David Sherwood  <david.sherwood@arm.com>
15365         * tree-vect-stmts.c (can_vectorize_live_stmts): New function,
15366         split out from...
15367         (vect_transform_stmt): ...here.
15368         (vect_analyze_stmt): Use it instead of calling
15369         vectorizable_live_operation directly.
15371 2017-09-18  Cesar Philippidis  <cesar@codesourcery.com>
15373         * omp-offload.c (oacc_xform_loop): Enable SIMD vectorization on
15374         non-SIMT targets in acc vector loops.
15376 2017-09-18  Claudiu Zissulescu  <claziss@synopsys.com>
15378         * configure.ac: Add arc and check if assembler supports gdwarf2.
15379         * configure: Regenerate.
15381 2017-09-18  Richard Biener  <rguenther@suse.de>
15383         PR tree-optimization/82220
15384         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Exclude
15385         epilogue niters from the min_profitable_iters compute.
15387 2017-09-18  Jakub Jelinek  <jakub@redhat.com>
15389         PR target/82145
15390         * config/i386/i386.c (ix86_init_large_pic_reg): Revert 2017-09-01
15391         changes.  Turn CODE_LABEL into NOTE_INSN_DELETED_LABEL immediately.
15392         (ix86_init_pic_reg): Revert 2017-09-01 changes.
15394 2017-09-18  Eric Botcazou  <ebotcazou@adacore.com>
15396         PR target/81361
15397         * dwarf2cfi.c (add_cfis_to_fde): Do not generate DW_CFA_set_loc after
15398         switching to a new text section.
15400 2017-09-18  Richard Biener  <rguenther@suse.de>
15402         * graphite-scop-detection.c (scop_detection::stmt_has_simple_data_ref):
15403         Simplify.
15404         (build_alias_set): Reject aliases with no access function.
15406 2017-09-18  Richard Biener  <rguenther@suse.de>
15408         PR tree-optimization/79622
15409         * graphite-scop-detection.c (build_cross_bb_scalars_def): Properly
15410         handle PHIs.
15411         (build_cross_bb_scalars_use): Likewise.
15413 2017-09-18  Pierre-Marie de Rodat  <derodat@adacore.com>
15415         * cgraph.h (cgraph_thunk_info): Fix a typo in a comment.
15417 2017-09-18  Alan Modra  <amodra@gmail.com>
15419         PR target/81996
15420         * gcc/config/rs6000/rs6000.c (rs6000_return_addr): Use
15421         stack_pointer_rtx for count 0.  Update comments.  Break up
15422         large rtl expression.
15424 2017-09-17  Daniel Santos  <daniel.santos@pobox.com>
15426         * config/i386/i386.c (xlogue_layout::STUB_NAME_MAX_LEN):
15427         Increase to 20 bytes.
15428         (xlogue_layout::s_stub_names): Add an additional size-2 diminsion.
15429         (xlogue_layout::get_stub_name): Modify to select the appropairate sse
15430         or avx version of the stub.
15432 2017-09-17  H.J. Lu  <hongjiu.lu@intel.com>
15434         PR target/82166
15435         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Properly
15436         compute the minimum stack alignment.  Also update preferred stack
15437         boundary for leaf functions.
15439 2017-09-16  Richard Sandiford  <richard.sandiford@linaro.org>
15441         PR tree-optimization/82228
15442         * tree-vect-loop.c (vectorizable_live_operation): Move initialization
15443         of ncopies.
15445 2017-09-16  Chung-Ju Wu  <jasonwucj@gmail.com>
15447         * common/config/nds32/nds32-common.c
15448         (nds32_option_optimization_table): Refine formatting.
15449         (nds32_option_optimization_table): Use -fsched-pressure and
15450         -fomit-frame-pointer for specific optimization level.
15452 2017-09-16  Chung-Ju Wu  <jasonwucj@gmail.com>
15454         * config/nds32/nds32.c: Refine formatting and comments.
15455         * config/nds32/nds32.h: Likewise.
15456         * config/nds32/nds32.md: Likewise.
15457         * config/nds32/nds32-cost.c: Likewise.
15458         * config/nds32/nds32-isr.c: Likewise.
15459         * config/nds32/nds32-md-auxiliary.c: Likewise.
15460         * config/nds32/nds32-multiple.md: Likewise.
15461         * config/nds32/nds32-predicates.c: Likewise.
15463 2017-09-15  Andrew Sutton  <andrew.n.sutton@gmail.com>
15464             Jakub Jelinek  <jakub@redhat.com>
15466         Add support for -std=c++2a.
15467         * doc/cpp.texi (__cplusplus): Document value for -std=c++2a
15468         or -std=gnu+2a.
15469         * doc/invoke.texi: Document -std=c++2a and -std=gnu++2a.
15471 2017-09-15  Steve Ellcey  <sellcey@cavium.com>
15473         PR target/82066
15474         * doc/extend.texi (Common Function Attributes): Add 
15475         references to ARM, AArch64, and S/390 specific attributes.
15476         (Function Specific Option Pragmas): Add AArch64 and S/390
15477         to list of back ends that support the target pragma.
15479 2017-09-15  Nathan Sidwell  <nathan@acm.org>
15481         * doc/standards.texi: Fix C++17 description.  Update URLs for
15482         C++11 & 14.
15484 2017-09-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
15486         * common.opt (Wcast-align=strict): New warning option.
15487         * doc/invoke.texi: Document -Wcast-align=strict. 
15489 2017-09-15  Pierre-Marie de Rodat  <derodat@adacore.com>
15491         * cgraph.h (cgraph_thunk_info): Add comments.
15492         * cgraph.c (cgraph_node::create_thunk): Adjust comment, make
15493         assert for VIRTUAL_* arguments stricter.
15495 2017-09-15  Jackson Woodruff  <jackson.woodruff@arm.com>
15497         PR tree-optimization/71026
15498         * match.pd: Move RDIV patterns from fold-const.c
15499         * fold-const.c (distribute_real_division): Removed.
15500         (fold_binary_loc): Remove calls to distribute_real_divison.
15502 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15504         * doc/invoke.texi: Document -std=c++17 and -std=gnu++17 and document
15505         c++1z and gnu++1z as deprecated.  Change other references to
15506         -std=c++1z to -std=c++17 and -std=gnu++1z to -std=gnu++17.
15507         Change -Wc++1z-compat to -Wc++17-compat.
15508         * doc/cpp.texi: Document -std=c++17 defines __cplusplus 201703L.
15509         * dwarf2out.c (highest_c_language): Handle C++17.
15510         (gen_compile_unit_die): Likewise.
15512 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15514         PR rtl-optimization/82192
15515         * combine.c (make_extraction): Don't look through non-paradoxical
15516         SUBREGs or TRUNCATE if pos + len is or might be bigger than
15517         inner's mode.
15519 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15520             Alan Hayward  <alan.hayward@arm.com>
15521             David Sherwood  <david.sherwood@arm.com>
15523         * target.def (function_arg_offset): New hook.
15524         * targhooks.h (default_function_arg_offset): Declare.
15525         * targhooks.c (default_function_arg_offset): New function.
15526         * function.c (locate_and_pad_parm): Use
15527         targetm.calls.function_arg_offset instead of FUNCTION_ARG_OFFSET.
15528         * doc/tm.texi.in (FUNCTION_ARG_OFFSET): Replace with...
15529         (TARGET_FUNCTION_ARG_OFFSET): ...this.
15530         * doc/tm.texi: Regenerate.
15531         * config/spu/spu.h (FUNCTION_ARG_OFFSET): Delete.
15532         * config/spu/spu.c (spu_function_arg_offset): New function.
15533         (TARGET_FUNCTION_ARG_OFFSET): Redefine.
15534         * system.h (FUNCTION_ARG_OFFSET): Poison.
15536 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15537             Alan Hayard  <alan.hayward@arm.com>
15538             David Sherwood  <david.sherwood@arm.com>
15540         * target.def (truly_noop_truncation): New hook.
15541         (mode_rep_extended): Refer to TARGET_TRULY_NOOP_TRUNCATION rather
15542         than TRULY_NOOP_TRUNCATION.
15543         * hooks.h (hook_bool_uint_uint_true): Declare.
15544         * hooks.c (hook_bool_uint_uint_true): New function.
15545         * doc/tm.texi.in (TRULY_NOOP_TRUNCATION): Replace with...
15546         (TARGET_TRULY_NOOP_TRUNCATION): ...this.
15547         * doc/tm.texi: Regenerate.
15548         * combine.c (make_extraction): Refer to TARGET_TRULY_NOOP_TRUNCATION
15549         rather than TRULY_NOOP_TRUNCATION in comments.
15550         (simplify_comparison): Likewise.
15551         (record_truncated_value): Likewise.
15552         * expmed.c (extract_bit_field_1): Likewise.
15553         (extract_split_bit_field): Likewise.
15554         * convert.c (convert_to_integer_1): Use targetm.truly_noop_truncation
15555         instead of TRULY_NOOP_TRUNCATION.
15556         * function.c (assign_parm_setup_block): Likewise.
15557         * machmode.h (TRULY_NOOP_TRUNCATION_MODES_P): Likewise.
15558         * rtlhooks.c: Include target.h.
15559         * config/aarch64/aarch64.h (TRULY_NOOP_TRUNCATION): Delete.
15560         * config/alpha/alpha.h (TRULY_NOOP_TRUNCATION): Delete.
15561         * config/arc/arc.h (TRULY_NOOP_TRUNCATION): Delete.
15562         * config/arm/arm.h (TRULY_NOOP_TRUNCATION): Delete.
15563         * config/avr/avr.h (TRULY_NOOP_TRUNCATION): Delete.
15564         * config/bfin/bfin.h (TRULY_NOOP_TRUNCATION): Delete.
15565         * config/c6x/c6x.h (TRULY_NOOP_TRUNCATION): Delete.
15566         * config/cr16/cr16.h (TRULY_NOOP_TRUNCATION): Delete.
15567         * config/cris/cris.h (TRULY_NOOP_TRUNCATION): Delete.
15568         * config/epiphany/epiphany.h (TRULY_NOOP_TRUNCATION): Delete.
15569         * config/fr30/fr30.h (TRULY_NOOP_TRUNCATION): Delete.
15570         * config/frv/frv.h (TRULY_NOOP_TRUNCATION): Delete.
15571         * config/ft32/ft32.h (TRULY_NOOP_TRUNCATION): Delete.
15572         * config/h8300/h8300.h (TRULY_NOOP_TRUNCATION): Delete.
15573         * config/i386/i386.h (TRULY_NOOP_TRUNCATION): Delete.
15574         * config/ia64/ia64.h (TRULY_NOOP_TRUNCATION): Delete.
15575         * config/iq2000/iq2000.h (TRULY_NOOP_TRUNCATION): Delete.
15576         * config/lm32/lm32.h (TRULY_NOOP_TRUNCATION): Delete.
15577         * config/m32c/m32c.h (TRULY_NOOP_TRUNCATION): Delete.
15578         * config/m32r/m32r.h (TRULY_NOOP_TRUNCATION): Delete.
15579         * config/m68k/m68k.h (TRULY_NOOP_TRUNCATION): Delete.
15580         * config/mcore/mcore.h (TRULY_NOOP_TRUNCATION): Delete.
15581         * config/microblaze/microblaze.h (TRULY_NOOP_TRUNCATION): Delete.
15582         * config/mips/mips.h (TRULY_NOOP_TRUNCATION): Delete.
15583         * config/mips/mips.c (mips_truly_noop_truncation): New function.
15584         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15585         * config/mips/mips.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15586         rather than TRULY_NOOP_TRUNCATION in comments.
15587         * config/mmix/mmix.h (TRULY_NOOP_TRUNCATION): Delete.
15588         * config/mn10300/mn10300.h (TRULY_NOOP_TRUNCATION): Delete.
15589         * config/moxie/moxie.h (TRULY_NOOP_TRUNCATION): Delete.
15590         * config/msp430/msp430.h (TRULY_NOOP_TRUNCATION): Delete.
15591         * config/nds32/nds32.h (TRULY_NOOP_TRUNCATION): Delete.
15592         * config/nios2/nios2.h (TRULY_NOOP_TRUNCATION): Delete.
15593         * config/nvptx/nvptx.h (TRULY_NOOP_TRUNCATION): Delete.
15594         * config/pa/pa.h (TRULY_NOOP_TRUNCATION): Delete.
15595         * config/pdp11/pdp11.h (TRULY_NOOP_TRUNCATION): Delete.
15596         * config/powerpcspe/powerpcspe.h (TRULY_NOOP_TRUNCATION): Delete.
15597         * config/riscv/riscv.h (TRULY_NOOP_TRUNCATION): Delete.
15598         * config/riscv/riscv.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15599         rather than TRULY_NOOP_TRUNCATION in comments.
15600         * config/rl78/rl78.h (TRULY_NOOP_TRUNCATION): Delete.
15601         * config/rs6000/rs6000.h (TRULY_NOOP_TRUNCATION): Delete.
15602         * config/rx/rx.h (TRULY_NOOP_TRUNCATION): Delete.
15603         * config/s390/s390.h (TRULY_NOOP_TRUNCATION): Delete.
15604         * config/sh/sh.h (MAYBE_BASE_REGISTER_RTX_P): Remove
15605         TRULY_NOOP_TRUNCATION condition.
15606         (MAYBE_INDEX_REGISTER_RTX_P): Likewise.
15607         (TRULY_NOOP_TRUNCATION): Delete.
15608         * config/sparc/sparc.h (TRULY_NOOP_TRUNCATION): Delete.
15609         * config/spu/spu.h (TRULY_NOOP_TRUNCATION): Delete.
15610         * config/spu/spu.c (spu_truly_noop_truncation): New function.
15611         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15612         * config/stormy16/stormy16.h (TRULY_NOOP_TRUNCATION): Delete.
15613         * config/tilegx/tilegx.h (TRULY_NOOP_TRUNCATION): Delete.
15614         * config/tilegx/tilegx.c (tilegx_truly_noop_truncation): New fuction.
15615         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15616         * config/tilegx/tilegx.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15617         rather than TRULY_NOOP_TRUNCATION in comments.
15618         * config/tilepro/tilepro.h (TRULY_NOOP_TRUNCATION): Delete.
15619         * config/v850/v850.h (TRULY_NOOP_TRUNCATION): Delete.
15620         * config/vax/vax.h (TRULY_NOOP_TRUNCATION): Delete.
15621         * config/visium/visium.h (TRULY_NOOP_TRUNCATION): Delete.
15622         * config/xtensa/xtensa.h (TRULY_NOOP_TRUNCATION): Delete.
15623         * system.h (TRULY_NOOP_TRUNCATION): Poison.
15625 2017-09-15  Christophe Lyon  <christophe.lyon@linaro.org>
15627         PR target/67591
15628         * config/arm/arm.md (*cmp_and): Add enabled_for_depr_it attribute.
15629         (*cmp_ior): Likewise.
15630         (*ior_scc_scc): Add alternative for enabled_for_depr_it attribute.
15631         (*ior_scc_scc_cmp): Likewise.
15632         (*and_scc_scc): Likewise.
15633         (*and_scc_scc_cmp): Likewise.
15635 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15636             Alan Hayard  <alan.hayward@arm.com>
15637             David Sherwood  <david.sherwood@arm.com>
15639         * target.def (can_change_mode_class): New hook.
15640         (mode_rep_extended): Refer to it instead of CANNOT_CHANGE_MODE_CLASS.
15641         (hard_regno_nregs): Likewise.
15642         * hooks.h (hook_bool_mode_mode_reg_class_t_true): Declare.
15643         * hooks.c (hook_bool_mode_mode_reg_class_t_true): New function.
15644         * doc/tm.texi.in (CANNOT_CHANGE_MODE_CLASS): Replace with...
15645         (TARGET_CAN_CHANGE_MODE_CLASS): ...this.
15646         (LOAD_EXTEND_OP): Update accordingly.
15647         * doc/tm.texi: Regenerate.
15648         * doc/rtl.texi: Refer to TARGET_CAN_CHANGE_MODE_CLASS instead of
15649         CANNOT_CHANGE_MODE_CLASS.
15650         * hard-reg-set.h (REG_CANNOT_CHANGE_MODE_P): Replace with...
15651         (REG_CAN_CHANGE_MODE_P): ...this new macro.
15652         * combine.c (simplify_set): Update accordingly.
15653         * emit-rtl.c (validate_subreg): Likewise.
15654         * recog.c (general_operand): Likewise.
15655         * regcprop.c (mode_change_ok): Likewise.
15656         * reload1.c (choose_reload_regs): Likewise.
15657         (inherit_piecemeal_p): Likewise.
15658         * rtlanal.c (simplify_subreg_regno): Likewise.
15659         * postreload.c (reload_cse_simplify_set): Use REG_CAN_CHANGE_MODE_P
15660         instead of CANNOT_CHANGE_MODE_CLASS.
15661         (reload_cse_simplify_operands): Likewise.
15662         * reload.c (push_reload): Use targetm.can_change_mode_class
15663         instead of CANNOT_CHANGE_MODE_CLASS.
15664         (push_reload): Likewise.  Also use REG_CAN_CHANGE_MODE_P instead of
15665         REG_CANNOT_CHANGE_MODE_P.
15666         * config/alpha/alpha.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15667         * config/alpha/alpha.c (alpha_can_change_mode_class): New function.
15668         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15669         * config/arm/arm.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15670         * config/arm/arm.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15671         (arm_can_change_mode_class): New function.
15672         * config/arm/neon.md: Refer to TARGET_CAN_CHANGE_MODE_CLASS rather
15673         than CANNOT_CHANGE_MODE_CLASS in comments.
15674         * config/i386/i386.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15675         * config/i386/i386-protos.h (ix86_cannot_change_mode_class): Delete.
15676         * config/i386/i386.c (ix86_cannot_change_mode_class): Replace with...
15677         (ix86_can_change_mode_class): ...this new function, inverting the
15678         sense of the return value.
15679         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15680         * config/ia64/ia64.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15681         * config/ia64/ia64.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15682         (ia64_can_change_mode_class): New function.
15683         * config/m32c/m32c.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15684         * config/m32c/m32c-protos.h (m32c_cannot_change_mode_class): Delete.
15685         * config/m32c/m32c.c (m32c_cannot_change_mode_class): Replace with...
15686         (m32c_can_change_mode_class): ...this new function, inverting the
15687         sense of the return value.
15688         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15689         * config/mips/mips.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15690         * config/mips/mips-protos.h (mips_cannot_change_mode_class): Delete.
15691         * config/mips/mips.c (mips_cannot_change_mode_class): Replace with...
15692         (mips_can_change_mode_class): ...this new function, inverting the
15693         sense of the return value.
15694         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15695         * config/msp430/msp430.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15696         * config/msp430/msp430.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15697         (msp430_can_change_mode_class): New function.
15698         * config/nvptx/nvptx.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15699         * config/nvptx/nvptx.c (nvptx_can_change_mode_class): New function.
15700         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15701         * config/pa/pa32-regs.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15702         * config/pa/pa64-regs.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15703         * config/pa/pa-protos.h (pa_cannot_change_mode_class): Delete.
15704         * config/pa/pa.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15705         (pa_cannot_change_mode_class): Replace with...
15706         (pa_can_change_mode_class): ...this new function, inverting the
15707         sense of the return value.
15708         (pa_modes_tieable_p): Refer to TARGET_CAN_CHANGE_MODE_CLASS rather
15709         than CANNOT_CHANGE_MODE_CLASS in comments.
15710         * config/pdp11/pdp11.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15711         * config/pdp11/pdp11-protos.h (pdp11_cannot_change_mode_class): Delete.
15712         * config/pdp11/pdp11.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15713         (pdp11_cannot_change_mode_class): Replace with...
15714         (pdp11_can_change_mode_class): ...this new function, inverting the
15715         sense of the return value.
15716         * config/powerpcspe/powerpcspe.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15717         * config/powerpcspe/powerpcspe-protos.h
15718         (rs6000_cannot_change_mode_class_ptr): Delete.
15719         * config/powerpcspe/powerpcspe.c
15720         (rs6000_cannot_change_mode_class_ptr): Delete.
15721         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15722         (rs6000_option_override_internal): Assign to
15723         targetm.can_change_mode_class instead of
15724         rs6000_cannot_change_mode_class_ptr.
15725         (rs6000_cannot_change_mode_class): Replace with...
15726         (rs6000_can_change_mode_class): ...this new function, inverting the
15727         sense of the return value.
15728         (rs6000_debug_cannot_change_mode_class): Replace with...
15729         (rs6000_debug_can_change_mode_class): ...this new function.
15730         * config/riscv/riscv.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15731         * config/riscv/riscv.c (riscv_can_change_mode_class): New function.
15732         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15733         * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15734         * config/rs6000/rs6000-protos.h (rs6000_cannot_change_mode_class_ptr):
15735         Delete.
15736         * config/rs6000/rs6000.c (rs6000_cannot_change_mode_class_ptr): Delete.
15737         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15738         (rs6000_option_override_internal): Assign to
15739         targetm.can_change_mode_class instead of
15740         rs6000_cannot_change_mode_class_ptr.
15741         (rs6000_cannot_change_mode_class): Replace with...
15742         (rs6000_can_change_mode_class): ...this new function, inverting the
15743         sense of the return value.
15744         (rs6000_debug_cannot_change_mode_class): Replace with...
15745         (rs6000_debug_can_change_mode_class): ...this new function.
15746         * config/s390/s390.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15747         * config/s390/s390-protos.h (s390_cannot_change_mode_class): Delete.
15748         * config/s390/s390.c (s390_cannot_change_mode_class): Replace with...
15749         (s390_can_change_mode_class): ...this new function, inverting the
15750         sense of the return value.
15751         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15752         * config/sh/sh.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15753         * config/sh/sh-protos.h (sh_cannot_change_mode_class): Delete.
15754         * config/sh/sh.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15755         (sh_cannot_change_mode_class): Replace with...
15756         (sh_can_change_mode_class): ...this new function, inverting the
15757         sense of the return value.
15758         * config/sparc/sparc.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15759         * config/sparc/sparc.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15760         (sparc_can_change_mode_class): New function.
15761         * config/spu/spu.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15762         * config/spu/spu.c (spu_can_change_mode_class): New function.
15763         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15764         * config/visium/visium.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15765         * config/visium/visium.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15766         (visium_can_change_mode_class): New function.
15767         * system.h (CANNOT_CHANGE_MODE_CLASS): Poison.
15769 2017-09-15  Richard Biener  <rguenther@suse.de>
15771         PR tree-optimization/82217
15772         * tree-ssa-sccvn.c (visit_phi): Properly handle all VN_TOP
15773         but not undefined case.
15775 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15777         PR target/82145
15778         * postreload.c (reload_cse_simplify_operands): Skip
15779         NOTE_INSN_DELETED_LABEL similarly to skipping CODE_LABEL.
15781 2017-09-15  Richard Biener  <rguenther@suse.de>
15783         PR tree-optimization/68823
15784         * graphite-scop-detection.c (build_alias_set): If we have a
15785         possible dependence check whether we can handle them by just
15786         looking at the DRs DR_ACCESS_FNs.
15787         (build_scops): If build_alias_set fails, fail the SCOP.
15789 2017-09-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
15791         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_1_HW): New macros
15792         to support float128 built-in functions that require the ISA 3.0
15793         hardware.
15794         (BU_FLOAT128_3_HW): Likewise.
15795         (SQRTF128): Add support for the IEEE 128-bit square root and fma
15796         built-in functions.
15797         (FMAF128): Likewise.
15798         (FMAQ): Likewise.
15799         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add
15800         support for built-in functions that need the ISA 3.0 IEEE 128-bit
15801         floating point instructions.
15802         (rs6000_invalid_builtin): Likewise.
15803         (rs6000_builtin_mask_names): Likewise.
15804         * config/rs6000/rs6000.h (MASK_FLOAT128_HW): Likewise.
15805         (RS6000_BTM_FLOAT128_HW): Likewise.
15806         (RS6000_BTM_COMMON): Likewise.
15807         * config/rs6000/rs6000.md (fma<mode>4_hw): Add a generator
15808         function.
15809         * doc/extend.texi (RS/6000 built-in functions): Document the
15810         IEEE 128-bit floating point square root and fused multiply-add
15811         built-in functions.
15813 2017-09-14  Pat Haugen  <pthaugen@us.ibm.com>
15815         * config/rs6000/rs6000.c (rs6000_set_up_by_prologue): Make sure the TOC
15816         reg (r2) isn't in the set of registers defined in the prologue.
15818 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15819             Alan Hayward  <alan.hayward@arm.com>
15820             David Sherwood  <david.sherwood@arm.com>
15822         * tree-vectorizer.h (_loop_vec_info): Add max_vectorization_factor.
15823         (LOOP_VINFO_MAX_VECT_FACTOR): New macro.
15824         (LOOP_VINFO_ORIG_VECT_FACTOR): Replace with...
15825         (LOOP_VINFO_ORIG_MAX_VECT_FACTOR): ...this new macro.
15826         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Update
15827         accordingly.
15828         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15829         max_vectorization_factor.
15830         (vect_analyze_loop_2): Set LOOP_VINFO_MAX_VECT_FACTOR.
15832 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15833             Alan Hayward  <alan.hayward@arm.com>
15834             David Sherwood  <david.sherwood@arm.com>
15836         * tree-vectorizer.h (vect_min_worthwhile_factor): Delete.
15837         (vect_worthwhile_without_simd_p): Declare.
15838         * tree-vect-loop.c (vect_worthwhile_without_simd_p): New function.
15839         (vectorizable_reduction): Use it.
15840         * tree-vect-stmts.c (vectorizable_shift): Likewise.
15841         (vectorizable_operation): Likewise.
15843 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15844             Alan Hayward  <alan.hayward@arm.com>
15845             David Sherwood  <david.sherwood@arm.com>
15847         * tree-vectorizer.h (vect_get_num_copies): New function.
15848         * tree-vect-data-refs.c (vect_get_data_access_cost): Use it.
15849         * tree-vect-loop.c (vectorizable_reduction): Likewise.
15850         (vectorizable_induction): Likewise.
15851         (vectorizable_live_operation): Likewise.
15852         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
15853         (vectorizable_bswap): Likewise.
15854         (vectorizable_call): Likewise.
15855         (vectorizable_conversion): Likewise.
15856         (vectorizable_assignment): Likewise.
15857         (vectorizable_shift): Likewise.
15858         (vectorizable_operation): Likewise.
15859         (vectorizable_store): Likewise.
15860         (vectorizable_load): Likewise.
15861         (vectorizable_condition): Likewise.
15862         (vectorizable_comparison): Likewise.
15863         (vect_analyze_stmt): Pass the slp node to vectorizable_live_operation.
15865 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15866             Alan Hayward  <alan.hayward@arm.com>
15867             David Sherwood  <david.sherwood@arm.com>
15869         * tree-vect-loop.c (vectorizable_induction): Use gimple_build instead
15870         of vect_init_vector.
15872 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15873             Alan Hayward  <alan.hayward@arm.com>
15874             David Sherwood  <david.sherwood@arm.com>
15876         * gimple-fold.h (gimple_build_vector_from_val): Declare, and provide
15877         an inline wrapper that provides a location.
15878         (gimple_build_vector): Likewise.
15879         * gimple-fold.c (gimple_build_vector_from_val): New function.
15880         (gimple_build_vector): Likewise.
15881         * tree-vect-loop.c (get_initial_def_for_reduction): Use the new
15882         functions to build the initial value.  Always return a gimple value.
15883         (get_initial_defs_for_reduction): Likewise.  Only compute
15884         neutral_vec once.
15885         (vect_create_epilog_for_reduction): Don't call force_gimple_operand or
15886         vect_init_vector on the results from get_initial_def(s)_for_reduction.
15887         (vectorizable_induction): Use gimple_build_vector rather than
15888         vect_init_vector.
15890 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15891             Alan Hayward  <alan.hayward@arm.com>
15892             David Sherwood  <david.sherwood@arm.com>
15894         * target.h (vec_perm_indices): New typedef.
15895         (auto_vec_perm_indices): Likewise.
15896         * optabs-query.h: Include target.h
15897         (can_vec_perm_p): Take a vec_perm_indices *.
15898         * optabs-query.c (can_vec_perm_p): Likewise.
15899         (can_mult_highpart_p): Update accordingly.  Use auto_vec_perm_indices.
15900         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15901         * tree-vect-generic.c (lower_vec_perm): Likewise.
15902         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15903         (vect_grouped_load_supported): Likewise.
15904         (vect_shift_permute_load_chain): Likewise.
15905         (vect_permute_store_chain): Use auto_vec_perm_indices.
15906         (vect_permute_load_chain): Likewise.
15907         * fold-const.c (fold_vec_perm): Take vec_perm_indices.
15908         (fold_ternary_loc): Update accordingly.  Use auto_vec_perm_indices.
15909         Update uses of can_vec_perm_p.
15910         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Replace the
15911         mode with a number of elements.  Take a vec_perm_indices *.
15912         (vect_create_epilog_for_reduction): Update accordingly.
15913         Use auto_vec_perm_indices.
15914         (have_whole_vector_shift): Likewise.  Update call to can_vec_perm_p.
15915         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15916         (vect_transform_slp_perm_load): Likewise.
15917         (vect_schedule_slp_instance): Use auto_vec_perm_indices.
15918         * tree-vectorizer.h (vect_gen_perm_mask_any): Take a vec_perm_indices.
15919         (vect_gen_perm_mask_checked): Likewise.
15920         * tree-vect-stmts.c (vect_gen_perm_mask_any): Take a vec_perm_indices.
15921         (vect_gen_perm_mask_checked): Likewise.
15922         (vectorizable_mask_load_store): Use auto_vec_perm_indices.
15923         (vectorizable_store): Likewise.
15924         (vectorizable_load): Likewise.
15925         (perm_mask_for_reverse): Likewise.  Update call to can_vec_perm_p.
15926         (vectorizable_bswap): Likewise.
15928 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15929             Alan Hayward  <alan.hayward@arm.com>
15930             David Sherwood  <david.sherwood@arm.com>
15932         * tree.h (build_vector): Take a vec<tree> instead of a tree *.
15933         * tree.c (build_vector): Likewise.
15934         (build_vector_from_ctor): Update accordingly.
15935         (build_vector_from_val): Likewise.
15936         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
15937         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15938         * tree-vect-generic.c (add_rshift): Likewise.
15939         (expand_vector_divmod): Likewise.
15940         (optimize_vector_constructor): Likewise.
15941         * tree-vect-slp.c (vect_get_constant_vectors): Likewise.
15942         (vect_transform_slp_perm_load): Likewise.
15943         (vect_schedule_slp_instance): Likewise.
15944         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
15945         (vectorizable_call): Likewise.
15946         (vect_gen_perm_mask_any): Likewise.  Add elements in order.
15947         * expmed.c (make_tree): Likewise.
15948         * fold-const.c (fold_negate_expr_1): Use auto_vec<tree> when building
15949         a vector passed to build_vector.
15950         (fold_convert_const): Likewise.
15951         (exact_inverse): Likewise.
15952         (fold_ternary_loc): Likewise.
15953         (fold_relational_const): Likewise.
15954         (const_binop): Likewise.  Use VECTOR_CST_ELT directly when operating
15955         on VECTOR_CSTs, rather than going through vec_cst_ctor_to_array.
15956         (const_unop): Likewise.  Store the reduction accumulator in a
15957         variable rather than an array.
15958         (vec_cst_ctor_to_array): Take the number of elements as a parameter.
15959         (fold_vec_perm): Update calls accordingly.  Use auto_vec<tree> for
15960         the new vector, rather than constructing it after the input arrays.
15961         (native_interpret_vector): Use auto_vec<tree> when building
15962         a vector passed to build_vector.  Add elements in order.
15963         * tree-vect-loop.c (get_initial_defs_for_reduction): Use
15964         auto_vec<tree> when building a vector passed to build_vector.
15965         (vect_create_epilog_for_reduction): Likewise.
15966         (vectorizable_induction): Likewise.
15967         (get_initial_def_for_reduction): Likewise.  Fix indentation of
15968         case statements.
15969         * config/sparc/sparc.c (sparc_handle_vis_mul8x16): Change n_elts
15970         to a vec<tree> *.
15971         (sparc_fold_builtin): Use auto_vec<tree> when building a vector
15972         passed to build_vector.
15974 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15975             Alan Hayward  <alan.hayward@arm.com>
15976             David Sherwood  <david.sherwood@arm.com>
15978         * tree-core.h (tree_base::u): Add an "nelts" field.
15979         (tree_vector): Use VECTOR_CST_NELTS as the length.
15980         * tree.c (tree_size): Likewise.
15981         (make_vector): Initialize VECTOR_CST_NELTS.
15982         * tree.h (VECTOR_CST_NELTS): Use the u.nelts field.
15983         * cfgexpand.c (expand_debug_expr): Use VECTOR_CST_NELTS instead of
15984         TYPE_VECTOR_SUBPARTS.
15985         * expr.c (const_vector_mask_from_tree): Consistently use "units"
15986         as the number of units, setting it from VECTOR_CST_NELTS.
15987         (const_vector_from_tree): Likewise.
15988         * fold-const.c (negate_expr_p): Use VECTOR_CST_NELTS instead of
15989         TYPE_VECTOR_SUBPARTS for the number of elements in a VECTOR_CST.
15990         (fold_negate_expr_1): Likewise.
15991         (fold_convert_const): Likewise.
15992         (const_binop): Likewise.  Differentiate the number of output and
15993         input elements.
15994         (const_unop): Likewise.
15995         (fold_ternary_loc): Use VECTOR_CST_NELTS for the number of elements
15996         in a VECTOR_CST, asserting that it is the same as TYPE_VECTOR_SUBPARTS
15997         in cases that did the opposite.
15999 2017-09-14  Richard Biener  <rguenther@suse.de>
16001         * tree-ssa-sccvn.c (visit_phi): Merge undefined values similar
16002         to VN_TOP.
16004 2017-09-14  Eric Botcazou  <ebotcazou@adacore.com>
16006         * dwarf2out.c (dwarf2out_source_line): Remove superfluous test.
16008 2017-09-14  Jakub Jelinek  <jakub@redhat.com>
16010         PR target/81325
16011         * cfgbuild.c (find_bb_boundaries): Ignore debug insns in decisions
16012         if and where to split a bb, except for splitting before debug insn
16013         sequences followed by non-label real insn.  Delete debug insns
16014         in between basic blocks.
16016         * combine.c (make_compound_operation_int): Formatting fixes.
16018         * config/alpha/elf.h (LINK_EH_SPEC): Add -static-pie support.
16019         * config/alpha/linux.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
16020         * config/netbsd.h (LINK_EH_SPEC): Likewise.
16021         * config/sol2.h (LINK_EH_SPEC): Likewise.
16022         * config/arm/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
16023         * config/s390/linux.h (LINK_SPEC): Likewise.
16024         * config/freebsd.h (LINK_EH_SPEC): Likewise.
16025         * config/openbsd.h (LINK_EH_SPEC): Likewise.
16026         * config/lm32/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
16027         * config/aarch64/aarch64-linux.h (LINUX_TARGET_LINK_SPEC): Likewise.
16028         * config/powerpcspe/sysv4.h (LINK_EH_SPEC): Likewise.
16029         * config/bfin/linux.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
16030         * config/i386/gnu-user64.h (GNU_USER_TARGET_LINK_SPEC): Fix a typo.
16031         * config/i386/gnu-user.h (GNU_USER_TARGET_LINK_SPEC): Formatting fix.
16033 2017-09-13  Jakub Jelinek  <jakub@redhat.com>
16035         * config/rs6000/sysv4.h (STARTFILE_LINUX_SPEC): Add -static-pie
16036         support.
16037         (ENDFILE_LINUX_SPEC): Likewise.
16038         (LINK_EH_SPEC): Likewise.
16039         * config/rs6000/linux64.h (LINK_SHLIB_SPEC): Likewise.
16040         (LINK_OS_LINUX_SPEC32): Likewise.
16041         (LINK_OS_LINUX_SPEC64): Likewise.
16042         * config/rs6000/linux.h (LINK_SHLIB_SPEC): Likewise.
16043         (LINK_OS_LINUX_SPEC): Likewise.
16045 2017-09-13  Martin Liska  <mliska@suse.cz>
16047         PR middle-end/82154
16048         * stmt.c (expand_sjlj_dispatch_table): Use CASE_LOW when
16049         CASE_HIGH is NULL_TREE.
16051 2017-09-13  Richard Sandiford  <richard.sandiford@linaro.org>
16052             Alan Hayward  <alan.hayward@arm.com>
16053             David Sherwood  <david.sherwood@arm.com>
16055         * target.def (secondary_memory_needed): New hook.
16056         (secondary_reload): Refer to TARGET_SECONDARY_MEMORY_NEEDED
16057         instead of SECONDARY_MEMORY_NEEDED.
16058         (secondary_memory_needed_mode): Likewise.
16059         * hooks.h (hook_bool_mode_reg_class_t_reg_class_t_false): Declare.
16060         * hooks.c (hook_bool_mode_reg_class_t_reg_class_t_false): New function.
16061         * doc/tm.texi.in (SECONDARY_MEMORY_NEEDED): Replace with...
16062         (TARGET_SECONDARY_MEMORY_NEEDED): ...this.
16063         (SECONDARY_MEMORY_NEEDED_RTX): Update reference accordingly.
16064         * doc/tm.texi: Regenerate.
16065         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED): Delete.
16066         * config/alpha/alpha.c (alpha_secondary_memory_needed): New function.
16067         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16068         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED): Delete.
16069         * config/i386/i386-protos.h (ix86_secondary_memory_needed): Delete.
16070         * config/i386/i386.c (inline_secondary_memory_needed): Put the
16071         mode argument first and change the reg_class arguments to reg_class_t.
16072         (ix86_secondary_memory_needed): Likewise.  Remove the strict parameter.
16073         Make static.  Update the call to inline_secondary_memory_needed.
16074         (ix86_register_move_cost): Update the call to
16075         inline_secondary_memory_needed.
16076         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16077         * config/ia64/ia64.h (SECONDARY_MEMORY_NEEDED): Delete commented-out
16078         definition.
16079         * config/ia64/ia64.c (spill_xfmode_rfmode_operand): Refer to
16080         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
16081         in comment.
16082         * config/mips/mips.h (SECONDARY_MEMORY_NEEDED): Delete.
16083         * config/mips/mips-protos.h (mips_secondary_memory_needed): Delete.
16084         * config/mips/mips.c (mips_secondary_memory_needed): Make static
16085         and match hook interface.  Add comment from mips.h.
16086         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16087         * config/mmix/mmix.md (truncdfsf2): Refer to
16088         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
16089         in comment.
16090         * config/pa/pa-64.h (SECONDARY_MEMORY_NEEDED): Rename to...
16091         (PA_SECONDARY_MEMORY_NEEDED): ...this, and put the mode argument first.
16092         * config/pa/pa.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16093         (pa_secondary_memory_needed): New function.
16094         * config/pdp11/pdp11.h (SECONDARY_MEMORY_NEEDED): Delete.
16095         * config/pdp11/pdp11-protos.h (pdp11_secondary_memory_needed): Delete.
16096         * config/pdp11/pdp11.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16097         (pdp11_secondary_memory_needed): Make static and match hook interface.
16098         * config/powerpcspe/powerpcspe.h (SECONDARY_MEMORY_NEEDED): Delete.
16099         * config/powerpcspe/powerpcspe-protos.h
16100         (rs6000_secondary_memory_needed_ptr): Delete.
16101         * config/powerpcspe/powerpcspe.c (rs6000_secondary_memory_needed_ptr):
16102         Delete.
16103         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16104         (rs6000_option_override_internal): Assign to
16105         targetm.secondary_memory_needed rather than
16106         rs6000_secondary_memory_needed_ptr.
16107         (rs6000_secondary_memory_needed): Match hook interface.
16108         (rs6000_debug_secondary_memory_needed): Likewise.
16109         * config/riscv/riscv.h (SECONDARY_MEMORY_NEEDED): Delete.
16110         * config/riscv/riscv.c (riscv_secondary_memory_needed): New function.
16111         (riscv_register_move_cost): Use it instead of SECONDARY_MEMORY_NEEDED.
16112         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16113         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED): Delete.
16114         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_ptr):
16115         Delete.
16116         * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_ptr): Delete.
16117         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16118         (rs6000_option_override_internal): Assign to
16119         targetm.secondary_memory_needed rather than
16120         rs6000_secondary_memory_needed_ptr.
16121         (rs6000_secondary_memory_needed): Match hook interface.
16122         (rs6000_debug_secondary_memory_needed): Likewise.
16123         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED): Delete.
16124         * config/s390/s390.c (s390_secondary_memory_needed): New function.
16125         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16126         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED): Delete.
16127         * config/sparc/sparc.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16128         (sparc_secondary_memory_needed): New function.
16129         * lra-constraints.c (check_and_process_move): Refer to
16130         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
16131         in comment.
16132         (curr_insn_transform): Likewise.
16133         (process_alt_operands): Use targetm.secondary_memory_needed
16134         instead of TARGET_SECONDARY_MEMORY_NEEDED.
16135         (check_secondary_memory_needed_p): Likewise.
16136         (choose_split_class): Likewise.
16137         * reload.c: Unconditionally include code that was previously
16138         conditional on SECONDARY_MEMORY_NEEDED.
16139         (push_secondary_reload): Use targetm.secondary_memory_needed
16140         instead of TARGET_SECONDARY_MEMORY_NEEDED.
16141         (push_reload): Likewise.
16142         * reload1.c: Unconditionally include code that was previously
16143         conditional on SECONDARY_MEMORY_NEEDED.
16144         (choose_reload_regs): Use targetm.secondary_memory_needed
16145         instead of TARGET_SECONDARY_MEMORY_NEEDED.
16146         (gen_reload): Likewise.
16147         * system.h (SECONDARY_MEMORY_NEEDED): Poison.
16149 2017-09-13  Richard Sandiford  <richard.sandiford@linaro.org>
16150             Alan Hayward  <alan.hayward@arm.com>
16151             David Sherwood  <david.sherwood@arm.com>
16153         * target.def (secondary_memory_needed_mode): New hook:
16154         * targhooks.c (default_secondary_memory_needed_mode): Declare.
16155         * targhooks.h (default_secondary_memory_needed_mode): New function.
16156         * doc/tm.texi.in (SECONDARY_MEMORY_NEEDED_MODE): Replace with...
16157         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): ...this.
16158         * doc/tm.texi: Regenerate.
16159         * lra-constraints.c (check_and_process_move): Use
16160         targetm.secondary_memory_needed_mode instead of
16161         TARGET_SECONDARY_MEMORY_NEEDED_MODE.
16162         (curr_insn_transform): Likewise.
16163         * reload.c (get_secondary_mem): Likewise.
16164         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16165         * config/alpha/alpha.c (alpha_secondary_memory_needed_mode): New
16166         function.
16167         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16168         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16169         * config/i386/i386.c (ix86_secondary_memory_needed_mode): New function.
16170         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16171         * config/powerpcspe/powerpcspe.h (SECONDARY_MEMORY_NEEDED_MODE):
16172         Delete.
16173         * config/powerpcspe/powerpcspe-protos.h
16174         (rs6000_secondary_memory_needed_mode): Delete.
16175         * config/powerpcspe/powerpcspe.c
16176         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16177         (rs6000_secondary_memory_needed_mode): Make static.
16178         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16179         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_mode):
16180         Delete.
16181         * config/rs6000/rs6000.c (TARGET_SECONDARY_MEMORY_NEEDED_MODE):
16182         Redefine.
16183         (rs6000_secondary_memory_needed_mode): Make static.
16184         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16185         * config/s390/s390.c (s390_secondary_memory_needed_mode): New function.
16186         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16187         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16188         * config/sparc/sparc.c (TARGET_SECONDARY_MEMORY_NEEDED_MODE):
16189         Redefine.
16190         (sparc_secondary_memory_needed_mode): New function.
16191         * system.h (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Poison.
16193 2017-09-13  Jackson Woodruff  <jackson.woodruff@arm.com>
16195         * config/aarch64/constraints.md (Umq): New constraint.
16196         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>):
16197         Change to use Umq.
16198         (mov<mode>): Update condition.
16200 2017-09-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16202         * gimple-ssa-store-merging.c (sort_by_bitpos): Compare store order
16203         when bitposition is the same.
16205 2017-09-13  Richard Biener  <rguenther@suse.de>
16207         * dwarf2out.c (output_die_symbol): Remove.
16208         (output_die): Do not output a DIEs symbol.
16210 2017-09-13  Richard Biener  <rguenther@suse.de>
16212         PR middle-end/82128
16213         * gimple-fold.c (gimple_fold_call): Update SSA name in-place to
16214         default-def to avoid breaking iterator update with the weird
16215         interaction with cgraph_update_edges_for_call_stmt_node.
16217 2017-09-13  Richard Biener  <rguenther@suse.de>
16219         * tree-cfg.c (verify_gimple_assign_binary): Add verification
16220         for WIDEN_SUM_EXPR, VEC_WIDEN_MULT_{HI,LO,EVEN,ODD}_EXPR,
16221         VEC_PACK_{TRUNC,SAT,FIX_TRUNC}_EXPR.
16222         (verify_gimple_assign_ternary): Add verification for DOT_PROD_EXPR.
16224 2017-09-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
16226         * config/aarch64/aarch64.c (aarch64_override_options_after_change_1):
16227         Disable pc relative literal load irrespective of
16228         TARGET_FIX_ERR_A53_84341 for default.
16230 2017-09-12  Eric Botcazou  <ebotcazou@adacore.com>
16232         * config/sparc/sparc.c (output_return): Output the source location of
16233         the insn in the delay slot, if any.
16234         (output_sibcall): Likewise.
16236 2017-09-12  H.J. Lu  <hongjiu.lu@intel.com>
16238         PR driver/81498
16239         * common.opt (-static-pie): New alias.
16240         (shared): Negate static-pie.
16241         (-no-pie): Update help text.
16242         (-pie): Likewise.
16243         (static-pie): New option.
16244         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Add
16245         -static-pie support.
16246         (GNU_USER_TARGET_ENDFILE_SPEC): Likewise.
16247         (LINK_EH_SPEC): Likewise.
16248         (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
16249         * config/i386/gnu-user.h (GNU_USER_TARGET_LINK_SPEC): Likewise.
16250         * config/i386/gnu-user64.h (GNU_USER_TARGET_LINK_SPEC): Likewise.
16251         * gcc.c (LINK_COMMAND_SPEC): Likewise.
16252         (init_gcc_specs): Likewise.
16253         (init_spec): Likewise.
16254         (display_help): Update help message for -pie.
16255         * doc/invoke.texi: Update -pie, -no-pie and -static.  Document
16256         -static-pie.
16258 2017-09-12  Wilco Dijkstra  <wdijkstr@arm.com>
16260         * config/aarch64/aarch64.md (movsi_aarch64): Remove all '*'.
16261         (movdi_aarch64): Likewise.
16262         (movti_aarch64): Likewise.
16264 2017-09-12 Simon Wright <simon@pushface.org>
16266         PR target/80204
16267         * config/darwin-driver.c (darwin_find_version_from_kernel): Eliminate
16268         calculation of the minor version, always output as 0.
16270 2017-09-12  Jakub Jelinek  <jakub@redhat.com>
16272         PR target/82112
16273         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): For
16274         ALTIVEC_BUILTIN_VEC_LD if arg1 has array type call default_conversion
16275         on it early, rather than manual conversion late.  For
16276         ALTIVEC_BUILTIN_VEC_ST if arg2 has array type call default_conversion
16277         instead of performing manual conversion.
16279 2017-09-12  Carl Love  <cel@us.ibm.com>
16281         * config/rs6000/altivec.md (vec_widen_umult_even_v4si,
16282         vec_widen_smult_even_v4si): Add define expands for vmuleuw, vmulesw,
16283         vmulouw, vmulosw.
16284         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
16285         VMULOSW): Add definitions.
16286         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
16287         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
16288         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
16289         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
16290         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
16292 2017-09-12  James Greenhalgh  <james.greenhalgh@arm.com>
16294         * config/aarch64/aarch64.md (movdi_aarch64): Set load/store
16295         types correctly.
16296         (movti_aarch64): Likewise.
16297         (movdf_aarch64): Likewise.
16298         (movtf_aarch64): Likewise.
16299         (load_pairdi): Likewise.
16300         (store_pairdi): Likewise.
16301         (load_pairdf): Likewise.
16302         (store_pairdf): Likewise.
16303         (loadwb_pair<GPI:mode>_<P:mode>): Likewise.
16304         (storewb_pair<GPI:mode>_<P:mode>): Likewise.
16305         (ldr_got_small_<mode>): Likewise.
16306         (ldr_got_small_28k_<mode>): Likewise.
16307         (ldr_got_tiny): Likewise.
16308         * config/aarch64/iterators.md (ldst_sz): New.
16309         (ldpstp_sz): Likewise.
16310         * config/aarch64/thunderx.md (thunderx_storepair): Split store_8
16311         to store_16.
16312         (thunderx_load): Split load_8 to load_16.
16313         * config/aarch64/thunderx2t99.md (thunderx2t99_loadpair): Split
16314         load_8 to load_16.
16315         (thunderx2t99_storepair_basic): Split store_8 to store_16.
16316         * config/arm/xgene1.md (xgene1_load_pair): Split load_8 to load_16.
16317         (xgene1_store_pair): Split store_8 to store_16.
16318         * config/aarch64/falkor.md (falkor_ld_3_ld): Split load_8 to load_16.
16319         (falkor_st_0_st_sd): Split store_8 to store_16.
16321 2017-09-12  James Greenhalgh  <james.greenhalgh@arm.com>
16323         * config/arm/types.md (type): Rename load1/2/3/4 to load_4/8/12/16
16324         and store1/2/3/4 to store_4/8/12/16.
16325         * config/aarch64/aarch64.md: Update for rename.
16326         * config/arm/arm.md: Likewise.: Likewise.
16327         * config/arm/arm.c: Likewise.
16328         * config/arm/thumb1.md: Likewise.
16329         * config/arm/thumb2.md: Likewise.
16330         * config/arm/vfp.md: Likewise.
16331         * config/arm/arm-generic.md: Likewise.
16332         * config/arm/arm1020e.md: Likewise.
16333         * config/arm/arm1026ejs.md: Likewise.
16334         * config/arm/arm1136jfs.md: Likewise.
16335         * config/arm/arm926ejs.md: Likewise.
16336         * config/arm/cortex-a15.md: Likewise.
16337         * config/arm/cortex-a17.md: Likewise.
16338         * config/arm/cortex-a5.md: Likewise.
16339         * config/arm/cortex-a53.md: Likewise.
16340         * config/arm/cortex-a57.md: Likewise.
16341         * config/arm/cortex-a7.md: Likewise.
16342         * config/arm/cortex-a8.md: Likewise.
16343         * config/arm/cortex-a9.md: Likewise.
16344         * config/arm/cortex-m4.md: Likewise.
16345         * config/arm/cortex-m7.md: Likewise.
16346         * config/arm/cortex-r4.md: Likewise.
16347         * config/arm/exynos-m1.md: Likewise.
16348         * config/arm/fa526.md: Likewise.
16349         * config/arm/fa606te.md: Likewise.
16350         * config/arm/fa626te.md: Likewise.
16351         * config/arm/fa726te.md: Likewise.
16352         * config/arm/fmp626.md: Likewise.
16353         * config/arm/iwmmxt.md: Likewise.
16354         * config/arm/ldmstm.md: Likewise.
16355         * config/arm/marvell-pj4.md: Likewise.
16356         * config/arm/xgene1.md: Likewise.
16357         * config/aarch64/thunderx.md: Likewise.
16358         * config/aarch64/thunderx2t99.md: Likewise.
16359         * config/aarch64/falkor.md: Likewise.
16361 2017-09-12  Martin Liska  <mliska@suse.cz>
16363         * attribs.c (private_lookup_attribute): New function.
16364         * attribs.h (private_lookup_attribute): Declared here.
16365         (lookup_attribute): Called from this place.
16367 2017-09-12  Richard Biener  <rguenther@suse.de>
16369         PR tree-optimization/82157
16370         * tree-ssa-pre.c (remove_dead_inserted_code): Do not remove
16371         stmts with side-effects.
16373 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16374             Alan Hayward  <alan.hayward@arm.com>
16375             David Sherwood <david.sherwood@arm.com>
16377         * target.def (hard_regno_nregs): New hook.
16378         (class_max_nregs): Refer to it instead of HARD_REGNO_NREGS.
16379         * targhooks.h (default_hard_regno_nregs): Declare.
16380         * targhooks.c (default_hard_regno_nregs): New function.
16381         * doc/tm.texi.in (HARD_REGNO_NREGS): Replace with...
16382         (TARGET_HARD_REGNO_NREGS): ...this hook.
16383         (HARD_REGNO_NREGS_HAS_PADDING): Update accordingly.
16384         (CLASS_MAX_NREGS): Likewise.
16385         * doc/tm.texi: Regenerate.
16386         * reginfo.c (init_reg_modes_target): Use targetm.hard_regno_nregs
16387         instead of HARD_REGNO_NREGS.
16388         * rtl.h (REG_NREGS): Refer to TARGET_HARD_REGNO_NREGS rather than
16389         HARD_REGNO_NREGS in the comment.
16390         * config/aarch64/aarch64.h (HARD_REGNO_NREGS): Delete.
16391         * config/aarch64/aarch64-protos.h (aarch64_hard_regno_nregs): Delete.
16392         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs): Make static.
16393         Return an unsigned int.
16394         (TARGET_HARD_REGNO_NREGS): Redefine.
16395         * config/alpha/alpha.h (HARD_REGNO_NREGS): Delete.
16396         * config/arc/arc.h (HARD_REGNO_NREGS): Delete.
16397         * config/arc/arc.c (TARGET_HARD_REGNO_NREGS): Redefine.
16398         (arc_hard_regno_nregs): New function.
16399         * config/arm/arm.h (HARD_REGNO_NREGS): Delete.
16400         * config/arm/arm.c (TARGET_HARD_REGNO_NREGS): Redefine.
16401         (arm_hard_regno_nregs): New function.
16402         * config/avr/avr.h (HARD_REGNO_NREGS): Delete.
16403         * config/bfin/bfin.h (HARD_REGNO_NREGS): Delete.
16404         * config/bfin/bfin.c (bfin_hard_regno_nregs): New function.
16405         (TARGET_HARD_REGNO_NREGS): Redefine.
16406         * config/c6x/c6x.h (HARD_REGNO_NREGS): Delete.
16407         * config/cr16/cr16.h (LONG_REG_P): Use targetm.hard_regno_nregs.
16408         (HARD_REGNO_NREGS): Delete.
16409         * config/cr16/cr16.c (TARGET_HARD_REGNO_NREGS): Redefine.
16410         (cr16_hard_regno_nregs): New function.
16411         (cr16_memory_move_cost): Use it instead of HARD_REGNO_NREGS.
16412         * config/cris/cris.h (HARD_REGNO_NREGS): Delete.
16413         * config/cris/cris.c (TARGET_HARD_REGNO_NREGS): Redefine.
16414         (cris_hard_regno_nregs): New function.
16415         * config/epiphany/epiphany.h (HARD_REGNO_NREGS): Delete.
16416         * config/fr30/fr30.h (HARD_REGNO_NREGS): Delete.
16417         (CLASS_MAX_NREGS): Use targetm.hard_regno_nregs.
16418         * config/frv/frv.h (HARD_REGNO_NREGS): Delete.
16419         (CLASS_MAX_NREGS): Remove outdated copy of documentation.
16420         * config/frv/frv-protos.h (frv_hard_regno_nregs): Delete.
16421         * config/frv/frv.c (TARGET_HARD_REGNO_NREGS): Redefine.
16422         (frv_hard_regno_nregs): Make static.  Take and return an
16423         unsigned int.
16424         (frv_class_max_nregs): Remove outdated copy of documentation.
16425         * config/ft32/ft32.h (HARD_REGNO_NREGS): Delete.
16426         * config/h8300/h8300.h (HARD_REGNO_NREGS): Delete.
16427         * config/h8300/h8300-protos.h (h8300_hard_regno_nregs): Delete.
16428         * config/h8300/h8300.c (h8300_hard_regno_nregs): Delete.
16429         * config/i386/i386.h (HARD_REGNO_NREGS): Delete.
16430         * config/i386/i386.c (ix86_hard_regno_nregs): New function.
16431         (TARGET_HARD_REGNO_NREGS): Redefine.
16432         * config/ia64/ia64.h (HARD_REGNO_NREGS): Delete.
16433         (CLASS_MAX_NREGS): Update comment.
16434         * config/ia64/ia64.c (TARGET_HARD_REGNO_NREGS): Redefine.
16435         (ia64_hard_regno_nregs): New function.
16436         * config/iq2000/iq2000.h (HARD_REGNO_NREGS): Delete.
16437         * config/lm32/lm32.h (HARD_REGNO_NREGS): Delete.
16438         * config/m32c/m32c.h (HARD_REGNO_NREGS): Delete.
16439         * config/m32c/m32c-protos.h (m32c_hard_regno_nregs): Delete.
16440         * config/m32c/m32c.c (m32c_hard_regno_nregs_1): Take and return
16441         an unsigned int.
16442         (m32c_hard_regno_nregs): Likewise.  Make static.
16443         (TARGET_HARD_REGNO_NREGS): Redefine.
16444         * config/m32r/m32r.h (HARD_REGNO_NREGS): Delete.
16445         * config/m68k/m68k.h (HARD_REGNO_NREGS): Delete.
16446         * config/m68k/m68k.c (TARGET_HARD_REGNO_NREGS): Redefine.
16447         (m68k_hard_regno_nregs): New function.
16448         * config/mcore/mcore.h (HARD_REGNO_NREGS): Delete.
16449         * config/microblaze/microblaze.h (HARD_REGNO_NREGS): Delete.
16450         * config/mips/mips.h (HARD_REGNO_NREGS): Delete.
16451         * config/mips/mips-protos.h (mips_hard_regno_nregs): Delete.
16452         * config/mips/mips.c (mips_hard_regno_nregs): Make static.
16453         Take and return an unsigned int.
16454         (TARGET_HARD_REGNO_NREGS): Redefine.
16455         * config/mmix/mmix.h (HARD_REGNO_NREGS): Delete.
16456         (CLASS_MAX_NREGS): Use targetm.hard_regno_nregs.
16457         * config/mn10300/mn10300.h (HARD_REGNO_NREGS): Delete.
16458         * config/moxie/moxie.h (HARD_REGNO_NREGS): Delete.
16459         * config/msp430/msp430.h (HARD_REGNO_NREGS): Delete.
16460         * config/msp430/msp430-protos.h (msp430_hard_regno_nregs): Delete.
16461         * config/msp430/msp430.c (TARGET_HARD_REGNO_NREGS): Redefine.
16462         (msp430_hard_regno_nregs): Make static.  Take and return an
16463         unsigned int.
16464         * config/nds32/nds32.h (HARD_REGNO_NREGS): Delete.
16465         * config/nds32/nds32-protos.h (nds32_hard_regno_nregs): Delete.
16466         * config/nds32/nds32.c (nds32_hard_regno_nregs): Delete.
16467         (nds32_hard_regno_mode_ok): Use targetm.hard_regno_nregs.
16468         * config/nios2/nios2.h (HARD_REGNO_NREGS): Delete.
16469         * config/nvptx/nvptx.h (HARD_REGNO_NREGS): Delete.
16470         * config/nvptx/nvptx.c (nvptx_hard_regno_nregs): New function.
16471         (TARGET_HARD_REGNO_NREGS): Redefine.
16472         * config/pa/pa32-regs.h (HARD_REGNO_NREGS): Rename to...
16473         (PA_HARD_REGNO_NREGS): ...this.
16474         * config/pa/pa64-regs.h (HARD_REGNO_NREGS): Rename to...
16475         (PA_HARD_REGNO_NREGS): ...this.
16476         * config/pa/pa.c (TARGET_HARD_REGNO_NREGS): Redefine.
16477         (pa_hard_regno_nregs): New function.
16478         * config/pdp11/pdp11.h (HARD_REGNO_NREGS): Delete.
16479         * config/pdp11/pdp11.c (TARGET_HARD_REGNO_NREGS): Redefine.
16480         (pdp11_hard_regno_nregs): New function.
16481         * config/powerpcspe/powerpcspe.h (HARD_REGNO_NREGS): Delete.
16482         * config/powerpcspe/powerpcspe.c (TARGET_HARD_REGNO_NREGS): Redefine.
16483         (rs6000_hard_regno_nregs_hook): New function.
16484         * config/riscv/riscv.h (HARD_REGNO_NREGS): Delete.
16485         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): Delete.
16486         * config/riscv/riscv.c (riscv_hard_regno_nregs): Make static.
16487         Take and return an unsigned int.  Move earlier in file.
16488         (TARGET_HARD_REGNO_NREGS): Redefine.
16489         * config/rl78/rl78.h (HARD_REGNO_NREGS): Delete.
16490         * config/rl78/rl78-protos.h (rl78_hard_regno_nregs): Delete.
16491         * config/rl78/rl78.c (TARGET_HARD_REGNO_NREGS): Reefine.
16492         (rl78_hard_regno_nregs): Make static.  Take and return an
16493         unsigned int.
16494         * config/rs6000/rs6000.h (HARD_REGNO_NREGS): Delete.
16495         * config/rs6000/rs6000.c (TARGET_HARD_REGNO_NREGS): Redefine.
16496         (rs6000_hard_regno_nregs_hook): New function.
16497         * config/rx/rx.h (HARD_REGNO_NREGS): Delete.
16498         * config/rx/rx.c (rx_hard_regno_nregs): New function.
16499         (TARGET_HARD_REGNO_NREGS): Redefine.
16500         * config/s390/s390.h (HARD_REGNO_NREGS): Delete.
16501         * config/s390/s390.c (REGNO_PAIR_OK): Use s390_hard_regno_nregs
16502         instead of HARD_REGNO_NREGS.
16503         (s390_hard_regno_nregs): New function.
16504         (s390_hard_regno_mode_ok): Add comment from s390.h.
16505         (TARGET_HARD_REGNO_NREGS): Redefine.
16506         * config/sh/sh.h (HARD_REGNO_NREGS): Delete.
16507         * config/sh/sh.c (TARGET_HARD_REGNO_NREGS): Redefine.
16508         (sh_hard_regno_nregs): New function.
16509         (sh_pass_in_reg_p): Use it.
16510         * config/sparc/sparc.h (HARD_REGNO_NREGS): Delete.
16511         * config/sparc/sparc.c (TARGET_HARD_REGNO_NREGS): Redefine.
16512         (sparc_hard_regno_nregs): New function.
16513         * config/spu/spu.h (HARD_REGNO_NREGS): Delete.
16514         * config/spu/spu.c (spu_hard_regno_nregs): New function.
16515         (spu_function_arg_advance): Use it, supplying a valid register number.
16516         (TARGET_HARD_REGNO_NREGS): Redefine.
16517         * config/stormy16/stormy16.h (HARD_REGNO_NREGS): Delete.
16518         * config/tilegx/tilegx.h (HARD_REGNO_NREGS): Delete.
16519         * config/tilepro/tilepro.h (HARD_REGNO_NREGS): Delete.
16520         * config/v850/v850.h (HARD_REGNO_NREGS): Delete.
16521         * config/vax/vax.h (HARD_REGNO_NREGS): Delete.
16522         * config/visium/visium.h (HARD_REGNO_NREGS): Delete.
16523         (CLASS_MAX_NREGS): Remove copy of old documentation.
16524         * config/visium/visium.c (TARGET_HARD_REGNO_NREGS): Redefine.
16525         (visium_hard_regno_nregs): New function.
16526         (visium_hard_regno_mode_ok): Use it instead of HARD_REGNO_NREGS.
16527         * config/xtensa/xtensa.h (HARD_REGNO_NREGS): Delete.
16528         * config/xtensa/xtensa.c (TARGET_HARD_REGNO_NREGS): Redefine.
16529         xtensa_hard_regno_nregs): New function.
16530         * system.h (HARD_REGNO_NREGS): Poison.
16532 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16534         * config/arm/arm.h (THUMB_SECONDARY_INPUT_RELOAD_CLASS): Use
16535         hard_regno_nregs instead of HARD_REGNO_NREGS.
16536         (THUMB_SECONDARY_OUTPUT_RELOAD_CLASS): Likewise.
16537         * config/c6x/c6x.c (c6x_expand_prologue): Likewise.
16538         (c6x_expand_epilogue): Likewise.
16539         * config/frv/frv.c (frv_alloc_temp_reg): Likewise.
16540         (frv_read_iacc_argument): Likewise.
16541         * config/sh/sh.c: Include regs.h.
16542         (sh_print_operand): Use hard_regno_nregs instead of HARD_REGNO_NREGS.
16543         (regs_used): Likewise.
16544         (output_stack_adjust): Likewise.
16545         * config/xtensa/xtensa.c (xtensa_copy_incoming_a7): Likewise.
16546         * expmed.c: Include regs.h.
16547         (store_bit_field_1): Use hard_regno_nregs instead of HARD_REGNO_NREGS.
16548         * ree.c: Include regs.h.
16549         (combine_reaching_defs): Use hard_regno_nregs instead of
16550         HARD_REGNO_NREGS.
16551         (add_removable_extension): Likewise.
16553 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16555         * regs.h (hard_regno_nregs): Turn into a function.
16556         (end_hard_regno): Update accordingly.
16557         * caller-save.c (setup_save_areas): Likewise.
16558         (save_call_clobbered_regs): Likewise.
16559         (replace_reg_with_saved_mem): Likewise.
16560         (insert_restore): Likewise.
16561         (insert_save): Likewise.
16562         * combine.c (can_change_dest_mode): Likewise.
16563         (move_deaths): Likewise.
16564         (distribute_notes): Likewise.
16565         * config/mips/mips.c (mips_hard_regno_call_part_clobbered): Likewise.
16566         * config/powerpcspe/powerpcspe.c (rs6000_cannot_change_mode_class)
16567         (rs6000_split_multireg_move): Likewise.
16568         (rs6000_register_move_cost): Likewise.
16569         (rs6000_memory_move_cost): Likewise.
16570         * config/rs6000/rs6000.c (rs6000_cannot_change_mode_class): Likewise.
16571         (rs6000_split_multireg_move): Likewise.
16572         (rs6000_register_move_cost): Likewise.
16573         (rs6000_memory_move_cost): Likewise.
16574         * cselib.c (cselib_reset_table): Likewise.
16575         (cselib_lookup_1): Likewise.
16576         * emit-rtl.c (set_mode_and_regno): Likewise.
16577         * function.c (aggregate_value_p): Likewise.
16578         * ira-color.c (setup_profitable_hard_regs): Likewise.
16579         (check_hard_reg_p): Likewise.
16580         (calculate_saved_nregs): Likewise.
16581         (assign_hard_reg): Likewise.
16582         (improve_allocation): Likewise.
16583         (calculate_spill_cost): Likewise.
16584         * ira-emit.c (modify_move_list): Likewise.
16585         * ira-int.h (ira_hard_reg_set_intersection_p): Likewise.
16586         (ira_hard_reg_in_set_p): Likewise.
16587         * ira.c (setup_reg_mode_hard_regset): Likewise.
16588         (clarify_prohibited_class_mode_regs): Likewise.
16589         (check_allocation): Likewise.
16590         * lra-assigns.c (find_hard_regno_for_1): Likewise.
16591         (lra_setup_reg_renumber): Likewise.
16592         (setup_try_hard_regno_pseudos): Likewise.
16593         (spill_for): Likewise.
16594         (assign_hard_regno): Likewise.
16595         (setup_live_pseudos_and_spill_after_risky_transforms): Likewise.
16596         * lra-constraints.c (in_class_p): Likewise.
16597         (lra_constraint_offset): Likewise.
16598         (simplify_operand_subreg): Likewise.
16599         (lra_constraints): Likewise.
16600         (split_reg): Likewise.
16601         (split_if_necessary): Likewise.
16602         (invariant_p): Likewise.
16603         (inherit_in_ebb): Likewise.
16604         * lra-lives.c (process_bb_lives): Likewise.
16605         * lra-remat.c (reg_overlap_for_remat_p): Likewise.
16606         (get_hard_regs): Likewise.
16607         (do_remat): Likewise.
16608         * lra-spills.c (assign_spill_hard_regs): Likewise.
16609         * mode-switching.c (create_pre_exit): Likewise.
16610         * postreload.c (reload_combine_recognize_pattern): Likewise.
16611         * recog.c (peep2_find_free_register): Likewise.
16612         * regcprop.c (kill_value_regno): Likewise.
16613         (set_value_regno): Likewise.
16614         (copy_value): Likewise.
16615         (maybe_mode_change): Likewise.
16616         (find_oldest_value_reg): Likewise.
16617         (copyprop_hardreg_forward_1): Likewise.
16618         * regrename.c (check_new_reg_p): Likewise.
16619         (regrename_do_replace): Likewise.
16620         * reload.c (push_reload): Likewise.
16621         (combine_reloads): Likewise.
16622         (find_dummy_reload): Likewise.
16623         (operands_match_p): Likewise.
16624         (find_reloads): Likewise.
16625         (find_equiv_reg): Likewise.
16626         (reload_adjust_reg_for_mode): Likewise.
16627         * reload1.c (count_pseudo): Likewise.
16628         (count_spilled_pseudo): Likewise.
16629         (find_reg): Likewise.
16630         (clear_reload_reg_in_use): Likewise.
16631         (free_for_value_p): Likewise.
16632         (allocate_reload_reg): Likewise.
16633         (choose_reload_regs): Likewise.
16634         (reload_adjust_reg_for_temp): Likewise.
16635         (emit_reload_insns): Likewise.
16636         (delete_output_reload): Likewise.
16637         * rtlanal.c (subreg_get_info): Likewise.
16638         * sched-deps.c (sched_analyze_reg): Likewise.
16639         * sel-sched.c (init_regs_for_mode): Likewise.
16640         (mark_unavailable_hard_regs): Likewise.
16641         (choose_best_reg_1): Likewise.
16642         (verify_target_availability): Likewise.
16643         * valtrack.c (dead_debug_insert_temp): Likewise.
16644         * var-tracking.c (track_loc_p): Likewise.
16645         (emit_note_insn_var_location): Likewise.
16646         * varasm.c (make_decl_rtl): Likewise.
16647         * reginfo.c (choose_hard_reg_mode): Likewise.
16648         (init_reg_modes_target): Refer directly to
16649         this_target_regs->x_hard_regno_nregs.
16651 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16653         * ira-costs.c (record_operand_costs): Use in_hard_reg_set_p
16654         instead of hard_regno_nregs.
16656 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16658         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Use
16659         end_hard_regno instead of hard_regno_nregs.
16660         * config/s390/s390.c (s390_reg_clobbered_rtx): Likewise.
16661         * config/sparc/sparc.h (ASM_DECLARE_REGISTER_GLOBAL): Likewise.
16662         * config/visium/visium.c (visium_hard_regno_mode_ok): Likewise.
16663         * ira-color.c (improve_allocation): Likewise.
16664         * lra-assigns.c (find_hard_regno_for_1): Likewise.
16665         * lra-lives.c (mark_regno_live): Likewise.
16666         (mark_regno_dead): Likewise.
16667         * lra-remat.c (operand_to_remat): Likewise.
16668         * lra.c (collect_non_operand_hard_regs): Likewise.
16669         * postreload.c (reload_combine_note_store): Likewise.
16670         (move2add_valid_value_p): Likewise.
16671         * reload.c (regno_clobbered_p): Likewise.
16673 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16675         * config/frv/frv.c (FOR_EACH_REGNO): Use END_REGNO instead of
16676         hard_regno_nregs.
16677         * config/v850/v850.c (v850_reorg): Likewise.
16678         * reload.c (refers_to_regno_for_reload_p): Likewise.
16679         (find_equiv_reg): Likewise.
16680         * reload1.c (reload_reg_reaches_end_p): Likewise.
16682 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16684         * caller-save.c (add_used_regs): Use REG_NREGS instead of
16685         hard_regno_nregs.
16686         * config/aarch64/aarch64.c (aarch64_split_combinev16qi): Likewise.
16687         * config/arm/arm.c (output_move_neon): Likewise.
16688         (arm_attr_length_move_neon): Likewise.
16689         (neon_split_vcombine): Likewise.
16690         * config/c6x/c6x.c (c6x_mark_reg_read): Likewise.
16691         (c6x_mark_reg_written): Likewise.
16692         (c6x_dwarf_register_span): Likewise.
16693         * config/i386/i386.c (ix86_save_reg): Likewise.
16694         * config/ia64/ia64.c (mark_reg_gr_used_mask): Likewise.
16695         (rws_access_reg): Likewise.
16696         * config/s390/s390.c (s390_call_saved_register_used): Likewise.
16697         * mode-switching.c (create_pre_exit): Likewise.
16698         * ree.c (combine_reaching_defs): Likewise.
16699         (add_removable_extension): Likewise.
16700         * regcprop.c (find_oldest_value_reg): Likewise.
16701         (copyprop_hardreg_forward_1): Likewise.
16702         * reload.c (reload_inner_reg_of_subreg): Likewise.
16703         (push_reload): Likewise.
16704         (combine_reloads): Likewise.
16705         (find_dummy_reload): Likewise.
16706         (reload_adjust_reg_for_mode): Likewise.
16707         * reload1.c (find_reload_regs): Likewise.
16708         (forget_old_reloads_1): Likewise.
16709         (reload_reg_free_for_value_p): Likewise.
16710         (reload_adjust_reg_for_temp): Likewise.
16711         (emit_reload_insns): Likewise.
16712         (delete_output_reload): Likewise.
16713         * sel-sched.c (choose_best_reg_1): Likewise.
16714         (choose_best_pseudo_reg): Likewise.
16716 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16717             Alan Hayward  <alan.hayward@arm.com>
16718             David Sherwood <david.sherwood@arm.com>
16720         * defaults.h (SLOW_UNALIGNED_ACCESS): Delete.
16721         * target.def (slow_unaligned_access): New hook.
16722         * targhooks.h (default_slow_unaligned_access): Declare.
16723         * targhooks.c (default_slow_unaligned_access): New function.
16724         * doc/tm.texi.in (SLOW_UNALIGNED_ACCESS): Replace with...
16725         (TARGET_SLOW_UNALIGNED_ACCESS): ...this.
16726         * doc/tm.texi: Regenerate.
16727         * config/alpha/alpha.h (SLOW_UNALIGNED_ACCESS): Delete.
16728         * config/arm/arm.h (SLOW_UNALIGNED_ACCESS): Delete.
16729         * config/i386/i386.h (SLOW_UNALIGNED_ACCESS): Delete commented-out
16730         definition.
16731         * config/powerpcspe/powerpcspe.h (SLOW_UNALIGNED_ACCESS): Delete.
16732         * config/powerpcspe/powerpcspe.c (TARGET_SLOW_UNALIGNED_ACCESS):
16733         Redefine.
16734         (rs6000_slow_unaligned_access): New function.
16735         (rs6000_emit_move): Use it instead of SLOW_UNALIGNED_ACCESS.
16736         (expand_block_compare): Likewise.
16737         (expand_strn_compare): Likewise.
16738         (rs6000_rtx_costs): Likewise.
16739         * config/riscv/riscv.h (SLOW_UNALIGNED_ACCESS): Delete.
16740         (riscv_slow_unaligned_access): Likewise.
16741         * config/riscv/riscv.c (riscv_slow_unaligned_access): Rename to...
16742         (riscv_slow_unaligned_access_p): ...this and make static.
16743         (riscv_option_override): Update accordingly.
16744         (riscv_slow_unaligned_access): New function.
16745         (TARGET_SLOW_UNALIGNED_ACCESS): Redefine.
16746         * config/rs6000/rs6000.h (SLOW_UNALIGNED_ACCESS): Delete.
16747         * config/rs6000/rs6000.c (TARGET_SLOW_UNALIGNED_ACCESS): Redefine.
16748         (rs6000_slow_unaligned_access): New function.
16749         (rs6000_emit_move): Use it instead of SLOW_UNALIGNED_ACCESS.
16750         (rs6000_rtx_costs): Likewise.
16751         * config/rs6000/rs6000-string.c (expand_block_compare)
16752         (expand_strn_compare): Use targetm.slow_unaligned_access instead
16753         of SLOW_UNALIGNED_ACCESS.
16754         * config/tilegx/tilegx.h (SLOW_UNALIGNED_ACCESS): Delete.
16755         * config/tilepro/tilepro.h (SLOW_UNALIGNED_ACCESS): Delete.
16756         * calls.c (expand_call): Use targetm.slow_unaligned_access instead
16757         of SLOW_UNALIGNED_ACCESS.
16758         * expmed.c (simple_mem_bitfield_p): Likewise.
16759         * expr.c (alignment_for_piecewise_move): Likewise.
16760         (emit_group_load_1): Likewise.
16761         (emit_group_store): Likewise.
16762         (copy_blkmode_from_reg): Likewise.
16763         (emit_push_insn): Likewise.
16764         (expand_assignment): Likewise.
16765         (store_field): Likewise.
16766         (expand_expr_real_1): Likewise.
16767         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
16768         * lra-constraints.c (simplify_operand_subreg): Likewise.
16769         * stor-layout.c (bit_field_mode_iterator::next_mode): Likewise.
16770         * gimple-ssa-store-merging.c: Likewise in block comment at start
16771         of file.
16772         * tree-ssa-strlen.c: Include target.h.
16773         (handle_builtin_memcmp): Use targetm.slow_unaligned_access instead
16774         of SLOW_UNALIGNED_ACCESS.
16775         * system.h (SLOW_UNALIGNED_ACCESS): Poison.
16777 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16779         PR rtl-optimization/82185
16780         * expmed.c (emit_store_flag_int): Only test tem if it has been
16781         initialized.
16783 2017-09-12  Richard Biener  <rguenther@suse.de>
16785         PR middle-end/82149
16786         * match.pd ((FTYPE) N CMP CST): Fix typo.
16788 2017-09-12  Simon Atanasyan  <simon.atanasyan@imgtec.com>
16790         * config/mips/mips.c (mips_attribute_table): Add 'short_call'
16791         attribute.
16792         (mips_near_type_p): Add 'short_call' attribute as a synonym
16793         for 'near'.
16794         * doc/extend.texi (short_call): Document new function attribute.
16796 2017-09-12  Jakub Jelinek  <jakub@redhat.com>
16798         PR target/82112
16799         * c-common.c (sync_resolve_size): Instead of c_dialect_cxx ()
16800         assertion check that in the condition.
16801         (get_atomic_generic_size): Likewise.  Before testing if parameter
16802         has pointer type, if it has array type, call for C++
16803         default_conversion to perform array-to-pointer conversion.
16805 2017-09-12  Richard Biener  <rguenther@suse.de>
16807         * tree-vect-generic.c (expand_vector_operations_1): Do nothing
16808         for operations we cannot scalarize.
16810 2017-09-12  Aldy Hernandez  <aldyh@redhat.com>
16812         * tree-ssa-threadbackward.c (fsm_find_thread_path): Make GC
16813         vectors heap vectors.  Clean up comments.
16814         Make visited_bbs a reference.
16815         (profitable_jump_thread_path): Make GC
16816         vectors heap vectors.  Clean up comments.
16817         Misc cleanups.
16818         (convert_and_register_jump_thread_path): Make GC vectors heap
16819         vectors.
16820         (check_subpath_and_update_thread_path): Same.  Clean up comments.
16821         Make visited_bbs a reference.
16822         (handle_phi): Abstract common code to to
16823         register_jump_thread_path_if_profitable.
16824         Rename VAR_BB to DEF_BB.
16825         Update comments.
16826         Make GC vectors heap vectors.
16827         Make visited_bbs a reference.
16828         (handle_assignment): Same.
16829         (register_jump_thread_path_if_profitable): New.
16830         (fsm_find_control_statement_thread_paths): Rename VAR_BB to
16831         DEF_BB.
16832         Make GC vectors heap vectors.  Clean up comments.
16833         Make visited_bbs a reference.
16834         (find_jump_threads_backwards): Make visited_bbs live in the stack.
16835         * tree-ssa-threadupdate.c (delete_jump_thread_path): Fix typo in
16836         comment.
16838 2017-09-11  Max Filippov  <jcmvbkbc@gmail.com>
16840         PR target/82181
16841         * config/xtensa/xtensa.c (xtensa_mem_offset): Check that both
16842         words of E_DImode object are reachable by xtensa_uimm8x4 access.
16844 2017-09-11  Vidya Praveen  <vidyapraveen@arm.com>
16846         Revert r251800 and r251799.
16848 2017-09-11  Martin Jambor  <mjambor@suse.cz>
16850         PR hsa/82119
16851         * hsa-gen.c (gen_hsa_phi_from_gimple_phi): Process ADDR_EXPRs in
16852         arguments in advance.
16853         * hsa-regalloc.c (naive_process_phi): New parameter predecessors,
16854         use it to find predecessor edges.
16855         (naive_outof_ssa): Collect vector of predecessors.
16857 2017-09-08  Jason Merrill  <jason@redhat.com>
16859         PR c++/70029 - ICE with ref-qualifier and -flto
16860         * langhooks.h (struct lang_hooks_for_types): Add
16861         copy_lang_qualifiers.
16862         * attribs.c (build_type_attribute_qual_variant): Use it.
16863         * langhooks-def.h (LANG_HOOKS_COPY_LANG_QUALIFIERS): Default to
16864         NULL.
16865         (LANG_HOOKS_FOR_TYPES_INITIALIZER): Use it.
16866         * tree.c (verify_type): Re-enable TYPE_CANONICAL main variant check.
16868 2017-09-08  Eric Botcazou  <ebotcazou@adacore.com>
16870         PR target/81988
16871         * config/sparc/sparc.md (mulsi3): Rename into *mulsi3_sp32.
16872         (*mulsi3_sp64): New instruction.
16873         (mulsi3): New expander.
16875 2017-09-08  Uros Bizjak  <ubizjak@gmail.com>
16877         * config/alpha/alpha.c (alpha_print_operand) <case 'S'>: Remove.
16879 2017-09-08  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
16881         * sancov.c: Include memmodel.h.
16883 2017-09-07  Eric Botcazou  <ebotcazou@adacore.com>
16885         PR target/80897
16886         * config/sparc/sparc.c (sparc_emit_set_symbolic_const64): Deal with too
16887         large offsets.
16889 2017-09-07  Carl Love  <cel@us.ibm.com>
16891         * config/rs6000/vsx.md (define_insn "*stxvl"): Add missing argument to
16892         the sldi instruction.
16894 2017-09-07  David Edelsohn  <dje.gcc@gmail.com>
16896         * sancov.c: Include tm_p.h.
16898 2017-09-07  Jakub Jelinek  <jakub@redhat.com>
16900         PR target/81979
16901         * output.h (switch_to_other_text_partition): New declaration.
16902         * varasm.c (switch_to_other_text_partition): New function.
16903         * config/rs6000/rs6000.c (uses_TOC): Return 2 if
16904         NOTE_INSN_SWITCH_TEXT_SECTIONS is seen before finding load_toc_* insn.
16905         (rs6000_elf_declare_function_name): If uses_TOC returned 2, switch
16906         to the other text partition before emitting LCL label and switch back
16907         after emitting the word after it.
16909 2017-09-07  Richard Biener  <rguenther@suse.de>
16911         * passes.def (pass_split_crit_edges): Remove instance before PRE.
16912         * tree-ssa-pre.c (pass_pre::execute): Instead manually split
16913         critical edges here, after loop init.
16914         (pass_data_pre): Remove PROP_no_crit_edges flags.
16915         * tree-ssa-sccvn.c (vn_reference_lookup_3): Use vn_valueize
16916         for valueization of call args to avoid leaking VN_TOP.
16917         (visit_use): Assert we do not visit default defs.
16918         (init_scc_vn): Use build_decl for VN_TOP to make name nicer.
16919         Use error_mark_node to more easily detect leaking VN_TOP.
16920         All default-defs are varying, not VN_TOP.  Mark them visited.
16921         (run_scc_vn): Make code match comment.
16923 2017-09-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
16925         * config/rs6000/rs6000-cpus.def (OTHER_VSX_VECTOR_MASKS): Delete
16926         OPTION_MASK_FLOAT128_KEYWORD.
16927         (POWERPC_MASKS): Likewise.
16928         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Delete
16929         support for the -mfloat128-type option, and make -mfloat128
16930         default on PowerPC Linux systems.  Define or undefine
16931         __FLOAT128__ and  __FLOAT128_HARDWARE__ for the current options.
16932         Define __float128 to be __ieee128 if IEEE 128-bit support is
16933         enabled, or undefine it.
16934         (rs6000_cpu_cpp_builtins): Delete defining __FLOAT128__ here.
16935         Delete defining __FLOAT128_TYPE__.
16936         * config/rs6000/rs6000.opt (x_TARGET_FLOAT128_TYPE): Delete the
16937         -mfloat128-type option and make -mfloat128 default on PowerPC
16938         Linux systems.
16939         (TARGET_FLOAT128_TYPE): Likewise.
16940         (-mfloat128-type): Likewise.
16941         * config/rs6000/rs6000.c (rs6000_option_override_internal):
16942         Delete the -mfloat128-type option and make -mfloat128 default on
16943         PowerPC Linux systems.  Always use __ieee128 to be the keyword for
16944         the IEEE 128-bit type, and map __float128 to __ieee128 if IEEE
16945         128-bit floating point is enabled.  Change tests from using
16946         -mfloat128-type to -mfloat128.
16947         (rs6000_mangle_type): Use the correct mangling for the __float128
16948         type even if normal long double is restricted to 64-bits.
16949         (floatn_mode): Enable the _Float128 type by default on VSX Linux
16950         systems.
16951         * config/rs6000/rs6000.h (MASK_FLOAT128_TYPE): Delete.
16952         (MASK_FLOAT128_KEYWORD): Define new shortcut macro.
16953         (RS6000BTM_FLOAT128): Define in terms of -mfloat128, not
16954         -mfloat128-type.
16955         * doc/invoke.texi (RS/6000 and PowerPC Options): Update
16956         documentation for -mfloat128.
16958 2017-09-06  Olivier Hainque  <hainque@adacore.com>
16960         * config.gcc (powerpc-wrs-vxworksspe): Now match as vxworks*spe.
16962 2017-09-06  Wish Wu  <wishwu007@gmail.com>
16963             Jakub Jelinek  <jakub@redhat.com>
16965         * asan.c (initialize_sanitizer_builtins): Add
16966         BT_FN_VOID_UINT8_UINT8, BT_FN_VOID_UINT16_UINT16,
16967         BT_FN_VOID_UINT32_UINT32, BT_FN_VOID_UINT64_UINT64,
16968         BT_FN_VOID_FLOAT_FLOAT, BT_FN_VOID_DOUBLE_DOUBLE and
16969         BT_FN_VOID_UINT64_PTR variables.
16970         * builtin-types.def (BT_FN_VOID_UINT8_UINT8): New fn type.
16971         (BT_FN_VOID_UINT16_UINT16): Likewise.
16972         (BT_FN_VOID_UINT32_UINT32): Likewise.
16973         (BT_FN_VOID_FLOAT_FLOAT): Likewise.
16974         (BT_FN_VOID_DOUBLE_DOUBLE): Likewise.
16975         (BT_FN_VOID_UINT64_PTR): Likewise.
16976         * common.opt (flag_sanitize_coverage): New variable.
16977         (fsanitize-coverage=trace-pc): Remove.
16978         (fsanitize-coverage=): Add.
16979         * flag-types.h (enum sanitize_coverage_code): New enum.
16980         * fold-const.c (fold_range_test): Disable non-short-circuit
16981         optimization if flag_sanitize_coverage.
16982         (fold_truth_andor): Likewise.
16983         * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
16984         * opts.c (COVERAGE_SANITIZER_OPT): Define.
16985         (coverage_sanitizer_opts): New array.
16986         (get_closest_sanitizer_option): Add OPTS argument, handle also
16987         OPT_fsanitize_coverage_.
16988         (parse_sanitizer_options): Adjusted to also handle
16989         OPT_fsanitize_coverage_.
16990         (common_handle_option): Add OPT_fsanitize_coverage_.
16991         * sancov.c (instrument_comparison, instrument_switch): New function.
16992         (sancov_pass): Add trace-cmp support.
16993         * sanitizer.def (BUILT_IN_SANITIZER_COV_TRACE_CMP1,
16994         BUILT_IN_SANITIZER_COV_TRACE_CMP2, BUILT_IN_SANITIZER_COV_TRACE_CMP4,
16995         BUILT_IN_SANITIZER_COV_TRACE_CMP8,
16996         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP1,
16997         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP2,
16998         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP4,
16999         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP8,
17000         BUILT_IN_SANITIZER_COV_TRACE_CMPF, BUILT_IN_SANITIZER_COV_TRACE_CMPD,
17001         BUILT_IN_SANITIZER_COV_TRACE_SWITCH): New builtins.
17002         * doc/invoke.texi: Document -fsanitize-coverage=trace-cmp.
17004 2017-09-06  Richard Earnshaw  <rearnsha@arm.com>
17006         * config/arm/parsecpu.awk (fatal): Note that we've encountered an
17007         error.  Only quit immediately if parsing is complete.
17008         (BEGIN): Initialize fatal_err and parse_done.
17009         (begin fpu, end fpu): Check number of arguments.
17010         (begin arch, end arch): Likewise.
17011         (begin cpu, end cpu): Likewise.
17012         (cname, tune for, tune flags, architecture, fpu, option): Likewise.
17013         (optalias): Likewise.
17015 2017-09-06  Richard Earnshaw  <rearnsha@arm.com>
17017         * config.gcc (arm*-*-*): Don't add arm-isa.h to tm_p_file.
17018         * config/arm/arm-isa.h: Delete.  Move definitions to ...
17019         * arm-cpus.in: ... here.  Use new feature and fgroup values.
17020         * config/arm/arm.c (arm_option_override): Use lower case for feature
17021         bit names.
17022         * config/arm/arm.h (TARGET_HARD_FLOAT): Likewise.
17023         (TARGET_VFP3, TARGET_VFP5, TARGET_FMA): Likewise.
17024         * config/arm/parsecpu.awk (END): Add new command 'isa'.
17025         (isa_pfx): Delete.
17026         (print_isa_bits_for): New function.
17027         (gen_isa): New function.
17028         (gen_comm_data): Use print_isa_bits_for.
17029         (define feature): New keyword.
17030         (define fgroup): New keyword.
17031         * config/arm/t-arm (OPTIONS_H_EXTRA): Add arm-isa.h
17032         (arm-isa.h): Add rule to generate file.
17033         * common/config/arm/arm-common.c: (arm_canon_arch_option): Use lower
17034         case for feature bit names.
17036 2017-09-06  Richard Biener  <rguenther@suse.de>
17038         * tree-ssa-pre.c (NECESSARY): Remove.
17039         (create_expression_by_pieces): Do not touch pass-local flags.
17040         (insert_into_preds_of_block): Likewise.
17041         (do_pre_regular_insertion): Likewise.
17042         (eliminate_insert): Likewise.
17043         (eliminate_dom_walker::before_dom_children): Likewise.
17044         (fini_eliminate): Do not look at inserted_exprs.
17045         (mark_operand_necessary): Remove.
17046         (remove_dead_inserted_code): Replace with simple work-list
17047         algorithm based on inserted_exprs and SSA uses.
17048         (pass_pre::execute): Re-order fini_eliminate and
17049         remove_dead_inserted_code.
17051 2017-09-06  Olivier Hainque  <hainque@adacore.com>
17053         * config/powerpcspe/vxworks.h (VXCPU_FOR_8548): Correct definition
17054         for VxWorks 7.  Adjust surrounding comments.
17056 2017-09-06  Richard Biener  <rguenther@suse.de>
17058         * gimple-ssa-strength-reduction.c
17059         (find_candidates_dom_walker::before_dom_children): Also allow
17060         pointer types.
17062 2017-09-06  Richard Biener  <rguenther@suse.de>
17064         PR tree-optimization/82108
17065         * tree-vect-stmts.c (vectorizable_load): Fix pointer adjustment
17066         for gap in the non-permutation SLP case.
17068 2017-09-06  Martin Jambor  <mjambor@suse.cz>
17070         PR tree-optimization/82078
17071         * tree-sra.c (sort_and_splice_var_accesses): Move call to
17072         add_access_to_work_queue...
17073         (build_accesses_from_assign): ...here.
17074         (propagate_all_subaccesses): Make sure racc is the group
17075         representative, if there is one.
17077 2017-09-06  Jakub Jelinek  <jakub@redhat.com>
17079         PR middle-end/82095
17080         * varasm.c (categorize_decl_for_section): Use SECCAT_TBSS for TLS vars with
17081         NULL DECL_INITIAL.
17083 2017-09-06  Richard Biener  <rguenther@suse.de>
17085         * gimple-ssa-strength-reduction.c
17086         (find_candidates_dom_walker::before_doom_children): Use a
17087         type and not a mode check.
17089 2017-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
17091         PR target/77308
17092         * config/arm/predicates.md (arm_general_adddi_operand): Create new
17093         non-vfp predicate.
17094         * config/arm/arm.md (*arm_adddi3, *arm_subdi3): Use new predicates.
17096 2017-09-05  Jeff Law  <law@redhat.com>
17098         PR tree-optimization/64910
17099         * tree-ssa-reassoc.c (reassociate_bb): Restrict last change to
17100         cases where we have 3 or more operands.
17102 2017-09-05  Jakub Jelinek  <jakub@redhat.com>
17104         PR middle-end/81768
17105         * omp-low.c (lower_omp_for): Recompute tree invariant if
17106         gimple_omp_for_initial/final is ADDR_EXPR.
17108         PR middle-end/81768
17109         * omp-expand.c (expand_omp_simd): Force second operands of COND_EXPR
17110         into gimple val before gimplification fo the COND_EXPR.
17112 2017-09-05  Aldy Hernandez  <aldyh@redhat.com>
17114         * tree-ssa-threadupdate.c (duplicate_thread_path): Remove unused
17115         REGION_COPY argument.
17116         (thread_through_all_blocks): Remove unused argument to
17117         duplicate_thread_path.
17119 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17120             Alan Hayward  <alan.hayward@arm.com>
17121             David Sherwood  <david.sherwood@arm.com>
17123         * config/aarch64/aarch64-protos.h (aarch64_gen_adjusted_ldpstp):
17124         Take a scalar_mode rather than a machine_mode.
17125         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
17126         * config/aarch64/aarch64.c (aarch64_simd_container_mode): Likewise.
17127         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
17128         (aarch64_gen_adjusted_ldpstp): Likewise.
17129         (aarch64_expand_vector_init): Use scalar_mode instead of machine_mode.
17131 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17132             Alan Hayward  <alan.hayward@arm.com>
17133             David Sherwood  <david.sherwood@arm.com>
17135         * config/aarch64/aarch64-protos.h (aarch64_is_extend_from_extract):
17136         Take a scalar_int_mode instead of a machine_mode.
17137         (aarch64_mask_and_shift_for_ubfiz_p): Likewise.
17138         (aarch64_output_scalar_simd_mov_immediate): Likewise.
17139         (aarch64_simd_scalar_immediate_valid_for_move): Likewise.
17140         (aarch64_simd_attr_length_rglist): Delete.
17141         * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Take
17142         a scalar_int_mode instead of a machine_mode.
17143         (aarch64_add_offset): Likewise.
17144         (aarch64_internal_mov_immediate): Likewise
17145         (aarch64_add_constant_internal): Likewise.
17146         (aarch64_add_constant): Likewise.
17147         (aarch64_movw_imm): Likewise.
17148         (aarch64_rtx_arith_op_extract_p): Likewise.
17149         (aarch64_mask_and_shift_for_ubfiz_p): Likewise.
17150         (aarch64_simd_scalar_immediate_valid_for_move): Likewise.
17151         Remove assert that the mode isn't a vector.
17152         (aarch64_output_scalar_simd_mov_immediate): Likewise.
17153         (aarch64_expand_mov_immediate): Update calls after above changes.
17154         (aarch64_output_casesi): Use as_a <scalar_int_mode>.
17155         (aarch64_and_bitmask_imm): Check for scalar integer modes.
17156         (aarch64_move_imm): Likewise.
17157         (aarch64_can_const_movi_rtx_p): Likewise.
17158         (aarch64_strip_extend): Likewise.
17159         (aarch64_extr_rtx_p): Likewise.
17160         (aarch64_rtx_costs): Likewise, using wode_mode as the mode of
17161         a CONST_INT when the mode parameter is VOIDmode.
17162         (aarch64_float_const_rtx_p): Use scalar_int_mode for a temporary.
17164 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17166         * machmode.h (bitwise_mode_for_mode): Return opt_mode.
17167         * stor-layout.c (bitwise_mode_for_mode): Likewise.
17168         (bitwise_type_for_mode): Update accordingly.
17170 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17172         * stor-layout.h (mode_for_size_tree): Return an opt_mode.
17173         * stor-layout.c (mode_for_size_tree): Likewise.
17174         (mode_for_array): Update accordingly.
17175         (layout_decl): Likewise.
17176         (compute_record_mode): Likewise.  Only set the mode once.
17178 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17180         * target.def (get_mask_mode): Change return type to opt_mode.
17181         Expand commentary.
17182         * doc/tm.texi: Regenerate.
17183         * targhooks.h (default_get_mask_mode): Return an opt_mode.
17184         * targhooks.c (default_get_mask_mode): Likewise.
17185         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
17186         * optabs-query.c (can_vec_mask_load_store_p): Update use of
17187         targetm.get_mask_mode.
17188         * tree.c (build_truth_vector_type): Likewise.
17190 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17192         * machmode.h (mode_for_vector): Return an opt_mode.
17193         * stor-layout.c (mode_for_vector): Likewise.
17194         (mode_for_int_vector): Update accordingly.
17195         (layout_type): Likewise.
17196         * config/i386/i386.c (emit_memmov): Likewise.
17197         (ix86_expand_set_or_movmem): Likewise.
17198         (ix86_expand_vector_init): Likewise.
17199         (ix86_get_mask_mode): Likewise.
17200         * config/powerpcspe/powerpcspe.c (rs6000_expand_vec_perm_const_1):
17201         Likewise.
17202         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Likewise.
17203         * expmed.c (extract_bit_field_1): Likewise.
17204         * expr.c (expand_expr_real_2): Likewise.
17205         * optabs-query.c (can_vec_perm_p): Likewise.
17206         (can_vec_mask_load_store_p): Likewise.
17207         * optabs.c (expand_vec_perm): Likewise.
17208         * targhooks.c (default_get_mask_mode): Likewise.
17209         * tree-vect-stmts.c (vectorizable_store): Likewise.
17210         (vectorizable_load): Likewise.
17211         (get_vectype_for_scalar_type_and_size): Likewise.
17213 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17215         * machmode.h (mode_for_int_vector): New function.
17216         * stor-layout.c (mode_for_int_vector): Likewise.
17217         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Use it.
17218         * config/powerpcspe/powerpcspe.c (rs6000_do_expand_vec_perm): Likewise.
17219         * config/rs6000/rs6000.c (rs6000_do_expand_vec_perm): Likewise.
17220         * config/s390/s390.c (s390_expand_vec_compare_cc): Likewise.
17221         (s390_expand_vcond): Likewise.
17223 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17225         * machmode.h (opt_machine_mode): New type.
17226         (opt_mode<T>): Allow construction from anything that can be
17227         converted to a T.
17228         (is_a, as_a, dyn_cast): Add overloads for opt_mode.
17229         (mode_for_size): Return an opt_machine_mode.
17230         * stor-layout.c (mode_for_size): Likewise.
17231         (mode_for_size_tree): Update call accordingly.
17232         (bitwise_mode_for_mode): Likewise.
17233         (make_fract_type): Likewise.
17234         (make_accum_type): Likewise.
17235         * caller-save.c (replace_reg_with_saved_mem): Update call
17236         accordingly.
17237         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17238         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17239         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17240         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17241         * expmed.c (extract_bit_field_1): Likewise.
17242         * reload.c (get_secondary_mem): Likewise.
17243         * varasm.c (assemble_integer): Likewise.
17244         * lower-subreg.c (simplify_subreg_concatn): Likewise.  Move
17245         early-out.
17247 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17249         * machmode.h (decimal_float_mode_for_size): New function.
17250         * real.h (REAL_VALUE_TO_TARGET_LONG_DOUBLE): Use float_mode_for_size.
17251         (REAL_VALUE_TO_TARGET_DOUBLE): Likewise.
17252         (REAL_VALUE_TO_TARGET_SINGLE): Likewise.
17253         (REAL_VALUE_TO_TARGET_DECIMAL128): Use decimal_float_mode_for_size.
17254         (REAL_VALUE_TO_TARGET_DECIMAL64): Likewise.
17255         (REAL_VALUE_TO_TARGET_DECIMAL32): Likewise.
17257 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17259         * builtins.c (expand_builtin_powi): Use int_mode_for_size.
17260         (get_builtin_sync_mode): Likewise.
17261         (expand_ifn_atomic_compare_exchange): Likewise.
17262         (expand_builtin_atomic_clear): Likewise.
17263         (expand_builtin_atomic_test_and_set): Likewise.
17264         (fold_builtin_atomic_always_lock_free): Likewise.
17265         * calls.c (compute_argument_addresses): Likewise.
17266         (emit_library_call_value_1): Likewise.
17267         (store_one_arg): Likewise.
17268         * combine.c (combine_instructions): Likewise.
17269         * config/aarch64/aarch64.c (aarch64_function_value): Likewise.
17270         * config/arm/arm.c (arm_function_value): Likewise.
17271         (aapcs_allocate_return_reg): Likewise.
17272         * config/c6x/c6x.c (c6x_expand_movmem): Likewise.
17273         * config/i386/i386.c (construct_container): Likewise.
17274         (ix86_gimplify_va_arg): Likewise.
17275         (ix86_expand_sse_cmp): Likewise.
17276         (emit_memmov): Likewise.
17277         (emit_memset): Likewise.
17278         (expand_small_movmem_or_setmem): Likewise.
17279         (ix86_expand_pextr): Likewise.
17280         (ix86_expand_pinsr): Likewise.
17281         * config/lm32/lm32.c (lm32_block_move_inline): Likewise.
17282         * config/microblaze/microblaze.c (microblaze_block_move_straight):
17283         Likewise.
17284         * config/mips/mips.c (mips_function_value_1) Likewise.
17285         (mips_block_move_straight): Likewise.
17286         (mips_expand_ins_as_unaligned_store): Likewise.
17287         * config/powerpcspe/powerpcspe.c
17288         (rs6000_darwin64_record_arg_advance_flush): Likewise.
17289         (rs6000_darwin64_record_arg_flush): Likewise.
17290         * config/rs6000/rs6000.c
17291         (rs6000_darwin64_record_arg_advance_flush): Likewise.
17292         (rs6000_darwin64_record_arg_flush): Likewise.
17293         * config/sparc/sparc.c (sparc_function_arg_1): Likewise.
17294         (sparc_function_value_1): Likewise.
17295         * config/spu/spu.c (adjust_operand): Likewise.
17296         (spu_emit_branch_or_set): Likewise.
17297         (arith_immediate_p): Likewise.
17298         * emit-rtl.c (gen_lowpart_common): Likewise.
17299         * expr.c (expand_expr_real_1): Likewise.
17300         * function.c (assign_parm_setup_block): Likewise.
17301         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Likewise.
17302         * reload1.c (alter_reg): Likewise.
17303         * stor-layout.c (mode_for_vector): Likewise.
17304         (layout_type): Likewise.
17306 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17308         * config/spu/spu.c (exp2_immediate_p): Use int_mode_for_mode.
17309         (spu_convert_move): Likewise.
17310         * lower-subreg.c (resolve_simple_move): Likewise.
17312 2017-09-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
17314         PR target/81833
17315         * config/rs6000/altivec.md (altivec_vsum2sws): Convert from a
17316         define_insn to a define_expand.
17317         (altivec_vsum2sws_direct): New define_insn.
17318         (altivec_vsumsws): Convert from a define_insn to a define_expand.
17320 2017-09-05  Wilco Dijkstra  <wdijkstr@arm.com>
17322         * config/arm/arm.c (arm_option_params_internal): Improve setting of
17323         max_insns_skipped.
17325 2017-09-05  H.J. Lu  <hongjiu.lu@intel.com>
17327         PR target/59501
17328         PR target/81624
17329         PR target/81769
17330         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Don't
17331         realign stack if stack alignment needed is less than incoming
17332         stack boundary.
17334 2017-09-05  Marek Polacek  <polacek@redhat.com>
17336         PR sanitizer/82072
17337         * convert.c (convert_to_integer_1) <case NEGATE_EXPR>: Move the ubsan
17338         check earlier.
17340 2017-09-05  Wilco Dijkstra  <wdijkstr@arm.com>
17342         * explow.c (get_dynamic_stack_size): Improve dynamic alignment.
17344 2017-09-05  Richard Biener  <rguenther@suse.de>
17346         PR tree-optimization/82084
17347         * fold-const.c (can_native_encode_string_p): Handle wide characters.
17349 2017-09-05  Richard Biener  <rguenther@suse.de>
17351         PR tree-optimization/82102
17352         * tree-ssa-pre.c (fini_eliminate): Check if lhs is NULL.
17354 2017-09-05  Martin Liska  <mliska@suse.cz>
17356         PR tree-optimization/82032
17357         * tree-cfg.c (generate_range_test): New function.
17358         * tree-cfg.h (generate_range_test): Declared here.
17359         * tree-cfgcleanup.c (convert_single_case_switch): New function.
17360         (cleanup_control_expr_graph): Use it.
17361         * tree-switch-conversion.c (try_switch_expansion): Remove
17362         assert.
17363         (emit_case_nodes): Use generate_range_test.
17365 2017-09-04  Uros Bizjak  <ubizjak@gmail.com>
17367         PR target/82098
17368         * config/i386/i386.md (*<btsc><mode>_mask): Add
17369         TARGET_USE_BT to insn constraint.
17370         (*btr<mode>_mask): Ditto.
17372 2017-09-04  Wilco Dijkstra  <wdijkstr@arm.com>
17374         * config/arm/arm.c (arm_legitimate_index_p): Add comment.
17375         (thumb2_legitimate_index_p): Use correct range for DI/DF mode.
17377 2017-09-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
17379         PR target/77308
17380         * config/arm/arm.md (*arm_adddi3, *arm_subdi3): Split early except for
17381         TARGET_NEON and TARGET_IWMMXT.
17382         (anddi3, iordi3, xordi3, one_cmpldi2): Split while expanding except for
17383         TARGET_NEON and TARGET_IWMMXT.
17384         (*one_cmpldi2_insn): Moved the body of one_cmpldi2 here.
17386 2017-09-04  Uros Bizjak  <ubizjak@gmail.com>
17388         * config/i386/i386-protos.h (ix86_tls_address_pattern_p) New prototype.
17389         (ix86_rewrite_tls_address): Ditto.
17390         * config/i386/i386.c (ix86_tls_address_pattern_p) New function.
17391         (ix86_rewrite_tls_address_1): Ditto.
17392         (ix86_rewrite_tls_address): Ditto.
17393         * config/i386/predicates.md (tls_address_pattern): New predicate.
17394         * config/i386/i386.md (TLS address splitter): New splitter.
17396 2017-09-04  Richard Biener  <rguenther@suse.de>
17398         PR tree-optimization/82084
17399         * fold-const.h (can_native_encode_string_p): Declare.
17400         * fold-const.c (can_native_encode_string_p): Factor out from ...
17401         (native_encode_string): ... here.
17402         * tree-vect-stmts.c (vectorizable_store): Call it to avoid
17403         vectorizing stores from constants we later cannot handle.
17405 2017-09-04  Marek Polacek  <polacek@redhat.com>
17407         PR c/81783
17408         * doc/invoke.texi: Update -Wtautological-compare documentation.
17410 2017-09-04  Jeff Law  <law@redhat.com>
17412         PR tree-optimization/64910
17413         * tree-ssa-reassoc.c (reassociate_bb): For bitwise binary ops,
17414         swap the first and last operand if the last is a constant.
17416 2017-09-04  Marek Polacek  <polacek@redhat.com>
17418         PR sanitizer/82072
17419         * convert.c (do_narrow): When sanitizing signed integer overflows,
17420         bail out for signed types.
17421         (convert_to_integer_1) <case NEGATE_EXPR>: Likewise.
17423 2017-09-04  Richard Biener  <rguenther@suse.de>
17425         PR tree-optimization/82060
17426         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
17427         Move devirtualization after stmt folding and before EH/AB/noreturn
17428         cleanup to get the stmt refs canonicalized.  Use a bool instead
17429         of gimple_modified_p since that doesn't work for NOPs.  Schedule
17430         NOPs generated by folding for removal.
17432 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17433             Alan Hayward  <alan.hayward@arm.com>
17434             David Sherwood  <david.sherwood@arm.com>
17436         * coretypes.h (pad_direction): New enum.
17437         * defaults.h (DEFAULT_FUNCTION_ARG_PADDING): Delete.
17438         (FUNCTION_ARG_PADDING): Likewise.
17439         * target.def (function_arg_padding): New hook.
17440         * targhooks.h (default_function_arg_padding): Declare.
17441         * targhooks.c (default_function_arg_padding): New function.
17442         * doc/tm.texi.in (FUNCTION_ARG_PADDING): Replace with...
17443         (TARGET_FUNCTION_ARG_PADDING): ...this.
17444         * doc/tm.texi: Regenerate.
17445         * calls.c (store_unaligned_arguments_into_pseudos): Use pad_direction
17446         instead of direction.
17447         (compute_argument_addresses): Likewise.
17448         (load_register_parameters): Likewise.
17449         (emit_library_call_value_1): Likewise.
17450         (store_one_arg): Use targetm.calls.function_arg_padding instead
17451         of FUNCTION_ARG_PADDING.
17452         (must_pass_in_stack_var_size_or_pad): Likewise.
17453         * expr.c (emit_group_load_1): Use pad_direction instead of direction.
17454         (emit_group_store): Likewise.
17455         (emit_single_push_insn_1): Use targetm.calls.function_arg_padding
17456         instead of FUNCTION_ARG_PADDING.
17457         (emit_push_insn): Likewise, and propagate enum change throughout
17458         function.
17459         * function.h (direction): Delete.
17460         (locate_and_pad_arg_data::where_pad): Use pad_direction instead
17461         of direction.
17462         * function.c (assign_parm_find_stack_rtl): Likewise.
17463         (assign_parm_setup_block_p): Likewise.
17464         (assign_parm_setup_block): Likewise.
17465         (gimplify_parameters): Likewise.
17466         (locate_and_pad_parm): Use targetm.calls.function_arg_padding
17467         instead of FUNCTION_ARG_PADDING, and propagate enum change throughout
17468         function.
17469         * config/aarch64/aarch64.h (FUNCTION_ARG_PADDING): Delete.
17470         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17471         * config/aarch64/aarch64-protos.h (aarch64_pad_arg_upward): Delete.
17472         * config/aarch64/aarch64.c (aarch64_pad_arg_upward): Replace with...
17473         (aarch64_function_arg_padding): ...this new function.
17474         (aarch64_gimplify_va_arg_expr): Use pad_direction instead of direction.
17475         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17476         * config/arm/arm.h (FUNCTION_ARG_PADDING): Delete.
17477         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17478         * config/arm/arm-protos.h (arm_pad_arg_upward): Delete.
17479         * config/arm/arm.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17480         (arm_pad_arg_upward): Replace with...
17481         (arm_function_arg_padding): ...this new function.
17482         * config/c6x/c6x.h (BLOCK_REG_PADDING): Use pad_direction instead
17483         of direction.
17484         * config/ia64/hpux.h (FUNCTION_ARG_PADDING): Delete.
17485         * config/ia64/ia64-protos.h (ia64_hpux_function_arg_padding): Delete.
17486         * config/ia64/ia64.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17487         (ia64_hpux_function_arg_padding): Replace with...
17488         (ia64_function_arg_padding): ...this new function.  Use pad_direction
17489         instead of direction.  Check for TARGET_HPUX.
17490         * config/iq2000/iq2000.h (FUNCTION_ARG_PADDING): Delete.
17491         * config/iq2000/iq2000.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17492         (iq2000_function_arg_padding): New function.
17493         * config/mips/mips-protos.h (mips_pad_arg_upward): Delete.
17494         * config/mips/mips.c (mips_pad_arg_upward): Replace with...
17495         (mips_function_arg_padding): ...this new function.
17496         (mips_pad_reg_upward): Update accordingly.
17497         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17498         * config/mips/mips.h (PAD_VARARGS_DOWN): Use
17499         targetm.calls.function_arg_padding.
17500         (FUNCTION_ARG_PADDING): Delete.
17501         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17502         * config/nios2/nios2.h (FUNCTION_ARG_PADDING): Delete.
17503         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17504         * config/nios2/nios2-protos.h (nios2_function_arg_padding): Delete.
17505         (nios2_block_reg_padding): Return pad_direction instead of direction.
17506         * config/nios2/nios2.c (nios2_block_reg_padding): Return pad_direction
17507         instead of direction.
17508         (nios2_function_arg_padding): Likewise.  Make static.
17509         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17510         * config/pa/pa.h (FUNCTION_ARG_PADDING): Delete.
17511         (BLOCK_REG_PADDING): Use targetm.calls.function_arg_padding.
17512         * config/pa/pa-protos.h (pa_function_arg_padding): Delete.
17513         * config/pa/pa.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17514         (pa_function_arg_padding): Make static.  Return pad_direction instead
17515         of direction.
17516         * config/powerpcspe/powerpcspe.h (FUNCTION_ARG_PADDING): Delete.
17517         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17518         * config/powerpcspe/aix.h (BLOCK_REG_PADDING): Use pad_direction
17519         instead of direction.  Use targetm.calls.function_arg_padding.
17520         * config/powerpcspe/darwin.h (BLOCK_REG_PADDING): Likewise.
17521         * config/powerpcspe/freebsd64.h (BLOCK_REG_PADDING): Likewise.
17522         * config/powerpcspe/linux64.h (BLOCK_REG_PADDING): Likewise.
17523         * config/powerpcspe/powerpcspe-protos.h (function_arg_padding): Delete.
17524         * config/powerpcspe/powerpcspe.c (TARGET_FUNCTION_ARG_PADDING):
17525         Redefine.
17526         (function_arg_padding): Rename to...
17527         (rs6000_function_arg_padding): ...this.  Make static.  Return
17528         pad_direction instead of direction.
17529         (rs6000_return_in_msb): Use rs6000_function_arg_padding.
17530         * config/rs6000/rs6000.h (FUNCTION_ARG_PADDING): Delete.
17531         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17532         * config/rs6000/aix.h (BLOCK_REG_PADDING): Use pad_direction
17533         instead of direction.  Use targetm.calls.function_arg_padding.
17534         * config/rs6000/darwin.h (BLOCK_REG_PADDING): Likewise.
17535         * config/rs6000/freebsd64.h (BLOCK_REG_PADDING): Likewise.
17536         * config/rs6000/linux64.h (BLOCK_REG_PADDING): Likewise.
17537         * config/rs6000/rs6000-protos.h (function_arg_padding): Delete.
17538         * config/rs6000/rs6000.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17539         (function_arg_padding): Rename to...
17540         (rs6000_function_arg_padding): ...this.  Make static.  Return
17541         pad_direction instead of direction.
17542         (rs6000_return_in_msb): Use rs6000_function_arg_padding.
17543         * config/s390/s390.h (FUNCTION_ARG_PADDING): Delete.
17544         * config/s390/s390.c (s390_function_arg_padding): New function.
17545         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17546         * config/sparc/sparc.h (FUNCTION_ARG_PADDING): Delete.
17547         * config/sparc/sparc-protos.h (function_arg_padding): Delete.
17548         * config/sparc/sparc.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17549         (function_arg_padding): Rename to...
17550         (sparc_function_arg_padding): ...this.  Make static.  Return
17551         pad_direction instead of direction.
17552         * config/spu/spu.h (FUNCTION_ARG_PADDING): Delete.
17553         * config/spu/spu.c (spu_function_arg_padding): New function.
17554         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17555         * system.h (FUNCTION_ARG_PADDING): Poison.
17557 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17558             Alan Hayward  <alan.hayward@arm.com>
17559             David Sherwood  <david.sherwood@arm.com>
17561         * target.def (modes_tieable_p): New hook.
17562         * doc/tm.texi (MODES_TIEABLE_P): Replace with...
17563         (TARGET_MODES_TIEABLE_P): ...this.
17564         * doc/tm.texi.in: Regenerate.
17565         * hooks.h (hook_bool_mode_mode_true): Declare.
17566         * hooks.c (hook_bool_mode_mode_true): New function.
17567         * combine.c (subst): Use targetm.modes_tieable_p instead of
17568         MODES_TIEABLE_P.
17569         * dse.c (find_shift_sequence): Likewise.
17570         * expmed.c (extract_low_bits): Likewise.
17571         * lower-subreg.c: Include target.h.
17572         (find_decomposable_subregs): Use targetm.modes_tieable_p instead of
17573         MODES_TIEABLE_P.
17574         * rtlanal.c (rtx_cost): Likewise.
17575         * config/aarch64/aarch64.h (MODES_TIEABLE_P): Delete.
17576         * config/aarch64/aarch64-protos.h (aarch64_modes_tieable_p): Delete.
17577         * config/aarch64/aarch64.c (aarch64_modes_tieable_p): Make static.
17578         (TARGET_MODES_TIEABLE_P): Redefine.
17579         * config/alpha/alpha.h (MODES_TIEABLE_P): Delete.
17580         * config/alpha/alpha.c (alpha_modes_tieable_p): New function.
17581         (TARGET_MODES_TIEABLE_P): Redefine.
17582         * config/arc/arc.h (MODES_TIEABLE_P): Delete.
17583         * config/arc/arc.c (TARGET_MODES_TIEABLE_P): Redefine.
17584         (arc_modes_tieable_p): New function.
17585         * config/arm/arm.h (MODES_TIEABLE_P): Delete.
17586         * config/arm/arm-protos.h (arm_modes_tieable_p): Delete.
17587         * config/arm/arm.c (TARGET_MODES_TIEABLE_P): Redefine.
17588         (arm_modes_tieable_p): Make static.
17589         * config/avr/avr.h (MODES_TIEABLE_P): Delete.
17590         * config/bfin/bfin.h (MODES_TIEABLE_P): Delete.
17591         * config/bfin/bfin.c (bfin_modes_tieable_p): New function.
17592         (TARGET_MODES_TIEABLE_P): Redefine.
17593         * config/c6x/c6x.h (MODES_TIEABLE_P): Delete.
17594         * config/c6x/c6x.c (c6x_modes_tieable_p): New function.
17595         (TARGET_MODES_TIEABLE_P): Redefine.
17596         * config/cr16/cr16.h (MODES_TIEABLE_P): Delete.
17597         * config/cr16/cr16.c (TARGET_MODES_TIEABLE_P): Redefine.
17598         (cr16_modes_tieable_p): New function.
17599         * config/cris/cris.h (MODES_TIEABLE_P): Delete.
17600         * config/epiphany/epiphany.h (MODES_TIEABLE_P): Delete.
17601         * config/fr30/fr30.h (MODES_TIEABLE_P): Delete.
17602         (TRULY_NOOP_TRUNCATION): Update comment.
17603         * config/frv/frv.h (MODES_TIEABLE_P): Delete.
17604         (TRULY_NOOP_TRUNCATION): Update comment.
17605         * config/frv/frv.c (TARGET_MODES_TIEABLE_P): Redefine.
17606         (frv_modes_tieable_p): New function.
17607         * config/ft32/ft32.h (MODES_TIEABLE_P): Delete.
17608         * config/h8300/h8300.h (MODES_TIEABLE_P): Delete.
17609         * config/h8300/h8300.c (h8300_modes_tieable_p): New function.
17610         (TARGET_MODES_TIEABLE_P): Redefine.
17611         * config/i386/i386.h (MODES_TIEABLE_P): Delete.
17612         * config/i386/i386-protos.h (ix86_modes_tieable_p): Delete.
17613         * config/i386/i386.c (ix86_modes_tieable_p): Make static.
17614         (TARGET_MODES_TIEABLE_P): Redefine.
17615         * config/ia64/ia64.h (MODES_TIEABLE_P): Delete.
17616         * config/ia64/ia64.c (TARGET_MODES_TIEABLE_P): Redefine.
17617         (ia64_modes_tieable_p): New function.
17618         * config/iq2000/iq2000.h (MODES_TIEABLE_P): Delete.
17619         * config/iq2000/iq2000.c (TARGET_MODES_TIEABLE_P): Redefine.
17620         (iq2000_modes_tieable_p): New function.
17621         * config/lm32/lm32.h (MODES_TIEABLE_P): Delete.
17622         * config/lm32/lm32.c (TARGET_MODES_TIEABLE_P): Redefine.
17623         (lm32_modes_tieable_p): New function.
17624         * config/m32c/m32c.h (MODES_TIEABLE_P): Delete.
17625         * config/m32c/m32c-protos.h (m32c_modes_tieable_p): Delete.
17626         * config/m32c/m32c.c (m32c_modes_tieable_p): Make static.
17627         (TARGET_MODES_TIEABLE_P): Redefine.
17628         * config/m32r/m32r.h (MODES_TIEABLE_P): Delete.
17629         * config/m32r/m32r.c (TARGET_MODES_TIEABLE_P): Redefine.
17630         (m32r_modes_tieable_p): New function.
17631         * config/m68k/m68k.h (MODES_TIEABLE_P): Delete.
17632         * config/m68k/m68k.c (TARGET_MODES_TIEABLE_P): Redefine.
17633         (m68k_modes_tieable_p): New function.
17634         * config/mcore/mcore.h (MODES_TIEABLE_P): Delete.
17635         * config/mcore/mcore.c (TARGET_MODES_TIEABLE_P): Redefine.
17636         (mcore_modes_tieable_p): New function.
17637         * config/microblaze/microblaze.h (MODES_TIEABLE_P): Delete.
17638         * config/microblaze/microblaze.c (microblaze_modes_tieable_p): New
17639         function.
17640         (TARGET_MODES_TIEABLE_P): Redefine.
17641         * config/mips/mips.h (MODES_TIEABLE_P): Delete.
17642         * config/mips/mips-protos.h (mips_modes_tieable_p): Delete.
17643         * config/mips/mips.c (mips_modes_tieable_p): Make static.
17644         (TARGET_MODES_TIEABLE_P): Redefine.
17645         * config/mmix/mmix.h (MODES_TIEABLE_P): Delete.
17646         * config/mn10300/mn10300.h (MODES_TIEABLE_P): Delete.
17647         * config/mn10300/mn10300-protos.h (mn10300_modes_tieable): Delete.
17648         * config/mn10300/mn10300.c (mn10300_modes_tieable): Rename to...
17649         (mn10300_modes_tieable_p): ...this and make static.
17650         (TARGET_MODES_TIEABLE_P): Redefine.
17651         * config/moxie/moxie.h (MODES_TIEABLE_P): Delete.
17652         * config/msp430/msp430.h (MODES_TIEABLE_P): Delete.
17653         * config/msp430/msp430-protos.h (msp430_modes_tieable_p): Delete.
17654         * config/msp430/msp430.c (TARGET_MODES_TIEABLE_P): Redefine.
17655         (msp430_modes_tieable_p): Make static.
17656         * config/nds32/nds32.h (MODES_TIEABLE_P): Delete.
17657         * config/nds32/nds32.c (nds32_modes_tieable_p): New function.
17658         (TARGET_MODES_TIEABLE_P): Redefine.
17659         * config/nios2/nios2.h (MODES_TIEABLE_P): Delete.
17660         * config/nvptx/nvptx.h (MODES_TIEABLE_P): Delete.
17661         * config/nvptx/nvptx.c (nvptx_modes_tieable_p): New function.
17662         (TARGET_MODES_TIEABLE_P): Redefine.
17663         * config/pa/pa.h (MODES_TIEABLE_P): Delete.
17664         * config/pa/pa-protos.h (pa_modes_tieable_p): Delete.
17665         * config/pa/pa.c (pa_modes_tieable_p): Make static.
17666         (TARGET_MODES_TIEABLE_P): Redefine.
17667         * config/pdp11/pdp11.h (MODES_TIEABLE_P): Delete.
17668         * config/pdp11/pdp11.c (TARGET_MODES_TIEABLE_P): Redefine.
17669         (pdp11_modes_tieable_p): New function.
17670         * config/powerpcspe/powerpcspe.h (MODES_TIEABLE_P): Delete.
17671         * config/powerpcspe/powerpcspe.c (TARGET_MODES_TIEABLE_P): Redefine.
17672         (rs6000_modes_tieable_p): New function.
17673         (rs6000_debug_reg_global): Use it instead of MODES_TIEABLE_P.
17674         * config/powerpcspe/powerpcspe.md: Update comment.
17675         * config/riscv/riscv.h (MODES_TIEABLE_P): Delete.
17676         * config/riscv/riscv.c (riscv_modes_tieable_p): New function.
17677         (TARGET_MODES_TIEABLE_P): Redefine.
17678         * config/rl78/rl78.h (MODES_TIEABLE_P): Delete.
17679         * config/rl78/rl78.c (TARGET_MODES_TIEABLE_P): Redefine.
17680         (rl78_modes_tieable_p): New function.
17681         * config/rs6000/rs6000.h (MODES_TIEABLE_P): Delete.
17682         * config/rs6000/rs6000.c (TARGET_MODES_TIEABLE_P): Redefine.
17683         (rs6000_modes_tieable_p): New function.
17684         (rs6000_debug_reg_global): Use it instead of MODES_TIEABLE_P.
17685         * config/rs6000/rs6000.md: Update comment.
17686         * config/rx/rx.h (MODES_TIEABLE_P): Delete.
17687         * config/rx/rx.c (rx_modes_tieable_p): New function.
17688         (TARGET_MODES_TIEABLE_P): Redefine.
17689         * config/s390/s390.h (MODES_TIEABLE_P): Delete.
17690         * config/s390/s390.c (s390_modes_tieable_p): New function.
17691         (TARGET_MODES_TIEABLE_P): Redefine.
17692         * config/sh/sh.h (MODES_TIEABLE_P): Delete.
17693         * config/sh/sh.c (TARGET_MODES_TIEABLE_P): Redefine.
17694         (sh_modes_tieable_p): New function.
17695         * config/sparc/sparc.h (MODES_TIEABLE_P): Delete.
17696         * config/sparc/sparc-protos.h (sparc_modes_tieable_p): Delete.
17697         * config/sparc/sparc.c (TARGET_MODES_TIEABLE_P): Redefine.
17698         (sparc_modes_tieable_p): Make static.
17699         * config/spu/spu.h (MODES_TIEABLE_P): Delete.
17700         * config/spu/spu.c (spu_modes_tieable_p): New function.
17701         (TARGET_MODES_TIEABLE_P): Redefine.
17702         * config/stormy16/stormy16.h (MODES_TIEABLE_P): Delete.
17703         * config/stormy16/stormy16.c (xstormy16_modes_tieable_p): New function.
17704         (TARGET_MODES_TIEABLE_P): Redefine.
17705         * config/tilegx/tilegx.h (MODES_TIEABLE_P): Delete.
17706         * config/tilepro/tilepro.h (MODES_TIEABLE_P): Delete.
17707         * config/v850/v850.h (MODES_TIEABLE_P): Delete.
17708         * config/v850/v850.c (v850_modes_tieable_p): New function.
17709         (TARGET_MODES_TIEABLE_P): Redefine.
17710         * config/vax/vax.h (MODES_TIEABLE_P): Delete.
17711         * config/visium/visium.h (MODES_TIEABLE_P): Delete.
17712         * config/visium/visium.c (TARGET_MODES_TIEABLE_P): Redefine.
17713         (visium_modes_tieable_p): New function.
17714         * config/xtensa/xtensa.h (MODES_TIEABLE_P): Delete.
17715         * config/xtensa/xtensa.c (TARGET_MODES_TIEABLE_P): Redefine.
17716         (xtensa_modes_tieable_p): New function.
17717         * system.h (MODES_TIEABLE_P): Poison.
17719 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17720             Alan Hayward  <alan.hayward@arm.com>
17721             David Sherwood  <david.sherwood@arm.com>
17723         * target.def (hard_regno_mode_ok): New hook.
17724         * doc/tm.texi (HARD_REGNO_MODE_OK): Replace with...
17725         (TARGET_HARD_REGNO_MODE_OK): ...this.
17726         * doc/tm.texi.in: Regenerate.
17727         * hooks.h (hook_bool_uint_mode_true): Declare.
17728         * hooks.c (hook_bool_uint_mode_true): New function.
17729         * doc/md.texi: Refer to targetm.hard_regno_mode_ok instead of
17730         HARD_REGNO_MODE_OK.
17731         * genpreds.c (write_insn_preds_c): Add an include of target.h.
17732         * alias.c (init_alias_target): Use targetm.hard_regno_mode_ok
17733         instead of HARD_REGNO_MODE_OK.
17734         * caller-save.c: Include target.h.
17735         (reg_save_code): Use targetm.hard_regno_mode_ok instead of
17736         HARD_REGNO_MODE_OK.
17737         * combine.c (can_combine_p): Likewise.
17738         (combinable_i3pat): Likewise.
17739         (can_change_dest_mode): Likewise.
17740         * expr.c (init_expr_target): Likewise.
17741         (convert_move): Likewise.
17742         (convert_modes): Likewise.
17743         * ira.c (setup_prohibited_class_mode_regs): Likewise.
17744         (setup_prohibited_mode_move_regs): Likewise.
17745         * ira.h (target_ira): Likewise.
17746         * lra-assigns.c (find_hard_regno_for_1): Likewise.
17747         * lra-constraints.c (process_alt_operands): Likewise.
17748         (split_reg): Likewise.
17749         * recog.c (peep2_find_free_register): Likewise.
17750         * ree.c (combine_reaching_defs): Likewise.
17751         * regcprop.c (maybe_mode_change): Likewise.
17752         * reginfo.c (init_reg_sets_1): Likewise.
17753         (choose_hard_reg_mode): Likewise.
17754         (simplifiable_subregs): Likewise.
17755         * regrename.c (check_new_reg_p): Likewise.
17756         * reload.c (find_valid_class): Likewise.
17757         (find_valid_class_1): Likewise.
17758         (reload_inner_reg_of_subreg): Likewise.
17759         (push_reload): Likewise.
17760         (combine_reloads): Likewise.
17761         (find_dummy_reload): Likewise.
17762         (find_reloads): Likewise.
17763         * reload1.c (find_reg): Likewise.
17764         (set_reload_reg): Likewise.
17765         (allocate_reload_reg): Likewise.
17766         (choose_reload_regs): Likewise.
17767         (reload_adjust_reg_for_temp): Likewise.
17768         * rtlanal.c (subreg_size_offset_from_lsb): Likewise.
17769         (simplify_subreg_regno): Likewise.
17770         * sel-sched.c (init_regs_for_mode): Likewise.
17771         * varasm.c (make_decl_rtl): Likewise.
17772         * config/aarch64/aarch64.h (HARD_REGNO_MODE_OK): Delete.
17773         (MODES_TIEABLE_P): Use targetm.hard_regno_mode_ok instead of
17774         HARD_REGNO_MODE_OK.
17775         * config/aarch64/aarch64-protos.h (aarch64_hard_regno_mode_ok): Delete.
17776         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Make static.
17777         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17778         * config/alpha/alpha.h (HARD_REGNO_MODE_OK): Delete.
17779         * config/alpha/alpha.c (alpha_hard_regno_mode_ok): New function.
17780         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17781         * config/arc/arc.h (arc_hard_regno_mode_ok): Delete.
17782         (arc_mode_class): Delete.
17783         (HARD_REGNO_MODE_OK): Delete.
17784         * config/arc/arc.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17785         (arc_hard_regno_mode_ok): Rename old array to...
17786         (arc_hard_regno_mode_ok_modes): ...this.
17787         (arc_conditional_register_usage): Update accordingly.
17788         (arc_mode_class): Make static.
17789         (arc_hard_regno_mode_ok): New function.
17790         * config/arm/arm.h (HARD_REGNO_MODE_OK): Delete.
17791         * config/arm/arm-protos.h (arm_hard_regno_mode_ok): Delete.
17792         * config/arm/arm.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17793         (arm_hard_regno_mode_ok): Make static.
17794         * config/arm/arm.md (movdi): Use targetm.hard_regno_mode_ok instead of
17795         HARD_REGNO_MODE_OK.
17796         * config/avr/avr-protos.h (avr_hard_regno_mode_ok): Delete.
17797         * config/avr/avr.h (HARD_REGNO_MODE_OK): Delete.
17798         * config/avr/avr.c (avr_hard_regno_mode_ok): Make static and
17799         return a bool.
17800         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17801         * config/bfin/bfin-protos.h (hard_regno_mode_ok): Delete.
17802         * config/bfin/bfin.h (HARD_REGNO_MODE_OK): Delete.
17803         * config/bfin/bfin.c (hard_regno_mode_ok): Rename to...
17804         (bfin_hard_regno_mode_ok): ...this.  Make static and return a bool.
17805         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17806         * config/bfin/predicates.md (valid_reg_operand): Use
17807         targetm.hard_regno_mode_ok instead of HARD_REGNO_MODE_OK.
17808         * config/c6x/c6x.h (HARD_REGNO_MODE_OK): Delete.
17809         * config/c6x/c6x.c (c6x_hard_regno_mode_ok): New function.
17810         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17811         * config/cr16/cr16.h (HARD_REGNO_MODE_OK): Delete.
17812         * config/cr16/cr16-protos.h (cr16_hard_regno_mode_ok): Delete.
17813         * config/cr16/cr16.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17814         (cr16_hard_regno_mode_ok): Make static and return a bool.
17815         * config/cris/cris.h (HARD_REGNO_MODE_OK): Delete.
17816         * config/cris/cris.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17817         (cris_hard_regno_mode_ok): New function.
17818         * config/epiphany/epiphany.h (epiphany_hard_regno_mode_ok): Delete.
17819         (epiphany_mode_class): Delete.
17820         (HARD_REGNO_MODE_OK): Delete.
17821         * config/epiphany/epiphany-protos.h (hard_regno_mode_ok): Delete.
17822         * config/epiphany/epiphany.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17823         (hard_regno_mode_ok): Rename to...
17824         (epiphany_hard_regno_mode_ok): ...this.  Make static and return a bool.
17825         * config/fr30/fr30.h (HARD_REGNO_MODE_OK): Delete.
17826         * config/fr30/fr30.md: Refer to targetm.hard_regno_mode_ok instead of
17827         HARD_REGNO_MODE_OK.
17828         * config/frv/frv.h (HARD_REGNO_MODE_OK): Delete.
17829         * config/frv/frv-protos.h (frv_hard_regno_mode_ok): Delete.
17830         * config/frv/frv.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17831         (frv_hard_regno_mode_ok): Make static and return a bool.
17832         * config/frv/frv.md: Refer to targetm.hard_regno_mode_ok instead of
17833         HARD_REGNO_MODE_OK.
17834         * config/ft32/ft32.h (HARD_REGNO_MODE_OK): Delete.
17835         * config/h8300/h8300.h (HARD_REGNO_MODE_OK): Delete.
17836         * config/h8300/h8300-protos.h (h8300_hard_regno_mode_ok): Delete.
17837         * config/h8300/h8300.c (h8300_hard_regno_mode_ok): Make static
17838         and return a bool.
17839         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17840         * config/i386/i386.h (HARD_REGNO_MODE_OK): Delete.
17841         * config/i386/i386-protos.h (ix86_hard_regno_mode_ok): Delete.
17842         * config/i386/i386.c (ix86_hard_regno_mode_ok): Make static and
17843         return a bool.
17844         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17845         * config/ia64/ia64.h (HARD_REGNO_MODE_OK): Delete.
17846         * config/ia64/ia64.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17847         (ia64_hard_regno_mode_ok): New function.
17848         * config/iq2000/iq2000.h (HARD_REGNO_MODE_OK): Delete.
17849         * config/iq2000/iq2000.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17850         (iq2000_hard_regno_mode_ok): New function.
17851         * config/lm32/lm32.h (HARD_REGNO_MODE_OK): Delete.
17852         * config/lm32/lm32.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17853         (lm32_hard_regno_mode_ok): New function.
17854         * config/m32c/m32c.h (HARD_REGNO_MODE_OK): Delete.
17855         * config/m32c/m32c-protos.h (m32c_hard_regno_ok): Delete.
17856         * config/m32c/m32c.c (class_can_hold_mode): Use m32c_hard_regno_mode_ok
17857         instead of HARD_REGNO_MODE_OK.
17858         (m32c_hard_regno_ok): Rename to...
17859         (m32c_hard_regno_mode_ok): ...this.  Make static and return a bool.
17860         (m32c_cannot_change_mode_class): Update accordingly.
17861         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17862         * config/m32r/m32r.h (m32r_hard_regno_mode_ok): Delete.
17863         (m32r_mode_class): Delete.
17864         (HARD_REGNO_MODE_OK): Delete.
17865         * config/m32r/m32r.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17866         (m32r_hard_regno_mode_ok): Rename to...
17867         (m32r_hard_regno_modes): ...this.
17868         (m32r_mode_class): Make static.
17869         (m32r_hard_regno_mode_ok): New function.
17870         * config/m68k/m68k.h (HARD_REGNO_MODE_OK): Delete.
17871         * config/m68k/m68k-protos.h (m68k_regno_mode_ok): Delete.
17872         * config/m68k/m68k.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17873         (m68k_hard_regno_mode_ok): Make static.
17874         * config/mcore/mcore.h (HARD_REGNO_MODE_OK): Delete.
17875         * config/mcore/mcore.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17876         (mcore_hard_regno_mode_ok): New function.
17877         * config/microblaze/microblaze.h (microblaze_hard_regno_mode_ok)
17878         (HARD_REGNO_MODE_OK): Delete.
17879         * config/microblaze/microblaze.c (microblaze_hard_regno_mode_ok):
17880         Rename to...
17881         (microblaze_hard_regno_mode_ok_p): ...this and make static.
17882         (microblaze_hard_regno_mode_ok): New function.
17883         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17884         * config/mips/mips.h (HARD_REGNO_MODE_OK): Delete.
17885         (mips_hard_regno_mode_ok): Delete.
17886         * config/mips/mips.c (mips_hard_regno_mode_ok): Rename to...
17887         (mips_hard_regno_mode_ok_p): ...this and make static.
17888         (mips_hard_regno_mode_ok_p): Rename to...
17889         (mips_hard_regno_mode_ok_uncached): ...this.
17890         (mips_hard_regno_mode_ok): New function.
17891         (mips_class_max_nregs): Use mips_hard_regno_mode_ok instead
17892         of HARD_REGNO_MODE_OK.
17893         (mips_option_override): Update after above name changes.
17894         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17895         * config/mmix/mmix.h (HARD_REGNO_MODE_OK): Delete.
17896         * config/mn10300/mn10300.h (HARD_REGNO_MODE_OK): Delete.
17897         * config/mn10300/mn10300-protos.h (mn10300_hard_regno_mode_ok): Delete.
17898         * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Make static.
17899         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17900         * config/moxie/moxie.h (HARD_REGNO_MODE_OK): Delete.
17901         * config/msp430/msp430.h (HARD_REGNO_MODE_OK): Delete.
17902         * config/msp430/msp430-protos.h (msp430_hard_regno_mode_ok): Delete.
17903         * config/msp430/msp430.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17904         (msp430_hard_regno_mode_ok): Make static and return a bool.
17905         * config/nds32/nds32.h (HARD_REGNO_MODE_OK): Delete.
17906         * config/nds32/nds32-protos.h (nds32_hard_regno_mode_ok): Delete.
17907         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Make static
17908         and return a bool.
17909         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17910         * config/nios2/nios2.h (HARD_REGNO_MODE_OK): Delete.
17911         * config/nvptx/nvptx.h (HARD_REGNO_MODE_OK): Delete.
17912         * config/pa/pa.h (MODES_TIEABLE_P): Update commentary.
17913         * config/pa/pa32-regs.h (HARD_REGNO_MODE_OK): Rename to...
17914         (PA_HARD_REGNO_MODE_OK): ...this
17915         * config/pa/pa64-regs.h (HARD_REGNO_MODE_OK): Rename to...
17916         (PA_HARD_REGNO_MODE_OK): ...this.
17917         * config/pa/pa.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17918         (pa_hard_regno_mode_ok): New function.
17919         * config/pdp11/pdp11.h (HARD_REGNO_MODE_OK): Delete.
17920         * config/pdp11/pdp11.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17921         (pdp11_hard_regno_mode_ok): New function.
17922         * config/powerpcspe/powerpcspe.h (HARD_REGNO_MODE_OK): Delete.
17923         * config/powerpcspe/powerpcspe-protos.h (rs6000_hard_regno_mode_ok_p):
17924         Delete.
17925         * config/powerpcspe/powerpcspe.c (rs6000_hard_regno_mode_ok_p):
17926         Make static.
17927         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17928         (rs6000_hard_regno_mode_ok): Rename to...
17929         (rs6000_hard_regno_mode_ok_uncached): ...this.
17930         (rs6000_init_hard_regno_mode_ok): Update accordingly.
17931         (rs6000_hard_regno_mode_ok): New function.
17932         * config/riscv/riscv.h (HARD_REGNO_MODE_OK): Delete.
17933         * config/riscv/riscv-protos.h (riscv_hard_regno_mode_ok_p): Delete.
17934         * config/riscv/riscv.c (riscv_hard_regno_mode_ok_p): Rename to...
17935         (riscv_hard_regno_mode_ok): ...this and make static.
17936         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17937         * config/rl78/rl78.h (HARD_REGNO_MODE_OK): Delete.
17938         * config/rl78/rl78-protos.h (rl78_hard_regno_mode_ok): Delete.
17939         * config/rl78/rl78.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17940         (rl78_hard_regno_mode_ok): Make static and return bool.
17941         * config/rs6000/rs6000.h (HARD_REGNO_MODE_OK): Delete.
17942         * config/rs6000/rs6000-protos.h (rs6000_hard_regno_mode_ok_p):
17943         Delete.
17944         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok_p): Make static.
17945         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17946         (rs6000_hard_regno_mode_ok): Rename to...
17947         (rs6000_hard_regno_mode_ok_uncached): ...this.
17948         (rs6000_init_hard_regno_mode_ok): Update accordingly.
17949         (rs6000_hard_regno_mode_ok): New function.
17950         * config/rx/rx.h (HARD_REGNO_MODE_OK): Delete.
17951         * config/rx/rx.c (rx_hard_regno_mode_ok): New function.
17952         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17953         * config/s390/s390.h (HARD_REGNO_MODE_OK): Delete.
17954         * config/s390/s390-protos.h (s390_hard_regno_mode_ok): Delete.
17955         * config/s390/s390.c (s390_hard_regno_mode_ok): Make static.
17956         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17957         * config/sh/sh.h (HARD_REGNO_MODE_OK): Delete.
17958         * config/sh/sh-protos.h (sh_hard_regno_mode_ok): Delete.
17959         * config/sh/sh.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17960         (sh_hard_regno_mode_ok): Make static.
17961         * config/sparc/constraints.md: Refer to targetm.hard_regno_mode_ok
17962         instead of HARD_REGNO_MODE_OK.
17963         * config/sparc/sparc.h (hard_regno_mode_classes): Delete.
17964         (sparc_mode_class): Delete.
17965         (HARD_REGNO_MODE_OK): Delete.
17966         * config/sparc/sparc.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17967         (hard_regno_mode_classes): Make static.
17968         (sparc_mode_class): Likewise.
17969         (sparc_hard_regno_mode_ok): New function.
17970         * config/spu/spu.h (HARD_REGNO_MODE_OK): Delete.
17971         * config/stormy16/stormy16.h (HARD_REGNO_MODE_OK): Delete.
17972         * config/stormy16/stormy16.c (xstormy16_hard_regno_mode_ok): New
17973         function.
17974         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17975         * config/tilegx/tilegx.h (HARD_REGNO_MODE_OK): Delete.
17976         * config/tilepro/tilepro.h (HARD_REGNO_MODE_OK): Delete.
17977         * config/v850/v850.h (HARD_REGNO_MODE_OK): Delete.
17978         * config/v850/v850.c (v850_hard_regno_mode_ok): New function.
17979         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17980         * config/vax/vax.h (HARD_REGNO_MODE_OK): Delete.
17981         * config/visium/visium.h (HARD_REGNO_MODE_OK): Delete.
17982         * config/visium/visium.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17983         (visium_hard_regno_mode_ok): New function.
17984         * config/visium/visium.md: Refer to targetm.hard_regno_mode_ok
17985         instead of HARD_REGNO_MODE_OK.
17986         * config/xtensa/xtensa.h (xtensa_hard_regno_mode_ok): Delete.
17987         (HARD_REGNO_MODE_OK): Delete.
17988         * config/xtensa/xtensa.c (xtensa_hard_regno_mode_ok): Rename to...
17989         (xtensa_hard_regno_mode_ok_p): ...this and make static.
17990         (xtensa_option_override): Update accordingly.
17991         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17992         (xtensa_hard_regno_mode_ok): New function.
17993         * system.h (HARD_REGNO_MODE_OK): Poison.
17995 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17996             Alan Hayward  <alan.hayward@arm.com>
17997             David Sherwood  <david.sherwood@arm.com>
17999         * target.def (hard_regno_call_part_clobbered): New hook.
18000         * doc/tm.texi.in (HARD_REGNO_CALL_PART_CLOBBERED): Replace with...
18001         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): ...this hook.
18002         * doc/tm.texi: Regenerate.
18003         * hooks.h (hook_bool_uint_mode_false): Declare.
18004         * hooks.c (hook_bool_uint_mode_false): New function.
18005         * regs.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18006         * cselib.c (cselib_process_insn): Use
18007         targetm.hard_regno_call_part_clobbered instead of
18008         HARD_REGNO_CALL_PART_CLOBBERED.
18009         * ira-conflicts.c (ira_build_conflicts): Likewise.
18010         * ira-costs.c (ira_tune_allocno_costs): Likewise.
18011         * lra-constraints.c (need_for_call_save_p): Likewise.
18012         * lra-lives.c: Include target.h.
18013         (check_pseudos_live_through_calls): Use
18014         targetm.hard_regno_call_part_clobbered instead of
18015         HARD_REGNO_CALL_PART_CLOBBERED.
18016         * regcprop.c: Include target.h.
18017         (copyprop_hardreg_forward_1): Use
18018         targetm.hard_regno_call_part_clobbered instead of
18019         HARD_REGNO_CALL_PART_CLOBBERED.
18020         * reginfo.c (choose_hard_reg_mode): Likewise.
18021         * regrename.c (check_new_reg_p): Likewise.
18022         * reload.c (find_equiv_reg): Likewise.
18023         * reload1.c (emit_reload_insns): Likewise.
18024         * sched-deps.c (deps_analyze_insn): Likewise.
18025         * sel-sched.c (init_regs_for_mode): Likewise.
18026         (mark_unavailable_hard_regs): Likewise.
18027         * targhooks.c (default_dwarf_frame_reg_mode): Likewise.
18028         * config/aarch64/aarch64.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18029         * config/aarch64/aarch64.c (aarch64_hard_regno_call_part_clobbered):
18030         New function.
18031         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18032         * config/avr/avr.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18033         * config/avr/avr-protos.h (avr_hard_regno_call_part_clobbered):
18034         Delete.
18035         * config/avr/avr.c (avr_hard_regno_call_part_clobbered): Make static
18036         and return a bool.
18037         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18038         * config/i386/i386.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18039         * config/i386/i386.c (ix86_hard_regno_call_part_clobbered): New
18040         function.
18041         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18042         * config/mips/mips.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18043         * config/mips/mips.c (mips_hard_regno_call_part_clobbered): New
18044         function.
18045         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18046         * config/powerpcspe/powerpcspe.h (HARD_REGNO_CALL_PART_CLOBBERED):
18047         Delete.
18048         * config/powerpcspe/powerpcspe.c
18049         (rs6000_hard_regno_call_part_clobbered): New function.
18050         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18051         * config/rs6000/rs6000.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18052         * config/rs6000/rs6000.c (rs6000_hard_regno_call_part_clobbered):
18053         New function.
18054         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18055         * config/s390/s390.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18056         * config/s390/s390.c (s390_hard_regno_call_part_clobbered): New
18057         function.
18058         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18059         * config/sh/sh.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18060         * system.h (HARD_REGNO_CALL_PART_CLOBBERED): Poison.
18062 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
18063             Alan Hayward  <alan.hayward@arm.com>
18064             David Sherwood  <david.sherwood@arm.com>
18066         * rtl.h (subreg_memory_offset): Declare.
18067         * emit-rtl.c (subreg_memory_offset): New function.
18068         * expmed.c (store_bit_field_1): Use it.
18069         * expr.c (undefined_operand_subword_p): Likewise.
18070         * simplify-rtx.c (simplify_subreg): Likewise.
18072 2017-09-04  Alexander Monakov  <amonakov@ispras.ru>
18074         PR rtl-optimization/57448
18075         PR target/67458
18076         PR target/81316
18077         * optabs.c (expand_atomic_load): Place compiler memory barriers if
18078         using atomic_load pattern.
18079         (expand_atomic_store): Likewise.
18081 2017-09-04  Jakub Jelinek  <jakub@redhat.com>
18083         PR sanitizer/81981
18084         * gimple-fold.c (gimple_fold_call): Optimize away useless UBSAN_PTR
18085         and UBSAN_BOUNDS internal calls.  Clean up IFN_UBSAN_OBJECT_SIZE
18086         handling.  Use replace_call_with_value with NULL instead of
18087         gsi_replace, unlink_stmt_vdef and release_defs.
18089         * gdbhooks.py (OptMachineModePrinter.to_string): Use 8 spaces
18090         instead of tab.
18092         * lra-remat.c (reg_overlap_for_remat_p): Fix a pasto.
18094 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
18096         PR bootstrap/82045
18097         * rtl.h (emit_library_call_value_1): Declare.
18098         (emit_library_call): Replace declaration with a series of overloads.
18099         Remove the parameter count argument.
18100         (emit_library_call_value): Likewise.
18101         * calls.c (emit_library_call_value_1): Make global.  Replace varargs
18102         with an "rtx_mode_t *".
18103         (emit_library_call_value): Delete.
18104         (emit_library_call): Likewise.
18105         * asan.c (asan_emit_stack_protection): Update calls accordingly.
18106         (asan_emit_allocas_unpoison): Likewise.
18107         * builtins.c (expand_builtin_powi): Likewise.
18108         (expand_asan_emit_allocas_unpoison): Likewise.
18109         * cfgexpand.c (expand_main_function): Likewise.
18110         * config/aarch64/aarch64.c (aarch64_trampoline_init): Likewise.
18111         * config/aarch64/aarch64.h (PROFILE_HOOK): Likewise.
18112         * config/alpha/alpha.c (alpha_trampoline_init): Likewise.
18113         * config/arm/arm.c (arm_trampoline_init): Likewise.
18114         (arm_call_tls_get_addr): Likewise.
18115         (arm_expand_divmod_libfunc): Likewise.
18116         * config/bfin/bfin.md (umulsi3_highpart): Likewise.
18117         (smulsi3_highpart): Likewise.
18118         * config/c6x/c6x.c (c6x_initialize_trampoline): Likewise.
18119         (c6x_expand_compare): Likewise.
18120         (c6x_expand_movmem): Likewise.
18121         * config/frv/frv.c (frv_trampoline_init): Likewise.
18122         * config/i386/i386.c (ix86_trampoline_init): Likewise.
18123         (ix86_expand_divmod_libfunc): Likewise.
18124         * config/ia64/ia64.c (ia64_expand_tls_address): Likewise.
18125         (ia64_expand_compare): Likewise.
18126         (ia64_profile_hook): Likewise.
18127         * config/ia64/ia64.md (save_stack_nonlocal): Likewise.
18128         (nonlocal_goto): Likewise.
18129         (restore_stack_nonlocal): Likewise.
18130         * config/m32r/m32r.c (block_move_call): Likewise.
18131         (m32r_trampoline_init): Likewise.
18132         * config/m68k/linux.h (FINALIZE_TRAMPOLINE): Likewise.
18133         * config/m68k/m68k.c (m68k_call_tls_get_addr): Likewise.
18134         (m68k_call_m68k_read_tp): Likewise.
18135         * config/microblaze/microblaze.c (microblaze_call_tls_get_addr)
18136         (microblaze_expand_divide): Likewise.
18137         * config/mips/mips.h (mips_args): Likewise.
18138         * config/mips/sdemtk.h (mips_sync_icache): Likewise.
18139         (MIPS_ICACHE_SYNC): Likewise.
18140         * config/nios2/nios2.c (nios2_emit_expensive_div): Likewise.
18141         (nios2_trampoline_init): Likewise.
18142         * config/pa/pa.c (hppa_tls_call): Likewise.
18143         (pa_trampoline_init): Likewise.
18144         * config/pa/pa.md (canonicalize_funcptr_for_compare): Likewise.
18145         * config/powerpcspe/powerpcspe.c (rs6000_legitimize_tls_address)
18146         (expand_strn_compare): Likewise.
18147         (rs6000_generate_compare): Likewise.
18148         (rs6000_expand_float128_convert): Likewise.
18149         (output_profile_hook): Likewise.
18150         (rs6000_trampoline_init): Likewise.
18151         * config/powerpcspe/powerpcspe.md (neg<mode>2): Likewise.
18152         * config/riscv/riscv.h (PROFILE_HOOK): Likewise.
18153         * config/rs6000/rs6000-string.c (expand_strn_compare): Likewise.
18154         * config/rs6000/rs6000.c (rs6000_legitimize_tls_address): Likewise.
18155         (rs6000_generate_compare): Likewise.
18156         (rs6000_expand_float128_convert): Likewise.
18157         (output_profile_hook): Likewise.
18158         (rs6000_trampoline_init): Likewise.
18159         * config/rs6000/rs6000.md (neg<mode>2): Likewise.
18160         * config/sh/sh.c (sh_trampoline_init): Likewise.
18161         * config/sparc/sparc.c (emit_soft_tfmode_libcall): Likewise.
18162         (sparc_emit_float_lib_cmp): Likewise.
18163         (sparc32_initialize_trampoline): Likewise.
18164         (sparc64_initialize_trampoline): Likewise.
18165         (sparc_profile_hook): Likewise.
18166         * config/spu/spu.c (ea_load_store): Likewise.
18167         * config/spu/spu.md (floatunssidf2): Likewise.
18168         * config/tilegx/tilegx.c (tilegx_trampoline_init): Likewise.
18169         * config/tilepro/tilepro.c (tilepro_trampoline_init): Likewise.
18170         * config/visium/visium.c (expand_block_move_4): Likewise.
18171         (expand_block_move_2): Likewise.
18172         (expand_block_move_1): Likewise.
18173         (expand_block_set_4): Likewise.
18174         (expand_block_set_2): Likewise.
18175         (expand_block_set_1): Likewise.
18176         (visium_trampoline_init): Likewise.
18177         (visium_profile_hook): Likewise.
18178         * config/xtensa/xtensa.c (xtensa_expand_nonlocal_goto): Likewise.
18179         (xtensa_setup_frame_addresses): Likewise.
18180         (xtensa_trampoline_init): Likewise.
18181         * except.c (sjlj_emit_function_enter): Likewise.
18182         (sjlj_emit_function_exit): Likewise.
18183         * explow.c (allocate_dynamic_stack_space): Likewise.
18184         (probe_stack_range): Likewise.
18185         * expr.c (convert_mode_scalar): Likewise.
18186         * optabs.c (expand_binop): Likewise.
18187         (expand_twoval_binop_libfunc): Likewise.
18188         (expand_unop): Likewise.
18189         (prepare_cmp_insn): Likewise.
18190         (prepare_float_lib_cmp): Likewise.
18191         (expand_float): Likewise.
18192         (expand_fix): Likewise.
18193         (expand_fixed_convert): Likewise.
18194         (maybe_emit_sync_lock_test_and_set): Likewise.
18195         (expand_atomic_compare_and_swap): Likewise.
18196         (expand_mem_thread_fence): Likewise.
18197         (expand_atomic_fetch_op): Likewise.
18199 2017-09-03  Gerald Pfeifer  <gerald@pfeifer.com>
18201         * doc/generic.texi (OpenACC): Adjust URL.
18202         * doc/invoke.texi (C Dialect Options): Ditto.
18204 2017-09-03  Uros Bizjak  <ubizjak@gmail.com>
18206         * config/i386/i386.md (*bt<mode>): Use nonimmediate_operand
18207         predicate for operand 1.  Add (m,<S>) constraint.
18208         (*jcc_bt<mode>): Use nonimmediate_operand predicate for operand 1.
18209         Prevent memory operand 1 with register operand 2.
18211 2017-09-01  Segher Boessenkool  <segher@kernel.crashing.org>
18213         PR rtl-optimization/82024
18214         * combine.c (try_combine): If the combination result is a PARALLEL,
18215         and we only need to retain the SET in there that would be placed
18216         at I2, check that we can place that at I3 instead, before doing so.
18218 2017-09-01  Jakub Jelinek  <jakub@redhat.com>
18220         PR target/81766
18221         * config/i386/i386.c (ix86_init_large_pic_reg): Return label
18222         instead of void.
18223         (ix86_init_pic_reg): Remember label from ix86_init_large_pic_reg,
18224         if non-NULL and preceded by NOTE_INSN_BASIC_BLOCK, swap the note
18225         and label.
18227 2017-09-01  Joerg Sonnenberger  <joerg@bec.de>
18228             Jeff Law  <law@redhat.com>
18230         * varasm.c (bss_initializer_p): Do not put constants into .bss
18231         (categorize_decl_for_section): Handle bss_initializer_p returning
18232         false when DECL_INITIAL is NULL.
18234 2017-09-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
18236         PR target/82012
18237         * config/s390/s390.c (s390_can_inline_p): New function.
18239 2017-09-01  Jeff Law  <law@redhat.com>
18241         PR tree-optimization/82052
18242         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
18243         Always initialize the returned slot after a hash table miss
18244         when INSERT is true.
18246 2017-09-01  Alexander Monakov  <amonakov@ispras.ru>
18248         * config/s390/s390.md (mem_signal_fence): Remove.
18249         * doc/md.texi (mem_signal_fence): Remove.
18250         * optabs.c (expand_mem_signal_fence): Remove uses of mem_signal_fence.
18251         Update comments.
18252         * target-insns.def (mem_signal_fence): Remove.
18254 2017-09-01  Jakub Jelinek  <jakub@redhat.com>
18256         PR sanitizer/81902
18257         * doc/invoke.texi: Document -fsanitize=pointer-overflow.
18259         PR sanitizer/81923
18260         * asan.c (create_odr_indicator): Strip name encoding from assembler
18261         name before appending it after __odr_asan_.
18263 2017-09-01  Martin Liska  <mliska@suse.cz>
18265         PR tree-optimization/82059
18266         * gimple-ssa-isolate-paths.c (isolate_path): Add profile and
18267         frequency only when an edge is redirected.
18269 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18271         * config/arc/arc-c.c (__ARC_LPC_WIDTH__): Add builtin define.
18272         * config/arc/arc.c (ARC_MAX_LOOP_LENGTH): Define.
18273         (arc_conditional_register_usage): Remove ARC600 lp_count
18274         exception.
18275         (arc_file_start): Emit Tag_ARC_CPU_variation.
18276         (arc_can_use_doloop_p): New conditions to use ZOLs.
18277         (hwloop_fail): New function.
18278         (hwloop_optimize): Likewise.
18279         (hwloop_pattern_reg): Likewise.
18280         (arc_doloop_hooks): New struct, to be used with reorg_loops.
18281         (arc_reorg_loops): New function, calls reorg_loops.
18282         (arc_reorg): Call arc_reorg_loops.  Remove old ZOL handling.
18283         (arc600_corereg_hazard): Remove ZOL checking, case handled by
18284         hwloop_optimize.
18285         (arc_loop_hazard): Remove function, functionality moved into
18286         hwloop_optimize.
18287         (arc_hazard): Remove arc_loop_hazard call.
18288         (arc_adjust_insn_length): Remove ZOL handling, functionality moved
18289         into hwloop_optimize.
18290         (arc_label_align): Remove ZOL handling.
18291         * config/arc/arc.h (LOOP_ALIGN): Changed to 0.
18292         * config/arc/arc.md (doloop_begin): Remove pattern.
18293         (doloop_begin_i): Likewise.
18294         (doloop_end_i): Likewise.
18295         (doloop_fallback): Likewise.
18296         (doloop_fallback_m): Likewise.
18297         (doloop_end): Reimplement expand.
18298         (arc_lp): New pattern for LP instruction.
18299         (loop_end): New pattern.
18300         (loop_fail): Likewise.
18301         (decrement_and_branch_until_zero): Likewise.
18302         * config/arc/arc.opt (mlpc-width): New option.
18303         * doc/invoke.texi (mlpc-width): Document option.
18305 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18307         * config/arc/arc.c (arc_ifcvt): Remove use of merge_blocks call.
18308         (arc_ccfsm_advance): Fix checking for delay slots.
18309         (arc_reorg): Add rtl dump after each call to arc_ifcvt.
18311 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18313         * config/arc/arc.md (movqi_insn): Add stores to save constant long
18314         immediates.
18315         (movhi_insn): Update store instruction constraint which are saving
18316         6-bit short immediates.
18317         (movsi_insn): Consider also short scaled load operations.
18318         (zero_extendhisi2_i): Use Usd constraint instead of T.
18319         (extendhisi2_i): Add q constraint.
18320         (arc_clzsi2): Add type and length attributes.
18321         (arc_ctzsi2): Likewise.
18322         * config/arc/constraints.md (Usc): Update constraint, the
18323         assembler can parse two relocations for a single instruction.
18325 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18327         * config/arc/arc.c (arc_use_anchors_for_symbol_p): New function.
18328         (TARGET_USE_ANCHORS_FOR_SYMBOL_P): Define.
18330 2017-08-31  Olivier Hainque  <hainque@adacore.com>
18332         * config.gcc (powerpc-wrs-vxworks|vxworksae|vxworksmils): Now
18333         match as powerpc-wrs-vxworks*.
18335 2017-08-31  James Greenhalgh  <james.greenhalgh@arm.com>
18337         * config/aarch64/aarch64-simd.md (aarch64_mla_elt_merge<mode>): Fix
18338         register constraint for by-element operand.
18339         (aarch64_mls_elt_merge<mode>): Likewise.
18341 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18343         * config/arc/arc.c (arc_can_follow_jump): Check for short
18344         branches.
18346 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18348         * config.gcc: Use g.opt for arc.
18349         * config/arc/arc.c (LEGITIMATE_SCALED_ADDRESS_P): Deleted,
18350         functionality moved to ...
18351         (legitimate_scaled_address_p): New function, ...here.
18352         (LEGITIMATE_SMALL_DATA_OFFSET_P): New define.
18353         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Use the above define.
18354         (legitimate_offset_address_p): Delete TARGET_NO_SDATA_SET
18355         condition.
18356         (arc_override_options): Handle G option.
18357         (arc_output_pic_addr_const): Correct function definition.
18358         (arc_legitimate_address_p): Use legitimate_scaled_address_p.
18359         (arc_decl_anon_ns_mem_p): Delete.
18360         (arc_in_small_data_p): Overhaul this function to take into
18361         consideration the value given via G option.
18362         (arc_rewrite_small_data_1): Renamed and corrected old
18363         arc_rewrite_small_data function.
18364         (arc_rewrite_small_data): New function.
18365         (small_data_pattern): Don't use pic_offset_table_rtx.
18366         * config/arc/arc.h (CC1_SPEC): Recognize G option.
18367         * config/arc/simdext.md (movmisalignv2hi): Use
18368         prepare_move_operands function.
18369         (mov*): Likewise.
18370         (movmisalign*): Likewise.
18371         * doc/invoke.texi (ARC options): Document -G option.
18373 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18375         * config/arc/arc-protos.h (compact_sda_memory_operand): Update
18376         prototype.
18377         * config/arc/arc.c (arc_print_operand): Output scalled address for
18378         sdata whenever is possible.
18379         (arc_in_small_data_p): Allow sdata for 64bit datum when double
18380         load/stores are available.
18381         (compact_sda_memory_operand): Check for the alignment required by
18382         code density instructions.
18383         * config/arc/arc.md (movsi_insn): Use newly introduced Us0
18384         constraint.
18385         * config/arc/constraints.md (Usd): Update constraint.
18386         (Us0): New constraint.
18387         (Usc): Update constraint.
18389 2017-08-31  Richard Biener  <rguenther@suse.de>
18391         PR middle-end/82054
18392         * dwarf2out.c (dwarf2out_early_global_decl): Process each
18393         function only once.
18395 2017-08-31  Tamar Christina  <tamar.christina@arm.com>
18397         * config/aarch64/aarch64-builtins.c (aarch64_init_simd_builtins):
18398         Resize type_signature.
18400 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18401             Alan Hayward  <alan.hayward@arm.com>
18402             David Sherwood  <david.sherwood@arm.com>
18404         * config/aarch64/aarch64.c (aarch64_base_register_rtx_p): Only allow
18405         subregs whose inner modes can be stored in GPRs.
18406         (aarch64_classify_index): Likewise.
18408 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18409             Alan Hayward  <alan.hayward@arm.com>
18410             David Sherwood  <david.sherwood@arm.com>
18412         * config/aarch64/iterators.md (V_cmp_result): Rename to...
18413         (V_INT_EQUIV): ...this.
18414         (v_cmp_result): Rename to...
18415         (v_int_equiv): ...this.
18416         * config/aarch64/aarch64.md (xorsign<mode>3): Update accordingly.
18417         * config/aarch64/aarch64-simd.md (xorsign<mode>3): Likewise.
18418         (copysign<mode>3): Likewise.
18419         (aarch64_simd_bsl<mode>_internal): Likewise.
18420         (aarch64_simd_bsl<mode>): Likewise.
18421         (vec_cmp<mode><mode>): Likewise.
18422         (vcond<mode><mode>): Likewise.
18423         (vcond<v_cmp_mixed><mode>): Likewise.
18424         (vcondu<mode><v_cmp_mixed>): Likewise.
18425         (aarch64_cm<optab><mode>): Likewise.
18426         (aarch64_cmtst<mode>): Likewise.
18427         (aarch64_fac<optab><mode>): Likewise.
18428         (vec_perm_const<mode>): Likewise.
18429         (vcond_mask_<mode><v_cmp_result>): Rename to...
18430         (vcond_mask_<mode><v_int_equiv>): ...this.
18431         (vec_cmp<mode><v_cmp_result>): Rename to...
18432         (vec_cmp<mode><v_int_equiv>): ...this.
18434 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18435             Alan Hayward  <alan.hayward@arm.com>
18436             David Sherwood  <david.sherwood@arm.com>
18438         * config/aarch64/aarch64-modes.def: Remove 32-, 48- and 64-byte
18439         vector modes.
18440         * config/aarch64/iterators.md (VRL2, VRL3, VRL4): Delete.
18441         * config/aarch64/aarch64.md (UNSPEC_LD2_DREG, UNSPEC_LD3_DREG)
18442         (UNSPEC_LD4_DREG): New unspecs.
18443         * config/aarch64/aarch64-simd.md (aarch64_ld2<mode>_dreg_le)
18444         (aarch64_ld2<mode>_dreg_be): Replace with...
18445         (aarch64_ld2<mode>_dreg): ...this pattern and use the new DREG
18446         unspec.
18447         (aarch64_ld3<mode>_dreg_le)
18448         (aarch64_ld3<mode>_dreg_be): Replace with...
18449         (aarch64_ld3<mode>_dreg): ...this pattern and use the new DREG
18450         unspec.
18451         (aarch64_ld4<mode>_dreg_le)
18452         (aarch64_ld4<mode>_dreg_be): Replace with...
18453         (aarch64_ld4<mode>_dreg): ...this pattern and use the new DREG
18454         unspec.
18456 2017-08-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
18458         PR tree-optimization/81987
18459         * gimple-ssa-strength-reduction.c (insert_initializers): Don't
18460         insert an initializer in a location not dominated by the stride
18461         definition.
18463 2017-08-30  Eric Botcazou  <ebotcazou@adacore.com>
18465         * tree-eh.c (lower_try_finally_switch): Set the location of the finally
18466         on the entire header of the finally block in the fallthru case.
18468 2017-08-30  Eric Botcazou  <ebotcazou@adacore.com>
18470         * varasm.c (decode_addr_const): Deal with INDIRECT_REF <INTEGER_CST>.
18472 2017-08-30  Pat Haugen  <pthaugen@us.ibm.com>
18474         * config/rs6000/rs6000.c (rs6000_emit_prologue_move_from_cr): Rename from
18475         rs6000_emit_move_from_cr and call renamed function.
18476         (rs6000_emit_prologue): Call renamed functions.
18477         * config/rs6000/rs6000.md (prologue_movesi_from_cr): Rename from
18478         movesi_from_cr, remove volatile CRs.
18480 2017-08-30  Jon Beniston  <jon@beniston.com>
18481             Richard Biener  <rguenther@suse.de>
18483         * tree-vect-patterns.c (vect_pattern_recog_1): Use VECTOR_TYPE_P instead
18484         of VECTOR_MODE_P check.
18485         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Allow single
18486         element vector types.
18488 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18490         * df.h (df_read_modify_subreg_p): Remove in favor of...
18491         * rtl.h (read_modify_subreg_p): ...this new function.  Take a
18492         const_rtx instead of an rtx.
18493         * cprop.c (local_cprop_find_used_regs): Update accordingly.
18494         * df-problems.c (df_word_lr_mark_ref): Likewise.
18495         * ira-lives.c (mark_pseudo_reg_live): Likewise.
18496         (mark_pseudo_reg_dead): Likewise.
18497         (mark_ref_dead): Likewise.
18498         * reginfo.c (init_subregs_of_mode): Likewise.
18499         * sched-deps.c (sched_analyze_1): Likewise.
18500         * df-scan.c (df_def_record_1): Likewise.
18501         (df_uses_record): Likewise.
18502         (df_read_modify_subreg_p): Remove in favor of...
18503         * rtlanal.c (read_modify_subreg_p): ...this new function.  Take a
18504         const_rtx instead of an rtx.
18506 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18507             Alan Hayward  <alan.hayward@arm.com>
18508             David Sherwood  <david.sherwood@arm.com>
18510         * rtl.h (partial_subreg_p): New function.
18511         * caller-save.c (save_call_clobbered_regs): Use it.
18512         * calls.c (expand_call): Likewise.
18513         * combine.c (combinable_i3pat): Likewise.
18514         (simplify_set): Likewise.
18515         (make_extraction): Likewise.
18516         (make_compound_operation_int): Likewise.
18517         (gen_lowpart_or_truncate): Likewise.
18518         (force_to_mode): Likewise.
18519         (make_field_assignment): Likewise.
18520         (reg_truncated_to_mode): Likewise.
18521         (record_truncated_value): Likewise.
18522         (move_deaths): Likewise.
18523         * cse.c (record_jump_cond): Likewise.
18524         (cse_insn): Likewise.
18525         * cselib.c (cselib_lookup_1): Likewise.
18526         * expmed.c (extract_bit_field_using_extv): Likewise.
18527         * function.c (assign_parm_setup_reg): Likewise.
18528         * ifcvt.c (noce_convert_multiple_sets): Likewise.
18529         * ira-build.c (create_insn_allocnos): Likewise.
18530         * lra-coalesce.c (merge_pseudos): Likewise.
18531         * lra-constraints.c (match_reload): Likewise.
18532         (simplify_operand_subreg): Likewise.
18533         (curr_insn_transform): Likewise.
18534         * lra-lives.c (process_bb_lives): Likewise.
18535         * lra.c (new_insn_reg): Likewise.
18536         (lra_substitute_pseudo): Likewise.
18537         * regcprop.c (mode_change_ok): Likewise.
18538         (maybe_mode_change): Likewise.
18539         (copyprop_hardreg_forward_1): Likewise.
18540         * reload.c (push_reload): Likewise.
18541         (find_reloads): Likewise.
18542         (find_reloads_subreg_address): Likewise.
18543         * reload1.c (alter_reg): Likewise.
18544         (eliminate_regs_1): Likewise.
18545         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
18547 2017-08-30  David Edelsohn  <dje.gcc@gmail.com>
18549         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Revert
18550         back to if statements, including unpack.
18552 2017-08-30  Martin Liska  <mliska@suse.cz>
18554         PR inline-asm/82001
18555         * ipa-icf-gimple.c (func_checker::compare_tree_list_operand):
18556         Rename to ...
18557         (func_checker::compare_asm_inputs_outputs): ... this function.
18558         (func_checker::compare_gimple_asm): Use the function to compare
18559         also ASM constrains.
18560         * ipa-icf-gimple.h: Rename the function.
18562 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18563             Alan Hayward  <alan.hayward@arm.com>
18564             David Sherwood  <david.sherwood@arm.com>
18566         * coretypes.h (complex_mode): New type.
18567         * gdbhooks.py (build_pretty_printer): Handle it.
18568         * machmode.h (complex_mode): New class.
18569         (complex_mode::includes_p): New function.
18570         (is_complex_int_mode): Likewise.
18571         (is_complex_float_mode): Likewise.
18572         * genmodes.c (get_mode_class): Handle complex mode classes.
18573         * function.c (expand_function_end): Use is_complex_int_mode.
18575 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18576             Alan Hayward  <alan.hayward@arm.com>
18577             David Sherwood  <david.sherwood@arm.com>
18579         * coretypes.h (scalar_mode_pod): New typedef.
18580         * gdbhooks.py (build_pretty_printer): Handle it.
18581         * machmode.h (gt_ggc_mx, gt_pch_nx): New functions.
18582         * fixed-value.h (fixed_value::mode): Change type to scalar_mode_pod.
18583         * fold-const.c (fold_convert_const_int_from_fixed): Use scalar_mode.
18584         * tree-streamer-in.c (unpack_ts_fixed_cst_value_fields): Use
18585         as_a <scalar_mode>.
18587 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18588             Alan Hayward  <alan.hayward@arm.com>
18589             David Sherwood  <david.sherwood@arm.com>
18591         * machmode.h (mode_for_vector): Take a scalar_mode instead
18592         of a machine_mode.
18593         * stor-layout.c (mode_for_vector): Likewise.
18594         * explow.c (promote_mode): Use as_a <scalar_mode>.
18595         * sdbout.c (sdbout_parms): Use is_a <scalar_mode>.
18597 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18598             Alan Hayward  <alan.hayward@arm.com>
18599             David Sherwood  <david.sherwood@arm.com>
18601         * target.def (preferred_simd_mode): Take a scalar_mode
18602         instead of a machine_mode.
18603         * targhooks.h (default_preferred_simd_mode): Likewise.
18604         * targhooks.c (default_preferred_simd_mode): Likewise.
18605         * config/arc/arc.c (arc_preferred_simd_mode): Likewise.
18606         * config/arm/arm.c (arm_preferred_simd_mode): Likewise.
18607         * config/c6x/c6x.c (c6x_preferred_simd_mode): Likewise.
18608         * config/epiphany/epiphany.c (epiphany_preferred_simd_mode): Likewise.
18609         * config/i386/i386.c (ix86_preferred_simd_mode): Likewise.
18610         * config/mips/mips.c (mips_preferred_simd_mode): Likewise.
18611         * config/nvptx/nvptx.c (nvptx_preferred_simd_mode): Likewise.
18612         * config/powerpcspe/powerpcspe.c (rs6000_preferred_simd_mode):
18613         Likewise.
18614         * config/rs6000/rs6000.c (rs6000_preferred_simd_mode): Likewise.
18615         * config/s390/s390.c (s390_preferred_simd_mode): Likewise.
18616         * config/sparc/sparc.c (sparc_preferred_simd_mode): Likewise.
18617         * config/aarch64/aarch64.c (aarch64_preferred_simd_mode): Likewise.
18618         (aarch64_simd_scalar_immediate_valid_for_move): Update accordingly.
18619         * doc/tm.texi: Regenerate.
18620         * optabs-query.c (can_vec_mask_load_store_p): Return false for
18621         non-scalar modes.
18623 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18624             Alan Hayward  <alan.hayward@arm.com>
18625             David Sherwood  <david.sherwood@arm.com>
18627         * target.def (scalar_mode_supported_p): Take a scalar_mode
18628         instead of a machine_mode.
18629         * targhooks.h (default_scalar_mode_supported_p): Likewise.
18630         * targhooks.c (default_scalar_mode_supported_p): Likewise.
18631         * config/aarch64/aarch64.c (aarch64_scalar_mode_supported_p): Likewise.
18632         * config/alpha/alpha.c (alpha_scalar_mode_supported_p): Likewise.
18633         * config/arm/arm.c (arm_scalar_mode_supported_p): Likewise.
18634         * config/avr/avr.c (avr_scalar_mode_supported_p): Likewise.
18635         * config/c6x/c6x.c (c6x_scalar_mode_supported_p): Likewise.
18636         * config/i386/i386.c (ix86_scalar_mode_supported_p): Likewise.
18637         * config/ia64/ia64.c (ia64_scalar_mode_supported_p): Likewise.
18638         * config/mips/mips.c (mips_scalar_mode_supported_p): Likewise.
18639         * config/msp430/msp430.c (msp430_scalar_mode_supported_p): Likewise.
18640         * config/pa/pa.c (pa_scalar_mode_supported_p): Likewise.
18641         * config/pdp11/pdp11.c (pdp11_scalar_mode_supported_p): Likewise.
18642         * config/powerpcspe/powerpcspe.c (rs6000_scalar_mode_supported_p):
18643         Likewise.
18644         * config/rs6000/rs6000.c (rs6000_scalar_mode_supported_p): Likewise.
18645         * config/s390/s390.c (s390_scalar_mode_supported_p): Likewise.
18646         * config/spu/spu.c (spu_scalar_mode_supported_p): Likewise.
18647         * config/tilegx/tilegx.c (tilegx_scalar_mode_supported_p): Likewise.
18648         * config/tilepro/tilepro.c (tilepro_scalar_mode_supported_p):
18649         Likewise.
18650         * doc/tm.texi: Regenerate.
18652 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18653             Alan Hayward  <alan.hayward@arm.com>
18654             David Sherwood  <david.sherwood@arm.com>
18656         * coretypes.h (opt_scalar_mode): New typedef.
18657         * gdbhooks.py (build_pretty_printers): Handle it.
18658         * machmode.h (mode_iterator::get_2xwider): Add overload for
18659         opt_mode<T>.
18660         * emit-rtl.c (init_emit_once): Use opt_scalar_mode when iterating
18661         over scalar modes.
18662         * expr.c (convert_mode_scalar): Likewise.
18663         * omp-low.c (omp_clause_aligned_alignment): Likewise.
18664         * optabs.c (expand_float): Likewise.
18665         (expand_fix): Likewise.
18666         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
18668 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18669             Alan Hayward  <alan.hayward@arm.com>
18670             David Sherwood  <david.sherwood@arm.com>
18672         * optabs.c (expand_float): Explicitly check for scalars before
18673         using a branching expansion.
18674         (expand_fix): Likewise.
18676 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18677             Alan Hayward  <alan.hayward@arm.com>
18678             David Sherwood  <david.sherwood@arm.com>
18680         * expr.c (convert_mode): Split scalar handling out into...
18681         (convert_mode_scalar): ...this new function.  Treat the modes
18682         as scalar_modes.
18684 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18685             Alan Hayward  <alan.hayward@arm.com>
18686             David Sherwood  <david.sherwood@arm.com>
18688         * omp-expand.c (expand_omp_atomic): Use is_int_mode, is_float_mode
18689         and scalar_mode.
18690         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Likewise.
18692 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18693             Alan Hayward  <alan.hayward@arm.com>
18694             David Sherwood  <david.sherwood@arm.com>
18696         * fixed-value.h (fixed_from_double_int): Take a scalar_mode
18697         rather than a machine_mode.
18698         (fixed_from_string): Likewise.
18699         (fixed_convert): Likewise.
18700         (fixed_convert_from_int): Likewise.
18701         (fixed_convert_from_real): Likewise.
18702         (real_convert_from_fixed): Likewise.
18703         * fixed-value.c (fixed_from_double_int): Likewise.
18704         (fixed_from_string): Likewise.
18705         (fixed_convert): Likewise.
18706         (fixed_convert_from_int): Likewise.
18707         (fixed_convert_from_real): Likewise.
18708         (real_convert_from_fixed): Likewise.
18709         * config/avr/avr.c (avr_out_round): Use as_a <scalar_mode>.
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         * emit-rtl.c (immed_double_const): Use is_a <scalar_mode> instead
18716         of separate mode class checks.  Do not allow vector modes here.
18717         (immed_wide_int_const): Use as_a <scalar_mode>.
18718         * explow.c (trunc_int_for_mode): Likewise.
18719         * rtl.h (wi::int_traits<rtx_mode_t>::get_precision): Likewise.
18720         (wi::shwi): Likewise.
18721         (wi::min_value): Likewise.
18722         (wi::max_value): Likewise.
18723         * dwarf2out.c (loc_descriptor): Likewise.
18724         * simplify-rtx.c (simplify_immed_subreg): Fix rtx_mode_t argument
18725         for CONST_WIDE_INT.
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         * tree.h (SCALAR_TYPE_MODE): New macro.
18732         * expr.c (expand_expr_addr_expr_1): Use it.
18733         (expand_expr_real_2): Likewise.
18734         * fold-const.c (fold_convert_const_fixed_from_fixed): Likeise.
18735         (fold_convert_const_fixed_from_int): Likewise.
18736         (fold_convert_const_fixed_from_real): Likewise.
18737         (native_encode_fixed): Likewise
18738         (native_encode_complex): Likewise
18739         (native_encode_vector): Likewise.
18740         (native_interpret_fixed): Likewise.
18741         (native_interpret_real): Likewise.
18742         (native_interpret_complex): Likewise.
18743         (native_interpret_vector): Likewise.
18744         * omp-simd-clone.c (simd_clone_adjust_return_type): Likewise.
18745         (simd_clone_adjust_argument_types): Likewise.
18746         (simd_clone_init_simd_arrays): Likewise.
18747         (simd_clone_adjust): Likewise.
18748         * stor-layout.c (layout_type): Likewise.
18749         * tree.c (build_minus_one_cst): Likewise.
18750         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
18751         * tree-inline.c (estimate_move_cost): Likewise.
18752         * tree-ssa-math-opts.c (convert_plusminus_to_widen): Likewise.
18753         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise.
18754         (vectorizable_reduction): Likewise.
18755         * tree-vect-patterns.c (vect_recog_widen_mult_pattern): Likewise.
18756         (vect_recog_mixed_size_cond_pattern): Likewise.
18757         (check_bool_pattern): Likewise.
18758         (adjust_bool_pattern): Likewise.
18759         (search_type_for_mask_1): Likewise.
18760         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
18761         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
18762         (vectorizable_load): Likewise.
18763         (vectorizable_store): Likewise.
18764         * ubsan.c (ubsan_encode_value): Likewise.
18765         * varasm.c (output_constant): Likewise.
18767 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18768             Alan Hayward  <alan.hayward@arm.com>
18769             David Sherwood  <david.sherwood@arm.com>
18771         * coretypes.h (scalar_mode): New class.
18772         * machmode.h (scalar_mode): Likewise.
18773         (scalar_mode::includes_p): New function.
18774         (mode_to_inner): Return a scalar_mode rather than a machine_mode.
18775         * gdbhooks.py (build_pretty_printers): Handle scalar_mode.
18776         * genmodes.c (get_mode_class): Handle remaining scalar modes.
18777         * cfgexpand.c (expand_debug_expr): Use scalar_mode.
18778         * expmed.c (store_bit_field_1): Likewise.
18779         (extract_bit_field_1): Likewise.
18780         * expr.c (write_complex_part): Likewise.
18781         (read_complex_part): Likewise.
18782         (emit_move_complex_push): Likewise.
18783         (expand_expr_real_2): Likewise.
18784         * function.c (assign_parm_setup_reg): Likewise.
18785         (assign_parms_unsplit_complex): Likewise.
18786         * optabs.c (expand_binop): Likewise.
18787         * rtlanal.c (subreg_get_info): Likewise.
18788         * simplify-rtx.c (simplify_immed_subreg): Likewise.
18789         * varasm.c (output_constant_pool_2): Likewise.
18791 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18792             Alan Hayward  <alan.hayward@arm.com>
18793             David Sherwood  <david.sherwood@arm.com>
18795         * expmed.c (extract_high_half): Use scalar_int_mode and remove
18796         assertion.
18797         (expmed_mult_highpart_optab): Likewise.
18798         (expmed_mult_highpart): Likewise.
18800 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18801             Alan Hayward  <alan.hayward@arm.com>
18802             David Sherwood  <david.sherwood@arm.com>
18804         * builtins.h (builtin_strncpy_read_str): Take a scalar_int_mode
18805         instead of a machine_mode.
18806         (builtin_memset_read_str): Likewise.
18807         * builtins.c (c_readstr): Likewise.
18808         (builtin_memcpy_read_str): Likewise.
18809         (builtin_strncpy_read_str): Likewise.
18810         (builtin_memset_read_str): Likewise.
18811         (builtin_memset_gen_str): Likewise.
18812         (expand_builtin_signbit): Use scalar_int_mode for local variables.
18813         * cfgexpand.c (convert_debug_memory_address): Take a scalar_int_mode
18814         instead of a machine_mode.
18815         * combine.c (simplify_if_then_else): Use scalar_int_mode for local
18816         variables.
18817         (make_extraction): Likewise.
18818         (try_widen_shift_mode): Take and return scalar_int_modes instead
18819         of machine_modes.
18820         * config/aarch64/aarch64.c (aarch64_libgcc_cmp_return_mode): Return
18821         a scalar_int_mode instead of a machine_mode.
18822         * config/avr/avr.c (avr_addr_space_address_mode): Likewise.
18823         (avr_addr_space_pointer_mode): Likewise.
18824         * config/cr16/cr16.c (cr16_unwind_word_mode): Likewise.
18825         * config/msp430/msp430.c (msp430_addr_space_pointer_mode): Likewise.
18826         (msp430_unwind_word_mode): Likewise.
18827         * config/spu/spu.c (spu_unwind_word_mode): Likewise.
18828         (spu_addr_space_pointer_mode): Likewise.
18829         (spu_addr_space_address_mode): Likewise.
18830         (spu_libgcc_cmp_return_mode): Likewise.
18831         (spu_libgcc_shift_count_mode): Likewise.
18832         * config/rl78/rl78.c (rl78_addr_space_address_mode): Likewise.
18833         (rl78_addr_space_pointer_mode): Likewise.
18834         (fl78_unwind_word_mode): Likewise.
18835         (rl78_valid_pointer_mode): Take a scalar_int_mode instead of a
18836         machine_mode.
18837         * config/alpha/alpha.c (vms_valid_pointer_mode): Likewise.
18838         * config/ia64/ia64.c (ia64_vms_valid_pointer_mode): Likewise.
18839         * config/mips/mips.c (mips_mode_rep_extended): Likewise.
18840         (mips_valid_pointer_mode): Likewise.
18841         * config/tilegx/tilegx.c (tilegx_mode_rep_extended): Likewise.
18842         * config/ft32/ft32.c (ft32_valid_pointer_mode): Likewise.
18843         (ft32_addr_space_pointer_mode): Return a scalar_int_mode instead
18844         of a machine_mode.
18845         (ft32_addr_space_address_mode): Likewise.
18846         * config/m32c/m32c.c (m32c_valid_pointer_mode): Take a
18847         scalar_int_mode instead of a machine_mode.
18848         (m32c_addr_space_pointer_mode): Return a scalar_int_mode instead
18849         of a machine_mode.
18850         (m32c_addr_space_address_mode): Likewise.
18851         * config/powerpcspe/powerpcspe.c (rs6000_abi_word_mode): Likewise.
18852         (rs6000_eh_return_filter_mode): Likewise.
18853         * config/rs6000/rs6000.c (rs6000_abi_word_mode): Likewise.
18854         (rs6000_eh_return_filter_mode): Likewise.
18855         * config/s390/s390.c (s390_libgcc_cmp_return_mode): Likewise.
18856         (s390_libgcc_shift_count_mode): Likewise.
18857         (s390_unwind_word_mode): Likewise.
18858         (s390_valid_pointer_mode): Take a scalar_int_mode rather than a
18859         machine_mode.
18860         * target.def (mode_rep_extended): Likewise.
18861         (valid_pointer_mode): Likewise.
18862         (addr_space.valid_pointer_mode): Likewise.
18863         (eh_return_filter_mode): Return a scalar_int_mode rather than
18864         a machine_mode.
18865         (libgcc_cmp_return_mode): Likewise.
18866         (libgcc_shift_count_mode): Likewise.
18867         (unwind_word_mode): Likewise.
18868         (addr_space.pointer_mode): Likewise.
18869         (addr_space.address_mode): Likewise.
18870         * doc/tm.texi: Regenerate.
18871         * dojump.c (prefer_and_bit_test): Take a scalar_int_mode rather than
18872         a machine_mode.
18873         (do_jump): Use scalar_int_mode for local variables.
18874         * dwarf2cfi.c (init_return_column_size): Take a scalar_int_mode
18875         rather than a machine_mode.
18876         * dwarf2out.c (convert_descriptor_to_mode): Likewise.
18877         (scompare_loc_descriptor_wide): Likewise.
18878         (scompare_loc_descriptor_narrow): Likewise.
18879         * emit-rtl.c (adjust_address_1): Use scalar_int_mode for local
18880         variables.
18881         * except.c (sjlj_emit_dispatch_table): Likewise.
18882         (expand_builtin_eh_copy_values): Likewise.
18883         * explow.c (convert_memory_address_addr_space_1): Likewise.
18884         Take a scalar_int_mode rather than a machine_mode.
18885         (convert_memory_address_addr_space): Take a scalar_int_mode rather
18886         than a machine_mode.
18887         (memory_address_addr_space): Use scalar_int_mode for local variables.
18888         * expmed.h (expand_mult_highpart_adjust): Take a scalar_int_mode
18889         rather than a machine_mode.
18890         * expmed.c (mask_rtx): Likewise.
18891         (init_expmed_one_conv): Likewise.
18892         (expand_mult_highpart_adjust): Likewise.
18893         (extract_high_half): Likewise.
18894         (expmed_mult_highpart_optab): Likewise.
18895         (expmed_mult_highpart): Likewise.
18896         (expand_smod_pow2): Likewise.
18897         (expand_sdiv_pow2): Likewise.
18898         (emit_store_flag_int): Likewise.
18899         (adjust_bit_field_mem_for_reg): Use scalar_int_mode for local
18900         variables.
18901         (extract_low_bits): Likewise.
18902         * expr.h (by_pieces_constfn): Take a scalar_int_mode rather than
18903         a machine_mode.
18904         * expr.c (pieces_addr::adjust):  Likewise.
18905         (can_store_by_pieces): Likewise.
18906         (store_by_pieces): Likewise.
18907         (clear_by_pieces_1): Likewise.
18908         (expand_expr_addr_expr_1): Likewise.
18909         (expand_expr_addr_expr): Use scalar_int_mode for local variables.
18910         (expand_expr_real_1): Likewise.
18911         (try_casesi): Likewise.
18912         * final.c (shorten_branches): Likewise.
18913         * fold-const.c (fold_convert_const_int_from_fixed): Change the
18914         type of "mode" to machine_mode.
18915         * internal-fn.c (expand_arith_overflow_result_store): Take a
18916         scalar_int_mode rather than a machine_mode.
18917         (expand_mul_overflow): Use scalar_int_mode for local variables.
18918         * loop-doloop.c (doloop_modify): Likewise.
18919         (doloop_optimize): Likewise.
18920         * optabs.c (expand_subword_shift): Take a scalar_int_mode rather
18921         than a machine_mode.
18922         (expand_doubleword_shift_condmove): Likewise.
18923         (expand_doubleword_shift): Likewise.
18924         (expand_doubleword_clz): Likewise.
18925         (expand_doubleword_popcount): Likewise.
18926         (expand_doubleword_parity): Likewise.
18927         (expand_absneg_bit): Use scalar_int_mode for local variables.
18928         (prepare_float_lib_cmp): Likewise.
18929         * rtl.h (convert_memory_address_addr_space_1): Take a scalar_int_mode
18930         rather than a machine_mode.
18931         (convert_memory_address_addr_space): Likewise.
18932         (get_mode_bounds): Likewise.
18933         (get_address_mode): Return a scalar_int_mode rather than a
18934         machine_mode.
18935         * rtlanal.c (get_address_mode): Likewise.
18936         * stor-layout.c (get_mode_bounds): Take a scalar_int_mode rather
18937         than a machine_mode.
18938         * targhooks.c (default_mode_rep_extended): Likewise.
18939         (default_valid_pointer_mode): Likewise.
18940         (default_addr_space_valid_pointer_mode): Likewise.
18941         (default_eh_return_filter_mode): Return a scalar_int_mode rather
18942         than a machine_mode.
18943         (default_libgcc_cmp_return_mode): Likewise.
18944         (default_libgcc_shift_count_mode): Likewise.
18945         (default_unwind_word_mode): Likewise.
18946         (default_addr_space_pointer_mode): Likewise.
18947         (default_addr_space_address_mode): Likewise.
18948         * targhooks.h (default_eh_return_filter_mode): Likewise.
18949         (default_libgcc_cmp_return_mode): Likewise.
18950         (default_libgcc_shift_count_mode): Likewise.
18951         (default_unwind_word_mode): Likewise.
18952         (default_addr_space_pointer_mode): Likewise.
18953         (default_addr_space_address_mode): Likewise.
18954         (default_mode_rep_extended): Take a scalar_int_mode rather than
18955         a machine_mode.
18956         (default_valid_pointer_mode): Likewise.
18957         (default_addr_space_valid_pointer_mode): Likewise.
18958         * tree-ssa-address.c (addr_for_mem_ref): Use scalar_int_mode for
18959         local variables.
18960         * tree-ssa-loop-ivopts.c (get_shiftadd_cost): Take a scalar_int_mode
18961         rather than a machine_mode.
18962         * tree-switch-conversion.c (array_value_type): Use scalar_int_mode
18963         for local variables.
18964         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
18965         * var-tracking.c (use_narrower_mode): Take a scalar_int_mode rather
18966         than a machine_mode.
18968 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18969             Alan Hayward  <alan.hayward@arm.com>
18970             David Sherwood  <david.sherwood@arm.com>
18972         * dojump.c (do_jump_by_parts_greater_rtx): Change the type of
18973         the mode argument to scalar_int_mode.
18974         (do_jump_by_parts_zero_rtx): Likewise.
18975         (do_jump_by_parts_equality_rtx): Likewise.
18976         (do_jump_by_parts_greater): Take a mode argument.
18977         (do_jump_by_parts_equality): Likewise.
18978         (do_jump_1): Update calls accordingly.
18980 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18981             Alan Hayward  <alan.hayward@arm.com>
18982             David Sherwood  <david.sherwood@arm.com>
18984         * is-a.h (safe_dyn_cast): New function.
18985         * rtl.h (rtx_jump_table_data::get_data_mode): New function.
18986         (jump_table_for_label): Likewise.
18987         * final.c (final_addr_vec_align): Take an rtx_jump_table_data *
18988         instead of an rtx_insn *.
18989         (shorten_branches): Use dyn_cast instead of LABEL_P and
18990         JUMP_TABLE_DATA_P.  Use jump_table_for_label and
18991         rtx_jump_table_data::get_data_mode.
18992         (final_scan_insn): Likewise.
18994 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18995             Alan Hayward  <alan.hayward@arm.com>
18996             David Sherwood  <david.sherwood@arm.com>
18998         * combine.c (try_combine): Use is_a <scalar_int_mode> when
18999         trying to combine a full-register integer set with a subreg
19000         integer set.
19002 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19003             Alan Hayward  <alan.hayward@arm.com>
19004             David Sherwood  <david.sherwood@arm.com>
19006         * expr.c (expand_expr_addr_expr): Add a new_tmode local variable
19007         that is always either address_mode or pointer_mode.
19009 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19010             Alan Hayward  <alan.hayward@arm.com>
19011             David Sherwood  <david.sherwood@arm.com>
19013         * expr.c (expand_expr_real_2): Use word_mode instead of innermode
19014         when the two are known to be equal.
19016 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19017             Alan Hayward  <alan.hayward@arm.com>
19018             David Sherwood  <david.sherwood@arm.com>
19020         * simplify-rtx.c (simplify_const_unary_operation): Use
19021         is_a <scalar_int_mode> instead of checking for a nonzero
19022         precision.  Forcibly convert op_mode to a scalar_int_mode
19023         in that case.  More clearly differentiate the operand and
19024         result modes and use the former when deciding what the value
19025         of a count-bits operation should be.  Use is_int_mode instead
19026         of checking for a MODE_INT.  Remove redundant check for whether
19027         this mode has a zero precision.
19029 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19030             Alan Hayward  <alan.hayward@arm.com>
19031             David Sherwood  <david.sherwood@arm.com>
19033         * optabs.c (widen_leading): Change the type of the mode argument
19034         to scalar_int_mode.  Use opt_scalar_int_mode for the mode iterator.
19035         (widen_bswap): Likewise.
19036         (expand_parity): Likewise.
19037         (expand_ctz): Change the type of the mode argument to scalar_int_mode.
19038         (expand_ffs): Likewise.
19039         (epand_unop): Check for scalar integer modes before calling the
19040         above routines.
19042 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19043             Alan Hayward  <alan.hayward@arm.com>
19044             David Sherwood  <david.sherwood@arm.com>
19046         * expr.c (const_scalar_mask_from_tree): Add a mode argument.
19047         Expand commentary.
19048         (expand_expr_real_1): Update call accordingly.
19050 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19051             Alan Hayward  <alan.hayward@arm.com>
19052             David Sherwood  <david.sherwood@arm.com>
19054         * expmed.c (store_bit_field_using_insv): Add op0_mode and
19055         value_mode arguments.  Use scalar_int_mode internally.
19056         (store_bit_field_1): Rename the new integer mode from imode
19057         to op0_mode and use it instead of GET_MODE (op0).  Update calls
19058         to store_split_bit_field, store_bit_field_using_insv and
19059         store_fixed_bit_field.
19060         (store_fixed_bit_field): Add op0_mode and value_mode arguments.
19061         Use scalar_int_mode internally.  Use a bit count rather than a mode
19062         when calculating the largest bit size for get_best_mode.
19063         Update calls to store_split_bit_field and store_fixed_bit_field_1.
19064         (store_fixed_bit_field_1): Add mode and value_mode arguments.
19065         Remove assertion that OP0 has a scalar integer mode.
19066         (store_split_bit_field): Add op0_mode and value_mode arguments.
19067         Update calls to extract_fixed_bit_field.
19068         (extract_bit_field_using_extv): Add an op0_mode argument.
19069         Use scalar_int_mode internally.
19070         (extract_bit_field_1): Rename the new integer mode from imode to
19071         op0_mode and use it instead of GET_MODE (op0).  Update calls to
19072         extract_split_bit_field, extract_bit_field_using_extv and
19073         extract_fixed_bit_field.
19074         (extract_fixed_bit_field): Add an op0_mode argument.  Update calls
19075         to extract_split_bit_field and extract_fixed_bit_field_1.
19076         (extract_fixed_bit_field_1): Add a mode argument.  Remove assertion
19077         that OP0 has a scalar integer mode.  Use as_a <scalar_int_mode>
19078         on the target mode.
19079         (extract_split_bit_field): Add an op0_mode argument.  Update call
19080         to extract_fixed_bit_field.
19082 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19083             Alan Hayward  <alan.hayward@arm.com>
19084             David Sherwood  <david.sherwood@arm.com>
19086         * cse.c (cse_insn): Use opt_scalar_int_mode for the mode iterator.
19087         * explow.c (hard_function_value): Likewise.
19088         * expmed.c (extract_fixed_bit_field_1): Likewise.  Move the
19089         convert_to_mode call outside the loop.
19090         * expr.c (alignment_for_piecewise_move): Use opt_scalar_int_mode
19091         for the mode iterator.  Require the mode specified by max_pieces
19092         to exist.
19093         (emit_block_move_via_movmem): Use opt_scalar_int_mode for the
19094         mode iterator.
19095         (copy_blkmode_to_reg): Likewise.
19096         (set_storage_via_setmem): Likewise.
19097         * optabs.c (prepare_cmp_insn): Likewise.
19098         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
19099         * stor-layout.c (finish_bitfield_representative): Likewise.
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         * rtl.h (subreg_unpromoted_mode, subreg_promoted_mode): New functions.
19106         * expr.c (convert_move): Use them.
19107         (convert_modes): Likewise.
19108         (store_expr_with_bounds): Likewise.
19110 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19111             Alan Hayward  <alan.hayward@arm.com>
19112             David Sherwood  <david.sherwood@arm.com>
19114         * rtl.h (rtl_hooks::reg_nonzero_bits): Add a scalar_int_mode
19115         parameter for the mode of "x".  Remove the "known_x", "known_mode"
19116         and "known_ret" arguments.  Change the type of the mode argument
19117         to scalar_int_mode.
19118         (rtl_hooks:reg_num_sign_bit_copies): Likewise.
19119         * combine.c (reg_nonzero_bits_for_combine): Update accordingly.
19120         (reg_num_sign_bit_copies_for_combine): Likewise.
19121         * rtlanal.c (nonzero_bits1): Likewise.
19122         (num_sign_bit_copies1): Likewise.
19123         * rtlhooks-def.h (reg_nonzero_bits_general): Likewise.
19124         (reg_num_sign_bit_copies_general): Likewise.
19125         * rtlhooks.c (reg_num_sign_bit_copies_general): Likewise.
19126         (reg_nonzero_bits_general): Likewise.
19128 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19129             Alan Hayward  <alan.hayward@arm.com>
19130             David Sherwood  <david.sherwood@arm.com>
19132         * rtlanal.c (num_sign_bit_copies): Handle VOIDmode here rather
19133         than in subroutines.  Return 1 for non-integer modes.
19134         (cached_num_sign_bit_copies): Change the type of the mode parameter
19135         to scalar_int_mode.
19136         (num_sign_bit_copies1): Likewise.  Remove early exit for other mode
19137         classes.  Handle CONST_INT_P first and then check whether X also
19138         has a scalar integer mode.  Check the same thing for inner registers
19139         of a SUBREG and for values that are being extended or truncated.
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         * rtlanal.c (nonzero_bits): Handle VOIDmode here rather than
19146         in subroutines.  Return the mode mask for non-integer modes.
19147         (cached_nonzero_bits): Change the type of the mode parameter
19148         to scalar_int_mode.
19149         (nonzero_bits1): Likewise.  Remove early exit for other mode
19150         classes.  Handle CONST_INT_P first and then check whether X
19151         also has a scalar integer mode.
19153 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19154             Alan Hayward  <alan.hayward@arm.com>
19155             David Sherwood  <david.sherwood@arm.com>
19157         * expr.c (widest_int_mode_for_size): Make the comment match the code.
19158         Return a scalar_int_mode and assert that the size is greater than
19159         one byte.
19160         (by_pieces_ninsns): Update accordingly and remove VOIDmode handling.
19161         (op_by_pieces_d::op_by_pieces_d): Likewise.
19162         (op_by_pieces_d::run): Likewise.
19163         (can_store_by_pieces): Likewise.
19165 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19166             Alan Hayward  <alan.hayward@arm.com>
19167             David Sherwood  <david.sherwood@arm.com>
19169         * combine.c (extract_left_shift): Add a mode argument and update
19170         recursive calls.
19171         (make_compound_operation_int): Change the type of the mode parameter
19172         to scalar_int_mode and update the call to extract_left_shift.
19174 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19175             Alan Hayward  <alan.hayward@arm.com>
19176             David Sherwood  <david.sherwood@arm.com>
19178         * combine.c (simplify_and_const_int): Change the type of the mode
19179         parameter to scalar_int_mode.
19180         (simplify_and_const_int_1): Likewise.  Update recursive call.
19182 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19183             Alan Hayward  <alan.hayward@arm.com>
19184             David Sherwood  <david.sherwood@arm.com>
19186         * combine.c (simplify_compare_const): Check that the mode is a
19187         scalar_int_mode (rather than VOIDmode) before testing its
19188         precision.
19189         (simplify_comparison): Move COMPARISON_P handling out of the
19190         loop and restrict the latter part of the loop to scalar_int_modes.
19191         Check is_a <scalar_int_mode> before calling HWI_COMPUTABLE_MODE_P
19192         and when considering SUBREG_REGs.  Use is_int_mode instead of
19193         checking GET_MODE_CLASS against MODE_INT.
19195 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19196             Alan Hayward  <alan.hayward@arm.com>
19197             David Sherwood  <david.sherwood@arm.com>
19199         * combine.c (try_widen_shift_mode): Move check for equal modes to...
19200         (simplify_shift_const_1): ...here.  Use scalar_int_mode for
19201         shift_unit_mode and for modes involved in scalar shifts.
19203 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19204             Alan Hayward  <alan.hayward@arm.com>
19205             David Sherwood  <david.sherwood@arm.com>
19207         * combine.c (force_int_to_mode): New function, split out from...
19208         (force_to_mode): ...here.  Keep xmode up-to-date and use it
19209         instead of GET_MODE (x).
19211 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19212             Alan Hayward  <alan.hayward@arm.com>
19213             David Sherwood  <david.sherwood@arm.com>
19215         * optabs-query.h (extraction_insn::struct_mode): Change type to
19216         opt_scalar_int_mode and update comment.
19217         (extraction_insn::field_mode): Change type to scalar_int_mode.
19218         (extraction_insn::pos_mode): Likewise.
19219         * combine.c (make_extraction): Update accordingly.
19220         * optabs-query.c (get_traditional_extraction_insn): Likewise.
19221         (get_optab_extraction_insn): Likewise.
19222         * recog.c (simplify_while_replacing): Likewise.
19223         * expmed.c (narrow_bit_field_mem): Change the type of the mode
19224         parameter to opt_scalar_int_mode.
19226 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19227             Alan Hayward  <alan.hayward@arm.com>
19228             David Sherwood  <david.sherwood@arm.com>
19230         * machmode.h (bit_field_mode_iterator::next_mode): Take a pointer
19231         to a scalar_int_mode instead of a machine_mode.
19232         (bit_field_mode_iterator::m_mode): Change type to opt_scalar_int_mode.
19233         (get_best_mode): Return a boolean and use a pointer argument to store
19234         the selected mode.  Replace the limit mode parameter with a bit limit.
19235         * expmed.c (adjust_bit_field_mem_for_reg): Use scalar_int_mode
19236         for the values returned by bit_field_mode_iterator::next_mode.
19237         (store_bit_field): Update call to get_best_mode.
19238         (store_fixed_bit_field): Likewise.
19239         (extract_fixed_bit_field): Likewise.
19240         * expr.c (optimize_bitfield_assignment_op): Likewise.
19241         * fold-const.c (optimize_bit_field_compare): Likewise.
19242         (fold_truth_andor_1): Likewise.
19243         * stor-layout.c (bit_field_mode_iterator::next_mode): As above.
19244         Update for new type of m_mode.
19245         (get_best_mode): As above.
19247 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19248             Alan Hayward  <alan.hayward@arm.com>
19249             David Sherwood  <david.sherwood@arm.com>
19251         * expmed.c (strict_volatile_bitfield_p): Change the type of fieldmode
19252         to scalar_int_mode.  Remove check for SCALAR_INT_MODE_P.
19253         (store_bit_field): Check is_a <scalar_int_mode> before calling
19254         strict_volatile_bitfield_p.
19255         (extract_bit_field): Likewise.
19257 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19258             Alan Hayward  <alan.hayward@arm.com>
19259             David Sherwood  <david.sherwood@arm.com>
19261         * target.def (cstore_mode): Return a scalar_int_mode.
19262         * doc/tm.texi: Regenerate.
19263         * config/sparc/sparc.c (sparc_cstore_mode): Return a scalar_int_mode.
19264         * targhooks.h (default_cstore_mode): Likewise.
19265         * targhooks.c (default_cstore_mode): Likewise, using a forced
19266         conversion.
19267         * expmed.c (emit_cstore): Expect the target of the cstore to be
19268         a scalar_int_mode.
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         * cfgloop.h (rtx_iv): Change type of extend_mode and mode to
19275         scalar_int_mode.
19276         (niter_desc): Likewise mode.
19277         (iv_analyze): Add a mode parameter.
19278         (biv_p): Likewise.
19279         (iv_analyze_expr): Pass the mode paraeter before the rtx it describes
19280         and change its type to scalar_int_mode.
19281         * loop-iv.c: Update commentary at head of file.
19282         (iv_constant): Pass the mode paraeter before the rtx it describes
19283         and change its type to scalar_int_mode.  Remove VOIDmode handling.
19284         (iv_subreg): Change the type of the mode parameter to scalar_int_mode.
19285         (iv_extend): Likewise.
19286         (shorten_into_mode): Likewise.
19287         (iv_add): Use scalar_int_mode.
19288         (iv_mult): Likewise.
19289         (iv_shift): Likewise.
19290         (canonicalize_iv_subregs): Likewise.
19291         (get_biv_step_1): Pass the outer_mode parameter before the rtx
19292         it describes and change its mode to scalar_int_mode.   Also change
19293         the type of the returned inner_mode to scalar_int_mode.
19294         (get_biv_step): Likewise, turning outer_mode from a pointer
19295         into a direct parameter.  Update call to get_biv_step_1.
19296         (iv_analyze_biv): Add an outer_mode parameter.  Update calls to
19297         iv_constant and get_biv_step.
19298         (iv_analyze_expr): Pass the mode parameter before the rtx it describes
19299         and change its type to scalar_int_mode.  Don't initialise iv->mode
19300         to VOIDmode and remove later checks for its still being VOIDmode.
19301         Update calls to iv_analyze_op and iv_analyze_expr.  Check
19302         is_a <scalar_int_mode> when changing the mode under consideration.
19303         (iv_analyze_def): Ignore registers that don't have a scalar_int_mode.
19304         Update call to iv_analyze_expr.
19305         (iv_analyze_op): Add a mode parameter.  Reject subregs whose
19306         inner register is not also a scalar_int_mode.  Update call to
19307         iv_analyze_biv.
19308         (iv_analyze): Add a mode parameter.  Update call to iv_analyze_op.
19309         (biv_p): Add a mode parameter.  Update call to iv_analyze_biv.
19310         (iv_number_of_iterations): Use is_a <scalar_int_mode> instead of
19311         separate mode class checks.  Update calls to iv_analyze.  Remove
19312         fix-up of VOIDmodes after iv_analyze_biv.
19313         * loop-unroll.c (analyze_iv_to_split_insn): Reject registers that
19314         don't have a scalar_int_mode.  Update call to biv_p.
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         * cfgexpand.c (convert_debug_memory_address): Use
19321         as_a <scalar_int_mode>.
19322         * combine.c (expand_compound_operation): Likewise.
19323         (make_extraction): Likewise.
19324         (change_zero_ext): Likewise.
19325         (simplify_comparison): Likewise.
19326         * cse.c (cse_insn): Likewise.
19327         * dwarf2out.c (minmax_loc_descriptor): Likewise.
19328         (mem_loc_descriptor): Likewise.
19329         (loc_descriptor): Likewise.
19330         * expmed.c (init_expmed_one_mode): Likewise.
19331         (synth_mult): Likewise.
19332         (emit_store_flag_1): Likewise.
19333         (expand_divmod): Likewise.  Use HWI_COMPUTABLE_MODE_P instead
19334         of a comparison with size.
19335         * expr.c (expand_assignment): Use as_a <scalar_int_mode>.
19336         (reduce_to_bit_field_precision): Likewise.
19337         * function.c (expand_function_end): Likewise.
19338         * internal-fn.c (expand_arith_overflow_result_store): Likewise.
19339         * loop-doloop.c (doloop_modify): Likewise.
19340         * optabs.c (expand_binop): Likewise.
19341         (expand_unop): Likewise.
19342         (expand_copysign_absneg): Likewise.
19343         (prepare_cmp_insn): Likewise.
19344         (maybe_legitimize_operand): Likewise.
19345         * recog.c (const_scalar_int_operand): Likewise.
19346         * rtlanal.c (get_address_mode): Likewise.
19347         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19348         (simplify_cond_clz_ctz): Likewise.
19349         * tree-nested.c (get_nl_goto_field): Likewise.
19350         * tree.c (build_vector_type_for_mode): Likewise.
19351         * var-tracking.c (use_narrower_mode): Likewise.
19353 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19354             Alan Hayward  <alan.hayward@arm.com>
19355             David Sherwood  <david.sherwood@arm.com>
19357         * tree.h (SCALAR_INT_TYPE_MODE): New macro.
19358         * builtins.c (expand_builtin_signbit): Use it.
19359         * cfgexpand.c (expand_debug_expr): Likewise.
19360         * dojump.c (do_jump): Likewise.
19361         (do_compare_and_jump): Likewise.
19362         * dwarf2cfi.c (expand_builtin_init_dwarf_reg_sizes): Likewise.
19363         * expmed.c (make_tree): Likewise.
19364         * expr.c (expand_expr_real_2): Likewise.
19365         (expand_expr_real_1): Likewise.
19366         (try_casesi): Likewise.
19367         * fold-const-call.c (fold_const_call_ss): Likewise.
19368         * fold-const.c (unextend): Likewise.
19369         (extract_muldiv_1): Likewise.
19370         (fold_single_bit_test): Likewise.
19371         (native_encode_int): Likewise.
19372         (native_encode_string): Likewise.
19373         (native_interpret_int): Likewise.
19374         * gimple-fold.c (gimple_fold_builtin_memset): Likewise.
19375         * internal-fn.c (expand_addsub_overflow): Likewise.
19376         (expand_neg_overflow): Likewise.
19377         (expand_mul_overflow): Likewise.
19378         (expand_arith_overflow): Likewise.
19379         * match.pd: Likewise.
19380         * stor-layout.c (layout_type): Likewise.
19381         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
19382         * tree-ssa-math-opts.c (convert_mult_to_widen): Likewise.
19383         * tree-ssanames.c (get_range_info): Likewise.
19384         * tree-switch-conversion.c (array_value_type) Likewise.
19385         * tree-vect-patterns.c (vect_recog_rotate_pattern): Likewise.
19386         (vect_recog_divmod_pattern): Likewise.
19387         (vect_recog_mixed_size_cond_pattern): Likewise.
19388         * tree-vrp.c (extract_range_basic): Likewise.
19389         (simplify_float_conversion_using_ranges): Likewise.
19390         * tree.c (int_fits_type_p): Likewise.
19391         * ubsan.c (instrument_bool_enum_load): Likewise.
19392         * varasm.c (mergeable_string_section): Likewise.
19393         (narrowing_initializer_constant_valid_p): Likewise.
19394         (output_constant): Likewise.
19396 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19397             Alan Hayward  <alan.hayward@arm.com>
19398             David Sherwood  <david.sherwood@arm.com>
19400         * machmode.h (NARROWEST_INT_MODE): New macro.
19401         * expr.c (alignment_for_piecewise_move): Use it instead of
19402         GET_CLASS_NARROWEST_MODE (MODE_INT).
19403         (push_block): Likewise.
19404         * stor-layout.c (bit_field_mode_iterator::bit_field_mode_iterator):
19405         Likewise.
19406         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19408 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19409             Alan Hayward  <alan.hayward@arm.com>
19410             David Sherwood  <david.sherwood@arm.com>
19412         * postreload.c (move2add_valid_value_p): Change the type of the
19413         mode parameter to scalar_int_mode.
19414         (move2add_use_add2_insn): Add a mode parameter and use it instead
19415         of GET_MODE (reg).
19416         (move2add_use_add3_insn): Likewise.
19417         (reload_cse_move2add): Update accordingly.
19419 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19420             Alan Hayward  <alan.hayward@arm.com>
19421             David Sherwood  <david.sherwood@arm.com>
19423         * expr.c (expand_expr_real_2): Use scalar_int_mode for the
19424         double-word mode.
19425         * lower-subreg.c (resolve_shift_zext): Use is_a <scalar_int_mode>.
19426         * optabs.c (expand_unop): Likewise.
19428 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19429             Alan Hayward  <alan.hayward@arm.com>
19430             David Sherwood  <david.sherwood@arm.com>
19432         * dwarf2out.c (typed_binop): Change mode parameter to scalar_int_mode.
19433         (clz_loc_descriptor): Likewise.  Remove SCALAR_INT_MODE_P check.
19434         (popcount_loc_descriptor): Likewise.
19435         (bswap_loc_descriptor): Likewise.
19436         (rotate_loc_descriptor): Likewise.
19437         (mem_loc_descriptor): Add is_a <scalar_int_mode> checks before
19438         calling the functions above.
19440 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19441             Alan Hayward  <alan.hayward@arm.com>
19442             David Sherwood  <david.sherwood@arm.com>
19444         * combine.c (sign_extend_short_imm): Add is_a <scalar_int_mode>
19445         checks.
19446         (try_combine): Likewise.
19447         (simplify_if_then_else): Likewise.
19448         * cse.c (cse_insn): Likewise.
19449         * dwarf2out.c (mem_loc_descriptor): Likewise.
19450         * emit-rtl.c (gen_lowpart_common): Likewise.
19451         * simplify-rtx.c (simplify_truncation): Likewise.
19452         (simplify_binary_operation_1): Likewise.
19453         (simplify_const_relational_operation): Likewise.
19454         (simplify_ternary_operation): Likewise.
19455         * tree-ssa-loop-ivopts.c (force_expr_to_var_cost): Likewise.
19457 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19458             Alan Hayward  <alan.hayward@arm.com>
19459             David Sherwood  <david.sherwood@arm.com>
19461         * cse.c (cse_insn): Add is_a <scalar_int_mode> checks.
19462         * reload.c (push_reload): Likewise.
19463         (find_reloads): Likewise.
19465 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19466             Alan Hayward  <alan.hayward@arm.com>
19467             David Sherwood  <david.sherwood@arm.com>
19469         * combine.c (find_split_point): Add is_a <scalar_int_mode> checks.
19470         (make_compound_operation_int): Likewise.
19471         (change_zero_ext): Likewise.
19472         * expr.c (convert_move): Likewise.
19473         (convert_modes): Likewise.
19474         * fwprop.c (forward_propagate_subreg): Likewise.
19475         * loop-iv.c (get_biv_step_1): Likewise.
19476         * optabs.c (widen_operand): Likewise.
19477         * postreload.c (move2add_valid_value_p): Likewise.
19478         * recog.c (simplify_while_replacing): Likewise.
19479         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19480         (simplify_binary_operation_1): Likewise.  Remove redundant
19481         mode equality check.
19483 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19484             Alan Hayward  <alan.hayward@arm.com>
19485             David Sherwood  <david.sherwood@arm.com>
19487         * combine.c (combine_simplify_rtx): Add checks for
19488         is_a <scalar_int_mode>.
19489         (simplify_if_then_else): Likewise.
19490         (make_field_assignment): Likewise.
19491         (simplify_comparison): Likewise.
19492         * ifcvt.c (noce_try_bitop): Likewise.
19493         * loop-invariant.c (canonicalize_address_mult): Likewise.
19494         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19496 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19497             Alan Hayward  <alan.hayward@arm.com>
19498             David Sherwood  <david.sherwood@arm.com>
19500         * gimple-fold.c (gimple_fold_builtin_memory_op): Use
19501         is_a <scalar_int_mode> instead of != BLKmode.
19503 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19504             Alan Hayward  <alan.hayward@arm.com>
19505             David Sherwood  <david.sherwood@arm.com>
19507         * cfgexpand.c (expand_debug_expr): Use is_a <scalar_int_mode>
19508         instead of != VOIDmode.
19509         * combine.c (if_then_else_cond): Likewise.
19510         (change_zero_ext): Likewise.
19511         * dwarf2out.c (mem_loc_descriptor): Likewise.
19512         (loc_descriptor): Likewise.
19513         * rtlanal.c (canonicalize_condition): Likewise.
19514         * simplify-rtx.c (simplify_relational_operation_1): 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         * simplify-rtx.c (simplify_binary_operation_1): Use
19521         is_a <scalar_int_mode> instead of !VECTOR_MODE_P.
19523 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19524             Alan Hayward  <alan.hayward@arm.com>
19525             David Sherwood  <david.sherwood@arm.com>
19527         * wide-int.h (int_traits<unsigned char>) New class.
19528         (int_traits<unsigned short>) Likewise.
19529         * cfgexpand.c (expand_debug_expr): Use is_a <scalar_int_mode>.
19530         Use GET_MODE_UNIT_PRECISION and remove redundant test for
19531         SCALAR_INT_MODE_P.
19532         * combine.c (set_nonzero_bits_and_sign_copies): Use
19533         is_a <scalar_int_mode>.
19534         (find_split_point): Likewise.
19535         (combine_simplify_rtx): Likewise.
19536         (simplify_logical): Likewise.
19537         (expand_compound_operation): Likewise.
19538         (expand_field_assignment): Likewise.
19539         (make_compound_operation): Likewise.
19540         (extended_count): Likewise.
19541         (change_zero_ext): Likewise.
19542         (simplify_comparison): Likewise.
19543         * dwarf2out.c (scompare_loc_descriptor): Likewise.
19544         (ucompare_loc_descriptor): Likewise.
19545         (minmax_loc_descriptor): Likewise.
19546         (mem_loc_descriptor): Likewise.
19547         (loc_descriptor): Likewise.
19548         * expmed.c (init_expmed_one_mode): Likewise.
19549         * lra-constraints.c (lra_constraint_offset): Likewise.
19550         * optabs.c (prepare_libcall_arg): Likewise.
19551         * postreload.c (move2add_note_store): Likewise.
19552         * reload.c (operands_match_p): Likewise.
19553         * rtl.h (load_extend_op): Likewise.
19554         * rtlhooks.c (gen_lowpart_general): Likewise.
19555         * simplify-rtx.c (simplify_truncation): Likewise.
19556         (simplify_unary_operation_1): Likewise.
19557         (simplify_binary_operation_1): Likewise.
19558         (simplify_const_binary_operation): Likewise.
19559         (simplify_const_relational_operation): Likewise.
19560         (simplify_subreg): Likewise.
19561         * stor-layout.c (bitwise_mode_for_mode): Likewise.
19562         * var-tracking.c (adjust_mems): Likewise.
19563         (prepare_call_arguments): Likewise.
19565 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19566             Alan Hayward  <alan.hayward@arm.com>
19567             David Sherwood  <david.sherwood@arm.com>
19569         * machmode.h (is_int_mode): New fuction.
19570         * combine.c (find_split_point): Use it.
19571         (combine_simplify_rtx): Likewise.
19572         (simplify_if_then_else): Likewise.
19573         (simplify_set): Likewise.
19574         (simplify_shift_const_1): Likewise.
19575         (simplify_comparison): Likewise.
19576         * config/aarch64/aarch64.c (aarch64_rtx_costs): Likewise.
19577         * cse.c (notreg_cost): Likewise.
19578         (cse_insn): Likewise.
19579         * cselib.c (cselib_lookup_1): Likewise.
19580         * dojump.c (do_jump_1): Likewise.
19581         (do_compare_rtx_and_jump): Likewise.
19582         * dse.c (get_call_args): Likewise.
19583         * dwarf2out.c (rtl_for_decl_init): Likewise.
19584         (native_encode_initializer): Likewise.
19585         * expmed.c (emit_store_flag_1): Likewise.
19586         (emit_store_flag): Likewise.
19587         * expr.c (convert_modes): Likewise.
19588         (store_field): Likewise.
19589         (expand_expr_real_1): Likewise.
19590         * fold-const.c (fold_read_from_constant_string): Likewise.
19591         * gimple-ssa-sprintf.c (get_format_string): Likewise.
19592         * optabs-libfuncs.c (gen_int_libfunc): Likewise.
19593         * optabs.c (expand_binop): Likewise.
19594         (expand_unop): Likewise.
19595         (expand_abs_nojump): Likewise.
19596         (expand_one_cmpl_abs_nojump): Likewise.
19597         * simplify-rtx.c (mode_signbit_p): Likewise.
19598         (val_signbit_p): Likewise.
19599         (val_signbit_known_set_p): Likewise.
19600         (val_signbit_known_clear_p): Likewise.
19601         (simplify_relational_operation_1): Likewise.
19602         * tree.c (vector_type_mode): Likewise.
19604 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19605             Alan Hayward  <alan.hayward@arm.com>
19606             David Sherwood  <david.sherwood@arm.com>
19608         * machmode.h (smallest_mode_for_size): Fix formatting.
19609         (smallest_int_mode_for_size): New function.
19610         * cfgexpand.c (expand_debug_expr): Use smallest_int_mode_for_size
19611         instead of smallest_mode_for_size.
19612         * combine.c (make_extraction): Likewise.
19613         * config/arc/arc.c (arc_expand_movmem): Likewise.
19614         * config/arm/arm.c (arm_expand_divmod_libfunc): Likewise.
19615         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
19616         * config/s390/s390.c (s390_expand_insv): Likewise.
19617         * config/sparc/sparc.c (assign_int_registers): Likewise.
19618         * config/spu/spu.c (spu_function_value): Likewise.
19619         (spu_function_arg): Likewise.
19620         * coverage.c (get_gcov_type): Likewise.
19621         (get_gcov_unsigned_t): Likewise.
19622         * dse.c (find_shift_sequence): Likewise.
19623         * expmed.c (store_bit_field_1): Likewise.
19624         * expr.c (convert_move): Likewise.
19625         (store_field): Likewise.
19626         * internal-fn.c (expand_arith_overflow): Likewise.
19627         * optabs-query.c (get_best_extraction_insn): Likewise.
19628         * optabs.c (expand_twoval_binop_libfunc): Likewise.
19629         * stor-layout.c (layout_type): Likewise.
19630         (initialize_sizetypes): Likewise.
19631         * targhooks.c (default_get_mask_mode): Likewise.
19632         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Likewise.
19634 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19635             Alan Hayward  <alan.hayward@arm.com>
19636             David Sherwood  <david.sherwood@arm.com>
19638         * machmode.h (opt_mode::else_blk): New function.
19639         (int_mode_for_mode): Declare.
19640         * stor-layout.c (int_mode_for_mode): Return an opt_scalar_int_mode.
19641         * builtins.c (expand_builtin_signbit): Adjust for new int_mode_for_mode
19642         return type.
19643         * cfgexpand.c (expand_debug_expr): Likewise.
19644         * combine.c (gen_lowpart_or_truncate): Likewise.
19645         (gen_lowpart_for_combine): Likewise.
19646         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Likewise.
19647         * config/avr/avr.c (avr_to_int_mode): Likewise.
19648         (avr_out_plus_1): Likewise.
19649         (avr_out_plus): Likewise.
19650         (avr_out_round): Likewise.
19651         * config/i386/i386.c (ix86_split_to_parts): Likewise.
19652         * config/s390/s390.c (s390_expand_vec_compare_cc): Likewise.
19653         (s390_expand_vcond): Likewise.
19654         * config/spu/spu.c (spu_split_immediate): Likewise.
19655         (spu_expand_mov): Likewise.
19656         * dse.c (get_stored_val): Likewise.
19657         * expmed.c (store_bit_field_1): Likewise.
19658         (convert_extracted_bit_field): Use int_mode_for_mode instead of
19659         int_mode_for_size.
19660         (extract_bit_field_1): Adjust for new int_mode_for_mode return type.
19661         (extract_low_bits): Likewise.
19662         * expr.c (emit_group_load_1): Likewise.  Separate out the BLKmode
19663         handling rather than repeating the check.
19664         (emit_group_store): Likewise.
19665         (emit_move_via_integer): Adjust for new int_mode_for_mode return type.
19666         * optabs.c (expand_absneg_bit): Likewise.
19667         (expand_copysign_absneg): Likewise.
19668         (expand_copysign_bit): Likewise.
19669         * tree-if-conv.c (ifcvt_can_use_mask_load_store): Likewise.
19670         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19671         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
19672         * var-tracking.c (prepare_call_arguments):  Likewise.
19673         * config/powerpcspe/powerpcspe.c (rs6000_do_expand_vec_perm): Use
19674         int_mode_for_mode instead of mode_for_size.
19675         * config/rs6000/rs6000.c (rs6000_do_expand_vec_perm): Likewise.
19677 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19678             Alan Hayward  <alan.hayward@arm.com>
19679             David Sherwood  <david.sherwood@arm.com>
19681         * machmode.h (int_mode_for_size): New function.
19682         * builtins.c (set_builtin_user_assembler_name): Use int_mode_for_size
19683         instead of mode_for_size.
19684         * calls.c (save_fixed_argument_area): Likewise.  Make use of BLKmode
19685         explicit.
19686         * combine.c (expand_field_assignment): Use int_mode_for_size
19687         instead of mode_for_size.
19688         (make_extraction): Likewise.
19689         (simplify_shift_const_1): Likewise.
19690         (simplify_comparison): Likewise.
19691         * dojump.c (do_jump): Likewise.
19692         * dwarf2out.c (mem_loc_descriptor): Likewise.
19693         * emit-rtl.c (init_derived_machine_modes): Likewise.
19694         * expmed.c (flip_storage_order): Likewise.
19695         (convert_extracted_bit_field): Likewise.
19696         * expr.c (copy_blkmode_from_reg): Likewise.
19697         * graphite-isl-ast-to-gimple.c (max_mode_int_precision): Likewise.
19698         * internal-fn.c (expand_mul_overflow): Likewise.
19699         * lower-subreg.c (simple_move): Likewise.
19700         * optabs-libfuncs.c (init_optabs): Likewise.
19701         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19702         * tree.c (vector_type_mode): Likewise.
19703         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
19704         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Likewise.
19705         * tree-vect-generic.c (expand_vector_parallel): Likewise.
19706         * tree-vect-stmts.c (vectorizable_load): Likewise.
19707         (vectorizable_store): Likewise.
19709 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19710             Alan Hayward  <alan.hayward@arm.com>
19711             David Sherwood  <david.sherwood@arm.com>
19713         * coretypes.h (pod_mode): New type.
19714         (scalar_int_mode_pod): New typedef.
19715         * machmode.h (pod_mode): New class.
19716         (int_n_data_t::m): Change type to scalar_int_mode_pod.
19717         * genmodes.c (emit_mode_int_n): Update accordingly.
19718         * lower-subreg.h (target_lower_subreg): Change type to
19719         scalar_int_mode_pod.
19720         * gdbhooks.py (build_pretty_printer): Handle pod_mode and
19721         scalar_int_mode_pod.
19723 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19724             Alan Hayward  <alan.hayward@arm.com>
19725             David Sherwood  <david.sherwood@arm.com>
19727         * config/powerpcspe/powerpcspe.h (rs6000_pmode): Change type from
19728         machine_mode to scalar_int_mode.
19729         * config/powerpcspe/powerpcspe.c (rs6000_pmode): Likewise.
19730         (rs6000_option_override_internal): Remove cast to int.
19731         * config/rs6000/rs6000.h (rs6000_pmode): Change type from
19732         machine_mode to scalar_int_mode.
19733         * config/rs6000/rs6000.c (rs6000_pmode): Likewise.
19734         (rs6000_option_override_internal): Remove cast to int.
19735         * config/s390/s390.h (Pmode): Remove cast to machine_mode.
19736         * config/epiphany/epiphany.h (RTX_OK_FOR_OFFSET_P): Add cast
19737         to machine_mode.
19738         * config/s390/s390.c (s390_expand_builtin): Likewise.
19739         * coretypes.h (scalar_int_mode): New type.
19740         (opt_scalar_int_mode): New typedef.
19741         * machmode.h (scalar_int_mode): New class.
19742         (scalar_int_mode::includes_p): New function.
19743         (byte_mode): Change type to scalar_int_mode.
19744         (word_mode): Likewise.
19745         (ptr_mode): Likewise.
19746         * emit-rtl.c (byte_mode): Likewise.
19747         (word_mode): Likewise.
19748         (ptr_mode): Likewise.
19749         (init_derived_machine_modes): Update accordingly.
19750         * genmodes.c (get_mode_class): Return scalar_int_mode for MODE_INT
19751         and MODE_PARTIAL_INT.
19752         * gdbhooks.py (build_pretty_printer): Handle scalar_int_mode and
19753         opt_scalar_int_mode.
19755 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19756             Alan Hayward  <alan.hayward@arm.com>
19757             David Sherwood  <david.sherwood@arm.com>
19759         * target.def (libgcc_floating_mode_supported_p): Take a
19760         scalar_float_mode.
19761         * doc/tm.texi: Regenerate.
19762         * targhooks.h (default_libgcc_floating_mode_supported_p): Take a
19763         scalar_float_mode.
19764         * targhooks.c (default_libgcc_floating_mode_supported_p): Likewise.
19765         * config/aarch64/aarch64.c (aarch64_libgcc_floating_mode_supported_p):
19766         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         * target.def (default_floatn_mode): Return an opt_scalar_float_mode.
19773         * doc/tm.texi: Regenerate.
19774         * config/arm/arm.c (arm_floatn_mode): Return an opt_scalar_float_mode.
19775         * config/powerpcspe/powerpcspe.c (rs6000_floatn_mode): Likewise.
19776         * config/rs6000/rs6000.c (rs6000_floatn_mode): Likewise.
19777         * targhooks.h (default_floatn_mode): Likewise.
19778         * targhooks.c (default_floatn_mode): Likewise.
19779         * tree.c (build_common_tree_nodes): Update accordingly.
19781 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19782             Alan Hayward  <alan.hayward@arm.com>
19783             David Sherwood  <david.sherwood@arm.com>
19785         * machmode.h (mode_iterator::start): Provide overload for opt_modes.
19786         (mode_iterator::iterate_p): Likewise.
19787         (mode_iterator::get_wider): Likewise.
19788         * expr.c (init_expr_target): Use opt_scalar_float_mode.
19790 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19791             Alan Hayward  <alan.hayward@arm.com>
19792             David Sherwood  <david.sherwood@arm.com>
19794         * coretypes.h (opt_scalar_float_mode): New typedef.
19795         * machmode.h (float_mode_for_size): New function.
19796         * emit-rtl.c (double_mode): Delete.
19797         (init_emit_once): Use float_mode_for_size.
19798         * stor-layout.c (layout_type): Likewise.
19799         * gdbhooks.py (build_pretty_printer): Handle opt_scalar_float_mode.
19801 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19802             Alan Hayward  <alan.hayward@arm.com>
19803             David Sherwood  <david.sherwood@arm.com>
19805         * output.h (assemble_real): Take a scalar_float_mode.
19806         * config/arm/arm.c (arm_assemble_integer): Update accordingly.
19807         * config/arm/arm.md (consttable_4): Likewise.
19808         (consttable_8): Likewise.
19809         (consttable_16): Likewise.
19810         * config/mips/mips.md (consttable_float): Likewise.
19811         * config/s390/s390.c (s390_output_pool_entry): Likewise.
19812         * varasm.c (assemble_real): Take a scalar_float_mode.
19813         (output_constant_pool_2): Update accordingly.
19814         (output_constant): Likewise.
19816 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19817             Alan Hayward  <alan.hayward@arm.com>
19818             David Sherwood  <david.sherwood@arm.com>
19820         * tree.h (SCALAR_FLOAT_TYPE_MODE): New macro.
19821         * builtins.c (expand_builtin_signbit): Use it instead of TYPE_MODE.
19822         * fold-const.c (fold_convert_const_real_from_fixed): Likewise.
19823         (native_encode_real): Likewise.
19824         (native_interpret_real): Likewise.
19825         * hsa-brig.c (emit_immediate_scalar_to_buffer): Likewise.
19826         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19828 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19829             Alan Hayward  <alan.hayward@arm.com>
19830             David Sherwood  <david.sherwood@arm.com>
19832         * optabs-libfuncs.c (gen_trunc_conv_libfunc): Use is_a
19833         <scalar_float_mode>.  Simplify.
19834         (gen_extend_conv_libfunc): Likewise.
19836 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19837             Alan Hayward  <alan.hayward@arm.com>
19838             David Sherwood  <david.sherwood@arm.com>
19840         * coretypes.h (scalar_float_mode): New type.
19841         * machmode.h (mode_traits::from_int): Use machine_mode if
19842         USE_ENUM_MODES is defined.
19843         (is_a): New function.
19844         (as_a): Likewise.
19845         (dyn_cast): Likewise.
19846         (scalar_float_mode): New class.
19847         (scalar_float_mode::includes_p): New function.
19848         (is_float_mode): Likewise.
19849         * gdbhooks.py (MachineModePrinter): New class.
19850         (build_pretty_printer): Use it for scalar_float_mode.
19851         * real.h (FLOAT_MODE_FORMAT): Use as_a <scalar_float_mode>.
19852         (format_helper::format_helper): Turn into a template.
19853         * genmodes.c (get_mode_class): New function.
19854         (emit_insn_modes_h): Give modes the class returned by get_mode_class,
19855         or machine_mode if none.
19856         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Use
19857         as_a <scalar_float_mode>.
19858         * dwarf2out.c (mem_loc_descriptor): Likewise.
19859         (insert_float): Likewise.
19860         (add_const_value_attribute): Likewise.
19861         * simplify-rtx.c (simplify_immed_subreg): Likewise.
19862         * optabs.c (expand_absneg_bit): Take a scalar_float_mode.
19863         (expand_unop): Update accordingly.
19864         (expand_abs_nojump): Likewise.
19865         (expand_copysign_absneg): Take a scalar_float_mode.
19866         (expand_copysign_bit): Likewise.
19867         (expand_copysign): Update accordingly.
19869 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19870             Alan Hayward  <alan.hayward@arm.com>
19871             David Sherwood  <david.sherwood@arm.com>
19873         * coretypes.h (opt_mode): New class.
19874         * machmode.h (opt_mode): Likewise.
19875         (opt_mode::else_void): New function.
19876         (opt_mode::require): Likewise.
19877         (opt_mode::exists): Likewise.
19878         (GET_MODE_WIDER_MODE): Turn into a function and return an opt_mode.
19879         (GET_MODE_2XWIDER_MODE): Likewise.
19880         (mode_iterator::get_wider): Update accordingly.
19881         (mode_iterator::get_2xwider): Likewise.
19882         (mode_iterator::get_known_wider): Likewise, turning into a template.
19883         * combine.c (make_extraction): Update use of GET_MODE_WIDER_MODE,
19884         forcing a wider mode to exist.
19885         * config/cr16/cr16.h (LONG_REG_P): Likewise.
19886         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
19887         * config/c6x/c6x.c (c6x_rtx_costs): Update use of
19888         GET_MODE_2XWIDER_MODE, forcing a wider mode to exist.
19889         * lower-subreg.c (init_lower_subreg): Likewise.
19890         * optabs-libfuncs.c (init_sync_libfuncs_1): Likewise, but not
19891         on the final iteration.
19892         * config/i386/i386.c (ix86_expand_set_or_movmem): Check whether
19893         a wider mode exists before asking for a move pattern.
19894         (get_mode_wider_vector): Update use of GET_MODE_WIDER_MODE,
19895         forcing a wider mode to exist.
19896         (expand_vselect_vconcat): Update use of GET_MODE_2XWIDER_MODE,
19897         returning false if no such mode exists.
19898         * config/ia64/ia64.c (expand_vselect_vconcat): Likewise.
19899         * config/mips/mips.c (mips_expand_vselect_vconcat): Likewise.
19900         * expmed.c (init_expmed_one_mode): Update use of GET_MODE_WIDER_MODE.
19901         Avoid checking for a MODE_INT if we already know the mode is not a
19902         SCALAR_INT_MODE_P.
19903         (extract_high_half): Update use of GET_MODE_WIDER_MODE,
19904         forcing a wider mode to exist.
19905         (expmed_mult_highpart_optab): Likewise.
19906         (expmed_mult_highpart): Likewise.
19907         * expr.c (expand_expr_real_2): Update use of GET_MODE_WIDER_MODE,
19908         using else_void.
19909         * lto-streamer-in.c (lto_input_mode_table): Likewise.
19910         * optabs-query.c (find_widening_optab_handler_and_mode): Likewise.
19911         * stor-layout.c (bit_field_mode_iterator::next_mode): Likewise.
19912         * internal-fn.c (expand_mul_overflow): Update use of
19913         GET_MODE_2XWIDER_MODE.
19914         * omp-low.c (omp_clause_aligned_alignment): Likewise.
19915         * tree-ssa-math-opts.c (convert_mult_to_widen): Update use of
19916         GET_MODE_WIDER_MODE.
19917         (convert_plusminus_to_widen): Likewise.
19918         * tree-switch-conversion.c (array_value_type): Likewise.
19919         * var-tracking.c (emit_note_insn_var_location): Likewise.
19920         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19921         Return false inside rather than outside the loop if no wider mode
19922         exists
19923         * optabs.c (expand_binop): Update use of GET_MODE_WIDER_MODE
19924         and GET_MODE_2XWIDER_MODE
19925         (can_compare_p): Use else_void.
19926         * gdbhooks.py (OptMachineModePrinter): New class.
19927         (build_pretty_printer): Use it for opt_mode.
19929 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19930             Alan Hayward  <alan.hayward@arm.com>
19931             David Sherwood  <david.sherwood@arm.com>
19933         * tree-switch-conversion.c (array_value_type): Only read TYPE_MODE
19934         once.  Use get_narrowest_mode instead of GET_CLASS_NARROWEST_MODE.
19936 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19937             Alan Hayward  <alan.hayward@arm.com>
19938             David Sherwood  <david.sherwood@arm.com>
19940         * machmode.h (mode_traits): New structure.
19941         (get_narrowest_mode): New function.
19942         (mode_iterator::start): Likewise.
19943         (mode_iterator::iterate_p): Likewise.
19944         (mode_iterator::get_wider): Likewise.
19945         (mode_iterator::get_known_wider): Likewise.
19946         (mode_iterator::get_2xwider): Likewise.
19947         (FOR_EACH_MODE_IN_CLASS): New mode iterator.
19948         (FOR_EACH_MODE): Likewise.
19949         (FOR_EACH_MODE_FROM): Likewise.
19950         (FOR_EACH_MODE_UNTIL): Likewise.
19951         (FOR_EACH_WIDER_MODE): Likewise.
19952         (FOR_EACH_2XWIDER_MODE): Likewise.
19953         * builtins.c (expand_builtin_strlen): Use new mode iterators.
19954         * combine.c (simplify_comparison): Likewise
19955         * config/i386/i386.c (type_natural_mode): Likewise.
19956         * cse.c (cse_insn): Likewise.
19957         * dse.c (find_shift_sequence): Likewise.
19958         * emit-rtl.c (init_derived_machine_modes): Likewise.
19959         (init_emit_once): Likewise.
19960         * explow.c (hard_function_value): Likewise.
19961         * expmed.c (extract_fixed_bit_field_1): Likewise.
19962         (extract_bit_field_1): Likewise.
19963         (expand_divmod): Likewise.
19964         (emit_store_flag_1): Likewise.
19965         * expr.c (init_expr_target): Likewise.
19966         (convert_move): Likewise.
19967         (alignment_for_piecewise_move): Likewise.
19968         (widest_int_mode_for_size): Likewise.
19969         (emit_block_move_via_movmem): Likewise.
19970         (copy_blkmode_to_reg): Likewise.
19971         (set_storage_via_setmem): Likewise.
19972         (compress_float_constant): Likewise.
19973         * omp-low.c (omp_clause_aligned_alignment): Likewise.
19974         * optabs-query.c (get_best_extraction_insn): Likewise.
19975         * optabs.c (expand_binop): Likewise.
19976         (expand_twoval_unop): Likewise.
19977         (expand_twoval_binop): Likewise.
19978         (widen_leading): Likewise.
19979         (widen_bswap): Likewise.
19980         (expand_parity): Likewise.
19981         (expand_unop): Likewise.
19982         (prepare_cmp_insn): Likewise.
19983         (prepare_float_lib_cmp): Likewise.
19984         (expand_float): Likewise.
19985         (expand_fix): Likewise.
19986         (expand_sfix_optab): Likewise.
19987         * postreload.c (move2add_use_add2_insn): Likewise.
19988         * reg-stack.c (reg_to_stack): Likewise.
19989         * reginfo.c (choose_hard_reg_mode): Likewise.
19990         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
19991         * stor-layout.c (mode_for_size): Likewise.
19992         (smallest_mode_for_size): Likewise.
19993         (mode_for_vector): Likewise.
19994         (finish_bitfield_representative): Likewise.
19995         * tree-ssa-math-opts.c (target_supports_divmod_p): Likewise.
19996         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
19997         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
19998         * var-tracking.c (prepare_call_arguments): Likewise.
20000 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20001             Alan Hayward  <alan.hayward@arm.com>
20002             David Sherwood  <david.sherwood@arm.com>
20004         * genconditions.c (write_header): Add a "#define USE_ENUM_MODES".
20005         * genmodes.c (emit_insn_modes_h): Define FOOmode to E_FOOmode if
20006         USE_ENUM_MODES is defined and to ((void) 0, E_FOOmode) otherwise.
20007         * machmode.h (mode_size): Move earlier in file.
20008         (mode_precision): Likewise.
20009         (mode_inner): Likewise.
20010         (mode_nunits): Likewise.
20011         (mode_unit_size): Likewise.
20012         (unit_unit_precision): Likewise.
20013         (mode_wider): Likewise.
20014         (mode_2xwider): Likewise.
20015         (machine_mode): New class.
20016         (mode_to_bytes): New function.
20017         (mode_to_bits): Likewise.
20018         (mode_to_precision): Likewise.
20019         (mode_to_inner): Likewise.
20020         (mode_to_unit_size): Likewise.
20021         (mode_to_unit_precision): Likewise.
20022         (mode_to_nunits): Likewise.
20023         (GET_MODE_SIZE): Use mode_to_bytes.
20024         (GET_MODE_BITSIZE): Use mode_to_bits.
20025         (GET_MODE_PRECISION): Use mode_to_precision.
20026         (GET_MODE_INNER): Use mode_to_inner.
20027         (GET_MODE_UNIT_SIZE): Use mode_to_unit_size.
20028         (GET_MODE_UNIT_PRECISION): Use mode_to_unit_precision.
20029         (GET_MODE_NUNITS): Use mode_to_nunits.
20030         * system.h (ALWAYS_INLINE): New macro.
20031         * config/powerpcspe/powerpcspe-c.c
20032         (altivec_resolve_overloaded_builtin): Use machine_mode instead of
20033         int for arg1_mode and arg2_mode.
20035 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20036             Alan Hayward  <alan.hayward@arm.com>
20037             David Sherwood  <david.sherwood@arm.com>
20039         * config/aarch64/aarch64-builtins.c (aarch64_simd_builtin_std_type):
20040         Prefix mode names with E_ in case statements.
20041         * config/aarch64/aarch64-elf.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20042         * config/aarch64/aarch64.c (aarch64_split_simd_combine): Likewise.
20043         (aarch64_split_simd_move): Likewise.
20044         (aarch64_gen_storewb_pair): Likewise.
20045         (aarch64_gen_loadwb_pair): Likewise.
20046         (aarch64_gen_store_pair): Likewise.
20047         (aarch64_gen_load_pair): Likewise.
20048         (aarch64_get_condition_code_1): Likewise.
20049         (aarch64_constant_pool_reload_icode): Likewise.
20050         (get_rsqrte_type): Likewise.
20051         (get_rsqrts_type): Likewise.
20052         (get_recpe_type): Likewise.
20053         (get_recps_type): Likewise.
20054         (aarch64_gimplify_va_arg_expr): Likewise.
20055         (aarch64_simd_container_mode): Likewise.
20056         (aarch64_emit_load_exclusive): Likewise.
20057         (aarch64_emit_store_exclusive): Likewise.
20058         (aarch64_expand_compare_and_swap): Likewise.
20059         (aarch64_gen_atomic_cas): Likewise.
20060         (aarch64_emit_bic): Likewise.
20061         (aarch64_emit_atomic_swap): Likewise.
20062         (aarch64_emit_atomic_load_op): Likewise.
20063         (aarch64_evpc_trn): Likewise.
20064         (aarch64_evpc_uzp): Likewise.
20065         (aarch64_evpc_zip): Likewise.
20066         (aarch64_evpc_ext): Likewise.
20067         (aarch64_evpc_rev): Likewise.
20068         (aarch64_evpc_dup): Likewise.
20069         (aarch64_gen_ccmp_first): Likewise.
20070         (aarch64_gen_ccmp_next): Likewise.
20071         * config/alpha/alpha.c (alpha_scalar_mode_supported_p): Likewise.
20072         (alpha_emit_xfloating_libcall): Likewise.
20073         (emit_insxl): Likewise.
20074         (alpha_arg_type): Likewise.
20075         * config/arc/arc.c (arc_vector_mode_supported_p): Likewise.
20076         (arc_preferred_simd_mode): Likewise.
20077         (arc_secondary_reload): Likewise.
20078         (get_arc_condition_code): Likewise.
20079         (arc_print_operand): Likewise.
20080         (arc_legitimate_constant_p): Likewise.
20081         * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20082         * config/arc/arc.md (casesi_load): Likewise.
20083         (casesi_compact_jump): Likewise.
20084         * config/arc/predicates.md (proper_comparison_operator): Likewise.
20085         (cc_use_register): Likewise.
20086         * config/arm/aout.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20087         * config/arm/arm-builtins.c (arm_simd_builtin_std_type): Likewise.
20088         (arm_init_iwmmxt_builtins): Likewise.
20089         * config/arm/arm.c (thumb1_size_rtx_costs): Likewise.
20090         (neon_expand_vector_init): Likewise.
20091         (arm_attr_length_move_neon): Likewise.
20092         (maybe_get_arm_condition_code): Likewise.
20093         (arm_emit_vector_const): Likewise.
20094         (arm_preferred_simd_mode): Likewise.
20095         (arm_output_iwmmxt_tinsr): Likewise.
20096         (thumb1_output_casesi): Likewise.
20097         (thumb2_output_casesi): Likewise.
20098         (arm_emit_load_exclusive): Likewise.
20099         (arm_emit_store_exclusive): Likewise.
20100         (arm_expand_compare_and_swap): Likewise.
20101         (arm_evpc_neon_vuzp): Likewise.
20102         (arm_evpc_neon_vzip): Likewise.
20103         (arm_evpc_neon_vrev): Likewise.
20104         (arm_evpc_neon_vtrn): Likewise.
20105         (arm_evpc_neon_vext): Likewise.
20106         (arm_validize_comparison): Likewise.
20107         * config/arm/neon.md (neon_vc<cmp_op><mode>): Likewise.
20108         * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Likewise.
20109         * config/avr/avr.c (avr_rtx_costs_1): Likewise.
20110         * config/c6x/c6x.c (c6x_vector_mode_supported_p): Likewise.
20111         (c6x_preferred_simd_mode): Likewise.
20112         * config/epiphany/epiphany.c (get_epiphany_condition_code): Likewise.
20113         (epiphany_rtx_costs): Likewise.
20114         * config/epiphany/predicates.md (proper_comparison_operator):
20115         Likewise.
20116         * config/frv/frv.c (condexec_memory_operand): Likewise.
20117         (frv_emit_move): Likewise.
20118         (output_move_single): Likewise.
20119         (output_condmove_single): Likewise.
20120         (frv_hard_regno_mode_ok): Likewise.
20121         (frv_matching_accg_mode): Likewise.
20122         * config/h8300/h8300.c (split_adds_subs): Likewise.
20123         (h8300_rtx_costs): Likewise.
20124         (h8300_print_operand): Likewise.
20125         (compute_mov_length): Likewise.
20126         (output_logical_op): Likewise.
20127         (compute_logical_op_length): Likewise.
20128         (compute_logical_op_cc): Likewise.
20129         (h8300_shift_needs_scratch_p): Likewise.
20130         (output_a_shift): Likewise.
20131         (compute_a_shift_length): Likewise.
20132         (compute_a_shift_cc): Likewise.
20133         (expand_a_rotate): Likewise.
20134         (output_a_rotate): Likewise.
20135         * config/i386/i386.c (classify_argument): Likewise.
20136         (function_arg_advance_32): Likewise.
20137         (function_arg_32): Likewise.
20138         (function_arg_64): Likewise.
20139         (function_value_64): Likewise.
20140         (ix86_gimplify_va_arg): Likewise.
20141         (ix86_legitimate_constant_p): Likewise.
20142         (put_condition_code): Likewise.
20143         (split_double_mode): Likewise.
20144         (ix86_avx256_split_vector_move_misalign): Likewise.
20145         (ix86_expand_vector_logical_operator): Likewise.
20146         (ix86_split_idivmod): Likewise.
20147         (ix86_expand_adjust_ufix_to_sfix_si): Likewise.
20148         (ix86_build_const_vector): Likewise.
20149         (ix86_build_signbit_mask): Likewise.
20150         (ix86_match_ccmode): Likewise.
20151         (ix86_cc_modes_compatible): Likewise.
20152         (ix86_expand_branch): Likewise.
20153         (ix86_expand_sse_cmp): Likewise.
20154         (ix86_expand_sse_movcc): Likewise.
20155         (ix86_expand_int_sse_cmp): Likewise.
20156         (ix86_expand_vec_perm_vpermi2): Likewise.
20157         (ix86_expand_vec_perm): Likewise.
20158         (ix86_expand_sse_unpack): Likewise.
20159         (ix86_expand_int_addcc): Likewise.
20160         (ix86_split_to_parts): Likewise.
20161         (ix86_vectorize_builtin_gather): Likewise.
20162         (ix86_vectorize_builtin_scatter): Likewise.
20163         (avx_vpermilp_parallel): Likewise.
20164         (inline_memory_move_cost): Likewise.
20165         (ix86_tieable_integer_mode_p): Likewise.
20166         (x86_maybe_negate_const_int): Likewise.
20167         (ix86_expand_vector_init_duplicate): Likewise.
20168         (ix86_expand_vector_init_one_nonzero): Likewise.
20169         (ix86_expand_vector_init_one_var): Likewise.
20170         (ix86_expand_vector_init_concat): Likewise.
20171         (ix86_expand_vector_init_interleave): Likewise.
20172         (ix86_expand_vector_init_general): Likewise.
20173         (ix86_expand_vector_set): Likewise.
20174         (ix86_expand_vector_extract): Likewise.
20175         (emit_reduc_half): Likewise.
20176         (ix86_emit_i387_round): Likewise.
20177         (ix86_mangle_type): Likewise.
20178         (ix86_expand_round_sse4): Likewise.
20179         (expand_vec_perm_blend): Likewise.
20180         (canonicalize_vector_int_perm): Likewise.
20181         (ix86_expand_vec_one_operand_perm_avx512): Likewise.
20182         (expand_vec_perm_1): Likewise.
20183         (expand_vec_perm_interleave3): Likewise.
20184         (expand_vec_perm_even_odd_pack): Likewise.
20185         (expand_vec_perm_even_odd_1): Likewise.
20186         (expand_vec_perm_broadcast_1): Likewise.
20187         (ix86_vectorize_vec_perm_const_ok): Likewise.
20188         (ix86_expand_vecop_qihi): Likewise.
20189         (ix86_expand_mul_widen_hilo): Likewise.
20190         (ix86_expand_sse2_abs): Likewise.
20191         (ix86_expand_pextr): Likewise.
20192         (ix86_expand_pinsr): Likewise.
20193         (ix86_preferred_simd_mode): Likewise.
20194         (ix86_simd_clone_compute_vecsize_and_simdlen): Likewise.
20195         * config/i386/sse.md (*andnot<mode>3): Likewise.
20196         (<mask_codefor><code><mode>3<mask_name>): Likewise.
20197         (*<code><mode>3): Likewise.
20198         * config/ia64/ia64.c (ia64_expand_vecint_compare): Likewise.
20199         (ia64_expand_atomic_op): Likewise.
20200         (ia64_arg_type): Likewise.
20201         (ia64_mode_to_int): Likewise.
20202         (ia64_scalar_mode_supported_p): Likewise.
20203         (ia64_vector_mode_supported_p): Likewise.
20204         (expand_vec_perm_broadcast): Likewise.
20205         * config/iq2000/iq2000.c (iq2000_move_1word): Likewise.
20206         (iq2000_function_arg_advance): Likewise.
20207         (iq2000_function_arg): Likewise.
20208         * config/m32c/m32c.c (m32c_preferred_reload_class): Likewise.
20209         * config/m68k/m68k.c (output_dbcc_and_branch): Likewise.
20210         (m68k_libcall_value): Likewise.
20211         (m68k_function_value): Likewise.
20212         (sched_attr_op_type): Likewise.
20213         * config/mcore/mcore.c (mcore_output_move): Likewise.
20214         * config/microblaze/microblaze.c (microblaze_function_arg_advance):
20215         Likewise.
20216         (microblaze_function_arg): Likewise.
20217         * config/mips/mips.c (mips16_build_call_stub): Likewise.
20218         (mips_print_operand): Likewise.
20219         (mips_mode_ok_for_mov_fmt_p): Likewise.
20220         (mips_vector_mode_supported_p): Likewise.
20221         (mips_preferred_simd_mode): Likewise.
20222         (mips_expand_vpc_loongson_even_odd): Likewise.
20223         (mips_expand_vec_unpack): Likewise.
20224         (mips_expand_vi_broadcast): Likewise.
20225         (mips_expand_vector_init): Likewise.
20226         (mips_expand_vec_reduc): Likewise.
20227         (mips_expand_msa_cmp): Likewise.
20228         * config/mips/mips.md (casesi_internal_mips16_<mode>): Likewise.
20229         * config/mn10300/mn10300.c (mn10300_print_operand): Likewise.
20230         (cc_flags_for_mode): Likewise.
20231         * config/msp430/msp430.c (msp430_print_operand): Likewise.
20232         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Likewise.
20233         (nds32_output_casesi_pc_relative): Likewise.
20234         * config/nds32/nds32.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20235         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Likewise.
20236         (nvptx_gen_unpack): Likewise.
20237         (nvptx_gen_pack): Likewise.
20238         (nvptx_gen_shuffle): Likewise.
20239         (nvptx_gen_wcast): Likewise.
20240         (nvptx_preferred_simd_mode): Likewise.
20241         * config/pa/pa.c (pa_secondary_reload): Likewise.
20242         * config/pa/predicates.md (base14_operand): Likewise.
20243         * config/powerpcspe/powerpcspe-c.c
20244         (altivec_resolve_overloaded_builtin): Likewise.
20245         * config/powerpcspe/powerpcspe.c (rs6000_setup_reg_addr_masks):
20246         Likewise.
20247         (rs6000_preferred_simd_mode): Likewise.
20248         (output_vec_const_move): Likewise.
20249         (rs6000_expand_vector_extract): Likewise.
20250         (rs6000_split_vec_extract_var): Likewise.
20251         (reg_offset_addressing_ok_p): Likewise.
20252         (rs6000_legitimate_offset_address_p): Likewise.
20253         (rs6000_legitimize_address): Likewise.
20254         (rs6000_emit_set_const): Likewise.
20255         (rs6000_const_vec): Likewise.
20256         (rs6000_emit_move): Likewise.
20257         (spe_build_register_parallel): Likewise.
20258         (rs6000_darwin64_record_arg_recurse): Likewise.
20259         (swap_selector_for_mode): Likewise.
20260         (spe_init_builtins): Likewise.
20261         (paired_init_builtins): Likewise.
20262         (altivec_init_builtins): Likewise.
20263         (do_load_for_compare): Likewise.
20264         (rs6000_generate_compare): Likewise.
20265         (rs6000_expand_float128_convert): Likewise.
20266         (emit_load_locked): Likewise.
20267         (emit_store_conditional): Likewise.
20268         (rs6000_output_function_epilogue): Likewise.
20269         (rs6000_handle_altivec_attribute): Likewise.
20270         (rs6000_function_value): Likewise.
20271         (emit_fusion_gpr_load): Likewise.
20272         (emit_fusion_p9_load): Likewise.
20273         (emit_fusion_p9_store): Likewise.
20274         * config/powerpcspe/predicates.md (easy_fp_constant): Likewise.
20275         (fusion_gpr_mem_load): Likewise.
20276         (fusion_addis_mem_combo_load): Likewise.
20277         (fusion_addis_mem_combo_store): Likewise.
20278         * config/rs6000/predicates.md (easy_fp_constant): Likewise.
20279         (fusion_gpr_mem_load): Likewise.
20280         (fusion_addis_mem_combo_load): Likewise.
20281         (fusion_addis_mem_combo_store): Likewise.
20282         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
20283         Likewise.
20284         * config/rs6000/rs6000-string.c (do_load_for_compare): Likewise.
20285         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Likewise.
20286         (rs6000_preferred_simd_mode): Likewise.
20287         (output_vec_const_move): Likewise.
20288         (rs6000_expand_vector_extract): Likewise.
20289         (rs6000_split_vec_extract_var): Likewise.
20290         (reg_offset_addressing_ok_p): Likewise.
20291         (rs6000_legitimate_offset_address_p): Likewise.
20292         (rs6000_legitimize_address): Likewise.
20293         (rs6000_emit_set_const): Likewise.
20294         (rs6000_const_vec): Likewise.
20295         (rs6000_emit_move): Likewise.
20296         (rs6000_darwin64_record_arg_recurse): Likewise.
20297         (swap_selector_for_mode): Likewise.
20298         (paired_init_builtins): Likewise.
20299         (altivec_init_builtins): Likewise.
20300         (rs6000_expand_float128_convert): Likewise.
20301         (emit_load_locked): Likewise.
20302         (emit_store_conditional): Likewise.
20303         (rs6000_output_function_epilogue): Likewise.
20304         (rs6000_handle_altivec_attribute): Likewise.
20305         (rs6000_function_value): Likewise.
20306         (emit_fusion_gpr_load): Likewise.
20307         (emit_fusion_p9_load): Likewise.
20308         (emit_fusion_p9_store): Likewise.
20309         * config/rx/rx.c (rx_gen_move_template): Likewise.
20310         (flags_from_mode): Likewise.
20311         * config/s390/predicates.md (s390_alc_comparison): Likewise.
20312         (s390_slb_comparison): Likewise.
20313         * config/s390/s390.c (s390_handle_vectorbool_attribute): Likewise.
20314         (s390_vector_mode_supported_p): Likewise.
20315         (s390_cc_modes_compatible): Likewise.
20316         (s390_match_ccmode_set): Likewise.
20317         (s390_canonicalize_comparison): Likewise.
20318         (s390_emit_compare_and_swap): Likewise.
20319         (s390_branch_condition_mask): Likewise.
20320         (s390_rtx_costs): Likewise.
20321         (s390_secondary_reload): Likewise.
20322         (__SECONDARY_RELOAD_CASE): Likewise.
20323         (s390_expand_cs): Likewise.
20324         (s390_preferred_simd_mode): Likewise.
20325         * config/s390/vx-builtins.md (vec_packsu_u<mode>): Likewise.
20326         * config/sh/sh.c (sh_print_operand): Likewise.
20327         (dump_table): Likewise.
20328         (sh_secondary_reload): Likewise.
20329         * config/sh/sh.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20330         * config/sh/sh.md (casesi_worker_1): Likewise.
20331         (casesi_worker_2): Likewise.
20332         * config/sparc/predicates.md (icc_comparison_operator): Likewise.
20333         (fcc_comparison_operator): Likewise.
20334         * config/sparc/sparc.c (sparc_expand_move): Likewise.
20335         (emit_soft_tfmode_cvt): Likewise.
20336         (sparc_preferred_simd_mode): Likewise.
20337         (output_cbranch): Likewise.
20338         (sparc_print_operand): Likewise.
20339         (sparc_expand_vec_perm_bmask): Likewise.
20340         (vector_init_bshuffle): Likewise.
20341         * config/spu/spu.c (spu_scalar_mode_supported_p): Likewise.
20342         (spu_vector_mode_supported_p): Likewise.
20343         (spu_expand_insv): Likewise.
20344         (spu_emit_branch_or_set): Likewise.
20345         (spu_handle_vector_attribute): Likewise.
20346         (spu_builtin_splats): Likewise.
20347         (spu_builtin_extract): Likewise.
20348         (spu_builtin_promote): Likewise.
20349         (spu_expand_sign_extend): Likewise.
20350         * config/tilegx/tilegx.c (tilegx_scalar_mode_supported_p): Likewise.
20351         (tilegx_simd_int): Likewise.
20352         * config/tilepro/tilepro.c (tilepro_scalar_mode_supported_p): Likewise.
20353         (tilepro_simd_int): Likewise.
20354         * config/v850/v850.c (const_double_split): Likewise.
20355         (v850_print_operand): Likewise.
20356         (ep_memory_offset): Likewise.
20357         * config/vax/vax.c (vax_rtx_costs): Likewise.
20358         (vax_output_int_move): Likewise.
20359         (vax_output_int_add): Likewise.
20360         (vax_output_int_subtract): Likewise.
20361         * config/visium/predicates.md (visium_branch_operator): Likewise.
20362         * config/visium/visium.c (rtx_ok_for_offset_p): Likewise.
20363         (visium_print_operand_address): Likewise.
20364         * config/visium/visium.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20365         * config/xtensa/xtensa.c (xtensa_mem_offset): Likewise.
20366         (xtensa_expand_conditional_branch): Likewise.
20367         (xtensa_copy_incoming_a7): Likewise.
20368         (xtensa_output_literal): Likewise.
20369         * dfp.c (decimal_real_maxval): Likewise.
20370         * targhooks.c (default_libgcc_floating_mode_supported_p): Likewise.
20372 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20373             Alan Hayward  <alan.hayward@arm.com>
20374             David Sherwood  <david.sherwood@arm.com>
20376         * genmodes.c (mode_size_inline): Add an E_ prefix to mode names.
20377         (mode_nunits_inline): Likewise.
20378         (mode_inner_inline): Likewise.
20379         (mode_unit_size_inline): Likewise.
20380         (mode_unit_precision_inline): Likewise.
20381         (emit_insn_modes_h): Likewise.  Also emit a #define of the
20382         unprefixed name.
20383         (emit_mode_wider): Add an E_ prefix to mode names.
20384         (emit_mode_complex): Likewise.
20385         (emit_mode_inner): Likewise.
20386         (emit_mode_adjustments): Likewise.
20387         (emit_mode_int_n): Likewise.
20388         * config/aarch64/aarch64-builtins.c (v8qi_UP, v4hi_UP, v4hf_UP)
20389         (v2si_UP, v2sf_UP, v1df_UP, di_UP, df_UP, v16qi_UP, v8hi_UP, v8hf_UP)
20390         (v4si_UP, v4sf_UP, v2di_UP, v2df_UP, ti_UP, oi_UP, ci_UP, xi_UP)
20391         (si_UP, sf_UP, hi_UP, hf_UP, qi_UP): Likewise.
20392         (CRC32_BUILTIN, ENTRY): Likewise.
20393         * config/aarch64/aarch64.c (aarch64_push_regs): Likewise.
20394         (aarch64_pop_regs): Likewise.
20395         (aarch64_process_components): Likewise.
20396         * config/alpha/alpha.c (alpha_emit_conditional_move): Likewise.
20397         * config/arm/arm-builtins.c (v8qi_UP, v4hi_UP, v4hf_UP, v2si_UP)
20398         (v2sf_UP, di_UP, v16qi_UP, v8hi_UP, v8hf_UP, v4si_UP, v4sf_UP)
20399         (v2di_UP, ti_UP, ei_UP, oi_UP, hf_UP, si_UP, void_UP): Likewise.
20400         * config/arm/arm.c (arm_init_libfuncs): Likewise.
20401         * config/i386/i386-builtin-types.awk (ix86_builtin_type_vect_mode):
20402         Likewise.
20403         * config/i386/i386-builtin.def (pcmpestr): Likewise.
20404         (pcmpistr): Likewise.
20405         * config/microblaze/microblaze.c (double_memory_operand): Likewise.
20406         * config/mmix/mmix.c (mmix_output_condition): Likewise.
20407         * config/powerpcspe/powerpcspe.c (rs6000_init_hard_regno_mode_ok):
20408         Likewise.
20409         * config/rl78/rl78.c (mduc_regs): Likewise.
20410         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
20411         (htm_expand_builtin): Likewise.
20412         * config/sh/sh.h (REGISTER_NATURAL_MODE): Likewise.
20413         * config/sparc/sparc.c (emit_save_or_restore_regs): Likewise.
20414         * config/xtensa/xtensa.c (print_operand): Likewise.
20415         * expmed.h (NUM_MODE_PARTIAL_INT): Likewise.
20416         (NUM_MODE_VECTOR_INT): Likewise.
20417         * genoutput.c (null_operand): Likewise.
20418         (output_operand_data): Likewise.
20419         * genrecog.c (print_parameter_value): Likewise.
20420         * lra.c (debug_operand_data): Likewise.
20422 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20424         * dwarf2out.c (scompare_loc_descriptor_wide)
20425         (scompare_loc_descriptor_narrow): New functions, split out from...
20426         (scompare_loc_descriptor): ...here.
20427         * expmed.c (emit_store_flag_int): New function, split out from...
20428         (emit_store_flag): ...here.
20430 2017-08-30  Richard Biener  <rguenther@suse.de>
20432         * dwarf2out.c (dwarf2out_finish): Remove setting AT_pubnames.
20433         (dwarf2out_early_finish): Move setting of AT_pubnames from
20434         early debug output to early finish.
20436 2017-08-30  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
20438         * gcc/config/msp430/msp430.h: Pass -mcode/data-region to the linker
20439         and -mdata-region to the assembler.
20441 2017-08-30  Richard Biener  <rguenther@suse.de>
20443         * dwarf2out.c (add_dwarf_attr): Check we don't add duplicate
20444         attributes.
20445         (gen_subprogram_die): Add DW_AT_object_pointer only early.
20446         (dwarf2out_early_global_decl): Only generate a DIE for the
20447         abstract origin if it doesn't already exist or is a declaration DIE.
20448         (resolve_addr): Do not add the linkage name twice when
20449         generating a stub DIE for the DW_TAG_GNU_call_site target.
20451 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20453         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
20454         Use machine_mode rather than int for arg1_mode.
20456 2017-08-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
20458         PR target/82015
20459         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Insure
20460         that the second argument of the built-in functions to unpack
20461         128-bit scalar types to 64-bit values is 0 or 1.  Change to use a
20462         switch statement instead a lot of if statements.
20463         * config/rs6000/rs6000.md (unpack<mode>, FMOVE128_VSX iterator):
20464         Allow 64-bit values to be in Altivec registers as well as
20465         traditional floating point registers.
20466         (pack<mode>, FMOVE128_VSX iterator): Likewise.
20468 2017-08-29  Alexander Monakov  <amonakov@ispras.ru>
20470         * ira-costs.c (record_address_regs): Handle both operands of PLUS for
20471         MAX_REGS_PER_ADDRESS == 1.
20473 2017-08-29  Uros Bizjak  <ubizjak@gmail.com>
20475         * config/i386/i386.opt (flag_fentry): Do not init to -1.
20476         * config/i386/i386.c (ix86_option_override_internal): Simplify
20477         setting of opts->x_flag_entry.
20479 2017-08-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20480             Jakub Jelinek  <jakub@redhat.com>
20481             Richard Biener  <rguenther@suse.de>
20483         PR tree-optimization/81503
20484         * gimple-ssa-strength-reduction.c (replace_mult_candidate): Ensure
20485         folded constant fits in the target type; reorder tests for clarity.
20487 2017-08-29  Martin Liska  <mliska@suse.cz>
20489         * passes.def: Include pass_lower_switch.
20490         * stmt.c (dump_case_nodes): Remove and move to
20491         tree-switch-conversion.
20492         (case_values_threshold): Likewise.
20493         (expand_switch_as_decision_tree_p): Likewise.
20494         (emit_case_decision_tree): Likewise.
20495         (expand_case): Likewise.
20496         (balance_case_nodes): Likewise.
20497         (node_has_low_bound): Likewise.
20498         (node_has_high_bound): Likewise.
20499         (node_is_bounded): Likewise.
20500         (emit_case_nodes): Likewise.
20501         (struct simple_case_node): New struct.
20502         (add_case_node): Remove.
20503         (emit_case_dispatch_table): Use vector instead of case_list.
20504         (reset_out_edges_aux): Remove.
20505         (compute_cases_per_edge): Likewise.
20506         (expand_case): Build list of simple_case_node.
20507         (expand_sjlj_dispatch_table): Use it.
20508         * tree-switch-conversion.c (struct case_node): Moved from
20509         stmt.c and adjusted.
20510         (emit_case_nodes): Likewise.
20511         (node_has_low_bound): Likewise.
20512         (node_has_high_bound): Likewise.
20513         (node_is_bounded): Likewise.
20514         (case_values_threshold): Likewise.
20515         (reset_out_edges_aux): Likewise.
20516         (compute_cases_per_edge): Likewise.
20517         (add_case_node): Likewise.
20518         (dump_case_nodes): Likewise.
20519         (balance_case_nodes): Likewise.
20520         (expand_switch_as_decision_tree_p): Likewise.
20521         (emit_jump): Likewise.
20522         (emit_case_decision_tree): Likewise.
20523         (try_switch_expansion): Likewise.
20524         (do_jump_if_equal): Likewise.
20525         (emit_cmp_and_jump_insns): Likewise.
20526         (fix_phi_operands_for_edge): New function.
20527         (record_phi_operand_mapping): Likewise.
20528         (class pass_lower_switch): New pass.
20529         (pass_lower_switch::execute): New function.
20530         (make_pass_lower_switch): Likewise.
20531         (conditional_probability):
20532         * timevar.def: Add TV_TREE_SWITCH_LOWERING.
20533         * tree-pass.h: Add make_pass_lower_switch.
20535 2017-08-29  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
20537         PR target/80993
20538         * gcc/config/msp430/msp430.c (msp430_attr): Mark interrupt
20539         handlers as used.
20541 2017-08-29  Richard Biener  <rguenther@suse.de>
20543         * dwarf2out.c (add_dwarf_attr): When checking is enabled verify
20544         we do not add a DW_AT_inline attribute twice.
20545         (gen_subprogram_die): Remove code setting DW_AT_inline on
20546         DECL_ABSTRACT_P nodes.
20548 2017-08-29  Richard Sandiford  <richard.sandiford@linaro.org>
20550         * gimplify.c (gimplify_call_expr): Copy the nothrow flag to
20551         calls to internal functions.
20552         (gimplify_modify_expr): Likewise.
20553         * tree-call-cdce.c (use_internal_fn): Likewise.
20554         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
20555         (convert_to_divmod): Set the nothrow flag.
20556         * tree-if-conv.c (predicate_mem_writes):  Likewise.
20557         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
20558         (vectorizable_call): Likewise.
20559         (vectorizable_store): Likewise.
20560         (vectorizable_load): Likewise.
20561         * tree-vect-patterns.c (vect_recog_pow_pattern): Likewise.
20562         (vect_recog_mask_conversion_pattern): Likewise.
20564 2017-08-29  Martin Liska  <mliska@suse.cz>
20566         PR other/39851
20567         * gcc.c (driver_handle_option): Add new argument.
20568         * opts-common.c (handle_option): Pass
20569         target_option_override_hook.
20570         * opts-global.c (lang_handle_option): Add new option.
20571         (set_default_handlers):  Add new argument.
20572         (decode_options): Likewise.
20573         * opts.c (target_handle_option): Likewise.
20574         (common_handle_option): Call target_option_override_hook.
20575         * opts.h (struct cl_option_handler_func): Add hook for
20576         target option override.
20577         (struct cl_option_handlers): Likewise.
20578         (set_default_handlers): Add new argument.
20579         (decode_options): Likewise.
20580         (common_handle_option): Likewise.
20581         (target_handle_option): Likewise.
20582         * toplev.c (toplev::main): Pass targetm.target_option.override
20583         hook.
20585 2017-08-29  Richard Biener  <rguenther@suse.de>
20586         Dominik Infuehr <dominik.infuehr@theobroma-systems.com>
20588         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Properly confine
20589         life to the active subtree.
20591 2017-08-28  Jeff Law  <law@redhat.com>
20593         * tree-ssa-dom.c (edge_info::record_simple_equiv): Call
20594         derive_equivalences.
20595         (derive_equivalences_from_bit_ior, record_temporary_equivalences):
20596         Code moved into....
20597         (edge_info::derive_equivalences): New private member function
20599         * tree-ssa-dom.c (class edge_info): Changed from a struct
20600         to a class.  Add ctor/dtor, methods and data members.
20601         (edge_info::edge_info): Renamed from allocate_edge_info.
20602         Initialize additional members.
20603         (edge_info::~edge_info): New.
20604         (free_dom_edge_info): Delete the edge info.
20605         (record_edge_info): Use new class & associated member functions.
20606         Tighten forms for testing for edge equivalences.
20607         (record_temporary_equivalences): Iterate over the simple
20608         equivalences rather than assuming there's only one per edge.
20609         (cprop_into_successor_phis): Iterate over the simple
20610         equivalences rather than assuming there's only one per edge.
20611         (optimize_stmt): Use operand_equal_p rather than pointer
20612         equality for mini-DSE code.
20614 2017-08-28  Nathan Sidwell  <nathan@acm.org>
20616         * gcc.c (execute): Fold SIGPIPE handling into switch
20617         statement.  Adjust internal error message.
20619 2017-08-28  Richard Biener  <rguenther@suse.de>
20621         PR debug/81993
20622         * dwarf2out.c (gen_remaining_tmpl_value_param_die_attributes):
20623         Do nothing for removed DIEs.
20625 2017-08-28  Richard Biener  <rguenther@suse.de>
20627         PR tree-optimization/81977
20628         * tree-ssa-sccvn.c (vn_reference_lookup_3): Fix look through
20629         memcpy.
20631 2017-08-28  Alexander Monakov  <amonakov@ispras.ru>
20633         PR target/80640
20634         * doc/md.texi (mem_thread_fence): Remove mention of mode.  Rewrite.
20635         * optabs.c (expand_mem_thread_fence): Emit a compiler barrier when
20636         using targetm.gen_mem_thread_fence.
20638 2017-08-27  Uros Bizjak  <ubizjak@gmail.com>
20640         PR target/81995
20641         * config/i386/i386.md (*<btsc><mode>): Change operand 2
20642         predicate to register_operand.  Reorder operands.
20643         (*btr<mode>): Ditto.
20644         (*<btsc><mode>_mask): Change operand 3 predicate to register_operand.
20645         (*btr<mode>_mask): Ditto.
20647 2017-08-25  Steven Munroe  <munroesj@gcc.gnu.org>
20649         * config.gcc (powerpc*-*-*): Add xmmintrin.h and mm_malloc.h.
20650         * config/rs6000/xmmintrin.h: New file.
20651         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include xmmintrin.h.
20653 2017-08-25  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20655         PR target/81504
20656         * config/rs6000/rs6000-p8swap.c (find_alignment_op): Add reference
20657         parameter and_insn and return it.
20658         (recombine_lvx_pattern): Insert a copy to ensure availability of
20659         the base register of the copied masking operation at the point of
20660         the instruction replacement.
20661         (recombine_stvx_pattern): Likewise.
20663 2017-08-25  Michael Meissner  <meissner@linux.vnet.ibm.com>
20665         * config/rs6000/rs6000.opt (-mpower9-dform-scalar): Delete
20666         undocumented switches.
20667         (-mpower9-dform-vector): Likewise.
20668         (-mpower9-dform): Likewise.
20669         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
20670         comments to delete references to -mpower9-dform* switches.
20671         * config/rs6000/predicates.md (vsx_quad_dform_memory_operand):
20672         Delete reference to -mpower9-dform* switches, test for
20673         -mpower9-vector instead.
20674         * config/rs6000/rs6000-cpus.def (ISA_3_0_MASKS_SERVER): Likewise.
20675         (OTHER_P9_VECTOR_MASKS): Likewise.
20676         (POWERPC_MASKS): Likewise.
20677         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Change
20678         tests against -mpower9-dform* to -mpower9-vector.  Delete code
20679         that checked for -mpower9-dform* consistancy with other options.
20680         Add test for -mpower9-misc to enable other power9 switches.
20681         (rs6000_init_hard_regno_mode_ok): Likewise.
20682         (rs6000_option_override_internal): Likewise.
20683         (rs6000_emit_prologue): Likewise.
20684         (rs6000_emit_epilogue): Likewise.
20685         (rs6000_opt_masks): Delete -mpower9-dform-{scalar,vector}.
20686         (rs6000_disable_incompatiable_switches): Delete -mpower9-dform.
20687         (emit_fusion_p9_load): Change tests for -mpower9-dform-scalar
20688         -mpower9-vector.
20689         (emit_fusion_p9_store): Likewise.
20690         * config/rs6000/rs6000.h (TARGET_P9_DFORM_SCALAR): Delete
20691         resetting these macros if the assembler does not support ISA 3.0
20692         instructions.
20693         (TARGET_P9_DFORM_VECTOR): Likewise.
20694         * config/rs6000/rs6000.md (peepholes to optimize altivec memory):
20695         Change to use -mpower9-vector instead of -mpower9-dform-scalar.
20697 2017-08-25  Alan Modra  <amodra@gmail.com>
20699         PR rtl-optimization/81747
20700         * cse.c (cse_extended_basic_block): Don't attempt to record
20701         equivalences for degenerate conditional branches that branch
20702         to their fall-through.
20704 2017-08-24  Martin Sebor  <msebor@redhat.com>
20706         PR middle-end/81908
20707         * gimple-fold.c (size_must_be_zero_p): New function.
20708         (gimple_fold_builtin_memory_op): Call it.
20710 2017-08-24  Steven Munroe  <munroesj@gcc.gnu.org>
20712         * config/rs6000/mm_malloc.h: New file.
20714 2017-08-24  Bin Cheng  <bin.cheng@arm.com>
20716         PR tree-optimization/81913
20717         * tree-ssa-loop-niter.c (number_of_iterations_cond): Skip niter
20718         analysis when either IVs in condition can wrap.
20720 2017-08-24  Uros Bizjak  <ubizjak@gmail.com>
20722         * dwarf2out.c (MAX_ARTIFICIAL_LABEL_BYTES): Increase to 40.
20723         * dwarf2cfi.c (MAX_ARTIFICIAL_LABEL_BYTES): Remove.
20725 2017-08-24  Richard Biener  <rguenther@suse.de>
20727         PR target/81921
20728         * targhooks.c (default_target_can_inline_p): Properly
20729         use target_option_default_node when no DECL_FUNCTION_SPECIFIC_TARGET
20730         is present and always compare.
20731         * config/i386/i386.c (ix86_valid_target_attribute_tree): Do not
20732         infer -mfpmath=sse from TARGET_SSE_P.
20733         (ix86_can_inline_p): Properly use target_option_default_node when
20734         no DECL_FUNCTION_SPECIFIC_TARGET is present and always compare.
20736 2017-08-24  Richard Biener  <rguenther@suse.de>
20738         PR debug/81936
20739         * dwarf2out.c (output_die): Handle flag_generate_offload like
20740         flag_generate_lto.
20741         (output_comp_unit): Likewise.
20742         (gen_array_type_die): Likewise.
20743         (dwarf2out_early_finish): Likewise.
20744         (note_variable_value_in_expr): Likewise.
20745         (dwarf2out_finish): Likewise.  Adjust assert.
20746         * cgraphunit.c (symbol_table::compile): Move setting of
20747         flag_generate_offload earlier ...
20748         (symbol_table::finalize_compilation_unit): ... here, before
20749         early debug finalization.
20751 2017-08-24  Richard Biener  <rguenther@suse.de>
20753         * config/i386/i386.c: Include symbol-summary.h, ipa-prop.h
20754         and ipa-fnsummary.h.
20755         (ix86_can_inline_p): When ix86_fpmath flags do not match
20756         check whether the callee uses FP math at all.
20758 2017-08-24  Aldy Hernandez  <aldyh@redhat.com>
20760         PR middle-end/81931
20761         * tree-ssanames.c (get_nonzero_bits): Use element_precision
20762         instead of TYPE_PRECISION.
20764 2017-08-24  Richard Sandiford  <richard.sandiford@linaro.org>
20765             Alan Hayward  <alan.hayward@arm.com>
20766             David Sherwood  <david.sherwood@arm.com>
20768         * combine.c (make_extraction): Use subreg_offset_from_lsb.
20770 2017-08-23  Daniel Santos  <daniel.santos@pobox.com>
20772         * config/i386/i386.h (ix86_frame::stack_realign_allocate_offset):
20773         Remove field.
20774         (ix86_frame::stack_realign_allocate): New field.
20775         (struct machine_frame_state): Modify comments.
20776         (machine_frame_state::sp_realigned_fp_end): New field.
20777         * config/i386/i386.c (ix86_compute_frame_layout): Rework stack frame
20778         layout calculation.
20779         (sp_valid_at): Add assertion to assure no attempt to access invalid
20780         offset of a realigned stack.
20781         (fp_valid_at): Likewise.
20782         (choose_baseaddr): Modify comments.
20783         (ix86_emit_outlined_ms2sysv_save): Adjust to changes in
20784         ix86_expand_prologue.
20785         (ix86_expand_prologue): Modify stack realignment and allocation.
20786         (ix86_expand_epilogue): Modify comments.
20787         * doc/sourcebuild.texi: Add documentation for target selectors avx2,
20788         avx2_runtime, avx512f, and avx512f_runtime.
20790 2017-08-23  Uros Bizjak  <ubizjak@gmail.com>
20792         * config/i386/i386.opt: Remove unneeded Init(0) initializations.
20793         (mstackrealign): Do not init to -1.
20794         * config/i386/i386.c (ix86_option_override_internal):
20795         Check opts_set, not opts when setting default value of
20796         opts->x_ix86_force_align_arg_pointer.
20798 2017-08-23  Richard Biener  <rguenther@suse.de>
20800         * function.c (fndecl_name): Use verbosity 1 (no arguments) for
20801         lang_hooks.decl_printable_name.
20802         * print-rtl-function.c (print_rtx_function): Likewise.
20803         * tree-pretty-print.c (dump_function_header): Likewise.
20805 2017-08-23  Richard Biener  <rguenther@suse.de>
20807         PR lto/81940
20808         * dwarf2out.c (dwarf2out_abstract_function): Handle LTO with
20809         -g0 at compile-time.
20811 2017-08-23  Tamar Christina  <tamar.christina@arm.com>
20813         PR middle-end/19706
20814         * doc/sourcebuild.texi (Other hardware attributes):
20815         Document xorsign.
20817 2017-08-23  Tamar Christina  <tamar.christina@arm.com>
20819         PR middle-end/19706
20820         * tree-ssa-math-opts.c (convert_expand_mult_copysign):
20821         Fix single-use check.
20823 2017-08-23  Thomas Preud'homme  <thomas.preudhomme@arm.com>
20825         * gcc.c (execute): Only test for SIGKILL and SIGQUIT if available.
20827 2017-08-22  Daniel Santos  <daniel.santos@pobox.com>
20829         * doc/install.texi: Modify to add more details on running selected
20830         tests.
20832 2017-08-22  Daniel Santos  <daniel.santos@pobox.com>
20834         * config/i386/i386.c (ix86_option_override_internal): Error when -mx32
20835         is combined with -mabi=ms.
20836         (ix86_function_type_abi): Limit errors for mixing -mx32 with attribute
20837         ms_abi.
20839 2017-08-22  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20841         PR tree-optimization/81488
20842         * gimple-ssa-strength-reduction (struct slsr_cand_d): Add visited
20843         and cached_basis fields.
20844         (MAX_SPREAD): New constant.
20845         (alloc_cand_and_find_basis): Initialize new fields.
20846         (clear_visited): New function.
20847         (create_phi_basis_1): Rename from create_phi_basis, set visited
20848         and cached_basis fields.
20849         (create_phi_basis): New wrapper function.
20850         (phi_add_costs_1): Rename from phi_add_costs, add spread
20851         parameter, set visited field, short-circuit when limits reached.
20852         (phi_add_costs): New wrapper function.
20853         (record_phi_increments_1): Rename from record_phi_increments, set
20854         visited field.
20855         (record_phi_increments): New wrapper function.
20856         (phi_incr_cost_1): Rename from phi_incr_cost, set visited field.
20857         (phi_incr_cost): New wrapper function.
20858         (all_phi_incrs_profitable_1): Rename from
20859         all_phi_incrs_profitable, set visited field.
20860         (all_phi_incrs_profitable): New wrapper function.
20862 2017-08-22  Richard Sandiford  <richard.sandiford@linaro.org>
20863             Alan Hayward  <alan.hayward@arm.com>
20864             David Sherwood  <david.sherwood@arm.com>
20866         * rtl.h (paradoxical_subreg_p): Define inline, and add a version
20867         that takes the outer and inner modes.
20868         * doc/rtl.texi: Use paradoxical_subreg_p instead of a GET_MODE_SIZE
20869         comparison as the canonical test for a paradoxical subreg.
20870         * combine.c (simplify_set): Use paradoxical_subreg_p.
20871         (make_extraction): Likewise.
20872         (force_to_mode): Likewise.
20873         (rtx_equal_for_field_assignment_p): Likewise.
20874         (gen_lowpart_for_combine): Likewise.
20875         (simplify_comparison): Likewise.
20876         * cse.c (equiv_constant): Likewise.
20877         * expmed.c (store_bit_field_1): Likewise.
20878         * final.c (alter_subreg): Likewise.
20879         * fwprop.c (propagate_rtx): Likewise.
20880         (forward_propagate_subreg): Likewise.
20881         * ira-conflicts.c (ira_build_conflicts): Likewise.
20882         * lower-subreg.c (simplify_gen_subreg_concatn): Likewise.
20883         * lra-constraints.c (curr_insn_transform): Likewise.
20884         (split_reg): Likewise.
20885         * lra-eliminations.c (move_plus_up): Likewise.
20886         (lra_eliminate_regs_1): Likewise.
20887         * recog.c (general_operand): Likewise.
20888         * ree.c (combine_reaching_defs): Likewise.
20889         * reload.c (push_reload): Likewise.
20890         (find_reloads): Likewise.
20891         * reload1.c (elimination_effects): Likewise.
20892         (compute_reload_subreg_offset): Likewise.
20893         (choose_reload_regs): Likewise.
20894         * rtlanal.c (subreg_lsb_1): Likewise.
20895         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
20896         (simplify_subreg): Likewise.
20897         * var-tracking.c (track_loc_p): Likewise.
20898         * emit-rtl.c (byte_lowpart_offset): Likewise.
20899         (paradoxical_subreg_p): Delete out-of-line definition.
20901 2017-08-22  Jeff Law  <law@redhat.com>
20903         PR tree-optimization/81741
20904         PR tree-optimization/71947
20905         * tree-ssa-dom.c: Include tree-inline.h.
20906         (record_temporary_equivalences): Only record SSA_NAME = SSA_NAME
20907         equivalences if one is more expensive to compute than the other.
20908         * tree-ssa-scopedtables.h (class const_or_copies): Make
20909         record_const_or_copy_raw method private.
20910         (class avail_exprs_stack): New method simplify_binary_operation.
20911         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr): Call
20912         avail_exprs_stack::simplify_binary_operation as needed.
20913         (avail_exprs_stack::simplify_binary_operation): New function.
20915 2017-08-22  Sebastian Huber  <sebastian.huber@embedded-brains.de>
20917         * config.gcc (powerpc-*-rtems*): Add rs6000/linux64.opt.
20918         * config/rs6000/rtems.h (ASM_PREFERRED_EH_DATA_FORMAT): New define.
20919         (DOT_SYMBOLS): Likewise.
20920         (MINIMAL_TOC_SECTION_ASM_OP): Likewise.
20921         (RELOCATABLE_NEEDS_FIXUP): Likewise.
20922         (RS6000_ABI_NAME): Likewise.
20923         (TARGET_CMODEL): Likewise.
20924         (TOC_SECTION_ASM_OP): Likewise.
20925         (SET_CMODEL): New macro.
20926         (SUBSUBTARGET_OVERRIDE_OPTIONS): Evaluate cmodel options.
20928 2017-08-22  Richard Biener  <rguenther@suse.de>
20930         * tree-inline.c (remap_type_1): Change asserts on TYPE_SIZE[_UNIT]
20931         to allow for free-lang-data replacements similar to verify_type_variant.
20933 2017-08-22  Richard Sandiford  <richard.sandiford@linaro.org>
20934             Alan Hayward  <alan.hayward@arm.com>
20935             David Sherwood  <david.sherwood@arm.com>
20937         * config/aarch64/aarch64.md (casesi): Use DImode rather than
20938         VOIDmode for the LABEL_REF.
20940 2017-08-22  Richard Biener  <rguenther@suse.de>
20942         * tree-cfg.c (gimple_split_edge): Avoid reallocating target PHI nodes.
20944 2017-08-22  Richard Biener  <rguenther@suse.de>
20946         * common.opt (feliminate-dwarf2-dups): Ignore.
20947         * doc/invoke.texi (feliminate-dwarf2-dups): Remove documentation.
20948         * dwarf2out.c (push_new_compile_unit, pop_compile_unit,
20949         same_die_p_wrap, compute_section_prefix,
20950         is_symbol_die, assign_symbol_names, break_out_includes): Remove.
20951         (comdat_symbol_id, comdat_symbol_number): Likewise.
20952         (cu_hash_table_entry, cu_hash_table_entry_hasher, cu_hash_type):
20953         Likewise.
20954         (check_duplicate_cu, record_comdat_symbol_number): Likewise.
20955         (output_die): Mark unreachable path unreachable.
20956         (dwarf2out_start_source_file): Do not create DW_TAG_GNU_BINCL.
20957         (dwarf2out_end_source_file): Do not create DW_TAG_GNU_EINCL.
20958         (dwarf2out_init): Remove code handling flag_eliminate_dwarf2_dups.
20959         (dwarf2out_early_finish): Likewise.
20961 2017-08-22  Aldy Hernandez  <aldyh@redhat.com>
20963         * wide-int.h (hwi_with_prec::hwi_with_prec): Sign extend.
20965 2017-08-22  Georg-Johann Lay  <avr@gjlay.de>
20967         PR target/81910
20968         * config/avr/avr.c (avr_handle_addr_attribute): Early return if
20969         not VAR_P. Filter attribute warnings with OPT_Wattributes.
20970         (avr_attribute_table) <io, io_low, address>: Initialize
20971         .decl_required with true.
20973 2017-08-21  Michael Meissner  <meissner@linux.vnet.ibm.com>
20975         * config/rs6000/rs6000-cpus.def (-mvsx-scalar-float): Delete
20976         undocumented debugging options.
20977         (-mvsx-scalar-double): Likewise.
20978         (-mallow-df-permute): Likewise.
20979         (-mvectorize-builtins): Likewise.
20980         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
20981         (rs6000_builtin_vectorized_function): Likewise.
20982         (rs6000_builtin_md_vectorized_function): Likewise.
20983         (rs6000_opt_vars): Likewise.
20985 2017-08-21  Uros Bizjak  <ubizjak@gmail.com>
20987         PR target/46091
20988         * config/i386/i386.md (*btsq_imm): Rename from *btsq.
20989         (*btrq_imm): Rename from *btrq.
20990         (*btcq_imm): Rename from *btcq.
20991         (btsc): New code attribute.
20992         (*<btsc><mode>): New insn pattern.
20993         (*btr<mode>): Ditto.
20994         (*<btsc><mode>_mask): New insn_and_split pattern.
20995         (*btr<mode>_mask): Ditto.
20997 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
20998             Alan Hayward  <alan.hayward@arm.com>
20999             David Sherwood  <david.sherwood@arm.com>
21001         * function.c (pad_below): Simplify padding calculation.
21003 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21004             Alan Hayward  <alan.hayward@arm.com>
21005             David Sherwood  <david.sherwood@arm.com>
21007         * target.def (function_prologue): Remove frame size argument.
21008         (function_epilogue): Likewise.
21009         * doc/tm.texi: Regenerate.
21010         * final.c (final_start_function): Update call to function_prologue.
21011         (final_end_function): Update call to function_epilogue.
21012         (default_function_pro_epilogue): Remove frame size argument.
21013         * output.h (default_function_pro_epilogue): Likewise.
21014         * config/arm/arm.c (arm_output_function_epilogue): Likewise.
21015         (arm_output_function_prologue): Likewise.
21016         * config/frv/frv.c (frv_function_prologue): Likewise.
21017         (frv_function_epilogue): Likewise.
21018         * config/i386/i386.c (ix86_output_function_epilogue): Likewise.
21019         * config/ia64/ia64.c (ia64_output_function_prologue): Likewise.
21020         (ia64_output_function_epilogue): Likewise.
21021         * config/m32r/m32r.c (m32r_output_function_prologue): Likewise.
21022         (m32r_output_function_epilogue): Likewise.
21023         * config/microblaze/microblaze.c (microblaze_function_prologue)
21024         (microblaze_function_epilogue): Likewise.
21025         * config/mips/mips.c (mips_output_function_prologue): Likewise.
21026         (mips_output_function_epilogue): Likewise.
21027         * config/mmix/mmix.c (mmix_target_asm_function_prologue): Likewise.
21028         (mmix_target_asm_function_epilogue): Likewise.
21029         * config/msp430/msp430.c (msp430_start_function): Likewise.
21030         * config/nds32/nds32.c (nds32_asm_function_prologue): Likewise.
21031         (nds32_asm_function_epilogue): Likewise.
21032         * config/nios2/nios2.c (nios2_asm_function_prologue): Likewise.
21033         * config/pa/pa.c (pa_output_function_prologue): Likewise.
21034         (pa_output_function_epilogue): Likewise.
21035         * config/powerpcspe/powerpcspe.c (rs6000_output_function_prologue)
21036         (rs6000_output_function_epilogue): Likewise.
21037         * config/rl78/rl78.c (rl78_start_function): Likewise.
21038         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Likewise.
21039         (rs6000_output_function_epilogue): Likewise.
21040         * config/rx/rx.c (rx_output_function_prologue): Likewise.
21041         * config/sh/sh.c (sh_output_function_epilogue): Likewise.
21042         * config/sparc/sparc.c (sparc_asm_function_prologue): Likewise.
21043         (sparc_asm_function_epilogue): Likewise.
21045 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21047         * tree.h (type_has_mode_precision_p): New function.
21048         * convert.c (convert_to_integer_1): Use it.
21049         * expr.c (expand_expr_real_2): Likewise.
21050         (expand_expr_real_1): Likewise.
21051         * fold-const.c (fold_single_bit_test_into_sign_test): Likewise.
21052         * match.pd: Likewise.
21053         * tree-ssa-forwprop.c (simplify_rotate): Likewise.
21054         * tree-ssa-math-opts.c (convert_mult_to_fma): Likewise.
21055         * tree-tailcall.c (process_assignment): Likewise.
21056         * tree-vect-loop.c (vectorizable_reduction): Likewise.
21057         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern)
21058         (vect_recog_mult_pattern, vect_recog_divmod_pattern): Likewise.
21059         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
21060         (vectorizable_assignment): Likewise.
21061         (vectorizable_shift): Likewise.
21062         (vectorizable_operation): Likewise.
21063         * tree-vrp.c (register_edge_assert_for_2): Likewise.
21065 2017-08-21  Wilco Dijkstra  <wdijkstr@arm.com>
21067         * match.pd: Add pow (C, x) simplification.
21069 2017-08-21  Richard Biener  <rguenther@suse.de>
21071         PR tree-optimization/81900
21072         * tree-ssa-pre.c (compute_antic_aux): Properly compute changed
21073         for blocks with abnormal predecessors.
21074         (compute_antic): Do not set visited flag prematurely.
21076 2017-08-21  Georg-Johann Lay  <avr@gjlay.de>
21078         PR target/79883
21079         * config/avr/avr.c (avr_set_current_function): Typo in diagnostic.
21081 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21083         * stor-layout.h (vector_type_mode): Move to...
21084         * tree.h (vector_type_mode): ...here.
21085         * stor-layout.c (vector_type_mode): Move to...
21086         * tree.c (vector_type_mode): ...here.  Include rtl.h and regs.h.
21088 2017-08-21  Richard Biener  <rguenther@suse.de>
21090         * debug.h (struct gcc_debug_hooks): Add die_ref_for_decl and
21091         register_external_die hooks.
21092         (debug_false_tree_charstarstar_uhwistar): Declare.
21093         (debug_nothing_tree_charstar_uhwi): Likewise.
21094         * debug.c (do_nothing_debug_hooks): Adjust.
21095         (debug_false_tree_charstarstar_uhwistar): New do nothing.
21096         (debug_nothing_tree_charstar_uhwi): Likewise.
21097         * dbxout.c (dbx_debug_hooks): Adjust.
21098         (xcoff_debug_hooks): Likewise.
21099         * sdbout.c (sdb_debug_hooks): Likewise.
21100         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
21101         * dwarf2out.c (macinfo_label_base): New global.
21102         (dwarf2out_register_external_die): New function for the
21103         register_external_die hook.
21104         (dwarf2out_die_ref_for_decl): Likewise for die_ref_for_decl.
21105         (dwarf2_debug_hooks): Use them.
21106         (dwarf2_lineno_debug_hooks): Adjust.
21107         (struct die_struct): Add with_offset flag.
21108         (DEBUG_LTO_DWO_INFO_SECTION, DEBUG_LTO_INFO_SECTION,
21109         DEBUG_LTO_DWO_ABBREV_SECTION, DEBUG_LTO_ABBREV_SECTION,
21110         DEBUG_LTO_DWO_MACINFO_SECTION, DEBUG_LTO_MACINFO_SECTION,
21111         DEBUG_LTO_DWO_MACRO_SECTION, DEBUG_LTO_MACRO_SECTION,
21112         DEBUG_LTO_LINE_SECTION, DEBUG_LTO_DWO_STR_OFFSETS_SECTION,
21113         DEBUG_LTO_STR_DWO_SECTION, DEBUG_STR_LTO_SECTION): New macros
21114         defining section names for the early LTO debug variants.
21115         (reset_indirect_string): New helper.
21116         (add_AT_external_die_ref): Helper for dwarf2out_register_external_die.
21117         (print_dw_val): Add support for offsetted symbol references.
21118         (get_ultimate_context): Split out from is_cxx.
21119         (is_cxx): Use get_ultimate_context.
21120         (is_fortran): Add decl overload.
21121         (compute_comp_unit_symbol): Split out worker from
21122         compute_section_prefix.
21123         (compute_section_prefix): Call compute_comp_unit_symbol and
21124         set comdat_type_p here.
21125         (output_die): Skip DIE symbol output for the LTO added one.
21126         Handle DIE symbol references with offset.
21127         (output_comp_unit): Guard section name mangling properly.
21128         For LTO debug sections emit a symbol at the section beginning
21129         which we use to refer to its DIEs.
21130         (add_abstract_origin_attribute): For DIEs registered via
21131         dwarf2out_register_external_die directly refer to the early
21132         DIE rather than indirectly through the shadow one we created.
21133         Remove obsolete call to dwarf2out_abstract_function for
21134         non-function/block origins.
21135         (gen_array_type_die): When generating early LTO debug do
21136         not emit DW_AT_string_length.
21137         (gen_formal_parameter_die): Do not re-create DIEs for PARM_DECLs
21138         late when in LTO.  As suggested place a gcc_unreachable for
21139         the DECL_ABSTRACT_P case.
21140         (gen_subprogram_die): Avoid another specification DIE
21141         for early built declarations/definitions for the late LTO case.
21142         (gen_variable_die): Add type references for late duplicated VLA dies
21143         when in late LTO.
21144         (gen_inlined_subroutine_die): Do not call dwarf2out_abstract_function,
21145         we have the abstract instance already.
21146         (process_scope_var): Adjust decl DIE contexts in LTO which
21147         first puts them in limbo.
21148         (gen_decl_die): Do not generate type DIEs late apart from
21149         types for VLAs or for decls we do not yet have a DIE.  Do not
21150         call dwarf2out_abstract_function late.
21151         (dwarf2out_early_global_decl): Make sure to create DIEs
21152         for abstract instances of a decl first.
21153         (dwarf2out_late_global_decl): Adjust comment.
21154         (output_macinfo_op): With multiple macro sections use
21155         macinfo_label_base to distinguish labels.
21156         (output_macinfo): Likewise.  Update macinfo_label_base.
21157         Pass in the line info label.
21158         (note_variable_value_in_expr): When generating LTO resolve
21159         all variable values here by generating DIEs as needed.
21160         (init_sections_and_labels): Add early LTO debug flag parameter
21161         and generate different sections and names if set.  Add generation
21162         counter for the labels so we can have multiple of them.
21163         (reset_dies): Helper to allow DIEs to be output multiple times.
21164         (dwarf2out_finish): When outputting DIEs to the fat part of an
21165         LTO object first reset DIEs.
21166         (dwarf2out_early_finish): Output early DIEs when generating LTO.
21167         (modified_type_die): Check for decl_ultimate_origin being self
21168         before recursing.
21169         (gen_type_die_with_usage): Likewise.
21170         (gen_typedef_die): Allow decl_ultimate_origin being self.
21171         (set_decl_abstract_flags): Remove.
21172         (set_block_abstract_flags): Likewise.
21173         (dwarf2out_abstract_function): Treat the early generated DIEs
21174         as the abstract copy and only add DW_AT_inline and
21175         DW_AT_artificial here and call set_decl_origin_self.
21176         If the DIE has an abstract origin don't do anything.
21177         * tree.c (free_lang_data): Build a dummy TRANSLATION_UNIT_DECL
21178         if we have none yet (Go fails to build one, PR78628).
21179         (variably_modified_type_p): Prevent endless recursion for Ada
21180         cyclic pointer types.
21181         * lto-streamer-in.c: Include debug.h.
21182         (dref_queue): New global.
21183         (lto_read_tree_1): Stream in DIE references.
21184         (lto_input_tree): Register DIE references.
21185         (input_function): Stream DECL_DEBUG_ARGS.
21186         * lto-streamer-out.c: Include debug.h.
21187         (lto_write_tree_1): Output DIE references.
21188         (DFS::DFS_write_tree_body): Follow DECL_ABSTRACT_ORIGIN.
21189         Force a TRANSLATION_UNIT_DECL DECL_CONTEXT for file-scope decls.
21190         (output_function): Stream DECL_DEBUG_ARGS.
21191         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
21192         Stream DECL_ABSTRACT_ORIGIN.
21193         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
21194         (write_ts_decl_minimal_tree_pointers): Force a TRANSLATION_UNIT_DECL
21195         DECL_CONTEXT for file-scope decls.
21196         * lto-streamer.h (struct dref_entry): Declare.
21197         (dref_queue): Likewise.
21198         * cfgexpand.c (pass_expand::execute): Do not call the
21199         outlining_inline_function hook here.
21200         * lto-wrapper.c (debug_obj): New global.
21201         (tool_cleanup): Unlink it if required.
21202         (debug_objcopy): New function.
21203         (run_gcc): Handle early debug sections in the IL files by
21204         extracting them to separate files, partially linkin them and
21205         feeding the result back as result to the linker.
21206         * config/darwin.h (DEBUG_LTO_INFO_SECTION, DEBUG_LTO_ABBREV_SECTION,
21207         DEBUG_LTO_MACINFO_SECTION, DEBUG_LTO_LINE_SECTION,
21208         DEBUG_STR_LTO_SECTION, DEBUG_LTO_MACRO_SECTION): Put early debug
21209         sections into a separate segment.
21210         * config/darwin.c (darwin_asm_named_section): Handle __GNU_DWARF_LTO
21211         segments.
21212         (darwin_asm_dwarf_section): Likewise.
21213         (darwin_asm_output_dwarf_offset): Likewise.
21214         * config/i386/i386.c (make_resolver_func): Set DECL_IGNORED_P.
21216 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21217             Alan Hayward  <alan.hayward@arm.com>
21218             David Sherwood  <david.sherwood@arm.com>
21220         * read-md.h (md_reader::record_potential_iterator_use): Replace
21221         pointer argument with an rtx and an index.
21222         * read-rtl.c (iterator_group::apply_iterator): Likewise.
21223         (apply_mode_iterator): Likewise.
21224         (apply_code_iterator): Likewise.
21225         (apply_int_iterator): Likewise.
21226         (apply_subst_iterator): Likewise.
21227         (record_iterator_use): Likewise.
21228         (record_attribute_use): Likewise.
21229         (md_reader::record_potential_iterator_use): Likewise.  Update calls
21230         to record_iterator_use and apply_iterator.
21231         (iterator_use): Replace ptr with x and index.
21232         (attribute_use): Likewise.
21233         (apply_attribute_uses): Update calls to apply_iterator.
21234         (apply_iterators): Likewise.  Update initialization of iterator_use.
21235         (rtx_reader::read_rtx_code): Update calls to record_iterator_use
21236         and record_potential_iterator_use.
21237         (rtx_reader::read_rtx_operand): Likewise.
21239 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21240             Alan Hayward  <alan.hayward@arm.com>
21241             David Sherwood  <david.sherwood@arm.com>
21243         * varasm.c (const_rtx_hash_1): Don't hash in the mode of a
21244         CONST_WIDE_INT.
21246 2017-08-21  Richard Biener  <rguenther@suse.de>
21248         PR middle-end/81884
21249         * tree-ssa-alias.c (stmt_kills_ref_p): Handle array accesses
21250         at struct end conservatively when comparing common bases.
21252 2017-08-21  Richard Biener  <rguenther@suse.de>
21254         * tree-ssa-loop-im.c (struct lim_aux_data): Add ref index member.
21255         (mem_ref_in_stmt): Remove.
21256         (determine_max_movement): Use ref index to get at the reference.
21257         (invariantness_dom_walker::before_dom_children): Deal with
21258         lim data already initialized.
21259         (gather_mem_refs_stmt): Initialize lim data and record ref index.
21261 2017-08-19  Uros Bizjak  <ubizjak@gmail.com>
21263         * config/i386/i386.h (OPTION_MASK_ISA_ROUND): Remove.
21264         (TARGET_ISA_ROUND): Ditto.
21265         (TARGET_ROUND): Ditto.
21266         * config/i386/i386.c: Substitute TARGET_ROUND with TARGET_SSE4_1.
21267         * config/i386/i386.md: Ditto.
21268         * config/i386/sse.md: Ditto.
21269         * config/i386/i386-builtin.def: Substitute OPTION_MASK_ISA_ROUND
21270         with OPTION_MASK_ISA_SSE4_1.
21272 2017-08-19  Uros Bizjak  <ubizjak@gmail.com>
21274         PR target/81894
21275         * doc/extend.texi (x86 Built-in Functions): Correct the name of
21276         __builtin_ia32_lzcnt_u16.
21278 2017-08-18  Peter Bergner  <bergner@vnet.ibm.com>
21280         PR target/80210
21281         * config/rs6000/rs6000.c (rs6000_activate_target_options): New function.
21282         (rs6000_set_current_function): Rewrite function to use it.
21284 2017-08-18  H.J. Lu  <hongjiu.lu@intel.com>
21286         PR c/53037
21287         * print-tree.c (print_node): Support DECL_WARN_IF_NOT_ALIGN
21288         and TYPE_WARN_IF_NOT_ALIGN.
21289         * stor-layout.c (do_type_align): Merge DECL_WARN_IF_NOT_ALIGN.
21290         (handle_warn_if_not_align): New.
21291         (place_union_field): Call handle_warn_if_not_align.
21292         (place_field): Call handle_warn_if_not_align.
21293         Copy TYPE_WARN_IF_NOT_ALIGN.
21294         (finish_builtin_struct): Copy TYPE_WARN_IF_NOT_ALIGN.
21295         (layout_type): Likewise.
21296         * tree-core.h (tree_type_common): Add warn_if_not_align.  Set
21297         spare to 18.
21298         (tree_decl_common): Add warn_if_not_align.
21299         * tree.c (build_range_type_1): Copy TYPE_WARN_IF_NOT_ALIGN.
21300         * tree.h (TYPE_WARN_IF_NOT_ALIGN): New.
21301         (SET_TYPE_WARN_IF_NOT_ALIGN): Likewise.
21302         (DECL_WARN_IF_NOT_ALIGN): Likewise.
21303         (SET_DECL_WARN_IF_NOT_ALIGN): Likewise.
21304         * doc/extend.texi: Document warn_if_not_aligned attribute.
21305         * doc/invoke.texi: Document -Wif-not-aligned and -Wpacked-not-aligned.
21307 2017-08-17  Martin Liska  <mliska@suse.cz>
21309         PR bootstrap/81864
21310         * tree-loop-distribution.c (ddrs_table): Change type to pointer type.
21311         (get_data_dependence): Use it as pointer type.
21312         (distribute_loop): Likewise.
21314 2017-08-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21316         * config/rs6000/altivec.md (UNSPEC_VMRGOW_DIRECT): New constant.
21317         (p8_vmrgew_v4sf_direct): Generalize to p8_vmrgew_<mode>_direct.
21318         (p8_vmrgow_<mode>_direct): New define_insn.
21319         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Properly
21320         handle endianness for vmrgew and vmrgow permute patterns.
21322 2017-08-17  Peter Bergner  <bergner@vnet.ibm.com>
21324         * config/rs6000/altivec.md (VParity): Remove TARGET_VSX_TIMODE.
21325         * config/rs6000/rs6000-cpus.def: Remove comment.
21326         (ISA_2_7_MASKS_SERVER): Delete OPTION_MASK_VSX_TIMODE;
21327         (POWERPC_MASKS): Likewise.
21328         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Remove unneeded
21329         use of TARGET_VSX_TIMODE.
21330         (rs6000_setup_reg_addr_masks): Change TARGET_VSX_TIMODE to TARGET_VSX.
21331         (rs6000_init_hard_regno_mode_ok): Remove unneeded uses of
21332         TARGET_VSX_TIMODE.  Change use of TARGET_VSX_TIMODE to TARGET_VSX.
21333         (rs6000_option_override_internal): Remove dead code.
21334         (rs6000_legitimize_address): Change TARGET_VSX_TIMODE to TARGET_VSX.
21335         (rs6000_legitimize_reload_address): Likewise.
21336         (rs6000_legitimate_address_p): Likewise.
21337         (rs6000_opt_masks): Delete "vsx-timode".
21338         (rs6000_disable_incompatible_switches): Remove mention of -mvsx-timode
21339         from function comment.
21340         * config/rs6000/rs6000.h (MASK_VSX_TIMODE): Delete.
21341         * config/rs6000/rs6000.md (FMOVE128_GPR): Remove TARGET_VSX_TIMODE.
21342         (V16QI, V8HI, V4SI, V4SF, V2DI, V2DF, V1TI): Remove useless empty
21343         condition.
21344         * config/rs6000/rs6000.opt (mvsx-timode): Replace with stub.
21345         * config/rs6000/vector.md (VEC_IP): Remove TARGET_VSX_TIMODE.
21346         * config/rs6000/vsx.md (VSX_LE_128): Likewise.
21347         (VSX_TI): Likewise.
21348         (VSX_M): Likewise.
21349         (define_peephole2): Likewise.
21351 2017-08-17  Martin Sebor  <msebor@redhat.com>
21353         PR c/81859
21354         * pretty-print.c (pp_format): Use strnlen in %.*s to avoid reading
21355         past the end of an array.
21356         (test_pp_format): Add test cases.
21358 2017-08-17  Richard Sandiford  <richard.sandiford@linaro.org>
21360         * internal-fn.def (CLRSB, CLZ, CTZ, FFS, PARITY, POPCOUNT): Add
21361         missing ECF_NOTHROW flags.
21363 2017-08-17  Peter Bergner  <bergner@vnet.ibm.com>
21365         PR target/72804
21366         * config/rs6000/vsx.md (*vsx_le_permute_<mode>): Add support for
21367         operands residing in integer registers.
21368         (*vsx_le_perm_load_<mode>): Likewise.
21369         (*vsx_le_perm_store_<mode>): Likewise.
21370         (define_peephole2): Add peepholes to optimize the above.
21372 2017-08-17  Marek Polacek  <polacek@redhat.com>
21374         PR middle-end/81814
21375         * fold-const.c (operand_equal_for_comparison_p): Remove code that used
21376         to mimic what shorten_compare did.  Change the return type to bool.
21377         (fold_cond_expr_with_comparison): Update call to
21378         operand_equal_for_comparison_p.
21379         (fold_ternary_loc): Likewise.
21381 2017-08-17  Jackson Woodruff  <jackson.woodruff@arm.com>
21383         * aarch64-simd.md (mov<mode>): No longer force zero immediate into
21384         register.
21385         (*aarch64_simd_mov<mode>): Add new case for stp using zero immediate.
21387 2017-08-17  Richard Biener  <rguenther@suse.de>
21389         * tree-ssa-structalias.c (solve_graph): When propagating
21390         to successors update the graphs succ edges and avoid duplicate work.
21392 2017-08-17  Maxim Ostapenko  <m.ostapenko@samsung.com>
21394         PR target/81861
21395         * config/i386/i386.c (ix86_option_override_internal): Save target
21396         specific options after ix86_stack_protector_guard_reg was changed.
21398 2017-08-17  Richard Biener  <rguenther@suse.de>
21400         PR tree-optimization/81827
21401         * tree-ssa-structalias.c (struct variable_info): Add is_reg_var flag.
21402         (new_var_info): Initialize it conservatively.
21403         (get_call_vi): Mark register vars.
21404         (new_scalar_tmp_constraint_exp): Likewise.
21405         (handle_rhs_call): Likewise.
21406         (handle_const_call): Likewise.
21407         (create_function_info_for): Likewise.
21408         (solve_constraints): Sort varinfos to separate register from
21409         non-register vars to pack points-to solution bitmaps during
21410         iteration.
21412 2017-08-17  Marek Polacek  <polacek@redhat.com>
21414         * gimplify.c (gimplify_adjust_omp_clauses): Compare with 0 instead of 1.
21416 2017-08-17  Richard Biener  <rguenther@suse.de>
21418         * tree-vrp.c (vrp_int_const_binop): Do not set *overflow_p
21419         to true when overflow is undefined and we saturated the result.
21421 2017-08-17  Alan Modra  <amodra@gmail.com>
21423         PR target/80938
21424         * config/rs6000/rs6000.c (rs6000_savres_strategy): Revert 2017-08-09.
21425         Don't use store multiple if only one reg needs saving.
21426         (interesting_frame_related_regno): New function.
21427         (rs6000_frame_related): Don't emit frame info for regs that
21428         don't need saving.
21429         (rs6000_emit_epilogue): Likewise.
21431 2017-08-16  Nathan Sidwell  <nathan@acm.org>
21433         * tree-core.h (tree_type_non_common): Rename binfo to lang_1.
21434         * tree.h (TYPE_BINFO): Use type_non_common.maxval.
21435         (TYPE_LANG_SLOT_1): Use type_non_common.lang_1, for any type.
21436         * tree.c (free_lang_data_in_type): Use else-if chain.  Always
21437         clear TYPE_LANG_1.  Remove obsolete member-function stripping.
21438         (find_decls_types_r): Comment about TYPE_MAX_VALUES_RAW.
21439         (verify_type): Adjust for TYPE_BINFO move.
21440         * lto-streamer-out.c (DFS::DFS_write_tree_body): No need to
21441         process TYPE_BINFO directly.
21442         (hash_tree): Likewise.
21443         * tree-streamer-in.c (lto_input_ts_type_non_common_tree_pointers):
21444         Likewise.
21445         * tree-streamer-out.c (write_ts_type_non_common_tree_pointers):
21446         Likewise.
21448 2017-08-16  David Malcolm  <dmalcolm@redhat.com>
21450         * diagnostic-show-locus.c (colorizer::m_caret): Remove unused field.
21452 2017-08-16  Uros Bizjak  <ubizjak@gmail.com>
21454         PR target/46091
21455         * config/i386/i386.md (*anddi_1_btr): Change predicates of
21456         operand 0 and operand 1 to nomimmediate_operand. Add "m" constraint.
21457         Add ix86_binary_operator_ok to insn constraint.
21458         (*iordi_1_bts): Ditto.
21459         (*xordi_1_btc): Ditto.
21460         (*btsq): Change predicate of operand 0 to nonimmediate_operand.
21461         Update corresponding peephole2 pattern.
21462         (*btrq): Ditto.
21463         (*btcq): Ditto.
21465 2017-08-16  Bin Cheng  <bin.cheng@arm.com>
21467         PR tree-optimization/81832
21468         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Don't
21469         copy loop header which has IFN_LOOP_DIST_ALIAS call.
21471 2017-08-16  Marek Polacek  <polacek@redhat.com>
21473         PR middle/81695
21474         * fold-const.c (fold_indirect_ref_1): Restore original behavior
21475         regarding size_zero_node.
21477 2017-08-16  Martin Liska  <mliska@suse.cz>
21479         PR target/81753
21480         * config.gcc: Respect previously set extra_objs in case
21481         of darwin target.
21483 2017-08-16  Richard Sandiford  <richard.sandiford@linaro.org>
21485         PR tree-optimization/81835
21486         * tree-vect-loop.c (vect_is_simple_reduction): Simply checks for
21487         the phi SSA_NAME.  Check that the condition in a COND_EXPR does
21488         not depend on the phi.
21490 2017-08-16  Alan Modra  <amodra@gmail.com>
21492         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Delete
21493         dead code.
21495 2017-08-16  Alan Modra  <amodra@gmail.com>
21497         * config/rs6000/rs6000.c (rs6000_reg_live_or_pic_offset_p): Merge..
21498         (save_reg_p): ..into this.  Update all callers.
21499         (first_reg_to_save): Simplify.
21501 2017-08-16  Alan Modra  <amodra@gmail.com>
21503         * config/rs6000/rs6000.c (rs6000_savres_strategy): Don't restore
21504         fixed regs.
21506 2017-08-15  Joseph Myers  <joseph@codesourcery.com>
21508         PR target/78460
21509         PR target/67712
21510         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Only unroll for
21511         constant count if that count is less than 32.
21513 2017-08-15  Nathan Sidwell  <nathan@acm.org>
21515         * gcc.c (execute): Emit friendlier message if inferior is killed
21516         by an external cause.
21518 2017-08-15  Richard Biener  <rguenther@suse.de>
21520         PR tree-optimization/81790
21521         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Handle both
21522         CONSTRUCTORs from simplifying and VN.
21524 2017-08-14  Martin Sebor  <msebor@redhat.com>
21526         * builtin-attrs.def: Add comments.
21528 2017-08-14  Martin Sebor  <msebor@redhat.com>
21530         PR c/81117
21531         * doc/extend.texi (attribute nonstring): Document new attribute.
21533 2017-08-14  Martin Sebor  <msebor@redhat.com>
21535         PR c/81117
21536         * tree-diagnostic.c (default_tree_printer): Handle %G.
21537         * gimple-pretty-print.h (percent_G_format): Declare new function.
21538         * gimple-pretty-print.c (percent_G_format): Define.
21539         * tree-pretty-print.c (percent_K_format): Add argument.
21541 2017-08-14  Martin Sebor  <msebor@redhat.com>
21543         PR translation/79998
21544         * gimple-ssa-sprintf.c (pass_sprintf_length::handle_gimple_call):
21545         Remove a stray space.
21547 2017-08-14  Uros Bizjak  <ubizjak@gmail.com>
21549         PR target/46091
21550         * config/i386/i386.md (*anddi_1_btr): New insn_and_split pattern.
21551         (*iordi_1_bts): Ditto.
21552         (*xordi_1_btc): Ditto.
21554 2017-08-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21556         PR target/79845
21557         * config/rs6000/linux64.h (INVALID_64BIT): Use quoted strings.
21558         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
21559         Likewise.
21560         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Use
21561         quoted strings, and make more translator-friendly.
21562         (darwin_rs6000_override_options): Likewise.
21563         (rs6000_option_override_internal): Likewise.
21564         (rs6000_return_in_memory): Fix overlong line.
21565         (init_cmulative_args): Use quoted strings, and make more
21566         translator-friendly.
21567         (rs6000_pass_by_reference): Fix overlong line.
21568         (def_builtin): Use quoted strings.
21569         (altivec_expand_predicate_builtin): Use quoted strings, and make
21570         more translator-friendly.
21571         (htm_expand_builtin): Use quoted strings.
21572         (cpu_expand_builtin): Use quoted strings, and make more
21573         translator-friendly.
21574         (altivec_expand_builtin): Likewise.
21575         (paired_expand_predicate_builtin): Likewise.
21576         (rs6000_invalid_builtin): Likewise.
21577         (builtin_function_type): Use quoted strings.
21578         (rs6000_expand_split_stack_prologue): Use quoted strings, and make
21579         more translator-friendly.
21580         (rs6000_trampoline_init): Likewise.
21581         (rs6000_handle_altivec_attribute): Likewise.
21582         (rs6000_inner_target_options): Use quoted strings.
21583         (rs6000_disable_incompatible_switches): Likewise.
21584         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Use quoted
21585         strings, and make more translator-friendly.
21586         (SUBSUBTARGET_OVERRIDE_OPTIONS): Use quoted strings.
21588 2017-08-14  Bin Cheng  <bin.cheng@arm.com>
21590         PR tree-optimization/81799
21591         * tree-loop-distribution.c (version_loop_by_alias_check): Force
21592         cond_expr to simple gimple operand.
21594 2017-08-14  Wilco Dijkstra  <wdijkstr@arm.com>
21596         PR middle-end/46932
21597         * doc/sourcebuild.texi (autoincdec): Add autoincdec description.
21599 2017-08-14  Georg-Johann Lay  <avr@gjlay.de>
21601         PR target/81754
21602         PR target/81268
21603         * config/avr/avr.opt (mgas-isr-prologues): New Var avr_gasisr_prologues.
21604         * config/avr/avr.md (gasisr, *gasisr): Use it instead of
21605         TARGET_GASISR_PROLOGUES.
21606         * config/avr/avr.c (avr_option_override): Same.
21607         (avr_pass_pre_proep::execute): Same.
21609 2017-08-13  H.J. Lu  <hongjiu.lu@intel.com>
21611         PR target/81820
21612         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Replace
21613         frame pointer with stack pointer - UNITS_PER_WORD in debug insns.
21615 2017-08-13  Uros Bizjak  <ubizjak@gmail.com>
21617         * config/i386/i386.md (*load_tp_<mode>): Redefine as
21618         define_insn_and_split.  Split to a memory load from 0 in
21619         DEFAULT_TLS_SEG_REG address space.  Merge with *load_tp_x32
21620         using PTR mode iterator.
21621         (*load_tp_x32_zext"): Redefine as define_insn_and_split.
21622         Split to a memory load from 0 in DEFAULT_TLS_SEG_REG address space.
21623         (*add_tp_<mode>): Redefine as define_insn_and_split.
21624         Split to an add with a memory load from 0 in DEFAULT_TLS_SEG_REG
21625         address space.  Merge with *add_tp_x32 using PTR mode iterator.
21626         (*add_tp_x32_zext"): Redefine as define_insn_and_split.
21627         Split to an add with a  memory load from 0 in
21628         DEFAULT_TLS_SEG_REG address space.
21630 2017-08-12  Andrew Pinski  <apinski@cavium.com>
21632         * config/aarch64/aarch64-option-extensions.def (rdma):
21633         Fix feature string to what Linux prints out in /proc/cpuinfo.
21635 2017-08-12  Pierre-Marie de Rodat  <derodat@adacore.com>
21637         PR ada/79542
21638         * dwarf2out.c (modified_type_die): For C typedef types that have
21639         an ultimate origin, process the ultimate origin instead of the
21640         input type.
21641         (gen_typedef_die): Assert that input DECLs have no ultimate
21642         origin.
21643         (gen_type_die_with_usage): For typedef variants that have an
21644         ultimate origin, just call gen_decl_die on the original DECL.
21645         (process_scope_var): Avoid creating DIEs for local typedefs and
21646         concrete static variables.
21648 2017-08-12  Alan Modra  <amodra@gmail.com>
21650         PR target/81170
21651         PR target/81295
21652         * config/rs6000/sysv4.h (STARTFILE_LINUX_SPEC): Upgrade to
21653         match gnu-user.h startfile.
21654         (ENDFILE_LINUX_SPEC): Similarly.
21656 2017-08-11  Thomas Schwinge  <thomas@codesourcery.com>
21658         PR lto/81430
21659         * config/nvptx/nvptx.c (nvptx_override_options_after_change):
21660         Remove function.
21661         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Remove definition.
21663 2017-08-11  Tamar Christina  <tamar.christina@arm.com>
21664         * config/aarch64/aarch64.md (mov<mode>): Change.
21665         (*movhf_aarch64, *movsf_aarch64, *movdf_aarch64):
21666         aarch64_reg_or_fp_float into aarch64_reg_or_fp_zero.
21667         * config/aarch64/predicates.md (aarch64_reg_or_fp_float): Removed.
21669 2017-08-11  Eric Botcazou  <ebotcazou@adacore.com>
21671         * tree-sra.c (build_access_from_expr_1): Use more precise diagnostics
21672         for storage order barriers.
21674 2017-08-11  Martin Liska  <mliska@suse.cz>
21676         PR tree-opt/79987
21677         * tree-chkp.c (chkp_get_bounds_for_decl_addr): Do not instrument
21678         variables of void type.
21680 2017-08-11  Martin Liska  <mliska@suse.cz>
21682         * asan.c (asan_protect_global): Replace ASM_OUTPUT_DEF with
21683         TARGET_SUPPORTS_ALIASES.
21684         * cgraph.c (cgraph_node::create_same_body_alias): Likewise.
21685         * ipa-visibility.c (can_replace_by_local_alias): Likewise.
21686         (optimize_weakref): Likewise.
21687         * symtab.c (symtab_node::noninterposable_alias): Likewise.
21688         * varpool.c (varpool_node::create_extra_name_alias): Likewise.
21689         * defaults.h: Introduce TARGET_SUPPORTS_ALIASES.
21691 2017-08-11  Martin Liska  <mliska@suse.cz>
21693         PR ipa/81213
21694         * config/i386/i386.c (make_resolver_func): Do complete
21695         refactoring of the function.
21697 2017-08-10  Uros Bizjak  <ubizjak@gmail.com>
21699         PR target/81708
21700         * config/i386/i386.opt (mstack-protector-guard-symbol=): New option
21701         * config/i386/i386.c (ix86_stack_protect_guard): Use
21702         ix86_stack_protect_guard_symbol_str to generate varible declaration.
21703         * doc/invoke.texi (x86 Options): Document
21704         -mstack-protector-guard-symbol= option.
21706 2017-08-10  Uros Bizjak  <ubizjak@gmail.com>
21708         * config/i386/i386-protos.h (ix86_split_stack_guard): New prototype.
21709         * config/i386/i386.c (ix86_split_stack_guard): New function.
21710         (ix86_xpand_split_stack_prologue): Call ix86_split_stack_guard.
21711         (ix86_legitimate_address_p) <case UNSPEC_STACK_CHECK>: Remove.
21712         (i386_asm_output_addr_const_extra) <case UNSPEC_STACK_CHECK>: Ditto.
21713         (optput_pic_addr_const): Remove UNSPEC_STACK_CHECK handling.
21714         * config/i386/i386.md (unspec): Remove UNSPEC_STACK_CHECK.
21715         (split_stack_space_check): Call ix86_split_stack_guard.
21717 2017-08-10  Martin Sebor  <msebor@redhat.com>
21719         * print-tree.c (print_node): Print location using the established
21720         format %s:%i%i.
21721         Replace spaces with colons.
21722         (debug_raw, debug): Ditto.
21724 2017-08-10  Martin Sebor  <msebor@redhat.com>
21726         PR c++/81586
21727         * pretty-print.c (pp_format): Correct the handling of %s precision.
21729 2017-08-10  H.J. Lu  <hongjiu.lu@intel.com>
21731         PR target/81736
21732         * config/i386/i386.c (ix86_finalize_stack_realign_flags): Renamed
21733         to ...
21734         (ix86_finalize_stack_frame_flags): This.  Also clear
21735         frame_pointer_needed if -fno-omit-frame-pointer is used without
21736         stack access.
21737         (ix86_expand_prologue): Replace ix86_finalize_stack_realign_flags
21738         with ix86_finalize_stack_frame_flags.
21739         (ix86_expand_epilogue): Likewise.
21740         (ix86_expand_split_stack_prologue): Likewise.
21741         * doc/invoke.texi: Add a note for -fno-omit-frame-pointer.
21743 2017-08-10  Martin Liska  <mliska@suse.cz>
21745         PR c++/81355
21746         * c-attribs.c (handle_target_attribute):
21747         Report warning for an empty string argument of target attribute.
21749 2017-08-09  Jakub Jelinek  <jakub@redhat.com>
21751         PR c/81687
21752         * omp-low.c (omp_copy_decl): Don't remap FORCED_LABEL or DECL_NONLOCAL
21753         LABEL_DECLs.
21754         * tree-cfg.c (move_stmt_op): Don't adjust DECL_CONTEXT of FORCED_LABEL
21755         or DECL_NONLOCAL labels.
21756         (move_stmt_r) <case GIMPLE_LABEL>: Adjust DECL_CONTEXT of FORCED_LABEL
21757         or DECL_NONLOCAL labels here.
21759 2017-08-09  Will Schmidt  <will_schmidt@vnet.ibm.com>
21761         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add blurb
21762         to indicate when early gimple folding has been disabled.
21763         (rs6000_gimple_fold_builtin): Add debug content.
21764         (rs6000_invalid_builtin): Fix whitespace.
21765         (rs6000_expand_builtin): Fix whitespace.
21766         * config/rs6000/rs6000.opt: Add option for -mfold-gimple.
21768 2017-08-09  Segher Boessenkool  <segher@kernel.crashing.org>
21770         PR target/80938
21771         * config/rs6000/rs6000.c (rs6000_savres_strategy): Don't use
21772         SAVE_MULTIPLE if not all the registers that saves, should be saved.
21774 2017-08-09  Jim Wilson  <jim.wilson@linaro.org>
21776         * config/aarch64/aarch64-cores.def (falkor): Use falkor pipeline.
21777         (qdf24xx): Likewise.
21778         * config/aarch64/aarch64.md: Include falkor.md.
21779         * config/aarch64/falkor.md: New.
21781 2017-08-09  Marek Polacek  <polacek@redhat.com>
21783         PR c/81233
21784         * diagnostic-core.h (emit_diagnostic_valist): Add declaration.
21785         * diagnostic.c (emit_diagnostic): Add a comment.
21786         (emit_diagnostic_valist): New function.
21788 2017-08-09  Marek Polacek  <polacek@redhat.com>
21790         PR c/81417
21791         * input.c (make_location): New overload.
21792         * input.h (make_location): Declare.
21794 2017-08-08  Alan Modra  <amodra@gmail.com>
21795             H.J. Lu  <hongjiu.lu@intel.com>
21797         PR driver/81523
21798         * gcc.c (NO_PIE_SPEC): Delete.
21799         (PIE_SPEC): Define as !no-pie/pie.  Move static|shared|r
21800         exclusion..
21801         (LINK_PIE_SPEC): ..to here.
21802         (LINK_COMMAND_SPEC): Support -no-pie.
21803         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Correct
21804         chain of crtbegin*.o selection, update for PIE_SPEC changes and
21805         format.
21806         (GNU_USER_TARGET_ENDFILE_SPEC): Similarly.
21807         * config/sol2.h (STARTFILE_CRTBEGIN_SPEC): Similarly.
21808         (ENDFILE_CRTEND_SPEC): Similarly.
21810 2017-08-08  Uros Bizjak  <ubizjak@gmail.com>
21812         PR target/81708
21813         * config/i386/i386.opt (mstack-protector-guard-reg=): New option
21814         (mstack-protector-guard-offset=): Ditto.
21815         * config/i386/i386.c (ix86_option_override): Handle
21816         -mstack-protector-guard-reg= and -mstack-protector-guard-offset=
21817         options.
21818         (ix86_stack_protect_guard): Use ix86_stack_protect_guard_reg and
21819         ix86_stack_protect_guard_offset variables.
21820         (TARGET_STACK_PROTECT_GUARD): Always define.
21821         * doc/invoke.texi (x86 Options): Document -mstack-protector-guard-reg=
21822         and -mstack-protector-guard-offset= options.
21824 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21826         * tree-ssa-loop-ivopts.c (relate_compare_use_with_all_cands): Handle
21827         boundary case for the last candidate.
21829 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21831         * doc/invoke.texi: Document -ftree-loop-distribution for O3.
21832         * opts.c (default_options_table): Add OPT_ftree_loop_distribution.
21834 2017-08-08  Tamar Christina  <tamar.christina@arm.com>
21836         PR middle-end/19706
21837         * config/aarch64/aarch64.md (xorsign<mode>3): New optabs.
21838         * config/aarch64/aarch64-builtins.c
21839         (aarch64_builtin_vectorized_function): Added CASE_CFN_XORSIGN.
21840         * config/aarch64/aarch64-simd-builtins.def: Added xorsign BINOP.
21841         * config/aarch64/aarch64-simd.md: Added xorsign<mode>3
21843 2017-08-08  Tamar Christina  <tamar.christina@arm.com>
21844             Andrew Pinski <pinskia@gmail.com>
21846         PR middle-end/19706
21847         * internal-fn.def (XORSIGN): New.
21848         * optabs.def (xorsign_optab): New.
21849         * tree-ssa-math-opts.c (is_copysign_call_with_1): New.
21850         (convert_expand_mult_copysign): New.
21851         (pass_optimize_widening_mul::execute): Call
21852         convert_expand_mult_copysign.
21854 2017-08-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21856         PR tree-optimization/81354
21857         * gimple-ssa-strength-reduction.c (create_add_on_incoming_edge):
21858         Insert on edges rather than explicitly creating landing pads.
21859         (analyze_candidates_and_replace): Commit edge inserts.
21861 2017-08-08  Richard Biener  <rguenther@suse.de>
21863         PR middle-end/81719
21864         * tree-ssa-loop-niter.c: Include tree-dfa.h.
21865         (expand_simple_operations): Also look through ADDR_EXPRs with
21866         MEM_REF bases treating them as POINTER_PLUS_EXPR.
21868 2017-08-08  Richard Biener  <rguenther@suse.de>
21870         PR tree-optimization/81723
21871         * tree-vect-slp.c (struct bst_traits): New hash traits.
21872         (bst_fail): New global.
21873         (vect_build_slp_tree_2): New worker, split out from ...
21874         (vect_build_slp_tree): ... this now wrapping it with using
21875         bst_fail set to cache SLP tree build fails.  Properly handle
21876         max_tree_size.
21877         (vect_analyze_slp_instance): Allocate and free bst_fail.
21879 2017-08-08  Martin Liska  <mliska@suse.cz>
21881         PR tree-opt/81696
21882         * ipa-icf-gimple.c (func_checker::compare_cst_or_decl): Consider
21883         LABEL_DECLs that can be from a different function.
21885 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21887         PR tree-optimization/81744
21888         * tree-predcom.c (prepare_finalizers_chain): Deep copy expr of
21889         loop's number of iterations.
21891 2017-08-08  Martin Liska  <mliska@suse.cz>
21893         * asan.c: Include header files.
21894         * attribs.c (build_decl_attribute_variant): New function moved
21895         from tree.[ch].
21896         (build_type_attribute_qual_variant): Likewise.
21897         (cmp_attrib_identifiers): Likewise.
21898         (simple_cst_list_equal): Likewise.
21899         (omp_declare_simd_clauses_equal): Likewise.
21900         (attribute_value_equal): Likewise.
21901         (comp_type_attributes): Likewise.
21902         (build_type_attribute_variant): Likewise.
21903         (lookup_ident_attribute): Likewise.
21904         (remove_attribute): Likewise.
21905         (merge_attributes): Likewise.
21906         (merge_type_attributes): Likewise.
21907         (merge_decl_attributes): Likewise.
21908         (merge_dllimport_decl_attributes): Likewise.
21909         (handle_dll_attribute): Likewise.
21910         (attribute_list_equal): Likewise.
21911         (attribute_list_contained): Likewise.
21912         * attribs.h (lookup_attribute): New function moved from tree.[ch].
21913         (lookup_attribute_by_prefix): Likewise.
21914         * bb-reorder.c: Include header files.
21915         * builtins.c: Likewise.
21916         * calls.c: Likewise.
21917         * cfgexpand.c: Likewise.
21918         * cgraph.c: Likewise.
21919         * cgraphunit.c: Likewise.
21920         * convert.c: Likewise.
21921         * dwarf2out.c: Likewise.
21922         * final.c: Likewise.
21923         * fold-const.c: Likewise.
21924         * function.c: Likewise.
21925         * gimple-expr.c: Likewise.
21926         * gimple-fold.c: Likewise.
21927         * gimple-pretty-print.c: Likewise.
21928         * gimple.c: Likewise.
21929         * gimplify.c: Likewise.
21930         * hsa-common.c: Likewise.
21931         * hsa-gen.c: Likewise.
21932         * internal-fn.c: Likewise.
21933         * ipa-chkp.c: Likewise.
21934         * ipa-cp.c: Likewise.
21935         * ipa-devirt.c: Likewise.
21936         * ipa-fnsummary.c: Likewise.
21937         * ipa-inline.c: Likewise.
21938         * ipa-visibility.c: Likewise.
21939         * ipa.c: Likewise.
21940         * lto-cgraph.c: Likewise.
21941         * omp-expand.c: Likewise.
21942         * omp-general.c: Likewise.
21943         * omp-low.c: Likewise.
21944         * omp-offload.c: Likewise.
21945         * omp-simd-clone.c: Likewise.
21946         * opts-global.c: Likewise.
21947         * passes.c: Likewise.
21948         * predict.c: Likewise.
21949         * sancov.c: Likewise.
21950         * sanopt.c: Likewise.
21951         * symtab.c: Likewise.
21952         * toplev.c: Likewise.
21953         * trans-mem.c: Likewise.
21954         * tree-chkp.c: Likewise.
21955         * tree-eh.c: Likewise.
21956         * tree-into-ssa.c: Likewise.
21957         * tree-object-size.c: Likewise.
21958         * tree-parloops.c: Likewise.
21959         * tree-profile.c: Likewise.
21960         * tree-ssa-ccp.c: Likewise.
21961         * tree-ssa-live.c: Likewise.
21962         * tree-ssa-loop.c: Likewise.
21963         * tree-ssa-sccvn.c: Likewise.
21964         * tree-ssa-structalias.c: Likewise.
21965         * tree-ssa.c: Likewise.
21966         * tree-streamer-in.c: Likewise.
21967         * tree-vectorizer.c: Likewise.
21968         * tree-vrp.c: Likewise.
21969         * tsan.c: Likewise.
21970         * ubsan.c: Likewise.
21971         * varasm.c: Likewise.
21972         * varpool.c: Likewise.
21973         * tree.c: Remove functions moved to attribs.[ch].
21974         * tree.h: Likewise.
21975         * config/aarch64/aarch64.c: Add attrs.h header file.
21976         * config/alpha/alpha.c: Likewise.
21977         * config/arc/arc.c: Likewise.
21978         * config/arm/arm.c: Likewise.
21979         * config/avr/avr.c: Likewise.
21980         * config/bfin/bfin.c: Likewise.
21981         * config/c6x/c6x.c: Likewise.
21982         * config/cr16/cr16.c: Likewise.
21983         * config/cris/cris.c: Likewise.
21984         * config/darwin.c: Likewise.
21985         * config/epiphany/epiphany.c: Likewise.
21986         * config/fr30/fr30.c: Likewise.
21987         * config/frv/frv.c: Likewise.
21988         * config/ft32/ft32.c: Likewise.
21989         * config/h8300/h8300.c: Likewise.
21990         * config/i386/winnt.c: Likewise.
21991         * config/ia64/ia64.c: Likewise.
21992         * config/iq2000/iq2000.c: Likewise.
21993         * config/lm32/lm32.c: Likewise.
21994         * config/m32c/m32c.c: Likewise.
21995         * config/m32r/m32r.c: Likewise.
21996         * config/m68k/m68k.c: Likewise.
21997         * config/mcore/mcore.c: Likewise.
21998         * config/microblaze/microblaze.c: Likewise.
21999         * config/mips/mips.c: Likewise.
22000         * config/mmix/mmix.c: Likewise.
22001         * config/mn10300/mn10300.c: Likewise.
22002         * config/moxie/moxie.c: Likewise.
22003         * config/msp430/msp430.c: Likewise.
22004         * config/nds32/nds32-isr.c: Likewise.
22005         * config/nds32/nds32.c: Likewise.
22006         * config/nios2/nios2.c: Likewise.
22007         * config/nvptx/nvptx.c: Likewise.
22008         * config/pa/pa.c: Likewise.
22009         * config/pdp11/pdp11.c: Likewise.
22010         * config/powerpcspe/powerpcspe.c: Likewise.
22011         * config/riscv/riscv.c: Likewise.
22012         * config/rl78/rl78.c: Likewise.
22013         * config/rx/rx.c: Likewise.
22014         * config/s390/s390.c: Likewise.
22015         * config/sh/sh.c: Likewise.
22016         * config/sol2.c: Likewise.
22017         * config/sparc/sparc.c: Likewise.
22018         * config/spu/spu.c: Likewise.
22019         * config/stormy16/stormy16.c: Likewise.
22020         * config/tilegx/tilegx.c: Likewise.
22021         * config/tilepro/tilepro.c: Likewise.
22022         * config/v850/v850.c: Likewise.
22023         * config/vax/vax.c: Likewise.
22024         * config/visium/visium.c: Likewise.
22025         * config/xtensa/xtensa.c: Likewise.
22027 2017-08-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
22029         PR target/81593
22030         * config/rs6000/vsx.md (vsx_concat_<mode>, VSX_D): Cleanup
22031         constraints since the -mupper-regs-* switches have been
22032         eliminated.
22033         (vsx_concat_<mode>_1): New combiner insns to recognize inserting
22034         into a vector from a double word element that was extracted from
22035         another vector, and eliminate extra XXPERMDI instructions.
22036         (vsx_concat_<mode>_2): Likewise.
22037         (vsx_concat_<mode>_3): Likewise.
22038         (vsx_set_<mode>, VSX_D): Rewrite vector set in terms of vector
22039         concat to allow optimizing inserts from previous extracts.
22041 2017-08-07  Uros Bizjak  <ubizjak@gmail.com>
22043         * config/i386/i386.c (ix86_stack_protect_guard): Generate
22044         memory reference to a SSP offset in TLS address space.
22045         (ix86_print_operand) <case '@'>: Remove.
22046         (ix86_print_operand_punct_valid_p): Remove '@' code.
22047         * config/i386/i386.md (unspec): Remove UNSPEC_SP_TLS_SET and
22048         UNSPEC_SP_TLS_TEST.
22049         (stack_tls_protect_set_<mode>): Remove.
22050         (stack_protect_set): Do not call gen_stack_tls_protect_set_<mode>.
22051         (stack_tls_protect_test_<mode>): Remove.
22052         (stack_protect_test): Do not call gen_stack_tls_protect_test_<mode>.
22054 2017-08-07  Olivier Hainque  <hainque@adacore.com>
22056         PR target/81755
22057         * config/vxworksae.h (VXWORKS_HAVE_TLS): Define.
22059 2017-08-07  Douglas Rupp  <rupp@adacore.com>
22061         * Makefile.in (install-mkheaders): Fix typo, where the multi_dir
22062         variable was referenced as multidir in command.
22064 2017-08-07  Jakub Jelinek  <jakub@redhat.com>
22066         PR c/69389
22067         * gimplify.c (goa_stabilize_expr): Handle BIT_INSERT_EXPR and
22068         BIT_FIELD_REF.
22070 2017-08-07  Martin Liska  <mliska@suse.cz>
22072         * config/m32c/m32c.c: Add include of stringpool.h and attribs.h.
22073         * config/rl78/rl78.c: Add include of attribs.h.
22074         * config/sh/sh.c: Likewise.
22075         * config/v850/v850.c: Likewise.
22077 2017-08-07  Tom de Vries  <tom@codesourcery.com>
22079         PR middle-end/78266
22080         * omp-expand.c (expand_oacc_for): Ensure diff_type is large enough.
22082 2017-08-07  Martin Liska  <mliska@suse.cz>
22084         * config/mips/mips.c: Include attribs.h.
22086 2017-08-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
22088         PR fortran/68829
22089         * doc/invoke.texi: Document change in behvaior for -Ofast for
22090         Fortran.
22092 2017-08-07  Wilco Dijkstra  <wdijkstr@arm.com>
22094         * config/aarch64/aarch64.c (aarch64_pushwb_single_reg):
22095         Use gen_frame_mem.
22096         (aarch64_pop_regs): Likewise.
22097         (aarch64_gen_load_pair): Likewise.
22098         (aarch64_save_callee_saves): Likewise.
22099         (aarch64_restore_callee_saves): Likewise.
22101 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
22103         * config/i386/i386.c: Revert the last change.
22105 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
22107         PR target/81736
22108         * config/i386/i386.c (ix86_finalize_stack_realign_flags): Renamed
22109         to ...
22110         (ix86_finalize_stack_frame_flags): This.  Also clear
22111         frame_pointer_needed if -fno-omit-frame-pointer is used without
22112         stack access.
22113         (ix86_expand_prologue): Replace ix86_finalize_stack_realign_flags
22114         with ix86_finalize_stack_frame_flags.
22115         (ix86_expand_epilogue): Likewise.
22116         (ix86_expand_split_stack_prologue): Likewise.
22118 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
22120         PR target/81743
22121         * config/i386/i386.c (get_builtin_code_for_version): Set priority
22122         to P_AES for Westmere.
22124 2017-08-07  Jonathan Yong  <10walls@gmail.com>
22126         * config/i386/mingw.opt (fset-stack-executable): Removed.
22127         * config/i386/cygming.opt (fset-stack-executable): Moved
22128         from mingw.opt.
22129         * config/i386/cygwin.h: Define CHECK_EXECUTE_STACK_ENABLED.
22131 2017-08-07  Segher Boessenkool  <segher@kernel.crashing.org>
22133         * print-rtl.c (print_exp): Print NOT as "~" instead of as "!".
22135 2017-08-07  Marek Polacek  <polacek@redhat.com>
22137         PR middle-end/81737
22138         * fold-const.c (fold_indirect_ref_1): Check type_domain.
22140 2017-08-07  Martin Liska  <mliska@suse.cz>
22142         * attribs.h (canonicalize_attr_name): New function.
22143         (cmp_attribs): Move from c-format.c and adjusted.
22144         (is_attribute_p): Moved from tree.h.
22145         * tree-inline.c: Add new includes.
22146         * tree.c (cmp_attrib_identifiers): Use cmp_attribs.
22147         (private_is_attribute_p): Remove.
22148         (private_lookup_attribute): Likewise.
22149         (private_lookup_attribute_by_prefix): Simplify.
22150         (remove_attribute): Use is_attribute_p.
22151         * tree.h: Remove removed declarations.
22153 2017-08-07  Jakub Jelinek  <jakub@redhat.com>
22155         PR middle-end/81698
22156         * stmt.c (emit_case_dispatch_table): Add DEFAULT_EDGE argument,
22157         instead of computing it in the function.  Formatting fix.
22158         (expand_case): Don't rely on default_edge being the first edge,
22159         clear it if removing it, pass default_edge to
22160         emit_case_dispatch_table.
22161         (expand_sjlj_dispatch_table): Pass NULL as DEFAULT_EDGE, formatting
22162         fix.
22164 2017-08-06  Uros Bizjak  <ubizjak@gmail.com>
22166         * config/alpha/alpha.c (alpha_reorg): If trap is the last active
22167         insn in the function, emit NOP after the insn.
22169 2017-08-06  Tom de Vries  <tom@codesourcery.com>
22171         * omp-expand.c (expand_oacc_for): Add missing edge probability for tile
22172         and element loops.
22174 2017-08-06  Tom de Vries  <tom@codesourcery.com>
22176         * omp-expand.c (expand_oacc_for): Add missing edge probability for chunk
22177         loop.
22179 2017-08-04  Yury Gribov  <tetra2005@gmail.com>
22181         PR tree-optimization/57371
22182         * match.pd: New pattern.
22184 2017-08-04  Marek Polacek  <polacek@redhat.com>
22186         PR middle-end/81695
22187         * fold-const.c (fold_indirect_ref_1): For ((int *)&a + 4 -> a[1],
22188         perform the computation in offset_int.
22190 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22192         PR tree-optimization/81136
22193         * tree-vectorizer.h: Include tree-hash-traits.h.
22194         (vec_base_alignments): New typedef.
22195         (vec_info): Add a base_alignments field.
22196         (vect_record_base_alignments): Declare.
22197         * tree-data-ref.h (data_reference): Add an is_conditional_in_stmt
22198         field.
22199         (DR_IS_CONDITIONAL_IN_STMT): New macro.
22200         (create_data_ref): Add an is_conditional_in_stmt argument.
22201         * tree-data-ref.c (create_data_ref): Likewise.  Use it to initialize
22202         the is_conditional_in_stmt field.
22203         (data_ref_loc): Add an is_conditional_in_stmt field.
22204         (get_references_in_stmt): Set the is_conditional_in_stmt field.
22205         (find_data_references_in_stmt): Update call to create_data_ref.
22206         (graphite_find_data_references_in_stmt): Likewise.
22207         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Likewise.
22208         * tree-vect-data-refs.c (vect_analyze_data_refs): Likewise.
22209         (vect_record_base_alignment): New function.
22210         (vect_record_base_alignments): Likewise.
22211         (vect_compute_data_ref_alignment): Adjust base_addr and aligned_to
22212         for nested statements even if we fail to compute a misalignment.
22213         Use pooled base alignments for unconditional references.
22214         (vect_find_same_alignment_drs): Compare base addresses instead
22215         of base objects.
22216         (vect_analyze_data_refs_alignment): Call vect_record_base_alignments.
22217         * tree-vect-slp.c (vect_slp_analyze_bb_1): Likewise.
22219 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22221         * tree-vectorizer.h (vec_info): Add a constructor and destructor.
22222         Add an explicit name for the enum.  Use auto_vec for slp_instances
22223         and grouped_stores.
22224         (_loop_vec_info): Add a constructor and destructor.  Use auto_vec
22225         for all vectors.
22226         (_bb_vec_info): Add a constructor and destructor.
22227         (vinfo_for_stmt): Return NULL for uids of -1 as well.
22228         (destroy_loop_vec_info): Delete.
22229         (vect_destroy_datarefs): Likewise.
22230         * tree-vectorizer.c (vect_destroy_datarefs): Delete.
22231         (vec_info::vec_info): New function.
22232         (vec_info::~vec_info): Likewise.
22233         (vectorize_loops): Use delete instead of destroy_loop_vec_info.
22234         * tree-parloops.c (gather_scalar_reductions): Use delete instead of
22235         destroy_loop_vec_info.
22236         * tree-vect-loop.c (new_loop_vec_info): Replace with...
22237         (_loop_vec_info::_loop_vec_info): ...this.
22238         (destroy_loop_vec_info): Replace with...
22239         (_loop_vec_info::~_loop_vec_info): ...this.  Unconditionally delete
22240         the stmt_vec_infos.  Leave handling of vec_info information to its
22241         destructor.  Remove explicit vector releases.
22242         (vect_analyze_loop_form): Use new instead of new_loop_vec_info.
22243         (vect_analyze_loop): Use delete instead of destroy_loop_vec_info.
22244         * tree-vect-slp.c (new_bb_vec_info): Replace with...
22245         (_bb_vec_info::_bb_vec_info): ...this.  Don't reserve space in
22246         BB_VINFO_GROUPED_STORES or BB_VINFO_SLP_INSTANCES.
22247         (destroy_bb_vec_info): Replace with...
22248         (_bb_vec_info::~_bb_vec_info): ...this.  Leave handling of vec_info
22249         information to its destructor.
22250         (vect_slp_analyze_bb_1): Use new and delete instead of
22251         new_bb_vec_info and destroy_bb_vec_info.
22252         (vect_slp_bb): Replace 2 calls to destroy_bb_vec_info with a
22253         single delete.
22255 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22257         * tree-data-ref.h (subscript): Add access_fn field.
22258         (data_dependence_relation): Add could_be_independent_p.
22259         (SUB_ACCESS_FN, DDR_COULD_BE_INDEPENDENT_P): New macros.
22260         (same_access_functions): Move to tree-data-ref.c.
22261         * tree-data-ref.c (ref_contains_union_access_p): New function.
22262         (access_fn_component_p): Likewise.
22263         (access_fn_components_comparable_p): Likewise.
22264         (dr_analyze_indices): Add a reference to access_fn_component_p.
22265         (dump_data_dependence_relation): Use SUB_ACCESS_FN instead of
22266         DR_ACCESS_FN.
22267         (constant_access_functions): Likewise.
22268         (add_other_self_distances): Likewise.
22269         (same_access_functions): Likewise.  (Moved from tree-data-ref.h.)
22270         (initialize_data_dependence_relation): Use XCNEW and remove
22271         explicit zeroing of DDR_REVERSED_P.  Look for a subsequence
22272         of access functions that have the same type.  Allow the
22273         subsequence to end with different bases in some circumstances.
22274         Record the chosen access functions in SUB_ACCESS_FN.
22275         (build_classic_dist_vector_1): Replace ddr_a and ddr_b with
22276         a_index and b_index.  Use SUB_ACCESS_FN instead of DR_ACCESS_FN.
22277         (subscript_dependence_tester_1): Likewise dra and drb.
22278         (build_classic_dist_vector): Update calls accordingly.
22279         (subscript_dependence_tester): Likewise.
22280         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Check
22281         DDR_COULD_BE_INDEPENDENT_P.
22282         * tree-vectorizer.h (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Test
22283         comp_alias_ddrs instead of may_alias_ddrs.
22284         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
22285         New function.
22286         (vect_analyze_data_ref_dependence): Use it if
22287         DDR_COULD_BE_INDEPENDENT_P, but fall back to using the recorded
22288         distance vectors if that fails.
22289         (dependence_distance_ge_vf): New function.
22290         (vect_prune_runtime_alias_test_list): Use it.  Don't clear
22291         LOOP_VINFO_MAY_ALIAS_DDRS.
22293 2017-08-04  Richard Biener  <rguenther@suse.de>
22295         PR middle-end/81705
22296         * fold-const.c (fold_binary_loc): Properly restrict
22297         minus_var0 && minus_var1 case when associating undefined overflow
22298         entities.
22300 2017-08-04  Tom de Vries  <tom@codesourcery.com>
22302         * omp-simd-clone.c (simd_clone_adjust): Add missing edge probability.
22304 2017-08-03  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
22306         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
22307         Don't start diagnostic messages with a capital letter.
22308         * config/rs6000/rs6000.c (rs6000_option_override_internal):
22309         Likewise.
22310         (rs6000_invalid_builtin): Likewise.
22311         (rs6000_trampoline_init): Likewise.
22313 2017-08-03  Jakub Jelinek  <jakub@redhat.com>
22315         PR target/81621
22316         * bb-reorder.c (pass_partition_blocks::execute): Return TODO_df_finish
22317         after setting changeable df flags.
22319 2017-08-03  Richard Biener  <rguenther@suse.de>
22321         * tree-ssa-reassoc.c (should_break_up_subtract): Also break
22322         up if the use is in USE - X.
22324 2017-08-03  Alexander Monakov  <amonakov@ispras.ru>
22326         * toplev.c (dumpfile.h): New include.
22327         (internal_error_reentered): New static function.  Use it...
22328         (internal_error_function): ...here to handle reentered internal_error.
22330 2017-08-03  Richard Biener  <rguenther@suse.de>
22332         PR middle-end/81148
22333         * fold-const.c (split_tree): Add minus_var and minus_con
22334         arguments, remove unused loc arg.  Never generate NEGATE_EXPRs
22335         here but always use minus_*.
22336         (associate_trees): Assert we never associate with MINUS_EXPR
22337         and NULL first operand.  Do not recurse for PLUS_EXPR operands
22338         when associating as MINUS_EXPR either.
22339         (fold_binary_loc): Track minus_var and minus_con.
22341 2017-08-03  Tom de Vries  <tom@codesourcery.com>
22343         PR lto/81430
22344         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers): If
22345         ACCEL_COMPILER, apply finish_options on
22346         DECL_FUNCTION_SPECIFIC_OPTIMIZATION.
22348 2017-08-03  Tom de Vries  <tom@codesourcery.com>
22350         PR target/81662
22351         * config/nvptx/nvptx.c (nvptx_option_override): Emit sorry if
22352         function_entry_patch_area_size > 0.
22354 2017-08-03  Jakub Jelinek  <jakub@redhat.com>
22356         PR driver/81650
22357         * calls.c (alloc_max_size): Use HOST_WIDE_INT_UC (10??)
22358         instead of 10??LU, perform unit multiplication in wide_int,
22359         don't change alloc_object_size_limit if the limit is larger
22360         than SSIZE_MAX.
22362         PR tree-optimization/81655
22363         PR tree-optimization/81588
22364         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Handle also
22365         the case when ranges[i].low and high are 1 for unsigned type with
22366         precision 1.
22368         PR middle-end/81052
22369         * omp-low.c (diagnose_sb_0): Handle flag_openmp_simd like flag_openmp.
22370         (pass_diagnose_omp_blocks::gate): Enable also for flag_openmp_simd.
22372 2017-08-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
22374         * tree-vrp.h: Add include guard.
22376 2017-08-02  Uros Bizjak  <ubizjak@gmail.com>
22378         PR target/81644
22379         * config/i386/i386.md (unspecv): Add UNSPECV_UD2.
22380         (ud2): New insn pattern.
22381         * config/i386/i386.c (ix86_expand_epilogue):
22382         For naked functions, generate ud2 instead of trap insn.
22384 2017-08-02  Marek Polacek  <polacek@redhat.com>
22386         PR other/81667
22387         * alloc-pool.h (base_pool_allocator): Initialize m_elt_size.
22389 2017-08-02  Tom de Vries  <tom@codesourcery.com>
22390             Cesar Philippidis  <cesar@codesourcery.com>
22392         * config/nvptx/nvptx.c (nvptx_lockless_update, nvptx_lockfull_update):
22393         Add missing edge probabilities.
22395 2017-08-02  Tamar Christina  <tamar.christina@arm.com>
22397         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
22398         Correct endianness.
22400 2017-08-02  Jakub Jelinek  <jakub@redhat.com>
22402         PR middle-end/79499
22403         * function.c (thread_prologue_and_epilogue_insns): Determine blocks
22404         for find_many_sub_basic_blocks bitmap by looking up BLOCK_FOR_INSN
22405         of first NONDEBUG_INSN_P in each of the split_prologue_seq and
22406         prologue_seq sequences - if any.
22408 2017-08-02  Richard Biener  <rguenther@suse.de>
22410         * tree-vect-stmts.c (vectorizable_store): Perform vector extracts
22411         via vectors if supported, integer extracts via punning if supported
22412         or otherwise vector extracts.
22414 2017-08-02  Richard Biener  <rguenther@suse.de>
22416         * tree-ssa-pre.c (bitmap_insert_into_set_1): Remove and inline
22417         into ...
22418         (bitmap_insert_into_set): ... this.
22420 2017-08-02  Richard Biener  <rguenther@suse.de>
22422         PR tree-optimization/81633
22423         Revert
22424         2015-08-17  Alan Hayward  <alan.hayward@arm.com>
22426         PR tree-optimization/71752
22427         * tree-vect-slp.c (vect_get_slp_defs): Handle null operands.
22429 2017-08-01  Daniel Santos  <daniel.santos@pobox.com>
22431         * config/i386/i386.h (ix86_frame::outlined_save_offset): Remove field.
22432         (machine_function::call_ms2sysv_pad_out): Remove field.
22433         * config/i386/i386.c (xlogue_layout::get_stack_space_used): Modify.
22434         (ix86_compute_frame_layout): Likewise.
22436 2017-08-01  H.J. Lu  <hongjiu.lu@intel.com>
22438         PR target/81654
22439         * config/i386/i386.c (ix86_set_func_type): Disallow naked
22440         attribute with interrupt attribute.
22442 2017-08-01  Andrew Pinski  <apinski@cavium.com>
22444         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Check
22445         BIT_INSERT_EXPR's operand 1
22446         to see if the types precision matches.
22448 2017-08-01  Martin Liska  <mliska@suse.cz>
22450         PR middle-end/70140
22451         * builtins.c (expand_builtin_memcpy_args): Remove.
22452         (expand_builtin_memcpy): Call newly added function
22453         expand_builtin_memory_copy_args.
22454         (expand_builtin_memcpy_with_bounds): Likewise.
22455         (expand_builtin_mempcpy): Remove last argument.
22456         (expand_builtin_mempcpy_with_bounds): Likewise.
22457         (expand_builtin_memory_copy_args): New function created from
22458         expand_builtin_mempcpy_args with small modifications.
22459         (expand_builtin_mempcpy_args): Remove.
22460         (expand_builtin_stpcpy): Remove unused argument.
22461         (expand_builtin): Likewise.
22462         (expand_builtin_with_bounds): Likewise.
22464 2017-08-01  Martin Liska  <mliska@suse.cz>
22466         Revert r250771
22467         Make mempcpy more optimal (PR middle-end/70140).
22469 2017-08-01  Jakub Jelinek  <jakub@redhat.com>
22471         PR target/81622
22472         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): For
22473         __builtin_vec_cmpne verify both arguments are compatible vectors
22474         before looking at TYPE_MODE on the element type.  For __builtin_vec_ld
22475         verify arg1_type is a pointer or array type.  For __builtin_vec_st,
22476         move computation of aligned to after checking the argument types.
22477         Formatting fixes.
22479         PR target/80846
22480         * config/rs6000/vsx.md (vextract_fp_from_shorth,
22481         vextract_fp_from_shortl): Add element mode after mode in gen_vec_init*
22482         calls.
22484 2017-08-01  Jerome Lambourg  <lambourg@adacore.com>
22485             Doug Rupp  <rupp@adacore.com>
22486             Olivier Hainque  <hainque@adacore.com>
22488         * config.gcc (arm-wrs-vxworks*): Rework to handle arm-wrs-vxworks7 as
22489         well as arm-wrs-vxworks. Update target_cpu_name from arm6 (arch v3) to
22490         arm8 (arch v4).
22491         * config/arm/vxworks.h (MAYBE_TARGET_BPABI_CPP_BUILTINS): New, helper
22492         for TARGET_OS_CPP_BUILTIN.
22493         (TARGET_OS_CPP_BUILTIN): Invoke MAYBE_TARGET_BPABI_CPP_BUILTINS(),
22494         refine CPU definitions for arm_arch5 and add those for arm_arch6 and
22495         arm_arch7.
22496         (MAYBE_ASM_ABI_SPEC): New, helper for SUBTARGET_EXTRA_ASM_SPEC,
22497         passing required abi options to the assembler for EABI configurations.
22498         (EXTRA_CC1_SPEC): New macro, to help prevent the implicit production
22499         of .text.hot and .text.unlikely sections for kernel modules when
22500         using ARM style exceptions.
22501         (CC1_SPEC): Remove obsolete attempt at mimicking Diab toolchain
22502         options. Add EXTRA_CC1_SPEC.
22503         (VXWORKS_ENDIAN_SPEC): Adjust comment and remove handling of Diab
22504         toolchain options.
22505         (DWARF2_UNWIND_INFO): Redefine to handle the pre/post VxWorks 7
22506         transition.
22507         (ARM_TARGET2_DWARF_FORMAT): Define.
22508         * config/arm/t-vxworks: Adjust multilib control to removal of the
22509         Diab command line options.
22511 2017-08-01  Martin Liska  <mliska@suse.cz>
22513         PR gcov-profile/81561
22514         * gcov.c (unblock): Make unblocking safe as we need to preserve
22515         index correspondence of blocks and block_lists.
22517 2017-08-01  Richard Biener  <rguenther@suse.de>
22519         PR tree-optimization/81181
22520         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
22521         (compute_antic): ... end of iteration here.
22523 2017-08-01  James Greenhalgh  <james.greenhalgh@arm.com>
22525         * common.opt (ftree-vectorize): No longer set flag_tree_vectorize.
22526         (ftree-loop-vectorize): Set as EnabledBy ftree-vectorize.
22527         (ftree-slp-vectorize): Likewise.
22528         * omp-expand (expand_omp_simd): Remove flag_tree_vectorize, as it
22529         can no longer be set independent of flag_tree_loop_vectorize.
22530         * omp-general.c (emp_max_vf): Likewise.
22531         * opts.c (enable_fdo_optimizations): Remove references to
22532         flag_tree_vectorize, these are now implicit.
22533         (common_handle_option): Remove handling for OPT_ftree_vectorize,
22534         and leave it for the options machinery.
22536 2017-08-01  Martin Liska  <mliska@suse.cz>
22538         PR middle-end/70140
22539         * builtins.c (expand_builtin_memcpy_args): Remove.
22540         (expand_builtin_memcpy): Call newly added function
22541         expand_builtin_memory_copy_args.
22542         (expand_builtin_memcpy_with_bounds): Likewise.
22543         (expand_builtin_mempcpy): Remove last argument.
22544         (expand_builtin_mempcpy_with_bounds): Likewise.
22545         (expand_builtin_memory_copy_args): New function created from
22546         expand_builtin_mempcpy_args with small modifications.
22547         (expand_builtin_mempcpy_args): Remove.
22548         (expand_builtin_stpcpy): Remove unused argument.
22549         (expand_builtin): Likewise.
22550         (expand_builtin_with_bounds): Likewise.
22552 2017-08-01  Uros Bizjak  <ubizjak@gmail.com>
22554         PR target/81641
22555         * config/i386/i386.c (ix86_print_operand_address_as): For -masm=intel
22556         print "ds:" only for immediates in generic address space.
22558 2017-08-01  Uros Bizjak  <ubizjak@gmail.com>
22560         PR target/81639
22561         * config/i386/i386.c (ix86_funciton_naked): New prototype.
22562         (ix86_function_ok_for_sibcall): Return false for naked functions.
22564 2017-08-01  Richard Biener  <rguenther@suse.de>
22566         * tree-ssa-pre.c (print_pre_expr): Handle NULL expr.
22567         (compute_antic): Seed worklist with exit block predecessors.
22568         * cfganal.c (dfs_find_deadend): For a cycle return the source
22569         of the edge closing it.
22571 2017-08-01  Tamar Christina  <tamar.christina@arm.com>
22573         * config/aarch64/aarch64.c
22574         (aarch64_can_const_movi_rtx_p): Move 0 check.
22576 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22578         * tree.h (POINTER_TYPE_OVERFLOW_UNDEFINED): Delete.
22579         * fold-const.c (fold_comparison, fold_binary_loc): Delete use of
22580         above macro.
22581         * match.pd: Ditto in address comparison pattern.
22583 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22585         PR tree-optimization/81627
22586         * tree-predcom.c (prepare_finalizers): Always rewrite into loop
22587         closed ssa form for store-store chain.
22589 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22591         PR tree-optimization/81620
22592         * tree-predcom.c (add_ref_to_chain): Don't set has_max_use_after
22593         for store-store chain.
22595 2017-08-01  Jakub Jelinek  <jakub@redhat.com>
22597         PR tree-optimization/81588
22598         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If
22599         ranges[i].in_p, invert comparison code ccode.  For >/>=,
22600         swap rhs1 and rhs2 and comparison code unconditionally,
22601         for </<= don't do that.  Don't swap rhs1/rhs2 again if
22602         ranges[i].in_p, instead invert comparison code ccode if
22603         opcode or oe->rank is BIT_IOR_EXPR.
22605         PR target/80846
22606         * optabs.def (vec_extract_optab, vec_init_optab): Change from
22607         a direct optab to conversion optab.
22608         * optabs.c (expand_vector_broadcast): Use convert_optab_handler
22609         with GET_MODE_INNER as last argument instead of optab_handler.
22610         * expmed.c (extract_bit_field_1): Likewise.  Use vector from
22611         vector extraction if possible and optab is available.
22612         * expr.c (store_constructor): Use convert_optab_handler instead
22613         of optab_handler.  Use vector initialization from smaller
22614         vectors if possible and optab is available.
22615         * tree-vect-stmts.c (vectorizable_load): Likewise.
22616         * doc/md.texi (vec_extract, vec_init): Document that the optabs
22617         now have two modes.
22618         * config/i386/i386.c (ix86_expand_vector_init): Handle expansion
22619         of vec_init from half-sized vectors with the same element mode.
22620         * config/i386/sse.md (ssehalfvecmode): Add V4TI case.
22621         (ssehalfvecmodelower, ssescalarmodelower): New mode attributes.
22622         (reduc_plus_scal_v8df, reduc_plus_scal_v4df, reduc_plus_scal_v2df,
22623         reduc_plus_scal_v16sf, reduc_plus_scal_v8sf, reduc_plus_scal_v4sf,
22624         reduc_<code>_scal_<mode>, reduc_umin_scal_v8hi): Add element mode
22625         after mode in gen_vec_extract* calls.
22626         (vec_extract<mode>): Renamed to ...
22627         (vec_extract<mode><ssescalarmodelower>): ... this.
22628         (vec_extract<mode><ssehalfvecmodelower>): New expander.
22629         (rotl<mode>3, rotr<mode>3, <shift_insn><mode>3, ashrv2di3): Add
22630         element mode after mode in gen_vec_init* calls.
22631         (VEC_INIT_HALF_MODE): New mode iterator.
22632         (vec_init<mode>): Renamed to ...
22633         (vec_init<mode><ssescalarmodelower>): ... this.
22634         (vec_init<mode><ssehalfvecmodelower>): New expander.
22635         * config/i386/mmx.md (vec_extractv2sf): Renamed to ...
22636         (vec_extractv2sfsf): ... this.
22637         (vec_initv2sf): Renamed to ...
22638         (vec_initv2sfsf): ... this.
22639         (vec_extractv2si): Renamed to ...
22640         (vec_extractv2sisi): ... this.
22641         (vec_initv2si): Renamed to ...
22642         (vec_initv2sisi): ... this.
22643         (vec_extractv4hi): Renamed to ...
22644         (vec_extractv4hihi): ... this.
22645         (vec_initv4hi): Renamed to ...
22646         (vec_initv4hihi): ... this.
22647         (vec_extractv8qi): Renamed to ...
22648         (vec_extractv8qiqi): ... this.
22649         (vec_initv8qi): Renamed to ...
22650         (vec_initv8qiqi): ... this.
22651         * config/rs6000/vector.md (VEC_base_l): New mode attribute.
22652         (vec_init<mode>): Renamed to ...
22653         (vec_init<mode><VEC_base_l>): ... this.
22654         (vec_extract<mode>): Renamed to ...
22655         (vec_extract<mode><VEC_base_l>): ... this.
22656         * config/rs6000/paired.md (vec_initv2sf): Renamed to ...
22657         (vec_initv2sfsf): ... this.
22658         * config/rs6000/altivec.md (splitter, altivec_copysign_v4sf3,
22659         vec_unpacku_hi_v16qi, vec_unpacku_hi_v8hi, vec_unpacku_lo_v16qi,
22660         vec_unpacku_lo_v8hi, mulv16qi3, altivec_vreve<mode>2): Add
22661         element mode after mode in gen_vec_init* calls.
22662         * config/aarch64/aarch64-simd.md (vec_init<mode>): Renamed to ...
22663         (vec_init<mode><Vel>): ... this.
22664         (vec_extract<mode>): Renamed to ...
22665         (vec_extract<mode><Vel>): ... this.
22666         * config/aarch64/iterators.md (Vel): New mode attribute.
22667         * config/s390/s390.c (s390_expand_vec_strlen, s390_expand_vec_movstr):
22668         Add element mode after mode in gen_vec_extract* calls.
22669         * config/s390/vector.md (non_vec_l): New mode attribute.
22670         (vec_extract<mode>): Renamed to ...
22671         (vec_extract<mode><non_vec_l>): ... this.
22672         (vec_init<mode>): Renamed to ...
22673         (vec_init<mode><non_vec_l>): ... this.
22674         * config/s390/s390-builtins.def (s390_vlgvb, s390_vlgvh, s390_vlgvf,
22675         s390_vlgvf_flt, s390_vlgvg, s390_vlgvg_dbl): Add element mode after
22676         vec_extract mode.
22677         * config/arm/iterators.md (V_elem_l): New mode attribute.
22678         * config/arm/neon.md (vec_extract<mode>): Renamed to ...
22679         (vec_extract<mode><V_elem_l>): ... this.
22680         (vec_extractv2di): Renamed to ...
22681         (vec_extractv2didi): ... this.
22682         (vec_init<mode>): Renamed to ...
22683         (vec_init<mode><V_elem_l>): ... this.
22684         (reduc_plus_scal_<mode>, reduc_plus_scal_v2di, reduc_smin_scal_<mode>,
22685         reduc_smax_scal_<mode>, reduc_umin_scal_<mode>,
22686         reduc_umax_scal_<mode>, neon_vget_lane<mode>, neon_vget_laneu<mode>):
22687         Add element mode after gen_vec_extract* calls.
22688         * config/mips/mips-msa.md (vec_init<mode>): Renamed to ...
22689         (vec_init<mode><unitmode>): ... this.
22690         (vec_extract<mode>): Renamed to ...
22691         (vec_extract<mode><unitmode>): ... this.
22692         * config/mips/loongson.md (vec_init<mode>): Renamed to ...
22693         (vec_init<mode><unitmode>): ... this.
22694         * config/mips/mips-ps-3d.md (vec_initv2sf): Renamed to ...
22695         (vec_initv2sfsf): ... this.
22696         (vec_extractv2sf): Renamed to ...
22697         (vec_extractv2sfsf): ... this.
22698         (reduc_plus_scal_v2sf, reduc_smin_scal_v2sf, reduc_smax_scal_v2sf):
22699         Add element mode after gen_vec_extract* calls.
22700         * config/mips/mips.md (unitmode): New mode iterator.
22701         * config/spu/spu.c (spu_expand_prologue, spu_allocate_stack,
22702         spu_builtin_extract): Add element mode after gen_vec_extract* calls.
22703         * config/spu/spu.md (inner_l): New mode attribute.
22704         (vec_init<mode>): Renamed to ...
22705         (vec_init<mode><inner_l>): ... this.
22706         (vec_extract<mode>): Renamed to ...
22707         (vec_extract<mode><inner_l>): ... this.
22708         * config/sparc/sparc.md (veltmode): New mode iterator.
22709         (vec_init<VMALL:mode>): Renamed to ...
22710         (vec_init<VMALL:mode><VMALL:veltmode>): ... this.
22711         * config/ia64/vect.md (vec_initv2si): Renamed to ...
22712         (vec_initv2sisi): ... this.
22713         (vec_initv2sf): Renamed to ...
22714         (vec_initv2sfsf): ... this.
22715         (vec_extractv2sf): Renamed to ...
22716         (vec_extractv2sfsf): ... this.
22717         * config/powerpcspe/vector.md (VEC_base_l): New mode attribute.
22718         (vec_init<mode>): Renamed to ...
22719         (vec_init<mode><VEC_base_l>): ... this.
22720         (vec_extract<mode>): Renamed to ...
22721         (vec_extract<mode><VEC_base_l>): ... this.
22722         * config/powerpcspe/paired.md (vec_initv2sf): Renamed to ...
22723         (vec_initv2sfsf): ... this.
22724         * config/powerpcspe/altivec.md (splitter, altivec_copysign_v4sf3,
22725         vec_unpacku_hi_v16qi, vec_unpacku_hi_v8hi, vec_unpacku_lo_v16qi,
22726         vec_unpacku_lo_v8hi, mulv16qi3): Add element mode after mode in
22727         gen_vec_init* calls.
22729 2017-08-01  Richard Biener  <rguenther@suse.de>
22731         PR tree-optimization/81297
22732         * tree-vrp.c (get_single_symbol): Remove assert, instead drop
22733         TREE_OVERFLOW from INTEGER_CSTs.
22735 2017-07-31  Segher Boessenkool  <segher@kernel.crashing.org>
22737         * config/rs6000/rs6000.c (enum rs6000_reg_type): Delete trailing comma.
22739 2017-07-31  Carl Love  <cel@us.ibm.com>
22741         * config/rs6000/rs6000-c: Add support for built-in functions
22742         vector signed char vec_xl_be (signed long long, signed char *);
22743         vector unsigned char vec_xl_be (signed long long, unsigned char *);
22744         vector signed int vec_xl_be (signed long long, signed int *);
22745         vector unsigned int vec_xl_be (signed long long, unsigned int *);
22746         vector signed long long vec_xl_be (signed long long, signed long long *);
22747         vector unsigned long long vec_xl_be (signed long long, unsigned long long *);
22748         vector signed short vec_xl_be (signed long long, signed short *);
22749         vector unsigned short vec_xl_be (signed long long, unsigned short *);
22750         vector double vec_xl_be (signed long long, double *);
22751         vector float vec_xl_be (signed long long, float *);
22752         * config/rs6000/altivec.h (vec_xl_be): Add #define.
22753         * config/rs6000/rs6000-builtin.def (XL_BE_V16QI, XL_BE_V8HI, XL_BE_V4SI)
22754         XL_BE_V2DI, XL_BE_V4SF, XL_BE_V2DF, XL_BE): Add definitions
22755         for the builtins.
22756         * config/rs6000/rs6000.c (altivec_expand_xl_be_builtin): Add function.
22757         (altivec_expand_builtin): Add switch statement to call
22758         altivec_expand_xl_be for each builtin.
22759         (altivec_init_builtins): Add def_builtin for _builtin_vsx_le_be_v8hi,
22760         __builtin_vsx_le_be_v4si, __builtin_vsx_le_be_v2di,
22761         __builtin_vsx_le_be_v4sf, __builtin_vsx_le_be_v2df,
22762         __builtin_vsx_le_be_v16qi.
22763         * doc/extend.texi: Update the built-in documentation file for the
22764         new built-in functions.
22766 2017-07-31  Uros Bizjak  <ubizjak@gmail.com>
22768         PR target/25967
22769         * config/i386/i386.c (ix86_allocate_stack_slots_for_args):
22770         New function.
22771         (TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS): Define.
22773 2017-07-31  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
22775         * config.gcc: Add z14.
22776         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
22777         CPU model numbers for z13s and z14.
22778         * config/s390/s390-c.c (s390_resolve_overloaded_builtin): Replace
22779         arch12 with z14.
22780         * config/s390/s390-opts.h (enum processor_type): Rename
22781         PROCESSOR_ARCH12 to PROCESSOR_3906_Z14.
22782         * config/s390/s390.c (processor_table): Add field for CPU name to
22783         be passed to Binutils.
22784         (s390_asm_output_machine_for_arch): Use the new field in
22785         processor_table for Binutils.
22786         (s390_expand_builtin): Replace arch12 with z14.
22787         (s390_issue_rate): Rename PROCESSOR_ARCH12 to PROCESSOR_3906_Z14.
22788         (s390_get_sched_attrmask): Likewise.
22789         (s390_get_unit_mask): Likewise.
22790         * config/s390/s390.opt: Add z14 to processor_type enum.
22792 2017-07-31  Martin Jambor  <mjambor@suse.cz>
22794         PR hsa/81477
22795         * ipa-fnsummary.c (ipa_fn_summary_generate): Set versionable
22796         regardless of optimization level.
22798 2017-07-31  Jan Hubicka <hubicka@ucw.cz>
22799             Martin Liska  <mliska@suse.cz>
22801         * predict.def: Remove old comment and adjust probability.
22802         * gimplify.c (should_warn_for_implicit_fallthrough): Ignore
22803         PREDICT statements.
22805 2017-07-31  Uros Bizjak  <ubizjak@gmail.com>
22807         PR target/25967
22808         * config/i386/i386.c (ix86_function_naked): New function.
22809         (ix86_can_use_return_insn_p): Return false for naked functions.
22810         (ix86_expand_prologue): Skip prologue for naked functions.
22811         (ix86_expand_epilogue): Skip epilogue for naked functions
22812         and emit trap instruction.
22813         (ix86_warn_func_return): New function.
22814         (ix86_attribute_table): Add "naked" attribute specification.
22815         (TARGET_WARN_FUNC_RETURN): Define.
22816         * doc/extend.texi (x86 Function Attributes) <naked>: Document it.
22818 2017-07-31  Martin Liska  <mliska@suse.cz>
22820         * gimple-pretty-print.c (dump_gimple_label): Never dump BB info.
22821         (dump_gimple_bb_header): Always dump BB info.
22822         (pp_cfg_jump): Do not append info about BB when dumping a jump.
22824 2017-07-31  Martin Liska  <mliska@suse.cz>
22826         PR sanitize/81530
22827         * convert.c (convert_to_integer_1): Guard condition with flag_sanitize_p
22828         also with current_function_decl non-null equality.
22830 2017-07-31  Jakub Jelinek  <jakub@redhat.com>
22832         PR sanitizer/81604
22833         * ubsan.c (ubsan_type_descriptor): For UBSAN_PRINT_ARRAY don't
22834         change type to the element type, instead add eltype variable and
22835         use it where we are interested in the element type.
22837         PR tree-optimization/81603
22838         * ipa-polymorphic-call.c
22839         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Perform
22840         offset arithmetic in offset_int, bail out if the resulting bit offset
22841         doesn't fit into shwi.
22843 2017-07-31  Martin Liska  <mliska@suse.cz>
22845         * gimplify.c (mostly_copy_tree_r): Remove Java specific hunk.
22846         (gimplify_save_expr): Fix comment.
22848 2017-07-30  H.J. Lu  <hongjiu.lu@intel.com>
22850         PR target/79793
22851         * config/i386/i386.c (ix86_function_arg): Update arguments for
22852         exception handler.
22853         (ix86_compute_frame_layout): Set the initial stack offset to
22854         INCOMING_FRAME_SP_OFFSET.  Update red-zone offset with
22855         INCOMING_FRAME_SP_OFFSET.
22856         (ix86_expand_epilogue): Don't pop the 'ERROR_CODE' off the
22857         stack before exception handler returns.
22858         * config/i386/i386.h (INCOMING_FRAME_SP_OFFSET): Add the
22859         the 'ERROR_CODE' for exception handler.
22861 2017-07-30  Uros Bizjak  <ubizjak@gmail.com>
22863         * config/i386/i386.h (ASM_PRINTF_EXTENSIONS): New macro.
22864         (ASM_OUTPUT_REG_PUSH): Rewrite with new operand modifiers.
22865         (ASM_OUTPUT_REG_POP): Ditto.
22866         * config/i386/i386.c (ix86_asm_output_function_label): Use fputs
22867         instead of asm_fprintf to output pure string.
22869 2017-07-29  Jakub Jelinek  <jakub@redhat.com>
22871         * debug.h (struct gcc_debug_hooks): Add IMPLICIT argument
22872         to imported_module_or_decl hook.
22873         (debug_nothing_tree_tree_tree_bool): Remove.
22874         (debug_nothing_tree_tree_tree_bool_bool): New declaration.
22875         * debug.c (do_nothing_debug_hooks): Use
22876         debug_nothing_tree_tree_tree_bool_bool instead of
22877         debug_nothing_tree_tree_tree_bool.
22878         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
22879         * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Likewise.
22880         * sdbout.c (sdb_debug_hooks): Likewise.
22881         * dwarf2out.c (dwarf2_lineno_debug_hooks): Likewise.
22882         (gen_namespace_die): Add DW_AT_export_symbols attribute if
22883         langhook wants it.
22884         (dwarf2out_imported_module_or_decl): Add IMPLICIT argument,
22885         if true, -gdwarf-5 and decl will have DW_AT_export_symbols
22886         attribute, don't add anything.
22888 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22890         * fold-const.c (fold_build1_stat_loc): Remove _stat from name.
22891         (fold_build2_stat_loc): Likewise.
22892         (fold_build3_stat_loc): Likewise.
22893         * fold-const.h (fold_build1, fold_build2, fold_build3): Adjust.
22894         (fold_build1_loc): Remove macro.
22895         (fold_build2_loc): Likewise.
22896         (fold_build3_loc): Likewise.
22898 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22900         * gimple.c (gimple_build_debug_bind_stat): Remove _stat from name.
22901         (gimple_build_debug_bind_source_stat): Likewise.
22902         * gimple.h (gimple_build_debug_bind): Remove macro.
22903         (gimple_build_debug_bind_source): Likewise.
22905 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22907         * bitmap.c (bitmap_alloc): Adjust.
22908         (bitmap_gc_alloc): Likewise.
22909         * bitmap.h (bitmap_initialize_stat): Remove _stat from name.
22911 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22913         * bitmap.c (bitmap_obstack_alloc_stat): Rename to bitmap_alloc.
22914         (bitmap_gc_alloc_stat): Rename to bitmap_gc_alloc.
22915         * bitmap.h (bitmap_obstack_alloc_stat): Adjust prototype.
22916         (bitmap_gc_alloc_stat): Likewise.
22917         (BITMAP_ALLOC, BITMAP_GGC_ALLOC): Adjust.
22919 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22921         * rtl.c (shallow_copy_rtx_stat): Remove _stat from name.
22922         * rtl.h (shallow_copy_rtx): Remove macro.
22924 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22926         * emit-rtl.c (gen_raw_REG): Adjust.
22927         * gengenrtl.c (gendef): Likewise.
22928         * rtl.c (rtx_alloc_stat): Remove _stat from name.
22929         * rtl.h (rtx_alloc): Remove macro.
22931 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22933         * tree.c (build_tree_list_vec_stat): Remove _stat from name.
22934         (build_tree_list_stat): Likewise.
22935         * tree.h (build_tree_list): Remove macro.
22936         (build_tree_list_vec): Likewise.
22938 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22940         * tree.c (make_vector_stat): Remove _stat from name.
22941         (build_vector_stat): Likewise.
22942         * tree.h (make_vector_stat): Remove macro.
22943         (build_vector_stat): Likewise.
22945 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22947         * tree.h (build_var_debug_value): Remove prototype.
22949 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22951         * tree.c (tree_cons_stat): Remove _stat from name.
22952         * tree.h (tree_cons): Remove macro.
22954 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22956         * tree.c (build_vl_exp_stat): Remove _stat from name.
22957         * tree.h (build_vl_exp): Remove macro.
22959 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22961         * tree.c (build_decl_stat): Remove _stat from name.
22962         * tree.h (build_decl): Remove macro.
22964 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22966         * gimple.c (gimple_build_with_ops_stat): Adjust.
22967         (gimple_alloc_stat): Remove _stat from name.
22968         * gimple.h (gimple_alloc): Remove macro.
22970 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22972         * tree.c (make_tree_vec_stat): Remove _stat from name.
22973         (grow_tree_vec_stat): Likewise.
22974         * tree.h (make_tree_vec_stat): Adjust prototype.
22975         (grow_tree_vec_stat): Likewise.
22976         (make_tree_vec): Remove macro.
22977         (grow_tree_vec): Likewise.
22979 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22981         * fold-const.c (fold_build1_stat_loc): Adjust.
22982         (fold_build2_stat_loc): Likewise.
22983         (fold_build3_stat_loc): Likewise.
22984         * tree.c (build0_stat): Remove _stat from name.
22985         (build1_stat): Likewise.
22986         (build2_stat): Likewise.
22987         (build3_stat): Likewise.
22988         (build4_stat): Likewise.
22989         (build5_stat): Likewise.
22990         * tree.h (build1_loc): Remove macro, and rename _stat function
22991         to this.
22992         (build2_loc): Likewise.
22993         (build3_loc): Likewise.
22994         (build4_loc): Likewise.
22995         (build5_loc): Likewise.
22997 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22999         * tree.c (make_int_cst_stat): Remove _stat from name.
23000         * tree.h (make_int_cst_stat): Adjust prototype.
23001         (make_int_cst): Remove macro.
23003 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
23005         * tree.c (make_tre_binfo_stat): Remove _stat from name.
23006         * tree.h (make_tree_binfo_stat): Adjust prototype.
23007         (make_tree_binfo): Remove.
23009 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
23011         * tree.c (copy_node_stat): Rename to copy_node.
23012         (build_distinct_type_copy): Adjust.
23013         * tree.h (copy_node_stat): Adjust prototype.
23014         (copy_node): Remove macro.
23016 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
23018         * tree.c (make_node_stat): rename to make_node.
23019         (build_tree_list_stat): Adjust.
23020         (build0_stat): Likewise.
23021         (build2_stat): Likewise.
23022         (build3_stat): Likewise.
23023         (build4_stat): Likewise.
23024         (build5_stat): Likewise.
23025         (build_decl_stat): Likewise.
23026         * tree.h (make_node_stat): Adjust prototype.
23027         (make_node): remove macro.
23029 2017-07-28  Peter Bergner  <bergner@vnet.ibm.com>
23031         * config/rs6000/ppc-auxv.h (PPC_FEATURE2_DARN): New define.
23032         (PPC_FEATURE2_SCV): Likewise.
23033         * config/rs6000/rs6000.c (cpu_supports_info): Use them.
23035 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
23037         * config/aarch64/aarch64.c
23038         (aarch64_internal_mov_immediate): Add new special pattern.
23039         * config/aarch64/aarch64.md (*movdi_aarch64):
23040         Add reg/32bit const mov case.
23042 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
23043             Richard Sandiford <richard.sandiford@linaro.org>
23045         * config/aarch64/aarch64.md (mov<mode>): Generalize.
23046         (*movhf_aarch64, *movsf_aarch64, *movdf_aarch64):
23047         Add integer and movi cases.
23048         (movi-split-hf-df-sf split, fp16): New.
23049         (enabled): Added TARGET_FP_F16INST.
23050         * config/aarch64/iterators.md (GPF_HF): New.
23051         * config/aarch64/predicates.md (aarch64_reg_or_fp_float): New.
23053 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
23055         * config/aarch64/aarch64.c
23056         (aarch64_simd_container_mode): Add prototype.
23057         (aarch64_expand_mov_immediate): Add HI support.
23058         (aarch64_reinterpret_float_as_int, aarch64_float_const_rtx_p: New.
23059         (aarch64_can_const_movi_rtx_p): New.
23060         (aarch64_preferred_reload_class):
23061         Remove restrictions of using FP registers for certain SIMD operations.
23062         (aarch64_rtx_costs): Added new cost for CONST_DOUBLE moves.
23063         (aarch64_valid_floating_const): Add integer move validation.
23064         (aarch64_simd_imm_scalar_p): Remove.
23065         (aarch64_output_scalar_simd_mov_immediate): Generalize function.
23066         (aarch64_legitimate_constant_p): Expand list of supported cases.
23067         * config/aarch64/aarch64-protos.h
23068         (aarch64_float_const_rtx_p, aarch64_can_const_movi_rtx_p): New.
23069         (aarch64_reinterpret_float_as_int): New.
23070         (aarch64_simd_imm_scalar_p): Remove.
23071         * config/aarch64/constraints.md (Uvi): New.
23072         (Dd): Split into Ds and new Dd.
23073         * config/aarch64/aarch64.md (*movsi_aarch64):
23074         Add SIMD mov case.
23075         (*movdi_aarch64): Add SIMD mov case.
23077 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23079         * tree-predcom.c: (struct chain): Handle store-store chain in which
23080         stores for elimination only store loop invariant values.
23081         (execute_pred_commoning_chain): Ditto.
23082         (prepare_initializers_chain_store_elim): Ditto.
23083         (prepare_finalizers): Ditto.
23084         (is_inv_store_elimination_chain): New function.
23085         (initialize_root_vars_store_elim_1): New function.
23087 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23089         * tree-predcom.c: Revise general description of the pass.
23090         (enum chain_type): New enum type for store elimination.
23091         (struct chain): New field supporting store elimination.
23092         (struct component): Ditto.
23093         (dump_chain): Dump store-stores chain.
23094         (release_chain): Release resources.
23095         (split_data_refs_to_components): Compute and create component
23096         contains only stores for elimination.
23097         (get_chain_last_ref_at): New function.
23098         (make_invariant_chain): Initialization.
23099         (make_rooted_chain): Specify chain type in parameter and record it.
23100         (add_looparound_copies): Skip for store-stores chain.
23101         (determine_roots_comp): Compute type of chain and pass it to
23102         make_rooted_chain.
23103         (initialize_root_vars_store_elim_2): New function.
23104         (finalize_eliminated_stores): New function.
23105         (remove_stmt): Handle store for elimination.
23106         (execute_pred_commoning_chain): Execute predictive commoning on
23107         store-store chains.
23108         (determine_unroll_factor): Skip unroll for store-stores chain.
23109         (prepare_initializers_chain_store_elim): New function.
23110         (prepare_initializers_chain): Hanlde store-store chain.
23111         (prepare_finalizers_chain, prepare_finalizers): New function.
23112         (tree_predictive_commoning_loop): Return integer value indicating
23113         if loop is unrolled or lcssa form is corrupted.
23114         (tree_predictive_commoning): Rewrite for lcssa form if necessary.
23116 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23118         * tree-predcom.c (initialize_root): Delete.
23119         (execute_pred_commoning_chain): Initialize root vars and replace
23120         reference of non-combined chain directly, rather than call above
23121         function.
23123 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23125         * tree-predcom.c (ref_at_iteration): Add parameter NITERS.  Compute
23126         memory reference to DR at (NITERS + ITERS)-th iteration of loop.
23128 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23130         * tree-predcom.c (struct chain): New field init_seq.
23131         (release_chain): Release init_seq.
23132         (prepare_initializers_chain): Record intialization stmts in above
23133         field.
23134         (insert_init_seqs): New function.
23135         (tree_predictive_commoning_loop): Call insert_init_seqs.
23137 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23139         * tree-predcom.c (determine_roots_comp): Skip trivial components.
23141 2017-07-28  Richard Biener  <rguenther@suse.de>
23143         * match.pd: Remove superfluous :c.
23144         * genmatch.c (simplify::id): Add member.
23145         (lower_commutative, lower_opt_convert, lower_cond, lower_for):
23146         Copy id.
23147         (current_id): New global.
23148         (dt_node::parent): Move from ...
23149         (dt_operand::parent): ... here.  Add for_id member.
23150         (is_a_helper <dt_operand *>::test): DT_TRUE is also a dt_operand.
23151         (decision_tree::find_node): Relax order requirement when
23152         merging DT_TRUE nodes to ones inbetween the current simplify
23153         and the one we try to merge with.  Add diagnostic whenever
23154         we need to enforce pattern order by not merging.
23155         (decision_tree::insert): Set current_id.
23156         (decision_tree::print_node): Dump parent node and for_id.
23157         (parser::last_id): Add member.
23158         (parser::push_simplify): Assign unique id.
23159         (parser::parser): Initialize last_id.
23161 2017-07-28  Martin Liska  <mliska@suse.cz>
23163         PR sanitizer/81340
23164         * sanopt.c (sanitize_rewrite_addressable_params): Set VALUE_EXPR after
23165         gimple_build_debug_bind.
23167 2017-07-28  Richard Biener  <rguenther@suse.de>
23169         PR tree-optimization/81502
23170         * match.pd: Add pattern combining BIT_INSERT_EXPR with
23171         BIT_FIELD_REF.
23172         * tree-cfg.c (verify_expr): Verify types of BIT_FIELD_REF
23173         size/pos operands.
23174         (verify_gimple_assign_ternary): Likewise for BIT_INSERT_EXPR pos.
23175         * gimple-fold.c (maybe_canonicalize_mem_ref_addr): Use bitsizetype
23176         for BIT_FIELD_REF args.
23177         * fold-const.c (make_bit_field_ref): Likewise.
23178         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
23180 2017-07-28  Jakub Jelinek  <jakub@redhat.com>
23182         PR sanitizer/80998
23183         * sanopt.c (pass_sanopt::execute): Handle IFN_UBSAN_PTR.
23184         * tree-ssa-alias.c (call_may_clobber_ref_p_1): Likewise.
23185         * flag-types.h (enum sanitize_code): Add SANITIZER_POINTER_OVERFLOW.
23186         Or it into SANITIZER_UNDEFINED.
23187         * ubsan.c: Include gimple-fold.h and varasm.h.
23188         (ubsan_expand_ptr_ifn): New function.
23189         (instrument_pointer_overflow): New function.
23190         (maybe_instrument_pointer_overflow): New function.
23191         (instrument_object_size): Formatting fix.
23192         (pass_ubsan::execute): Call instrument_pointer_overflow
23193         and maybe_instrument_pointer_overflow.
23194         * internal-fn.c (expand_UBSAN_PTR): New function.
23195         * ubsan.h (ubsan_expand_ptr_ifn): Declare.
23196         * sanitizer.def (__ubsan_handle_pointer_overflow,
23197         __ubsan_handle_pointer_overflow_abort): New builtins.
23198         * tree-ssa-tail-merge.c (merge_stmts_p): Handle IFN_UBSAN_PTR.
23199         * internal-fn.def (UBSAN_PTR): New internal function.
23200         * opts.c (sanitizer_opts): Add pointer-overflow.
23201         * lto-streamer-in.c (input_function): Handle IFN_UBSAN_PTR.
23202         * fold-const.c (build_range_check): Compute pointer range check in
23203         integral type if pointer arithmetics would be needed.  Formatting
23204         fixes.
23206 2017-07-28  Martin Liska  <mliska@suse.cz>
23208         PR sanitizer/81460
23209         * sanopt.c (sanitize_rewrite_addressable_params): Do not rewrite
23210         parameters that are of a variable-length.
23212 2017-07-28  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23214         * config.gcc (powerpc-*-rtems*): Remove rs6000/eabi.h.  Add
23215         rs6000/biarch64.h.
23216         * config/rs6000/rtems.h (ASM_DECLARE_FUNCTION_SIZE): New macro.
23217         (ASM_OUTPUT_SPECIAL_POOL_ENTRY_P): Likewise.
23218         (CRT_CALL_STATIC_FUNCTION): Likewise.
23219         (ASM_DEFAULT_SPEC): New define.
23220         (ASM_SPEC32): Likewise.
23221         (ASM_SPEC64): Likewise.
23222         (ASM_SPEC_COMMON): Likewise.
23223         (ASM_SPEC): Likewise.
23224         (INVALID_64BIT): Likewise.
23225         (LINK_OS_DEFAULT_SPEC): Likewise.
23226         (LINK_OS_SPEC32): Likewise.
23227         (LINK_OS_SPEC64): Likewise.
23228         (POWERPC_LINUX): Likewise.
23229         (PTRDIFF_TYPE): Likewise.
23230         (RESTORE_FP_PREFIX): Likewise.
23231         (RESTORE_FP_SUFFIX): Likewise.
23232         (SAVE_FP_PREFIX): Likewise.
23233         (SAVE_FP_SUFFIX): Likewise.
23234         (SIZE_TYPE): Likewise.
23235         (SUBSUBTARGET_OVERRIDE_OPTIONS): Likewise.
23236         (TARGET_64BIT): Likewise.
23237         (TARGET_64BIT): Likewise.
23238         (TARGET_AIX): Likewise.
23239         (WCHAR_TYPE_SIZE): Likewise.
23240         (WCHAR_TYPE): Undefine.
23241         (TARGET_OS_CPP_BUILTINS): Add 64-bit PowerPC defines.
23242         (CPP_OS_DEFAULT_SPEC): Use previous CPP_OS_RTEMS_SPEC.
23243         (CPP_OS_RTEMS_SPEC): Delete.
23244         (SUBSUBTARGET_EXTRA_SPECS): Remove cpp_os_rtems.  Add
23245         asm_spec_common, asm_spec32, asm_spec64, link_os_spec32, and
23246         link_os_spec64.
23247         * config/rs6000/t-rtems: Add mcpu=e6500/m64 multilibs.
23249 2017-07-28  Jakub Jelinek  <jakub@redhat.com>
23251         PR tree-optimization/81578
23252         * tree-parloops.c (build_new_reduction): Bail out if
23253         reduction_code isn't one of the standard OpenMP reductions.
23254         Move the details printing after that decision.
23256 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23258         * config/rs6000/predicates.md (volatile_mem_operand): Remove code
23259         related to reload_in_progress.
23260         (splat_input_operand): Likewise.
23261         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_rtx):
23262         Delete prototype.
23263         * config/rs6000/rs6000.c (machine_function): Remove sdmode_stack_slot
23264         field.
23265         (TARGET_EXPAND_TO_RTL_HOOK): Delete.
23266         (TARGET_INSTANTIATE_DECLS): Likewise.
23267         (legitimate_indexed_address_p): Delete reload_in_progress code.
23268         (rs6000_debug_legitimate_address_p): Likewise.
23269         (rs6000_eliminate_indexed_memrefs): Likewise.
23270         (rs6000_emit_le_vsx_store): Likewise.
23271         (rs6000_emit_move_si_sf_subreg): Likewise.
23272         (rs6000_emit_move): Likewise.
23273         (register_to_reg_type): Likewise.
23274         (rs6000_pre_atomic_barrier): Likewise.
23275         (rs6000_machopic_legitimize_pic_address): Likewise.
23276         (rs6000_allocate_stack_temp): Likewise.
23277         (rs6000_address_for_fpconvert): Likewise.
23278         (rs6000_address_for_altivec): Likewise.
23279         (rs6000_secondary_memory_needed_rtx): Delete function.
23280         (rs6000_check_sdmode): Likewise.
23281         (rs6000_alloc_sdmode_stack_slot): Likewise.
23282         (rs6000_instantiate_decls): Likewise.
23283         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED_RTX): Delete.
23284         * config/rs6000/rs6000.md (splitter for *movsi_got_internal):
23285         Delete reload_in_progress.
23286         (*vec_reload_and_plus_<mptrsize>): Likewise.
23287         * config/rs6000/vsx.md (vsx_mul_v2di): Likewise.
23288         (vsx_div_v2di): Likewise.
23289         (vsx_udiv_v2di): Likewise.
23291 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23293         * config/rs6000/rs6000.opt (mlra): Replace with stub.
23294         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete OPTION_MASK_LRA.
23295         * config/rs6000/rs6000.c (TARGET_LRA_P): Delete.
23296         (rs6000_debug_reg_global): Delete print of LRA status.
23297         (rs6000_option_override_internal): Delete dead LRA related code.
23298         (rs6000_lra_p): Delete function.
23299         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mlra.
23301 2017-07-27  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23303         * config.gcc (riscv*-*-elf*): Add (riscv*-*-rtems*).
23304         * config/riscv/rtems.h: New file.
23306 2017-07-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23307             Sudakshina Das  <sudi.das@arm.com>
23309         * config/aarch64/aarch64.md
23310         (define_split for and<mode>3nr_compare): Move
23311         non aarch64_logical_operand to a register.
23312         (define_split for and_<SHIFT:optab><mode>3nr_compare0): Move non
23313         register immediate operand to a register.
23314         * config/aarch64/predicates.md (aarch64_mov_imm_operand): New.
23316 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23318         PR middle-end/81564
23319         * tree-cfg.c (group_case_labels_stmt): Handle already deleted blocks.
23321 2017-07-27  Richard Biener  <rguenther@suse.de>
23323         PR tree-optimization/81573
23324         PR tree-optimization/81494
23325         * tree-vect-loop.c (vect_create_epilog_for_reduction): Handle
23326         multi defuse cycle case.
23328 2017-07-27  Richard Biener  <rguenther@suse.de>
23330         PR tree-optimization/81571
23331         * tree-vect-slp.c (vect_build_slp_tree): Properly verify reduction
23332         PHIs.
23334 2017-07-27  Eric Botcazou  <ebotcazou@adacore.com>
23336         * config/sparc/sparc.c (sparc_option_override): Set MASK_FSMULD flag
23337         earlier and only if MASK_FPU is set.  Adjust formatting.
23339 2017-07-27  Martin Liska  <mliska@suse.cz>
23341         * opt-functions.awk: Add validation of value of Init.
23342         * optc-gen.awk: Pass new argument.
23344 2017-07-27  Martin Liska  <mliska@suse.cz>
23346         * auto-profile.c (autofdo_source_profile::update_inlined_ind_target):
23347         Fix wrong condition.
23349 2017-07-27  Martin Liska  <mliska@suse.cz>
23351         * auto-profile.c (afdo_annotate_cfg): Assign zero counts to
23352         BBs and edges seen by autoFDO.
23354 2017-07-27  Richard Biener  <rguenther@suse.de>
23356         PR tree-optimization/81502
23357         * tree-ssa.c (non_rewritable_lvalue_p): Handle BIT_INSERT_EXPR
23358         with incompatible but same sized type.
23359         (execute_update_addresses_taken): Likewise.
23361 2017-07-27  James Greenhalgh  <james.greenhalgh@arm.com>
23363         * tree-ssa-loop-ch.c (pass_ch::process_loop_p): Guard on
23364         flag_tree_loop_vectorize rather than flag_tree_vectorize.
23366 2017-07-27  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
23368         PR target/81534
23369         * config/s390/s390.md ("*atomic_compare_and_swap<mode>_1")
23370         ("*atomic_compare_and_swapdi_2", "*atomic_compare_and_swapsi_3"):
23371         Change s_operand to memory_operand.
23373 2017-07-27  Richard Sandiford  <richard.sandiford@linaro.org>
23375         * config/rs6000/rs6000-protos.h (rs6000_emit_le_vsx_permute): Declare.
23376         * config/rs6000/rs6000.c (rs6000_gen_le_vsx_permute): Replace with...
23377         (rs6000_emit_le_vsx_permute): ...this.  Take the destination as input.
23378         Emit instructions rather than returning an expression.  Handle TFmode
23379         and KFmode by casting to TImode.
23380         (rs6000_emit_le_vsx_load): Update to use rs6000_emit_le_vsx_permute.
23381         (rs6000_emit_le_vsx_store): Likewise.
23382         * config/rs6000/vsx.md (VSX_TI): New iterator.
23383         (*vsx_le_permute_<mode>): Use it instead of VSX_LE_128.
23384         (*vsx_le_undo_permute_<mode>): Likewise.
23385         (*vsx_le_perm_load_<mode>): Use rs6000_emit_le_vsx_permute to
23386         emit the split sequence.
23387         (*vsx_le_perm_store_<mode>): Likewise.
23389 2017-07-27  Jakub Jelinek  <jakub@redhat.com>
23391         PR tree-optimization/81555
23392         PR tree-optimization/81556
23393         * tree-ssa-reassoc.c (rewrite_expr_tree): Add NEXT_CHANGED argument,
23394         if true, force CHANGED for the recursive invocation.
23395         (reassociate_bb): Remember original length of ops array, pass
23396         len != orig_len as NEXT_CHANGED in rewrite_expr_tree call.
23398         * attribs.c (decl_attributes): Imply noinline, noclone and no_icf
23399         attributes for noipa attribute.  For naked attribute use
23400         lookup_attribute first before lookup_attribute_spec.
23401         * final.c (rest_of_handle_final): Disable IPA RA for functions with
23402         noipa attribute.
23403         * ipa-visibility.c (non_local_p): Fix comment typos.  Return true
23404         for functions with noipa attribute.
23405         (cgraph_externally_visible_p): Return true for functions with noipa
23406         attribute.
23407         * cgraph.c (cgraph_node::get_availability): Return AVAIL_INTERPOSABLE
23408         for functions with noipa attribute.
23409         * doc/extend.texi: Document noipa function attribute.
23410         * tree-ssa-structalias.c (refered_from_nonlocal_fn): Set *nonlocal_p
23411         also for functions with noipa attribute.
23412         (ipa_pta_execute): Set nonlocal_p also for nodes with noipa attribute.
23414 2017-07-26  Andrew Pinski  <apinski@cavium.com>
23416         * config/aarch64/aarch64.c (thunderx_vector_cost): Decrease cost of
23417         vec_unalign_load_cost and vec_unalign_store_cost.
23419 2017-07-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
23421         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Delete
23422         -mvsx-small-integer option.
23423         (ISA_3_0_MASKS_IEEE): Likewise.
23424         (OTHER_VSX_VECTOR_MASKS): Likewise.
23425         (POWERPC_MASKS): Likewise.
23426         * config/rs6000/rs6000.opt (-mvsx-small-integer): Likewise.
23427         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Simplify
23428         code, only testing for DImode being allowed in non-VSX floating
23429         point registers.
23430         (rs6000_init_hard_regno_mode_ok): Change TARGET_VSX_SMALL_INTEGER
23431         to TARGET_P8_VECTOR test.  Remove redundant VSX test inside of
23432         another VSX test.
23433         (rs6000_option_override_internal): Delete -mvsx-small-integer.
23434         (rs6000_expand_vector_set): Change TARGET_VSX_SMALL_INTEGER to
23435         TARGET_P8_VECTOR test.
23436         (rs6000_secondary_reload_simple_move): Likewise.
23437         (rs6000_preferred_reload_class): Delete TARGET_VSX_SMALL_INTEGER,
23438         since TARGET_P9_VECTOR was already tested.
23439         (rs6000_opt_masks): Remove -mvsx-small-integer.
23440         * config/rs6000/vsx.md (vsx_extract_<mode>): Delete
23441         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23442         used.
23443         (vsx_extract_<mode>_p9): Delete TARGET_VSX_SMALL_INTEGER, since a
23444         test for TARGET_VEXTRACTUB was used, and that uses
23445         TARGET_P9_VECTOR.
23446         (p9 extract splitter): Likewise.
23447         (vsx_extract_<mode>_di_p9): Likewise.
23448         (vsx_extract_<mode>_store_p9): Likewise.
23449         (vsx_extract_si): Delete TARGET_VSX_SMALL_INTEGER, since a test
23450         for TARGET_P9_VECTOR was used.  Delete code that is now dead with
23451         the elimination of TARGET_VSX_SMALL_INTEGER.
23452         (vsx_extract_<mode>_p8): Likewise.
23453         (vsx_ext_<VSX_EXTRACT_I:VS_scalar>_fl_<FL_CONV:mode>): Likewise.
23454         (vsx_ext_<VSX_EXTRACT_I:VS_scalar>_ufl_<FL_CONV:mode>): Likewise.
23455         (vsx_set_<mode>_p9): Likewise.
23456         (vsx_set_v4sf_p9): Likewise.
23457         (vsx_set_v4sf_p9_zero): Likewise.
23458         (vsx_insert_extract_v4sf_p9): Likewise.
23459         (vsx_insert_extract_v4sf_p9_2): Likewise.
23460         * config/rs6000/rs6000.md (sign extend splitter): Change
23461         TARGET_VSX_SMALL_INTEGER to TARGET_P8_VECTOR test.
23462         (floatsi<mode>2_lfiwax_mem): Likewise.
23463         (floatunssi<mode>2_lfiwzx_mem): Likewise.
23464         (float<QHI:mode><FP_ISA3:mode>2): Delete TARGET_VSX_SMALL_INTEGER,
23465         since a test for TARGET_P9_VECTOR was used.
23466         (float<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23467         (floatuns<QHI:mode><FP_ISA3:mode>2): Likewise.
23468         (floatuns<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23469         (fix_trunc<mode>si2): Change TARGET_VSX_SMALL_INTEGER to
23470         TARGET_P8_VECTOR test.
23471         (fix_trunc<mode>si2_stfiwx): Likewise.
23472         (fix_trunc<mode>si2_internal): Likewise.
23473         (fix_trunc<SFDF:mode><QHI:mode>2): Delete
23474         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23475         used.
23476         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
23477         (fixuns_trunc<mode>si2): Change TARGET_VSX_SMALL_INTEGER to
23478         TARGET_P8_VECTOR test.
23479         (fixuns_trunc<mode>si2_stfiwx): Likewise.
23480         (fixuns_trunc<SFDF:mode><QHI:mode>2): Delete
23481         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23482         used.
23483         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
23484         (fctiw<u>z_<mode>_smallint): Delete TARGET_VSX_SMALL_INTEGER,
23485         since a test for TARGET_P9_VECTOR was used.
23486         (splitter for loading small constants): Likewise.
23488 2017-07-26  Andrew Pinski  <apinski@cavium.com>
23490         * config/aarch64/aarch64.c (thunderx_vector_cost): Fix
23491         vec_fp_stmt_cost.
23493 2017-07-26  H.J. Lu  <hongjiu.lu@intel.com>
23495         PR target/81563
23496         * config/i386/i386.c (sp_valid_at): Properly check CFA offset.
23497         (fp_valid_at): Likewise.
23499 2017-07-26  James Greenhalgh  <james.greenhalgh@arm.com>
23501         * config/aarch64/aarch64.c (cortexa57_addrcost_table): Remove.
23502         (qdf24xx_addrcost_table): Likewise.
23503         (cortexa57_tunings): Update to use generic_branch_cost.
23504         (cortexa72_tunings): Likewise.
23505         (cortexa73_tunings): Likewise.
23506         (qdf24xx_tunings): Likewise.
23508 2017-07-26  James Greenhalgh  <james.greenhalgh@arm.com>
23510         * config/aarch64/aarch64.c (cortexa57_branch_cost): Remove.
23511         (thunderx2t99_branch_cost): Likewise.
23512         (cortexa35_tunings): Update to use generic_branch_cost.
23513         (cortexa53_tunings): Likewise.
23514         (cortexa57_tunings): Likewise.
23515         (cortexa72_tunings): Likewise.
23516         (cortexa73_tunings): Likewise.
23517         (thunderx2t99_tunings): Likewise.
23519 2017-07-26  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23521         * config/sparc/sparc.c (dump_target_flag_bits): Dump MASK_FSMULD.
23522         (sparc_option_override): Honour MASK_FSMULD.
23523         * config/sparc/sparc.h (MASK_FEATURES): Add MASK_FSMULD.
23524         * config/sparc/sparc.md (muldf3_extend): Use TARGET_FSMULD.
23525         * config/sparc/sparc.opt (mfsmuld): New option.
23526         * doc/invoke.texi (mfsmuld): Document option.
23528 2017-07-26  Marek Polacek  <polacek@redhat.com>
23530         PR middle-end/70992
23531         * tree.c (build2_stat): Don't set TREE_CONSTANT on divisions by zero.
23533 2017-07-26  Richard Biener  <rguenther@suse.de>
23535         * gimple-match-head.c (do_valueize): Return OP if valueize
23536         returns NULL_TREE.
23537         (get_def): New helper to get at the def stmt of a SSA name
23538         if valueize allows.
23539         * genmatch.c (dt_node::gen_kids_1): Use get_def instead of
23540         do_valueize to get at the def stmt.
23541         (dt_operand::gen_gimple_expr): Simplify do_valueize calls.
23543 2017-07-26  Wilco Dijkstra  <wdijkstr@arm.com>
23545         PR middle-end/46932
23546         * auto-inc-dec.c (parse_add_or_inc): Block autoinc on sfp.
23548 2017-07-26  Martin Liska  <mliska@suse.cz>
23550         PR sanitize/81186
23551         * function.c (expand_function_start): Make expansion of
23552         nonlocal_goto_save_area after parm_birth_insn.
23554 2017-07-26  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23556         * config/sparc/sparc.c (sparc_option_override): Remove MASK_FPU
23557         from all CPU target flags enable members.
23559 2017-07-26  Richard Biener  <rguenther@suse.de>
23561         * genmatch.c (dt_simplify::gen): Make iterator vars const.
23562         (decision_tree::gen): Make 'type' const.
23563         (write_predicate): Likewise.
23565 2017-07-24  Michael Meissner  <meissner@linux.vnet.ibm.com>
23567         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok):
23568         Eliminate TARGET_UPPER_REGS_{DF,DI,SF} usage.
23569         (rs6000_option_override_internal): Likewise.
23570         (rs6000_expand_vector_set): Likewise.
23571         * config/rs6000/rs6000.h (TARGET_UPPER_REGS_DF): Delete.
23572         (TARGET_UPPER_REGS_SF): Likewise.
23573         (TARGET_UPPER_REGS_DI): Likewise.
23574         (TARGET_VEXTRACTUB): Eliminate TARGET_UPPER_REGS_{DF,DI,SF}.
23575         (TARGET_DIRECT_MOVE_64BIT): Likewise.
23576         * config/rs6000/rs6000.md (ALTIVEC_DFORM): Likewise.
23577         (float<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23578         (Splitters for DI constants in Altivec registers): Likewise.
23579         * config/rs6000/vsx.md (vsx_set_<mode>_p9): Likewise.
23580         (vsx_set_v4sf_p9): Likewise.
23581         (vsx_set_v4sf_p9_zero): Likewise.
23582         (vsx_insert_extract_v4sf_p9): Likewise.
23583         (vsx_insert_extract_v4sf_p9_2): Likewise.
23585 2017-07-25  Carl Love  <cel@us.ibm.com>
23587         * doc/extend.texi: Update the built-in documentation file for the
23588         existing built-in functions
23589         vector signed char vec_cnttz (vector signed char);
23590         vector unsigned char vec_cnttz (vector unsigned char);
23591         vector signed short vec_cnttz (vector signed short);
23592         vector unsigned short vec_cnttz (vector unsigned short);
23593         vector signed int vec_cnttz (vector signed int);
23594         vector unsigned int vec_cnttz (vector unsigned int);
23595         vector signed long long vec_cnttz (vector signed long long);
23596         vector unsigned long long vec_cnttz (vector unsigned long long);
23598 2017-07-25  Andrew Pinski  <apinski@cavium.com>
23600         * tree-ssa-uninit.c (warn_uninitialized_vars): Don't warn about memory
23601         accesses where the use is for the first operand of a BIT_INSERT.
23603 2017-07-25  Jim Wilson  <jim.wilson@linaro.org>
23605         PR bootstrap/81521
23606         * config/i386/winnt-cxx.c (i386_pe_adjust_class_at_definition): Look
23607         for FUNCTION_DECLs in TYPE_FIELDS rather than TYPE_METHODS.
23609 2017-07-25  Jim Wilson  <jim.wilson@linaro.org>
23611         * config/i386/gstabs.h: Delete.
23612         * config/i386/openbsd.h, config/i386/t-openbsd: Likewise.
23614 2017-07-25  Uros Bizjak  <ubizjak@gmail.com>
23616         * config/i386/i386.c (ix86_decompose_address): Do not check for
23617         register RTX when looking at index_reg or base_reg.
23618         * config/i386/i386.h (INCOMING_RETURN_ADDR_RTX): Use stack_pointer_rtx.
23620 2017-07-25  Eric Botcazou  <ebotcazou@adacore.com>
23622         * gimple.c (gimple_assign_set_rhs_with_ops): Do not ask gsi_replace
23623         to update EH info here.
23625 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23627         * match.pd ((X * CST1) * CST2): Simplify to X * (CST1 * CST2).
23629 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23631         * match.pd ((X * CST) * Y): Reassociate to (X * Y) * CST.
23633 2017-07-25  Torsten Duwe  <duwe@suse.de>
23635         * common.opt: Introduce -fpatchable-function-entry
23636         command line option, and its variables function_entry_patch_area_size
23637         and function_entry_patch_area_start.
23638         * opts.c (common_handle_option): Add -fpatchable_function_entry_ case,
23639         including a two-value parser.
23640         * target.def (print_patchable_function_entry): New target hook.
23641         * targhooks.h (default_print_patchable_function_entry): New function.
23642         * targhooks.c (default_print_patchable_function_entry): Likewise.
23643         * toplev.c (process_options): Switch off IPA-RA if
23644         patchable function entries are being generated.
23645         * varasm.c (assemble_start_function): Look at the
23646         patchable-function-entry command line switch and current
23647         function attributes and maybe generate NOP instructions by
23648         calling the print_patchable_function_entry hook.
23649         * doc/extend.texi: Document patchable_function_entry attribute.
23650         * doc/invoke.texi: Document -fpatchable_function_entry
23651         command line option.
23652         * doc/tm.texi.in (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY):
23653         New target hook.
23654         * doc/tm.texi: Re-generate.
23656 2017-07-25  Jakub Jelinek  <jakub@redhat.com>
23658         PR target/81532
23659         * config/i386/constraints.md (Yd, Ye): Use ALL_SSE_REGS for
23660         TARGET_AVX512DQ rather than TARGET_AVX512BW.
23662 2017-07-25  Tamar Christina  <tamar.christina@arm.com>
23664         * config/arm/parsecpu.awk (all_cores): Remove duplicates.
23666 2017-07-25  Richard Biener  <rguenther@suse.de>
23668         PR tree-optimization/81455
23669         * tree-ssa-loop-unswitch.c (find_loop_guard): Make sure to
23670         not walk in cycles when looking for guards.
23672 2017-07-25  Richard Biener  <rguenther@suse.de>
23674         PR tree-optimization/81529
23675         * tree-vect-stmts.c (process_use): Disregard live induction PHIs
23676         when optimizing backedge uses.
23678 2017-07-25  David Edelsohn  <dje.gcc@gmail.com>
23680         * dwarf2asm.c (dw2_asm_output_nstring): Encode double quote
23681         character for AIX.
23682         * dwarf2out.c (output_macinfo): Copy debug_line_section_label
23683         to dl_section_ref.  On AIX, append an expression to subtract
23684         the size of the section length to dl_section_ref.
23686 2017-07-25  Segher Boessenkool  <segher@kernel.crashing.org>
23688         * configure.ac: If any of the config.* scripts fail, exit 1.
23689         * configure: Regenerate.
23691 2017-07-25  Richard Biener  <rguenther@suse.de>
23693         PR middle-end/81546
23694         * tree-ssa-operands.c (verify_imm_links): Remove cap on number
23695         of immediate uses, be more verbose on errors.
23697 2017-07-25  Richard Biener  <rguenther@suse.de>
23699         PR tree-optimization/81510
23700         * tree-vect-loop.c (vect_is_simple_reduction): When the
23701         reduction stmt is not inside the loop bail out.
23703 2017-07-25  Richard Biener  <rguenther@suse.de>
23705         PR tree-optimization/81303
23706         * tree-vect-loop-manip.c (vect_loop_versioning): Build
23707         profitability check against LOOP_VINFO_NITERSM1.
23709 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23711         * domwalk.c (cmp_bb_postorder): Simplify.
23712         (sort_bbs_postorder): New function.  Use it...
23713         (dom_walker::walk): ...here to optimize common cases.
23715 2017-07-25  Martin Liska  <mliska@suse.cz>
23717         PR ipa/81520
23718         * ipa-visibility.c (function_and_variable_visibility): Make the
23719         redirection just on target that supports aliasing.
23720         Fix GNU coding style.
23722 2017-07-25  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23724         PR libgcc/61152
23725         * config/aarch64/rtems.h: Add GCC Runtime Library Exception.
23726         Format changes.
23727         * config/arm/rtems.h: Likewise.
23728         * config/bfin/rtems.h: Likewise.
23729         * config/i386/rtemself.h: Likewise.
23730         * config/lm32/rtems.h: Likewise.
23731         * config/m32c/rtems.h: Likewise.
23732         * config/m68k/rtemself.h: Likewise.
23733         * config/microblaze/rtems.h: Likewise.
23734         * config/mips/rtems.h: Likewise.
23735         * config/moxie/rtems.h: Likewise.
23736         * config/nios2/rtems.h: Likewise.
23737         * config/powerpcspe/rtems.h: Likewise.
23738         * config/rs6000/rtems.h: Likewise.
23739         * config/rtems.h: Likewise.
23740         * config/sh/rtems.h: Likewise.
23741         * config/sh/rtemself.h: Likewise.
23742         * config/sparc/rtemself.h: Likewise.
23744 2017-07-25  Georg-Johann Lay  <avr@gjlay.de>
23746         PR 81487
23747         * hsa-brig.c (brig_init): Use xasprintf instead of asprintf.
23748         * gimple-pretty-print.c (dump_profile, dump_probability): Same.
23749         * tree-ssa-structalias.c (alias_get_name): Same.
23751 2017-07-25  Bin Cheng  <bin.cheng@arm.com>
23753         PR target/81414
23754         * config/aarch64/cortex-a57-fma-steering.c (analyze): Skip fmul/fmac
23755         instructions if no du chain is found.
23757 2017-07-25  Georg-Johann Lay  <avr@gjlay.de>
23759         * config/avr/avr-log.c (avr_log_vadump) ['T']: Print NULL-TREE.
23761 2017-07-25  Richard Biener  <rguenther@suse.de>
23763         PR middle-end/81505
23764         * fold-const.c (fold_negate_const): TREE_OVERFLOW should be
23765         sticky.
23767 2017-07-24  Michael Meissner  <meissner@linux.vnet.ibm.com>
23769         * config/rs6000/rs6000-cpus.def (ISA_2_6_MASKS_SERVER): Delete
23770         upper-regs options.
23771         (ISA_2_7_MASKS_SERVER): Likewise.
23772         (ISA_3_0_MASKS_IEEE): Likewise.
23773         (OTHER_P8_VECTOR_MASKS): Likewise.
23774         (OTHER_VSX_VECTOR_MASKS): Likewise.
23775         (POWERPC_MASKS): Likewise.
23776         (power7 cpu): Use ISA_2_6_MASKS_SERVER instead of using a
23777         duplicate list of options.
23778         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove
23779         explicit -mupper-regs options.
23780         * config/rs6000/rs6000.opt (-mvsx-scalar-memory): Delete
23781         -mupper-regs* options.  Delete -mvsx-scalar-memory, which was an
23782         alias for -mupper-regs-df.
23783         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Likewise.
23784         (rs6000_init_hard_regno_mode_ok): Likewise.
23785         (rs6000_option_override_internal): Likewise.
23786         (rs6000_opt_masks): Likewise.
23787         * config/rs6000/rs6000.h (TARGET_UPPER_REGS_DF): Define upper regs
23788         options in terms of whether -mvsx or -mpower8-vector was used.
23789         (TARGET_UPPER_REGS_DI): Likewise.
23790         (TARGET_UPPER_REGS_SF): Likewise.
23791         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete the
23792         -mupper-regs-* options.
23794 2017-07-24  Segher Boessenkool  <segher@kernel.crashing.org>
23796         * passes.c (emergency_dump_function): Print some empty lines and a
23797         header before the RTL dump.
23799 2017-07-24  Segher Boessenkool  <segher@kernel.crashing.org>
23801         * cfgrtl.c (rtl_dump_bb): Don't call NEXT_INSN on NULL.
23803 2017-07-24  Wilco Dijkstra  <wdijkstr@arm.com>
23805         PR target/79041
23806         * config/aarch64/aarch64.c (aarch64_classify_symbol):
23807         Avoid SYMBOL_SMALL_ABSOLUTE for literals with pc-relative literals.
23809 2017-07-24  Carl Love  <cel@us.ibm.com>
23811         * config/rs6000/rs6000-c.c: Add support for built-in functions
23812         vector float vec_extract_fp32_from_shorth (vector unsigned short);
23813         vector float vec_extract_fp32_from_shortl (vector unsigned short);
23814         * config/rs6000/altivec.h (vec_extract_fp_from_shorth,
23815         vec_extract_fp_from_shortl): Add defines for the two builtins.
23816         * config/rs6000/rs6000-builtin.def (VEXTRACT_FP_FROM_SHORTH,
23817         VEXTRACT_FP_FROM_SHORTL): Add BU_P9V_OVERLOAD_1 and BU_P9V_VSX_1
23818         new builtins.
23819         * config/rs6000/vsx.md vsx_xvcvhpsp): Add define_insn.
23820         (vextract_fp_from_shorth, vextract_fp_from_shortl): Add define_expands.
23821         * doc/extend.texi: Update the built-in documentation file for the
23822         new built-in function.
23824 2017-07-24  Jakub Jelinek  <jakub@redhat.com>
23826         PR bootstrap/81521
23827         * tree.def: Remove TYPE_METHODS documentation, adjust TYPE_FIELDS
23828         documentation.
23829         * doc/generic.texi: Likewise.
23830         * config/i386/winnt-cxx.c (i386_pe_adjust_class_at_definition): Look
23831         for FUNCTION_DECLs in TYPE_FIELDS rather than TYPE_METHODS.
23833 2017-07-24  Jackson Woodruff  <jackson.woodruff@arm.com>
23835         * config/aarch64/aarch64-simd.md (aarch64_mla_elt_merge<mode>): New.
23836         (aarch64_mls_elt_merge<mode>): Likewise.
23838 2017-07-23  Krister Walfridsson  <krister.walfridsson@gmail.com>
23840         * config.gcc (*-*-netbsd*): Remove check for NetBSD versions not
23841         having __cxa_atexit.
23843 2017-07-23  Michael Collison  <michael.collison@arm.com>
23845         * config/arm/arm.c (arm_option_override): Deprecate
23846         use of -mstructure-size-boundary.
23847         * config/arm/arm.opt: Deprecate -mstructure-size-boundary.
23848         * doc/invoke.texi: Deprecate -mstructure-size-boundary.
23850 2017-07-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23852         PR target/80695
23853         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost):
23854         Reduce cost estimate for direct moves.
23856 2017-07-23  Uros Bizjak  <ubizjak@gmail.com>
23858         PR target/80569
23859         * config/i386/i386.c (ix86_option_override_internal): Disable
23860         BMI, BMI2 and TBM instructions for -m16.
23862 2017-07-21  Carl Love  <cel@us.ibm.com>
23864         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
23865         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
23866         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
23867         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
23868         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
23869         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
23870         VMULOSW): New enum "unspec" values.
23871         (altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
23872         altivec_vmulosw): New patterns.
23873         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
23874         VMULOSW): Add definitions.
23876 2017-07-21  Jim Wilson  <jim.wilson@linaro.org>
23878         * config/aarch64/aarch64-cores.def (falkor): Add AARCH64_FL_RDMA.
23879         (qdf24xx): Likewise.
23880         * config/aarch64/aarch64-options-extensions.def (rdma); New.
23881         * config/aarch64/aarch64.h (AARCH64_FL_RDMA): New.
23882         (AARCH64_FL_V8_1): Renumber.
23883         (AARCH64_FL_FOR_ARCH8_1): Add AARCH64_FL_RDMA.
23884         (AARCH64_ISA_RDMA): Use AARCH64_FL_RDMA.
23885         * config/aarch64/arm_neon.h: Use +rdma instead of arch=armv8.1-a.
23886         * doc/invoke.texi (AArch64 Options): Mention +rmda in -march docs.  Add
23887         rdma to feature modifiers list.
23889 2017-07-21  Yury Gribov  <tetra2005@gmail.com>
23891         PR middle-end/56727
23892         * ipa-visibility (function_and_variable_visibility): Convert
23893         recursive PLT call to direct call if appropriate.
23895 2017-07-21  Andrew Pinski  <apinski@cavium.com>
23897         * tree-ssa-sccvn.c (vn_nary_op_eq): Check BIT_INSERT_EXPR's
23898         operand 1 to see if the types precision matches.
23899         * fold-const.c (operand_equal_p): Likewise.
23901 2017-07-21  Richard Biener  <rguenther@suse.de>
23903         PR tree-optimization/81303
23904         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs): Pass
23905         in datarefs vector.  Allow NULL dr0 for no peeling cost estimate.
23906         (vect_peeling_hash_get_lowest_cost): Adjust.
23907         (vect_enhance_data_refs_alignment): Likewise.  Use
23908         vect_get_peeling_costs_all_drs to compute the penalty for no
23909         peeling to match up costs.
23911 2017-07-21  Richard Biener  <rguenther@suse.de>
23913         PR tree-optimization/81500
23914         * tree-vect-loop.c (vect_is_simple_reduction): Properly fail if
23915         we didn't identify a reduction path.
23917 2017-07-21  Tom de Vries  <tom@codesourcery.com>
23918             Cesar Philippidis  <cesar@codesourcery.com>
23920         PR gcov-profile/81442
23921         * config/nvptx/nvptx.c (nvptx_goacc_reduction_init): Add missing edge
23922         probabilities.
23924 2017-07-21  Tom de Vries  <tom@codesourcery.com>
23926         PR lto/81430
23927         * config/nvptx/nvptx.c (nvptx_override_options_after_change): New
23928         function.
23929         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define to
23930         nvptx_override_options_after_change.
23932 2017-07-21  Ulrich Drepper  <drepper@redhat.com>
23934         * dwarf2out.c (output_file_names): Avoid double testing for
23935         dwarf_version >= 5.
23937 2017-07-21  Georg-Johann Lay  <avr@gjlay.de>
23939         * doc/invoke.texi (AVR Built-in Functions): Re-layout section.
23941 2016-07-21  Jan Hubicka  <hubicka@ucw.cz>
23943         * cfgcleanup.c (flow_find_cross_jump): Do not crossjump across
23944         hot/cold regions.
23945         (try_crossjump_to_edge): Do not punt on partitioned functions.
23947 2016-07-21  Jan Hubicka  <hubicka@ucw.cz>
23949         * bb-reorder.c (find_rarely_executed_basic_blocks_and_crossing_edges):
23950         Put all BBs reachable only via paths crossing cold region to cold
23951         region.
23952         * cfgrtl.c (find_bbs_reachable_by_hot_paths): New function.
23954 2016-07-21  Richard Biener  <rguenther@suse.de>
23956         PR tree-optimization/81303
23957         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Take
23958         into account prologue and epilogue iterations when raising
23959         min_profitable_iters to sth at least covering one vector iteration.
23961 2017-07-21  Tamar Christina  <tamar.christina@arm.com>
23963         * config/arm/arm.c (arm_test_cpu_arch_dat):
23964         Check for overlap.
23966 2017-07-20  Nathan Sidwell  <nathan@acm.org>
23968         Remove TYPE_METHODS.
23969         * tree.h (TYPE_METHODS): Delete.
23970         * dwarf2out.c (gen_member_die): Member fns are on TYPE_FIELDS.
23971         * dbxout.c (dbxout_type_fields): Ignore FUNCTION_DECLs.
23972         (dbxout_type_methods): Scan TYPE_FIELDS.
23973         (dbxout_type): Don't check TYPE_METHODS here.
23974         * function.c (use_register_for_decl): Always ignore register for
23975         class types when not optimizing.
23976         * ipa-devirt.c (odr_types_equivalent_p): Delete TYPE_METHODS scan.
23977         * tree.c (free_lang_data_in_type): Stitch out member functions and
23978         templates from TYPE_FIELDS.
23979         (build_distinct_type_copy, verify_type_variant,
23980         verify_type): Member fns are on TYPE_FIELDS.
23981         * tree-dump.c (dequeue_and_dump): No TYPE_METHODS.
23982         * tree-pretty-print.c (dump_generic_node): Likewise.
23984 2017-07-20  Jakub Jelinek  <jakub@redhat.com>
23986         PR target/80846
23987         * config/i386/i386.c (ix86_expand_vector_init_general): Handle
23988         V2TImode and V4TImode.
23989         (ix86_expand_vector_extract): Likewise.
23990         * config/i386/sse.md (VMOVE): Enable V4TImode even for just
23991         TARGET_AVX512F, instead of only for TARGET_AVX512BW.
23992         (ssescalarmode): Handle V4TImode and V2TImode.
23993         (VEC_EXTRACT_MODE): Add V4TImode and V2TImode.
23994         (*vec_extractv2ti, *vec_extractv4ti): New insns.
23995         (VEXTRACTI128_MODE): New mode iterator.
23996         (splitter for *vec_extractv?ti first element): New.
23997         (VEC_INIT_MODE): New mode iterator.
23998         (vec_init<mode>): Consolidate 3 expanders into one using
23999         VEC_INIT_MODE mode iterator.
24001 2017-07-20  Alexander Monakov  <amonakov@ispras.ru>
24003         * lra-assigns.c (pseudo_compare_func): Fix comparison step based on
24004         non_spilled_static_chain_regno_p.
24006 2017-07-20  Alexander Monakov  <amonakov@ispras.ru>
24008         * gimple-ssa-store-merging.c (sort_by_bitpos): Return 0 on equal bitpos.
24010 2017-07-20  Jan Hubicka  <hubicka@ucw.cz>
24012         * bb-reorder.c (connect_traces): Allow copying of blocks within
24013         single partition.
24015 2017-07-20  Richard Biener  <rguenther@suse.de>
24017         * gimple.h (gimple_phi_result): Add gphi * overload.
24018         (gimple_phi_result_ptr): Likewise.
24019         (gimple_phi_arg): Likewise.  Adjust index assert to only
24020         allow actual argument accesses rather than all slots available
24021         by capacity.
24022         (gimple_phi_arg_def): Add gphi * overload.
24023         * tree-phinodes.c (make_phi_node): Initialize only actual
24024         arguments.
24025         (resize_phi_node): Clear memory not covered by old node,
24026         do not initialize excess argument slots.
24027         (reserve_phi_args_for_new_edge): Initialize new argument slot
24028         completely.
24030 2017-07-20  Bin Cheng  <bin.cheng@arm.com>
24032         PR tree-optimization/81388
24033         Revert r238585:
24034         2016-07-21  Bin Cheng  <bin.cheng@arm.com>
24036         * tree-ssa-loop-niter.c (number_of_iterations_lt_to_ne): Clean up
24037         by removing computation of may_be_zero.
24039 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
24040             Tom de Vries  <tom@codesourcery.com>
24042         PR middle-end/81030
24043         * cfgbuild.c (find_many_sub_basic_blocks): Update REG_BR_PROB note
24044         when gimple level profile disagrees with what RTL expander did.
24046 2017-07-20  Richard Biener  <rguenther@suse.de>
24048         PR tree-optimization/61171
24049         * tree-vectorizer.h (slp_instance): Add reduc_phis member.
24050         (vect_analyze_stmt): Add slp instance parameter.
24051         (vectorizable_reduction): Likewise.
24052         * tree-vect-loop.c (vect_analyze_loop_operations): Adjust.
24053         (vect_is_simple_reduction): Deal with chains not detected
24054         as SLP reduction chain, specifically not properly associated
24055         chains containing a mix of plus/minus.
24056         (get_reduction_op): Remove.
24057         (get_initial_defs_for_reduction): Simplify, pass in whether
24058         this is a reduction chain, pass in the SLP node for the PHIs.
24059         (vect_create_epilog_for_reduction): Get the SLP instance as
24060         arg and adjust.
24061         (vectorizable_reduction): Get the SLP instance as arg.
24062         During analysis remember the SLP node with the PHIs in the
24063         instance.  Simplify getting at the vectorized reduction PHIs.
24064         * tree-vect-slp.c (vect_slp_analyze_node_operations): Pass
24065         through SLP instance.
24066         (vect_slp_analyze_operations): Likewise.
24067         * tree-vect-stms.c (vect_analyze_stmt): Likewise.
24068         (vect_transform_stmt): Likewise.
24070 2017-07-20  Tom de Vries  <tom@codesourcery.com>
24072         PR tree-optimization/81489
24073         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior): Move
24074         read of phi arg location to before loop that modifies phi.
24076 2017-07-20  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
24078         * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0):
24079         New pattern.
24081 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
24083         PR middle-end/81331
24084         * except.c (execute): Fix ordering issue.
24086 2018-07-19  Segher Boessenkool  <segher@kernel.crashing.org>
24088         PR rtl-optimization/81423
24089         * combine.c (make_compound_operation_int): Don't try to optimize
24090         the AND of a SUBREG of an LSHIFTRT if that SUBREG is paradoxical.
24092 2017-07-19  Segher Boessenkool  <segher@kernel.crashing.org>
24094         PR rtl-optimization/81423
24095         * simplify-rtx.c (simplify_truncation): Handle truncating an IOR
24096         with a constant that is -1 in the truncated to mode.
24098 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
24100         * predict.c (propagate_unlikely_bbs_forward): Break out from ...
24101         (determine_unlikely_bbs): ... here.
24102         * predict.h (propagate_unlikely_bbs_forward): Declare.
24103         * cfgexpand.c (pass_expand::execute): Use it.
24104         * bb-reorder.c (sanitize_hot_paths): Do not consider known to be
24105         unlikely edges.
24106         (find_rarely_executed_basic_blocks_and_crossing_edges): Use
24107         propagate_unlikely_bbs_forward.
24109 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
24111         PR middle-end/81331
24112         * except.c (maybe_add_nop_after_section_switch): New function.
24113         (execute): Use it.
24115 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24117         * gimple.h (gimple_phi_set_arg): Make assert more strict.
24119 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24121         * gimple.h (gimple_phi_arg): Make assert more strict.
24123 2017-07-19  Steven Munroe  <munroesj@gcc.gnu.org>
24125         * config.gcc (powerpc*-*-*): Add mmintrin.h.
24126         * config/rs6000/mmintrin.h: New file.
24127         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include mmintrin.h.
24129 2017-07-19  Jakub Jelinek  <jakub@redhat.com>
24131         PR tree-optimization/81346
24132         * match.pd: Optimize (X - 1U) <= INT_MAX-1U into (int) X > 0.
24134 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24136         * config/nvptx/nvptx.md (VECIM): Add V2DI.
24138 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24140         * config/nvptx/nvptx-modes.def: Add V2DImode.
24141         * config/nvptx/nvptx-protos.h (nvptx_data_alignment): Declare.
24142         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Handle V2DImode.
24143         (nvptx_output_mov_insn): Handle lack of mov.b128.
24144         (nvptx_print_operand): Handle 'H' and 'L' codes.
24145         (nvptx_vector_mode_supported): Allow V2DImode.
24146         (nvptx_preferred_simd_mode): New function.
24147         (nvptx_data_alignment): New function.
24148         (TARGET_VECTORIZE_PREFERRED_SIMD_MODE): Redefine to
24149         nvptx_preferred_simd_mode.
24150         * config/nvptx/nvptx.h (STACK_BOUNDARY, BIGGEST_ALIGNMENT): Change from
24151         64 to 128 bits.
24152         (DATA_ALIGNMENT): Define.  Set to nvptx_data_alignment.
24154 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24156         * config/nvptx/nvptx-modes.def: New file.  Add V2SImode.
24157         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Handle V2SImode.
24158         (nvptx_vector_mode_supported): New function.  Allow V2SImode.
24159         (TARGET_VECTOR_MODE_SUPPORTED_P): Redefine to nvptx_vector_mode_supported.
24160         * config/nvptx/nvptx.md (VECIM): New mode iterator. Add V2SI.
24161         (mov<VECIM>_insn): New define_insn.
24162         (define_expand "mov<VECIM>): New define_expand.
24164 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24166         * config/nvptx/nvptx.c (nvptx_print_operand): Handle v2 vector mode.
24168 2017-07-19  Jakub Jelinek  <jakub@redhat.com>
24170         PR tree-optimization/81346
24171         * fold-const.h (fold_div_compare, range_check_type): Declare.
24172         * fold-const.c (range_check_type): New function.
24173         (build_range_check): Use range_check_type.
24174         (fold_div_compare): No longer static, rewritten into
24175         a match.pd helper function.
24176         (fold_comparison): Don't call fold_div_compare here.
24177         * match.pd (X / C1 op C2): New optimization using fold_div_compare
24178         as helper function.
24180 2017-07-19  Nathan Sidwell  <nathan@acm.org>
24182         * tree.h (TYPE_MINVAL, TYPE_MAXVAL): Rename to ...
24183         (TYPE_MIN_VALUE_RAW, TYPE_MAX_VALUE_RAW): ... these.
24184         * tree.c (find_decls_types_r, verify_type): Use
24185         TYPE_{MIN,MAX}_VALUE_RAW.
24186         * lto-streamer-out.c (DFS::DFS_write_tree_body): Likewise.
24187         (hash_tree): Likewise.
24188         * tree-streamer-in.c (lto_input_ts_type_non_common_tree_pointers):
24189         Likewise.
24190         * tree-streamer-out.c (write_ts_type_non_common_tree_pointers):
24191         Likewise.
24193 2017-07-18  Tom de Vries  <tom@codesourcery.com>
24195         PR middle-end/81464
24196         * omp-expand.c (expand_omp_for_static_chunk): Handle
24197         equal-argument loop exit phi.
24199 2017-07-18  Uros Bizjak  <ubizjak@gmail.com>
24201         PR target/81471
24202         * config/i386/i386.md (rorx_immediate_operand): New mode attribute.
24203         (*bmi2_rorx<mode>3_1): Use rorx_immediate_operand as
24204         operand 2 predicate.
24205         (*bmi2_rorxsi3_1_zext): Use const_0_to_31_operand as
24206         operand 2 predicate.
24207         (ror,rol -> rorx splitters): Use const_int_operand as
24208         operand 2 predicate.
24210 2017-06-18  Richard Biener  <rguenther@suse.de>
24212         PR tree-optimization/81410
24213         * tree-vect-stmts.c (vectorizable_load): Properly adjust for
24214         the gap in the ! slp_perm SLP case after each group.
24216 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
24218         PR middle-end/81463
24219         * cfgloopmanip.c (scale_loop_profile): Watch out for zero frequency
24220         again.
24222 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
24224         PR middle-end/81462
24225         * predict.c (set_even_probabilities): Cleanup; do not affect
24226         probabilities that are already known.
24227         (combine_predictions_for_bb): Call even when count is set.
24229 2017-07-18  Nathan Sidwell  <nathan@acm.org>
24231         * tree-parloops.c (try_transform_to_exit_first_loop_alt): Use
24232         TYPE_MAX_VALUE.
24234 2017-07-18  Bin Cheng  <bin.cheng@arm.com>
24236         PR target/81408
24237         * tree-ssa-loop-niter.c (number_of_iterations_exit): Dump missed
24238         optimization for loop niter analysis.
24240 2017-07-18  Georg-Johann Lay  <avr@gjlay.de>
24242         PR target/81473
24243         * config/avr/avr.c (avr_optimize_casesi): Don't use
24244         INT8_MIN, INT8_MAX, UINT8_MAX, INT16_MIN, INT16_MAX, UINT16_MAX.
24246 2017-07-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
24248         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Remove
24249         body_cost_vec from _vect_peel_extended_info.
24250         (vect_peeling_hash_get_lowest_cost): Do not set body_cost_vec.
24251         (vect_peeling_hash_choose_best_peeling): Remove body_cost_vec and
24252         npeel.
24254 2017-07-18  Bin Cheng  <bin.cheng@arm.com>
24256         * config/arm/arm.c (emit_unlikely_jump): Remove unused var.
24258 2017-07-18  Richard Biener  <rguenther@suse.de>
24260         PR tree-optimization/80620
24261         PR tree-optimization/81403
24262         * tree-ssa-pre.c (phi_translate_1): Clear range and points-to
24263         info when re-using a VN table entry.
24265 2017-07-18  Richard Biener  <rguenther@suse.de>
24267         PR tree-optimization/81418
24268         * tree-vect-loop.c (vectorizable_reduction): Properly compute
24269         vectype_in.  Verify that with lane-reducing reduction operations
24270         we have a single def-use cycle.
24272 2017-07-17  Carl Love  <cel@us.ibm.com>
24274         Revert commit r249424   2017-06-20  Carl Love  <cel@us.ibm.com>
24276         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
24277         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
24278         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
24279         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
24280         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
24281         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
24282         VMULOSW): New enum "unspec" values.
24283         (vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
24284         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si,
24285         altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
24286         altivec_vmulosw): New patterns.
24287         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
24288         VMULOSW): Add definitions.
24290 2017-07-17  Uros Bizjak  <ubizjak@gmail.com>
24292         * config/alpha/alpha.c: Include predict.h.
24294 2017-07-17  Yury Gribov  <tetra2005@gmail.com>
24296         * tree-vrp.c (compare_assert_loc): Fix comparison function
24297         to return predictable results.
24299 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24301         * config/arc/arc.md (adddi3): Remove support for mexpand-adddi
24302         option.
24303         (subdi3): Likewise.
24304         * config/arc/arc.opt (mexpand-adddi): Deprecate it.
24305         * doc/invoke.texi (mexpand-adddi): Update text.
24307 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24309         * config/arc/arc.md (clzsi2): Expand to an arc_clzsi2 instruction
24310         that also clobbers the CC register. The old expand code is moved
24311         to ...
24312         (*arc_clzsi2): ... here.
24313         (ctzsi2): Expand to an arc_ctzsi2 instruction that also clobbers
24314         the CC register. The old expand code is moved to ...
24315         (arc_ctzsi2): ... here.
24317 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24319         * config/arc/arc.opt (mindexed-loads): Use initial value
24320         TARGET_INDEXED_LOADS_DEFAULT.
24321         (mauto-modify-reg): Use initial value
24322         TARGET_AUTO_MODIFY_REG_DEFAULT.
24323         * config/arc/elf.h (TARGET_INDEXED_LOADS_DEFAULT): Define.
24324         (TARGET_AUTO_MODIFY_REG_DEFAULT): Likewise.
24325         * config/arc/linux.h (TARGET_INDEXED_LOADS_DEFAULT): Define.
24326         (TARGET_AUTO_MODIFY_REG_DEFAULT): Likewise.
24328 2017-07-17  Martin Liska  <mliska@suse.cz>
24330         PR sanitizer/81302
24331         * opts.c (finish_options): Do not allow -fgnu-tm
24332         w/ -fsanitize={kernel-,}address.  Say sorry.
24334 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24336         PR target/81369
24337         * tree-loop-distribution.c (classify_partition): Only assert on
24338         numer of iterations.
24339         (merge_dep_scc_partitions): Delete prameter.  Update function call.
24340         (distribute_loop): Remove code handling loop with unknown niters.
24341         (pass_loop_distribution::execute): Skip loop with unknown niters.
24343 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24345         PR target/81369
24346         * tree-loop-distribution.c (merge_dep_scc_partitions): Sink call to
24347         function sort_partitions_by_post_order.
24349 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24351         PR tree-optimization/81374
24352         * tree-loop-distribution.c (pass_loop_distribution::execute): Record
24353         the max index of basic blocks, rather than number of basic blocks.
24355 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24357         * config/arc/arc-protos.h (arc_legitimate_pc_offset_p): Remove
24358         proto.
24359         (arc_legitimate_pic_operand_p): Likewise.
24360         * config/arc/arc.c (arc_legitimate_pic_operand_p): Remove
24361         function.
24362         (arc_needs_pcl_p): Likewise.
24363         (arc_legitimate_pc_offset_p): Likewise.
24364         (arc_legitimate_pic_addr_p): Remove LABEL_REF case, as this
24365         function is also used in constrains.md.
24366         (arc_legitimate_constant_p): Use arc_legitimate_pic_addr_p to
24367         validate pic constants. Handle CONST_INT, CONST_DOUBLE, MINUS and
24368         PLUS.  Only return true/false in known cases, otherwise assert.
24369         (arc_legitimate_address_p): Remove arc_legitimate_pic_addr_p as it
24370         is already called in arc_legitimate_constant_p.
24371         * config/arc/arc.h (CONSTANT_ADDRESS_P): Consider also LABEL for
24372         pic addresses.
24373         (LEGITIMATE_PIC_OPERAND_P): Use
24374         arc_raw_symbolic_reference_mentioned_p function.
24375         * config/arc/constraints.md (Cpc): Use arc_legitimate_pic_addr_p
24376         function.
24377         (Cal): Likewise.
24378         (C32): Likewise.
24380 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24381         Andrew Burgess  <andrew.burgess@embecosm.com>
24383         * config/arc/arc-protos.h (arc_compute_function_type): Change prototype.
24384         (arc_return_address_register): New function.
24385         * config/arc/arc.c (arc_handle_fndecl_attribute): New function.
24386         (arc_handle_fndecl_attribute): Add naked attribute.
24387         (TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS): Define.
24388         (TARGET_WARN_FUNC_RETURN): Likewise.
24389         (arc_allocate_stack_slots_for_args): New function.
24390         (arc_warn_func_return): Likewise.
24391         (machine_function): Change type fn_type.
24392         (arc_compute_function_type): Consider new naked function type,
24393         change function return type.
24394         (arc_must_save_register): Adapt to handle new
24395         arc_compute_function_type's return type.
24396         (arc_expand_prologue): Likewise.
24397         (arc_expand_epilogue): Likewise.
24398         (arc_return_address_regs): Delete.
24399         (arc_return_address_register): New function.
24400         (arc_epilogue_uses): Use above function.
24401         * config/arc/arc.h (arc_return_address_regs): Delete prototype.
24402         (arc_function_type): Change encoding, add naked type.
24403         (ARC_INTERRUPT_P): Change to handle the new encoding.
24404         (ARC_FAST_INTERRUPT_P): Likewise.
24405         (ARC_NORMAL_P): Define.
24406         (ARC_NAKED_P): Likewise.
24407         (arc_compute_function_type): Delete prototype.
24408         * config/arc/arc.md (in_ret_delay_slot): Use
24409         arc_return_address_register function.
24410         (simple_return): Likewise.
24411         (p_return_i): Likewise.
24413 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24415         PR tree-optimization/81428
24416         * match.pd (X / X -> one): Don't optimize _Fract divisions, as 1
24417         can't be built for those types.
24419 2017-07-17  Georg-Johann Lay  <avr@gjlay.de>
24421         Remove stuff dead since r239246.
24423         * config/avr/avr-arch.h (avr_inform_devices): Remove dead proto.
24424         * config/avr/avr-devices.c (mcu_name, comparator, avr_mcus_str)
24425         (avr_inform_devices): Remove dead stuff.
24427 2017-07-17  Tamar Christina  <tamar.christina@arm.com>
24429         * config/arm/arm_neon.h: Fix softp typo.
24431 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24433         PR tree-optimization/81365
24434         * tree-ssa-phiprop.c (propagate_with_phi): When considering hoisting
24435         aggregate moves onto bb predecessor edges, make sure there are no
24436         loads that could alias the lhs in between the start of bb and the
24437         loads from *phi.
24439 2017-07-17  Georg-Johann Lay  <avr@gjlay.de>
24441         PR 80929
24442         * config/avr/avr.c (avr_mul_highpart_cost): New static function.
24443         (avr_rtx_costs_1) [TRUNCATE]: Use it to compute mul_highpart cost.
24444         [LSHIFTRT, outer_code = TRUNCATE]: Same.
24446 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24448         PR tree-optimization/81396
24449         * tree-ssa-math-opts.c (struct symbolic_number): Add n_ops field.
24450         (init_symbolic_number): Initialize it to 1.
24451         (perform_symbolic_merge): Add n_ops from both operands into the new
24452         n_ops.
24453         (find_bswap_or_nop): Don't consider n->n == cmpnop computations
24454         without base_addr as useless if they need more than one operation.
24455         (bswap_replace): Handle !bswap case for NULL base_addr.
24457 2017-07-17  Tom de Vries  <tom@codesourcery.com>
24459         PR target/81069
24460         * config/nvptx/nvptx.c (nvptx_single): Insert diverging branch as late
24461         as possible.
24463 2017-07-17  Sebastian Huber  <sebastian.huber@embedded-brains.de>
24465         * config/sparc/rtemself.h (TARGET_OS_CPP_BUILTINS): Add
24466         conditional builtin define __FIX_LEON3FT_B2BST.
24468 2017-07-17  Daniel Cederman  <cederman@gaisler.com>
24470         * config/sparc/t-rtems: Add mfix-gr712rc multilibs. Replace
24471         MULTILIB_EXCEPTIONS with MULTILIB_REQUIRED. Match -mfix-gr712rc
24472         with -mfix-ut700.
24474 2017-07-16  Eric Botcazou  <ebotcazou@adacore.com>
24476         PR rtl-optimization/81424
24477         * optabs.c (prepare_cmp_insn): Use copy_to_reg instead of force_reg
24478         to remove potential trapping from operands if -fnon-call-exceptions.
24480 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24482         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Use
24483         profile_proability for scalling.
24484         * scale_profile_for_vect_loop.c (scale_profile_for_vect_loop): Likewise.
24486 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24488         * cgraph.c (cgraph_edge::redirect_call_stmt_to_caller): Cleanup.
24490 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24492         * cfgloopmanip.c (scale_loop_profile): Avoid use of REG_BR_PROB_BASE
24493         fixpoint arithmetics.
24495 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24497         * tree-ssa-loop-unswitch.c (hoist_guard): Avoid use of REG_BR_PROB_BASE
24498         fixpoint arithmetics.
24500 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24502         * asan.c (create_cond_insert_point): Avoid use of REG_BR_PROB_BASE
24503         fixpoint arithmetics.
24505 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24507         * profile-count.h (profile_probability::from_reg_br_prob_note,
24508         profile_probability::to_reg_br_prob_note): New functions.
24509         * doc/rtl.texi (REG_BR_PROB_NOTE): Update documentation.
24510         * reg-notes.h (REG_BR_PROB, REG_BR_PRED): Update docs.
24511         * predict.c (probability_reliable_p): Update.
24512         (edge_probability_reliable_p): Update.
24513         (br_prob_note_reliable_p): Update.
24514         (invert_br_probabilities): Update.
24515         (add_reg_br_prob_note): New function.
24516         (combine_predictions_for_insn): Update.
24517         * asan.c (asan_clear_shadow): Update.
24518         * cfgbuild.c (compute_outgoing_frequencies): Update.
24519         * cfgrtl.c (force_nonfallthru_and_redirect): Update.
24520         (update_br_prob_note): Update.
24521         (rtl_verify_edges): Update.
24522         (purge_dead_edges): Update.
24523         (fixup_reorder_chain): Update.
24524         * emit-rtl.c (try_split): Update.
24525         * ifcvt.c (cond_exec_process_insns): Update.
24526         (cond_exec_process_if_block): Update.
24527         (dead_or_predicable): Update.
24528         * internal-fn.c (expand_addsub_overflow): Update.
24529         (expand_neg_overflow): Update.
24530         (expand_mul_overflow): Update.
24531         * loop-doloop.c (doloop_modify): Update.
24532         * loop-unroll.c (compare_and_jump_seq): Update.
24533         * optabs.c (emit_cmp_and_jump_insn_1): Update.
24534         * predict.h: Update.
24535         * reorg.c (mostly_true_jump): Update.
24536         * rtl.h: Update.
24537         * config/aarch64/aarch64.c (aarch64_emit_unlikely_jump): Update.
24538         * config/alpha/alpha.c (emit_unlikely_jump): Update.
24539         * config/arc/arc.c: (emit_unlikely_jump): Update.
24540         * config/arm/arm.c: (emit_unlikely_jump): Update.
24541         * config/bfin/bfin.c (cbranch_predicted_taken_p): Update.
24542         * config/frv/frv.c (frv_print_operand_jump_hint): Update.
24543         * config/i386/i386.c (ix86_expand_split_stack_prologue): Update.
24544         (ix86_print_operand): Update.
24545         (ix86_split_fp_branch): Update.
24546         (predict_jump): Update.
24547         * config/ia64/ia64.c (ia64_print_operand): Update.
24548         * config/mmix/mmix.c (mmix_print_operand): Update.
24549         * config/powerpcspe/powerpcspe.c (output_cbranch): Update.
24550         (rs6000_expand_split_stack_prologue): Update.
24551         * config/rs6000/rs6000.c: Update.
24552         * config/s390/s390.c (s390_expand_vec_strlen): Update.
24553         (s390_expand_vec_movstr): Update.
24554         (s390_expand_cs_tdsi): Update.
24555         (s390_expand_split_stack_prologue): Update.
24556         * config/sh/sh.c (sh_print_operand): Update.
24557         (expand_cbranchsi4): Update.
24558         (expand_cbranchdi4): Update.
24559         * config/sparc/sparc.c (output_v9branch): Update.
24560         * config/spu/spu.c (get_branch_target): Update.
24561         (ea_load_store_inline): Update.
24562         * config/tilegx/tilegx.c (cbranch_predicted_p): Update.
24563         * config/tilepro/tilepro.c: Update.
24565 2017-07-16  Eric Botcazou  <ebotcazou@adacore.com>
24567         * gimplify.c (mostly_copy_tree_r): Revert latest change.
24568         (gimplify_save_expr): Likewise.
24570 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24572         * ipa-visibility.c (function_and_variable_visibility): Fix pasto.
24574 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24576         * ipa-fnsummary.c (pass_data_ipa_fn_summary): Use
24577         TV_IPA_FNSUMMARY.
24578         * timevar.def (TV_IPA_FNSUMMARY): Define.
24580 2017-07-16  Daniel Cederman  <cederman@gaisler.com>
24582         * config/sparc/sparc.md (divdf3_fix): Add NOP to prevent back
24583         to back store errata sensitive sequence from being generated.
24584         (sqrtdf2_fix): Likewise.
24586 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24588         * tree-ssa-threadupdate.c (compute_path_counts,
24589         update_joiner_offpath_counts): Use profile_probability.
24591 2017-07-15  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24593         Revert:
24594         2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24596         * config/arm/arm-c.c (arm_cpu_builtins): Define
24597         __ARM_FEATURE_NUMERIC_MAXMIN solely based on TARGET_VFP5.
24599 2017-07-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
24601         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
24602         array entries to represent __ieee128 versions of the
24603         scalar_test_data_class, scalar_test_neg, scalar_extract_exp,
24604         scalar_extract_sig, and scalar_insert_exp built-in functions.
24605         (altivec_resolve_overloaded_builtin): Add special case handling
24606         for the __builtin_scalar_insert_exp function, as represented by
24607         the P9V_BUILTIN_VEC_VSIEDP constant.
24608         * config/rs6000/rs6000-builtin.def (VSEEQP): Add scalar extract
24609         exponent support for __ieee128 argument.
24610         (VSESQP): Add scalar extract signature support for __ieee128
24611         argument.
24612         (VSTDCNQP): Add scalar test negative support for __ieee128
24613         argument.
24614         (VSIEQP): Add scalar insert exponent support for __int128 argument
24615         with __ieee128 result.
24616         (VSIEQPF): Add scalar insert exponent support for __ieee128
24617         argument with __ieee128 result.
24618         (VSTDCQP): Add scalar test data class support for __ieee128
24619         argument.
24620         (VSTDCNQP): Add overload support for scalar test negative with
24621         __ieee128 argument.
24622         (VSTDCQP): Add overload support for scalar test data class
24623         __ieee128 argument.
24624         * config/rs6000/vsx.md (UNSPEC_VSX_SXSIG) Replace
24625         UNSPEC_VSX_SXSIGDP.
24626         (UNSPEC_VSX_SIEXPQP): New constant.
24627         (xsxexpqp): New insn for VSX scalar extract exponent quad
24628         precision.
24629         (xsxsigqp): New insn for VSX scalar extract significand quad
24630         precision.
24631         (xsiexpqpf): New insn for VSX scalar insert exponent quad
24632         precision with floating point argument.
24633         (xststdcqp): New expand for VSX scalar test data class quad
24634         precision.
24635         (xststdcnegqp): New expand for VSX scalar test negative quad
24636         precision.
24637         (xststdcqp): New insn to match expansions for VSX scalar test data
24638         class quad precision and VSX scalar test negative quad precision.
24639         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Add
24640         special case operand checking to enforce that second operand of
24641         VSX scalar test data class with quad precision argument is a 7-bit
24642         unsigned literal.
24643         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
24644         prototypes and descriptions of __ieee128 versions of
24645         scalar_extract_exp, scalar_extract_sig, scalar_insert_exp,
24646         scalar_test_data_class, and scalar_test_neg built-in functions.
24648 2016-07-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24650         PR tree-optimization/81162
24651         * gimple-ssa-strength-reduction.c (replace_mult_candidate): Don't
24652         replace a negate with an add.
24654 2017-07-14  James Greenhalgh  <james.greenhalgh@arm.com>
24656         * doc/invoke.texi (arm/-mcpu): Document +crypto.
24658 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24660         * config/arm/arm-c.c (arm_cpu_builtins): Define
24661         __ARM_FEATURE_NUMERIC_MAXMIN solely based on TARGET_VFP5.
24663 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24665         * config/arm/arm-cpus.in (cortex-r52): Add new entry.
24666         (armv8-r): Set ARM Cortex-R52 as default CPU.
24667         * config/arm/arm-tables.opt: Regenerate.
24668         * config/arm/arm-tune.md: Regenerate.
24669         * config/arm/driver-arm.c (arm_cpu_table): Add entry for ARM
24670         Cortex-R52.
24671         * doc/invoke.texi: Mention -mtune=cortex-r52 and availability of fp.dp
24672         extension for -mcpu=cortex-r52.
24674 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24676         * config/arm/arm-isa.h (isa_bit_FP_ARMv8): Delete enumerator.
24677         (ISA_FP_ARMv8): Define as ISA_FPv5 and ISA_FP_D32.
24678         * config/arm/arm-cpus.in (armv8-r): Define fp.sp as enabling FPv5.
24679         (fp-armv8): Define it as FP_ARMv8 only.
24680         config/arm/arm.h (TARGET_FPU_ARMV8): Delete.
24681         (TARGET_VFP_FP16INST): Define using TARGET_VFP5 rather than
24682         TARGET_FPU_ARMV8.
24683         config/arm/arm.c (arm_rtx_costs_internal): Replace checks against
24684         TARGET_FPU_ARMV8 by checks against TARGET_VFP5.
24685         * config/arm/arm-builtins.c (arm_builtin_vectorized_function): Define
24686         first ARM_CHECK_BUILTIN_MODE definition using TARGET_VFP5 rather
24687         than TARGET_FPU_ARMV8.
24688         * config/arm/arm-c.c (arm_cpu_builtins): Likewise for
24689         __ARM_FEATURE_NUMERIC_MAXMIN macro definition.
24690         * config/arm/arm.md (cmov<mode>): Condition on TARGET_VFP5 rather than
24691         TARGET_FPU_ARMV8.
24692         * config/arm/neon.md (neon_vrint): Likewise.
24693         (neon_vcvt): Likewise.
24694         (neon_<fmaxmin_op><mode>): Likewise.
24695         (<fmaxmin><mode>3): Likewise.
24696         * config/arm/vfp.md (l<vrint_pattern><su_optab><mode>si2): Likewise.
24697         * config/arm/predicates.md (arm_cond_move_operator): Check against
24698         TARGET_VFP5 rather than TARGET_FPU_ARMV8 and fix spacing.
24700 2017-07-14  Jackson Woodruff  <jackson.woodruff@arm.com>
24702         * config/aarch64/aarch64.c (aarch64_print_operand): Move comments
24703         to top of function.
24705 2017-07-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24707         * gimple-ssa-store-merging.c (clear_bit_region): Replace reference to
24708         loop in comment with memset.
24710 2017-07-14  Martin Liska  <mliska@suse.cz>
24712         * cfgexpand.c (expand_gimple_basic_block): Remove dead comment.
24713         * dwarf2out.c (is_java): Remove the function.
24714         (output_pubname): Remove usage of the function.
24715         (lower_bound_default): Remove usage of DW_LANG_Java.
24716         (gen_compile_unit_die): Likewise.
24717         * gcc.c: Remove compiler defaults for .java and .zip files.
24718         * gimple-expr.c (remove_suffix): Change as there's no longer
24719         extension than 4-letter one.
24720         * gimplify.c (mostly_copy_tree_r): Remove Java-special part.
24721         (gimplify_save_expr): Likewise.
24722         * ipa-utils.h (polymorphic_type_binfo_p): Remove the comment
24723         as it's possible even for other languages than Java.
24724         * langhooks.h (struct lang_hooks): Remove Java from a comment.
24725         * lto-opts.c (lto_write_options): Remove reference to Java.
24726         * opts.c (strip_off_ending): Update file extension handling.
24727         * tree-cfg.c (verify_gimple_call): Remove comment with Java.
24728         * tree-eh.c (lower_resx): Likewise.
24729         * tree.c (free_lang_data_in_type): Remove dead code.
24730         (find_decls_types_r): Likewise.
24731         (build_common_builtin_nodes): Remove Java from a comment.
24732         (verify_type): Remove dead code.
24733         * varasm.c (assemble_external): Remove Java from a comment.
24735 2017-07-14  Martin Liska  <mliska@suse.cz>
24737         * opts.c (finish_options): Add quotes.
24738         (common_handle_option): Likewise.
24740 2017-07-14  Martin Liska  <mliska@suse.cz>
24742         * dbxout.c (get_lang_number): Do not handle GNU Pascal.
24743         * dbxout.h (extern void dbxout_stab_value_internal_label_diff):
24744         Remove N_SO_PASCAL.
24745         * dwarf2out.c (lower_bound_default): Do not handle
24746         DW_LANG_Pascal83.
24747         (gen_compile_unit_die): Likewise.
24748         * gcc.c: Remove default extension binding for GNU Pascal.
24749         * stmt.c: Remove Pascal language from a comment.
24750         * xcoffout.c: Likewise.
24752 2017-07-13  David Malcolm  <dmalcolm@redhat.com>
24754         PR c/81405
24755         * diagnostic-show-locus.c (fixit_cmp): New function.
24756         (layout::layout): Sort m_fixit_hints.
24757         (column_range::column_range): Assert that the values are valid.
24758         (struct char_span): New struct.
24759         (correction::overwrite): New method.
24760         (struct source_line): New struct.
24761         (line_corrections::add_hint): Add assertions.  Reimplement memcpy
24762         calls in terms of classes source_line and char_span, and
24763         correction::overwrite.
24764         (selftest::test_overlapped_fixit_printing_2): New function.
24765         (selftest::diagnostic_show_locus_c_tests): Call it.
24767 2017-07-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
24769         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Return
24770         early if there is no lhs.
24772 2017-07-13  Martin Liska  <mliska@suse.cz>
24774         * dwarf2out.c (gen_pointer_type_die): Remove dead code.
24775         (gen_reference_type_die): Likewise.
24776         * stor-layout.c: Remove Pascal-related comment.
24778 2017-07-13  Martin Liska  <mliska@suse.cz>
24780         * opts.c (finish_options): Add quotes to error messages.
24781         (parse_sanitizer_options): Likewise.
24783 2017-07-13  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24785         * doc/invoke.texi (armv8-r): Document +fp.sp ARMv8-R extension.
24787 2017-07-13  Richard Earnshaw  <rearnsha@arm.com>
24789         * config/arm/vxworks.h (TARGET_ENDIAN_DEFAULT): Define.
24791 2017-07-13  Maxim Ostapenko  <m.ostapenko@samsung.com>
24793         * asan.c (asan_emit_allocas_unpoison): Use ptr_mode for arguments
24794         during expansion.
24795         * builtins.c (expand_asan_emit_allocas_unpoison): Likewise.
24797 2017-07-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
24799         PR target/81193
24800         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If GLIBC
24801         provides the hardware capability bits, define the macro
24802         __BUILTIN_CPU_SUPPORTS__.
24803         * config/rs6000/rs6000.c (cpu_expand_builtin): Generate a warning
24804         if GLIBC does not provide the hardware capability bits.  Add a
24805         gcc_unreachable call if the built-in cpu function is neither
24806         __builtin_cpu_is nor __builtin_cpu_supports.
24807         (rs6000_get_function_versions_dispatcher): Change the warning
24808         that an old GLIBC is used which does not export the capability
24809         bits to be an error.
24810         * doc/extend.texi (target_clones attribute): Document the
24811         restriction that GLIBC 2.23 or newer is needed on the PowerPC.
24812         (PowerPC built-in functions): Document that GLIBC 2.23 or newer is
24813         needed by __builtin_cpu_is and __builtin_cpu_supports.  Document
24814         the macros defined by GCC if the newer GLIBC is available.
24816 2017-07-12  Jeff Law  <law@redhat.com>
24818         * config/riscv/riscv.c: Remove unnecessary includes.  Reorder
24819         remaining includes slightly.
24820         * config/riscv/riscv-builtins.c: Include profile-count.h.
24822 2017-07-12  Georg-Johann Lay  <avr@gjlay.de>
24824         PR target/79883
24825         * config/avr/avr.c (avr_set_current_function): In diagnostic
24826         messages: Quote keywords and (parts of) identifiers.
24827         [WITH_AVRLIBC]: Warn for functions named "ISR", "SIGNAL" or
24828         "INTERUPT".
24830 2017-07-12  Carl Love  <cel@us.ibm.com>
24832         * config/rs6000/rs6000-c.c: Add support for built-in functions
24833         vector bool char vec_revb (vector bool char);
24834         vector bool short vec_revb (vector short char);
24835         vector bool int vec_revb (vector bool int);
24836         vector bool long long vec_revb (vector bool long long);
24837         * doc/extend.texi: Update the built-in documentation file for the
24838         new built-in functions.
24840 2017-07-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24842         * config/s390/s390.md: Remove movcc splitter.
24844 2017-07-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24846         * config/s390/s390.c (s390_rtx_costs): Return proper costs for
24847         load/store on condition.
24849 2017-07-12  Georg-Johann Lay  <avr@gjlay.de>
24851         PR target/81407
24852         * config/avr/avr.c (avr_encode_section_info)
24853         [progmem && !TREE_READONLY]: Error if progmem object needs
24854         constructing.
24856 2017-07-11  Michael Collison  <michael.collison@arm.com>
24858         * config/aarch64/aarch64-simd.md (aarch64_sub<mode>_compare0):
24859         New pattern.
24861 2017-07-11  Carl Love  <cel@us.ibm.com>
24863         * config/rs6000/rs6000-c.c: Add support for builtins
24864         vector unsigned int vec_parity_lsbb (vector signed int);
24865         vector unsigned int vec_parity_lsbb (vector unsigned int);
24866         vector unsigned __int128 vec_parity_lsbb (vector signed __int128);
24867         vector unsigned __int128 vec_parity_lsbb (vector unsigned __int128);
24868         vector unsigned long long vec_parity_lsbb (vector signed long long);
24869         vector unsigned long long vec_parity_lsbb (vector unsigned long long);
24870         * config/rs6000/rs6000-builtin.def (VPARITY_LSBB): Add BU_P9V_OVERLOAD1.
24871         * config/rs6000/altivec.h (vec_parity_lsbb): Add define.
24872         * doc/extend.texi: Update the built-in documentation file for the
24873         new built-in functions.
24875 2017-07-11  David Malcolm  <dmalcolm@redhat.com>
24877         * diagnostic-show-locus.c: Include "gcc-rich-location.h".
24878         (layout::m_primary_loc): New field.
24879         (layout::layout): Initialize new field.  Move location filtering
24880         logic from here to...
24881         (layout::maybe_add_location_range): ...this new method.  Add
24882         support for filtering to just the lines already specified by other
24883         locations.
24884         (layout::will_show_line_p): New method.
24885         (gcc_rich_location::add_location_if_nearby): New method.
24886         (selftest::test_add_location_if_nearby): New test function.
24887         (selftest::diagnostic_show_locus_c_tests): Call it.
24888         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
24889         New method.
24891 2017-07-11  Tom de Vries  <tom@codesourcery.com>
24893         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG): New macro.
24894         (bb_first_real_insn): New function.
24895         (nvptx_single): Add extra initialization of broadcasted condition
24896         variables.
24898 2017-07-11  Nathan Sidwell  <nathan@acm.org>
24900         * dwarf2out.c (gen_member_die): Remove useless check for anon ctors.
24902 2017-07-11  Georg-Johann Lay  <avr@gjlay.de>
24904         * doc/extend.texi (AVR Function Attributes): Remove weblink to
24905         Binutils doc as TEXI will mess them up.
24906         * doc/invoke.texi (AVR Options): Same here.
24908 2017-07-11  Daniel Cederman  <cederman@gaisler.com>
24910         * config/sparc/sparc.opt (mfix-ut700): New option.
24911         (mfix-gr712rc): Likewise.
24912         (sparc_fix_b2bst): New variable.
24913         * doc/invoke.texi (SPARC options): Document them.
24914         (ARM options): Fix warnings.
24915         * config/sparc/sparc.c (sparc_do_work_around_errata): Insert NOP
24916         instructions to prevent sequences that can trigger the store-store
24917         errata for certain LEON3FT processors.
24918         (pass_work_around_errata::gate): Also test sparc_fix_b2bst.
24919         (sparc_option_override): Set sparc_fix_b2bst appropriately.
24920         * config/sparc/sparc.md (fix_b2bst): New attribute.
24921         (in_branch_delay): Prevent stores in delay slot if fix_b2bst.
24923 2017-07-10  Uros Bizjak  <ubizjak@gmail.com>
24925         PR target/81375
24926         * config/i386/i386.md (divsf3): Add TARGET_SSE to TARGET_SSE_MATH.
24927         (rcpps): Ditto.
24928         (*rsqrtsf2_sse): Ditto.
24929         (rsqrtsf2): Ditto.
24930         (div<mode>3): Macroize insn from divdf3 and divsf3
24931         using MODEF mode iterator.
24933 2017-07-10  Martin Sebor  <msebor@redhat.com>
24935         PR tree-optimization/80397
24936         * gimple-ssa-sprintf.c (format_integer): Use INTEGRAL_TYPE_P()
24937         instead of testing for equality to INTEGER_TYPE.
24939 2017-07-10  Vineet Gupta <vgupta@synopsys.com>
24941         * config.gcc: Remove uclibc from arc target spec.
24943 2017-07-10  Claudiu Zissulescu  <claziss@synopsys.com>
24945         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Define.
24947 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24949         PR lto/80838
24950         * lto-wrapper.c (remove_option): New function.
24951         (merge_and_complain): Merge PIC/PIE options more realistically.
24953 2017-07-10  Georg-Johann Lay  <avr@gjlay.de>
24955         Better ISR prologues by supporting GASes __gcc_isr pseudo insn.
24957         PR target/20296
24958         PR target/81268
24959         * configure.ac [target=avr]: Add GAS check for -mgcc-isr.
24960         (HAVE_AS_AVR_MGCCISR_OPTION):  If so, AC_DEFINE it.
24961         * config.in: Regenerate.
24962         * configure: Regenerate.
24963         * doc/extend.texi (AVR Function Attributes) <no_gccisr>: Document it.
24964         * doc/invoke.texi (AVR Options) <-mgas-isr-prologues>: Document it.
24965         * config/avr/avr.opt (-mgas-isr-prologues): New option and...
24966         (TARGET_GASISR_PROLOGUES): ...target mask.
24967         * common/config/avr/avr-common.c
24968         (avr_option_optimization_table) [OPT_LEVELS_1_PLUS_NOT_DEBUG]:
24969         Set -mgas-isr-prologues.
24970         * config/avr/avr-passes.def (avr_pass_pre_proep): Add
24971         INSERT_PASS_BEFORE for it.
24972         * config/avr/avr-protos.h (make_avr_pass_pre_proep): New proto.
24973         * config/avr/avr.c (avr_option_override)
24974         [!HAVE_AS_AVR_MGCCISR_OPTION]: Unset TARGET_GASISR_PROLOGUES.
24975         (avr_no_gccisr_function_p, avr_hregs_split_reg): New static functions.
24976         (avr_attribute_table) <no_gccisr>: Add new function attribute.
24977         (avr_set_current_function) <is_no_gccisr>: Init machine field.
24978         (avr_pass_data_pre_proep, avr_pass_pre_proep): New pass data
24979         and rtl_opt_pass.
24980         (make_avr_pass_pre_proep): New function.
24981         (emit_push_sfr) <treg>: Add argument to function and use it
24982         instead of TMP_REG.
24983         (avr_expand_prologue) [machine->gasisr.maybe]: Emit gasisr insn
24984         and set machine->gasisr.yes.
24985         (avr_expand_epilogue) [machine->gasisr.yes]: Similar.
24986         (avr_asm_function_end_prologue) [machine->gasisr.yes]: Add
24987         __gcc_isr.n_pushed to .L__stack_usage.
24988         (TARGET_ASM_FINAL_POSTSCAN_INSN): Define to...
24989         (avr_asm_final_postscan_insn): ...this new static function.
24990         * config/avr/avr.h (machine_function)
24991         <is_no_gccisr, use_L__stack_usage>: New fields.
24992         <gasisr, gasisr.yes, gasisr.maybe, gasisr.regno>: New fields.
24993         * config/avr/avr.md (UNSPECV_GASISR): Add unspecv enum.
24994         (GASISR_Prologue, GASISR_Epilogue, GASISR_Done): New define_constants.
24995         (gasisr, *gasisr): New expander and insn.
24996         * config/avr/gen-avr-mmcu-specs.c (print_mcu)
24997         [HAVE_AS_AVR_MGCCISR_OPTION]: Print asm_gccisr spec.
24998         * config/avr/specs.h (ASM_SPEC) <asm_gccisr>: Add sub spec.
25000 2017-07-10  Richard Earnshaw  <rearnsha@arm.com>
25002         * config/arm/parsecpu.awk (gen_comm_data): Do not escape single quotes
25003         in quoted strings.
25005 2017-07-10  Georg-Johann Lay  <avr@gjlay.de>
25007         Move jump-tables out of .text again.
25009         PR target/81075
25010         * config/avr/avr.c (ASM_OUTPUT_ADDR_VEC_ELT): Remove function.
25011         (ASM_OUTPUT_ADDR_VEC): New function.
25012         (avr_adjust_insn_length) [JUMP_TABLE_DATA_P]: Return 0.
25013         (avr_final_prescan_insn) [avr_log.insn_addresses]: Dump
25014         INSN_ADDRESSes as asm comment.
25015         * config/avr/avr.h (JUMP_TABLES_IN_TEXT_SECTION): Adjust comment.
25016         (ASM_OUTPUT_ADDR_VEC_ELT): Remove define.
25017         (ASM_OUTPUT_ADDR_VEC): Define to avr_output_addr_vec.
25018         * config/avr/avr.md (*tablejump): Adjust comment.
25019         * config/avr/elf.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove.
25020         * config/avr/avr-log.c (avr_log_set_avr_log) <insn_addresses>:
25021         New detail.
25022         * config/avr/avr-protos.h (avr_output_addr_vec_elt): Remove proto.
25023         (avr_output_addr_vec): New proto.
25024         (avr_log_t) <insn_addresses>: New field.
25026 2017-07-09  H.J. Lu  <hongjiu.lu@intel.com>
25028         PR target/81313
25029         * config/i386/i386.c (ix86_function_arg_advance): Set
25030         outgoing_args_on_stack to true if there are outgoing arguments
25031         on stack.
25032         (ix86_function_arg): Likewise.
25033         (ix86_get_drap_rtx): Use DRAP only if there are outgoing
25034         arguments on stack and ACCUMULATE_OUTGOING_ARGS is false.
25035         * config/i386/i386.h (machine_function): Add
25036         outgoing_args_on_stack.
25038 2017-07-09  Krister Walfridsson  <krister.walfridsson@gmail.com>
25040         * config.gcc (*-*-netbsd*): Remove check for NetBSD versions not
25041         supporting pthreds.
25042         * config/netbsd.h (NETBSD_LIBGCC_SPEC): Always enable pthreads.
25044 2017-07-08  Richard Sandiford  <richard.sandiford@linaro.org>
25046         * Makefile.in (HOOKS_H, RTL_BASE_H, FUNCTION_H, EXPR_H, REGS_H)
25047         (REAL_H): Remove $(MACHMODE_H).
25048         (FIXED_VALUE_H, TREE_CORE_H, CFGLOOP_H): Remove $(MACHMODE_H) and
25049         double-int.h.
25050         (CORETYPES_H): Add signop.h, wide-int.h, wide-int-print.h,
25051         $(MACHMODE_H) and double-int.h.
25052         (build/min-insn-modes.o): Depend on $(CORETYPES_H) rather than
25053         $(MACHMODE_H).
25054         (gengtype-state.o, gengtype.o, build/gengtype.o): Don't depend on
25055         double-int.h.
25057 2017-07-07  Andrew Pinski  <apinski@cavium.com>
25059         * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Check
25060         prev_set and curr_set for AARCH64_FUSE_ALU_BRANCH.
25062 2017-07-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
25064         * config/rs6000/rs6000.c (rs6000_get_function_versions_dispatcher):
25065         Add warning if GCC was not configured to link against a GLIBC that
25066         exports the hardware capability bits.
25067         (make_resolver_func): Make resolver function private and not a
25068         COMDAT function.  Create the name with clone_function_name instead
25069         of make_unique_name.
25071         PR target/81348
25072         * config/rs6000/rs6000.md (HI sign_extend splitter): Use the
25073         correct operand in doing the split.
25075 2017-07-07 Carl Love  <cel@us.ibm.com>
25077         * config/rs6000/rs6000-c: Add support for built-in function
25078         vector unsigned short vec_pack_to_short_fp32 (vector float,
25079                                                       vector float).
25080         * config/rs6000/rs6000-builtin.def (CONVERT_4F32_8I16): Add
25081         BU_P9V_AV_2 and BU_P9V_OVERLOAD_2 definitions.
25082         * config/rs6000/altivec.h (vec_pack_to_short_fp32): Add define.
25083         * config/rs6000/altivec.md(UNSPEC_CONVERT_4F32_8I16): Add UNSPEC.
25084         (convert_4f32_8i16): Add define_expand.
25085         * doc/extend.texi: Update the built-in documentation file for the
25086         new built-in function.
25088 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25090         * config/sparc/m8.md: New file.
25091         * config/sparc/sparc.md: Include m8.md.
25093 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25095         * config/sparc/sparc.opt: New option -mvis4b.
25096         * config/sparc/sparc.c (dump_target_flag_bits): Handle MASK_VIS4B.
25097         (sparc_option_override): Handle VIS4B.
25098         (enum sparc_builtins): Define
25099         SPARC_BUILTIN_DICTUNPACK{8,16,32},
25100         SPARC_BUILTIN_FPCMP{LE,GT,EQ,NE}{8,16,32}SHL,
25101         SPARC_BUILTIN_FPCMPU{LE,GT}{8,16,32}SHL,
25102         SPARC_BUILTIN_FPCMPDE{8,16,32}SHL and
25103         SPARC_BUILTIN_FPCMPUR{8,16,32}SHL.
25104         (check_constant_argument): New function.
25105         (sparc_vis_init_builtins): Define builtins
25106         __builtin_vis_dictunpack{8,16,32},
25107         __builtin_vis_fpcmp{le,gt,eq,ne}{8,16,32}shl,
25108         __builtin_vis_fpcmpu{le,gt}{8,16,32}shl,
25109         __builtin_vis_fpcmpde{8,16,32}shl and
25110         __builtin_vis_fpcmpur{8,16,32}shl.
25111         (sparc_expand_builtin): Check that the constant operands to
25112         __builtin_vis_fpcmp*shl and _builtin_vis_dictunpack* are indeed
25113         constant and in range.
25114         * config/sparc/sparc-c.c (sparc_target_macros): Handle
25115         TARGET_VIS4B.
25116         * config/sparc/sparc.h (SPARC_IMM2_P): Define.
25117         (SPARC_IMM5_P): Likewise.
25118         * config/sparc/sparc.md (cpu_feature): Add new feagure "vis4b".
25119         (enabled): Handle vis4b.
25120         (UNSPEC_DICTUNPACK): New unspec.
25121         (UNSPEC_FPCMPSHL): Likewise.
25122         (UNSPEC_FPUCMPSHL): Likewise.
25123         (UNSPEC_FPCMPDESHL): Likewise.
25124         (UNSPEC_FPCMPURSHL): Likewise.
25125         (cpu_feature): New CPU feature `vis4b'.
25126         (dictunpack{8,16,32}): New insns.
25127         (FPCSMODE): New mode iterator.
25128         (fpcscond): New code iterator.
25129         (fpcsucond): Likewise.
25130         (fpcmp{le,gt,eq,ne}{8,16,32}{si,di}shl): New insns.
25131         (fpcmpu{le,gt}{8,16,32}{si,di}shl): Likewise.
25132         (fpcmpde{8,16,32}{si,di}shl): Likewise.
25133         (fpcmpur{8,16,32}{si,di}shl): Likewise.
25134         * config/sparc/constraints.md: Define constraints `q' for unsigned
25135         2-bit integer constants and `t' for unsigned 5-bit integer
25136         constants.
25137         * config/sparc/predicates.md (imm5_operand_dictunpack8): New
25138         predicate.
25139         (imm5_operand_dictunpack16): Likewise.
25140         (imm5_operand_dictunpack32): Likewise.
25141         (imm2_operand): Likewise.
25142         * doc/invoke.texi (SPARC Options): Document -mvis4b.
25143         * doc/extend.texi (SPARC VIS Built-in Functions): Document the
25144         ditunpack* and fpcmp*shl builtins.
25146 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25148         * config.gcc: Handle m8 in --with-{cpu,tune} options.
25149         * config.in: Add HAVE_AS_SPARC6 define.
25150         * config/sparc/driver-sparc.c (cpu_names): Add entry for the SPARC
25151         M8.
25152         * config/sparc/sol2.h (CPP_CPU64_DEFAULT_SPEC): Define for
25153         TARGET_CPU_m8.
25154         (ASM_CPU32_DEFAUILT_SPEC): Likewise.
25155         (CPP_CPU_SPEC): Handle m8.
25156         (ASM_CPU_SPEC): Likewise.
25157         * config/sparc/sparc-opts.h (enum processor_type): Add
25158         PROCESSOR_M8.
25159         * config/sparc/sparc.c (m8_costs): New struct.
25160         (sparc_option_override): Handle TARGET_CPU_m8.
25161         (sparc32_initialize_trampoline): Likewise.
25162         (sparc64_initialize_trampoline): Likewise.
25163         (sparc_issue_rate): Likewise.
25164         (sparc_register_move_cost): Likewise.
25165         * config/sparc/sparc.h (TARGET_CPU_m8): Define.
25166         (CPP_CPU64_DEFAULT_SPEC): Define for M8.
25167         (ASM_CPU64_DEFAULT_SPEC): Likewise.
25168         (CPP_CPU_SPEC): Handle M8.
25169         (ASM_CPU_SPEC): Likewise.
25170         (AS_M8_FLAG): Define.
25171         * config/sparc/sparc.md: Add m8 to the cpu attribute.
25172         * config/sparc/sparc.opt: New option -mcpu=m8 for sparc targets.
25173         * configure.ac (HAVE_AS_SPARC6): Check for assembler support for
25174         M8 instructions.
25175         * configure: Regenerate.
25176         * doc/invoke.texi (SPARC Options): Document -mcpu=m8 and
25177         -mtune=m8.
25179 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25181         * config/sparc/niagara7.md: Rework the DFA scheduler to use insn
25182         subtypes.
25183         * config/sparc/sparc.md: Remove the `v3pipe' insn attribute.
25184         ("*movdi_insn_sp32"): Do not set v3pipe.
25185         ("*movsi_insn"): Likewise.
25186         ("*movdi_insn_sp64"): Likewise.
25187         ("*movsf_insn"): Likewise.
25188         ("*movdf_insn_sp32"): Likewise.
25189         ("*movdf_insn_sp64"): Likewise.
25190         ("*zero_extendsidi2_insn_sp64"): Likewise.
25191         ("*sign_extendsidi2_insn"): Likewise.
25192         ("*mov<VM32:mode>_insn"): Likewise.
25193         ("*mov<VM64:mode>_insn_sp64"): Likewise.
25194         ("*mov<VM64:mode>_insn_sp32"): Likewise.
25195         ("<plusminus_insn><VADDSUB:mode>3"): Likewise.
25196         ("<vlop:code><VL:mode>3"): Likewise.
25197         ("*not_<vlop:code><VL:mode>3"): Likewise.
25198         ("*nand<VL:mode>_vis"): Likewise.
25199         ("*<vlnotop:code>_not1<VL:mode>_vis"): Likewise.
25200         ("*<vlnotop:code>_not2<VL:mode>_vis"): Likewise.
25201         ("one_cmpl<VL:mode>2"): Likewise.
25202         ("faligndata<VM64:mode>_vis"): Likewise.
25203         ("alignaddrsi_vis"): Likewise.
25204         ("alignaddrdi_vis"): Likweise.
25205         ("alignaddrlsi_vis"): Likewise.
25206         ("alignaddrldi_vis"): Likewise.
25207         ("fcmp<gcond:code><GCM:gcm_name><P:mode>_vis"): Likewise.
25208         ("bmaskdi_vis"): Likewise.
25209         ("bmasksi_vis"): Likewise.
25210         ("bshuffle<VM64:mode>_vis"): Likewise.
25211         ("cmask8<P:mode>_vis"): Likewise.
25212         ("cmask16<P:mode>_vis"): Likewise.
25213         ("cmask32<P:mode>_vis"): Likewise.
25214         ("pdistn<P:mode>_vis"): Likewise.
25215         ("<vis3_addsub_ss_patname><VASS:mode>3"): Likewise.
25217 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25219         * config/sparc/sparc.md ("subtype"): New insn attribute.
25220         ("*wrgsr_sp64"): Set insn subtype.
25221         ("*rdgsr_sp64"): Likewise.
25222         ("alignaddrsi_vis"): Likewise.
25223         ("alignaddrdi_vis"): Likewise.
25224         ("alignaddrlsi_vis"): Likewise.
25225         ("alignaddrldi_vis"): Likewise.
25226         ("<plusminus_insn><VADDSUB:mode>3"): Likewise.
25227         ("fexpand_vis"): Likewise.
25228         ("fpmerge_vis"): Likewise.
25229         ("faligndata<VM64:mode>_vis"): Likewise.
25230         ("bshuffle<VM64:mode>_vis"): Likewise.
25231         ("cmask8<P:mode>_vis"): Likewise.
25232         ("cmask16<P:mode>_vis"): Likewise.
25233         ("cmask32<P:mode>_vis"): Likewise.
25234         ("fchksm16_vis"): Likewise.
25235         ("v<vis3_shift_patname><GCM:mode>3"): Likewise.
25236         ("fmean16_vis"): Likewise.
25237         ("fp<plusminus_insn>64_vis"): Likewise.
25238         ("<plusminus_insn>v8qi3"): Likewise.
25239         ("<vis3_addsub_ss_patname><VASS:mode>3"): Likewise.
25240         ("<vis4_minmax_patname><VMMAX:mode>3"): Likewise.
25241         ("<vis4_uminmax_patname><VMMAX:mode>3"): Likewise.
25242         ("<vis3_addsub_ss_patname>v8qi3"): Likewise.
25243         ("<vis4_addsub_us_patname><VAUS:mode>3"): Likewise.
25244         ("*movqi_insn"): Likewise.
25245         ("*movhi_insn"): Likewise.
25246         ("*movsi_insn"): Likewise.
25247         ("movsi_pic_gotdata_op"): Likewise.
25248         ("*movdi_insn_sp32"): Likewise.
25249         ("*movdi_insn_sp64"): Likewise.
25250         ("movdi_pic_gotdata_op"): Likewise.
25251         ("*movsf_insn"): Likewise.
25252         ("*movdf_insn_sp32"): Likewise.
25253         ("*movdf_insn_sp64"): Likewise.
25254         ("*zero_extendhisi2_insn"): Likewise.
25255         ("*zero_extendqihi2_insn"): Likewise.
25256         ("*zero_extendqisi2_insn"): Likewise.
25257         ("*zero_extendqidi2_insn"): Likewise.
25258         ("*zero_extendhidi2_insn"): Likewise.
25259         ("*zero_extendsidi2_insn_sp64"): Likewise.
25260         ("ldfsr"): Likewise.
25261         ("prefetch_64"): Likewise.
25262         ("prefetch_32"): Likewise.
25263         ("tie_ld32"): Likewise.
25264         ("tie_ld64"): Likewise.
25265         ("*tldo_ldub_sp32"): Likewise.
25266         ("*tldo_ldub1_sp32"): Likewise.
25267         ("*tldo_ldub2_sp32"): Likewise.
25268         ("*tldo_ldub_sp64"): Likewise.
25269         ("*tldo_ldub1_sp64"): Likewise.
25270         ("*tldo_ldub2_sp64"): Likewise.
25271         ("*tldo_ldub3_sp64"): Likewise.
25272         ("*tldo_lduh_sp32"): Likewise.
25273         ("*tldo_lduh1_sp32"): Likewise.
25274         ("*tldo_lduh_sp64"): Likewise.
25275         ("*tldo_lduh1_sp64"): Likewise.
25276         ("*tldo_lduh2_sp64"): Likewise.
25277         ("*tldo_lduw_sp32"): Likewise.
25278         ("*tldo_lduw_sp64"): Likewise.
25279         ("*tldo_lduw1_sp64"): Likewise.
25280         ("*tldo_ldx_sp64"): Likewise.
25281         ("*mov<VM32:mode>_insn"): Likewise.
25282         ("*mov<VM64:mode>_insn_sp64"): Likewise.
25283         ("*mov<VM64:mode>_insn_sp32"): Likewise.
25285 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25287         * config/sparc/sparc.md ("type"): New insn type viscmp.
25288         ("fcmp<gcond:code><GCM:gcm_name><P:mode>_vis"): Set insn type to
25289         viscmp.
25290         ("fpcmp<gcond:code>8<P:mode>_vis"): Likewise.
25291         ("fucmp<gcond:code>8<P:mode>_vis"): Likewise.
25292         ("fpcmpu<gcond:code><GCM:gcm_name><P:mode>_vis"): Likewise.
25293         * config/sparc/niagara7.md ("n7_vis_logical_v3pipe"): Handle
25294         viscmp.
25295         ("n7_vis_logical_11cycle"): Likewise.
25296         * config/sparc/niagara4.md ("n4_vis_logical"): Likewise.
25297         * config/sparc/niagara2.md ("niag3_vis": Likewise.
25298         * config/sparc/niagara.md ("niag_vis"): Likewise.
25299         * config/sparc/ultra3.md ("us3_fga"): Likewise.
25300         * config/sparc/ultra1_2.md ("us1_fga_double"): Likewise.
25302 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25304         * config/sparc/sparc.md: New instruction type `bmask'.
25305         (bmaskdi_vis): Use the `bmask' type.
25306         (bmasksi_vis): Likewise.
25307         * config/sparc/ultra3.md (us3_array): Likewise.
25308         * config/sparc/niagara7.md (n7_array): Likewise.
25309         * config/sparc/niagara4.md (n4_array): Likewise.
25310         * config/sparc/niagara2.md (niag2_vis): Likewise.
25311         (niag3_vis): Likewise.
25312         * config/sparc/niagara.md (niag_vis): Likewise.
25314 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25316         * ipa-comdats.c: Remove optimize check from gate.
25317         * ipa-fnsummary.c (ipa_fn_summary_generate): do not generate summary
25318         for functions not optimized.
25319         (ipa_fn_summary_read): Skip optimize check.
25320         (ipa_fn_summary_write): Likewise.
25321         * ipa-inline-analysis.c (do_estimate_growth_1): Check that caller
25322         is optimized.
25323         * ipa-inline.c (can_inline_edge_p): Not optimized functions are
25324         uninlinable.
25325         (can_inline_edge_p): Check flag_pcc_struct_return for match.
25326         (check_callers): Give up on caller which is not optimized.
25327         (inline_small_functions): Likewise.
25328         (ipa_inline): Do not give up when not optimizing.
25329         * ipa-visbility.c (function_and_variable_visibility): Do not optimize
25330         away unoptimizes cdtors.
25331         (whole_program_function_and_variable_visibility): Do
25332         ipa_discover_readonly_nonaddressable_vars in LTO mode.
25333         * ipa.c (process_references): Do not check optimize.
25334         (symbol_table::remove_unreachable_nodes): Update optimize check.
25335         (set_writeonly_bit): Update optimize check.
25336         (pass_ipa_cdtor_merge::gate): Do not check optimize.
25337         (pass_ipa_single_use::gate): Remove.
25339 2017-07-06  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
25341         * config/rs6000/rs6000.c (union_defs, union_uses, insn_is_load_p,
25342         insn_is_store_p, insn_is_swap_p, const_load_sequence_p, v2df_reduction_p,
25343         rtx_is_swappable_p, insn_is_swappable_p, chain_contains_only_swaps,
25344         mark_swaps_for_removal, swap_const_vector_halves, adjust_subreg_index,
25345         permute_load, permute_store, adjust_extract, adjust_splat,
25346         adjust_xxpermdi, adjust_concat, adjust_vperm, handle_special_swappables,
25347         replace_swap_with_copy, dump_swap_insn_table,
25348         alignment_with_canonical_addr, alignment_mask, find_alignment_op,
25349         recombine_lvx_pattern, recombine_stvx_pattern,
25350         recombine_lvx_stvx_patterns, rs6000_analyze_swaps,
25351         make_pass_analyze_swaps): Move all code related to p8 swap optimizations
25352         to file rs6000-p8swap.c.
25353         * config/rs6000/rs6000-p8swap.c: New file.
25354         * config/rs6000/t-rs6000: Add rule to build rs6000-p8swap.o.
25355         * config.gcc: Add rs6000-p8swap.o to extra_objs for powerpc*-*-*
25356         and rs6000*-*-* targets.
25358 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25360         * Makefile.in (selftest): Remove dependency on s-selftest-c++.
25362 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25364         * lto-wrapper.c (merge_and_complain): Do not merge
25365         fexceptions, fnon_call_exceptions, ftrapv, ffp_contract_, fmath_errno,
25366         fsigned_zeros, ftrapping_math, fwrapv.
25367         (append_compiler_options): Do not track these options.
25368         (append_linker_options): Likewie
25370 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25372         * cgraphunit.c (cgraph_node::finalize_function): When
25373         !flag_toplevel_reorde set no_reorder flag.
25374         (varpool_node::finalize_decl): Likewise.
25375         (symbol_table::compile): Drop no toplevel reorder path.
25377 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25379         * bb-reorder.c (better_edge_p): Do not build traces across abnormal/eh
25380         edges; zero probability is not better than uninitialized.
25382 2017-07-06  Maxim Ostapenko  <m.ostapenko@samsung.com>
25384         * asan.h (asan_sanitize_allocas_p): Declare.
25385         * asan.c (asan_sanitize_allocas_p): New function.
25386         (handle_builtin_stack_restore): Bail out if !asan_sanitize_allocas_p.
25387         (handle_builtin_alloca): Likewise.
25388         * cfgexpand.c (expand_used_vars): Do not add allocas unpoisoning stuff
25389         if !asan_sanitize_allocas_p.
25390         * params.def (asan-instrument-allocas): Add new option.
25391         * params.h (ASAN_PROTECT_ALLOCAS): Define.
25392         * opts.c (common_handle_option): Disable allocas sanitization for
25393         KASan by default.
25395 2017-07-06  Maxim Ostapenko  <m.ostapenko@samsung.com>
25397         * asan.c: Include gimple-fold.h.
25398         (get_last_alloca_addr): New function.
25399         (handle_builtin_stackrestore): Likewise.
25400         (handle_builtin_alloca): Likewise.
25401         (asan_emit_allocas_unpoison): Likewise.
25402         (get_mem_refs_of_builtin_call): Add new parameter, remove const
25403         quallifier from first paramerer. Handle BUILT_IN_ALLOCA,
25404         BUILT_IN_ALLOCA_WITH_ALIGN and BUILT_IN_STACK_RESTORE builtins.
25405         (instrument_builtin_call): Pass gimple iterator to
25406         get_mem_refs_of_builtin_call.
25407         (last_alloca_addr): New global.
25408         * asan.h (asan_emit_allocas_unpoison): Declare.
25409         * builtins.c (expand_asan_emit_allocas_unpoison): New function.
25410         (expand_builtin): Handle BUILT_IN_ASAN_ALLOCAS_UNPOISON.
25411         * cfgexpand.c (expand_used_vars): Call asan_emit_allocas_unpoison
25412         if function calls alloca.
25413         * gimple-fold.c (replace_call_with_value): Remove static keyword.
25414         * gimple-fold.h (replace_call_with_value): Declare.
25415         * internal-fn.c: Include asan.h.
25416         * sanitizer.def (BUILT_IN_ASAN_ALLOCA_POISON,
25417         BUILT_IN_ASAN_ALLOCAS_UNPOISON): New builtins.
25419 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25421         * Makefile.in (SELFTEST_FLAGS): Drop "-x c", moving it to...
25422         (C_SELFTEST_FLAGS): New.
25423         (CPP_SELFTEST_FLAGS): New.
25424         (SELFTEST_DEPS): New, from deps of s-selftest.
25425         (C_SELFTEST_DEPS): New, from deps of s-selftest.
25426         (CPP_SELFTEST_DEPS): New.
25427         (selftest): Add dependency on s-selftest-c++.
25428         (s-selftest): Rename to...
25429         (s-selftest-c): ...this, moving deps to SELFTEST_DEPS
25430         and C_SELFTEST_DEPS, and using C_SELFTEST_FLAGS rather
25431         than SELFTEST_FLAGS.
25432         (selftest-gdb): Rename to...
25433         (selftest-c-gdb): ...this, using C_SELFTEST_DEPS and
25434         C_SELFTEST_FLAGS.
25435         (selftest-gdb): Reintroduce as an alias for selftest-c-gdb.
25436         (selftest-valgrind): Rename to...
25437         (selftest-c-valgrind): ...this, using C_SELFTEST_DEPS and
25438         C_SELFTEST_FLAGS.
25439         (selftest-valgrind): Reintroduce as an alias for
25440         selftest-c-valgrind.
25441         (s-selftest-c++): New.
25442         (selftest-c++-gdb): New.
25443         (selftest-c++-valgrind): New.
25445 2017-07-06  Olivier Hainque  <hainque@adacore.com>
25447         * gcc.c (process_command): When deciding if undefined variables
25448         should be ignored when processing specs, accept "gcc -v" as well.
25450 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25452         * auto-profile.c (afdo_set_bb_count, afdo_propagate_edge,
25453         afdo_annotate_cfg): Set counts/probabilities as determined by afdo.
25455 2017-07-06  Thomas Preud'homme  <thomas.preudhomme@arm.com>
25457         * config/arm/arm-cpus.in (armv8-r): Add new entry.
25458         * config/arm/arm-isa.h (ISA_ARMv8r): Define macro.
25459         * config/arm/arm-tables.opt: Regenerate.
25460         * config/arm/arm.h (enum base_architecture): Add BASE_ARCH_8R
25461         enumerator.
25462         * doc/invoke.texi: Mention -march=armv8-r and its extensions.
25464 2017-07-06  Carl Love  <cel@us.ibm.com>
25466         * ChangeLog: Clean up from mid air collision
25468 2017-07-06  Carl Love  <cel@us.ibm.com>
25470         * config/rs6000/rs6000-c.c: Add support for built-in functions
25471         vector signed int vec_subc (vector signed int, vector signed int);
25472         vector signed __int128 vec_subc (vector signed __int128,
25473                                          vector signed __int128);
25474         vector unsigned __int128 vec_subc (vector unsigned __int128,
25475                                            vector unsigned __int128);
25476         vector signed int vec_sube (vector signed int, vector signed int,
25477                                     vector signed int);
25478         vector unsigned int vec_sube (vector unsigned int,
25479                                       vector unsigned int,
25480                                       vector unsigned int);
25481         vector signed __int128 vec_sube (vector signed __int128,
25482                                          vector signed __int128,
25483                                          vector signed__int128);
25484         vector unsigned __int128 vec_sube (vector unsigned __int128,
25485                                            vector unsigned __int128,
25486                                            vector unsigned __int128);
25487         vector signed int vec_subec (vector signed int, vector signed int,
25488                                      vector signed int);
25489         vector unsigned int vec_subec (vector unsigned int,
25490                                        vector unsigned int,
25491                                        vector unsigned int);
25492         vector signed __int128 vec_subec (vector signed __int128,
25493                                           vector signed __int128,
25494                                           vector signed__int128);
25495         vector unsigned __int128 vec_subec (vector unsigned __int128,
25496                                             vector unsigned __int128,
25497                                             vector unsigned __int128);
25498         * config/rs6000/rs6000.c (ALTIVEC_BUILTIN_VEC_SUBE,
25499         ALTIVEC_BUILTIN_VEC_SUBEC): Add ef_builtins.
25500         * config/rs6000/rs6000-builtin.def (SUBE, SUBEC): Add
25501         BU_ALTIVEC_OVERLOAD_X definitions.
25502         * config/rs6000/altivec.h (vec_sube, vec_subec): Add builtin defines.
25503         * doc/extend.texi: Update the built-in documentation file for the new
25504         built-in functions.
25506 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25508         PR c++/79300
25509         * diagnostic-show-locus.c (layout::layout): Use start and finish
25510         spelling location for the start and finish of each range.
25511         * genmatch.c (linemap_client_expand_location_to_spelling_point):
25512         Add unused aspect param.
25513         * input.c (expand_location_1): Add "aspect" param, and use it
25514         to access the correct part of the location.
25515         (expand_location): Pass LOCATION_ASPECT_CARET to new param of
25516         expand_location_1.
25517         (expand_location_to_spelling_point): Likewise.
25518         (linemap_client_expand_location_to_spelling_point): Add "aspect"
25519         param, and pass it to expand_location_1.
25521 2017-07-06  Sebastian Peryt  <sebastian.peryt@intel.com>
25523         * config/i386/avx512fintrin.h (_mm_mask_getexp_round_ss,
25524         _mm_maskz_getexp_round_ss, _mm_mask_getexp_round_sd,
25525         _mm_maskz_getexp_round_sd, _mm_mask_getmant_round_sd,
25526         _mm_maskz_getmant_round_sd, _mm_mask_getmant_round_ss,
25527         _mm_maskz_getmant_round_ss, _mm_mask_getexp_ss, _mm_maskz_getexp_ss,
25528         _mm_mask_getexp_sd, _mm_maskz_getexp_sd, _mm_mask_getmant_sd,
25529         _mm_maskz_getmant_sd, _mm_mask_getmant_ss,
25530         _mm_maskz_getmant_ss): New intrinsics.
25531         (__builtin_ia32_getexpss128_mask): Changed to ...
25532         __builtin_ia32_getexpss128_round ... this.
25533         (__builtin_ia32_getexpsd128_mask): Changed to ...
25534         __builtin_ia32_getexpsd128_round ... this.
25535         * config/i386/i386-builtin-types.def
25536         ((V2DF, V2DF, V2DF, INT, V2DF, UQI, INT),
25537         (V4SF, V4SF, V4SF, INT, V4SF, UQI, INT)): New function type aliases.
25538         * config/i386/i386-builtin.def (__builtin_ia32_getexpsd_mask_round,
25539         __builtin_ia32_getexpss_mask_round,     __builtin_ia32_getmantsd_mask_round,
25540         __builtin_ia32_getmantss_mask_round): New builtins.
25541         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_INT_V2DF_UQI_INT,
25542         V4SF_FTYPE_V4SF_V4SF_INT_V4SF_UQI_INT): Handle new types.
25543         (CODE_FOR_avx512f_vgetmantv2df_mask_round,
25544         CODE_FOR_avx512f_vgetmantv4sf_mask_round): New cases.
25545         * config/i386/sse.md
25546         (avx512f_sgetexp<mode><round_saeonly_name>): Changed to ...
25547         avx512f_sgetexp<mode><mask_scalar_name>
25548         <round_saeonly_scalar_name> ... this.
25549         (vgetexp<ssescalarmodesuffix>\t{<round_saeonly_op3>%2, %1, %0|
25550         %0, %1, %2<round_saeonly_op3>}): Changed to ...
25551         vgetexp<ssescalarmodesuffix>
25552         \t{<round_saeonly_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25553         %0<mask_scalar_operand3>, %1, %2<round_saeonly_scalar_mask_op3>} ... this.
25554         (avx512f_vgetmant<mode><round_saeonly_name>): Changed to ...
25555         avx512f_vgetmant<mode><mask_scalar_name>
25556         <round_saeonly_scalar_name> ... this.
25557         (vgetmant<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
25558         %0, %1, %2<round_saeonly_op4>, %3}): Changed to ...
25559         vgetmant<ssescalarmodesuffix>
25560         \t{%3, <round_saeonly_scalar_mask_op4>%2, %1, %0<mask_scalar_operand4>|
25561         %0<mask_scalar_operand4>, %1, %2
25562         <round_saeonly_scalar_mask_op4>, %3} ... this.
25563         * config/i386/subst.md (mask_scalar_operand4,
25564         round_saeonly_scalar_mask_operand4,     round_saeonly_scalar_mask_op4,
25565         round_saeonly_scalar_nimm_predicate): New subst attributes.
25567 2017-07-06  Julia Koval  <julia.koval@intel.com>
25569         * config/i386/i386.c (ix86_erase_embedded_rounding):
25570         Remove code for old rounding pattern.
25572 2017-07-06  Richard Earnshaw  <rearnsha@arm.com>
25574         * config/arm/t-arm (GTM_H): Add arm-cpu.h.
25576 2017-07-06  Christophe Lyon  <christophe.lyon@linaro.org>
25578         * doc/sourcebuild.texi (Test Directives, Variants of
25579         dg-require-support): Add documentation for dg-require-stack-check.
25581 2017-07-05  Sebastian Peryt  <sebastian.peryt@intel.com>
25583         * config/i386/subst.md (mask_scalar, round_scalar,
25584         round_saeonly_scalar): New meta-templates.
25585         (mask_scalar_name, mask_scalar_operand3, round_scalar_name,
25586         round_scalar_mask_operand3, round_scalar_mask_op3,
25587         round_scalar_constraint, round_scalar_prefix, round_saeonly_scalar_name,
25588         round_saeonly_scalar_mask_operand3, round_saeonly_scalar_mask_op3,
25589         round_saeonly_scalar_constraint,
25590         round_saeonly_scalar_prefix): New subst attribute.
25591         * config/i386/sse.md
25592         (<sse>_vm<plusminus_insn><mode>3<mask_name><round_name>): Renamed to ...
25593         <sse>_vm<plusminus_insn><mode>3<mask_scalar_name>
25594         <round_scalar_name> ... this.
25595         (<sse>_vm<multdiv_mnemonic><mode>3<mask_name><round_name>): Renamed to ...
25596         <sse>_vm<multdiv_mnemonic><mode>3<mask_scalar_name>
25597         <round_scalar_name> ... this.
25598         (<sse>_vm<code><mode>3<mask_name><round_saeonly_name>): Renamed to ...
25599         <sse>_vm<code><mode>3<mask_scalar_name>
25600         <round_saeonly_scalar_name> ... this.
25601         (v<plusminus_mnemonic><ssescalarmodesuffix>
25602         \t{<round_mask_op3>%2, %1, %0<mask_operand3>|
25603         %0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}): Changed to ...
25604         v<plusminus_mnemonic><ssescalarmodesuffix>
25605         \t{<round_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25606         %0<mask_scalar_operand3>, %1, %<iptr>2<round_scalar_mask_op3>} ... this.
25607         (v<multdiv_mnemonic><ssescalarmodesuffix>
25608         \t{<round_mask_op3>%2, %1, %0<mask_operand3>|
25609         %0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}): Changed to ...
25610         v<multdiv_mnemonic><ssescalarmodesuffix>
25611         \t{<round_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25612         %0<mask_scalar_operand3>, %1, %<iptr>2<round_scalar_mask_op3>} ... this.
25613         (v<maxmin_float><ssescalarmodesuffix>
25614         \t{<round_saeonly_mask_op3>%2, %1, %0<mask_operand3>|
25615         %0<mask_operand3>, %1, %<iptr>2<round_saeonly_mask_op3>}): Changed to ...
25616         v<maxmin_float><ssescalarmodesuffix>
25617         \t{<round_saeonly_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25618         %0<mask_scalar_operand3>, %1, %<iptr>2
25619         <round_saeonly_scalar_mask_op3>} ... this.
25621 2017-07-05  Richard Earnshaw  <rearnsha@arm.com>
25623         * config/arm/arm.c (arm_fixed_condition_code_regs): New function.
25624         (TARGET_FIXED_CONDITION_CODE_REGS): Redefine.
25626 2017-07-05  Richard Sandiford  <richard.sandiford@linaro.org>
25627             Alan Hayward  <alan.hayward@arm.com>
25628             David Sherwood  <david.sherwood@arm.com>
25630         * combine.c (simplify_if_then_else): Remove "enum" before
25631         "machine_mode".
25632         * compare-elim.c (can_eliminate_compare): Likewise.
25633         * config/aarch64/aarch64-builtins.c (aarch64_simd_builtin_std_type):
25634         Likewise.
25635         (aarch64_lookup_simd_builtin_type): Likewise.
25636         (aarch64_simd_builtin_type): Likewise.
25637         (aarch64_init_simd_builtin_types): Likewise.
25638         (aarch64_simd_expand_args): Likewise.
25639         * config/aarch64/aarch64-protos.h (aarch64_simd_attr_length_rglist):
25640         Likewise.
25641         (aarch64_reverse_mask): Likewise.
25642         (aarch64_simd_emit_reg_reg_move): Likewise.
25643         (aarch64_gen_adjusted_ldpstp): Likewise.
25644         (aarch64_ccmp_mode_to_code): Likewise.
25645         (aarch64_operands_ok_for_ldpstp): Likewise.
25646         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
25647         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
25648         Likewise.
25649         (aarch64_min_divisions_for_recip_mul): Likewise.
25650         (aarch64_reassociation_width): Likewise.
25651         (aarch64_get_condition_code_1): Likewise.
25652         (aarch64_simd_emit_reg_reg_move): Likewise.
25653         (aarch64_simd_attr_length_rglist): Likewise.
25654         (aarch64_reverse_mask): Likewise.
25655         (aarch64_operands_ok_for_ldpstp): Likewise.
25656         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
25657         (aarch64_gen_adjusted_ldpstp): Likewise.
25658         * config/aarch64/cortex-a57-fma-steering.c (fma_node::rename):
25659         Likewise.
25660         * config/arc/arc.c (legitimate_offset_address_p): Likewise.
25661         * config/arm/arm-builtins.c (arm_simd_builtin_std_type): Likewise.
25662         (arm_lookup_simd_builtin_type): Likewise.
25663         (arm_simd_builtin_type): Likewise.
25664         (arm_init_simd_builtin_types): Likewise.
25665         (arm_expand_builtin_args): Likewise.
25666         * config/arm/arm-protos.h (arm_expand_builtin): Likewise.
25667         * config/ft32/ft32.c (ft32_libcall_value): Likewise.
25668         (ft32_setup_incoming_varargs): Likewise.
25669         (ft32_function_arg): Likewise.
25670         (ft32_function_arg_advance): Likewise.
25671         (ft32_pass_by_reference): Likewise.
25672         (ft32_arg_partial_bytes): Likewise.
25673         (ft32_valid_pointer_mode): Likewise.
25674         (ft32_addr_space_pointer_mode): Likewise.
25675         (ft32_addr_space_legitimate_address_p): Likewise.
25676         * config/i386/i386-protos.h (ix86_operands_ok_for_move_multiple):
25677         Likewise.
25678         * config/i386/i386.c (ix86_setup_incoming_vararg_bounds): Likewise.
25679         (ix86_emit_outlined_ms2sysv_restore): Likewise.
25680         (iamcu_alignment): Likewise.
25681         (canonicalize_vector_int_perm): Likewise.
25682         (ix86_noce_conversion_profitable_p): Likewise.
25683         (ix86_mpx_bound_mode): Likewise.
25684         (ix86_operands_ok_for_move_multiple): Likewise.
25685         * config/microblaze/microblaze-protos.h
25686         (microblaze_expand_conditional_branch_reg): Likewise.
25687         * config/microblaze/microblaze.c
25688         (microblaze_expand_conditional_branch_reg): Likewise.
25689         * config/powerpcspe/powerpcspe.c (rs6000_init_hard_regno_mode_ok):
25690         Likewise.
25691         (rs6000_reassociation_width): Likewise.
25692         (rs6000_invalid_binary_op): Likewise.
25693         (fusion_p9_p): Likewise.
25694         (emit_fusion_p9_load): Likewise.
25695         (emit_fusion_p9_store): Likewise.
25696         * config/riscv/riscv-protos.h (riscv_regno_mode_ok_for_base_p):
25697         Likewise.
25698         (riscv_hard_regno_mode_ok_p): Likewise.
25699         (riscv_address_insns): Likewise.
25700         (riscv_split_symbol): Likewise.
25701         (riscv_legitimize_move): Likewise.
25702         (riscv_function_value): Likewise.
25703         (riscv_hard_regno_nregs): Likewise.
25704         (riscv_expand_builtin): Likewise.
25705         * config/riscv/riscv.c (riscv_build_integer_1): Likewise.
25706         (riscv_build_integer): Likewise.
25707         (riscv_split_integer): Likewise.
25708         (riscv_legitimate_constant_p): Likewise.
25709         (riscv_cannot_force_const_mem): Likewise.
25710         (riscv_regno_mode_ok_for_base_p): Likewise.
25711         (riscv_valid_base_register_p): Likewise.
25712         (riscv_valid_offset_p): Likewise.
25713         (riscv_valid_lo_sum_p): Likewise.
25714         (riscv_classify_address): Likewise.
25715         (riscv_legitimate_address_p): Likewise.
25716         (riscv_address_insns): Likewise.
25717         (riscv_load_store_insns): Likewise.
25718         (riscv_force_binary): Likewise.
25719         (riscv_split_symbol): Likewise.
25720         (riscv_force_address): Likewise.
25721         (riscv_legitimize_address): Likewise.
25722         (riscv_move_integer): Likewise.
25723         (riscv_legitimize_const_move): Likewise.
25724         (riscv_legitimize_move): Likewise.
25725         (riscv_address_cost): Likewise.
25726         (riscv_subword): Likewise.
25727         (riscv_output_move): Likewise.
25728         (riscv_canonicalize_int_order_test): Likewise.
25729         (riscv_emit_int_order_test): Likewise.
25730         (riscv_function_arg_boundary): Likewise.
25731         (riscv_pass_mode_in_fpr_p): Likewise.
25732         (riscv_pass_fpr_single): Likewise.
25733         (riscv_pass_fpr_pair): Likewise.
25734         (riscv_get_arg_info): Likewise.
25735         (riscv_function_arg): Likewise.
25736         (riscv_function_arg_advance): Likewise.
25737         (riscv_arg_partial_bytes): Likewise.
25738         (riscv_function_value): Likewise.
25739         (riscv_pass_by_reference): Likewise.
25740         (riscv_setup_incoming_varargs): Likewise.
25741         (riscv_print_operand): Likewise.
25742         (riscv_elf_select_rtx_section): Likewise.
25743         (riscv_save_restore_reg): Likewise.
25744         (riscv_for_each_saved_reg): Likewise.
25745         (riscv_register_move_cost): Likewise.
25746         (riscv_hard_regno_mode_ok_p): Likewise.
25747         (riscv_hard_regno_nregs): Likewise.
25748         (riscv_class_max_nregs): Likewise.
25749         (riscv_memory_move_cost): Likewise.
25750         * config/rl78/rl78-protos.h (rl78_split_movsi): Likewise.
25751         * config/rl78/rl78.c (rl78_split_movsi): Likewise.
25752         (rl78_addr_space_address_mode): Likewise.
25753         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
25754         Likewise.
25755         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
25756         (rs6000_reassociation_width): Likewise.
25757         (rs6000_invalid_binary_op): Likewise.
25758         (fusion_p9_p): Likewise.
25759         (emit_fusion_p9_load): Likewise.
25760         (emit_fusion_p9_store): Likewise.
25761         * config/visium/visium-protos.h (prepare_move_operands): Likewise.
25762         (ok_for_simple_move_operands): Likewise.
25763         (ok_for_simple_move_strict_operands): Likewise.
25764         (ok_for_simple_arith_logic_operands): Likewise.
25765         (visium_legitimize_reload_address): Likewise.
25766         (visium_select_cc_mode): Likewise.
25767         (output_cbranch): Likewise.
25768         (visium_split_double_move): Likewise.
25769         (visium_expand_copysign): Likewise.
25770         (visium_expand_int_cstore): Likewise.
25771         (visium_expand_fp_cstore): Likewise.
25772         * config/visium/visium.c (visium_pass_by_reference): Likewise.
25773         (visium_function_arg): Likewise.
25774         (visium_function_arg_advance): Likewise.
25775         (visium_libcall_value): Likewise.
25776         (visium_setup_incoming_varargs): Likewise.
25777         (visium_legitimate_constant_p): Likewise.
25778         (visium_legitimate_address_p): Likewise.
25779         (visium_legitimize_address): Likewise.
25780         (visium_secondary_reload): Likewise.
25781         (visium_register_move_cost): Likewise.
25782         (visium_memory_move_cost): Likewise.
25783         (prepare_move_operands): Likewise.
25784         (ok_for_simple_move_operands): Likewise.
25785         (ok_for_simple_move_strict_operands): Likewise.
25786         (ok_for_simple_arith_logic_operands): Likewise.
25787         (visium_function_value_1): Likewise.
25788         (rtx_ok_for_offset_p): Likewise.
25789         (visium_legitimize_reload_address): Likewise.
25790         (visium_split_double_move): Likewise.
25791         (visium_expand_copysign): Likewise.
25792         (visium_expand_int_cstore): Likewise.
25793         (visium_expand_fp_cstore): Likewise.
25794         (visium_split_cstore): Likewise.
25795         (visium_select_cc_mode): Likewise.
25796         (visium_split_cbranch): Likewise.
25797         (output_cbranch): Likewise.
25798         (visium_print_operand_address): Likewise.
25799         * expmed.c (flip_storage_order): Likewise.
25800         * expmed.h (emit_cstore): Likewise.
25801         (flip_storage_order): Likewise.
25802         * genrecog.c (validate_pattern): Likewise.
25803         * hsa-gen.c (gen_hsa_addr): Likewise.
25804         * internal-fn.c (expand_arith_overflow): Likewise.
25805         * ira-color.c (allocno_copy_cost_saving): Likewise.
25806         * lra-assigns.c (find_hard_regno_for_1): Likewise.
25807         * lra-constraints.c (prohibited_class_reg_set_mode_p): Likewise.
25808         (process_invariant_for_inheritance): Likewise.
25809         * lra-eliminations.c (move_plus_up): Likewise.
25810         * omp-low.c (lower_oacc_reductions): Likewise.
25811         * simplify-rtx.c (simplify_subreg): Likewise.
25812         * target.def (TARGET_SETUP_INCOMING_VARARG_BOUNDS): Likewise.
25813         (TARGET_CHKP_BOUND_MODE): Likewise..
25814         * targhooks.c (default_chkp_bound_mode): Likewise.
25815         (default_setup_incoming_vararg_bounds): Likewise.
25816         * targhooks.h (default_chkp_bound_mode): Likewise.
25817         (default_setup_incoming_vararg_bounds): Likewise.
25818         * tree-ssa-math-opts.c (divmod_candidate_p): Likewise.
25819         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
25820         (have_whole_vector_shift): Likewise.
25821         * tree-vect-stmts.c (vectorizable_load): Likewise.
25822         * doc/tm.texi: Regenerate.
25824 2017-07-05  Georg-Johann Lay  <avr@gjlay.de>
25826         Graceful degrade if Binutils PR21472 is not available.
25828         PR target/81072
25829         * configure.ac [target=avr]: WARN instead of ERROR if avrxmega3
25830         .rodata in flash test fails.
25831         (HAVE_LD_AVR_AVRXMEGA3_RODATA_IN_FLASH): Define it if test passes.
25832         * confgure: Regenerate.
25833         * config.in: Regenerate.
25834         * config/avr/avr.c (avr_asm_named_section)
25835         [HAVE_LD_AVR_AVRXMEGA3_RODATA_IN_FLASH]: Only trigger
25836         __do_copy_data for stuff in .rodata if flash_pm_offset = 0.
25837         (avr_asm_init_sections): Same.
25839 2017-07-05  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
25841         * config/arm/neon.md (fma<VCVTF:mode>4): Remove %?.
25842         (fma<VH:mode>4_intrinsic): Likewise.
25843         (*fmsub<VCVTF:mode>4): Likewise.
25844         (*fmsub<VH:mode>4_intrinsic): Likewise.
25846 2017-07-05  Georg-Johann Lay  <avr@gjlay.de>
25848         PR target/81305
25849         * config/avr/avr.c (avr_out_movhi_mr_r_xmega) [CONSTANT_ADDRESS_P]:
25850         Don't depend on "optimize > 0".
25851         (out_movhi_r_mr, out_movqi_mr_r): Same.
25852         (out_movhi_mr_r, out_movqi_r_mr): Same.
25853         (avr_address_cost) [CONSTANT_ADDRESS_P]: Don't depend cost for
25854         io_address_operand on "optimize > 0".
25856 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25858         * tree-loop-distribution.c: Add general explanantion on the pass.
25859         (generate_loops_for_partition): Mark distributed loop.
25860         (pg_add_dependence_edges): New parameter.  Handle alias data
25861         dependence specially and record it in the parameter if asked.
25862         (struct pg_vdata, pg_edata, pg_edge_callback_data): New structs.
25863         (init_partition_graph_vertices, add_partition_graph_edge): New.
25864         (pg_skip_alias_edge, free_partition_graph_edata_cb): New.
25865         (free_partition_graph_vdata, build_partition_graph): New.
25866         (sort_partitions_by_post_order, merge_dep_scc_partitions): New.
25867         (pg_collect_alias_ddrs, break_alias_scc_partitions): New.
25868         (data_ref_segment_size, latch_dominated_by_data_ref): New.
25869         (compute_alias_check_pairs, version_loop_by_alias_check): New.
25870         (version_for_distribution_p, finalize_partitions): New.
25871         (distribute_loop): Handle alias data dependence specially.  Factor
25872         out loop fusion code as functions and call these functions.
25874 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25876         * tree-loop-distribution.c (classify_partition): New parameter and
25877         better handle reduction statement.
25878         (rdg_build_partitions): Revise comment.
25879         (distribute_loop): Compute statements in all partitions and pass it
25880         to classify_partition.
25882 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25884         * tree-loop-distribution.c (enum partition_type): New.
25885         (struct partition): New field type.
25886         (partition_merge_into): Add parameter.  Update partition type.
25887         (data_dep_in_cycle_p, update_type_for_merge): New functions.
25888         (build_rdg_partition_for_vertex): Compute partition type.
25889         (rdg_build_partitions): Dump partition type.
25890         (distribute_loop): Update calls to partition_merge_into.
25892 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25894         * tree-loop-distribution.c (struct ddr_hasher): New.
25895         (ddr_hasher::hash, ::equal, get_data_dependence): New function.
25896         (ddrs_table): New.
25897         (classify_partition): Call get_data_dependence.
25898         (pg_add_dependence_edges): Ditto.
25899         (distribute_loop): Release data dependence hash table.
25901 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25903         * tree-loop-distribution.c (ref_base_address): Delete.
25904         (similar_memory_accesses): Rename ...
25905         (share_memory_accesses): ... to this.  Check if partitions access
25906         the same memory reference.
25907         (distribute_loop): Call share_memory_accesses.
25909 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25911         * tree-loop-distribution.c (struct partition): New field recording
25912         its data reference.
25913         (partition_alloc, partition_free): Init and release data refs.
25914         (partition_merge_into): Merge data refs.
25915         (build_rdg_partition_for_vertex): Collect data refs for partition.
25916         (pg_add_dependence_edges): Change parameters from vector to bitmap.
25917         Update uses.
25918         (distribute_loop): Remve data refs from vertice data of partition
25919         graph.
25921 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25923         * tree-loop-distribution.c (params.h): Include header file.
25924         (MAX_DATAREFS_NUM, DR_INDEX): New macro.
25925         (datarefs_vec): New global var.
25926         (create_rdg_vertices): Use datarefs_vec directly.
25927         (free_rdg): Don't free data references.
25928         (build_rdg): Update use.  Don't free data references.
25929         (distribute_loop): Compute global variable for data references.
25930         Bail out if there are too many data references.
25932 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25934         * tree-loop-distribution.c (loop_nest): New global var.
25935         (build_rdg): Use loop directly, rather than loop nest.
25936         (pg_add_dependence_edges): Remove loop nest parameter.  Use global
25937         variable directly.
25938         (distribute_loop): Compute global variable loop nest.  Update use.
25940 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25942         * tree-loop-distribution.c (enum fuse_type, fuse_message): New.
25943         (partition_merge_into): New parameter.  Dump reason for fusion.
25944         (distribute_loop): Update use of partition_merge_into.
25946 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25948         * tree-loop-distribution.c (bb_top_order_index): New.
25949         (bb_top_order_index_size, bb_top_order_cmp): New.
25950         (stmts_from_loop): Use topological order.
25951         (pass_loop_distribution::execute): Compute and release topological
25952         order for basic blocks.
25954 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25956         * tree-loop-distribution.c (pass_loop_distribution::execute): Skip
25957         if no loops.
25959 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25961         * cfgloop.h (struct loop): Add comment.  New field orig_loop_num.
25962         * cfgloopmanip.c (lv_adjust_loop_entry_edge): Comment change.
25963         * internal-fn.c (expand_LOOP_DIST_ALIAS): New function.
25964         * internal-fn.def (LOOP_DIST_ALIAS): New.
25965         * tree-vectorizer.c (fold_loop_vectorized_call): Rename to ...
25966         (fold_loop_internal_call): ... this.
25967         (vect_loop_dist_alias_call): New function.
25968         (set_uid_loop_bbs): Call fold_loop_internal_call.
25969         (vectorize_loops): Fold IFN_LOOP_VECTORIZED and IFN_LOOP_DIST_ALIAS
25970         internal calls.
25972 2017-07-04  Uros Bizjak  <ubizjak@gmail.com>
25974         PR target/81300
25975         * config/i386/i386.md (setcc + movzbl/and to xor + setcc peepholes):
25976         Require dead FLAGS_REG at the beginning of a peephole.
25978 2017-07-04  Uros Bizjak  <ubizjak@gmail.com>
25980         PR target/81294
25981         * config/i386/adxintrin.h (_subborrow_u32): Swap _X and _Y
25982         arguments in the call to __builtin_ia32_sbb_u32.
25983         (_subborrow_u64): Swap _X and _Y arguments in the call to
25984         __builtin_ia32_sbb_u64.
25986 2017-07-04  Jakub Jelinek  <jakub@redhat.com>
25988         PR debug/81278
25989         * tree-vrp.c (compare_assert_loc): Turn into a function template
25990         with stable template parameter.  Only test if a->e is NULL,
25991         !a->e == !b->e has been verified already.  Use e == NULL or
25992         e != NULL instead of e or ! e tests.  If stable is true, don't use
25993         iterative_hash_expr, on the other side allow a or b or both NULL
25994         and sort the NULLs last.
25995         (process_assert_insertions): Sort using compare_assert_loc<false>
25996         instead of compare_assert_loc, later sort using
25997         compare_assert_loc<true> before calling process_assert_insertions_for
25998         in a loop.  Use break instead of continue once seen NULL pointer.
26000 2017-07-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
26002         * config/arm/driver-arm.c (arm_cpu_table): Add entry for ARM
26003         Cortex-R7 and Cortex-R8 processors.
26005 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
26007         * ipa-utils.c (ipa_merge_profiles): Fix merging when dst is
26008         uninitialized while src is not.
26010 2017-07-04  Richard Earnshaw  <rearnsha@arm.com>
26012         * common/config/arm/arm-common.c: Adjust include path for
26013         arm-cpu-cdata.h
26014         * config/arm/t-arm (TM_H): Adjust path for arm-cpu.h.
26015         (arm-cpu.h): Create in build directory.  Adjust dependency rules.
26016         (arm-cpu-data.h): Likewise.
26017         (arm-cpu-cdata.h): Likewise.
26018         * config/arm/arm-cpu.h: Delete.
26019         * config/arm/arm-cpu-cdata.h: Delete.
26020         * config/arm/arm-cpu-data.h: Delete.
26022 2017-07-04  James Greenhalgh  <james.greenhalgh@arm.com>
26024         * config/arm/arm-cpus.in (cortex-a55): New.
26025         (cortex-a75): Likewise.
26026         (cortex-a75.cortex-a55): Likewise.
26027         * config/arm/driver-arm.c (arm_cpu_table): Add cortex-a55 and
26028         cortex-a75.
26029         * doc/invoke.texi (-mcpu): Document cortex-a55 and cortex-a75.
26030         * config/arm/arm-cpu-cdata.h: Regenerate.
26031         * config/arm/arm-cpu-data.h: Regenerate.
26032         * config/arm/arm-cpu.h: Regenerate.
26033         * config/arm/arm-tables.opt: Regenerate.
26034         * config/arm/arm-tune.md: Regenerate.
26036 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
26038         * haifa-sched.c (sched_create_recovery_edges): Update profile.
26040 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
26042         * bb-reorder.c (better_edge_p): Fix handling of uninitialized
26043         probability.
26045 2017-07-04  Richard Sandiford  <richard.sandiford@linaro.org>
26047         PR tree-optimization/81292
26048         * tree-ssa-strlen.c (handle_builtin_strlen): When setting
26049         full_string_p, also call adjust_related_strinfos if the adjustment
26050         is simple, otherwise invalidate related strinfos.
26052 2017-07-04  Martin Liska  <mliska@suse.cz>
26054         PR sanitizer/81040
26055         * sanopt.c (sanitize_rewrite_addressable_params): Mark the
26056         newly created variable as DECL_IGNORED_P.
26058 2017-07-04  Martin Liska  <mliska@suse.cz>
26060         PR ipa/81293
26061         * ipa-inline.c (inline_small_functions):
26062         Use xstrdup_for_dump.
26064 2017-07-04  Tom de Vries  <tom@codesourcery.com>
26066         * graph.c (draw_cfg_edges): Save and restore EDGE_DFS_BACK.
26068 2017-07-03  Dominique d'Humieres  <dominiq@lps.ens.fr>
26070         PR target/81033
26071         * config/darwin.c (darwin_function_switched_text_sections):
26072         Fix spaces.
26074 2017-07-03  Jan Hubicka  <hubicka@ucw.cz>
26076         * tree-vect-loop-manip.c (vect_do_peeling): Fix scaling up.
26078 2017-07-03  Richard Earnshaw  <rearnsha@arm.com>
26080         * doc/invoke.texi (ARM Options): Add -mbe8 and -mbe32 to option summary.
26082 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26084         * tree-vect-loop.c (vect_analyze_loop_2): Treat min_scalar_loop_bound,
26085         min_profitable_iters, and th as inclusive lower bounds.
26086         Fix LOOP_VINFO_PEELING_FOR_GAPS condition.
26087         (vect_estimate_min_profitable_iters): Return inclusive lower bounds
26088         for min_profitable_iters and min_profitable_estimate.
26089         (vect_transform_loop): Treat th as an inclusive lower bound.
26090         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
26092 2017-07-03  Dominique d'Humieres  <dominiq@lps.ens.fr>
26094         PR target/81033
26095         * config/darwin.c (darwin_function_switched_text_sections):
26096         Replace DECL_NAME with DECL_ASSEMBLER_NAME, split assemble_name_raw
26097         in two pieces, and suppress the use of buf.
26099 2017-07-03  Nathan Sidwell  <nathan@acm.org>
26101         * hash-table.h (hash_table_mod1): Fix indentation.
26103 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26105         PR middle-end/81290
26106         * predict.c (force_edge_cold): Be more careful about propagation
26107         backward.
26108         * profile-count.h (profile_probability::guessed,
26109         profile_probability::fdo, profile_count::guessed, profile_count::fdo):
26110         New.
26111         * tree-ssa-threadupdate.c (recompute_probabilities): Result is guessed.
26113 2017-07-03  James Greenhalgh  <james.greenhalgh@arm.com>
26115         * doc/invoke.texi (rcpc architecture extension): Document it.
26117 2017-07-03  Richard Biener  <rguenther@suse.de>
26119         PR tree-optimization/60510
26120         * tree-vect-loop.c (vect_create_epilog_for_reduction): Pass in
26121         the scalar reduction PHI and use it.
26122         (vectorizable_reduction): Properly guard the single_defuse_cycle
26123         path for non-SLP reduction chains where we cannot use it.
26124         Rework reduc_def/index and vector type deduction.  Rework
26125         vector operand gathering during reduction op code-gen.
26126         * tree-vect-slp.c (vect_analyze_slp): For failed SLP reduction
26127         chains dissolve the chain and leave it to non-SLP reduction
26128         handling.
26130 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26132         * tree-data-ref.h (dr_alignment): Declare.
26133         * tree-data-ref.c (dr_alignment): New function.
26134         * tree-vectorizer.h (dataref_aux): Remove base_element_aligned.
26135         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
26136         set it.
26137         * tree-vect-stmts.c (vectorizable_store): Use dr_alignment.
26139 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26141         * tree-data-ref.h (innermost_loop_behavior): Add base_alignment
26142         and base_misalignment fields.
26143         (DR_BASE_ALIGNMENT, DR_BASE_MISALIGNMENT): New macros.
26144         * tree-data-ref.c: Include builtins.h.
26145         (dr_analyze_innermost): Set up the new innmost_loop_behavior fields.
26146         * tree-vectorizer.h (STMT_VINFO_DR_BASE_ALIGNMENT): New macro.
26147         (STMT_VINFO_DR_BASE_MISALIGNMENT): Likewise.
26148         * tree-vect-data-refs.c: Include tree-cfg.h.
26149         (vect_compute_data_ref_alignment): Use the new innermost_loop_behavior
26150         fields instead of calculating an alignment here.
26151         (vect_analyze_data_refs): Use dr_analyze_innermost.  Dump the new
26152         innermost_loop_behavior fields.
26154 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26156         * tree-data-ref.h (innermost_loop_behavior): Add a step_alignment
26157         field.
26158         (DR_STEP_ALIGNMENT): New macro.
26159         * tree-vectorizer.h (STMT_VINFO_DR_STEP_ALIGNMENT): Likewise.
26160         * tree-data-ref.c (dr_analyze_innermost): Initalize step_alignment.
26161         (create_data_ref): Print it.
26162         * tree-vect-stmts.c (vectorizable_load): Use the step alignment
26163         to tell whether the step preserves vector (mis)alignment.
26164         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Likewise.
26165         Move the check for an integer step and generalise to all INTEGER_CST.
26166         (vect_analyze_data_refs): Set DR_STEP_ALIGNMENT when setting DR_STEP.
26167         Print the outer step alignment.
26169 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26171         * tree-data-ref.h (innermost_loop_behavior): Replace aligned_to
26172         with offset_alignment.
26173         (DR_ALIGNED_TO): Delete.
26174         (DR_OFFSET_ALIGNMENT): New macro.
26175         * tree-vectorizer.h (STMT_VINFO_DR_ALIGNED_TO): Delete.
26176         (STMT_VINFO_DR_OFFSET_ALIGNMENT): New macro.
26177         * tree-data-ref.c (dr_analyze_innermost): Update after above changes.
26178         (create_data_ref): Likewise.
26179         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Likewise.
26180         (vect_analyze_data_refs): Likewise.
26181         * tree-if-conv.c (if_convertible_loop_p_1): Use memset before
26182         creating dummy innermost behavior.
26184 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26186         * tree-data-ref.h (dr_analyze_innermost): Replace the dr argument
26187         with a "innermost_loop_behavior *" and refeence tree.
26188         * tree-data-ref.c (dr_analyze_innermost): Likewise.
26189         (create_data_ref): Update call accordingly.
26190         * tree-predcom.c (find_looparound_phi): Likewise.
26192 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26194         * tree-vectorizer.h (_stmt_vec_info): Replace individual dr_*
26195         fields with dr_wrt_vec_loop.
26196         (STMT_VINFO_DR_BASE_ADDRESS, STMT_VINFO_DR_INIT, STMT_VINFO_DR_OFFSET)
26197         (STMT_VINFO_DR_STEP, STMT_VINFO_DR_ALIGNED_TO): Update accordingly.
26198         (STMT_VINFO_DR_WRT_VEC_LOOP): New macro.
26199         (vect_dr_behavior): New function.
26200         (vect_create_addr_base_for_vector_ref): Remove loop parameter.
26201         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
26202         vect_dr_behavior.  Use a step_preserves_misalignment_p boolean to
26203         track whether the step preserves the misalignment.
26204         (vect_create_addr_base_for_vector_ref): Remove loop parameter.
26205         Use vect_dr_behavior.
26206         (vect_setup_realignment): Update call accordingly.
26207         (vect_create_data_ref_ptr): Likewise.  Use vect_dr_behavior.
26208         * tree-vect-loop-manip.c (vect_gen_prolog_loop_niters): Update
26209         call to vect_create_addr_base_for_vector_ref.
26210         (vect_create_cond_for_align_checks): Likewise.
26211         * tree-vect-patterns.c (vect_recog_bool_pattern): Copy
26212         STMT_VINFO_DR_WRT_VEC_LOOP as a block.
26213         (vect_recog_mask_conversion_pattern): Likewise.
26214         * tree-vect-stmts.c (compare_step_with_zero): Use vect_dr_behavior.
26215         (new_stmt_vec_info): Remove redundant zeroing.
26217 2017-07-03  Richard Earnshaw  <rearnsha@arm.com>
26219         * common/config/arm/arm-common.c (arm_be8_option): New function.
26220         * config/arm/arm-isa.h (isa_feature): Add new feature bit isa_bit_be8.
26221         (ISA_ARMv6): Add isa_bit_be8.
26222         * config/arm/arm.h (arm_be8_option): Add prototype.
26223         (BE8_SPEC_FUNCTION): New define.
26224         (EXTRA_SPEC_FUNCTIONS): Add BE8_SPEC_FUNCTION.
26225         * config/arm/arm.opt (mbig-endian): Mark as Negative of mlittle-endian.
26226         (mlittle-endian): Similarly.
26227         (mbe8, mbe32): New options.
26228         * config/arm/bpabi.h (BE8_LINK_SPEC): Call arm_be8_option.
26229         * doc/invoke.texi (ARM Options): Document -mbe8 and -mbe32.
26231 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26233         * tree-cfg.c (gimple_find_sub_bbs): Fix profile updating.
26235 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26237         * tree-cfgcleanup.c (want_merge_blocks_p): New function.
26238         (cleanup_tree_cfg_bb): Use it.
26239         * profile-count.h (profile_count::of_for_merging, profile_count::merge):
26240         New functions.
26241         * tree-cfg.c (gimple_merge_blocks): Use profile_count::merge.
26243 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26245         PR bootstrap/81285
26246         * loop-doloop.c (add_test): Update profile.
26248 2017-07-03  Martin Liska  <mliska@suse.cz>
26250         PR sanitize/81040
26251         * sanopt.c (rewrite_usage_of_param): New function.
26252         (sanitize_rewrite_addressable_params): Likewise.
26253         (pass_sanopt::execute): Call rewrite_usage_of_param.
26255 2017-07-03  Richard Biener  <rguenther@suse.de>
26257         * tree-vect-loop.c (vect_create_epilog_for_reduction): Revert
26258         back to using VIEW_CONVERT_EXPR.
26260 2017-07-03  Martin Liska  <mliska@suse.cz>
26262         PR other/78366
26263         * doc/extend.texi: Document when a resolver function is
26264         generated for target_clones.
26266 2017-07-03  Martin Liska  <mliska@suse.cz>
26268         * asan.c (asan_emit_stack_protection): Unpoison just red zones
26269         and shadow memory of auto variables which are subject of
26270         use-after-scope sanitization.
26271         (asan_expand_mark_ifn): Add do set only when is_poison.
26273 2016-07-03  Richard Biener  <rguenther@suse.de>
26275         * tree-vect-loop.c (vect_analyze_loop_operations): Also analyze
26276         reduction PHIs.
26277         (vect_force_simple_reduction): Record reduction def -> phi mapping.
26278         (vectorizable_reduction): Perform reduction PHI creation when
26279         visiting a reduction PHI and adjust and simplify code generation
26280         phase of the reduction op.  Cache dts, use fold_binary, not fold_build2.
26281         (vect_transform_loop): Visit reduction PHIs.
26282         * tree-vect-slp.c (vect_get_and_check_slp_defs): Record reduction
26283         defs into the SLP tree.
26284         (vect_build_slp_tree): Reduction defs terminate the recursion.
26285         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Allow lookup
26286         of reduction defs.
26287         (vect_get_vec_defs_for_stmt_copy): Export.
26288         (vect_get_vec_defs): Likewise.
26289         * tree-vectorizer.h (struct _stmt_vec_info): Amend reduc_def
26290         purpose.
26291         (vect_get_vec_defs_for_stmt_copy): Declare.
26292         (vect_get_vec_defs): Likewise.
26294 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26296         * tree-data-ref.c (dr_analyze_innermost): Replace the "nest"
26297         parameter with a "loop" parameter and use it instead of the
26298         loop containing DR_STMT.  Don't check simple_iv when doing
26299         BB analysis.  Describe the two analysis modes in the comment.
26301 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26303         PR tree-optimization/69468
26304         * tree-ssa-tail-merge.c (ignore_edge_flags): New constant.
26305         (find_same_succ_bb): Handle ignore_edge_flags.
26307 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26309         PR tree-optimization/81192
26310         * tree-ssa-tail-merge.c (same_succ_hash): Use bb->loop_father->num in
26311         hash.
26312         (same_succ::equal): Don't find bbs to be equal if bb->loop_father
26313         differs.
26314         (find_same_succ_bb): Remove obsolete test on bb->loop_father->latch.
26316 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26318         PR tree-optimization/81192
26319         * tree-ssa-tail-merge.c (same_succ_flush_bb): Handle
26320         BB_SAME_SUCC (bb) == NULL.
26322 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26324         * cfgrtl.c (rtl_verify_edges): Enable checking of profile_probability
26325         consistency.
26327 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26329         * dumpfile.c: Include profile-count.h
26330         * tree-cfg.c (gimple_duplicate_sese_tail): Drop UNUSED attributes;
26331         update profile.
26332         (insert_cond_bb): Update profile.
26333         * tree-cfg.h (insert_cond_bb): Update prototype.
26334         * tree-chkp-opt.c (chkp_optimize_string_function_calls): Update.
26335         * tree-dump.c: Do not include tree-cfg.
26337 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26339         * bb-reorder.c (fix_up_crossing_landing_pad): Update profile.
26341 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26343         * expect.c (dw2_build_landing_pads): Update profile of the landing pad
26344         bb.
26346 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26348         * tree-complex.c (expand_complex_div_wide): update profile.
26350 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26351             Alan Hayward  <alan.hayward@arm.com>
26352             David Sherwood  <david.sherwood@arm.com>
26354         * Makefile.in (MACHMODE_H): Remove insn-modes.h
26355         (CORETYPES_H): New define.
26356         (MOSTLYCLEANFILES): Add insn-modes-inline.h.
26357         (insn-modes-inline.h, s-modes-inline-h): New rules.
26358         (generated_files): Add insn-modes-inline.h.
26359         (RTL_BASE_H, TREE_CORE_H): Use CORETYPES_H instead of coretypes.h.
26360         (build/gensupport.o, build/ggc-none.o, build/print-rtl.o): Likewise.
26361         (build/read-md.o, build/read-rtl.o, build/rtl.o): Likewise.
26362         (build/vec.o, build/hash-table.o, build/inchash.o): Likewise.
26363         (build/gencondmd.o, build/genattr.o, build/genattr-common.o): Likewise.
26364         (build/genattrtab.o, build/genautomata.o, build/gencheck.o): Likewise.
26365         (build/gencodes.o, build/genconditions.o): Likewise.
26366         (build/genconfig.o, build/genconstants.o, build/genemit.o): Likewise.
26367         (build/genenums.o, build/genextract.o, build/genflags.o): Likewise.
26368         (build/gentarget-def.o, build/genmddeps.o, build/genopinit.o)
26369         (build/genoutput.o, build/genpeep.o, build/genpreds.o): Likewise.
26370         (build/genrecog.o, build/genmddump.o, build/genmatch.o): Likewise.
26371         (build/gencfn-macros.o, build/gcov-iov.o): Likewise.
26372         * coretypes.h: Include everything up to real.h for generators.
26373         Include insn-modes.h first.  Include wide-int-print.h after
26374         wide-int.h.  Include insn-modes-inline.h and then machmode.h.
26375         * machmode.h: Don't include insn-modes.h here.
26376         * function-tests.c: Remove includes of signop.h, machmode.h,
26377         double-int.h and wide-int.h.
26378         * rtl.h: Likewise.
26379         * gcc-rich-location.c: Remove includes of machmode.h, double-int.h
26380         and wide-int.h.
26381         * optc-save-gen.awk: Likewise.
26382         * gencheck.c (BITS_PER_UNIT): Delete dummy definition.
26383         * godump.c: Remove include of wide-int-print.h.
26384         * pretty-print.h: Likewise.
26385         * wide-int-print.cc: Likewise.
26386         * wide-int.cc: Likewise.
26387         * hash-map-tests.c: Remove include of signop.h.
26388         * hash-set-tests.c: Likewise.
26389         * rtl-tests.c: Likewise.
26390         * mkconfig.sh: Remove include of machmode.h.
26391         * genmodes.c (emit_insn_modes_h): Split emission of inline functions
26392         into...
26393         (emit_insn_modes_inline_h): ...this new function.  Emit the code
26394         into an insn-modes-inline.h header file, adding appropriate
26395         include guards and end comments.
26396         (emit_insn_modes_c_header): Remove include of machmode.h.
26397         (emit_min_insn_modes_c_header): Include coretypes.h rather than
26398         machmode.h.
26399         (main): Handle -i flag and call emit_insn_modes_inline_h when
26400         it is passed.
26402 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26404         * tree-ssa-strlen.c (strinfo): Rename the length field to
26405         nonzero_chars.  Add a full_string_p field.
26406         (compare_nonzero_chars, zero_length_string_p): New functions.
26407         (get_addr_stridx): Add an offset_out parameter.
26408         Use compare_nonzero_chars.
26409         (get_stridx): Update accordingly.  Use compare_nonzero_chars.
26410         (new_strinfo): Update after above changes to strinfo.
26411         (set_endptr_and_length): Set full_string_p.
26412         (get_string_length): Update after above changes to strinfo.
26413         (unshare_strinfo): Update call to new_strinfo.
26414         (maybe_invalidate): Likewise.
26415         (get_stridx_plus_constant): Change off to unsigned HOST_WIDE_INT.
26416         Use compare_nonzero_chars and zero_string_p.  Treat nonzero_chars
26417         as a uhwi instead of an shwi.  Update after above changes to
26418         strinfo and new_strinfo.
26419         (zero_length_string): Assert that chainsi contains full strings.
26420         Use zero_length_string_p.  Update call to new_strinfo.
26421         (adjust_related_strinfos): Update after above changes to strinfo.
26422         Copy full_string_p from origsi.
26423         (adjust_last_stmt): Use zero_length_string_p.
26424         (handle_builtin_strlen): Update after above changes to strinfo and
26425         new_strinfo.  Install the lhs as the string length if the previous
26426         entry didn't describe a full string.
26427         (handle_builtin_strchr): Update after above changes to strinfo
26428         and new_strinfo.
26429         (handle_builtin_strcpy): Likewise.
26430         (handle_builtin_strcat): Likewise.
26431         (handle_builtin_malloc): Likewise.
26432         (handle_pointer_plus): Likewise.
26433         (handle_builtin_memcpy): Likewise.  Track nonzero characters
26434         that aren't necessarily followed by a nul terminator.
26435         (handle_char_store): Likewise.
26437 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26439         PR tree-optimization/80769
26440         * tree-ssa-strlen.c (strinfo): Document that "stmt" is also used
26441         for malloc and calloc.  Document the new invariant that all related
26442         strinfos have delayed lengths or none do.
26443         (verify_related_strinfos): Move earlier in file.
26444         (set_endptr_and_length): New function, split out from...
26445         (get_string_length): ...here.  Also set the lengths of related
26446         strinfos.
26447         (zero_length_string): Assert that chainsi has known (rather than
26448         delayed) lengths.
26449         (adjust_related_strinfos): Likewise.
26451 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26453         PR tree-optimization/81136
26454         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Only
26455         assert that two references with the same misalignment have the same
26456         compile-time misalignment if those compile-time misalignments
26457         are known.
26459 2017-07-01  Andi Kleen  <ak@linux.intel.com>
26461         * print-tree.c (print_node): Print all attributes.
26463 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26465         * cfg.c (scale_bbs_frequencies): New function.
26466         * cfg.h (scale_bbs_frequencies): Declare it.
26467         * cfgloopanal.c (single_likely_exit): Cleanup.
26468         * cfgloopmanip.c (scale_loop_frequencies): Take profile_probability
26469         as parameter.
26470         (scale_loop_profile): Likewise.
26471         (loop_version): Likewise.
26472         (create_empty_loop_on_edge): Update.
26473         * cfgloopmanip.h (scale_loop_frequencies, scale_loop_profile,
26474         scale_loop_frequencies, scale_loop_profile, loopify,
26475         loop_version): Update prototypes.
26476         * modulo-sched.c (sms_schedule): Update.
26477         * predict.c (unlikely_executed_edge_p): Also check probability.
26478         (probably_never_executed_edge_p): Fix typo.
26479         * tree-if-conv.c (version_loop_for_if_conversion): Update.
26480         * tree-parloops.c (gen_parallel_loop): Update.
26481         * tree-ssa-loop-ivcanon.c (try_peel_loop): Update.
26482         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
26483         * tree-ssa-loop-split.c (split_loop): Update.
26484         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Update.
26485         * tree-vect-loop-manip.c (vect_do_peeling): Update.
26486         (vect_loop_versioning): Update.
26487         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
26489 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26491         * trans-mem.c (split_bb_make_tm_edge): Update profile.
26493 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26495         * tree-if-conv.c (combine_blocks): Use make_single_succ_edge
26496         to keep profile consistent.
26498 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26500         * cfgrtl.c (rtl_flow_call_edges_add): Update profile.
26501         * tree-cfg.c (gimple_flow_call_edges_add): Likewise.
26502         * profile-count.h (max_safe_multiplier): Make unsigned.
26503         (profile_count::guessed_zero): New.
26505 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26507         * bb-reorder.c (fix_up_crossing_landing_pad,
26508         fix_crossing_conditional_branches): Use make_single_succ_edge
26509         to keep profile consistent.
26511 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26513         * tree-vect-loop.c (optimize_mask_stores): Use make_single_succ_edge
26514         to update profile.
26516 2017-07-01  Jakub Jelinek  <jakub@redhat.com>
26518         PR sanitizer/81262
26519         * bb-reorder.c (fix_up_fall_thru_edges): Move variable declarations to
26520         the right scopes, make sure cond_jump isn't preserved between multiple
26521         iterations.  Search for fallthru edge whenever there are 3+ edges and
26522         use find_fallthru_edge for it.
26524 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26526         Patch by Alexander Monakov <amonakov@ispras.ru>
26527         * sel-sched-ir.c (compute_succs_info): Handle uninitialized
26528         probabilities consistently.
26530 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26532         * pa.c (pa_expand_compare_and_swap_loop): Update call of
26533         emit_cmp_and_jump_insns.
26535 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26537         PR ipa/81261
26538         * tree-inline.c (expand_call_inline): Combine profile statuses.
26540 2017-06-30  Andrew Pinski  <apinski@cavium.com>
26542         * tree-if-conv.c (predicate_scalar_phi): Update new_stmt if
26543         fold_stmt returned true.
26545 2017-06-30  Nathan Sidwell  <nathan@acm.org>
26547         * ggc.h (empty_string): Delete.
26548         * cfgexpand.c (expand_asm_stmt): Use plain "".
26549         * optabs.c (expand_asm_memory_barrier): Likewise.
26550         * stringpool.c (empty_string): Delete.
26551         (digit_vector, digit_string): Delete.
26552         (ggc_alloc_string): Use plain "", don't optimize single digit
26553         strings.  Use ggc_alloc_atomic.
26555 2017-06-30  Richard Earnshaw  <rearnsha@arm.com>
26557         * rtlanal.c (insn_rtx_cost): If a parallel contains exactly one
26558         comparison set and one other set, use the cost of the non-comparison
26559         set.
26561 2017-06-30  Nathan Sidwell  <nathan@acm.org>
26563         * ggc.h: Replace all 'static inline' with plain 'inline'.  Fix
26564         some formatting.
26566 2017-06-30  Peter Bergner  <bergner@vnet.ibm.com>
26568         * tree-cfg.c (group_case_labels_stmt): Merge scanning and compressing
26569         loops.  Remove now unneeded calls to gimple_switch_set_label() that
26570         just set removed labels to NULL_TREE.
26572 2017-06-30  Aldy Hernandez  <aldyh@redhat.com>
26574         * tree-ssanames.c (set_range_info_raw): Abstract from ...
26575         (set_range_info): ...here.  Only call set_range_info_raw if domain
26576         is useful.
26577         (set_nonzero_bits): Call set_range_info_raw.
26578         * tree-ssanames.h (set_range_info_raw): New.
26580 2017-06-30  Jakub Jelinek  <jakub@redhat.com>
26582         PR target/81225
26583         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): For V8FI,
26584         V16FI and VI8F_256 iterators, use <store_mask_predicate> instead
26585         of nonimmediate_operand and <store_mask_constraint> instead of m
26586         for the input operand.  For V8FI iterator, always split if input
26587         is a MEM.  For V16FI and V8SF_256 iterators, don't test if both
26588         operands are MEM if <mask_applied>.  For VI4F_256 iterator, use
26589         <store_mask_predicate> instead of register_operand and
26590         <store_mask_constraint> instead of v for the input operand.  Make
26591         sure both operands aren't MEMs for if not <mask_applied>.
26593 2017-06-30  Sylvestre Ledru  <sylvestre@debian.org>
26595         * lto-wrapper.c (copy_file) Close both file descriptors before
26596         exiting normally.
26598 2017-06-30  Martin Liska  <mliska@suse.cz>
26600         PR ipa/81214
26601         * multiple_target.c (create_dispatcher_calls): Make ifunc
26602         also for function that don't have calls or are not referenced.
26604 2017-06-30  Richard Biener  <rguenther@suse.de>
26606         * tree-vect-slp.c (vect_slp_analyze_node_operations): Only
26607         analyze the first scalar stmt.  Move vector type computation
26608         for the BB case here from ...
26609         * tree-vect-stmts.c (vect_analyze_stmt): ... here.  Guard
26610         live operation processing in the SLP case properly.
26612 2017-06-30  Richard Biener  <rguenther@suse.de>
26614         * graph.c (draw_cfg_node_succ_edges): Fix broken dot syntax.
26616 2017-06-30  Martin Liska  <mliska@suse.cz>
26618         PR sanitizer/81021
26619         * tree-eh.c (lower_resx): Call BUILT_IN_ASAN_HANDLE_NO_RETURN
26620         before BUILT_IN_UNWIND_RESUME when ASAN is used.
26622 2017-06-30  Yvan Roux  <yvan.roux@linaro.org>
26624         * doc/invoke.texi (AArch64): Add missing options and remove redundant
26625         ones.
26627 2017-06-30  Richard Biener  <rguenther@suse.de>
26629         PR tree-optimization/81249
26630         * tree-vect-loop.c (vect_create_epilog_for_reduction): Convert
26631         condition reduction result to original scalar type.
26633 2017-06-30  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26635         * profile-count.h (enum profile_quality): Fix typos and whitespace
26636         issues.
26638 2017-06-30  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26640         * config/s390/s390.c (s390_expand_setmem): Adjust to the new data
26641         type for branch probabilities.
26643 2017-06-29  Julian Brown  <julian@codesourcery.com>
26644             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
26646         * config/aarch64/aarch64-fusion-pairs.def: Add ALU_BRANCH entry.
26647         * config/aarch64/aarch64.c (AARCH64_FUSE_ALU_BRANCH): New fusion type.
26648         (thunderx2t99_tunings): Set AARCH64_FUSE_ALU_BRANCH flag.
26649         (aarch_macro_fusion_pair_p): Add support for AARCH64_FUSE_ALU_BRANCH.
26651 2017-06-29  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
26653         * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Push the
26654         check for CC usage into AARCH64_FUSE_CMP_BRANCH.
26655         * config/i386/i386.c (ix86_macro_fusion_pair_p): Push the check for
26656         CC usage from generic code to here.
26657         * sched-deps.c (sched_macro_fuse_insns): Move the condition for
26658         CC usage into the target macros.
26660 2017-06-29  Maya Rashish  <coypu@sdf.org>
26662         * config/netbsd.h (NETBSD_LIB_SPEC): Add -lc when creating shared
26663         objects.
26665 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26667         * arm/arm-builtins.c: Include profile-count.h
26668         * except.c (sjlj_emit_function_enter): Use
26669         profile_probability::unlikely.
26671 2017-06-29  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
26673         * config/rs6000/rs6000.c (toc_relative_expr_p): Make tocrel_base
26674         and tocrel_offset be pointer args rather than implicitly using
26675         static versions.
26676         (legitimate_constant_pool_address_p, rs6000_emit_move,
26677         const_load_sequence_p, adjust_vperm): Add local tocrel_base and
26678         tocrel_offset and use in toc_relative_expr_p call.
26679         (print_operand, print_operand_address): Use static tocrel_base_oac
26680         and tocrel_offset_oac.
26681         (rs6000_output_addr_const_extra): Use static tocrel_base_oac and
26682         tocrel_offset_oac.
26684 2017-06-29  Maya Rashish  <coypu@sdf.org>
26686         * config/vax/builtins.md (ffssi2_internal): Correct constraint.
26688 2017-06-29  Eric Botcazou  <ebotcazou@adacore.com>
26690         * expr.c (expand_expr) <normal_inner_ref>: When testing for unaligned
26691         objects, take into account only the alignment of 'op0' and 'mode1' if
26692         'op0' is a MEM.
26694 2017-06-29  Steve Ellcey  <sellcey@cavium.com>
26696         * ccmp.c (ccmp_tree_comparison_p): New function.
26697         (ccmp_candidate_p): Update to use above function.
26698         (get_compare_parts): New function.
26699         (expand_ccmp_next): Update to use new functions.
26700         (expand_ccmp_expr_1): Take tree arg instead of gimple, update to use
26701         new functions.
26702         (expand_ccmp_expr): Pass tree instead of gimple to expand_ccmp_expr_1,
26703         take mode as argument.
26704         * ccmp.h (expand_ccmp_expr): Add mode as argument.
26705         * expr.c (expand_expr_real_1): Pass mode as argument.
26707 2017-06-29  Segher Boessenkool  <segher@kernel.crashing.org>
26709         * combine.c (combine_instructions): Print insns to dump_file, together
26710         with their costs.
26712 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26714         * asan.c (asan_emit_stack_protection): Update.
26715         (create_cond_insert_point): Update.
26716         * auto-profile.c (afdo_propagate_circuit): Update.
26717         * basic-block.h (struct edge_def): Turn probability to
26718         profile_probability.
26719         (EDGE_FREQUENCY): Update.
26720         * bb-reorder.c (find_traces_1_round): Update.
26721         (better_edge_p): Update.
26722         (sanitize_hot_paths): Update.
26723         * cfg.c (unchecked_make_edge): Initialize probability to uninitialized.
26724         (make_single_succ_edge): Update.
26725         (check_bb_profile): Update.
26726         (dump_edge_info): Update.
26727         (update_bb_profile_for_threading): Update.
26728         * cfganal.c (connect_infinite_loops_to_exit): Initialize new edge
26729         probabilitycount to 0.
26730         * cfgbuild.c (compute_outgoing_frequencies): Update.
26731         * cfgcleanup.c (try_forward_edges): Update.
26732         (outgoing_edges_match): Update.
26733         (try_crossjump_to_edge): Update.
26734         * cfgexpand.c (expand_gimple_cond): Update make_single_succ_edge.
26735         (expand_gimple_tailcall): Update.
26736         (construct_init_block): Use make_single_succ_edge.
26737         (construct_exit_block): Use make_single_succ_edge.
26738         * cfghooks.c (verify_flow_info): Update.
26739         (redirect_edge_succ_nodup): Update.
26740         (split_edge): Update.
26741         (account_profile_record): Update.
26742         * cfgloopanal.c (single_likely_exit): Update.
26743         * cfgloopmanip.c (scale_loop_profile): Update.
26744         (set_zero_probability): Remove.
26745         (duplicate_loop_to_header_edge): Update.
26746         * cfgloopmanip.h (loop_version): Update prototype.
26747         * cfgrtl.c (try_redirect_by_replacing_jump): Update.
26748         (force_nonfallthru_and_redirect): Update.
26749         (update_br_prob_note): Update.
26750         (rtl_verify_edges): Update.
26751         (purge_dead_edges): Update.
26752         (rtl_lv_add_condition_to_bb): Update.
26753         * cgraph.c: (cgraph_edge::redirect_call_stmt_to_calle): Update.
26754         * cgraphunit.c (init_lowered_empty_function): Update.
26755         (cgraph_node::expand_thunk): Update.
26756         * cilk-common.c: Include profile-count.h
26757         * dojump.c (inv): Remove.
26758         (jumpifnot): Update.
26759         (jumpifnot_1): Update.
26760         (do_jump_1): Update.
26761         (do_jump): Update.
26762         (do_jump_by_parts_greater_rtx): Update.
26763         (do_compare_rtx_and_jump): Update.
26764         * dojump.h (jumpifnot, jumpifnot_1, jumpif_1, jumpif, do_jump,
26765         do_jump_1. do_compare_rtx_and_jump): Update prototype.
26766         * dwarf2cfi.c: Include profile-count.h
26767         * except.c (dw2_build_landing_pads): Use make_single_succ_edge.
26768         (sjlj_emit_dispatch_table): Likewise.
26769         * explow.c: Include profile-count.h
26770         * expmed.c (emit_store_flag_force): Update.
26771         (do_cmp_and_jump): Update.
26772         * expr.c (compare_by_pieces_d::generate): Update.
26773         (compare_by_pieces_d::finish_mode): Update.
26774         (emit_block_move_via_loop): Update.
26775         (store_expr_with_bounds): Update.
26776         (store_constructor): Update.
26777         (expand_expr_real_2): Update.
26778         (expand_expr_real_1): Update.
26779         * expr.h (try_casesi, try_tablejump): Update prototypes.
26780         * gimple-pretty-print.c (dump_probability): Update.
26781         (dump_profile): New.
26782         (dump_gimple_label): Update.
26783         (dump_gimple_bb_header): Update.
26784         * graph.c (draw_cfg_node_succ_edges): Update.
26785         * hsa-gen.c (convert_switch_statements): Update.
26786         * ifcvt.c (cheap_bb_rtx_cost_p): Update.
26787         (find_if_case_1): Update.
26788         (find_if_case_2): Update.
26789         * internal-fn.c (expand_arith_overflow_result_store): Update.
26790         (expand_addsub_overflow): Update.
26791         (expand_neg_overflow): Update.
26792         (expand_mul_overflow): Update.
26793         (expand_vector_ubsan_overflow): Update.
26794         * ipa-cp.c (good_cloning_opportunity_p): Update.
26795         * ipa-split.c (split_function): Use make_single_succ_edge.
26796         * ipa-utils.c (ipa_merge_profiles): Update.
26797         * loop-doloop.c (add_test): Update.
26798         (doloop_modify): Update.
26799         * loop-unroll.c (compare_and_jump_seq): Update.
26800         (unroll_loop_runtime_iterations): Update.
26801         * lra-constraints.c (lra_inheritance): Update.
26802         * lto-streamer-in.c (input_cfg): Update.
26803         * lto-streamer-out.c (output_cfg): Update.
26804         * mcf.c (adjust_cfg_counts): Update.
26805         * modulo-sched.c (sms_schedule): Update.
26806         * omp-expand.c (expand_omp_for_init_counts): Update.
26807         (extract_omp_for_update_vars): Update.
26808         (expand_omp_ordered_sink): Update.
26809         (expand_omp_for_ordered_loops): Update.
26810         (expand_omp_for_generic): Update.
26811         (expand_omp_for_static_nochunk): Update.
26812         (expand_omp_for_static_chunk): Update.
26813         (expand_cilk_for): Update.
26814         (expand_omp_simd): Update.
26815         (expand_omp_taskloop_for_outer): Update.
26816         (expand_omp_taskloop_for_inner): Update.
26817         * omp-simd-clone.c (simd_clone_adjust): Update.
26818         * optabs.c (expand_doubleword_shift): Update.
26819         (expand_abs): Update.
26820         (emit_cmp_and_jump_insn_1): Update.
26821         (expand_compare_and_swap_loop): Update.
26822         * optabs.h (emit_cmp_and_jump_insns): Update prototype.
26823         * predict.c (predictable_edge_p): Update.
26824         (edge_probability_reliable_p): Update.
26825         (set_even_probabilities): Update.
26826         (combine_predictions_for_insn): Update.
26827         (combine_predictions_for_bb): Update.
26828         (propagate_freq): Update.
26829         (estimate_bb_frequencies): Update.
26830         (force_edge_cold): Update.
26831         * profile-count.c (profile_count::dump): Add missing space into dump.
26832         (profile_count::debug): Add newline.
26833         (profile_count::differs_from_p): Explicitly convert to unsigned.
26834         (profile_count::stream_in): Update.
26835         (profile_probability::dump): New member function.
26836         (profile_probability::debug): New member function.
26837         (profile_probability::differs_from_p): New member function.
26838         (profile_probability::differs_lot_from_p): New member function.
26839         (profile_probability::stream_in): New member function.
26840         (profile_probability::stream_out): New member function.
26841         * profile-count.h (profile_count_quality): Rename to ...
26842         (profile_quality): ... this one.
26843         (profile_probability): New.
26844         (profile_count): Update.
26845         * profile.c (compute_branch_probabilities): Update.
26846         * recog.c (peep2_attempt): Update.
26847         * sched-ebb.c (schedule_ebbs): Update.
26848         * sched-rgn.c (find_single_block_region): Update.
26849         (compute_dom_prob_ps): Update.
26850         (schedule_region): Update.
26851         * sel-sched-ir.c (compute_succs_info): Update.
26852         * stmt.c (struct case_node): Update.
26853         (do_jump_if_equal): Update.
26854         (get_outgoing_edge_probs): Update.
26855         (conditional_probability): Update.
26856         (emit_case_dispatch_table): Update.
26857         (expand_case): Update.
26858         (expand_sjlj_dispatch_table): Update.
26859         (emit_case_nodes): Update.
26860         * targhooks.c: Update.
26861         * tracer.c (better_p): Update.
26862         (find_best_successor): Update.
26863         * trans-mem.c (expand_transaction): Update.
26864         * tree-call-cdce.c: Update.
26865         * tree-cfg.c (gimple_split_edge): Upate.
26866         (move_sese_region_to_fn): Upate.
26867         * tree-cfgcleanup.c (cleanup_control_expr_graph): Upate.
26868         * tree-eh.c (lower_resx): Upate.
26869         (cleanup_empty_eh_move_lp): Upate.
26870         * tree-if-conv.c (version_loop_for_if_conversion): Update.
26871         * tree-inline.c (copy_edges_for_bb): Update.
26872         (copy_cfg_body): Update.
26873         * tree-parloops.c (gen_parallel_loop): Update.
26874         * tree-profile.c (gimple_gen_ic_func_profiler): Update.
26875         (gimple_gen_time_profiler): Update.
26876         * tree-ssa-dce.c (remove_dead_stmt): Update.
26877         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
26878         * tree-ssa-loop-im.c (execute_sm_if_changed): Update.
26879         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update.
26880         (unloop_loops): Update.
26881         (try_peel_loop): Update.
26882         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
26883         * tree-ssa-loop-split.c (connect_loops): Update.
26884         (split_loop): Update.
26885         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Update.
26886         (hoist_guard): Update.
26887         * tree-ssa-phionlycprop.c (propagate_rhs_into_lhs): Update.
26888         * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Update.
26889         (value_replacement): Update.
26890         * tree-ssa-reassoc.c (branch_fixup): Update.
26891         * tree-ssa-tail-merge.c (replace_block_by): Update.
26892         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges): Update.
26893         (create_edge_and_update_destination_phis): Update.
26894         (compute_path_counts): Update.
26895         (recompute_probabilities): Update.
26896         (update_joiner_offpath_counts): Update.
26897         (freqs_to_counts_path): Update.
26898         (duplicate_thread_path): Update.
26899         * tree-switch-conversion.c (hoist_edge_and_branch_if_true): Update.
26900         (struct switch_conv_info): Update.
26901         (gen_inbound_check): Update.
26902         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
26903         (vect_do_peeling): Update.
26904         (vect_loop_versioning): Update.
26905         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
26906         (optimize_mask_stores): Update.
26907         * ubsan.c (ubsan_expand_null_ifn): Update.
26908         * value-prof.c (gimple_divmod_fixed_value): Update.
26909         (gimple_divmod_fixed_value_transform): Update.
26910         (gimple_mod_pow2): Update.
26911         (gimple_mod_pow2_value_transform): Update.
26912         (gimple_mod_subtract): Update.
26913         (gimple_mod_subtract_transform): Update.
26914         (gimple_ic): Update.
26915         (gimple_stringop_fixed_value): Update.
26916         (gimple_stringops_transform): Update.
26917         * value-prof.h: Update.
26919 2017-06-29  Carl Love  <cel@us.ibm.com>
26921         * config/rs6000/rs6000-c.c: Add support for built-in functions
26922         vector signed int vec_signed (vector float);
26923         vector signed long long vec_signed (vector double);
26924         vector signed int vec_signed2 (vector double, vector double);
26925         vector signed int vec_signede (vector double);
26926         vector signed int vec_signedo (vector double);
26927         * config/rs6000/rs6000.c (rs6000_generate_vsigned2_code): Add
26928         instruction generator.
26929         * config/rs6000/vsx.md (UNSPEC_VSX_XVCVSPSXWS, UNSPEC_VSX_XVCVSPSXDS,
26930         UNSPEC_VSX_VSIGNED2): Add UNSPECS.
26931         (vsx_xvcvspsxws, vsx_xvcvdpuxds_scale, vsx_xvcvspuxws, vsigned2_v2df):
26932         Add define_insn.
26933         (vsignedo_v2df, vsignede_v2df, vunsigned2_v2df, vunsignedo_v2df,
26934         vunsignede_v2df): Add define_expands.
26935         * config/rs6000/rs6000-builtin.def (VEC_SIGNED, VEC_UNSIGNED,
26936         VEC_SIGNED2, VEC_UNSIGNED2, VEC_SIGNEDE, VEC_UNSIGNEDE, VEC_SIGNEDO,
26937         VEC_UNSIGNEDO): Add definitions.
26938         * config/vsx.md (UNSPEC_VSX_XVCVSPSXWS, UNSPEC_VSX_XVCVSPSXDS,
26939         UNSPEC_VSX_VSIGNED2): Add UNSPECs.
26940         (vsx_xvcvspsxws, vsx_xvcvspuxws): Add define_insn.
26941         (vsigned2_v2df, vsigendo_v2df, vsignede_v2df,
26942         vunsigned2_v2df, vunsignedo_v2df, vunsignede_v2df): Add define_expands.
26943         * config/rs6000/altivec.h (vec_signed, vec_signed2,
26944         vec_signede and vec_signedo, vec_unsigned, vec_unsigned2,
26945         vec_unsignede, vec_unsignedo): Add builtin defines.
26946         * config/rs6000-protos.h (rs6000_generate_vsigned2_code): Add extern
26947         declaration.
26948         * doc/extend.texi: Update the built-in documentation file for the
26949         new built-in functions.
26951 2017-06-29  Richard Biener  <rguenther@suse.de>
26953         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Do not add
26954         reduction chains to LOOP_VINFO_REDUCTIONS.
26955         * tree-vect-slp.c (vect_analyze_slp): Continue looking for
26956         SLP reductions after processing reduction chains.
26958 2017-06-29  Nathan Sidwell  <nathan@acm.org>
26960         * builtins.c (fold_builtin_FUNCTION): Use
26961         lang_hooks.decl_printable_name.
26963 2017-06-29  Peter Bergner  <bergner@vnet.ibm.com>
26965         PR middle-end/81194
26966         * cfgexpand.c (expand_gimple_stmt_1): Handle switch statements
26967         with only one label.
26968         * stmt.c (expand_case): Assert NCASES is greater than one.
26970 2017-06-29  Richard Biener  <rguenther@suse.de>
26972         * tree-cfg.c (group_case_labels_stmt): Return whether we changed
26973         anything.
26974         (group_case_labels): Likewise.
26975         (find_taken_edge): Push sanity checking on val to workers...
26976         (find_taken_edge_cond_expr): ... here
26977         (find_taken_edge_switch_expr): ... and here, handle cases
26978         with just a default label.
26979         * tree-cfg.h (group_case_labels_stmt): Adjust prototype.
26980         (group_case_labels): Likewise.
26981         * tree-cfgcleanup.c (execute_cleanup_cfg_post_optimizing): When
26982         group_case_labels does anything cleanup the CFG again.
26984 2017-06-29  Bin Cheng  <bin.cheng@arm.com>
26986         PR tree-optimization/81196
26987         * tree-ssa-loop-niter.c (number_of_iterations_cond): Handle loop
26988         exit condition comparing two IVs.
26990 2017-06-29  Richard Earnshaw  <rearnsha@arm.com>
26992         * config/arm/parsecpu.awk (gen_comm_data): Add initializer for
26993         profile to the dummy entry at the end of the list of architectures.
26994         * config/arm/arm-cpu-cdata.h: Regenerated.
26996 2017-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26997             Michael Collison <michael.collison@arm.com>
26999         PR target/70119
27000         * config/aarch64/aarch64.md (*aarch64_<optab>_reg_<mode>3_mask1):
27001         New pattern.
27002         (*aarch64_reg_<mode>3_neg_mask2): New pattern.
27003         (*aarch64_reg_<mode>3_minus_mask): New pattern.
27004         (*aarch64_<optab>_reg_di3_mask2): New pattern.
27005         * config/aarch64/aarch64.c (aarch64_rtx_costs): Account for cost
27006         of shift when the shift amount is masked with constant equal to
27007         the size of the mode.
27008         * config/aarch64/predicates.md (subreg_lowpart_operator): New
27009         predicate.
27011 2017-06-29  Martin Liska  <mliska@suse.cz>
27013         * config/i386/i386.opt: Change range from [1,5] to [0,5].
27015 2017-06-29  Yury Gribov  <tetra2005@gmail.com>
27017         PR bootstrap/80565
27018         * ipa-cp.c (allocate_and_init_ipcp_value): Add initialization
27019         code.
27020         * ipa-inline.h
27021         (edge_growth_cache_entry::edge_growth_cache_entry): New
27022         function.
27023         (reset_edge_growth_cache): Update to use constructor.
27025 2017-06-28  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
27027         * config/aarch64/aarch64.h (AARCH64_EXPAND_ALIGNMENT): New.
27028         (DATA_ALIGNMENT): Update to use AARCH64_EXPAND_ALIGNMENT.
27029         (LOCAL_ALIGNMENT): Update to use AARCH64_EXPAND_ALIGNMENT.
27031 2017-06-28  Sebastian Peryt  <sebastian.peryt@intel.com>
27033         * config/i386/avx512vlintrin.h (_mm256_permutexvar_epi64)
27034         (_mm256_permutexvar_epi32, _mm256_permutex_epi64): New intrinsics.
27036 2017-06-28  Szabolcs Nagy  <szabolcs.nagy@arm.com>
27038         * config.gcc (*-linux-musl*): Add t-musl tmake_file.
27039         (*-linux-uclibc*): Add t-uclibc tmake_file.
27040         * config/t-musl: New.
27041         * config/t-uclibc: New.
27043 2017-06-28  Richard Earnshaw  <rearnsha@arm.com>
27045         * config/arm/parsecpu.awk (profile): Parse new keyword in an arch
27046         context.
27047         (gen_comm_data): Emit architectural setting of arch_prof.
27048         * config/arm/arm-cpus.in (armv6-m, armv6s-m, armv7-a, armv7ve): Set the
27049         profile.
27050         (armv7-r, armv7-m, armv7e-m, armv8-a, armv8.1-a, armv8.2-a): Likewise.
27051         (armv8-m.base, armv8-m.main): Likewise.
27052         * arm-protos.h (arm_build_target): Add profile field.
27053         (arch_option): Likewise.
27054         * config/arm/arm.c (arm_configure_build_target): Copy the profile to
27055         the active target.
27056         * config/arm/arm.h (TARGET_ARM_ARCH_PROFILE): Use
27057         arm_active_target.profile.
27059 2017-06-28  Richard Biener  <rguenther@suse.de>
27061         PR middle-end/81227
27062         * fold-const.c (negate_expr_p): Use TYPE_UNSIGNED, not
27063         TYPE_OVERFLOW_WRAPS.
27064         * match.pd (negate_expr_p): Likewise.
27065         * tree-ssa-reassoc.c (optimize_range_tests_diff): Use
27066         fold_build2, not fold_binary.
27068 2017-06-28  Wilco Dijkstra  <wdijkstr@arm.com>
27070         * config/aarch64/aarch64 (aarch64_expand_mov_immediate):
27071         Convert memory address to Pmode.
27072         (aarch64_print_operand): Assert MEM operands are always Pmode.
27074 2017-06-28  Wilco Dijkstra  <wdijkstr@arm.com>
27076         PR target/79665
27077         * config/arm/aarch-common.c (arm_no_early_alu_shift_dep):
27078         Remove redundant if.
27079         (aarch_forward_to_shift_is_not_shifted_reg): Remove.
27080         * config/arm/aarch-common-protos.h
27081         (aarch_forward_to_shift_is_not_shifted_re): Remove.
27082         * config/arm/cortex-a53.md: Use arm_no_early_alu_shift_dep in bypass.
27084 2017-06-28  Michael Meissner  <meissner@linux.vnet.ibm.com>
27086         PR ipa/81238
27087         * multiple_target.c (create_dispatcher_calls): Set the default
27088         clone to be static, not public.
27090 2017-06-28  Richard Biener  <rguenther@suse.de>
27092         * tree-vect-loop.c (vectorizable_reduction): Move special
27093         cond reduction IV var creation ...
27094         (vect_create_epilog_for_reduction): ... here.  Remove induction_index
27095         parameter.  Use STMT_VINFO_VECTYPE.
27096         * tree-vect-slp.c (vect_get_constant_vectors): Properly reset
27097         constant_p.
27099 2017-06-28  Martin Liska  <mliska@suse.cz>
27101         PR ipa/81128
27102         * ipa-visibility.c (non_local_p): Handle visibility.
27104 2017-06-28  Martin Liska  <mliska@suse.cz>
27106         PR driver/79659
27107         * common.opt: Add IntegerRange to various options.
27108         * opt-functions.awk (integer_range_info): New function.
27109         * optc-gen.awk: Add integer_range_info to cl_options struct.
27110         * opts-common.c (decode_cmdline_option): Handle
27111         CL_ERR_INT_RANGE_ARG.
27112         (cmdline_handle_error): Likewise.
27113         * opts.c (print_filtered_help): Show valid interval in
27114         when --help is provided.
27115         * opts.h (struct cl_option): Add range_min and range_max fields.
27116         * config/i386/i386.opt: Add IntegerRange for -mbranch-cost.
27118 2017-06-28  Marc Glisse  <marc.glisse@inria.fr>
27120         * match.pd ((X & ~Y) | (~X & Y)): Generalize to + and ^.
27121         (x * C EQ/NE y * C): New transformation.
27123 2017-06-28  Christophe Lyon  <christophe.lyon@linaro.org>
27125         * genmultilib (combination_space): Accept '+' in option names.
27127 2017-06-28  Martin Liska  <mliska@suse.cz>
27129         PR sanitizer/81224
27130         * asan.c (instrument_derefs): Bail out inner references
27131         that are hard register variables.
27133 2017-06-28  Jakub Jelinek  <jakub@redhat.com>
27135         PR target/81175
27136         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use def_builtin
27137         rather than def_builtin_pure for __builtin_ia32_gatherpf*.
27139 2017-06-28  Richard Biener  <rguenther@suse.de>
27141         * tree-vectorizer.h (vect_get_vec_defs): Remove.
27142         (vect_get_slp_defs): Adjust.
27143         * tree-vect-loop.c (get_initial_defs_for_reduction): Split
27144         out from ...
27145         * tree-vect-slp.c (vect_get_constant_vectors): ... here and
27146         simplify.
27147         * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
27148         get_initial_defs_for_reduction instead of vect_get_vec_defs.
27149         (vectorizable_reduction): Adjust.
27150         * tree-vect-slp.c (vect_get_constant_vectors): Remove reduction
27151         handling.
27152         (vect_get_slp_defs): Likewise.
27153         * tree-vect-stmts.c (vect_get_vec_defs): Make static and adjust.
27154         (vectorizable_bswap): Adjust.
27155         (vectorizable_call): Likewise.
27156         (vectorizable_conversion): Likewise.
27157         (vectorizable_assignment): Likewise.
27158         (vectorizable_shift): Likewise.
27159         (vectorizable_operation): Likewise.
27160         (vectorizable_store): Likewise.
27161         (vectorizable_condition): Likewise.
27162         (vectorizable_comparison): Likewise.
27164 2017-06-28  Michael Collison  <michael.collison@arm.com>
27166         PR target/68535
27167         * config/arm/arm.c (gen_ldm_seq): Remove last unnecessary
27168         set of base_reg
27169         (arm_gen_movmemqi): Removed unused variable 'i'.
27170         Convert 'for' loop into 'while' loop.
27171         (arm_expand_prologue): Remove last unnecessary set of insn.
27172         (thumb_pop): Remove unused variable 'pushed_words'.
27173         (thumb_exit): Remove last unnecessary set of regs_to_pop.
27175 2017-06-28  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
27177         * config/s390/predicates.md: Use s390_rel_address_ok_p.
27178         * config/s390/s390-protos.h: Add prototype of
27179         s390_rel_address_ok_p.
27180         * config/s390/s390.c (s390_got_symbol): New function.
27181         (s390_rel_address_ok_p): New function.
27182         (legitimize_pic_address): Use s390_rel_address_ok_p.
27183         (s390_load_got): Use s390_got_symbol.
27184         (s390_option_override): Issue error if
27185         -mno-pic-data-is-text-relative is used without -fpic/-fPIC.
27186         * config/s390/s390.h (TARGET_DEFAULT_PIC_DATA_IS_TEXT_RELATIVE):
27187         New macro.
27188         * config/s390/s390.opt: New option mpic-data-is-text-relative.
27190 2017-06-27  Andrew Pinski  <apinski@cavium.com>
27192         * match.pd (X >/>=/</<= 0 ? 1.0 : -1.0): New patterns.
27193         (X * copysign (1.0, X)): New pattern.
27194         (X * copysign (1.0, -X)): New pattern.
27195         (copysign (-1.0, CST)): New pattern.
27197 2017-06-27  Joseph Myers  <joseph@codesourcery.com>
27199         * genmultilib (combination_space): Remove variable.
27200         Validate reuse rules against regular expression for any sequence
27201         of multilib options in any order.
27203 2017-06-27  Michael Collison  <michael.collison@arm.com>
27205         * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Directly
27206         call aarch64_split_simd_combine.
27207         * (aarch64_combine_internal<mode>): Delete pattern.
27208         * config/aarch64/aarch64.c (aarch64_split_simd_combine):
27209         Allow register and subreg operands.
27211 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27213         * config/i386/vxworks.h (ASM_SPEC): Remove definition. No target
27214         specific need, just fallback on defaults.
27215         (ASM_OUTPUT_ALIGNED_BSS): Add #undef before #define.
27217 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27218             Olivier Hainque  <hainque@adacore.com>
27220         * config/i386/vxworks.h (DBX_REGISTER_NUMBER): Pick distinct
27221         map for 64bits.
27222         (TARGET_OS_CPP_BUILTINS): builtin_define CPU to X86_64 for 64bit
27223         targets. Pick a default if no particular attempt applied.
27224         (STACK_CHECK_PROTECT): Double for 64bit targets, which have
27225         larger contexts.
27227 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27229         * config.gcc (i*86-wrs-vxworks7): Handle new acceptable triplet.
27230         (x86_64-wrs-vxworks7): Likewise.
27232 2017-06-27  Marek Polacek  <polacek@redhat.com>
27234         PR sanitizer/81223
27235         * ubsan.c (instrument_null): Check get_base_address's result for null.
27237 2017-06-27  Marc Glisse  <marc.glisse@inria.fr>
27239         * match.pd ((A+-B)+(C-A), (A+B)-(A-C)): New transformations.
27241 2017-06-27  Marc Glisse  <marc.glisse@inria.fr>
27243         * builtin-types.def (BT_FENV_T_PTR, BT_CONST_FENV_T_PTR,
27244         BT_FEXCEPT_T_PTR, BT_CONST_FEXCEPT_T_PTR): New primitive types.
27245         (BT_FN_INT_FENV_T_PTR, BT_FN_INT_CONST_FENV_T_PTR,
27246         BT_FN_INT_FEXCEPT_T_PTR_INT, BT_FN_INT_CONST_FEXCEPT_T_PTR_INT):
27247         New function types.
27248         * builtins.def (BUILT_IN_FECLEAREXCEPT, BUILT_IN_FEGETENV,
27249         BUILT_IN_FEGETEXCEPTFLAG, BUILT_IN_FEGETROUND,
27250         BUILT_IN_FEHOLDEXCEPT, BUILT_IN_FERAISEEXCEPT,
27251         BUILT_IN_FESETENV, BUILT_IN_FESETEXCEPTFLAG,
27252         BUILT_IN_FESETROUND, BUILT_IN_FETESTEXCEPT,
27253         BUILT_IN_FEUPDATEENV): New builtins.
27254         * tree-core.h (TI_FENV_T_PTR_TYPE, TI_CONST_FENV_T_PTR_TYPE,
27255         TI_FEXCEPT_T_PTR_TYPE, TI_CONST_FEXCEPT_T_PTR_TYPE): New entries.
27256         * tree.h (fenv_t_ptr_type_node, const_fenv_t_ptr_type_node,
27257         fexcept_t_ptr_type_node, const_fexcept_t_ptr_type_node): New
27258         macros.
27259         (builtin_structptr_types): Adjust size.
27260         * tree.c (builtin_structptr_types): Add four entries.
27262 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27263             Olivier Hainque  <hainque@adacore.com>
27265         * config/vxworks.h (VXWORKS_LIB_SPEC): Incorporate ...
27266         (TLS_SYM): New local macro, forcing reference to __tls__ on
27267         link command lines for VxWorks 7 RTPs, triggering initialization
27268         of tlsLib.
27269         (VXWORKS_HAVE_TLS): New macro. State whether the target VxWorks
27270         OS features TLS support, true for RTPs on VxWorks 7.
27271         * config/vxworks.c (vxworks_override_options): Setup emutls
27272         accordingly.
27274 2017-06-27  Jakub Jelinek  <jakub@redhat.com>
27276         * predict.c (test_prediction_value_range): Use -1U instead of -1
27277         to avoid narrowing conversion warning.
27278         * dumpfile.c (dump_options): Wrap all value into dump_flags_t cast
27279         to avoid narrowing conversion warning.
27280         * opt-functions.awk (var_ref): Return (unsigned short) -1 instead of
27281         -1.
27282         * optc-gen.awk (END): Expect (unsigned short) -1 instead of -1.
27284 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27286         * config/vxworks.h (VXWORKS_LIBS_RTP): Alternative definition for
27287         64bit configurations.
27288         (PTR_DIFF_TYPE): Alternative definition for TARGET_LP64.
27289         (SIZE_TYPE): Likewise.
27290         * config/vxworks.c (vxworks_emutls_var_fields): Use
27291         long_unsigned_type_node instead of unsigned_type_node as the offset
27292         field type, which is "pointer" mode in emutls.c.
27294 2017-06-27  Jakub Jelinek  <jakub@redhat.com>
27296         PR sanitizer/81209
27297         * ubsan.c (ubsan_encode_value): Initialize DECL_CONTEXT on var.
27299         PR middle-end/81207
27300         * gimple-fold.c (replace_call_with_call_and_fold): Handle
27301         gimple_vuse copying separately from gimple_vdef copying.
27303 2017-06-27  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
27305         * value-prof.c (free_hist): Remove call to memset and the enclosing if
27306         condition.
27308 2017-06-26  Jerome Lambourg  <lambourg@adacore.com>
27309             Olivier Hainque  <hainque@adacore.com>
27311         * config.gcc (*-*-vxworks*): Add TARGET_VXWORKS7=1 to tm_defines
27312         for all vxworks7 targets.
27313         * config/vxworks.h (TARGET_VXWORKS7): If not defined, define to 0.
27314         (VXWORKS_ADDITIONAL_CPP_SPEC): Alternative definition for VXWORKS7.
27315         (VXWORKS_LIBS_RTP, VXWORKS_LIBS_RTP_DIR): New macros, allowing
27316         variations for VX6/VX7 and 32/64bits later on in ...
27317         (VXWORKS_LIB_SPEC): Leverage new macros.
27318         (VXWORKS_OS_CPP_BUILTINS): Define _VSB_CONFIG_FILE for VXWORKS7,
27319         as well as _ALLOW_KEYWORD_MACROS when "inline" is not a keyword.
27321 2017-06-26  Jerome Lambourg  <lambourg@adacore.com>
27323         * config/vxworks.h (VXWORKS_OS_CPP_BUILTINS): builtin_define
27324         _VX_TOOL_FAMILY and _VX_TOOL to gnu.
27326 2017-06-26  Carl Love  <cel@us.ibm.com>
27328         * config/rs6000/rs6000-c.c: Add support for built-in functions
27329         vector bool char vec_reve (vector bool char);
27330         vector signed char vec_reve (vector signed char);
27331         vector unsigned char vec_reve (vector unsigned char);
27332         vector bool int vec_reve (vector bool int);
27333         vector signed int vec_reve (vector signed int);
27334         vector unsigned int vec_reve (vector unsigned int);
27335         vector bool long long vec_reve (vector bool long long);
27336         vector signed long long vec_reve (vector signed long long);
27337         vector unsigned long long vec_reve (vector unsigned long long);
27338         vector bool short vec_reve (vector bool short);
27339         vector signed short vec_reve (vector signed short);
27340         vector double vec_reve (vector double);
27341         vector float vec_reve (vector float);
27342         * config/rs6000/rs6000-builtin.def (VREVE_V2DI, VREVE_V4SI,
27343         VREVE_V8HI, VREVE_V16QI, VREVE_V2DF, VREVE_V4SF, VREVE): New builtin.
27344         * config/rs6000/altivec.md (UNSPEC_VREVEV): New UNSPEC.
27345         (altivec_vreve): New pattern.
27346         * config/rs6000/altivec.h (vec_reve): New define.
27347         * doc/extend.texi (vec_rev): Update the built-in documentation file
27348         for the new built-in functions.
27350 2016-06-26  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27352         PR tree-optimization/71815
27353         * gimple-ssa-strength-reduction.c (uses_consumed_by_stmt): New
27354         function.
27355         (find_basis_for_candidate): Call uses_consumed_by_stmt rather than
27356         has_single_use.
27357         (slsr_process_phi): Likewise.
27358         (replace_uncond_cands_and_profitable_phis): Don't replace a
27359         multiply candidate with a stride of 1 (copy or cast).
27360         (phi_incr_cost): Call uses_consumed_by_stmt rather than
27361         has_single_use.
27362         (lowest_cost_path): Likewise.
27363         (total_savings): Likewise.
27365 2017-06-26  Richard Biener  <rguenther@suse.de>
27367         PR target/81175
27368         * config/i386/i386.c (ix86_init_mmx_sse_builtins):
27369         Use def_builtin_pure for all gather builtins.
27371 2017-06-26  Richard Biener  <rguenther@suse.de>
27373         PR tree-optimization/81203
27374         * tree-tailcall.c (find_tail_calls): Do not move stmts into
27375         non-dominating BBs.
27377 2017-06-26  Marek Polacek  <polacek@redhat.com>
27379         PR c/80116
27380         * doc/invoke.texi: Document -Wmultistatement-macros.
27382 2017-06-26  Christophe Lyon  <christophe.lyon@linaro.org>
27384         * doc/sourcebuild.texi (ARM-specific attributes): Document new
27385         arm_neon_ok_no_float_abi effective target.
27387 2017-06-26  Richard Biener  <rguenther@suse.de>
27389         PR tree-optimization/80928
27390         * cfghooks.c (duplicate_block): Do not copy BB_DUPLICATED flag.
27391         (copy_bbs): Set BB_DUPLICATED flag early.
27392         (execute_on_growing_pred): Do not execute for BB_DUPLICATED
27393         marked blocks.
27394         (execute_on_shrinking_pred): Likewise.
27395         * tree-ssa.c (ssa_redirect_edge): Do not look for PHI args in
27396         BB_DUPLICATED blocks.
27397         * tree-ssa-phionlycoprop.c (eliminate_degenerate_phis_1): Properly
27398         iterate over all PHIs considering removal of *gsi.
27400 2017-06-23  Jim Wilson  <jim.wilson@linaro.org>
27402         * doc/invoke.texi (AArch64 Options, -mtune): Re-add falkor and
27403         qdf24xx.
27405 2017-06-23  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
27407         * config/rs6000/rs6000-string.c: (expand_block_clear,
27408         do_load_for_compare, select_block_compare_mode,
27409         compute_current_alignment, expand_block_compare,
27410         expand_strncmp_align_check, expand_strn_compare,
27411         expand_block_move, rs6000_output_load_multiple)
27412         Move functions related to string/block move/compare
27413         to a separate file.
27414         * config/rs6000/rs6000.c: Move above functions to rs6000-string.c.
27415         * config/rs6000/rs6000-protos.h (rs6000_emit_dot_insn): Add prototype
27416         for this function which is now used in two files.
27417         * config/rs6000/t-rs6000: Add rule to compile rs6000-string.o.
27418         * config.gcc: Add rs6000-string.o to extra_objs for
27419         targets powerpc*-*-* and rs6000*-*-*.
27421 2017-06-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
27423         PR target/80510
27424         * config/rs6000/rs6000.md (ALTIVEC_DFORM): Do not allow DImode in
27425         32-bit, since indexed is not valid for DImode.
27426         (mov<mode>_hardfloat32): Reorder ISA 2.07 load/stores before ISA
27427         3.0 d-form load/stores to be the same as mov<mode>_hardfloat64.
27428         (define_peephole2 for Altivec d-form load): Add 32-bit support.
27429         (define_peephole2 for Altivec d-form store): Likewise.
27431         PR ipa/81185
27432         * multiple_target.c (create_dispatcher_calls): Only create the
27433         dispatcher call if the function is the default clone of a
27434         versioned function.
27436 2017-06-23  Segher Boessenkool  <segher@kernel.crashing.org>
27438         PR middle-end/80902
27439         * builtins.c (expand_builtin_atomic_fetch_op): If emitting code after
27440         a call, force the call to not be a tail call.
27442 2017-06-23  Jeff Law  <law@redhat.com>
27444         * doc/contrib.texi: Add entry for Steven Pemberton's work on
27445         enquire.
27447 2017-06-23  Will Schmidt  <will_schmidt@vnet.ibm.com>
27449         * config/rs6000/rs6000.c: Add include of ssa-propagate.h for
27450         update_call_from_tree().  (rs6000_gimple_fold_builtin): Add
27451         handling for early expansion of vector shifts (sl,sr,sra,rl).
27452         (builtin_function_type): Add vector shift right instructions
27453         to the unsigned argument list.
27455 2017-06-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
27457         rtl-optimizatoin/79286
27458         * ira.c (update_equiv_regs): Revert to using may_trap_or_fault_p again.
27459         * rtlanal.c (rtx_addr_can_trap_p_1): SYMBOL_REF_FUNCTION_P can never
27460         trap.  PIC register plus a const unspec without offset can never trap.
27462 2017-06-23  Marc Glisse  <marc.glisse@inria.fr>
27464         * tree.h (builtin_structptr_type): New type.
27465         (builtin_structptr_types): Declare new array.
27466         * tree.c (builtin_structptr_types): New array.
27467         (free_lang_data, build_common_tree_nodes): Use it.
27469 2017-06-23  Jonathan Wakely  <jwakely@redhat.com>
27471         PR c++/81187
27472         * doc/invoke.texi (-Wnoexcept-type): Fix name of option, from
27473         -Wnoexcept.
27475 2017-06-22  Matt Turner  <mattst88@gmail.com>
27477         * config/i386/driver-i386.c (host_detect_local_cpu): Add Kaby
27478         Lake models to skylake case.  Assume skylake for unknown
27479         models with clflushopt.
27481 2017-06-22  Jeff Law  <law@redhat.com>
27483         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Handle
27484         frame sizes that do not satisfy aarch64_uimm12_shift.
27486 2017-06-22  Jan Hubicka <hubicka@ucw.cz>
27488         * profile-count.h (apply_probability,
27489         apply_scale, probability_in): Fix checks for zero.
27491 2017-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27493         * incpath.c (add_sysroot_to_chain): Allow for $SYSROOT prefix.
27494         * doc/cppdiropts.texi (-I @var{dir}): Document it.
27496 2016-06-22  Richard Biener  <rguenther@suse.de>
27498         * tree-vect-loop.c (vect_model_reduction_cost): Handle
27499         COND_REDUCTION and INTEGER_INDUC_COND_REDUCTION without
27500         REDUC_MAX_EXPR support.
27501         (vectorizable_reduction): Likewise.
27502         (vect_create_epilog_for_reduction): Likewise.
27504 2017-06-22  James Greenhalgh  <james.greenhalgh@arm.com>
27506         * match.pd (A / (1 << B) -> A >> B): New.
27507         * generic-match-head.c: Include optabs-tree.h.
27508         * gimple-match-head.c: Likewise.
27509         * optabs-tree.h (target_supports_op_p): New.
27510         * optabs-tree.c (target_supports_op_p): New.
27512 2017-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27514         * configure.ac (gcc_cv_ld_static_dynamic): Also check stderr for
27515         $gcc_cv_ld --help output.
27516         (gcc_cv_ld_demangle): Likewise.
27517         (gcc_cv_ld_eh_frame_hdr): Likewise.
27518         (gcc_cv_ld_pie): Likewise.
27519         (gcc_cv_ld_as_needed): Likewise.  Prefer native forms unless $gnu_ld.
27520         (gcc_cv_ld_buildid): Likewise.
27521         (gcc_cv_ld_sysroot): Likewise.
27522         (ld_bndplt_support): Likewise.
27523         (ld_pushpopstate_support): Likewise.
27524         * configure: Regenerate.
27525         * config/sol2.h [!USE_GLD] (SYSROOT_SPEC): Define.
27527 2017-06-21  Jakub Jelinek  <jakub@redhat.com>
27529         PR target/81151
27530         * config/i386/sse.md (round<mode>2): Renumber match_dup and
27531         operands indexes to avoid gap between operands and match_dups.
27533 2017-06-21  Andrew Pinski  <apinski@cavium.com>
27535         * config/aarch64/aarch64-cost-tables.h (thunderx_extra_costs):
27536         Increment Arith_shift and Arith_shift_reg by 1.
27537         * config/aarch64/aarch64-tuning-flags.def (cheap_shift_extend):
27538         New tuning flag.
27539         * config/aarch64/aarch64.c (thunderx_tunings): Enable
27540         AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
27541         (aarch64_strip_extend): Add new argument and test for it.
27542         (aarch64_cheap_mult_shift_p): New function.
27543         (aarch64_rtx_mult_cost): Call aarch64_cheap_mult_shift_p and don't
27544         add a cost if it is true.
27545         Update calls to aarch64_strip_extend.
27546         (aarch64_rtx_costs): Update calls to aarch64_strip_extend.
27548 2017-06-21  Andrew Pinski  <apinski@cavium.com>
27550         * config/aarch64/aarch64-cores.def (thunderxt88p1): Use thunderxt88
27551         tunings.
27552         (thunderxt88): Likewise.
27553         * config/aarch64/aarch64.c (thunderxt88_prefetch_tune): New variable.
27554         (thunderx_prefetch_tune): New variable.
27555         (thunderx2t99_prefetch_tune): Update for the correct values.
27556         (thunderxt88_tunings): New variable.
27557         (thunderx_tunings): Use thunderx_prefetch_tune instead of
27558         generic_prefetch_tune.
27559         (thunderx2t99_tunings): Use AUTOPREFETCHER_WEAK.
27561 2017-06-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27563         * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>_lse,
27564         SHORT): Relax operand 3 to aarch64_reg_or_zero and constraint to Z.
27565         (aarch64_compare_and_swap<mode>_lse, GPI): Likewise.
27566         (aarch64_atomic_cas<mode>, SHORT): Likewise for operand 2.
27567         (aarch64_atomic_cas<mode>, GPI): Likewise.
27569 2017-06-21  Martin Liska  <mliska@suse.cz>
27571         * gimplify.c (gimplify_label_expr): Insert GIMPLE_PREDICT
27572         statements on cold and hot labels.
27573         * predict.c (tree_estimate_probability_bb): Remove the
27574         prediction from this place.
27576 2017-06-21  Martin Liska  <mliska@suse.cz>
27578         PR tree-optimization/79489
27579         * gimplify.c (maybe_add_early_return_predict_stmt): New
27580         function.
27581         (gimplify_return_expr): Call the function.
27582         * predict.c (tree_estimate_probability_bb): Remove handling
27583         of early return.
27584         * predict.def: Update comment about early return predictor.
27585         * gimple-predict.h (is_gimple_predict): New function.
27586         * predict.def: Change default value of early return to 66.
27587         * tree-tailcall.c (find_tail_calls): Skip GIMPLE_PREDICT
27588         statements.
27589         * passes.def: Put pass_strip_predict_hints to the beginning of
27590         IPA passes.
27592 2017-06-21  Pierre-Marie de Rodat  <derodat@adacore.com>
27594         * dwarf2out.c (gen_decl_die): Remove the guard to skip file-scope
27595         FUNCTION_DECL declarations.
27596         (dwarf2out_early_global_decl): Remove the guard to skip FUNCTION_DECL
27597         declarations.
27598         (dwaf2out_decl): Likewise.
27599         * godump.c (go_early_global_decl): Skip call to the real debug hook
27600         for FUNCTION_DECL declarations.
27601         * passes.c (rest_of_decl_compilation): Skip call to the
27602         early_global_decl debug hook for FUNCTION_DECL declarations, unless
27603         -fdump-go-spec is passed.
27605 2017-06-21  Marc Glisse  <marc.glisse@inria.fr>
27607         * config/i386/i386.c (struct builtin_isa): New field pure_p.
27608         Reorder for compactness.
27609         (def_builtin, def_builtin2, ix86_add_new_builtins): Handle pure_p.
27610         (def_builtin_pure, def_builtin_pure2): New functions.
27611         (ix86_init_mmx_sse_builtins) [__builtin_ia32_stmxcsr]: Mark as pure.
27613 2017-06-21  Marc Glisse  <marc.glisse@inria.fr>
27615         * match.pd (nop_convert): New predicate.
27616         ((A +- CST1) +- CST2): Allow some NOP conversions.
27618 2017-06-21  Jakub Jelinek  <jakub@redhat.com>
27620         PR c++/81130
27621         * gimplify.c (omp_add_variable): Don't force GOVD_SEEN for types
27622         with ctors/dtors if GOVD_SHARED is set.
27624 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27626         * config/aarch64/aarch64.md (movti_aarch64):
27627         Emit mov rather than orr.
27628         (movtf_aarch64): Likewise.
27629         * config/aarch64/aarch64-simd.md (aarch64_simd_mov):
27630         Emit mov rather than orr.
27632 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27634         * config/aarch64/aarch64-simd.md (aarch64_simd_dup):
27635         Swap alternatives, make integer dup more expensive.
27637 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27639         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
27640         Return true for non-tls symbols.
27642 2017-06-21  James Greenhalgh  <james.greenhalgh@arm.com>
27644         * config/aarch64/aarch64-cores.def (cortex-a55): New.
27645         (cortex-a75): Likewise.
27646         (cortex-a75.cortex-a55): Likewise.
27647         * config/aarch64/aarch64-tune.md: Regenerate.
27648         * doc/invoke.texi (-mtune): Document new values for -mtune.
27650 2017-06-21  Tom de Vries  <tom@codesourcery.com>
27652         * doc/sourcebuild.texi (Add Options, Features for dg-add-options): Add
27653         stack_size feature.
27654         (Effective-Target Keywords, Other attributes): Suggest using
27655         dg-add-options stack_size feature to get stack limit in stack_size
27656         effective target documentation.
27658 2017-06-21  Julian Brown  <julian@codesourcery.com>
27659             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
27661         * config/aarch64/aarch64-simd.md (aarch64_crypto_pmulldi)
27662         (aarch64_crypto_pmullv2di): Change type attribute to crypto_pmull.
27663         * config/aarch64/thunderx2t99.md (thunderx2t99_pmull): New
27664         reservation.
27665         * config/arm/cortex-a53.md (cortex_a53_advsimd_type): Add crypto_pmull to
27666         attribute type list for neon_multiply.
27667         * config/arm/cortex-a57.md (cortex_a57_neon_type): Add crypto_pmull to
27668         attribute type list for neon_multiply.
27669         * config/arm/crypto.md (crypto_vmullp64): Change type to crypto_pmull.
27670         * config/arm/exynos-m1.md (exynos_m1_neon_type): Add crypto_pmull to
27671         attribute type list for neon_multiply.
27672         * config/arm/types.md (crypto_pmull): Add.
27673         * config/arm/xgene1.md (xgene1_neon_pmull): Add crypto_pmull to
27674         attribute type list.
27676 2017-06-20  Andreas Tobler  <andreast@gcc.gnu.org>
27678         * config.gcc (armv6*-*-freebsd*): Change the target_cpu_cname to
27679         arm1176jzf-s.
27681 2017-06-20  Jakub Jelinek  <jakub@redhat.com>
27683         * ira-costs.c (find_costs_and_classes): Initialize cost_classes later
27684         to make sure not to dereference a NULL cost_classes_ptr pointer.
27686 2017-06-20  Carl Love  <cel@us.ibm.com>
27688         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
27689         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
27690         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
27691         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
27692         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
27693         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
27694         VMULOSW): New enum "unspec" values.
27695         (vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
27696         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si,
27697         altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
27698         altivec_vmulosw): New patterns.
27699         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
27700         VMULOSW): Add definitions.
27702 2017-06-20  Julia Koval  <julia.koval@intel.com>
27704         * config/i386/i386.c: Fix rounding expand for new pattern.
27705         * config/i386/subst.md: Fix pattern (parallel -> unspec).
27707 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27709         * config/aarch64/aarch64-option-extensions.def (rcpc): New.
27710         * config/aarch64/aarch64.h (AARCH64_FL_RCPC): New.
27712 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27714         * config/aarch64/aarch64-option-extensions.def (fp16): Fix expected
27715         feature string.
27717 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27719         * config/aarch64/aarch64-cores.def: Rearrange to sort by
27720         architecture, then by implementer ID.
27721         * config/aarch64/aarch64-tune.md: Regenerate.
27723 2017-06-20  Richard Biener  <rguenther@suse.de>
27725         PR middle-end/81097
27726         * fold-const.c (split_tree): Fold to type before negating.
27728 2017-06-20  David Malcolm  <dmalcolm@redhat.com>
27730         * diagnostic-show-locus.c
27731         (selftest::test_fixit_deletion_affecting_newline): New function.
27732         (selftest::diagnostic_show_locus_c_tests): Call it.
27734 2017-06-20  Andreas Schwab  <schwab@suse.de>
27736         PR target/80970
27737         * config/m68k/m68k.md (bsetdreg, bchgdreg, bclrdreg): Use "=d"
27738         instead of "+d".
27740 2017-06-20  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
27742         * config/arm/arm-c.c (arm_cpu_builtins): New block to define
27743         __ARM_FEATURE_COPROC according to support.
27745 2017-06-20  Jakub Jelinek  <jakub@redhat.com>
27747         * tree-chkp.c (chkp_get_hard_register_var_fake_base_address):
27748         Rewritten to avoid overflow for > 32-bit pointers.
27750         PR sanitizer/81125
27751         * ubsan.h (ubsan_encode_value): Workaround buggy clang++ parser
27752         by removing enum keyword.
27753         (ubsan_type_descriptor): Likewise.  Formatting fix.
27755         PR target/81121
27756         * config/i386/i386.md (TARGET_USE_VECTOR_CONVERTS float si->{sf,df}
27757         splitter): Require TARGET_SSE2 in the condition.
27759 2017-06-20  Michael Meissner  <meissner@linux.vnet.ibm.com>
27761         PR target/79799
27762         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Add support
27763         for doing vector set of SFmode on ISA 3.0.
27764         * config/rs6000/vsx.md (vsx_set_v4sf_p9): Likewise.
27765         (vsx_set_v4sf_p9_zero): Special case setting 0.0f to a V4SF
27766         element.
27767         (vsx_insert_extract_v4sf_p9): Add an optimization for inserting a
27768         SFmode value into a V4SF variable that was extracted from another
27769         V4SF variable without converting the element to double precision
27770         and back to single precision vector format.
27771         (vsx_insert_extract_v4sf_p9_2): Likewise.
27773 2017-06-19  Jakub Jelinek  <jakub@redhat.com>
27775         * tree-ssa-structalias.c (get_constraint_for_ptr_offset): Multiply
27776         in UWHI to avoid undefined overflow.
27778         PR sanitizer/81125
27779         * ubsan.h (enum ubsan_encode_value_phase): New.
27780         (ubsan_encode_value): Change second argument to
27781         enum ubsan_encode_value_phase with default value of
27782         UBSAN_ENCODE_VALUE_GENERIC.
27783         * ubsan.c (ubsan_encode_value): Change second argument to
27784         enum ubsan_encode_value_phase PHASE from bool IN_EXPAND_P,
27785         adjust uses, for UBSAN_ENCODE_VALUE_GENERIC use just
27786         create_tmp_var_raw instead of create_tmp_var and use a
27787         TARGET_EXPR.
27788         (ubsan_expand_bounds_ifn, ubsan_build_overflow_builtin,
27789         instrument_bool_enum_load, ubsan_instrument_float_cast): Adjust
27790         ubsan_encode_value callers.
27792         PR sanitizer/81111
27793         * ubsan.c (ubsan_encode_value): If current_function_decl is NULL,
27794         use create_tmp_var_raw instead of create_tmp_var, mark it addressable
27795         just by setting TREE_ADDRESSABLE on the result and use a TARGET_EXPR.
27797 2017-06-19  Richard Biener  <rguenther@suse.de>
27799         PR middle-end/81118
27800         * tree-cfgcleanup.c (cleanup_tree_cfg_noloop): Clear niter
27801         estimates if we changed anything.
27803 2017-06-19  Richard Biener  <rguenther@suse.de>
27805         PR tree-optimization/80887
27806         * tree-ssa-sccvn.c (mprts_hook_cnt): New global.
27807         (vn_lookup_simplify_result): Allow only mprts_hook_cnt succesful
27808         simplified lookups, then reset mprts_hook.
27809         (vn_nary_build_or_lookup_1): Set mprts_hook_cnt to 9 before
27810         simplifying.
27811         (try_to_simplify): Likewise.
27813 2017-06-19  Martin Liska  <mliska@suse.cz>
27815         PR sanitizer/80879
27816         * gimplify.c (gimplify_switch_expr):
27817         Initialize live_switch_vars for SWITCH_BODY == STATEMENT_LIST.
27819 2017-06-19  Martin Liska  <mliska@suse.cz>
27821         * doc/install.texi: Document that PGO runs in 4 stages.
27823 2017-06-19  Martin Liska  <mliska@suse.cz>
27825         PR ipa/80732
27826         * attribs.c (make_dispatcher_decl): Do not append '.ifunc'
27827         to dispatcher function name.
27828         * multiple_target.c (replace_function_decl): New function.
27829         (create_dispatcher_calls): Redirect both edges and references.
27831 2017-06-19  Jan Hubicka <hubicka@ucw.cz>
27833         * profile-count.c (profile_count::dump): Dump quality.
27834         (profile_count::differs_from_p): Update for unsigned val.
27835         * profile-count.h (profile_count_quality): New enum.
27836         (profile_count): Turn m_val to 62bit unsigned, add quality tracking.
27838 2017-06-19  Richard Biener  <rguenther@suse.de>
27840         * tree-ssa-loop-niter.h (estimate_numbers_of_iterations): Take
27841         struct function as arg.
27842         (estimate_numbers_of_iterations): Export overload with loop arg.
27843         (free_numbers_of_iterations_estimates_loop): Use an overload of
27844         free_numbers_of_iterations_estimates instead.
27845         * tree-cfg.c (remove_bb): Adjust.
27846         * tree-cfgcleanup.c (remove_forwarder_block_with_phi): Likewise.
27847         * tree-parloops.c (gen_parallel_loop): Likewise.
27848         * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables):
27849         Likewise.
27850         (tree_unroll_loops_completely): Likewise.
27851         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations_loop):
27852         Use an overload instead and export.
27853         (estimated_loop_iterations): Adjust.
27854         (max_loop_iterations): Likewise.
27855         (likely_max_loop_iterations): Likewise.
27856         (estimate_numbers_of_iterations): Take struct function as arg
27857         and adjust.
27858         (loop_exits_before_overflow): Adjust.
27859         (free_numbers_of_iterations_estimates_loop): Use an overload.
27860         * tree-vect-loop.c (vect_analyze_loop_form): Adjust.
27861         * tree-vectorizer.c (vect_free_loop_info_assumptions): Likewise.
27863 2017-06-19  Richard Biener  <rguenther@suse.de>
27865         PR ipa/81112
27866         * ipa-prop.c (find_constructor_constant_at_offset): Handle
27867         RANGE_EXPR conservatively.
27869 2017-06-16  Carl Love  <cel@us.ibm.com>
27871         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
27872         definitions for vec_float, vec_float2, vec_floato,
27873         vec_floate built-ins.
27874         * config/rs6000/vsx.md (define_c_enum "unspec"): Add RTL code
27875         for instructions vsx_xvcvsxws vsx_xvcvuxwsp, float2, floato and
27876         floate.
27877         * config/rs6000/rs6000-builtin.def (FLOAT2_V2DI, FLOATE_V2DF,
27878         FLOATE_2DI, FLOATO_V2DF, FLOATEE_V2DI, XVCVSXWSP_V4SF,
27879         UNS_FLOATO_V2DI, UNS_FLOATE_V2DI): Add definitions.
27880         * config/altivec.md (define_insn "p8_vmrgew_<mode>",
27881         define_mode_attr VF_sxddp): Add V4SF type to p8_vmrgew.
27882         * config/rs6000/altivec.h (vec_float, vec_float2, vec_floate,
27883         vec_floato): Add builtin defines.
27884         * doc/extend.texi (vec_float, vec_float2, vec_floate, vec_floato):
27885         Update the built-in documentation file for the new built-in
27886         functions.
27888 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27890         * config/arm/arm.opt (marm): Mark as the negative of of -mthumb.
27891         (mthumb): Mark as the negative of -marm.
27893 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27895         * doc/invoke.texi (ARM Options, -mcpu): Document supported
27896         extension options.
27897         (ARM Options, -mtune): Document that this accepts the same
27898         extension options as -mcpu.
27899         (ARM Options, -mfpu): Document addition of -mfpu=auto.
27901 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27903         * doc/invoke.texi (ARM Options, -march=): Document new syntax and
27904         permitted extensions.
27906 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27908         * config/arm/arm-cpus.in (armv7): Add extension +nofp.
27909         (armv7-r): Add aliases vfpv3xd and vfpv3-d16.
27910         (armv8-m.main): Add option +nodsp.
27911         * config/arm/arm-cpu-cdata.h: Regenerated.
27913 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27915         * config/arm/t-fuchsia: New file.
27916         * config.gcc (arm*-*-fuchsia*): Use it.
27918 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27920         * config/arm/t-symbian: Rewrite for new option infrastructure.
27922 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27924         * config/arm/t-phoenix (MULTILIB_REUSE): Clear variable.
27925         (MULTILIB_REQUIRED): Likewise.
27927 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27929         * config/arm/t-linux-eabi (MULTILIB_EXCEPTIONS): Set to empty.
27930         (MULTILIB_RESUE): Likewise.
27931         (MULTILIB_MATCHES): Likewise.
27932         (MULTLIB_REQUIRED): Likewise.
27934 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27936         * config/arm/t-rtems: Rewrite for new option framework.
27938 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27940         * config/arm/t-aprofile (v7_a_nosimd_variants, v7_a_simd_variants)
27941         (v7ve_nosimd_variatns, v7ve_vfpv3_simd_variants)
27942         (v7ve_vfpv4_simd_variants, v8_a_nosimd_variants, v8_a_simd_variants)
27943         (v8_1_a_simd_variants, v8_2_a_simd_variants): Move to ...
27944         * config/arm/t-multilib: ... here.
27945         (MULTILIB_OPTIONS): Add armv7 and armv7+fp architectures.
27946         (MULTILIB_MATCHES): Use armv7 libraries for armv7-r.  Also use for
27947         armv7-a and armv8*-a when A-profile libraries have not been built.
27948         * config/arm/t-rmprofile: Rewrite.
27950 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27952         * genmultilib (multilib_reuse): Allow an explicit period to be escaped
27953         with a backslash.  Remove the backslash after substituting unescaped
27954         periods.
27955         * doc/fragments.texi (MULTILIB_REUSE): Document it.
27957 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27959         * config.gcc: (arm*-*-*): When building a-profile libraries, force
27960         the driver to pass through the default setting of -mfloat-abi.
27961         * common/config/arm/arm-common.c (arm_target_thumb_only): Return -marm
27962         rather than NULL.
27963         * config/arm/t-multilib (MULTILIB_REUSE): Initialize to empty.
27964         (all_feat_combs): New rule.
27965         (MULTILIB_OPTIONS): Use explicit ARM and Thumb directories.  Rework
27966         default libraries.
27967         * config/arm/t-aprofile: Rewrite.
27969 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27971         * config/arm/arm.h (FPUTYPE_AUTO): Define.
27972         * config/arm/arm.c (arm_option_override): Use FPUTYPE_AUTO if the
27973         fpu is not specified by the user/command-line.
27974         * config/arm/bpabi.h (FPUTYPE_DEFAULT): Delete.
27975         * config/arm/netbsd-elf.h (FPUTYPE_DEFAULT): Delete.
27976         * config/arm/linux-elf.h (FPUTYPE_DEFAULT): Delete.
27977         * config/arm/vxworks.h (FPUTYPE_DEFAULT): Delete.
27978         * common/config/arm/arm-common.c (arm_canon_arch_option): Use
27979         FPUTYPE_AUTO insted of FPUTYPE_DEFAULT.
27981 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27983         * config/arm/elf.h (MULTILIB_DEFAULTS): Delete.
27984         * config/arm/t-arm-elf: Rewritten.
27986 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27988         * config/arm/arm.h (TARGET_HARD_FLOAT): Also check that we
27989         have some floating-point instructions.
27990         (TARGET_SOFT_FLOAT): Define as inverse of TARGET_HARD_FLOAT.
27991         (TARGET_MAYBE_HARD_FLOAT): New macro.
27992         * config/arm/arm-builtins.c (arm_init_builtins): Use
27993         TARGET_MAYBE_HARD_FLOAT.
27994         * config/arm/arm.c (arm_option_override): Use TARGET_HARD_FLOAT_ABI.
27996 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27998         * common/config/arm/arm-common.c: Define INCLUDE_LIST.
27999         (configargs.h): Include it.
28000         (arm_print_hint_for_fpu_option): New function.
28001         (arm_parse_fpu_option): New function.
28002         (candidate_extension): New class.
28003         (arm_canon_for_multilib): New function.
28004         * config/arm/arm.h (CANON_ARCH_SPEC_FUNCTION): New macro.
28005         (EXTRA_SPEC_FUNCTIONS): Add CANON_ARCH_SPEC_FUNCTION.
28006         (ARCH_CANONICAL_SPECS): New macro.
28007         (DRIVER_SELF_SPECS): Add ARCH_CANONICAL_SPECS.
28009 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28011         * config.gcc (arm*-*-*): Ensure both target_cpu_cname and with_cpu
28012         are set after handling multilib fragments.  Set target_cpu_default2
28013         from with_cpu.
28015 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28017         * config.gcc (arm*-*-fucshia*): Set target_cpu_cname to the real
28018         cpu name.
28019         (arm*-*-*): Set target_cpu_default2 to a quoted string.
28020         * config/arm/parsecpu.awk (check_cpu): Validate any extension
28021         options.
28022         (check_arch): Likewise.
28023         * config/arm/arm.c (arm_configure_build_target): Handle
28024         TARGET_CPU_DEFAULT being a string constant.  Scan any feature
28025         options in the default.
28027 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28029         * config/arm/arm-protos.h (cpu_arch_extension): Add field to record
28030         when an option is an alias of another.
28031         * config/arm/parsecpu.awk (optalias): New parser token.
28032         (gen_comm_data): Mark non-alias options as such.  Emit entries
28033         for extension aliases.
28034         * config/arm/arm-cpus.in (armv5e): Make vfpv2 an alias.
28035         (armv5te, armv5tej, armv6, armv6j, armv6k, armv6z): Likewise.
28036         (armv6kz, armv6zk, armv6t2): Likewise.
28037         (armv7): Make vfpv3-d16 an alias.
28038         (armv7-a): Make vfpv3-d16, neon and neon-vfpv3 aliases.  Sort in
28039         canonical order.
28040         (armv7ve): Make vfpv4-d16, neon-vfpv3 and neon-vfpv4 aliases.
28041         Sort in canonical order.
28042         (armv8-a): Sort in canonical order.
28043         (armv8.1-a, armv8.2-a):  Likewise.
28044         (generic-armv7-a): Make neon and neon-vfpv3 aliases.  Sort in
28045         canonical order.
28046         (cortex-a9): Sort in canonical order.
28047         * config/arm/arm.c (selftests.h): Include it.
28048         (arm_test_cpu_arch_data): New function.
28049         (arm_run_self_tests): New function.
28050         (TARGET_RUN_TARGET_SELFTESTS): Redefine.
28051         (targetm): Move declaration to the end of the file.
28052         * arm-cpu-cdata.h: Regenerated.
28054 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28056         * config/arm/arm.h (TARGET_MODE_SPECS): Add additional parameter to
28057         call to target_mode_check describing the type of option passed.
28058         * common/config/arm/arm-common.c (arm_arch_core_flag): Delete.
28059         (arm_target_thumb_only): Use arm_parse_arch_option_name or
28060         arm_parse_cpu_option_name to match parameters against list of
28061         available targets.
28062         * config/arm/parsecpu.awk (gen_comm_data): Don't generate
28063         arm_arch_core_flags data structure.
28064         * config/arm/arm-cpu_cdata.h: Regenerated.
28066 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28068         * common/config/arm/arm-common.c (arm_initialize_isa): Moved here from
28069         config/arm/arm.c.
28070         (arm_print_hint_for_cpu_option): Likewise.
28071         (arm_print_hint_for_arch_option): Likewise.
28072         (arm_parse_cpu_option_name): Likewise.
28073         (arm_parse_arch_option_name): Likewise.
28074         * config/arm/arm.c (arm_identify_fpu_from_isa): Use the computed number
28075         of entries in the all_fpus list.
28076         * config/arm/arm-protos.h (all_architectures, all_cores): Declare.
28077         (arm_parse_cpu_option_name): Declare.
28078         (arm_parse_arch_option_name): Declare.
28079         (arm_parse_option_features): Declare.
28080         (arm_intialize_isa): Declare.
28081         * config/arm/parsecpu.awk (gen_data): Move CPU and architecture
28082         data tables to ...
28083         (gen_comm_data): ... here.  Make definitions non-static.
28084         * config/arm/arm-cpu-data.h: Regenerated.
28085         * config/arm/arm-cpu-cdata.h: Regenerated.
28087 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28089         * config/arm/arm-protos.h (arm_build_target): Remove arch_core.
28090         (cpu_arch_extension): New structure.
28091         (cpu_arch_option, arch_option, cpu_option): New structures.
28092         * config/arm/parsecpu.awk (gen_headers): Build an enumeration of
28093         architecture types.
28094         (gen_data): Generate new format data tables.
28095         * config/arm/arm.c (cpu_tune): New structure.
28096         (cpu_option, processors): Delete.
28097         (arm_print_hint_for_core_or_arch): Delete.  Replace with ...
28098         (arm_print_hint_for_cpu_option): ... this and ...
28099         (arm_print_hint_for_arch_option): ... this.
28100         (arm_parse_arch_cpu_name): Delete.  Replace with ...
28101         (arm_parse_cpu_option_name): ... this and ...
28102         (arm_parse_arch_option_name): ... this.
28103         (arm_unrecognized_feature): Change type of target parameter to
28104         cpu_arch_option.
28105         (arm_parse_arch_cpu_features): Delete.  Replace with ...
28106         (arm_parse_option_features): ... this.
28107         (arm_configure_build_target): Rework to use new configuration data
28108         tables.
28109         (arm_print_tune_info): Rework for new configuration data tables.
28110         * config/arm/arm-cpu-data.h: Regenerated.
28111         * config/arm/arm-cpu.h: Regenerated.
28113 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28115         * Makefile.in (OBJS): Move sbitmap.o from here ...
28116         (OBJS-libcommon): ... to here.
28118 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28120         * config/arm/arm-isa.h (ISA_ALL_FPU_INTERNAL): Renamed from ISA_ALL_FPU.
28121         (ISA_ALL_CRYPTO): New macro.
28122         (ISA_ALL_SIMD): New macro
28123         (ISA_ALL_FP): New macro.
28124         * config/arm/arm.c (fpu_bitlist): Update initializer.
28125         * config/arm/arm-cpus.in: Use new ISA_ALL macros to disable crypto,
28126         simd or fp.
28127         (arm9e): Add fpu.  Add option for nofp
28128         (arm946e-s, arm966e-s, arm968e-s, arm10e, arm1020e, arm1022e): Likewise.
28129         (arm926ej-s, arm1026ej-s): Likewise.
28130         (generic-armv7-a): Add fpu.  Add options for simd, vfpv3, vfpv3-d16,
28131         vfpv3-fp16, vfpv3-d16-fp16, vfpv4, vfpv4-d16, neon, neon-vfp3,
28132         neon-fp16, neon-vfpv4, nofp and nosimd.
28133         (cortex-a5, cortex-a7): Add fpu.  Add options for nosimd and nofp.
28134         (cortex-a8): Add fpu.  Add option for nofp.
28135         (cortex-a9): Add fpu.  Add options for nosimd and nofp.
28136         (cortex-a12, cortex-a15, cortex-a17): Add fpu.  Add option for nofp.
28137         (cortex-r4f): Add fpu.
28138         (cortex-r5): Add fpu.  Add options for nofp.dp and nofp.
28139         (cortex-r7): Use idiv option from architecture.  Add fpu.  Add option
28140         for nofp.
28141         (cortex-r8): Likewise.
28142         (cortex-m4): Add fpu.  Add option for nofp.
28143         (cortex-a15.cortex-a7): Add fpu.  Add option for nofp.
28144         (cortex-a17.cortex-a7): Likewise.
28145         (cortex-a32): Add fpu.  Add options for crypto and nofp.
28146         (cortex-a35, cortex-a53): Likewise.
28147         (cortex-a57): Add fpu.  Add option for crypto.
28148         (cortex-a72, cortex-a73): Likewise.
28149         (exynos-m1): Likewise.
28150         (cortex-a57.cortex-a53, cortex-a72.cortex-a53): Likewise.
28151         (cortex-a73.cortex-a35, cortex-a73.cortex-a53): Likewise.
28152         (cortex-m33): Add fpu.  Add option for nofp.
28153         * config/arm/arm-cpu-cdata.h: Regenerated
28154         * config/arm/arm-cpu-data.h: Regenerated.
28156 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28158         * arm-cpus.in (armv5e): Add options fp, vfpv2 and nofp.
28159         (armv5te, armv5tej): Likewise.
28160         (armv6, armv6j, armv6k, armv6z, armv6kz, armv6zk, armv6t2): Likewise.
28161         (armv7): Add options fp and vfpv3-d16.
28162         (armv7-a): Add options fp, simd, vfpv3, vfpv3-d16, vfpv3-d16-fp16,
28163         vfpv3-fp16, vfpv4, vfpv4-d16, neon, neon-vfpv3, neon-fp16, neon-vfpv4,
28164         nofp and nosimd.
28165         (armv7ve): Likewise.
28166         (armv7-r): Add options fp, fp.sp, idiv, nofp and noidiv.
28167         (armv7e-m): Add options fp, fpv5, fp.dp and nofp.
28168         (armv8-a): Add nocrypto option.
28169         (armv8.1-a, armv8.2-a): Likewise.
28170         (armv8-m.main): add options fp, fp.dp and nofp.
28172 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28174         * config/arm/arm-cpus.in (armv8-a): Add options crc, simd crypto and
28175         nofp.
28176         (armv8-a+crc): Delete.
28177         (armv8.1-a): Add options simd, crypto and nofp.
28178         (armv8.2-a): Add options fp16, simd, crypto and nofp.
28179         (armv8.2-a+fp16): Delete.
28180         (armv8-m.main): Add option dsp.
28181         (armv8-m.main+dsp): Delete.
28182         (cortex-a8): Add fpu.  Add nofp option.
28183         (cortex-a9): Add fpu.  Add nofp and nosimd options.
28184         * config/arm/parsecpu.awk (gen_data): Generate option tables and
28185         link to main cpu and architecture data structures.
28186         (gen_comm_data): Only put isa attributes from the main architecture
28187         in common tables.
28188         (option): New statement for architecture and CPU entries.
28189         * arm.c (struct cpu_option): New structure.
28190         (struct processors): Add entry for options.
28191         (arm_unrecognized_feature): New function.
28192         (arm_parse_arch_cpu_name): Ignore any characters after the first
28193         '+' character.
28194         (arm_parse_arch_cpu_feature): New function.
28195         (arm_configure_build_target): Separate out any CPU and architecture
28196         features and parse separately.  Don't error out if -mfpu=auto is
28197         used with only an architecture string.
28198         (arm_print_asm_arch_directives): New function.
28199         (arm_file_start): Call it.
28200         * config/arm/arm-cpu-cdata.h: Regenerated.
28201         * config/arm/arm-cpu-data.h: Likewise.
28202         * config/arm/arm-tables.opt: Likewise.
28204 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28206         * config/arm/elf.h (ASM_SPEC): Only pass -mfpu through to the
28207         assembler when it is not -mfpu=auto.
28209 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28211         * config/arm/arm.h (BIG_LITTLE_SPEC): Delete macro.
28212         (ASM_REWRITE_SPEC_FUNCTIONS): New macro.
28213         (BIG_LITTLE_CPU_SPEC_FUNCTIONS): Delete macro.
28214         (ASM_CPU_SPEC): Rewrite.
28215         (MCPU_MTUNE_NATIVE_FUNCTIONS): New macro.
28216         (EXTRA_SPEC_FUNCTIONS): Move outside of ifdef.  Use
28217         MCPU_MTUNE_NATIVE_FUNCTIONS and ASM_REWRITE_SPEC_FUNCTIONS.  Remove
28218         reference to BIG_LITTLE_CPU_SPEC_FUNCTIONS.
28219         * common/config/arm/arm-common.c (arm_rewrite_selected_cpu): Ensure
28220         copied string is NUL-terminated.  Also strip any characters prefixed
28221         by '+'.
28222         (arm_rewrite_selected_arch): New function.
28223         (arm_rewrite_march): New function.
28225 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28227         * config/arm/arm.opt (x_arm_arch_string): New TargetSave option.
28228         (x_arm_cpu_string, x_arm_tune_string): Likewise.
28229         (march, mcpu, mtune): Convert to string-based options.
28230         * config/arm/arm.c (arm_print_hint_for_core_or_arch): New function.
28231         (arm_parse_arch_cpu_name): New function.
28232         (arm_configure_build_target): Use arm_parse_arch_cpu_name to
28233         identify selected architecture or CPU.
28234         (arm_option_save): New function.
28235         (TARGET_OPTION_SAVE): Redefine.
28236         (arm_option_restore): Restore string options.
28237         (arm_option_print): Print string options.
28239 2017-06-16  Martin Sebor  <msebor@redhat.com>
28241         PR tree-optimization/80933
28242         PR tree-optimization/80934
28243         * builtins.c (fold_builtin_3): Do not handle bcmp here.
28244         * gimple-fold.c (gimple_fold_builtin_bcmp): New function.
28245         (gimple_fold_builtin_bcopy, gimple_fold_builtin_bzero): Likewise.
28246         (gimple_fold_builtin): Call them.
28248 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28250         * gimple-ssa-isolate-paths.c (isolate_path): Set edge leading to path
28251         as unlikely; update profile.
28253 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28255         * predict.c (force_edge_cold): Handle declaring edges impossible
28256         more aggresively.
28258 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28260         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update
28261         profile.
28262         (try_unroll_loop_completely): Fix reporting.
28264 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28266         * tree-ssa-tail-merge.c (replace_block_by): Fix profile updating.
28268 2017-06-16  James Greenhalgh  <james.greenhalgh@arm.com>
28270         PR target/71778
28271         * config/arm/arm-builtins.c (arm_expand_builtin_args): Return TARGET
28272         if given a non-constant argument for an intrinsic which requires a
28273         constant.
28275 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28277         * profile.c (compare_freqs): New function.
28278         (branch_prob): Sort edge list.
28279         (find_spanning_tree): Assume that the list is priority sorted.
28281 2017-06-16  Richard Biener  <rguenther@suse.de>
28283         PR tree-optimization/81090
28284         * passes.def (pass_record_bounds): Remove.
28285         * tree-pass.h (make_pass_record_bounds): Likewise.
28286         * tree-ssa-loop.c (pass_data_record_bounds, pass_record_bounds,
28287         make_pass_record_bounds): Likewise.
28288         * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables): Do
28289         not free niter estimates at the beginning but at the end.
28290         * tree-scalar-evolution.c (scev_finalize): Free niter estimates.
28292 2017-06-16  Richard Biener  <rguenther@suse.de>
28294         * tree-switch-conversion.c (emit_case_bit_tests): Adjust
28295         initializer to workaround ICE in host GCC 4.8.
28297 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28299         * ipa-inline-transform.c (update_noncloned_frequencies): Update also
28300         counts.
28301         (clone_inlined_nodes): Update.
28303 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28305         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune): Update
28306         prefetch settings, and enable prefetching by default at -O3.
28308 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28310         * config/aarch64/aarch64.c (aarch64_override_options_internal):
28311         Set flag_prefetch_loop_arrays according to tuning data.
28313 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28315         * config/aarch64/aarch64-protos.h (struct cpu_prefetch_tune):
28316         New tune structure.
28317         (struct tune_params): Use cpu_prefetch_tune instead of cache_line_size.
28318         [Unrelated to main purpose of the patch] Place the pointer field last
28319         to enable type checking errors when tune structure are wrongly merged.
28320         * config/aarch64/aarch64.c (generic_prefetch_tune,)
28321         (exynosm1_prefetch_tune, qdf24xx_prefetch_tune,)
28322         (thunderx2t99_prefetch_tune): New tune constants.
28323         (tune_params *_tunings): Update all tunings (no functional change).
28324         (aarch64_override_options_internal): Set PARAM_SIMULTANEOUS_PREFETCHES,
28325         PARAM_L1_CACHE_SIZE, PARAM_L1_CACHE_LINE_SIZE, and PARAM_L2_CACHE_SIZE
28326         from tunings structures.
28328 2017-06-16  Jakub Jelinek  <jakub@redhat.com>
28330         PR sanitizer/81094
28331         * ubsan.c (instrument_null): Add T argument, use it instead
28332         of computing it based on IS_LHS.
28333         (instrument_object_size): Likewise.
28334         (pass_ubsan::execute): Adjust instrument_null and
28335         instrument_object_size callers to pass gimple_get_lhs or
28336         gimple_assign_rhs1 result to it.  Use instrument_null instead of
28337         calling get_base_address and instrument_mem_ref.  Handle
28338         aggregate call arguments for object-size sanitization.
28340 2017-06-16  Yury Gribov  <tetra2005@gmail.com>
28342         PR tree-optimization/81089
28343         * tree-vrp.c (is_masked_range_test): Validate operands of
28344         subexpression.
28346 2017-06-15  Martin Sebor  <msebor@redhat.com>
28348         PR c++/80560
28349         * dumpfile.c (dump_register): Avoid calling memset to initialize
28350         a class with a default ctor.
28351         * gcc.c (struct compiler): Remove const qualification.
28352         * genattrtab.c (gen_insn_reserv): Replace memset with initialization.
28353         * hash-table.h: Ditto.
28354         * ipa-cp.c (allocate_and_init_ipcp_value): Replace memset with
28355           assignment.
28356         * ipa-prop.c (ipa_free_edge_args_substructures): Ditto.
28357         * omp-low.c (lower_omp_ordered_clauses): Replace memset with
28358         default ctor.
28359         * params.h (struct param_info): Make struct members non-const.
28360         * tree-switch-conversion.c (emit_case_bit_tests): Replace memset
28361         with default initialization.
28362         * vec.h (vec_copy_construct, vec_default_construct): New helper
28363         functions.
28364         (vec<T>::copy, vec<T>::splice, vec<T>::reserve): Replace memcpy
28365         with vec_copy_construct.
28366         (vect<T>::quick_grow_cleared): Replace memset with default ctor.
28367         (vect<T>::vec_safe_grow_cleared, vec_safe_grow_cleared): Same.
28368         * doc/invoke.texi (-Wclass-memaccess): Document.
28370 2017-06-15  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
28372         * emit-rtl.h (is_leaf): Update comment about local
28373         register allocator.
28375 2017-06-15  Jozef Lawrynowicz   <jozef.l@somniumtech.com>
28377         PR target/78818
28378         * config/msp430/msp430.c (msp430_data_attr): Check that it's possible
28379         for a variable to have a section before checking if the section has a
28380         name.
28381         Set section to.persistent if persistent attribute is set.
28382         Warn if .persistent attribute is used on an automatic variable.
28384 2017-06-15  Eric Botcazou  <ebotcazou@adacore.com>
28386         PR rtl-optimization/80474
28387         * reorg.c (update_block): Do not ignore instructions in a delay slot.
28389 2017-06-15  Segher Boessenkool  <segher@kernel.crashing.org>
28391         * config/rs6000/rs6000.md (add<mode>3): Use reg_or_subregno instead
28392         of REGNO.
28394 2017-06-14  Maciej W. Rozycki  <macro@imgtec.com>
28396         * config/mips/mips.md (MIPS16_T_REGNUM): Remove constant.
28397         (casesi): Emit bounds checking as RTL.
28398         (casesi_internal_mips16_<mode>): Remove bounds checking.
28400 2017-06-14  Max Filippov  <jcmvbkbc@gmail.com>
28402         * config/xtensa/xtensa.c (xtensa_option_override): Append
28403         MASK_CONST16 to target_flags in the absence of TARGET_L32R.
28404         (hwloop_optimize, hwloop_fail, hwloop_pattern_reg,
28405          xtensa_doloop_hooks): Define unconditionally.
28406         (xtensa_reorg_loops): Only call reorg_loops in the presence of
28407         TARGET_LOOPS.
28408         * config/xtensa/xtensa.h (TARGET_L32R): New definition.
28409         (TARGET_DEFAULT): Remove XCHAL_HAVE_L32R condition and account
28410         for it in xtensa_option_override.
28411         (HARD_FRAME_POINTER_IS_FRAME_POINTER,
28412          HARD_FRAME_POINTER_IS_ARG_POINTER): New definitions.
28414 2017-06-14  Boris Kolpackov  <boris@codesynthesis.com>
28416         * doc/cppopts.texi: Document '-' special value to -MF.
28418 2017-06-14  Wilco Dijkstra  <wdijkstr@arm.com>
28420         * config/arm/cortex-a53.md (cortex_a53_fpalu) Adjust latency.
28421         (cortex_a53_fconst): Likewise.
28422         (cortex_a53_fpmul): Likewise.
28423         (cortex_a53_f_load_64): Likewise.
28424         (cortex_a53_f_load_many): Likewise.
28425         (cortex_a53_advsimd_alu): Likewise.
28426         (cortex_a53_advsimd_alu_q): Likewise.
28427         (cortex_a53_advsimd_mul): Likewise.
28428         (cortex_a53_advsimd_mul_q): Likewise.
28429         (fpmac bypass): Add new bypass for fpmac-fpmac case.
28430         Add missing fmul, r2f_cvt and fconst cases.
28432 2017-06-14  Richard Biener  <rguenther@suse.de>
28434         PR middle-end/81088
28435         * fold-const.c (split_tree): Drop TREE_OVERFLOW flag from
28436         literal constants.
28437         (fold_binary_loc): When associating do not treat pre-existing
28438         TREE_OVERFLOW on literal constants as a reason to allow
28439         TREE_OVERFLOW on associated literal constants.
28441 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28443         * config/sparc/sparc.h (MASK_ISA): Add MASK_LEON and MASK_LEON3.
28444         (MASK_FEATURES): New macro.
28445         * config/sparc/sparc.c (sparc_option_override): Remove the special
28446         handling of -mfpu and generalize it to all MASK_FEATURES switches.
28448 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28450         * simplify-rtx.c (simplify_binary_operation_1) <UDIV>: Do not simplify
28451         a division of 0 if non-call exceptions are enabled.
28453 2017-06-14  Andrew Pinski  <apinski@cavium.com>
28454             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
28456         PR target/71663
28457         * config/aarch64/aarch64.c (aarch64_expand_vector_init):
28458         Improve vector initialization code gen for only variable case.
28460 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28462         * config/sparc/driver-sparc.c (cpu_names): Add SPARC-T5 entry.
28464 2017-06-14  Richard Biener  <rguenther@suse.de>
28466         PR tree-optimization/81083
28467         * tree-ssa-sccvn.c (vn_reference_lookup_3): Do not use abnormals
28468         as values.
28470 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28472         * config/rs6000/rs6000.c: Update all comments that mentioned SPE.
28473         (rs6000_expand_builtin): Remove RS6000_BTC_EVSEL.
28474         * config/rs6000/rs6000.h (RS6000_BTC_EVSEL): Delete.
28475         * config/rs6000/vxworks.h (VXCPU_FOR_8548): Delete.  Adjust former use.
28476         * config/rs6000/vxworksae.h (VXCPU_FOR_8548): Delete.
28477         * config/rs6000/vxworksmils.h (VXCPU_FOR_8548): Delete.
28479 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28481         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete VECTOR_SPE.
28482         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete VECTOR_SPE.
28484 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28486         * config/rs6000/rs6000.h (FIXED_SCRATCH): Delete.
28488 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28490         * config/rs6000/t-rtems: Don't handle SPE.
28492 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28494         * config/rs6000/t-linux: Don't handle SPE.
28496 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28498         * config/rs6000/eabispe.h: Delete file.
28500 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28502         * config/rs6000/t-spe: Delete file.
28504 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28506         * config/rs6000/rs6000.c (SPE_CONST_OFFSET_OK): Delete.
28507         (rs6000_legitimate_offset_address_p): Return false for anything in
28508         V2SImode or V2SFmode.
28510 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28512         * config/rs6000/rs6000-modes.def: Remove all 8-byte vector modes
28513         except V2SF and V2SI.  Rearrange the vector modes, and add comments.
28514         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove V8QImode
28515         and V4HImode.
28516         (reg_offset_addressing_ok_p): Remove V4HImode and V1DImode.
28517         (rs6000_legitimate_offset_address_p): Ditto.
28518         (rs6000_emit_move): Ditto.
28519         (rs6000_init_builtins): Remove V4HI_type_node.
28521 2017-06-13  Martin Liska  <mliska@suse.cz>
28523         PR sanitize/78204
28524         * asan.c (asan_sanitize_stack_p): Use sanitize_flags_p.
28525         (gate_asan): Likewise.
28526         * asan.h (asan_no_sanitize_address_p): Remove the function.
28527         (sanitize_flags_p): New function.
28528         * builtins.def: Fix coding style.
28529         * common.opt: Use renamed enum value.
28530         * convert.c (convert_to_integer_1): Use sanitize_flags_p.
28531         * doc/extend.texi: Document no_sanitize attribute.
28532         * flag-types.h (enum sanitize_code): Rename SANITIZE_NONDEFAULT
28533         to SANITIZE_UNDEFINED_NONDEFAULT.
28534         * gcc.c (sanitize_spec_function): Use the renamed enum value.
28535         * gimple-fold.c (optimize_atomic_compare_exchange_p):
28536         Use sanitize_flags_p.
28537         * gimplify.c (gimplify_function_tree): Likewise.
28538         * ipa-inline.c (sanitize_attrs_match_for_inline_p): Likewise.
28539         * opts.c (parse_no_sanitize_attribute): New function.
28540         (common_handle_option): Use renamed enum value.
28541         * opts.h (parse_no_sanitize_attribute): Declare.
28542         * tree.c (sanitize_flags_p): New function.
28543         * tree.h: Declared here.
28544         * tsan.c: Use sanitize_flags_p.
28545         * ubsan.c (ubsan_expand_null_ifn): Likewise.
28546         (instrument_mem_ref): Likewise.
28547         (instrument_bool_enum_load): Likewise.
28548         (do_ubsan_in_current_function): Remove the function.
28549         (pass_ubsan::execute): Use sanitize_flags_p.
28550         * ubsan.h: Remove do_ubsan_in_current_function
28551         * tree-cfg.c (print_no_sanitize_attr_value): New function.
28552         (dump_function_to_file): Use it here.
28554 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28556         PR tree-optimization/80803
28557         PR tree-optimization/81063
28558         * tree-sra.c (subtree_mark_written_and_enqueue): Move up in the file.
28559         (propagate_subaccesses_across_link): Enqueue subtree whenever
28560         necessary instead of relying on the caller.
28562 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28564         * tree-sra.c (add_access_to_work_queue): Only enqueue accesses
28565         that have a first_link.
28566         (sort_and_splice_var_accesses): Do not check first_link before
28567         enquing.
28568         (subtree_mark_written_and_enqueue): Likewise.
28569         (propagate_all_subaccesses): Likewise and do not stop at first
28570         parent with a first_link.
28572 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28574         * tree-sra.c (dump_access_tree_1): Fix accidental dumping to stderr
28575         instead of f.
28577 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28579         * match.pd: New pattern.
28581 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28583         * tree-vrp.c (is_masked_range_test): New function.
28584         (register_edge_assert_for): Determine ranges for
28585         some bit tests.
28587 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28589         PR tree-optimization/67328
28590         * fold-const.c (maskable_range_p): New function.
28591         (build_range_check): Generate bittests if possible.
28593 2017-06-13  Martin Liska  <mliska@suse.cz>
28595         * gimple-pretty-print.c (dump_probability): Add new argument.
28596         (dump_edge_probability): Dump both probability and count.
28597         (dump_gimple_label): Likewise.
28598         (dump_gimple_bb_header): Likewise.
28600 2017-06-13  Georg-Johann Lay  <avr@gjlay.de>
28602         PR target/81072
28603         * config/avr/avr-devices.c: Fix indentation.
28604         * config/avr/gen-avr-mmcu-specs.c: Dito.
28606 2017-06-13  Richard Biener  <rguenther@suse.de>
28608         * tree-vect-loop.c (vect_model_reduction_cost): Do not fail,
28609         instead get vector type from stmt_info.
28610         (vectorizable_reduction): Adjust.  Remove dead code.
28612 2017-06-13  Richard Biener  <rguenther@suse.de>
28614         PR middle-end/81065
28615         * fold-const.c (extract_muldiv_1): Remove bogus distribution
28616         case of C * (x * C2 + C3).
28617         (fold_addr_of_array_ref_difference): Properly fold index difference.
28619 2017-06-12  David S. Miller  <davem@davemloft.net>
28621         PR target/80968
28622         * config/sparc/sparc.md (return expander): Emit frame blockage if
28623         function uses alloca.
28625 2017-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
28627         * combine.c (make_field_assignment): Check len rather than the mode
28628         precision when calling force_to_mode.
28630 2017-06-12  Georg-Johann Lay  <avr@gjlay.de>
28632         Support multilibs and devices that see flash in RAM address range.
28634         PR target/81072
28635         * config/avr/avr-arch.h (avr_arch_id) <ARCH_AVRXMEGA3>: New enum.
28636         (avr_mcu_t) <flash_pm_offset>: New field.
28637         (avr_device_specific_features) <AVR_ISA_RCALL>: New enum.
28638         * config/avr/avr.h (AVR_SHORT_CALLS): New define.
28639         (AVR_HAVE_JMP_CALL): Don't set if AVR_SHORT_CALLS.
28640         (AVR_TINY_PM_OFFSET): Remove macro.
28641         * config/avr/avr.opt (-mshort-calls): New option.
28642         * config/avr/gen-avr-mmcu-specs.c (print_mcu)
28643         [*self_spec]: Add / remove -mshort-calls depending on AVR_ISA_RCALL.
28644         * config/avr/avr-c.c (avr_cpu_cpp_builtins)
28645         <__AVR_SHORT_CALLS__>: Built-in define if AVR_SHORT_CALLS.
28646         <__AVR_HAVE_JMP_CALL__>: Use AVR_HAVE_JMP_CALL as condition
28647         instead of avr_arch->have_jmp_call.
28648         <__AVR_PM_BASE_ADDRESS__>: Built-in define if avr_arch->flash_pm_offset.
28649         [AVR_TINY] <__AVR_TINY_PM_BASE_ADDRESS__>: Use
28650         avr_arch->flash_pm_offset to define.
28651         * config/avr/avr-devices.c (avr_arch_types): Add initializers for
28652         new field flash_pm_offset.  Add entry for avrxmega3.
28653         (avr_texinfo): Add entry for avrxmega3.
28654         * config/avr/avr-mcus.def: Add entries for: avrxmega3,
28655         attiny212, attiny214,
28656         attiny412, attiny414, attiny416, attiny417,
28657         attiny814, attiny816, attiny817,
28658         attiny1614, attiny1616, attiny1617,
28659         attiny3214, attiny3216, attiny3217.
28660         * config/avr/avr.c (avr_assemble_integer)[AVR_TINY]: Use
28661         avr_arch->flash_pm_offset instead of AVR_TINY_PM_OFFSET.
28662         (avr_print_operand_address) [AVR_TINY]: Same.
28663         (avr_asm_init_sections) <readonly_data_section>: Only patch
28664         callback if avr_arch->flash_pm_offset = 0.
28665         (avr_asm_named_section) <avr_need_copy_data_p>: Skip setting it
28666         for rodata if avr_arch->flash_pm_offset != 0.
28667         (avr_encode_section_info) [AVR_TINY]: Adjust comment.
28668         * config/avr/genmultilib.awk (dir_rcall, opt_rcall): New vars.
28669         (opts) [AVR_ISA_RCALL]: Append opt_rcall.
28670         (m_options): Append opt_rcall.
28671         (m_dirnames): Append dir_rcall.
28672         * config/avr/t-multilib: Regenerate.
28674         * configure.ac [target=avr]: Check whether avrxmega3 default
28675         linker description file works as needed.
28676         * configure: Regenerate.
28677         * doc/avr-mmcu.texi: Regenerate.
28678         * doc/invoke.texi (AVR Options) <-mshort-calls>: Document it.
28679         <__AVR_ARCH__>: Document avrxmega3 and 103.
28680         <__AVR_HAVE_JMP_CALL__>: Adjust documentation.
28681         <__AVR_SHORT_CALLS__>: Document it.
28682         <__AVR_PM_BASE_ADDRESS__>: Document it.
28683         * doc/extend.texi (AVR Options) <-mshort-calls>: Document it.
28684         (AVR Variable Attributes) <progmem>: Document this is
28685         not needed for avrxmega3.
28686         (AVR Named Address Spaces) <__flash>: Dito.
28688 2017-06-12  Jan Hubicka  <hubicka@ucw.cz>
28690         * cgraph.c (cgraph_node::dump): Complain about profile insanities.
28692 2017-06-12  Doug Rupp  <rupp@adacore.com>
28694         * config.gcc (*-*-vxworks*): Set use_gcc_stdint to "provide".
28695         Append vxworks-stdint.h to the tm_file list.
28696         * config/vxworks-stdint.h: New file.
28698 2017-06-12  Martin Liska  <mliska@suse.cz>
28700         PR tree-optimization/81041
28701         * tree-profile.c (gimple_gen_ic_func_profiler):
28702         Create an extra BB in profile-generate
28703         (gimple_gen_time_profiler): Likewise.
28705 2017-06-12  Jakub Jelinek  <jakub@redhat.com>
28707         PR tree-optimization/81003
28708         * tree-ssa-reassoc.c (force_into_ssa_name): New function.
28709         (update_range_test): Use it instead of force_gimple_operand_gsi.
28711 2017-06-12  Richard Biener  <rguenther@suse.de>
28713         PR tree-optimization/81053
28714         * tree-vect-loop.c (vect_is_simple_reduction): Handle PHI
28715         with backedge value not defined in loop.  Simplify def stmt
28716         compute.
28718 2017-06-11  Tom de Vries  <tom@codesourcery.com>
28720         PR target/79939
28721         * config/nvptx/nvptx.c (nvptx_cannot_force_const_mem): New function.
28722         Return true.
28723         (TARGET_CANNOT_FORCE_CONST_MEM): Redefine to
28724         nvptx_cannot_force_const_mem.
28726 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28728         * opts.c (finish_options): Move test for flag_split_stack after
28729         it has been initialized.
28731 2017-06-11  Jason Merrill  <jason@redhat.com>
28733         * tree.h (id_equal): New.
28734         * dwarf2out.c, hsa-gen.c, ipa-devirt.c, omp-expand.c,
28735         omp-simd-clone.c, read-rtl-function.c, tree-chkp.c, tree.c: Use it
28736         instead of strcmp of IDENTIFIER_POINTER.
28738 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28740         * ipa-inline-transform.c: Include function.h, cfg.h and basic-block.h
28741         (mark_all_inlined_calls_cdtor): Fix formating.
28742         (inline_transform): Rescale profile before inlining.
28744 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28746         * cgraph.h (cgraph_edge::clone): Update prototype.
28747         * cgraphclones.c (cgraph_edge::clone): Update profile scaling.
28748         (cgraph_node::create_clone): Update.
28749         (cgraph_node::create_version_clone): Update.
28750         * tree-inline.c (copy_bb): Update.
28751         (expand_call_inline): Update.
28753 2017-06-10  Segher Boessenkool  <segher@kernel.crashing.org>
28755         * config/rs6000/rs6000.c (emit_vrsave_prologue): New function,
28756         factored out from ...
28757         (rs6000_emit_prologue): ... here.
28759 2017-06-10  Segher Boessenkool  <segher@kernel.crashing.org>
28761         * config/rs6000/rs6000.c (emit_split_stack_prologue): New function,
28762         factored out from ...
28763         (rs6000_emit_prologue): ... here.
28765 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28767         * predict.c (drop_profile): Also drop individual bb/edge and cgraph
28768         edge counts.
28769         (handle_missing_profiles): Fix computation of tp_first_run.
28770         (counts_to_freqs): Do not touch freqs when count is 0.
28772 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28774         * cgraphbuild.c (cgraph_edge::rebuild_references): Do not touch
28775         profile.
28777 2017-06-10  Tom de Vries  <tom@codesourcery.com>
28779         * doc/sourcebuild.texi (Effective-Target Keywords, Environment
28780         attributes): Document signal effective target.
28782 2017-06-10  Tom de Vries  <tom@codesourcery.com>
28784         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
28785         Document effective target stack_size.
28787 2017-06-09  David Malcolm  <dmalcolm@redhat.com>
28789         * diagnostic.c (diagnostic_report_diagnostic): Only add fixits
28790         to the edit_context if they can be auto-applied.
28792 2017-06-9  Ian Lance Taylor  <iant@golang.org>
28794         * opts.c (finish_options): If -fsplit-stack, disable implicit
28795         -forder-blocks-and-partition.
28796         * doc/invoke.texi (Optimize Options): Document that when using
28797         -fsplit-stack -forder-blocks-and-partition is not implicitly
28798         enabled.
28800 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28802         * builtin-attrs.def (ATTR_NORETURN_NOTHROW_LEAF_COLD_LIST,
28803         ATTR_CONST_NORETURN_NOTHROW_LEAF_COLD_LIST,
28804         ATTR_TMPURE_NORETURN_NOTHROW_LEAF_COLD_LIST): New.
28805         * builtins.def (abort, trap, unreachable): Declare cold.
28806         * calls.c (flags_from_decl_or_type): Lookup ECF_COLD.
28807         * tree-core.h (ECF_COLD): New.
28808         * tree.c (set_call_expr_flags): Handle ECF_COLD.
28809         (build_common_builtin_nodes): Mark unreachable and abort as cold.
28811 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28813         * predict.c (unlikely_executed_stmt_p): Cleanup.
28815 2017-06-09  Richard Biener  <rguenther@suse.de>
28817         * tree-ssa-loop-im.c (execute_sm): Do not force multi-threaded
28818         model if the ref is always written to.
28820 2017-06-09  Tamar Christina  <tamar.christina@arm.com>
28822         * config/aarch64/aarch64.md (lrint<GPF:mode><GPI:mode>2): New.
28824 2017-06-09  Tamar Christina  <tamar.christina@arm.com>
28826         * config/arm/arm.c (arm_rtx_costs_internal): Make sdiv more expensive
28827         than udiv.
28829 2017-06-09  Tom de Vries  <tom@codesourcery.com>
28831         PR target/80855
28832         * config/nvptx/nvptx.md (define_expand "mov<QHSDISDFM>"): Error out with
28833         "target cannot support label values" when encountering LABEL_REF.
28835 2017-06-09  Martin Liska  <mliska@suse.cz>
28837         * tree-profile.c (gimple_gen_ic_profiler): Update comment.
28838         (gimple_gen_ic_func_profiler): Emit direct comparison
28839         of __gcov_indirect_call_callee with NULL.
28840         (gimple_gen_time_profiler): Change probability from
28841         PROB_VERY_UNLIKELY to PROB_UNLIKELY.
28843 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28845         * profile.c (edge_gcov_counts): Turn to pointer.
28846         (compute_branch_probabilities, compute_branch_probabilities): Update.
28847         (branch_prob): Do not clear edge_gcov_count.
28848         * profile.h (edge_gcov_counts): Turn to pointer.
28849         (edge_gcov_count): Update.
28851 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28853         * gimple.h (gimple_check_failed): Mark cold.
28855 2017-06-09  Richard Biener  <rguenther@suse.de>
28857         PR tree-optimization/66623
28858         * tree-vect-loop.c (vect_is_simple_reduction): Cleanup,
28859         refactor check_reduction into two parts, properly computing
28860         whether we have to check reduction validity for outer loop
28861         vectorization.
28863 2017-06-09  Richard Biener  <rguenther@suse.de>
28865         PR tree-optimization/79483
28866         * graphite-scop-detection.c (order): New global.
28867         (get_order): Compute bb to order mapping that satisfies code
28868         generation constraints.
28869         (cmp_pbbs): New helper.
28870         (build_scops): Start domwalk at entry block, sort generated
28871         pbbs.
28873 2017-06-09  Richard Biener  <rguenther@suse.de>
28875         PR middle-end/81007
28876         * ipa-polymorphic-call.c
28877         (ipa_polymorphic_call_context::restrict_to_inner_class):
28878         Skip FIELD_DECLs with error_mark_node type.
28879         * passes.def (all_lowering_passes): Run pass_build_cgraph_edges
28880         last again.
28882 2017-06-09  Martin Liska  <mliska@suse.cz>
28884         * predict.c (struct branch_predictor): New struct.
28885         (test_prediction_value_range): New test.
28886         (predict_c_tests): New function.
28887         * selftest-run-tests.c (selftest::run_tests): Run the function.
28888         * selftest.h: Declare new tests.
28890 2017-06-09  Segher Boessenkool  <segher@kernel.crashing.org>
28892         PR target/80966
28893         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack): Assert that
28894         gen_add3_insn did not fail.
28895         * config/rs6000/rs6000.md (add<mode>3): If asked to add a constant to
28896         r0, construct that number in a temporary reg and add that reg to r0.
28897         If asked to put the result in r0 as well, fail.
28899 2017-06-08  Will Schmidt  <will_schmidt@vnet.ibm.com>
28901         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
28902         for early expansion of vec_eqv.
28904 2017-06-08  Jakub Jelinek  <jakub@redhat.com>
28906         PR middle-end/81005
28907         * ubsan.c (instrument_null): Avoid pointless code temporary.
28908         (pass_ubsan::execute): Instrument aggregate arguments of calls.
28910 2017-06-08  Uros Bizjak  <ubizjak@gmail.com>
28912         PR target/81015
28913         Revert:
28914         2016-12-14  Uros Bizjak  <ubizjak@gmail.com>
28916         PR target/59874
28917         * config/i386/i386.md (*ctzhi2): New insn_and_split pattern.
28918         (*clzhi2): Ditto.
28920 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28922         * predict.c (unlikely_executed_edge_p): Move ahead.
28923         (probably_never_executed_edge_p): Use it.
28925 2017-06-08  Alexander Ivchenko  <alexander.ivchenko@intel.com>
28927         PR middle-end/79988
28928         * tree-chkp.c (chkp_gimple_call_builtin_p): Remove
28929         gimple_call_builtin_p call.
28931 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28933         * system.h (fancy_abort): Annotate by ATTRIBUTE_COLD.
28934         * rtl.h (rtl_check_failed_bounds, rtl_check_failed_type1,
28935         rtl_check_failed_type2, rtl_check_failed_code1,
28936         rtl_check_failed_code2, rtl_check_failed_code_mode,
28937         rtl_check_failed_block_symbol, cwi_check_failed_bounds,
28938         rtvec_check_failed_bounds, rtl_check_failed_flag,
28939         _fatal_insn_not_found, _fatal_insn): Likewise.
28940         * tree.h (tree_contains_struct_check_failed,
28941         tree_check_failed, tree_not_check_failed,
28942         tree_class_check_failed, tree_range_check_failed,
28943         tree_not_class_check_failed, tree_int_cst_elt_check_failed,
28944         tree_vec_elt_check_failed, phi_node_elt_check_failed,
28945         tree_operand_check_failed, omp_clause_check_failed,
28946         omp_clause_operand_check_failed, omp_clause_range_check_failed):
28947         Likewise.
28949 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28951         * cgraph.c (cgraph_edge::maybe_hot_p): Do not check
28952         flag_branch_probabilities.
28953         * ipa-inline.c (edge_badness): Likewise.
28954         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
28955         * postreload-gcse.c (eliminate_partially_redundant_load): Likewise.
28956         * predict.c (maybe_hot_frequency_p): Likewise.
28957         (probably_never_executed): Likewise.
28958         * sched-ebb.c (schedule_ebbs): Likewise.
28959         * sched-rgn.c (find_single_block_region): Likewise.
28960         * tracer.c (tail_duplicate): Likewise.
28962 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28964         * opts.c (finish_options): x_flag_reorder_blocks_and_partition no
28965         longer requires x_flag_profile_use.
28967 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28969         * cfgrtl.c (cfg_layout_initialize): Check crtl->has_bb_partition
28970         instead of flag_reorder_blocks_and_partition.
28971         * dbxout.c (dbxout_function_end): Likewise.
28972         * dwarf2out.c (gen_subprogram_die): Likewise.
28973         * haifa-sched.c (sched_create_recovery_edges): Likewise.
28974         * hw-doloop.c (reorg_loops): Likewise.
28975         * varasm.c (assemble_start_function,
28976         assemble_end_function): Likewise.
28977         (decide_function_section): Do not check for
28978         flag_reorder_blocks_and_partition.
28980 2017-06-08  Alexander Ivchenko  <alexander.ivchenko@intel.com>
28982         * tree-chkp.c (chkp_get_hard_register_var_fake_base_address):
28983         New function.
28984         (chkp_get_hard_register_fake_addr_expr): Ditto.
28985         (chkp_build_addr_expr): Add check for hard reg case.
28986         (chkp_parse_array_and_component_ref): Ditto.
28987         (chkp_find_bounds_1): Ditto.
28988         (chkp_process_stmt): Don't generate bounds store for
28989         hard reg case.
28991 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28993         * predict.c (maybe_hot_bb_p): Do not check profile status.
28994         (maybe_hot_edge_p): Likewise.
28995         (probably_never_executed): Check for zero counts even if profile
28996         is not read.
28997         (unlikely_executed_edge_p): New function.
28998         (unlikely_executed_stmt_p): New function.
28999         (unlikely_executed_bb_p): New function.
29000         (set_even_probabilities): Use unlikely predicates.
29001         (combine_predictions_for_bb): Likewise.
29002         (predict_paths_for_bb): Likewise.
29003         (predict_paths_leading_to_edge): Likewise.
29004         (determine_unlikely_bbs): New function.
29005         (estimate_bb_frequencies): Use it.
29006         (compute_function_frequency): Use zero counts even if profile is
29007         not read.
29008         * profile-count.h: Fix typo.
29010 2017-08-08  Julia Koval  <julia.koval@intel.com>
29012         * config/i386/avx512bwintrin.h (_mm512_mask_cvtepi16_storeu_epi8,
29013         _mm512_mask_cvtsepi16_storeu_epi8,
29014         _mm512_mask_cvtusepi16_storeu_epi8): New intrinsics.
29015         * config/i386/avx512vlbwintrin.h (_mm256_mask_cvtepi16_storeu_epi8,
29016         _mm_mask_cvtsepi16_storeu_epi8, _mm256_mask_cvtsepi16_storeu_epi8,
29017         _mm_mask_cvtusepi16_storeu_epi8, _mm256_mask_cvtusepi16_storeu_epi8,
29018         _mm_mask_cvtepi16_storeu_epi8): New intrinsics.
29019         * config/i386/i386-builtin-types.def (PV8Q, V8QI): New pointer type.
29020         (VOID_FTYPE_PV32QI_V32HI_USI, VOID_FTYPE_PV8QI_V8HI_UQI,
29021         VOID_FTYPE_PV16QI_V16HI_UHI): New function types.
29022         * config/i386/i386-builtin.def (__builtin_ia32_pmovwb128mem_mask,
29023         __builtin_ia32_pmovwb256mem_mask, __builtin_ia32_pmovswb128mem_mask,
29024         __builtin_ia32_pmovswb256mem_mask, __builtin_ia32_pmovuswb128mem_mask,
29025         __builtin_ia32_pmovuswb256mem_mask,
29026         __builtin_ia32_pmovuswb512mem_mask, __builtin_ia32_pmovswb512mem_mask)
29027         __builtin_ia32_pmovwb512mem_mask): New builtins.
29029 2017-08-08  Julia Koval  <julia.koval@intel.com>
29031         PR target/73350,80862
29032         * config/i386/subst.md (round): Fix round pattern.
29033         * config/i386/i386.c (ix86_erase_embedded_rounding):
29034         Fix erasing rounding for the fixed pattern.
29036 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
29038         * cfgbuild.c (find_many_sub_basic_blocks): Fix thinko.
29040 2017-06-08  Martin Liska  <mliska@suse.cz>
29042         PR gcov-profile/80911
29043         * gcov.c (block_info::block_info): New constructor.
29045 2017-06-07  Carl Love  <cel@us.ibm.com>
29047         * config/rs6000/rs6000-c: The return type of the following
29048         built-in functions was implemented as int not long long.  Fix sign
29049         of return value for the unsigned version of vec_mulo and vec_mule.
29050         vector unsigned long long vec_bperm (vector unsigned long long,
29051                                              vector unsigned char)
29052         vector signed long long vec_mule (vector signed int,
29053                                           vector signed int)
29054         vector unsigned long long vec_mule (vector unsigned int,
29055                                             vector unsigned int)
29056         vector signed long long vec_mulo (vector signed int,
29057                                           vector signed int)
29058         vector unsigned long long vec_mulo (vector unsigned int,
29059                                             vector unsigned int)
29060         * doc/extend.texi: Fix the documentation for the built-in
29061         functions.
29063 2017-06-07  Carl Love  <cel@us.ibm.com>
29065         PR target/80982
29066         * config/rs6000/altivec.md (double<mode>2): Fix the implementation of
29067         for BE.
29069 2017-06-07  Carl Love  <cel@us.ibm.com>
29071         * config/rs6000/altivec.md: Fix argument swizzle in vec_doublel
29072         support, Generate       doublehv for signed int/float for BE case only.
29074 2017-06-07  Alexander Monakov  <amonakov@ispras.ru>
29076         * doc/invoke.texi (mcx16): Rewrite.
29078 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29080         * config/rs6000/predicates.md (rs6000_nonimmediate_operand): Delete.
29081         * config/rs6000/rs6000.md (*movsi_internal1, movsi_from_sf,
29082         *mov<mode>_softfloat, and an anonymous splitter): Use
29083         nonimmediate_operand instead of rs6000_nonimmediate_operand.
29085 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29087         * config/rs6000/darwin.h (REGISTER_NAMES): Delete the SPE_ACC and
29088         SPEFSCR registers.
29089         * config/rs6000/rs6000.c (rs6000_reg_names, alt_reg_names): Ditto.
29090         (enum rs6000_reg_type): Delete SPE_ACC_TYPE and SPEFSCR_REG_TYPE.
29091         (rs6000_debug_reg_global): Adjust.
29092         (rs6000_init_hard_regno_mode_ok): Adjust.
29093         (rs6000_dbx_register_number): Adjust.
29094         * config/rs6000/rs6000.h (FIRST_PSEUDO_REGISTER): Change to 115.
29095         (FIXED_REGISTERS, CALL_USED_REGISTERS, CALL_REALLY_USED_REGISTERS):
29096         Remove SPE_ACC and SPEFSCR.
29097         (REG_ALLOC_ORDER): Ditto.
29098         (FRAME_POINTER_REGNUM): Change to 111.
29099         (enum reg_class): Remove the SPE_ACC and SPEFSCR registers.
29100         (REG_CLASS_NAMES): Ditto.
29101         (REG_CLASS_CONTENTS): Delete the SPE_ACC and SPEFSCR registers.
29102         (REGISTER_NAMES): Ditto.
29103         (ADDITIONAL_REG_NAMES): Ditto.
29104         (rs6000_reg_names): Ditto.
29105         * config/rs6000/rs6000.md: Renumber some register number
29106         define_constants.
29108 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29110         * config/rs6000/darwin.h (REGISTER_NAMES): Delete the SPE high
29111         registers.
29112         * config/rs6000/rs6000.c (rs6000_reg_names, alt_reg_names): Ditto.
29113         * config/rs6000/rs6000.h (FIRST_PSEUDO_REGISTER): Change from 149
29114         to 117.
29115         (DWARF_REG_TO_UNWIND_COLUMN): Do not define.
29116         (FIXED_REGISTERS, CALL_USED_REGISTERS, CALL_REALLY_USED_REGISTERS):
29117         Delete the SPE high registers.
29118         (REG_ALLOC_ORDER): Ditto.
29119         (enum reg_class): Remove SPE_HIGH_REGS.
29120         (REG_CLASS_NAMES): Ditto.
29121         (REG_CLASS_CONTENTS): Delete the SPE high registers.
29122         (REGISTER_NAMES): Ditto.
29123         (rs6000_reg_names): Ditto.
29124         * doc/tm.texi.in: Remove SPE as example.
29125         * doc/tm.texi: Regenerate.
29127 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29129         * config/rs6000/8540.md (ppc8540_brinc): Delete.
29130         * config/rs6000/e500mc.md (e500mc_brinc): Delete.
29131         * config/rs6000/e500mc64.md (e500mc64_brinc): Delete.
29132         * config/rs6000/rs6000.md (type): Remove "brinc".
29134 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29136         * config.gcc (powerpc*-*-*): Don't add spe.h to extra_headers.
29137         (powerpc*-linux*spe*): Use ${cpu_type} instead of rs6000.
29138         * config/rs6000/linuxspe.h: Delete file.
29139         * config/rs6000/rs6000.md: Don't include spe.md.
29140         * config/rs6000/spe.h: Delete file.
29141         * config/rs6000/spe.md: Delete file.
29142         * config/rs6000/t-rs6000: Remove spe.md.
29144 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29146         * config/rs6000/predicates.md (reg_or_mem_operand): Reformat.
29147         (reg_or_none500mem_operand): Delete.
29148         * config/rs6000/rs6000.md (extendsfdf2): Use reg_or_mem_operand
29149         instead of reg_or_none500mem_operand.
29151 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29153         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
29154         handling of SPE flags.
29155         * config/rs6000/rs6000.opt (-mspe, -mspe=no, -mspe=yes): Delete.
29157 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29159         * config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
29160         SPE ABI handling.
29161         * config/rs6000/paired.md (paired_negv2sf2): Rename to negv2sf2.
29162         (paired_absv2sf2, paired_addv2sf3, paired_subv2sf3, paired_mulv2sf3,
29163         paired_divv2sf3): Similar.
29164         * config/rs6000/predicates.md: Replace TARGET_SPE, TARGET_SPE_ABI,
29165         SPE_VECTOR_MODE and SPE_HIGH_REGNO_P by 0; simplify.
29166         * config/rs6000/rs6000-builtin.def: Delete RS6000_BUILTIN_E and
29167         RS6000_BUILTIN_S.
29168         Delete BU_SPE_1, BU_SPE_2, BU_SPE_3, BU_SPE_E, BU_SPE_P, and BU_SPE_X.
29169         Rename the paired_* instruction patterns.
29170         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Do not
29171         define __SPE__.
29172         * config/rs6000/rs6000-protos.h (invalid_e500_subreg): Delete.
29173         * config/rs6000/rs6000.c: Delete RS6000_BUILTIN_E and RS6000_BUILTIN_S.
29174         (struct rs6000_stack): Delete fields spe_gp_save_offset, spe_gp_size,
29175         spe_padding_size, and spe_64bit_regs_used.  Replace TARGET_SPE and
29176         TARGET_SPE_ABI with 0, simplify.  Replace SPE_VECTOR_MODE with
29177         PAIRED_VECTOR_MODE.
29178         (struct machine_function): Delete field spe_insn_chain_scanned_p.
29179         (spe_func_has_64bit_regs_p): Delete.
29180         (spe_expand_predicate_builtin): Delete.
29181         (spe_expand_evsel_builtin): Delete.
29182         (TARGET_DWARF_REGISTER_SPAN): Do not define.
29183         (TARGET_MEMBER_TYPE_FORCES_BLK): Do not define.
29184         (invalid_e500_subreg): Delete.
29185         (rs6000_legitimize_address): Always force_reg op2 as well, for
29186         paired single memory accesses.
29187         (rs6000_member_type_forces_blk): Delete.
29188         (rs6000_spe_function_arg): Delete.
29189         (rs6000_expand_unop_builtin): Delete SPE handling.
29190         (rs6000_expand_binop_builtin): Ditto.
29191         (spe_expand_stv_builtin): Delete.
29192         (bdesc_2arg_spe): Delete.
29193         (spe_expand_builtin): Delete.
29194         (spe_expand_predicate_builtin): Delete.
29195         (spe_expand_evsel_builtin): Delete.
29196         (rs6000_invalid_builtin): Remove RS6000_BTM_SPE handling.
29197         (spe_init_builtins): Delete.
29198         (spe_func_has_64bit_regs_p): Delete.
29199         (savres_routine_name): Delete "info" parameter.  Adjust callers.
29200         (rs6000_emit_stack_reset): Ditto.
29201         (rs6000_dwarf_register_span): Delete.
29202         * config/rs6000/rs6000.h (TARGET_SPE_ABI, TARGET_SPE,
29203         UNITS_PER_SPE_WORD, SPE_HIGH_REGNO_P, SPE_SIMD_REGNO_P,
29204         SPE_VECTOR_MODE, RS6000_BTM_SPE, RS6000_BUILTIN_E, RS6000_BUILTIN_S):
29205         Delete.
29206         * config/rs6000/rs6000.md (FIRST_SPE_HIGH_REGNO, LAST_SPE_HIGH_REGNO):
29207         Delete.
29208         * config/rs6000/rs6000.opt (-mabi=spe, -mabi=no-spe): Delete.
29209         * config/rs6000/spe.md: Delete every pattern that uses TARGET_SPE.
29210         * config/rs6000/vector.md (absv2sf2, negv2sf2, addv2sf3, subv2sf3,
29211         mulv2sf3, divv2sf3): Delete expanders.
29213 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29215         config/rs6000/rs6000.md (UNSPEC_MV_CR_GT): Delete.
29217 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29219         * config/rs6000/rs6000-protos.h (output_e500_flip_gt_bit): Delete.
29220         * config/rs6000/rs6000.c: Ditto.
29222 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29224         * config/rs6000/predicated.md (rs6000_cbranch_operator): Delete.
29225         * config/rs6000/rs6000.md: Replace rs6000_cbranch_operator by
29226         comparison_operator.
29228 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29230         * config/rs6000/rs6000.c: Remove everything related to -mfloat-gprs.
29231         * config/rs6000/rs6000.opt: Ditto.
29232         * config/rs6000/t-rtems: Ditto.
29234 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29236         * config/rs6000/predicates.md: Replace TARGET_E500_DOUBLE and
29237         TARGET_E500_SINGLE by 0, simplify.
29238         * config/rs6000/rs6000.c: Ditto.
29239         (rs6000_option_override_internal): Delete CHECK_E500_OPTIONS.
29240         (spe_build_register_parallel): Delete.
29241         * config/rs6000/rs6000.h: Delete TARGET_E500_SINGLE,
29242         TARGET_E500_DOUBLE, and CHECK_E500_OPTIONS.
29243         * config/rs6000/rs6000.md: Replace TARGET_E500_DOUBLE,
29244         TARGET_E500_SINGLE, and <E500_CONVERT> by 0, simplify.
29245         (E500_CONVERT): Delete.
29246         * config/rs6000/spe.md: Remove many patterns and all define_constants.
29248 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29250         * config/rs6000/darwin.md: Replace TARGET_FPRS by 1 and simplify.
29251         * config/rs6000/dfp.md: Ditto.
29252         (negdd2, *negdd2_fpr): Merge.
29253         (absdd2, *absdd2_fpr): Merge.
29254         (negtd2, *negtd2_fpr): Merge.
29255         (abstd2, *abstd2_fpr): Merge.
29256         * config/rs6000/e500.h: Delete file.
29257         * config/rs6000/predicates.md (rs6000_cbranch_operator): Replace
29258         TARGET_FPRS by 1 and simplify.
29259         * config/rs6000/rs6000-c.c: Ditto.
29260         * config/rs6000/rs6000.c: Ditto.  Also replace TARGET_SF_SPE and
29261         TARGET_DF_SPE by 0.
29262         * config/rs6000/rs6000.h: Ditto.  Delete TARGET_SF_SPE and
29263         TARGET_DF_SPE.
29264         * config/rs6000/rs6000.md: Ditto.
29265         (floatdidf2, *floatdidf2_fpr): Merge.
29266         (move_from_CR_gt_bit): Delete.
29267         * config/rs6000/spe.md: Replace TARGET_FPRS by 1 and simplify.
29268         (E500_CR_IOR_COMPARE): Delete.
29269         (All patterns that require !TARGET_FPRS): Delete.
29270         * config/rs6000/vsx.md: Replace TARGET_FPRS by 1 and simplify.
29272 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29274         * passes.def (pass_iv_canon): Move before pass_loop_distribution.
29276 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29278         * graphds.c (add_edge): Intitialize edge's attached data.
29279         (foll_in_subgraph, dfs_fst_edge, dfs_next_edge): New function
29280         pointer parameter.  Call pointed function on each edge during
29281         graph traversing.  Skip traversing the edge when the function
29282         returns true.
29283         (graphds_dfs, graphds_scc): Ditto.
29284         (for_each_edge): New parameter.  Pass the new parameter to callback
29285         function.
29286         * graphds.h (skip_edge_callback): New function pointer type.
29287         (graphds_dfs, graphds_scc): New function pointer parameter.
29288         (graphds_edge_callback, for_each_edge): New parameter.
29290 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29292         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test): Factor
29293         out code checking if runtime alias check is possible to below ...
29294         Call the new function.
29295         * tree-data-ref.c (runtime_alias_check_p): ... to new function.
29296         * tree-data-ref.h (runtime_alias_check_p): New decalaration.
29298 2017-06-07  Marek Polacek  <polacek@redhat.com>
29300         PR sanitizer/80932
29301         * fold-const.c (extract_muldiv_1) <case MINUS_EXPR>: Add
29302         TYPE_OVERFLOW_WRAPS check.
29304 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29306         * tree-vect-loop-manip.c (vect_do_peeling): Don't skip vector loop
29307         if versioning is required.
29308         * tree-vect-loop.c (vect_analyze_loop_2): Merge niter check for loop
29309         peeling with the check for versioning.
29311 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29313         * tree-vectorizer.h (vect_build_loop_niters): New parameter.
29314         * tree-vect-loop-manip.c (vect_build_loop_niters): New parameter.
29315         Set true to new parameter if new ssa variable is defined.
29316         (vect_gen_vector_loop_niters): Refactor.  Set range information
29317         for the new vector loop bound variable.
29318         (vect_do_peeling): Ditto.
29320 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29322         * tree-affine.c (ssa.h): Include header file.
29323         (tree_to_aff_combination): Handle (T1)(X - CST) when inner type
29324         has wrapping overflow behavior.
29326 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29328         * tree-affine.c (tree_to_aff_combination): Handle (T1)(X + X).
29330 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29332         (aff_combination_expand): Move (T1)(X *+- CST) simplification to ...
29333         (tree_to_aff_combination): ... here.
29335 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29337         * tree-ssa-loop-ivopts.c (ivopts_estimate_reg_pressure): New
29338         reg_pressure model function.
29339         (ivopts_global_cost_for_size): Delete.
29340         (determine_set_costs, iv_ca_recount_cost): Call new model function
29341         ivopts_estimate_reg_pressure.
29343 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29345         * config/aarch64/aarch64.c (aarch64_rtx_costs): Make sdiv more
29346         expensive than udiv.  Remove floating point cases from mod.
29348 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29350         * config/arm/aarch-cost-tables.h (cortexa53_extra_cost):
29351         Increase idiv cost.
29353 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29355         * config/aarch64/aarch64.md
29356         (copysignsf3): Fix mask generation.
29358 2017-06-07  Jakub Jelinek  <jakub@redhat.com>
29360         * dumpfile.h (enum tree_dump_index): Rename TDI_generic to
29361         TDI_gimple.
29362         (class dump_manager): Add register_dumps method.
29363         * dumpfile.c: Include langhooks.h.
29364         (dump_files): Use 0 instead of 3/4/5 for TDI_{original,gimple,nested}.
29365         (FIRST_AUTO_NUMBERED_DUMP): Decrease to 1.
29366         (FIRST_ME_AUTO_NUMBERED_DUMP): Define.
29367         (dump_manager::dump_register): Start with 512 entries instead of 32.
29368         (dump_manager::register_dumps): New method.
29369         * toplev.c (general_init): Instead of invoking register_dumps
29370         langhook, invoke register_dumps method on the dump manager.
29371         * gimplify.c (gimplify_function_tree): Use TDI_gimple instead of
29372         TDI_generic.
29374 2017-06-07  Richard Sandiford  <richard.sandiford@linaro.org>
29376         * doc/md.texi: Clarify the restrictions on a define_insn condition.
29377         Say that # requires an associated define_split to exist, and that
29378         the define_split must be suitable for use after register allocation.
29380 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29382         * cfgbuild.c (find_bb_boundaries): Initialize profile of split blocks.
29383         (compute_outgoing_frequencies): Also initialize zero counts.
29384         (find_many_sub_basic_blocks): Do not produce uninitialized profile
29385         around loops; preserve more of profile when nothing changes.
29387 2017-06-06  Jim Wilson  <jim.wilson@linaro.org>
29389         * config/aarch64/aarch64-cost-tables.h (qdf24xx_extra_costs): Move to
29390         here.
29391         * config/arm/aarch-cost-tables.h (qdf24xx_extra_costs): From here.
29392         * config/arm/arm-cpu-cdata.h: Regenerate.
29393         * config/arm/arm-cpu-data.h, config/arm/arm-cpu.h: Likewise.
29394         * config/arm/arm-tables.opt, config/arm/arm-tune.md: Likewise.
29395         * config/arm/arm-cpus.in: Delete falkor and qdf24xx entries.
29396         * config/arm/arm.c (arm_qdf24xx_tune): Delete.
29397         * config/arm/bpabi.h (BE8_LINK_SPEC): Delete falkor and qdf24xx
29398         support.
29399         * config/arm/t-aprofile (MULTILIB_MATCHES): Delete falkor and qdf24xx
29400         support.
29401         * config/arm/t-rmprofile: Likewise.
29402         * doc/invoke.texi (ARM Options): Drop falkor and qdf24xx support.
29404 2017-06-06  David S. Miller  <davem@davemloft.net>
29406         PR target/80968
29407         * config/sparc/sparc.c (sparc_expand_prologue): Emit frame
29408         blockage if function uses alloca.
29410 2017-06-06  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
29412         * tree-ssa-loop-prefetch.c (struct mem_ref_group, struct mem_ref):
29413         New "uid" fields to hold pretty-print IDs of group and ref.
29414         Memory references are now identified as <group_id>:<ref_id>
29415         instead of using [random] addresses.
29416         (dump_mem_details): Simplify, no functional change.
29417         (dump_mem_ref): Simplify and make output more concise.
29418         Replace couple of fprintf's throughout code with calls to dump_mem_ref.
29419         (find_or_create_group): Initialize group uid.
29420         (record_ref): Initialize ref uid.  Improve debug output.
29421         (prune_group_by_reuse, should_issue_prefetch_p,)
29422         (should_issue_prefetch_p, schedule_prefetches, issue_prefetch_ref,)
29423         (mark_nontemporal_store, determine_loop_nest_reuse):
29424         Improve debug output.
29426 2017-06-06  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
29428         * dbgcnt.def (prefetch): New debug counter.
29429         * tree-ssa-loop-prefetch.c (dbgcnt.h): New include.
29430         (schedule_prefetches): Stop issueing prefetches if debug counter
29431         tripped.
29433 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29435         * doc/sourcebuild.texi (Testsuites, C Language Testsuites,
29436         gcc.c-torture/compile): Remove mention of NO_LABEL_VALUES in fixme.
29438 2017-06-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29440         * config/aarch64/atomics.md (atomic_compare_and_swap<mode> expander):
29441         Use aarch64_reg_or_zero predicate for operand 4.
29442         (aarch64_compare_and_swap<mode> define_insn_and_split):
29443         Use aarch64_reg_or_zero predicate for operand 3.  Add 'Z' constraint.
29444         (aarch64_store_exclusive<mode>): Likewise for operand 2.
29446 2017-06-06  Thomas Preud'homme  <thomas.preudhomme@arm.com>
29448         * config/arm/arm.c (arm_compute_save_reg_mask): Rename into ...
29449         (arm_compute_save_core_reg_mask): This.
29450         (thumb1_compute_save_reg_mask): Rename into ...
29451         (thumb1_compute_save_core_reg_mask): This.
29452         (arm_compute_save_reg0_reg12_mask): Adapt comment.
29453         (arm_compute_frame_layout): Likewise.
29455 2017-06-06  Richard Biener  <rguenther@suse.de>
29457         PR tree-optimization/80974
29458         * tree-ssa-sccvn.c (set_ssa_val_to): Do not change but only
29459         keep or clear leaders SSA info.
29461 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29463         * config/nvptx/nvptx.c (split_mode_p): New function.
29464         (nvptx_declare_function_name, nvptx_print_operand): Use split_mode_p.
29466 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29468         * config/nvptx/nvptx.c (nvptx_print_operand): Use maybe_split_mode.
29470 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29472         PR bootstrap/80978
29473         * tree-cfg.c (execute_fixup_cfg): Fix condition on when to rescale
29474         profile.
29476 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29478         * shrink-wrap.c (handle_simple_exit): Update profile.
29479         (try_shrink_wrapping): Upate profile.
29481 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29483         * predict.c (tree_estimate_probability_bb): Add LOCAL_ONLY.
29484         (tree_guess_outgoing_edge_probabilities): New.
29485         * predict.h (tree_guess_outgoing_edge_probabilities): Declare.
29486         * tree-cfg.c (gimple_find_sub_bbs): Propagate profile.
29488 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29490         * ipa-split.c (split_function): Initialize return bb profile.
29492 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29494         * profile.c (compute_branch_probabilities): Also initialize
29495         EXIT_BLOCK profile.
29497 2017-06-06  Richard Biener  <rguenther@suse.de>
29499         PR tree-optimization/80928
29500         * tree-vect-loop.c (vect_update_vf_for_slp): Amend dumps.
29501         (vect_analyze_loop_operations): Properly guard analysis for
29502         pure SLP case.
29503         (vect_transform_loop): Likewise.
29504         (vect_analyze_loop_2): Also reset SLP type on PHIs.
29505         (vect_model_induction_cost): Do not cost for pure SLP.
29506         (vectorizable_induction): Pass in SLP node, implement SLP vectorization
29507         of induction in inner loop vectorization.
29508         * tree-vect-slp.c (vect_create_new_slp_node): Handle PHIs.
29509         (vect_get_and_check_slp_defs): Handle vect_induction_def.
29510         (vect_build_slp_tree): Likewise.  Handle PHIs as terminating the
29511         recursion.
29512         (vect_analyze_slp_cost_1): Cost induction.
29513         (vect_detect_hybrid_slp_stmts): Handle PHIs.
29514         (vect_get_slp_vect_defs): Likewise.
29515         * tree-vect-stmts.c (vect_analyze_stmt): Handle induction.
29516         (vect_transform_stmt): Handle SLP reductions.
29517         * tree-vectorizer.h (vectorizable_induction): Adjust.
29519 2017-06-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
29521         * config/rs6000/rs6000.c (make_resolver_func): Update
29522         init_lowered_empty_function call.
29524 2017-06-05  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29526         * doc/invoke.texi: Document the -fprofile-abs-path option.
29527         * common.opt (fprofile-abs-path): New option.
29528         * gcov-io.h (gcov_write_filename): Declare.
29529         * gcov-io.c (gcov_write_filename): New function.
29530         * coverage.c (coverage_begin_function): Use gcov_write_filename.
29531         * profile.c (output_location): Likewise.
29533 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29535         * shring-wrap.c: Revert accidental commit.
29537 2017-06-05  Volker Reichelt  <v.reichelt@netcologne.de>
29539         * doc/invoke.texi (-Wduplicated-branches): Add to warning list.
29541 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29543         * cfgexpand.c (expand_gimple_tailcall): Initialize profile of
29544         new edge.
29545         * ipa-inline.c (want_inline_self_recursive_call_p): Watch for missing
29546         profile in callgraph edge.
29547         * profile-count.h (apply_probability): If THIS is 0, then result is 0
29548         (apply_scale): Likewise.
29549         * tree-inline.c (copy_bb, copy_edges_for_bb, copy_cfg_body):
29550         Also scale profile when inlining function with zero profile.
29551         (initialize_cfun): Update exit block profile even when it is zero.
29552         * tree-ssa-threadupdate.c (clear_counts_path): Handle correctly case
29553         when profile is read.
29555 2017-06-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
29557         * config/rs6000/rs6000.c (toplevel): Include attribs.h.
29558         (CLONE_*): New constants to define the processors we can generate
29559         code for with the target_clone attribute.
29560         (rs6000_clone_map): New array to identify which clone processors
29561         the current program is running on.
29562         (TARGET_COMPARE_VERSION_PRIORITY): Define to enable the
29563         target_clone attribute.
29564         (TARGET_GENERATE_VERSION_DISPATCHER_BODY): Likewise.
29565         (TARGET_GET_FUNCTION_VERSIONS_DISPATCHER): Likewise.
29566         (TARGET_OPTION_FUNCTION_VERSIONS): Likewise.
29567         (cpu_expand_builtin): Add support for target_clone attribute.
29568         (rs6000_valid_attribute_p): Allow "default" attribute.
29569         (get_decl_name): New debug function to simplify printing the
29570         current function name in debugging statements.
29571         (rs6000_clone_priority): New functions to support the target_clone
29572         attribute, and be able to generate code to switch between ISA 2.05
29573         through ISA 3.0 (power6 through power9).
29574         (rs6000_compare_version_priority): Likewise.
29575         (rs6000_get_function_versions_dispatcher): Likewise.
29576         (make_resolver_func): Likewise.
29577         (add_condition_to_bb): Likewise.
29578         (dispatch_function_versions): Likewise.
29579         (rs6000_generate_version_dispatcher_body): Likewise.
29580         (rs6000_can_inline_p): Call get_decl_name for debugging usage.
29581         (fusion_gpr_load_p): Fix a spacing issue.
29582         * doc/extend.texi (Common Function Attributes): Document that the
29583         PowerPC supports the target_clone attribute.
29585 2017-06-05  Thomas Preud'homme  <thomas.preudhomme@arm.com>
29587         * config/arm/arm.h: explain F symbol found in description of ARM
29588         register allocation in its legend.
29590 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29592         * config/mips/frame-header-opt.c: Include profile-count.h.
29593         * config/riscv/riscv.c: Include profile-count.h
29595 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29597         * tree-ssa-loop-im.c (execute_sm_if_changed): Add FLAG_BBS parameter;
29598         update profile.
29599         (sm_set_flag_if_changed): Add bbs field.
29600         (execute_sm_if_changed_flag_set): Pass BBS.
29601         (execute_sm): Update.
29603 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29605         * config/aarch64/aarch64-simd.md (aarch64_store_lane0<mode>):
29606         New pattern.
29608 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29610         * config/aarch64/aarch64.md (sub<mode>3_compare1_imm): New define_insn.
29611         (peephole2): New peephole2 to emit the above.
29612         * config/aarch64/predicates.md (aarch64_sub_immediate): New predicate.
29614 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29616         * config/aarch64/aarch64.c (define_peephole2 above
29617         *sub_<shift>_<mode>): New peephole.
29619 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
29621         * config/i386/i386.c (make_resolver_func): Update.
29622         * Makefile.in: Add profile-count.h and profile-count.o
29623         * auto-profile.c (afdo_indirect_call): Update to new API.
29624         (afdo_set_bb_count): Update.
29625         (afdo_propagate_edge): Update.
29626         (afdo_propagate_circuit): Update.
29627         (afdo_calculate_branch_prob): Update.
29628         (afdo_annotate_cfg): Update.
29629         * basic-block.h: Include profile-count.h
29630         (struct edge_def): Turn count to profile_count.
29631         (struct basic_block_def): Likewie.
29632         (REG_BR_PROB_BASE): Move to profile-count.h
29633         (RDIV): Move to profile-count.h
29634         * bb-reorder.c (max_entry_count): Turn to profile_count.
29635         (find_traces): Update.
29636         (rotate_loop):Update.
29637         (connect_traces):Update.
29638         (sanitize_hot_paths):Update.
29639         * bt-load.c (migrate_btr_defs): Update.
29640         * cfg.c (RDIV): Remove.
29641         (init_flow): Use alloc_block.
29642         (alloc_block): Uninitialize count.
29643         (unchecked_make_edge): Uninitialize count.
29644         (check_bb_profile): Update.
29645         (dump_edge_info): Update.
29646         (dump_bb_info): Update.
29647         (update_bb_profile_for_threading): Update.
29648         (scale_bbs_frequencies_int): Update.
29649         (scale_bbs_frequencies_gcov_type): Update.
29650         (scale_bbs_frequencies_profile_count): New.
29651         * cfg.h (update_bb_profile_for_threading): Update.
29652         (scale_bbs_frequencies_profile_count): Declare.
29653         * cfgbuild.c (compute_outgoing_frequencies): Update.
29654         (find_many_sub_basic_blocks): Update.
29655         * cfgcleanup.c (try_forward_edges): Update.
29656         (try_crossjump_to_edge): Update.
29657         * cfgexpand.c (expand_gimple_tailcall): Update.
29658         (construct_exit_block): Update.
29659         * cfghooks.c (verify_flow_info): Update.
29660         (dump_bb_for_graph): Update.
29661         (split_edge): Update.
29662         (make_forwarder_block): Update.
29663         (duplicate_block): Update.
29664         (account_profile_record): Update.
29665         * cfgloop.c (find_subloop_latch_edge_by_profile): Update.
29666         (get_estimated_loop_iterations): Update.
29667         * cfgloopanal.c (expected_loop_iterations_unbounded): Update.
29668         (single_likely_exit): Update.
29669         * cfgloopmanip.c (scale_loop_profile): Update.
29670         (loopify): Update.
29671         (set_zero_probability): Update.
29672         (lv_adjust_loop_entry_edge): Update.
29673         * cfgrtl.c (force_nonfallthru_and_redirect): Update.
29674         (purge_dead_edges): Update.
29675         (rtl_account_profile_record): Update.
29676         * cgraph.c (cgraph_node::create): Uninitialize count.
29677         (symbol_table::create_edge): Uninitialize count.
29678         (cgraph_update_edges_for_call_stmt_node): Update.
29679         (cgraph_edge::dump_edge_flags): Update.
29680         (cgraph_node::dump): Update.
29681         (cgraph_edge::maybe_hot_p): Update.
29682         * cgraph.h: Include profile-count.h
29683         (create_clone), create_edge, create_indirect_edge): Update.
29684         (cgraph_node): Turn count to profile_count.
29685         (cgraph_edge0: Likewise.
29686         (make_speculative, clone): Update.
29687         (create_edge): Update.
29688         (init_lowered_empty_function): Update.
29689         * cgraphclones.c (cgraph_edge::clone): Update.
29690         (duplicate_thunk_for_node): Update.
29691         (cgraph_node::create_clone): Update.
29692         * cgraphunit.c (cgraph_node::analyze): Update.
29693         (cgraph_node::expand_thunk): Update.
29694         * final.c (dump_basic_block_info): Update.
29695         * gimple-streamer-in.c (input_bb): Update.
29696         * gimple-streamer-out.c (output_bb): Update.
29697         * graphite.c (print_global_statistics): Update.
29698         (print_graphite_scop_statistics): Update.
29699         * hsa-brig.c: Include basic-block.h.
29700         * hsa-dump.c: Include basic-block.h.
29701         * hsa-gen.c (T sum_slice): Update.
29702         (convert_switch_statements):Update.
29703         * hsa-regalloc.c: Include basic-block.h.
29704         * ipa-chkp.c (chkp_produce_thunks): Update.
29705         * ipa-cp.c (struct caller_statistics): Update.
29706         (init_caller_stats): Update.
29707         (gather_caller_stats): Update.
29708         (ipcp_cloning_candidate_p): Update.
29709         (good_cloning_opportunity_p): Update.
29710         (get_info_about_necessary_edges): Update.
29711         (dump_profile_updates): Update.
29712         (update_profiling_info): Update.
29713         (update_specialized_profile): Update.
29714         (perhaps_add_new_callers): Update.
29715         (decide_about_value): Update.
29716         (ipa_cp_c_finalize): Update.
29717         * ipa-devirt.c (struct odr_type_warn_count): Update.
29718         (struct decl_warn_count): Update.
29719         (struct final_warning_record): Update.
29720         (possible_polymorphic_call_targets): Update.
29721         (ipa_devirt): Update.
29722         * ipa-fnsummary.c (redirect_to_unreachable): Update.
29723         * ipa-icf.c (sem_function::merge): Update.
29724         * ipa-inline-analysis.c (do_estimate_edge_time): Update.
29725         * ipa-inline.c (compute_uninlined_call_time): Update.
29726         (compute_inlined_call_time): Update.
29727         (want_inline_small_function_p): Update.
29728         (want_inline_self_recursive_call_p): Update.
29729         (edge_badness): Update.
29730         (lookup_recursive_calls): Update.
29731         (recursive_inlining): Update.
29732         (inline_small_functions): Update.
29733         (dump_overall_stats): Update.
29734         (dump_inline_stats): Update.
29735         * ipa-profile.c (ipa_profile_generate_summary): Update.
29736         (ipa_propagate_frequency): Update.
29737         (ipa_profile): Update.
29738         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
29739         * ipa-utils.c (ipa_merge_profiles): Update.
29740         * loop-doloop.c (doloop_modify): Update.
29741         * loop-unroll.c (report_unroll): Update.
29742         (unroll_loop_runtime_iterations): Update.
29743         * lto-cgraph.c (lto_output_edge): Update.
29744         (lto_output_node): Update.
29745         (input_node): Update.
29746         (input_edge): Update.
29747         (merge_profile_summaries): Update.
29748         * lto-streamer-in.c (input_cfg): Update.
29749         * lto-streamer-out.c (output_cfg): Update.
29750         * mcf.c (create_fixup_graph): Update.
29751         (adjust_cfg_counts): Update.
29752         (sum_edge_counts): Update.
29753         * modulo-sched.c (sms_schedule): Update.
29754         * postreload-gcse.c (eliminate_partially_redundant_load): Update.
29755         * predict.c (maybe_hot_count_p): Update.
29756         (probably_never_executed): Update.
29757         (dump_prediction): Update.
29758         (combine_predictions_for_bb): Update.
29759         (propagate_freq): Update.
29760         (handle_missing_profiles): Update.
29761         (counts_to_freqs): Update.
29762         (rebuild_frequencies): Update.
29763         (force_edge_cold): Update.
29764         * predict.h: Include profile-count.h
29765         (maybe_hot_count_p, counts_to_freqs): UPdate.
29766         * print-rtl-function.c: Do not include cfg.h
29767         * print-rtl.c: Include basic-block.h
29768         * profile-count.c: New file.
29769         * profile-count.h: New file.
29770         * profile.c (is_edge_inconsistent): Update.
29771         (correct_negative_edge_counts): Update.
29772         (is_inconsistent): Update.
29773         (set_bb_counts): Update.
29774         (read_profile_edge_counts): Update.
29775         (compute_frequency_overlap): Update.
29776         (compute_branch_probabilities): Update; Initialize and deinitialize
29777         gcov_count tables.
29778         (branch_prob): Update.
29779         * profile.h (bb_gcov_counts, edge_gcov_counts): New.
29780         (edge_gcov_count): New.
29781         (bb_gcov_count): New.
29782         * shrink-wrap.c (try_shrink_wrapping): Update.
29783         * tracer.c (better_p): Update.
29784         * trans-mem.c (expand_transaction): Update.
29785         (ipa_tm_insert_irr_call): Update.
29786         (ipa_tm_insert_gettmclone_call): Update.
29787         * tree-call-cdce.c: Update.
29788         * tree-cfg.c (gimple_duplicate_sese_region): Update.
29789         (gimple_duplicate_sese_tail): Update.
29790         (gimple_account_profile_record): Update.
29791         (execute_fixup_cfg): Update.
29792         * tree-inline.c (copy_bb): Update.
29793         (copy_edges_for_bb): Update.
29794         (initialize_cfun): Update.
29795         (freqs_to_counts): Update.
29796         (copy_cfg_body): Update.
29797         (expand_call_inline): Update.
29798         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
29799         * tree-ssa-loop-ivcanon.c (unloop_loops): Update.
29800         (try_unroll_loop_completely): Update.
29801         (try_peel_loop): Update.
29802         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
29803         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations_loop): Update.
29804         * tree-ssa-loop-split.c (connect_loops): Update.
29805         * tree-ssa-loop-unswitch.c (hoist_guard): Update.
29806         * tree-ssa-reassoc.c (branch_fixup): Update.
29807         * tree-ssa-tail-merge.c (replace_block_by): Update.
29808         * tree-ssa-threadupdate.c (create_block_for_threading): Update.
29809         (compute_path_counts): Update.
29810         (update_profile): Update.
29811         (recompute_probabilities): Update.
29812         (update_joiner_offpath_counts): Update.
29813         (estimated_freqs_path): Update.
29814         (freqs_to_counts_path): Update.
29815         (clear_counts_path): Update.
29816         (ssa_fix_duplicate_block_edges): Update.
29817         (duplicate_thread_path): Update.
29818         * tree-switch-conversion.c (case_bit_test_cmp): Update.
29819         (struct switch_conv_info): Update.
29820         * tree-tailcall.c (decrease_profile): Update.
29821         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
29822         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
29823         * value-prof.c (check_counter): Update.
29824         (gimple_divmod_fixed_value): Update.
29825         (gimple_mod_pow2): Update.
29826         (gimple_mod_subtract): Update.
29827         (gimple_ic_transform): Update.
29828         (gimple_stringop_fixed_value): Update.
29829         * value-prof.h (gimple_ic): Update.
29831 2017-06-02  Carl Love  <cel@us.ibm.com>
29833         * config/rs6000/rs6000-c: Add support for built-in functions
29834         vector double vec_doublee (vector signed int);
29835         vector double vec_doublee (vector unsigned int);
29836         vector double vec_doublee (vector float);
29837         vector double vec_doubleh (vector signed int);
29838         vector double vec_doubleh (vector unsigned int);
29839         vector double vec_doubleh (vector float);
29840         vector double vec_doublel (vector signed int);
29841         vector double vec_doublel (vector unsigned int);
29842         vector double vec_doublel (vector float);
29843         vector double vec_doubleo (vector signed int);
29844         vector double vec_doubleo (vector unsigned int);
29845         vector double vec_doubleo (vector float);.
29846         * config/rs6000/rs6000-builtin.def: Add definitions for DOUBLEE,
29847         DOUBLEO, DOUBLEH, DOUBLEL, UNS_DOUBLEO, UNS_DOUBLEE, UNS_DOUBLEH,
29848         UNS_DOUBLEL.
29849         * config/rs6000/altivec.md: Add code generator for doublee<mode>2,
29850         unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2, doubleh<mode>2,
29851         unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2, add mode attribute
29852         VS_sxwsp.
29853         * config/rs6000/altivec.h: Add define for vec_doublee, vec_doubleo,
29854         vec_doublel, vec_doubleh.
29855         * doc/extend.texi: Update the built-in documentation file for the
29856         new built-in functions.
29858 2017-06-02  David Malcolm  <dmalcolm@redhat.com>
29860         PR jit/80954
29861         * ipa-inline-analysis.c (free_growth_caches): Set
29862         edge_removal_hook_holder to NULL after removing it.
29864 2017-06-02  Sudakshina Das  <sudi.das@arm.com>
29866         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Return CC_SWP for
29867         comparision with zero.
29869 2017-06-02  Will Schmidt  <will_schmidt@vnet.ibm.com>
29870         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
29871         for early expansion of vec_min and vec_max builtins.
29872         (builtin_function_type): Add min/max unsigned variants to those
29873         identified as having unsigned arguments.
29875 2017-06-02  Olivier Hainque  <hainque@adacore.com>
29877         * config/vx-common.h (DWARF_UNWIND_INFO): Switch #define to 1.
29879 2017-06-02  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29881         * config/aarch64/aarch64-simd.md (*aarch64_simd_vec_copy_lane<mode>):
29882         Use VALL_F16 iterator rather than VALL.
29884 2017-06-02  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29886         * config/aarch64/aarch64.c (aarch64_split_compare_and_swap):
29887         Emit CBNZ inside loop when doing a strong exchange and comparing
29888         against zero.  Generate the CC flags after the loop.
29890 2017-06-02  David Edelsohn  <dje.gcc@gmail.com>
29892         * dwarf2out.c (DWARF_INITIAL_LENGTH_SIZE_STR): New.
29893         (dl_section_ref): New.
29894         (dwarf2out_finish): Copy debug_line_section_label to dl_section_ref.
29895         On AIX, append an expression to subtract the size of the
29896         section length to dl_section_ref.
29898 2017-06-02  Will Schmidt  <will_schmidt@vnet.ibm.com>
29900         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
29901         for early expansion of vector absolute builtins.
29903 2017-06-02  Richard Biener  <rguenther@suse.de>
29905         * tree-vect-slp.c (vect_detect_hybrid_slp_2): Match up
29906         what we consider a relevant use stmt with vect_detect_hybrid_slp_stmts.
29908 2017-06-02  Richard Biener  <rguenther@suse.de>
29910         PR tree-optimization/80948
29911         * tree-tailcall.c (find_tail_calls): Track stmts to move in
29912         stmt order as well.
29914 2017-06-02  Richard Biener  <rguenther@suse.de>
29916         * tree-vect-loop.c (vect_analyze_loop_operations): Not relevant
29917         PHIs are ok.
29918         * tree-vect-stmts.c (process_use): Do not mark backedge defs
29919         for inductions as relevant.
29921 2017-06-02  Richard Biener  <rguenther@suse.de>
29923         * tree-vect-loop.c (get_initial_def_for_induction): Inline into ...
29924         (vectorizable_induction): ... this.  Remove dead code.
29926 2017-06-02  Eric Botcazou  <ebotcazou@adacore.com>
29928         * builtins. (expand_builtin_alloca): Remove second parameter and
29929         infer its value from the first parameter instead.
29930         (expand_builtin) <BUILT_IN_ALLOCA>: Adjust call to above.
29932 2017-06-02  Jakub Jelinek  <jakub@redhat.com>
29934         PR rtl-optimization/80903
29935         * loop-doloop.c (add_test): Unshare sequence.
29937 2017-06-02  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29939         * doc/invoke.texi: Document the -Wsizeof-pointer-div warning.
29941 2017-06-01  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29943         * config/i386/i386.c (x86_64_ms_sysv_extra_clobbered_registers): Make
29944         static.
29945         (xlogue_layout::get_stack_space_used, xlogue_layout::s_instances,
29946         xlogue_layout::get_instance, logue_layout::xlogue_layout,
29947         sp_valid_at, fp_valid_at, choose_basereg): Formatting.
29948         (xlogue_layout::get_stub_rtx): Make static.
29949         (xlogue_layout::get_stub_name): Avoid const-cast, make static.
29950         (xlogue_layout::compute_stub_managed_regs): Rename to...
29951         (xlogue_layout::count_stub_managed_regs): ...this.
29952         (xlogue_layout::is_stub_managed_reg): New function.
29953         (xlogue_layout::m_stub_names): Rename to...
29954         (xlogue_layout::s_stub_names): ...this, make static.
29955         (xlogue_layout::STUB_INDEX_OFFSET, xlogue_layout::MIN_REGS,
29956         xlogue_layout::MAX_REGS, xlogue_layout::MAX_EXTRA_REGS,
29957         xlogue_layout::VARIANT_COUNT, xlogue_layout::STUB_NAME_MAX_LEN,
29958         xlogue_layout::s_stub_names): Instantiate statics.
29959         (stub_managed_regs): Remove.
29960         (ix86_save_reg): Use xlogue_layout::compute_stub_managed_regs.
29961         (disable_call_ms2sysv_xlogues): Rename to...
29962         (warn_once_call_ms2sysv_xlogues): ...this, and warn only once.
29963         (ix86_initial_elimination_offset, ix86_expand_call): Fix call_ms2sysv
29964         warning logic.
29965         (ix86_static_chain): Make sure that ix86_static_chain_on_stack can't
29966         change after reload_completed.
29967         (ix86_can_use_return_insn_p): Use the ix86_frame data structure
29968         directly.
29969         (ix86_expand_prologue): Likewise.
29970         (ix86_expand_epilogue): Likewise.
29971         (ix86_expand_split_stack_prologue): Likewise.
29972         (ix86_compute_frame_layout): Remove frame parameter ...
29973         (TARGET_COMPUTE_FRAME_LAYOUT): ... and export it as a target hook.
29974         (ix86_finalize_stack_realign_flags): Call ix86_compute_frame_layout
29975         only if necessary.
29976         (ix86_init_machine_status): Don't set use_fast_prologue_epilogue_nregs.
29977         (ix86_frame): Move from here ...
29978         * config/i386/i386.h (ix86_frame): ... to here.
29979         (machine_function): Remove use_fast_prologue_epilogue_nregs, cache the
29980         complete ix86_frame data structure instead.  Remove some_ld_name.
29982 2017-06-01  Pierre-Marie de Rodat  <derodat@adacore.com>
29984         * dwarf2out.c (dwarf2out_late_global_decl): Add locations for
29985         symbols that hold a DECL_VALUE_EXPR.
29987 2017-06-01  Martin Jambor  <mjambor@suse.cz>
29989         PR tree-optimization/80898
29990         * tree-sra.c (process_subtree_disqualification): Removed.
29991         (disqualify_candidate): Do not acll
29992         process_subtree_disqualification.
29993         (subtree_mark_written_and_enqueue): New function.
29994         (propagate_all_subaccesses): Set grp_write of LHS subtree if the
29995         RHS has been disqualified and re-queue LHS if necessary.  Apart
29996         from that, ignore disqualified RHS.
29998 2017-06-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
30000         * config/s390/s390.c (s390_emit_epilogue): Disable early return
30001         address fetch for z10 or later.
30003 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30005         * config/arc/arc.md (tst_movb): Add guard when splitting.
30007 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30009         * config/arc/arc.c (arc_can_eliminate): Test against
30010         arc_frame_pointer_needed.
30012 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30014         * config/arc/arc.c (arc_expand_prologue): Emit a special barrier
30015         to prevent store reordering.
30016         * config/arc/arc.md (UNSPEC_ARC_STKTIE): Define.
30017         (type): Add block type.
30018         (stack_tie): Define special instruction to be used in
30019         expand_prologue.
30021 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30023         * config/arc/arc.md (commutative_binary_comparison): Remove 'I'
30024         constraint. It is not valid for the pattern.
30025         (noncommutative_binary_comparison): Likewise.
30027 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30029         * config/arc/simdext.md (movv2hi_insn): Change predicate to avoid
30030         scaled addresses.
30032 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30034         * config/arc/arc.c (arc_conditional_register_usage): Allow r30 to
30035         be used by the reg-alloc.
30037 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30039         * config/arc/arc.md (mulsi3): Avoid use of hard registers before
30040         reg-alloc when having mul64 or mul32x16 instructions.
30041         (mulsidi3): Likewise.
30042         (umulsidi3): Likewise.
30043         (mulsi32x16): New pattern.
30044         (mulsi64): Likewise.
30045         (mulsidi64): Likewise.
30046         (umulsidi64): Likewise.
30047         (MUL32x16_REG): Define.
30048         (mul64_600): Use MUL32x16_REG.
30049         (mac64_600): Likewise.
30050         (umul64_600): Likewise.
30051         (umac64_600): Likewise.
30053 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30055         * config/arc/arc.md (mulsi3_700): Make it commutative.
30057 2017-06-01  Jose E. Marchesi  <jose.marchesi@oracle.com>
30059         * config/sparc/sparc.md (*zero_extendsidi2_insn_sp64): Set insn
30060         type for movstouw.
30061         (*sign_extendsidi2_insn): Likewise for movstosw.
30063 2017-06-01  Pierre-Marie de Rodat  <derodat@adacore.com>
30065         * dwarf2out.c (get_discr_value): Call the get_debug_type hook on
30066         the type of the input discriminant value.  Convert the
30067         discriminant value of signedness vary.
30069 2017-06-01  Volker Reichelt  <v.reichelt@netcologne.de>
30071         * doc/invoke.texi (-Wcatch-value): Document new shortcut.
30072         Add to -Wall section.
30074 2017-06-01  Richard Biener  <rguenther@suse.de>
30076         PR middle-end/66313
30077         * fold-const.c (fold_plusminus_mult_expr): If the factored
30078         factor may be zero use a wrapping type for the inner operation.
30079         * tree-tailcall.c (independent_of_stmt_p): Pass in to_move bitmap
30080         and handle moved defs.
30081         (process_assignment): Properly guard the unary op case.  Return a
30082         tri-state indicating that moving the stmt before the call may allow
30083         to continue.  Pass through to_move.
30084         (find_tail_calls): Handle moving unrelated defs before
30085         the call.
30087 2017-05-31  Segher Boessenkool  <segher@kernel.crashing.org>
30089         PR target/80618
30090         * config/rs6000/vector.md (*vector_uneq<mode>): Write the nor in the
30091         splitter result in the canonical way.
30093 2017-05-31  Uros Bizjak  <ubizjak@gmail.com>
30095         * config/i386/i386.md (*zero_extendsidi2): Enable alternative (?r, *Yj)
30096         also for 32bit target.  Update insn attributes.
30097         (zero-extendsidi2 splitter): Allow all registers for operand 1.
30099 2017-05-31  Sebastian Peryt  <sebastian.peryt@intel.com>
30101         * config/i386/avx512fintrin.h (_mm_mask_max_sd)
30102         (_mm_maskz_max_sd, _mm_mask_max_ss, _mm_maskz_max_ss)
30103         (_mm_mask_min_sd, _mm_maskz_min_sd, _mm_mask_min_ss)
30104         (_mm_maskz_min_ss): New intrinsics.
30106 2017-05-31  Martin Liska  <mliska@suse.cz>
30108         * tree-vect-loop.c (vect_create_epilog_for_reduction):
30109         Change comment style to one we normally use.
30110         (vectorizable_reduction): Likewise.
30111         (vectorizable_induction): Likewise.
30112         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
30113         (vectorizable_call): Likewise.
30114         (vectorizable_simd_clone_call): Likewise.
30115         (vectorizable_conversion): Likewise.
30116         (vectorizable_assignment): Likewise.
30117         (vectorizable_shift): Likewise.
30118         (vectorizable_operation): Likewise.
30119         (vectorizable_store): Likewise.
30120         (vectorizable_load): Likewise.
30121         * tree-vectorizer.h: Likewise.
30123 2017-05-31  Alexander Monakov  <amonakov@ispras.ru>
30125         * passes.c (emergency_dump_function): New.
30126         * tree-pass.h (emergency_dump_function): Declare.
30127         * plugin.c (plugins_internal_error_function): Remove.
30128         * plugin.h (plugins_internal_error_function): Remove declaration.
30129         * toplev.c (internal_error_function): New static function.  Use it...
30130         (general_init): ...here.
30132 2017-05-31  Graham Markall  <graham.markall@embecosm.com>
30134         * config/arc/arc.c (arc_print_operand): Handle constant operands.
30135         (arc_rtx_costs): Add costs for new patterns.
30136         * config/arc/arc.md: Additional *add_n and *sub_n patterns.
30137         * config/arc/predicates.md: Add _1_2_3_operand predicate.
30139 2017-05-31  Richard Sandiford  <richard.sandiford@linaro.org>
30141         * tree-ssa-strlen.c (get_next_strinfo): New function.
30142         (get_stridx_plus_constant): Use it.
30143         (zero_length_string): Likewise.
30144         (adjust_related_strinfos): Likewise.
30145         (adjust_last_stmt): Likewise.
30147 2017-05-31  Richard Biener  <rguenther@suse.de>
30149         PR target/80880
30150         * config/i386/i386.c (ix86_expand_builtin): Remove assert
30151         for arg being an SSA name when expanding IX86_BUILTIN_BNDRET.
30153 2017-05-31  Richard Sandiford  <richard.sandiford@linaro.org>
30155         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Remove
30156         loop_vinfo argument and use of dependence distance vectors.
30157         Check instead whether the two references differ only in their
30158         initial value and assume that they have the same alignment if the
30159         difference is a multiple of the vector alignment.
30160         (vect_analyze_data_refs_alignment): Update call accordingly.
30162 2017-05-31  Martin Liska  <mliska@suse.cz>
30164         PR target/79155
30165         * config/i386/cpuid.h: Fix typo in a comment in cpuid.h.
30167 2017-05-31  Bin Cheng  <bin.cheng@arm.com>
30169         * tree-vect-loop-manip.c (create_intersect_range_checks_index)
30170         (create_intersect_range_checks): Move from ...
30171         * tree-data-ref.c (create_intersect_range_checks_index)
30172         (create_intersect_range_checks): ... to here.
30173         (create_runtime_alias_checks): New function factored from ...
30174         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks): ...
30175         here.  Call above function.
30176         * tree-data-ref.h (create_runtime_alias_checks): New function.
30178 2017-05-31  Bin Cheng  <bin.cheng@arm.com>
30180         * tree-data-ref.c (prune_runtime_alias_test_list): Relax minimal
30181         segment length for dr_b and compute it in wide_int.
30183 2017-05-31  Richard Biener  <rguenther@suse.de>
30185         PR tree-optimization/80906
30186         * graphite-isl-ast-to-gimple.c (copy_loop_close_phi_nodes): Get
30187         and pass through iv_map.
30188         (copy_bb_and_scalar_dependences): Adjust.
30189         (translate_pending_phi_nodes): Likewise.
30190         (copy_loop_close_phi_args): Handle code-generating IVs instead
30191         of ICEing.
30193 2017-05-30  David Malcolm  <dmalcolm@redhat.com>
30195         * diagnostic-color.c (color_dict): Add "type-diff".
30196         (parse_gcc_colors): Update comment.
30197         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
30198         -fdiagnostics-show-template-tree and -fno-elide-type.
30199         (GCC_COLORS): Add type-diff to example.
30200         (type-diff=): New.
30201         (-fdiagnostics-show-template-tree): New.
30202         (-fno-elide-type): New.
30203         * pretty-print.c (pp_format): Pass quote and formatters[argno] to
30204         the pp_format_decoder callback.  Call any m_format_postprocessor's
30205         "handle" method.
30206         (pretty_printer::pretty_printer): Initialize
30207         m_format_postprocessor.
30208         (pretty_printer::~pretty_printer): Delete any
30209         m_format_postprocessor.
30210         * pretty-print.h (printer_fn): Add bool and const char ** parameters.
30211         (class format_postprocessor): New class.
30212         (struct pretty_printer::format_decoder): Document the new parameters.
30213         (struct pretty_printer::m_format_postprocessor): New field.
30214         * tree-diagnostic.c (default_tree_printer): Update for new
30215         bool and const char ** params.
30216         * tree-diagnostic.h (default_tree_printer): Likewise.
30218 2017-05-30  Segher Boessenkool  <segher@kernel.crashing.org>
30220         * config/rs6000/predicates.md (cc_reg_not_micro_cr0_operand): Delete.
30221         (lwa_operand): Delete rs6000_gen_cell_microcode test.
30222         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
30223         rs6000_gen_cell_microcode code.
30224         (rs6000_final_prescan_insn): Delete.
30225         (rs6000_opt_vars): Delete the "gen-cell-microcode" and
30226         "warn-cell-microcode" entries.
30227         * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Delete.
30228         * config/rs6000/rs6000.md: Delete rs6000_gen_cell_microcode tests
30229         throughout.  Change cc_reg_not_micro_cr0_operand to
30230         cc_reg_not_cr0_operand throughout.
30231         (*extendhi<mode>2_noload): Delete.
30232         * config/rs6000/rs6000.opt (mgen-cell-microcode): Replace by stub.
30233         (mwarn-cell-microcode): Delete.
30234         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete
30235         -mgen-cell-microcode and -mwarn-cell-microcode.
30237 2017-05-30  Uros Bizjak  <ubizjak@gmail.com>
30239         PR target/80833
30240         * config/i386/constraints.md (Yd): New constraint.
30241         (Ye): Ditto.
30242         * config/i386/i386.md (*movti_internal): Add (?r, Ye)
30243         and (?Yd, r) alternatives.  Update insn attributes.
30244         * config/i386/i386.md (*movti_internal): Add (?r, *Ye)
30245         and (?*Yd, r) alternatives.  Update insn attributes.
30246         (double-mode inter-unit splitters): Add new GR<->XMM splitters.
30248 2017-05-30  Pierre-Marie de Rodat  <derodat@adacore.com>
30250         * gimplify.c (gimplify_modify_expr): Don't create a
30251         DECL_DEBUG_EXPR link if *FROM_P does not belong to the current
30252         function.
30254 2017-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
30256         * config/arm/arm-builtins.c (arm_expand_builtin): Remove const.
30258 2017-05-30  Richard Biener  <rguenther@suse.de>
30260         * tree-vectorizer.h (struct _stmt_vec_info): Add reduc_type
30261         and reduc_def fields.
30262         (STMT_VINFO_REDUC_TYPE): New define.
30263         (STMT_VINFO_REDUC_DEF): Likewise.
30264         (vect_force_simple_reduction): Adjust prototype.
30265         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Adjust.
30266         (vect_is_simple_reduction): Remove check_reduction argument.
30267         (vect_force_simple_reduction): Adjust and set
30268         STMT_VINFO_REDUC_TYPE and STMT_VINFO_REDUC_DEF.
30269         (vectorizable_reduction): Do not re-do reduction analysis
30270         but use STMT_VINFO_REDUC_TYPE and STMT_VINFO_REDUC_DEF.
30271         * tree-parloops.c (gather_scalar_reductions): Adjust.
30273 2017-05-30  Richard Biener  <rguenther@suse.de>
30275         PR middle-end/80901
30276         * cfgexpand.c (expand_gimple_cond): Match up loop fixup with
30277         split_edge code.
30279 2017-05-24  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30281         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs):
30282         Introduce unknown_misalignment parameter and remove vf.
30283         (vect_peeling_hash_get_lowest_cost):
30284         Pass unknown_misalignment parameter.
30285         (vect_enhance_data_refs_alignment):
30286         Fix unsupportable data ref treatment.
30288 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30290         * tree-vect-data-refs.c (vect_get_data_access_cost):
30291         Workaround for SLP handling.
30292         (vect_enhance_data_refs_alignment):
30293         Compute costs for doing no peeling at all, compare to the best
30294         peeling costs so far and avoid peeling if cheaper.
30296 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30298         * tree-vect-data-refs.c (vect_peeling_hash_choose_best_peeling):
30299         Return peeling info and set costs to zero for unlimited cost
30300         model.
30301         (vect_enhance_data_refs_alignment): Also inspect all datarefs
30302         with unknown misalignment. Compute and costs for unknown
30303         misalignment, compare them to the costs for known misalignment
30304         and choose the cheapest for peeling.
30306 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30308         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Rename.
30309         (vect_get_peeling_costs_all_drs): Create function.
30310         (vect_peeling_hash_get_lowest_cost):
30311         Use vect_get_peeling_costs_all_drs.
30312         (vect_peeling_supportable): Create function.
30313         (vect_enhance_data_refs_alignment): Use vect_peeling_supportable.
30315 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30317         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Create
30318         DR_HAS_NEGATIVE_STEP.
30319         (vect_update_misalignment_for_peel): Define DR_MISALIGNMENT.
30320         (vect_enhance_data_refs_alignment): Use.
30321         (vect_duplicate_ssa_name_ptr_info): Use.
30322         * tree-vectorizer.h (dr_misalignment): Use.
30323         (known_alignment_for_access_p): Use.
30325 2017-05-30  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
30327         PR target/78838
30328         * config/msp430/msp430.c (gen_prefix): Return NULL when section name is
30329         .lowtext.
30330         (has_section_name): New function.
30332 2017-05-30  Martin Liska  <mliska@suse.cz>
30334         PR other/80909
30335         * auto-profile.c (get_function_decl_from_block): Fix
30336         parenthesis.
30338 2017-05-30  Richard Biener  <rguenther@suse.de>
30340         PR middle-end/80876
30341         * cfgexpand.c (expand_gimple_cond): Fixup preserving loops again.
30343 2017-05-30  Martin Liska  <mliska@suse.cz>
30345         * dumpfile.c: Use newly added macro DUMP_FILE_INFO.
30346         * dumpfile.h (struct dump_file_info): Remove ctors.
30348 2017-05-30  Martin Liska  <mliska@suse.cz>
30350         * predict.def: Fix GNU coding style.
30352 2017-05-29  Max Filippov  <jcmvbkbc@gmail.com>
30354         * config/xtensa/xtensa.c (xtensa_initial_elimination_offset):
30355         Mark 'to' argument with ATTRIBUTE_UNUSED.
30357 2017-05-29  Max Filippov  <jcmvbkbc@gmail.com>
30359         * config/xtensa/xtensa.c (xtensa_emit_call): Use
30360         HOST_WIDE_INT_PRINT_HEX instead of 0x%lx format string.
30361         (print_operand): Use HOST_WIDE_INT_PRINT_DEC instead of %ld
30362         format string.
30364 2017-05-29  Eric Botcazou  <ebotcazou@adacore.com>
30366         * doc/install.texi (Options specification): Restore entry of
30367         --enable-sjlj-exceptions.
30369 2017-05-27  Michael Eager  <eager@eagercon.com>
30371         Revert:
30372         2016-01-21  Ajit Agarwal  <ajitkum@xilinx.com>
30374         See https://gcc.gnu.org/ml/gcc/2017-05/msg00221.html.
30376         * config/microblaze/microblaze.h
30377         (FIXED_REGISTERS): Update in macro.
30378         (CALL_USED_REGISTERS): Update in macro.
30380 2017-05-27  François-Xavier Coudett  <fxcoudert@gcc.gnu.org>
30382         * doc/install.texi: Add links to macOS binary distributions.
30384 2017-05-27  Jakub Jelinek  <jakub@redhat.com>
30386         PR bootstrap/80887
30387         Revert:
30388         2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30390         * match.pd ((A +- CST1) +- CST2): Allow some conversions.
30392 2017-05-26  Martin Liska  <mliska@suse.cz>
30394         * dumpfile.h (enum dump_kind): Renumber TDF_* flags to be contiguous.
30396 2017-05-26  Martin Liska  <mliska@suse.cz>
30398         * cfg.c (check_bb_profile): Do not use TDF_COMMENT and print
30399         always leading ';; '.
30400         (dump_bb_info): Likewise.
30401         (brief_dump_cfg): Likewise.
30402         * cfgrtl.c (print_rtl_with_bb): Do not use TDF_COMMENT.
30403         * dumpfile.c: Remove usage of TDF_VERBOSE.
30404         * dumpfile.h (enum dump_kind): Likewise.
30405         (dump_gimple_bb_header): Do not use TDF_COMMENT.
30406         * print-tree.c (debug_verbose): Remove.
30407         * tree-cfg.c (gimple_dump_cfg): Do not use TDF_COMMENT.
30408         (dump_function_to_file): Remove dumps guarded with TDF_VERBOSE.
30409         * tree-diagnostic.c (default_tree_printer): Replace
30410         TDF_DIAGNOSTIC with TDF_SLIM.
30412 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30414         * tree-vect-loop-manip.c (create_intersect_range_checks_index): Pass
30415         in parameter loop, rather than loop_vinfo.
30416         (create_intersect_range_checks): Ditto.
30417         (vect_create_cond_for_alias_checks): Update call to above functions.
30419 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30421         PR tree-optimization/80815
30422         * tree-data-ref.c (prune_runtime_alias_test_list): Simplify condition
30423         for merging runtime alias checks.  Handle negative DR_STEPs.
30425 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30427         * tree-vect-data-refs.c (Operator==, comp_dr_with_seg_len_pair):
30428         Move from ...
30429         * tree-data-ref.c (Operator==, comp_dr_with_seg_len_pair): To here.
30430         * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list): Factor
30431         out code pruning runtime alias checks.
30432         * tree-data-ref.c (prune_runtime_alias_test_list): New function
30433         factored out from above.
30434         * tree-vectorizer.h (struct dr_with_seg_len, dr_with_seg_len_pair_t):
30435         Move from ...
30436         * tree-data-ref.h (struct dr_with_seg_len, dr_with_seg_len_pair_t):
30437         ... to here.
30438         (prune_runtime_alias_test_list): New decalaration.
30440 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30442         * tree-vect-data-refs.c (compare_tree): Rename and move ...
30443         * tree-data-ref.c (data_ref_compare_tree): ... to here.
30444         * tree-data-ref.h (data_ref_compare_tree): New decalaration.
30445         * tree-vect-data-refs.c (dr_group_sort_cmp): Update uses.
30446         (operator==, comp_dr_with_seg_len_pair): Ditto.
30447         (vect_prune_runtime_alias_test_list): Ditto.
30449 2017-05-26  Martin Liska  <mliska@suse.cz>
30451         PR ipa/80663
30452         * params.def: Bound partial-inlining-entry-probability param.
30454 2017-05-26  Marek Polacek  <polacek@redhat.com>
30456         PR sanitizer/80875
30457         * fold-const.c (fold_binary_loc) <case MULT_EXPR>: Check if OP1
30458         can be negated.
30460 2017-05-26  Richard Biener  <rguenther@suse.de>
30462         PR tree-optimization/80842
30463         * tree-ssa-ccp.c (set_lattice_value): Always meet with the old
30464         value.
30466 2017-05-26  Richard Biener  <rguenther@suse.de>
30468         PR tree-optimization/80844
30469         * tree-vectorizer.c (adjust_simduid_builtins): Propagate results.
30471 2017-05-25  Sebastian Peryt  <sebastian.peryt@intel.com>
30473         * doc/md.texi (Machine Constraints): Update x86 family
30474         machine constraints section to match 'config/i386/constraints.md'.
30476 2017-05-25  Volker Reichelt  <v.reichelt@netcologne.de>
30478         * doc/invoke.texi (-Wcatch-value=): Document new warning option.
30480 2017-05-25  Nathan Sidwell  <nathan@acm.org>
30482         * doc/invoke.texi (--enable-languages): Update documentation.
30484 2017-05-25  Martin Liska  <mliska@suse.cz>
30486         * dumpfile.c: Add TDF_FOLDING.
30487         * dumpfile.h (enum dump_kind): Likewise.
30488         * genmatch.c (dt_simplify::gen_1): Use it.
30490 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30492         * match.pd (view_convert (convert@0 @1)): Handle zero-extension.
30494 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30496         * match.pd ((A +- CST1) +- CST2): Allow some conversions.
30497         * tree.c (drop_tree_overflow): Handle COMPLEX_CST and VECTOR_CST.
30499 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30501         * fold-const.c (fold_binary_loc) [(A & C) == D]: Remove transformation.
30502         * match.pd (X == C): Rewrite it here.
30503         (with_possible_nonzero_bits, with_possible_nonzero_bits2,
30504         with_certain_nonzero_bits2): New predicates.
30505         * tree-ssanames.c (get_nonzero_bits): Handle INTEGER_CST.
30507 2017-05-24  Nathan Sidwell  <nathan@acm.org>
30509         * lto-streamer-in.c (lto_input_data_block): Adjust T const cast to
30510         avoid warning.
30512         * auto-profile.c (afdo_propagate): Adjust T const cast to avoid
30513         warning.
30515 2017-05-24  Segher Boessenkool  <segher@kernel.crashing.org>
30517         * config/powerpcspe: New port.  Files are copied from the rs6000
30518         port, with "rs6000" in filenames replaced by "powerpcspe".
30520 2017-05-24  Wilco Dijkstra  <wdijkstr@arm.com>
30522         PR rtl-optimization/80754
30523         * lra-remat.c (do_remat): Add overlap checks for dst_regno.
30525 2017-05-24  Sheldon Lobo  <smlobo@sheldon.us.oracle.com>
30527         * config/sparc/sparc.md (length): Return the correct value for -mflat
30528         sibcalls to match output_sibcall.
30530 2017-05-24  Segher Boessenkool  <segher@kernel.crashing.org>
30532         PR bootstrap/80860
30533         PR bootstrap/80843
30534         * config/rs6000/rs6000.c (struct machine_function): Add new field
30535         n_components.
30536         (rs6000_get_separate_components): Init that field, use it.
30537         (rs6000_components_for_bb): Use the field.
30539 2017-05-24  Thomas Preud'homme  <thomas.preudhomme@arm.com>
30541         * config/arm/arm.c (arm_expand_prologue): Fix typo in comment.
30543 2017-05-24  Peter Bergner  <bergner@vnet.ibm.com>
30545         PR middle-end/80823
30546         * tree-cfg.c (group_case_labels_stmt): Delete increment of "i";
30548 2017-05-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
30550         PR target/80725
30551         * config/s390/s390.c (s390_check_qrst_address): Check incoming
30552         address against address_operand predicate.
30553         * config/s390/s390.md ("*indirect_jump"): Swap alternatives.
30555 2017-05-24  Eric Botcazou  <ebotcazou@adacore.com>
30557         * var-tracking.c (track_expr_p): Do not return 0 for tracked record
30558         parameters passed indirectly.
30560 2017-05-23  Uros Bizjak  <ubizjak@gmail.com>
30562         * config/i386/i386.md (*movdi_internal): Remove SSE4
30563         alternative 18 (?r, *v).  Update insn attributes.
30564         (*movsi_internal): Remove SSE4 alternative 13 (?r, *v).
30565         Update insn attributes.
30566         (*zero_extendsidi2): Remove SSE4 alternative (?r, *x).
30567         Update insn attributes.
30568         * config/i386/sse.md (vec_extract<ssevecmodelower>_0): Remove SSE4
30569         alternative 1 (r, v). Remove isa attribute.
30570         * config/i386/i386.c (dimode_scalar_chain::make_vector_copies):
30571         Always move value through stack for !TARGET_INTER_UNIT_MOVES_TO_VEC
30572         and !TARGET_INTER_UNIT_MOVES_TO_VEC targets.
30574 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30576         * doc/sourcebuild.texi (Directives, Verify compiler message): Document
30577         dg-line directive.
30579 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30581         * cgraphunit.c (symbol_table::process_new_functions): Update.
30582         * ipa-fnsummary.c (pass_data_inline_parameters): Remove.
30583         (inline_generate_summary): Rename to ...
30584         (ipa_fn_summary_generate): ... this one.
30585         (inline_read_summary): Rename to ...
30586         (ipa_fn_summary_read): ... this one.
30587         (inline_write_summary): Rename to ...
30588         (ipa_fn_summary_write): ... this one.
30589         (inline_free_summary): Rename to ...
30590         (ipa_free_fn_summary): ... this one.
30591         (pass_data_local_fn_summary, pass_local_fn_summary,
30592         make_pass_local_fn_summary, pass_data_ipa_free_fn_summary,
30593         pass_ipa_free_fn_summary, make_pass_ipa_free_fn_summary,
30594         pass_data_ipa_fn_summary, pass_ipa_fn_summary,
30595         make_pass_ipa_fn_summary): New.
30596         * ipa-fnsummary.h (inline_generate_summary, inline_read_summary,
30597         inline_write_summary, inline_free_summary): Remove.
30598         (ipa_free_fn_summary) : New.
30599         * ipa-inline.c (ipa_inline): Update.
30600         (pass_ipa_inline): Do not generate summaries.
30601         * ipa.c (pass_data_ipa_free_fn_summary, pass_ipa_free_fn_summary):
30602         Remove.
30603         * passes.def: Replace pass_inline_parameters by pass_local_fn_summary
30604         and add pass_ipa_fn_summary.
30605         * tree-pass.h (make_pass_ipa_fn_summary, make_pass_local_fn_summary):
30606         New.
30607         (make_pass_inline_parameters): Remove.
30609 2017-05-23  Thomas Schwinge  <thomas@codesourcery.com>
30611         * omp-low.c (struct omp_context): Remove "default_kind" member.
30612         Adjust all users.
30614         * omp-offload.c (execute_oacc_device_lower): Remove the
30615         parallelism dimensions function attributes for unparallelized
30616         OpenACC kernels constructs.
30618 2017-05-23  Martin Liska  <mliska@suse.cz>
30620         * cgraph.c (cgraph_node::get_create): Use symtab_node::dump_{asm_,}name
30621         functions.
30622         (cgraph_edge::make_speculative): Likewise.
30623         (cgraph_edge::resolve_speculation): Likewise.
30624         (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
30625         (cgraph_node::dump): Likewise.
30626         * cgraph.h: Likewise.
30627         * cgraphunit.c (analyze_functions): Likewise.
30628         (symbol_table::compile): Likewise.
30629         * ipa-cp.c (print_all_lattices): Likewise.
30630         (determine_versionability): Likewise.
30631         (initialize_node_lattices): Likewise.
30632         (ipcp_verify_propagated_values): Likewise.
30633         (estimate_local_effects): Likewise.
30634         (update_profiling_info): Likewise.
30635         (create_specialized_node): Likewise.
30636         (perhaps_add_new_callers): Likewise.
30637         (decide_about_value): Likewise.
30638         (decide_whether_version_node): Likewise.
30639         (identify_dead_nodes): Likewise.
30640         (ipcp_store_bits_results): Likewise.
30641         * ipa-devirt.c (dump_targets): Likewise.
30642         (ipa_devirt): Likewise.
30643         * ipa-icf.c (sem_item::dump): Likewise.
30644         (sem_function::equals): Likewise.
30645         (sem_variable::equals): Likewise.
30646         (sem_item_optimizer::read_section): Likewise.
30647         (sem_item_optimizer::execute): Likewise.
30648         (congruence_class::dump): Likewise.
30649         * ipa-inline-analysis.c (dump_ipa_call_summary): Likewise.
30650         (dump_inline_summary): Likewise.
30651         (estimate_node_size_and_time): Likewise.
30652         (inline_analyze_function): Likewise.
30653         * ipa-inline-transform.c (inline_call): Likewise.
30654         * ipa-inline.c (report_inline_failed_reason): Likewise.
30655         (want_early_inline_function_p): Likewise.
30656         (edge_badness): Likewise.
30657         (update_edge_key): Likewise.
30658         (inline_small_functions): Likewise.
30659         * ipa-profile.c (ipa_profile): Likewise.
30660         * ipa-prop.c (ipa_print_node_jump_functions): Likewise.
30661         (ipa_make_edge_direct_to_target): Likewise.
30662         (remove_described_reference): Likewise.
30663         (ipa_impossible_devirt_target): Likewise.
30664         (propagate_controlled_uses): Likewise.
30665         (ipa_print_node_params): Likewise.
30666         (ipcp_transform_function): Likewise.
30667         * ipa-pure-const.c (pure_const_read_summary): Likewise.
30668         (propagate_pure_const): Likewise.
30669         * ipa-reference.c (generate_summary): Likewise.
30670         (read_write_all_from_decl): Likewise.
30671         (propagate): Likewise.
30672         (ipa_reference_read_optimization_summary): Likewise.
30673         * ipa-utils.c (ipa_merge_profiles): Likewise.
30674         * ipa.c (walk_polymorphic_call_targets): Likewise.
30675         (symbol_table::remove_unreachable_nodes): Likewise.
30676         (ipa_single_use): Likewise.
30677         * passes.c (execute_todo): Likewise.
30678         * predict.c (drop_profile): Likewise.
30679         * symtab.c (symtab_node::get_dump_name): New function.
30680         (symtab_node::dump_name): Likewise.
30681         (symtab_node::dump_asm_name): Likewise.
30682         (symtab_node::dump_references): Likewise.
30683         (symtab_node::dump_referring): Likewise.
30684         (symtab_node::dump_base): Likewise.
30685         (symtab_node::debug_symtab): Likewise.
30686         * tree-sra.c (convert_callers_for_node): Likewise.
30687         * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
30688         * value-prof.c (init_node_map): Likewise.
30690 2017-05-23  Martin Liska  <mliska@suse.cz>
30692         * cgraph.h: Move symtab_node::dump_table to symbol_table::dump
30693         and symtab_node::debug_symtab to symbol_table::debug.
30694         * cgraphunit.c (analyze_functions): Use the renamed function.
30695         (symbol_table::compile): Likewise.
30696         * ipa-cp.c (ipcp_verify_propagated_values): Likewise.
30697         * ipa-icf.c (sem_item_optimizer::execute): Likewise.
30698         * passes.c (execute_todo): Likewise.
30699         * symtab.c (symbol_table::dump): New function.
30700         * tree-ssa-structalias.c (ipa_pta_execute): Use the renamed function.
30702 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30704         * ipa-fnsummary.c (estimate_node_size_and_time): Do not sanity check
30705         that nonconst implies exec.
30707 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30709         * ipa-inline-analysis.c (cgraph_2edge_hook_list, cgraph_edge_hook_list,
30710         inline_edge_removal_hook, inline_edge_duplication_hook): Remove.
30711         (inline_edge_summary_vec): Turn into ...
30712         (ipa_call_summaries): ... this one.
30713         (redirect_to_unreachable, edge_set_predicate,
30714         evaluate_properties_for_edge, inline_summary_alloc,
30715         reset_ipa_call_summary, reset_inline_summary,
30716         inline_summary_t::duplicate): Update.
30717         (inline_edge_duplication_hook): Turn to ...
30718         (ipa_call_summary_t::duplicate): ... this one.
30719         (inline_edge_removal_hook): Turn to ...
30720         (ipa_call_summary_t::remove): ... this one.
30721         (dump_inline_edge_summary): Turn to ...
30722         (dump_ipa_call_summary): ... this one.
30723         (estimate_function_body_sizes): Update.
30724         (inline_update_callee_summaries): Update.
30725         (remap_edge_change_prob): Update.
30726         (remap_edge_summaries): Update.
30727         (inline_merge_summary): Update.
30728         (do_estimate_edge_time): Update.
30729         (inline_generate_summary): Update.
30730         (inline_read_section): Update.
30731         (inline_read_summary): Update.
30732         (inline_free_summary): Update.
30733         * ipa-inline.c (can_inline_edge_p): Update.
30734         (compute_inlined_call_time): Update.
30735         (want_inline_small_function_p): Update.
30736         (edge_badness): Update.
30737         (early_inliner): Update.
30738         * ipa-inline.h (inline_edge_summary): Turn to ...
30739         (ipa_call_summary): ... this one.
30740         (ipa_call_summary_t): New class.
30741         (inline_edge_summary_t, inline_edge_summary_vec): Remove.
30742         (ipa_call_summaries): New.
30743         (inline_edge_summary): Remove.
30744         (estimate_edge_growth): Update.
30745         * ipa-profile.c (ipa_propagate_frequency_1): Update.
30746         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
30747         * ipa-split.c (execute_split_functions): Update.
30748         * ipa.c (symbol_table::remove_unreachable_nodes): Update.
30750 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30752         * doc/sourcebuild.texi (Effective-Target Keywords, Other hardware
30753         attributes): Document rdrand effective target.
30755 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30757         * doc/sourcebuild.texi (Effective-Target Keywords, Other hardware
30758         attributes): Sort alphabetically.
30760 2017-05-23  Georg-Johann Lay  <avr@gjlay.de>
30762         * config/avr/genmultilib.awk: Use gsub instead of gensub.
30764 2017-05-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
30766         PR target/80718
30767         * config/rs6000/vsx.md (vsx_splat_<mode>, VSX_D iterator): Split
30768         V2DF/V2DI splat into two separate patterns, one that handles
30769         registers, and the other that only handles memory.  Drop support
30770         for splatting from a GPR on ISA 2.07 and then splitting the
30771         splat into direct move and splat.
30772         (vsx_splat_<mode>_reg): Likewise.
30773         (vsx_splat_<mode>_mem): Likewise.
30775 2017-05-22  Segher Boessenkool  <segher@kernel.crashing.org>
30777         * cfgcleanup.c (bb_is_just_return): Allow CLOBBERs.
30779 2017-05-22  Jakub Jelinek  <jakub@redhat.com>
30781         PR middle-end/80809
30782         * omp-low.c (finish_taskreg_remap): New function.
30783         (finish_taskreg_scan): If unit size of ctx->record_type
30784         is non-constant, unshare the size expression and replace
30785         decls in it with possible outer var refs.
30787         PR middle-end/80809
30788         * gimplify.c (omp_add_variable): For GOVD_DEBUG_PRIVATE use
30789         GOVD_SHARED rather than GOVD_PRIVATE with it.
30790         (gimplify_adjust_omp_clauses_1, gimplify_adjust_omp_clauses): Expect
30791         GOVD_SHARED rather than GOVD_PRIVATE with GOVD_DEBUG_PRIVATE.
30793         PR middle-end/80853
30794         * omp-low.c (lower_reduction_clauses): Pass OMP_CLAUSE_PRIVATE
30795         as last argument to build_outer_var_ref for pointer bases of array
30796         section reductions.
30798 2017-05-19  Martin Sebor  <msebor@redhat.com>
30800         * print-tree.c (print_node): Print DECL_READ_P flag.
30802 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30804         * Makefile.in: Add ipa-fnsummary.o and ipa-fnsummary.h
30805         * auto-profile.c: Replace ipa-inline.h by ipa-fnsummary.h
30806         * cgraph.c: Likewise.
30807         * cgraphunit.c: Likewise.
30808         * gengtype.c: Likewise.
30809         * ipa-cp.c: Likewise.
30810         * ipa-devirt.c: Likewise.
30811         * ipa-icf.c: Likewise.
30812         * ipa-predicate.c: Likewise.
30813         * ipa-profile.c: Likewise.
30814         * ipa-prop.c: Likewise.
30815         * ipa-split.c: Likewise.
30816         * ipa.c: Likewise.
30817         * ipa-inline-analysis.c (inline_summaries, ipa_call_summaries,
30818         edge_predicate_pool, dump_inline_hints,
30819         inline_summary::account_size_time, redirect_to_unreachable,
30820         edge_set_predicate, set_hint_predicate,
30821         evaluate_conditions_for_known_args, evaluate_properties_for_edge,
30822         inline_summary_alloc, ipa_call_summary::reset, inline_summary::reset,
30823         inline_summary_t::remove, remap_hint_predicate_after_duplication,
30824         inline_summary_t::duplicate, ipa_call_summary_t::duplicate,
30825         ipa_call_summary_t::remove, initialize_growth_caches,
30826         free_growth_caches, dump_ipa_call_summary, dump_inline_summary,
30827         debug_inline_summary, dump_inline_summaries, initialize_inline_failed,
30828         mark_modified, unmodified_parm_1, unmodified_parm,
30829         unmodified_parm_or_parm_agg_item, eliminated_by_inlining_prob,
30830         set_cond_stmt_execution_predicate, set_switch_stmt_execution_predicate,
30831         compute_bb_predicates, will_be_nonconstant_expr_predicate,
30832         will_be_nonconstant_predicate, record_modified_bb_info,
30833         get_minimal_bb, record_modified, param_change_prob,
30834         phi_result_unknown_predicate, predicate_for_phi_result,
30835         array_index_predicate, clobber_only_eh_bb_p, fp_expression_p,
30836         estimate_function_body_sizes, compute_inline_parameters,
30837         compute_inline_parameters_for_curren, pass_data_inline_parameters,
30838         estimate_node_size_and_time, estimate_ipcp_clone_size_and_time,
30839         inline_update_callee_summaries, remap_edge_change_prob,
30840         remap_edge_summaries, remap_hint_predicate, inline_merge_summary,
30841         inline_update_overall_summary, inline_indirect_intraprocedural_analysis,
30842         inline_analyze_function, inline_summary_t::insert,
30843         inline_generate_summary, read_ipa_call_summary, inline_read_section,
30844         inline_read_summary, write_ipa_call_summary, inline_write_summary,
30845         inline_free_summary): Move to ipa-fnsummary.h
30846         (predicate_t): Remove.
30847         * ipa-fnsummary.c: New file.
30848         * ipa-inline.h:  Do not include sreal.h and ipa-predicate.h
30849         (enum inline_hints_vals, inline_hints, agg_position_info,
30850         INLINE_SIZE_SCALE, size_time_entry, inline_summary, inline_summary_t,
30851         inline_summaries, ipa_call_summary, ipa_call_summary_t,
30852         ipa_call_summaries, debug_inline_summary, dump_inline_summaries,
30853         dump_inline_summary, dump_inline_hints, inline_generate_summary,
30854         inline_read_summary, inline_write_summary, inline_free_summary,
30855         inline_analyze_function, initialize_inline_failed,
30856         inline_merge_summary, inline_update_overall_summary,
30857         compute_inline_parameters): Move to ipa-fnsummary.h
30858         * ipa-fnsummary.h: New file.
30859         * ipa-inline-transform.h: Include ipa-inline.h.
30860         * ipa-inline.c: LIkewise.
30862 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30864         * ipa-inline.c (edge_badness): Use inlined_time instead of
30865         inline_summaries->get.
30867 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30869         * ipa-inline.c (edge_badness): Use estimate_size_after_inlining.
30871 2017-05-22  Nathan Sidwell  <nathan@acm.org>
30873         * doc/invoke.texi (fdump-translation-unit): Delete documentation.
30874         (fdump-lang): Document 'raw' option.
30875         * dumpfile.h (TDI_tu): Delete.
30876         * dumpfile.c (dump_files): Remove translation-unit.
30877         (FIRST_AUTO_NUMBERED_DUMP): Decrement.
30879 2017-05-22  Georg-Johann Lay  <avr@gjlay.de>
30881         * config/avr/t-avr (AWK) [t-multilib]: Remove "-v FORMAT=Makefile"
30882         command option from $(AWK) call.
30883         * config/avr/genmultilib.awk: Simplify and rewrite so that it
30884         generates MULTILIB_REQUIRED instead of MULTILIB_EXCEPTIONS.
30885         [FORMAT]: Remove handling of variable.
30886         * config/avr/t-multilib: Regenerate.
30888 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30890         * ipa-inline-analysis.c (inline_summary::reset): Do not reset
30891         self_time.
30892         (dump_inline_summary): Do not print self_time.
30893         (estimate_function_body_sizes): Do not set self_time.
30894         (compute_inline_parameters): Likewise.
30895         (inline_read_section, inline_write_summary): Do not stream self_time.
30896         * ipa-inline.h (inline_summary): Drop self_time.
30898 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30900         * ipa-inline-analysis.c (account_size_time): Rename to ...
30901         (inline_summary::account_size_time): ... this one.
30902         (reset_ipa_call_summary): Turn to ...
30903         (ipa_call_summary::reset): ... this one.
30904         (reset_inline_summary): Turn to ...
30905         (inline_summary::reset): ... this one.
30906         (inline_summary_t::remove): Update.
30907         (inline_summary_t::duplicate): Update.
30908         (ipa_call_summary_t::remove): Update.
30909         (dump_inline_summary): Update.
30910         (estimate_function_body_sizes): Update.
30911         (compute_inline_parameters): Update.
30912         (estimate_node_size_and_time): Update.
30913         (inline_merge_summary): Update.
30914         (inline_update_overall_summary): Update.
30915         (inline_read_section): Update.
30916         (inline_write_summary): Update.
30917         * ipa-inline.h (inline_summary): Rename entry to size_time_table;
30918         add account_size_time and reset member functions.
30919         (ipa_call_summary): Add reset function.
30920         * ipa-predicate.h (predicate::operator &): Constify.
30922 2017-05-22  Richard Biener  <rguenther@suse.de>
30924         * df-scan.c (df_insn_refs_verify): Speedup when not verifying.
30926 2017-05-19  Jason Merrill  <jason@redhat.com>
30928         * tree.c (make_tree_vec_stat, grow_tree_vec_stat): Use size_t.
30930 2017-05-19  Marek Polacek  <polacek@redhat.com>
30932         PR sanitizer/80800
30933         * fold-const.c (extract_muldiv_1) <case TRUNC_DIV_EXPR>: Add
30934         TYPE_OVERFLOW_WRAPS checks.
30936 2017-05-19  Thomas Schwinge  <thomas@codesourcery.com>
30938         * tree-core.h (enum omp_clause_default_kind): Add
30939         "OMP_CLAUSE_DEFAULT_PRESENT".
30940         * tree-pretty-print.c (dump_omp_clause): Handle it.
30941         * gimplify.c (enum gimplify_omp_var_data): Add
30942         "GOVD_MAP_FORCE_PRESENT".
30943         (gimplify_adjust_omp_clauses_1): Map it to
30944         "GOMP_MAP_FORCE_PRESENT".
30945         (oacc_default_clause): Handle "OMP_CLAUSE_DEFAULT_PRESENT".
30947         * gimplify.c (oacc_default_clause): Clarify.
30949 2017-05-19  Nathan Sidwell  <nathan@acm.org>
30951         LANG_HOOK_REGISTER_DUMPS
30952         * toplev.c (general_init): Call register dump lang hook.
30953         * doc/invoke.texi: Document -fdump-lang option family.
30954         * dumpfile.c (dump_files): Remove class dump here.
30955         (FIRST_AUTO_NUMBERED_DUMP): Adjust.
30956         * dumpfile.h (tree_dump_index): Remove TDI_class.
30957         * langhooks-def.h (lhd_register_dumps): Declare.
30958         (LANG_HOOKS_REGISTER_DUMPS): Define.
30959         (LANG_HOOKS_INITIALIZER): Add it.
30960         * langhooks.c (lhd_register_dumps): Define.
30961         * langhooks.h (struct lang_hooks): Add register_dumps.
30963 2017-05-19  Nathan Sidwell  <nathan@acm.org>
30965         * context.h (context::set_passes): New.
30966         * context.c (context::context): Do not create pass manager.
30967         * toplev.c (general_init): Create pass manager here.
30969 2017-05-19  Segher Boessenkool  <segher@kernel.crashing.org>
30971         * config/rs6000/rs6000.md (splitter to load of -1 and mask): Don't
30972         use this splitter if two add or or instructions would also work for
30973         the constant we want to generate.
30975 2017-05-19  Richard Biener  <rguenther@suse.de>
30977         PR build/80821
30978         * genmatch.c (dt_node::gen_kids_1): Add missing scope around
30979         predicate evaluation.
30981 2017-05-19  Jan Hubicka  <hubicka@ucw.cz>
30983         * ipa-inline.h (ipa_call_summary): Turn sizes into signed;
30984         add ctor.
30985         * ipa-inline.c (want_inline_small_function_p): Do not cast to
30986         unsigned.
30988 2017-05-19  Jan Hubicka  <hubicka@ucw.cz>
30990         * ipa-inline-analysis.c (cgraph_2edge_hook_list, cgraph_edge_hook_list,
30991         inline_edge_removal_hook, inline_edge_duplication_hook): Remove.
30992         (inline_edge_summary_vec): Turn into ...
30993         (ipa_call_summaries): ... this one.
30994         (redirect_to_unreachable, edge_set_predicate,
30995         evaluate_properties_for_edge, inline_summary_alloc,
30996         reset_ipa_call_summary, reset_inline_summary,
30997         inline_summary_t::duplicate): Update.
30998         (inline_edge_duplication_hook): Turn to ...
30999         (ipa_call_summary_t::duplicate): ... this one.
31000         (inline_edge_removal_hook): Turn to ...
31001         (ipa_call_summary_t::remove): ... this one.
31002         (dump_inline_edge_summary): Turn to ...
31003         (dump_ipa_call_summary): ... this one.
31004         (estimate_function_body_sizes): Update.
31005         (inline_update_callee_summaries): Update.
31006         (remap_edge_change_prob): Update.
31007         (remap_edge_summaries): Update.
31008         (inline_merge_summary): Update.
31009         (do_estimate_edge_time): Update.
31010         (inline_generate_summary): Update.
31011         (inline_read_section): Update.
31012         (inline_read_summary): Update.
31013         (inline_free_summary): Update.
31014         * ipa-inline.c (can_inline_edge_p): Update.
31015         (compute_inlined_call_time): Update.
31016         (want_inline_small_function_p): Update.
31017         (edge_badness): Update.
31018         (early_inliner): Update.
31019         * ipa-inline.h (inline_edge_summary): Turn to ...
31020         (ipa_call_summary): ... this one.
31021         (ipa_call_summary_t): New class.
31022         (inline_edge_summary_t, inline_edge_summary_vec): Remove.
31023         (ipa_call_summaries): New.
31024         (inline_edge_summary): Remove.
31025         (estimate_edge_growth): Update.
31026         * ipa-profile.c (ipa_propagate_frequency_1): Update.
31027         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
31028         * ipa-split.c (execute_split_functions): Update.
31029         * ipa.c (symbol_table::remove_unreachable_nodes): Update.
31031 2017-05-19  Richard Biener  <rguenther@suse.de>
31033         PR middle-end/80764
31034         * cfgexpand.c (expand_gimple_cond): Fix loop fixup.
31036 2017-05-18  Segher Boessenkool  <segher@kernel.crashing.org>
31038         * config/rs6000/rs6000.c (struct machine_function): Add field
31039         fpr_is_wrapped_separately.
31040         (rs6000_get_separate_components): Use 64 components.  Handle the
31041         new FPR components.
31042         (rs6000_components_for_bb): Handle the FPR components.
31043         (rs6000_emit_prologue_components): Handle the FPR components.
31044         (rs6000_emit_epilogue_components): Handle the FPR components.
31045         (rs6000_set_handled_components): Handle the FPR components.
31046         (rs6000_emit_prologue): Don't output prologue code for those FPRs
31047         that are already separately shrink-wrapped.
31048         (rs6000_emit_epilogue): Don't output epilogue code for those FPRs
31049         that are already separately shrink-wrapped.
31051 2017-05-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
31053         PR target/80510
31054         * config/rs6000/predicates.md (simple_offsettable_mem_operand):
31055         New predicate.
31057         * config/rs6000/rs6000.md (ALTIVEC_DFORM): New iterator.
31058         (define_peephole2 for Altivec d-form load): Add peepholes to catch
31059         cases where the register allocator uses a move and an offsettable
31060         memory operation to/from a FPR register on ISA 2.06/2.07.
31061         (define_peephole2 for Altivec d-form store): Likewise.
31063 2017-05-18  Uros Bizjak  <ubizjak@gmail.com>
31065         PR target/80799
31066         * config/i386/mmx.md (*mov<mode>_internal): Enable
31067         alternatives 11, 12, 13 and 14 also for 32bit targets.
31068         Remove alternatives 15, 16, 17 and 18.
31069         * config/i386/sse.md (vec_concatv2di): Change
31070         alternative (!x, *y) to (x, ?!*Yn).
31072 2017-05-18  Paolo Carlini  <paolo.carlini@oracle.com>
31074         * dumpfile.h (enum dump_kind): Remove stray comma.
31076 2017-05-18  Jan Hubicka  <hubicka@ucw.cz>
31078         * Makefile.in: Add ipa-predicate.o and ipa-predicate.h
31079         * ipa-inline-analysis.c (NUM_CONDITIONS): turn into
31080         predicate::num_conditions
31081         (IS_NOT_CONSTANT): turn into predicate::is_not_constant.
31082         (CHANGED): turn into predicate::changed.
31083         (agg_position_info): Move to ipa-predicate.h
31084         (add_condition, predicate::add_clause, predicate::operator &=,
31085         predicate::or_with, predicate::evaluate, predicate::probability,
31086         dump_condition, dump_clause, predicate::dump,
31087         predicate::remap_after_duplication, predicate::remap_after_inlining,
31088         predicate::stream_in, predicate::stream_out): Move to ipa-predicate.c
31089         (evaluate_conditions_for_known_args): Update.
31090         (set_cond_stmt_execution_predicate): Update.
31091         * ipa-inline.h: Include ipa-predicate.h
31092         (condition, inline_param_summary, conditions, agg_position_info,
31093         predicate): Move to ipa-predicate.h
31094         * ipa-predicate.c: New file.
31095         * ipa-predicate.h: New file.
31097 2017-05-18  Wilco Dijkstra  <wdijkstr@arm.com>
31099         * final.c (leaf_function_p): Check we are not in a sequence.
31101 2017-05-18  Martin Liska  <mliska@suse.cz>
31103         * cfgrtl.c (rtl_verify_edges): Remove usage of TDF_RTL.
31104         * dumpfile.c (dump_register): Use new enum dump_kind.
31105         (get_dump_file_name): Likewise.
31106         (dump_enable_all): Likewise.
31107         (dump_switch_p_1): Likewise.
31108         (enable_rtl_dump_file): Remove usage of TDF_RTL.
31109         * dumpfile.h (enum dump_kind): New enum type.
31110         (struct dump_file_info): Create constructor and
31111         format fields and comments.
31112         * passes.c (pass_manager::register_one_dump_file):
31113         Use num dump_kind.
31114         * statistics.c (statistics_early_init): Likewise.
31115         * tree-ssa-loop-prefetch.c (dump_mem_details): Replace
31116         TDF_TREE with TDF_SLIM.
31117         (gather_memory_references_ref): Likewise.
31119 2017-05-18  Martin Liska  <mliska@suse.cz>
31121         * vec.h (struct vnull): Use it.
31123 2017-05-18  Jan Hubicka  <hubicka@ucw.cz>
31125         * ipa-inline-analysis.c (predicate_conditions): Move to ipa-inline.h
31126         (true_predicate, false_predicate, true_predicate_p,
31127         false_predicate_p): Remove.
31128         (single_cond_predicate, not_inlined_predicate): Turn to member function
31129         in ipa-inline.h
31130         (add_condition): Update.
31131         (add_clause): Turn to...
31132         (predicate::add_clause): ... this one; update; allow passing NULL
31133         as parameter.
31134         (and_predicates): Turn to ...
31135         (predicate::operator &=): ... this one.
31136         (predicates_equal_p): Move to predicate::operator == in ipa-inline.h
31137         (or_predicates): Turn to ...
31138         (predicate::or_with): ... this one.
31139         (evaluate_predicate): Turn to ...
31140         (predicate::evaluate): ... this one.
31141         (predicate_probability): Turn to ...
31142         (predicate::probability): ... this one.
31143         (dump_condition): Update.
31144         (dump_predicate): Turn to ...
31145         (predicate::dump): ... this one.
31146         (account_size_time): Update.
31147         (edge_set_predicate): Update.
31148         (set_hint_predicate): UPdate.
31149         (evaluate_conditions_for_known_args): Update.
31150         (evaluate_properties_for_edge): Update.
31151         (remap_predicate_after_duplication): Turn to...
31152         (predicate::remap_after_duplication): ... this one.
31153         (remap_hint_predicate_after_duplication): Update.
31154         (inline_summary_t::duplicate): UPdate.
31155         (dump_inline_edge_summary): Update.
31156         (dump_inline_summary): Update.
31157         (set_cond_stmt_execution_predicate): Update.
31158         (set_switch_stmt_execution_predicate): Update.
31159         (compute_bb_predicates): Update.
31160         (will_be_nonconstant_expr_predicate): Update.
31161         (will_be_nonconstant_predicate): Update.
31162         (phi_result_unknown_predicate): Update.
31163         (predicate_for_phi_result): Update.
31164         (array_index_predicate): Update.
31165         (estimate_function_body_sizes): Update.
31166         (estimate_node_size_and_time): Update.
31167         (estimate_ipcp_clone_size_and_time): Update.
31168         (remap_predicate): Rename to ...
31169         (predicate::remap_after_inlining): ... this one.
31170         (remap_hint_predicate): Update.
31171         (inline_merge_summary): Update.
31172         (inline_update_overall_summary): Update.
31173         (estimate_size_after_inlining): Update.
31174         (read_predicate): Rename to ...
31175         (predicate::stream_in): ... this one.
31176         (read_inline_edge_summary): Update.
31177         (write_predicate): Rename to ...
31178         (predicate::stream_out): ... this one.
31179         (write_inline_edge_summary): Update.
31180         * ipa-inline.h (MAX_CLAUSES): Turn to predicate::max_clauses.
31181         (clause_t): Turn to uint32_t
31182         (predicate): Turn to class; implement constructor and operators
31183         ==, !=, &
31184         (size_time_entry): Update.
31185         (inline_summary): Update.
31186         (inline_edge_summary): Update.
31188 2017-05-18  Marc Glisse  <marc.glisse@inria.fr>
31190         * fold-const.c (fold_binary_loc): Move transformation...
31191         * match.pd (C - X CMP X): ... here.
31193 2017-05-18  Sheldon Lobo  <sheldon.lobo@oracle.com>
31195         * config/sparc/sparc.c (sparc_option_override): Set function
31196         alignment for -mcpu=niagara7 to 64 to match the I$ line.
31197         * config/sparc/sparc.h (BRANCH_COST): Set the SPARC M7 branch
31198         latency to 1.
31199         * config/sparc/sparc.h (BRANCH_COST): Set the SPARC T4 branch
31200         latency to 2.
31201         * config/sparc/sol2.h: Fix a ASM_CPU32_DEFAULT_SPEC typo.
31203 2017-05-18  Marek Polacek  <polacek@redhat.com>
31205         PR sanitizer/80797
31206         * ubsan.c (instrument_null): Unwrap ADDR_EXPRs.
31207         (pass_ubsan::execute): Call gimple_assign_single_p instead of
31208         gimple_assign_load_p.
31210 2017-05-17  Segher Boessenkool  <segher@kernel.crashing.org>
31212         PR middle-end/80692
31213         * real.c (do_compare): Give decimal_do_compare preference over
31214         comparing just the signs.
31216 2017-05-17  Uros Bizjak  <ubizjak@gmail.com>
31218         * doc/md.texi (Canonicalization of Instructions): Describe the
31219         canonical form of instructions that inherently set a condition
31220         code register.
31222 2017-05-17  Peter Bergner  <bergner@vnet.ibm.com>
31224         PR middle-end/80775
31225         * tree-cfg.c: Move deletion of unreachable case statements to after
31226         the merging of consecutive case labels.
31228 2017-05-17  Thomas Preud'homme  <thomas.preudhomme@arm.com>
31230         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Refer
31231         readers to __gnu_cmse_nonsecure_call libcall for saving, clearing and
31232         restoring of callee-saved registers.
31234 2017-05-17  Eric Botcazou  <ebotcazou@adacore.com>
31236         * compare-elim.c (equivalent_reg_at_start): Adjust outdated comment.
31237         * config/visium/visium.c (single_set_and_flags): Likewise.
31238         * config/visium/visium.md (Substitutions): Likewise.
31240 2017-05-17  Martin Liska  <mliska@suse.cz>
31242         * cfg.c: Introduce dump_flags_t type and
31243         use it instead of int type.
31244         * cfg.h: Likewise.
31245         * cfghooks.c: Likewise.
31246         * cfghooks.h (struct cfg_hooks): Likewise.
31247         * cfgrtl.c: Likewise.
31248         * cfgrtl.h: Likewise.
31249         * cgraph.c (cgraph_node::get_body): Likewise.
31250         * coretypes.h: Likewise.
31251         * domwalk.c: Likewise.
31252         * domwalk.h: Likewise.
31253         * dumpfile.c (struct dump_option_value_info): Likewise.
31254         (dump_enable_all): Likewise.
31255         (dump_switch_p_1): Likewise.
31256         (opt_info_switch_p): Likewise.
31257         * dumpfile.h (enum tree_dump_index): Likewise.
31258         (struct dump_file_info): Likewise.
31259         * genemit.c: Likewise.
31260         * generic-match-head.c: Likewise.
31261         * gengtype.c (open_base_files): Likewise.
31262         * gimple-pretty-print.c: Likewise.
31263         * gimple-pretty-print.h: Likewise.
31264         * graph.c (print_graph_cfg): Likewise.
31265         * graphite-scop-detection.c (dot_all_sese): Likewise.
31266         * ipa-devirt.c (build_type_inheritance_graph): Likewise.
31267         * loop-unroll.c (report_unroll): Likewise.
31268         * passes.c (pass_manager::register_one_dump_file): Likewise.
31269         * print-tree.c: Likewise.
31270         * statistics.c: Likewise.
31271         * tree-cfg.c: Likewise.
31272         * tree-cfg.h: Likewise.
31273         * tree-dfa.c: Likewise.
31274         * tree-dfa.h: Likewise.
31275         * tree-dump.c (dump_function): Likewise.
31276         * tree-dump.h (struct dump_info): Likewise.
31277         * tree-pretty-print.c: Likewise.
31278         * tree-pretty-print.h: Likewise.
31279         * tree-ssa-live.c: Likewise.
31280         * tree-ssa-live.h: Likewise.
31281         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Likewise.
31282         * tree-vect-loop.c: Likewise.
31283         * tree-vect-slp.c: Likewise.
31285 2017-05-16  James Greenhalgh  <james.greenhalgh@arm.com>
31286             Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
31288         PR tree-optimization/80457
31289         * tree-vect-stmts.c (vect_model_simple_cost): Model the cost
31290         of all arguments to a statement as scalar_to_vec operations.
31291         (vectorizable_call): Adjust call to vect_model_simple_cost for
31292         new parameter.
31293         (vectorizable_conversion): Likewise.
31294         (vectorizable_assignment): Likewise.
31295         (vectorizable_shift): Likewise.
31296         (vectorizable_operation): Likewise.
31297         (vectorizable_comparison): Likewise.
31298         (vect_is_simple_cond): Record the def types for operands.
31299         (vectorizable_condition): Likewise, call vect_model_simple_cost.
31300         * tree-vectorizer.h (vect_model_simple_cost): Add new parameter
31301         for statement argument count.
31303 2017-05-16  Carl Love  <cel@us.ibm.com>
31305         * config/rs6000/rs6000-c: Add support for built-in functions
31306         vector unsigned long long vec_bperm (vector unsigned long long,
31307                                              vector unsigned char)
31308         vector signed long long vec_mule (vector signed int,
31309                                           vector signed int)
31310         vector unsigned long long vec_mule (vector unsigned int,
31311                                             vector unsigned int)
31312         vector signed long long vec_mulo (vector signed int,
31313                                           vector signed int)
31314         vector unsigned long long vec_mulo (vector unsigned int,
31315                                             vector unsigned int)
31316         vector signed char vec_sldw (vector signed char,
31317                                      vector signed char,
31318                                      const int)
31319         vector unsigned char vec_sldw (vector unsigned char,
31320                                        vector unsigned char,
31321                                        const int)
31322         vector signed short vec_sldw (vector signed short,
31323                                       vector signed short,
31324                                       const int)
31325         vector unsigned short vec_sldw (vector unsigned short,
31326                                         vector unsigned short,
31327                                         const int)
31328         vector signed int vec_sldw (vector signed int,
31329                                     vector signed int,
31330                                     const int)
31331         vector unsigned int vec_sldw (vector unsigned int,
31332                                       vector unsigned int,
31333                                       const int)
31334         vector signed long long vec_sldw (vector signed long long,
31335                                           vector signed long long,
31336                                           const int)
31337         vector unsigned long long vec_sldw (vector unsigned long long,
31338                                             vector unsigned long long,
31339                                             const int)
31340         * config/rs6000/rs6000-c: Add support for built-in functions
31341         * config/rs6000/rs6000-builtin.def: Add definition for SLDW.
31342         * config/rs6000/altivec.h: Add defintion for vec_sldw.
31343         * doc/extend.texi: Update the built-in documentation for the
31344         new built-in functions.
31346 2017-05-16  Marek Polacek  <polacek@redhat.com>
31348         PR sanitizer/80536
31349         PR sanitizer/80386
31350         * tree.c (save_expr): Don't fold the expression.
31352 2017-05-16  Uros Bizjak  <ubizjak@gmail.com>
31354         * config/i386/i386.md (*movsi_internal): Split (?rm,*y) alternative
31355         to (?r,*Yn) and (?m,*y) alternatives, and (?*y,rm) to (?*Ym,r)
31356         and (?*y,m).  Update insn attributes.
31358 2017-05-16  Martin Liska  <mliska@suse.cz>
31360         * cgraph.c (cgraph_edge::resolve_speculation): Add default value for
31361         flags argument of print_gimple_stmt, print_gimple_expr,
31362         print_generic_stmt and print_generic_expr.
31363         * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
31364         * coretypes.h: Likewise.
31365         * except.c (dump_eh_tree): Likewise.
31366         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
31367         * gimple-pretty-print.h: Likewise.
31368         * gimple-ssa-backprop.c (dump_usage_prefix): Likewise.
31369         (backprop::push_to_worklist): Likewise.
31370         (backprop::pop_from_worklist): Likewise.
31371         (backprop::process_use): Likewise.
31372         (backprop::intersect_uses): Likewise.
31373         (note_replacement): Likewise.
31374         * gimple-ssa-store-merging.c
31375         (pass_store_merging::terminate_all_aliasing_chains): Likewise.
31376         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
31377         (pass_store_merging::execute): Likewise.
31378         * gimple-ssa-strength-reduction.c (dump_candidate): Likewise.
31379         (ssa_base_cand_dump_callback): Likewise.
31380         (dump_incr_vec): Likewise.
31381         (replace_refs): Likewise.
31382         (replace_mult_candidate): Likewise.
31383         (create_add_on_incoming_edge): Likewise.
31384         (create_phi_basis): Likewise.
31385         (insert_initializers): Likewise.
31386         (all_phi_incrs_profitable): Likewise.
31387         (introduce_cast_before_cand): Likewise.
31388         (replace_one_candidate): Likewise.
31389         * gimplify.c (gimplify_expr): Likewise.
31390         * graphite-isl-ast-to-gimple.c (is_valid_rename): Likewise.
31391         (set_rename): Likewise.
31392         (rename_uses): Likewise.
31393         (copy_loop_phi_nodes): Likewise.
31394         (add_close_phis_to_merge_points): Likewise.
31395         (copy_loop_close_phi_args): Likewise.
31396         (copy_cond_phi_args): Likewise.
31397         (graphite_copy_stmts_from_block): Likewise.
31398         (translate_pending_phi_nodes): Likewise.
31399         * graphite-poly.c (print_pdr): Likewise.
31400         (dump_gbb_cases): Likewise.
31401         (dump_gbb_conditions): Likewise.
31402         (print_scop_params): Likewise.
31403         * graphite-scop-detection.c (build_cross_bb_scalars_def): Likewise.
31404         (build_cross_bb_scalars_use): Likewise.
31405         (gather_bbs::before_dom_children): Likewise.
31406         * hsa-dump.c (dump_hsa_immed): Likewise.
31407         * ipa-cp.c (print_ipcp_constant_value): Likewise.
31408         (get_replacement_map): Likewise.
31409         * ipa-inline-analysis.c (dump_condition): Likewise.
31410         (estimate_function_body_sizes): Likewise.
31411         * ipa-polymorphic-call.c (check_stmt_for_type_change): Likewise.
31412         (ipa_polymorphic_call_context::get_dynamic_type): Likewise.
31413         * ipa-prop.c (ipa_dump_param): Likewise.
31414         (ipa_print_node_jump_functions_for_edge): Likewise.
31415         (ipa_modify_call_arguments): Likewise.
31416         (ipa_modify_expr): Likewise.
31417         (ipa_dump_param_adjustments): Likewise.
31418         (ipa_dump_agg_replacement_values): Likewise.
31419         (ipcp_modif_dom_walker::before_dom_children): Likewise.
31420         * ipa-pure-const.c (check_stmt): Likewise.
31421         (pass_nothrow::execute): Likewise.
31422         * ipa-split.c (execute_split_functions): Likewise.
31423         * omp-offload.c (dump_oacc_loop_part): Likewise.
31424         (dump_oacc_loop): Likewise.
31425         * trans-mem.c (tm_log_emit): Likewise.
31426         (tm_memopt_accumulate_memops): Likewise.
31427         (dump_tm_memopt_set): Likewise.
31428         (dump_tm_memopt_transform): Likewise.
31429         * tree-cfg.c (gimple_verify_flow_info): Likewise.
31430         (print_loop): Likewise.
31431         * tree-chkp-opt.c (chkp_print_addr): Likewise.
31432         (chkp_gather_checks_info): Likewise.
31433         (chkp_get_check_result): Likewise.
31434         (chkp_remove_check_if_pass): Likewise.
31435         (chkp_use_outer_bounds_if_possible): Likewise.
31436         (chkp_reduce_bounds_lifetime): Likewise.
31437         * tree-chkp.c (chkp_register_addr_bounds): Likewise.
31438         (chkp_mark_completed_bounds): Likewise.
31439         (chkp_register_incomplete_bounds): Likewise.
31440         (chkp_mark_invalid_bounds): Likewise.
31441         (chkp_maybe_copy_and_register_bounds): Likewise.
31442         (chkp_build_returned_bound): Likewise.
31443         (chkp_get_bound_for_parm): Likewise.
31444         (chkp_build_bndldx): Likewise.
31445         (chkp_get_bounds_by_definition): Likewise.
31446         (chkp_generate_extern_var_bounds): Likewise.
31447         (chkp_get_bounds_for_decl_addr): Likewise.
31448         * tree-chrec.c (chrec_apply): Likewise.
31449         * tree-data-ref.c (dump_data_reference): Likewise.
31450         (dump_subscript): Likewise.
31451         (dump_data_dependence_relation): Likewise.
31452         (analyze_overlapping_iterations): Likewise.
31453         * tree-inline.c (expand_call_inline): Likewise.
31454         (tree_function_versioning): Likewise.
31455         * tree-into-ssa.c (dump_defs_stack): Likewise.
31456         (dump_currdefs): Likewise.
31457         (dump_names_replaced_by): Likewise.
31458         (dump_update_ssa): Likewise.
31459         (update_ssa): Likewise.
31460         * tree-object-size.c (pass_object_sizes::execute): Likewise.
31461         * tree-parloops.c (build_new_reduction): Likewise.
31462         (try_create_reduction_list): Likewise.
31463         (ref_conflicts_with_region): Likewise.
31464         (oacc_entry_exit_ok_1): Likewise.
31465         (oacc_entry_exit_single_gang): Likewise.
31466         * tree-pretty-print.h: Likewise.
31467         * tree-scalar-evolution.c (set_scalar_evolution): Likewise.
31468         (get_scalar_evolution): Likewise.
31469         (add_to_evolution): Likewise.
31470         (get_loop_exit_condition): Likewise.
31471         (analyze_evolution_in_loop): Likewise.
31472         (analyze_initial_condition): Likewise.
31473         (analyze_scalar_evolution): Likewise.
31474         (instantiate_scev): Likewise.
31475         (number_of_latch_executions): Likewise.
31476         (gather_chrec_stats): Likewise.
31477         (final_value_replacement_loop): Likewise.
31478         (scev_const_prop): Likewise.
31479         * tree-sra.c (dump_access): Likewise.
31480         (disqualify_candidate): Likewise.
31481         (create_access): Likewise.
31482         (reject): Likewise.
31483         (maybe_add_sra_candidate): Likewise.
31484         (create_access_replacement): Likewise.
31485         (analyze_access_subtree): Likewise.
31486         (analyze_all_variable_accesses): Likewise.
31487         (sra_modify_assign): Likewise.
31488         (initialize_constant_pool_replacements): Likewise.
31489         (find_param_candidates): Likewise.
31490         (decide_one_param_reduction): Likewise.
31491         (replace_removed_params_ssa_names): Likewise.
31492         * tree-ssa-ccp.c (ccp_fold_stmt): Likewise.
31493         * tree-ssa-copy.c (dump_copy_of): Likewise.
31494         (copy_prop_visit_cond_stmt): Likewise.
31495         * tree-ssa-dce.c (mark_operand_necessary): Likewise.
31496         * tree-ssa-dom.c (pass_dominator::execute): Likewise.
31497         (record_equivalences_from_stmt): Likewise.
31498         * tree-ssa-dse.c (compute_trims): Likewise.
31499         (delete_dead_call): Likewise.
31500         (delete_dead_assignment): Likewise.
31501         * tree-ssa-forwprop.c (forward_propagate_into_gimple_cond): Likewise.
31502         (forward_propagate_into_cond): Likewise.
31503         (pass_forwprop::execute): Likewise.
31504         * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
31505         * tree-ssa-loop-im.c (invariantness_dom_walker::before_dom_children):
31506         Likewise.
31507         (move_computations_worker): Likewise.
31508         (execute_sm): Likewise.
31509         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Likewise.
31510         (remove_exits_and_undefined_stmts): Likewise.
31511         (remove_redundant_iv_tests): Likewise.
31512         * tree-ssa-loop-ivopts.c (dump_use): Likewise.
31513         (adjust_iv_update_pos): Likewise.
31514         * tree-ssa-math-opts.c (bswap_replace): Likewise.
31515         * tree-ssa-phiopt.c (factor_out_conditional_conversion): Likewise.
31516         (value_replacement): Likewise.
31517         * tree-ssa-phiprop.c (phiprop_insert_phi): Likewise.
31518         * tree-ssa-pre.c (print_pre_expr): Likewise.
31519         (get_representative_for): Likewise.
31520         (create_expression_by_pieces): Likewise.
31521         (insert_into_preds_of_block): Likewise.
31522         (eliminate_insert): Likewise.
31523         (eliminate_dom_walker::before_dom_children): Likewise.
31524         (eliminate): Likewise.
31525         (remove_dead_inserted_code): Likewise.
31526         * tree-ssa-propagate.c (substitute_and_fold): Likewise.
31527         * tree-ssa-reassoc.c (get_rank): Likewise.
31528         (eliminate_duplicate_pair): Likewise.
31529         (eliminate_plus_minus_pair): Likewise.
31530         (eliminate_not_pairs): Likewise.
31531         (undistribute_ops_list): Likewise.
31532         (eliminate_redundant_comparison): Likewise.
31533         (update_range_test): Likewise.
31534         (optimize_range_tests_var_bound): Likewise.
31535         (optimize_vec_cond_expr): Likewise.
31536         (rewrite_expr_tree): Likewise.
31537         (rewrite_expr_tree_parallel): Likewise.
31538         (linearize_expr): Likewise.
31539         (break_up_subtract): Likewise.
31540         (linearize_expr_tree): Likewise.
31541         (attempt_builtin_powi): Likewise.
31542         (attempt_builtin_copysign): Likewise.
31543         (transform_stmt_to_copy): Likewise.
31544         (transform_stmt_to_multiply): Likewise.
31545         (dump_ops_vector): Likewise.
31546         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Likewise.
31547         (print_scc): Likewise.
31548         (set_ssa_val_to): Likewise.
31549         (visit_reference_op_store): Likewise.
31550         (visit_use): Likewise.
31551         (sccvn_dom_walker::before_dom_children): Likewise.
31552         (run_scc_vn): Likewise.
31553         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
31554         Likewise.
31555         (expr_hash_elt::print): Likewise.
31556         (const_and_copies::pop_to_marker): Likewise.
31557         (const_and_copies::record_const_or_copy_raw): Likewise.
31558         * tree-ssa-structalias.c (compute_dependence_clique): Likewise.
31559         * tree-ssa-uninit.c (collect_phi_def_edges): Likewise.
31560         (dump_predicates): Likewise.
31561         (find_uninit_use): Likewise.
31562         (warn_uninitialized_phi): Likewise.
31563         (pass_late_warn_uninitialized::execute): Likewise.
31564         * tree-ssa.c (verify_vssa): Likewise.
31565         (verify_ssa): Likewise.
31566         (maybe_optimize_var): Likewise.
31567         * tree-vrp.c (dump_value_range): Likewise.
31568         (dump_all_value_ranges): Likewise.
31569         (dump_asserts_for): Likewise.
31570         (register_edge_assert_for_2): Likewise.
31571         (vrp_visit_cond_stmt): Likewise.
31572         (vrp_visit_switch_stmt): Likewise.
31573         (vrp_visit_stmt): Likewise.
31574         (vrp_visit_phi_node): Likewise.
31575         (simplify_cond_using_ranges_1): Likewise.
31576         (fold_predicate_in): Likewise.
31577         (evrp_dom_walker::before_dom_children): Likewise.
31578         (evrp_dom_walker::push_value_range): Likewise.
31579         (evrp_dom_walker::pop_value_range): Likewise.
31580         (execute_early_vrp): Likewise.
31582 2017-05-16  Richard Biener  <rguenther@suse.de>
31584         * dwarf2out.c (loc_list_from_tree_1): Do not create
31585         DW_OP_GNU_variable_value for DECL_IGNORED_P decls.
31587 2017-05-16  Richard Biener  <rguenther@suse.de>
31589         * dwarf2out.c (resolve_variable_value_in_expr): Lookup DIE
31590         just generated.
31591         (note_variable_value_in_expr): If we resolved the decl ref
31592         do not push to the stack.
31594 2017-05-16  Matthew Wahab  <matthew.wahab@arm.com>
31596         * config/arm/arm_neon.h (vadd_f16): Use standard arithmetic
31597         operations in fast-math mode.
31598         (vaddq_f16): Likewise.
31599         (vmul_f16): Likewise.
31600         (vmulq_f16): Likewise.
31601         (vsub_f16): Likewise.
31602         (vsubq_f16): Likewise.
31603         * config/arm/neon.md (add<mode>3): New.
31604         (sub<mode>3): New.
31605         (fma:<VH:mode>3): New.  Also remove outdated comment.
31606         (mul<mode>3): New.
31608 2017-05-16  Martin Liska  <mliska@suse.cz>
31610         PR ipa/79849.
31611         PR ipa/79850.
31612         * ipa-devirt.c (warn_types_mismatch): Fix typo.
31613         (odr_types_equivalent_p): Likewise.
31615 2017-05-15  Sylvestre Ledru  <sylvestre@debian.org>
31617         * plugin.c (try_init_one_plugin): Fix ressource leaks (CID 726637).
31619 2017-05-15  Uros Bizjak  <ubizjak@gmail.com>
31621         PR target/80425
31622         * config/i386.i386.md (*zero_extendsidi2): Do not penalize
31623         non-interunit SSE move alternatives with '?'.
31624         (zero-extendsidi peephole2): New peephole to skip intermediate
31625         general register in SSE zero-extend sequence.
31627 2017-05-15  Jeff Law  <law@redhat.com>
31629         * reorg.c (relax_delay_slots): Create a new variable to hold
31630         the temporary target rather than clobbering TARGET_LABEL.
31632         * config/tilegx/tilegx.c (tilegx_expand_unaligned_load): Add
31633         missing argument to extract_bit_field call.
31634         * config/tilepro/tilepro.c (tilepro_expand_unaligned_load): Likewise.
31636 2017-05-15  Martin Liska  <mliska@suse.cz>
31638         PR driver/31468
31639         * gcc.c (process_command): Do not allow empty argument of -o option.
31641 2017-05-15  Renlin Li  <renlin.li@arm.com>
31643         * config/aarch64/aarch64-protos.h (aarch64_expand_call): Declare.
31644         * config/aarch64/aarch64.c (aarch64_expand_call): Define.
31645         * config/aarch64/constraints.md (Usf): Add long call check.
31646         * config/aarch64/aarch64.md (call): Use aarch64_expand_call.
31647         (call_value): Likewise.
31648         (sibcall): Likewise.
31649         (sibcall_value): Likewise.
31650         (call_insn): New.
31651         (call_value_insn): New.
31652         (sibcall_insn): Update rtx pattern.
31653         (sibcall_value_insn): Likewise.
31654         (call_internal): Remove.
31655         (call_value_internal): Likewise.
31656         (sibcall_internal): Likewise.
31657         (sibcall_value_internal): Likewise.
31658         (call_reg): Likewise.
31659         (call_symbol): Likewise.
31660         (call_value_reg): Likewise.
31661         (call_value_symbol): Likewise.
31663 2017-05-14  Krister Walfridsson  <krister.walfridsson@gmail.com>
31665         PR target/80600
31666         * config/netbsd.h (NETBSD_LIBGCC_SPEC): Always add -lgcc.
31668 2017-05-14  Uros Bizjak  <ubizjak@gmail.com>
31670         * config/i386.i386.c (ix86_cc_modes_compatible): CCNOmode is
31671         compatible with CCGOCmode and with CCZmode.
31673 2017-05-14  Martin Sebor  <msebor@redhat.com>
31675         PR middle-end/77671
31676         * gimple-fold.c (gimple_fold_builtin_sprintf): Make extern.
31677         (gimple_fold_builtin_snprintf): Same.
31678         * gimple-fold.h (gimple_fold_builtin_sprintf): Declare.
31679         (gimple_fold_builtin_snprintf): Same.
31680         * gimple-ssa-sprintf.c (get_format_string): Correct the detection
31681         of character types.
31682         (is_call_safe): New function.
31683         (try_substitute_return_value): Call it.
31684         (try_simplify_call): New function.
31685         (pass_sprintf_length::handle_gimple_call): Call it.
31687 2017-05-14  Martin Sebor  <msebor@redhat.com>
31689         PR middle-end/80669
31690         * builtins.c (expand_builtin_stpncpy): Simplify.
31692 2017-05-14  Daniel Santos  <daniel.santos@pobox.com>
31694         * config/i386/i386.opt: Add option -mcall-ms2sysv-xlogues.
31695         * config/i386/i386.h
31696         (x86_64_ms_sysv_extra_clobbered_registers): Change type to unsigned.
31697         (NUM_X86_64_MS_CLOBBERED_REGS): New macro.
31698         (struct machine_function): Add new members call_ms2sysv,
31699         call_ms2sysv_pad_in, call_ms2sysv_pad_out and call_ms2sysv_extra_regs.
31700         (struct machine_frame_state): New fields sp_realigned and
31701         sp_realigned_offset.
31702         * config/i386/i386.c
31703         (enum xlogue_stub): New enum.
31704         (enum xlogue_stub_sets): New enum.
31705         (class xlogue_layout): New class.
31706         (struct ix86_frame): New fields stack_realign_allocate_offset,
31707         stack_realign_offset and outlined_save_offset.  Modify comments to
31708         detail stack layout when using out-of-line stubs.
31709         (ix86_target_string): Add -mcall-ms2sysv-xlogues option.
31710         (ix86_option_override_internal): Add sorry() for TARGET_SEH and
31711         -mcall-ms2sysv-xlogues.
31712         (stub_managed_regs): New static variable.
31713         (ix86_save_reg): Add new parameter ignore_outlined to optionally omit
31714         registers managed by out-of-line stub.
31715         (disable_call_ms2sysv_xlogues): New function.
31716         (ix86_compute_frame_layout): Modify re-alignment calculations, disable
31717         m->call_ms2sysv when appropriate and compute frame layout for
31718         out-of-line stubs.
31719         (sp_valid_at, fp_valid_at): New inline functions.
31720         (choose_basereg): New function.
31721         (choose_baseaddr): Add align parameter, use choose_basereg and modify
31722         all callers.
31723         (ix86_emit_save_reg_using_mov, ix86_emit_restore_sse_regs_using_mov):
31724         Use align parameter of choose_baseaddr to generated aligned SSE movs
31725         when possible.
31726         (pro_epilogue_adjust_stack): Modify to track
31727         machine_frame_state::sp_realigned.
31728         (ix86_nsaved_regs): Modify to accommodate changes to ix86_save_reg.
31729         (ix86_nsaved_sseregs): Likewise.
31730         (ix86_emit_save_regs): Likewise.
31731         (ix86_emit_save_regs_using_mov): Likewise.
31732         (ix86_emit_save_sse_regs_using_mov): Likewise.
31733         (get_scratch_register_on_entry): Likewise.
31734         (gen_frame_set): New function.
31735         (gen_frame_load): Likewise.
31736         (gen_frame_store): Likewise.
31737         (emit_outlined_ms2sysv_save): Likewise.
31738         (emit_outlined_ms2sysv_restore): Likewise.
31739         (ix86_expand_prologue): Modify stack re-alignment code and call
31740         emit_outlined_ms2sysv_save when appropriate.
31741         (ix86_emit_leave): Clear machine_frame_state::sp_realigned.  Add
31742         parameter rtx_insn *insn, which allows the function to be used to only
31743         generate the notes.
31744         (ix86_expand_epilogue): Modify validity checks of frame and stack
31745         pointers, and call emit_outlined_ms2sysv_restore when appropriate.
31746         (ix86_expand_call): Modify to enable m->call_ms2sysv when appropriate.
31747         * config/i386/predicates.md
31748         (save_multiple): New predicate.
31749         (restore_multiple): Likewise.
31750         * config/i386/sse.md
31751         (save_multiple<mode>): New pattern.
31752         (save_multiple_realign<mode>): Likewise.
31753         (restore_multiple<mode>): Likewise.
31754         (restore_multiple_and_return<mode>): Likewise.
31755         (restore_multiple_leave_return<mode>): Likewise.
31756         * Makefile.in: Export HOSTCXX and HOSTCXXFLAGS to site.exp
31758 2017-05-14  Julia Koval  <julia.koval@intel.com>
31760         * config/i386/i386-builtin-types.def (VOID_FTYPE_INT_INT64): New type.
31761         * config/i386/i386-builtin.def (__builtin_ia32_xgetbv)
31762         (__builtin_ia32_xsetbv): New builtins.
31763         * config/i386/i386.c (ix86_expand_special_args_builtin):
31764         Process new types.
31765         (ix86_expand_builtin): Special expand for new intrinsics.
31766         * config/i386/i386.md (UNSPECV_XGETBV, UNSPECV_XSETBV): New.
31767         (xsetbv, xsetbv_rex64, xgetbv, xgetbv_rex64): New insn patterns.
31768         * config/i386/xsaveintrin.h (_xsetbv, _getbv): New intrinsics.
31770 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31772         * cfganal.c (inverted_post_order_compute): Change argument type
31773         to vec *.
31774         * cfganal.h (inverted_post_order_compute): Adjust prototype.
31775         * df-core.c (rest_of_handle_df_initialize): Adjust.
31776         (rest_of_handle_df_finish): Likewise.
31777         (df_analyze_1): Likewise.
31778         (df_analyze): Likewise.
31779         (loop_inverted_post_order_compute): Change argument to be a vec *.
31780         (df_analyze_loop): Adjust.
31781         (df_get_n_blocks): Likewise.
31782         (df_get_postorder): Likewise.
31783         * df.h (struct df_d): Change field to be a vec.
31784         * lcm.c (compute_laterin): Adjust.
31785         (compute_available): Likewise.
31786         * lra-lives.c (lra_create_live_ranges_1): Likewise.
31787         * tree-ssa-dce.c (remove_dead_stmt): Likewise.
31788         * tree-ssa-pre.c (compute_antic): Likewise.
31790 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31792         * cfganal.c (connect_infinite_loops_to_exit): Adjust.
31793         (depth_first_search::depth_first_search): Change structure init
31794         function to this constructor.
31795         (depth_first_search::add_bb): Rename function to this member.
31796         (depth_first_search::execute): Likewise.
31797         (flow_dfs_compute_reverse_finish): Adjust.
31799 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31801         * ddg.c (find_nodes_on_paths): Use auto_sbitmap.
31802         (longest_simple_path): Likewise.
31803         * shrink-wrap.c (spread_components): Likewise.
31804         (disqualify_problematic_components): Likewise.
31805         (emit_common_heads_for_components): Likewise.
31806         (emit_common_tails_for_components): Likewise.
31807         (insert_prologue_epilogue_for_components): Likewise.
31809 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31811         * tree-ssa-dse.c (dse_dom_walker): Make m_live_byes a
31812         auto_sbitmap.
31814 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31816         * df-core.c (df_set_blocks): Start using auto_bitmap.
31817         (df_compact_blocks): Likewise.
31818         * df-problems.c (df_rd_confluence_n): Likewise.
31819         * df-scan.c (df_insn_rescan_all): Likewise.
31820         (df_process_deferred_rescans): Likewise.
31821         (df_update_entry_block_defs): Likewise.
31822         (df_update_exit_block_uses): Likewise.
31823         (df_entry_block_bitmap_verify): Likewise.
31824         (df_exit_block_bitmap_verify): Likewise.
31825         (df_scan_verify): Likewise.
31826         * lra-constraints.c (lra_constraints): Likewise.
31827         (undo_optional_reloads): Likewise.
31828         (lra_undo_inheritance): Likewise.
31829         * lra-remat.c (calculate_gen_cands): Likewise.
31830         (do_remat): Likewise.
31831         * lra-spills.c (assign_spill_hard_regs): Likewise.
31832         (spill_pseudos): Likewise.
31833         * tree-ssa-pre.c (bitmap_set_and): Likewise.
31834         (bitmap_set_subtract_values): Likewise.
31836 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31838         * haifa-sched.c (estimate_shadow_tick): Replace manual bitmap
31839         management with auto_bitmap.
31840         (fix_inter_tick): Likewise.
31841         (fix_recovery_deps): Likewise.
31842         * ira.c (add_store_equivs): Likewise.
31843         (find_moveable_pseudos): Likewise.
31844         (split_live_ranges_for_shrink_wrap): Likewise.
31845         * print-rtl.c (rtx_reuse_manager::rtx_reuse_manager): Likewise.
31846         (rtx_reuse_manager::seen_def_p): Likewise.
31847         (rtx_reuse_manager::set_seen_def): Likewise.
31848         * print-rtl.h (class rtx_reuse_manager): Likewise.
31850 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31852         * bt-load.c (combine_btr_defs): Use auto_bitmap to manage bitmap
31853         lifetime.
31854         (migrate_btr_def): Likewise.
31855         * cfgloop.c (get_loop_body_in_bfs_order): Likewise.
31856         * df-core.c (loop_post_order_compute): Likewise.
31857         (loop_inverted_post_order_compute): Likewise.
31858         * hsa-common.h: Likewise.
31859         * hsa-gen.c (hsa_bb::~hsa_bb): Likewise.
31860         * init-regs.c (initialize_uninitialized_regs): Likewise.
31861         * ipa-inline.c (resolve_noninline_speculation): Likewise.
31862         (inline_small_functions): Likewise.
31863         * ipa-reference.c (ipa_reference_write_optimization_summary): Likewise.
31864         * ira.c (combine_and_move_insns): Likewise.
31865         (build_insn_chain): Likewise.
31866         * loop-invariant.c (find_invariants): Likewise.
31867         * lower-subreg.c (propagate_pseudo_copies): Likewise.
31868         * predict.c (tree_predict_by_opcode): Likewise.
31869         (predict_paths_leading_to): Likewise.
31870         (predict_paths_leading_to_edge): Likewise.
31871         (estimate_loops_at_level): Likewise.
31872         (estimate_loops): Likewise.
31873         * shrink-wrap.c (try_shrink_wrapping): Likewise.
31874         (spread_components): Likewise.
31875         * tree-cfg.c (remove_edge_and_dominated_blocks): Likewise.
31876         * tree-loop-distribution.c (rdg_build_partitions): Likewise.
31877         * tree-predcom.c (tree_predictive_commoning_loop): Likewise.
31878         * tree-ssa-coalesce.c (coalesce_ssa_name): Likewise.
31879         * tree-ssa-phionlycprop.c (pass_phi_only_cprop::execute): Likewise.
31880         * tree-ssa-pre.c (remove_dead_inserted_code): Likewise.
31881         * tree-ssa-sink.c (nearest_common_dominator_of_uses): Likewise.
31882         * tree-ssa-threadupdate.c (compute_path_counts): Likewise.
31883         (mark_threaded_blocks): Likewise.
31884         (thread_through_all_blocks): Likewise.
31885         * tree-ssa.c (verify_ssa): Likewise.
31886         (execute_update_addresses_taken): Likewise.
31887         * tree-ssanames.c (verify_ssaname_freelists): Likewise.
31889 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31891         * cfganal.c (mark_dfs_back_edges): Replace manual stack with
31892         auto_vec.
31893         (post_order_compute): Likewise.
31894         (inverted_post_order_compute): Likewise.
31895         (pre_and_rev_post_order_compute_fn): Likewise.
31897 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31899         * genrecog.c (int_set::int_set): Explicitly construct our
31900         auto_vec base class.
31901         * vec.h (auto_vec::auto_vec): New constructor.
31903 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31905         * bitmap.h (class auto_bitmap): New constructor taking
31906         bitmap_obstack * argument.
31908 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31910         * bitmap.h (class auto_bitmap): Change type of m_bits to
31911         bitmap_head, and adjust ctor / dtor and member operators.
31913 2017-05-13  Uros Bizjak  <ubizjak@gmail.com>
31915         * compare-elim.c (equivalent_reg_at_start): Return NULL_RTX
31916         when returned register mode doesn't match original mode.
31918 2017-05-12  Jeff Law  <law@redhat.com>
31919             Jakub Jelinek  <jakub@redhat.com>
31921         * config/mn10300/mn10300.c (mn10300_match_ccmode): Fix where
31922         we look for cc setter after the compare-elim changes.
31923         * config/mn10300/mn10300.md (addsi3_flags): Fix order of patterns
31924         within the vector to match what compare-elim now expects.
31925         (subsi3_flags, andsi3_flags, iorsi3_flags): Likewise.
31926         (xorsi3_flags, one_cmplsi2_flags): Likewise.
31928         * config/rx/rx.c (rx_match_ccmode): Fix where we look cc setter
31929         after the compare-elim changes.
31930         * config/rx/rx.md (abssi2_flags): Fix order of patterns within
31931         the vector to match what compare-elim now expects.
31932         (addsi3_flags, adc_flags, addsi3_flags peepholes): Likewise.
31933         (andsi3_flags, negsi2_flags, one_cmplsi2_flags): Likewise.
31934         (iorsi3_flags, rotlsi3_flags, rotrsi3_flags): Likewise.
31935         (ashrsi3_flags, lshrsi3_flags, ashlsi3_flags): Likewise.
31936         (ssaddsi3, subsi3_flags, sbb_flags, xorsi3_flags): Likewise.
31938         * config/visium/visium.c (single_set_and_flags): Fix where
31939         we look for cc setter after the compare-elim changes.
31940         * config/visium/visium.md (flags_subst_logic): Fix order of patterns
31941         with the vector to match what compare-elim now expects.
31942         (flags_subst_arith, add<mode>3_insn_set_carry): Likewise.
31943         (add<mode>3_insn_set_overflow, addsi3_insn_set_carry): Likewise.
31944         (addsi3_insn_set_overflow, sub<mode>3_insn_set_carry): Likewise.
31945         (sub<mode>3_insn_set_overflow, subsi3_insn_set_carry): Likewise.
31946         (subsi3_insn_set_overflow, negsi2_insn_set_carry): Likewise.
31947         (neg<mode>2_insn_set_overflow): Likewise.
31949 2017-05-12  Jim Wilson  <jim.wilson@linaro.org>
31951         PR middle-end/79794
31952         * expmed.c (extract_bit_field_1): Add alt_rtl argument.  Before
31953         maybe_expand_insn call, set ops[0].target.  If still set after call,
31954         set alt_rtl.  Add extra arg to recursive calls.
31955         (extract_bit_field): Add alt_rtl argument.  Pass to
31956         extract_bit_field.
31957         * expmed.h (extract_bit_field): Fix prototype.
31958         * expr.c (emit_group_load_1, copy_blkmode_from_reg)
31959         (copy_blkmode_to_reg, read_complex_part, store_field): Pass extra NULL
31960         to extract_bit_field_calls.
31961         (expand_expr_real_1): Pass alt_rtl to expand_expr_real instead of 0.
31962         Pass alt_rtl to extract_bit_field calls.
31963         * calls.c (store_unaligned_arguments_into_psuedos)
31964         load_register_parameters): Pass extra NULL to extract_bit_field calls.
31965         * optabs.c (maybe_legitimize_operand): Clear op->target when call
31966         gen_reg_rtx.
31967         * optabs.h (struct expand_operand): Add target bitfield.
31969 2017-05-12  Uros Bizjak  <ubizjak@gmail.com>
31971         * compare-elim.c (try_eliminate_compare): Canonicalize
31972         operation with embedded compare to
31973         [(set (reg:CCM) (compare:CCM (operation) (immediate)))
31974          (set (reg) (operation)].
31976         * config/i386/i386.c (TARGET_FLAGS_REGNUM): New define.
31978 2017-05-12  Uros Bizjak  <ubizjak@gmail.com>
31980         PR target/80723
31981         * config/i386/i386.c (ix86_rtx_cost) [case PLUS]: Ignore the
31982         cost of adding a carry flag for ADC instruction.
31983         [case MINUS]: Ignore the cost of subtracting a carry flag
31984         for SBB instruction.
31986 2017-05-12  Steven Munroe  <munroesj@gcc.gnu.org>
31988         * config.gcc (powerpc*-*-*): Add bmi2intrin.h, bmiintrin.h,
31989         and x86intrin.h
31990         * config/rs6000/bmiintrin.h: New file.
31991         * config/rs6000/bmi2intrin.h: New file.
31992         * config/rs6000/x86intrin.h: New file.
31994 2017-05-12  Jeff Law  <law@redhat.com>
31996         * tree-vrp.c (vrp_dom_walker::before_dom_children): Push unwinding
31997         markers.
31999 2017-05-12  Peter Bergner  <bergner@vnet.ibm.com>
32001         PR middle-end/80707
32002         * tree-cfg.c: Remove cfg edges of unreachable case statements.
32004 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
32006         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling for
32007         early expansion of vector divide builtins.
32008         (builtin_function_type): Add VSX_BUILTIN_UDIV_V2DI to the list of
32009         builtins identified as having unsigned arguments.
32011 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
32013         * config/rs6000/rs6000.c (gimple-fold.h): New #include.
32014         (rs6000_gimple_fold_builtin): Add handling for early GIMPLE
32015         expansion of vector logical operations (and, andc, or, xor,
32016         nor, orc, nand).
32018 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
32020         * gimple-fold.c (create_tmp_reg_or_ssa_name): Remove static declaration.
32021         * gimple-fold.h (create_tmp_reg_or_ssa_name): New prototype.
32023 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
32025         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling for
32026         early GIMPLE expansion of vector multiplies.
32028 2017-05-12  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
32030         * config/arm/arm.md (movsi): Add TARGET_32BIT in addition to the
32031         TARGET_HAVE_MOVT conditional.
32032         (movt splitter): Likewise.
32034 2017-05-12  Richard Biener  <rguenther@suse.de>
32036         * tree-ssa-sccvn.h (has_VN_INFO): Declare.
32037         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
32038         Fold all stmts not inplace.
32040 2017-05-12  Richard Biener  <rguenther@suse.de>
32042         PR tree-optimization/80713
32043         * tree-ssa-pre.c (remove_dead_inserted_code): Clear
32044         inserted_exprs bit for not removed stmts.
32046 2017-05-12  Thomas Schwinge  <thomas@codesourcery.com>
32048         PR middle-end/69921
32049         * tree-parloops.c (create_parallel_loop): Set "oacc kernels
32050         parallelized" attribute for parallelized OpenACC kernels.
32051         * omp-offload.c (execute_oacc_device_lower): Use it.
32053         * omp-expand.c (expand_omp_target) <GF_OMP_TARGET_KIND_OACC_KERNELS>:
32054         Set "oacc kernels" attribute.
32055         * omp-general.c (oacc_set_fn_attrib): Remove is_kernel formal
32056         parameter.  Adjust all users.
32057         (oacc_fn_attrib_kernels_p): Remove function.
32058         * omp-offload.c (execute_oacc_device_lower): Look for "oacc
32059         kernels" attribute instead of calling oacc_fn_attrib_kernels_p.
32060         * tree-ssa-loop.c (gate_oacc_kernels): Likewise.
32061         * tree-parloops.c (create_parallel_loop): If oacc_kernels_p,
32062         assert "oacc kernels" attribute is set.
32064 2017-05-11  Carl Love  <cel@us.ibm.com>
32066         * config/rs6000/rs6000-c: Add support for built-in functions
32067         vector unsigned char vec_popcnt (vector signed char)
32068         vector unsigned char vec_popcnt (vector unsigned char)
32069         vector unsigned short vec_popcnt (vector signed short)
32070         vector unsigned short vec_popcnt (vector unsigned short)
32071         vector unsigned int vec_popcnt (vector signed int)
32072         vector unsigned int vec_popcnt (vector unsigned int)
32073         vector unsigned long long vec_popcnt (vector signed long long)
32074         vector unsigned long long vec_popcnt (vector unsigned long long)
32075         vector signed long long vec_slo (vector signed long long,
32076                                          vector signed char)
32077         vector signed long long vec_slo (vector signed long long,
32078                                          vector unsigned char)
32079         vector unsigned long long vec_slo (vector unsigned long long,
32080                                            vector signed char)
32081         vector unsigned long long vec_slo (vector unsigned long long,
32082                                            vector unsigned char)
32083         * config/rs6000/rs6000-builtin.def: Add definitions for VPOPCNTUB,
32084         VPOPCNTUH, VPOPCNTUW, and VPOPCNTUD overloads.
32085         * config/rs6000/altivec.h: Add define for vec_popcnt, vec_popcntb,
32086         vec_popcnth, vec_popcntw and vec_popcntd built-in functions.
32087         * doc/extend.texi: Update the built-in documentation file for the
32088         new built-in functions.
32090 2017-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
32092         * attribs.h (sorted_attr_string): Move machine independent
32093         functions for target clone support from the i386 port to common
32094         code.  Rename ix86_function_versions to common_function_versions.
32095         Rename make_name to make_unique_name.
32096         (common_function_versions): Likewise.
32097         (make_unique_name): Likewise.
32098         (make_dispatcher_decl): Likewise.
32099         (is_function_default_version): Likewise.
32100         * attribs.c (attr_strcmp): Likewise.
32101         (sorted_attr_string): Likewise.
32102         (common_function_versions): Likewise.
32103         (make_unique_name): Likewise.
32104         (make_dispatcher_decl): Likewise.
32105         (is_function_default_version): Likewise.
32106         * config/i386/i386.c (attr_strcmp): Likewise.
32107         (sorted_attr_string): Likewise.
32108         (ix86_function_versions): Likewise.
32109         (make_name): Likewise.
32110         (make_dispatcher_decl): Likewise.
32111         (is_function_default_version): Likewise.
32112         (TARGET_OPTION_FUNCTION_VERSIONS): Update target function hook.
32114 2017-05-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32116         PR target/80695
32117         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost):
32118         Account for direct move costs for vec_construct of integer
32119         vectors.
32121 2017-05-11  Uros Bizjak  <ubizjak@gmail.com>
32123         PR target/80706
32124         * config/i386/sync.md (UNSPEC_LDX_ATOMIC): New unspec.
32125         (UNSPEC_STX_ATOMIC): Ditto.
32126         (loaddi_via_sse): New insn.
32127         (storedi_via_sse): Ditto.
32128         (atomic_loaddi_fpu): Emit loaddi_via_sse and storedi_via_sse.
32129         Update corresponding peephole2 patterns.
32130         (atomic_storedi_fpu): Ditto.
32132 2017-05-11  Julia Koval  <julia.koval@intel.com>
32134         * config/i386/avx512fintrin.h (_mm_mask_rsqrt14_sd)
32135         (_mm_maskz_rsqrt14_sd, _mm_mask_rsqrt14_ss, _mm_maskz_rsqrt14_ss):
32136         New intrinsics.
32137         * config/i386/i386-builtin.def (__builtin_ia32_rsqrt14sd_mask)
32138         (__builtin_ia32_rsqrt14ss_mask): New builtins.
32139         * config/i386/sse.md (rsqrt14_<mode>_mask): New pattern.
32141 2017-05-11  Nathan Sidwell  <nathan@acm.org>
32143         * graphite-poly.c: Include dumpfile.h.
32145         * dumpfle.h (dump_function): Declare here ...
32146         * tree-dump.h (dump_function): ... not here.
32147         * dumpfile.c: #include tree-cfg.h.
32148         (dump_function): Move here from ...
32149         * tree-dump.c (dump_function): ... here.
32150         * gimplify.c: #include splay-tree.h, not tree-dump.h.
32151         * graphite-poly.c: Don't include tree-dump.h.
32152         * cgraphclones.c: Include dumpfile.h not tree-dump.h.
32153         * print-tree.c: Likewise.
32154         * stor-layout.c: Likewise.
32155         * tree-nested.c: Likewise.
32157         * dumpfile.c (dump_start): Use TDF_FLAGS.
32158         (dump_enable_all): Fix TDF_KIND check thinko.
32160 2017-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
32162         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
32163         array entries to represent two legal parameterizations of the
32164         overloaded __builtin_cmpb function, as represented by the
32165         P6_OV_BUILTIN_CMPB constant.
32166         (altivec_resolve_overloaded_builtin): Add special case handling
32167         for the __builtin_cmpb function, as represented by the
32168         P6_OV_BUILTIN_CMPB constant.
32169         * config/rs6000/rs6000-builtin.def (BU_P6_2): New macro.
32170         (BU_P6_64BIT_2): New macro.
32171         (BU_P6_OVERLOAD_2): New macro
32172         (CMPB_32): Add 32-bit compare-bytes support for 32-bit only targets.
32173         (CMPB): Add 64-bit compare-bytes support for 32-bit and 64-bit targets.
32174         (CMPB): Add overload support to represent both 32-bit and 64-bit
32175         compare-bytes function.
32176         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add
32177         support for TARGET_CMPB.
32178         * config/rs6000/rs6000.h: Add support for RS6000_BTM_CMPB.
32179         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
32180         documentation of the __builtin_cmpb overloaded built-in function.
32182 2017-05-11  Richard Biener  <rguenther@suse.de>
32184         PR tree-optimization/80705
32185         * tree-vect-data-refs.c (vect_analyze_data_refs): DECL_NONALIASED
32186         bases are not vectorizable.
32188 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32190         * tree-ssa-loop-ivopts.c (determine_set_costs): Skip non-interger
32191         when counting register pressure.
32193 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32195         * tree-ssa-loop-ivopts.c (compare_cost_pair): New.
32196         (iv_ca_more_deps): Renamed to ...
32197         (iv_ca_compare_deps): ... this.
32198         (iv_ca_extend): Extend iv_ca if NEW_CP is cheaper than OLD_CP.
32200 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32202         * tree-ssa-loop-ivopts.c (find_interesting_uses): Move inv vars dump
32203         to ...
32204         (determine_group_iv_costs): ... here.
32205         (find_inv_vars_cb): Record inv var if it's not recorded before.
32207 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32209         * tree-ssa-loop-ivopts.c (allow_ip_end_pos_p): Refine comments.
32210         (get_shiftadd_cost): Ditto.
32212 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32214         * tree-ssa-address.c: Include header file.
32215         (move_hint_to_base): Return TRUE if BASE_HINT is moved to memory
32216         address.
32217         (add_to_parts): Refactor.
32218         (addr_to_parts): New parameter.  Update use of move_hint_to_base.
32219         (create_mem_ref): Update use of addr_to_parts.  Re-associate addr
32220         in new order.
32222 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32224         PR tree-optimization/53090
32225         * tree-ssa-loop-ivopts.c (enum comp_iv_rewrite): New enum value
32226         COMP_IV_EXPR_2.
32227         (extract_cond_operands): Detect condition with IV on both sides
32228         and return COMP_IV_EXPR_2.
32229         (find_interesting_uses_cond): Add iv_use for both IVs in condition.
32230         (rewrite_use_compare): Simplify by removing call to function
32231         extract_cond_operands.
32233 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32235         * tree-ssa-loop-ivopts.c (enum comp_iv_rewrite): New.
32236         (extract_cond_operands): Detect condition comparing against non-
32237         invariant bound and return appropriate enum value.
32238         (find_interesting_uses_cond): Update use of extract_cond_operands.
32239         Handle its return value accordingly.
32240         (determine_group_iv_cost_cond, rewrite_use_compare): Ditto.
32242 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32244         * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Re-associate
32245         nonlinear iv_use computation in loop invariant sensitive way.
32247 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32249         * tree-ssa-loop-ivopts.c (relate_compare_use_with_all_cands): New.
32250         (find_iv_candidates): Call relate_compare_use_with_all_cands.
32252 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32254         * tree-ssa-loop-ivopts.c (struct iv_cand): New field inv_exprs.
32255         (dump_cand): Support iv_cand.inv_exprs.
32256         (add_candidate_1): Record invariant exprs in iv_cand.inv_exprs
32257         for candidates.
32258         (iv_ca_set_no_cp, iv_ca_set_cp, free_loop_data): Support
32259         iv_cand.inv_exprs.
32261 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32263         * tree-ssa-loop-ivopts.c (multiplier_allowed_in_address_p): Move
32264         from ...
32265         * tree-ssa-address.c (multiplier_allowed_in_address_p): ... to here
32266         as local function.  Include necessary header files.
32267         * tree-ssa-loop-ivopts.h (multiplier_allowed_in_address_p): Delete.
32269 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32271         * tree-ssa-loop-ivopts.c (autoinc_possible_for_pair): Simplify.
32273 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32275         * tree-ssa-loop-ivopts.c (force_expr_to_var_cost): Handle more
32276         operators: TRUNC_DIV_EXPR, BIT_AND_EXPR, BIT_IOR_EXPR, LSHIFT_EXPR,
32277         RSHIFT_EXPR and BIT_NOT_EXPR.
32279 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32281         * tree-ssa-loop-ivopts.c (get_loop_invariant_expr): Simplify.
32282         (adjust_setup_cost): New parameter supporting round up adjustment.
32283         (struct address_cost_data): Delete.
32284         (force_expr_to_var_cost): Don't bound cost with spill_cost.
32285         (split_address_cost, ptr_difference_cost): Delete.
32286         (difference_cost, compare_aff_trees, record_inv_expr): Delete.
32287         (struct ainc_cost_data): New struct.
32288         (get_address_cost_ainc): New function.
32289         (get_address_cost, get_computation_cost): Reimplement.
32290         (determine_group_iv_cost_address): Record inv_expr for all uses of
32291         a group.
32292         (determine_group_iv_cost_cond): Call get_loop_invariant_expr.
32293         (iv_ca_has_deps): Reimplemented to ...
32294         (iv_ca_more_deps): ... this.  Check if NEW_CP introduces more deps
32295         than OLD_CP.
32296         (iv_ca_extend): Call iv_ca_more_deps.
32298 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32300         * tree-ssa-address.c (struct mem_address): Move to header file.
32301         (valid_mem_ref_p, move_fixed_address_to_symbol): Make it global.
32302         * tree-ssa-address.h (struct mem_address): Move from C file.
32303         (valid_mem_ref_p, move_fixed_address_to_symbol): Declare.
32305 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32307         * tree-affine.h (aff_combination_type): New interface.
32308         (aff_combination_zero_p): Remove static.
32309         (aff_combination_const_p): New interface.
32310         (aff_combination_singleton_var_p): New interfaces.
32312 2017-05-11  Richard Biener  <rguenther@suse.de>
32314         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
32315         Skip unreachable blocks and destinations.
32316         (eliminate): Move stmt removal and fixup ...
32317         (fini_eliminate): ... here.  Skip inserted exprs.
32318         (pass_pre::execute): Move fini_pre after fini_eliminate.
32319         * tree-ssa-tailmerge.c: Include tree-cfgcleanup.h.
32320         (tail_merge_optimize): Run cleanup_tree_cfg if requested by
32321         PRE to get rid of dead code that has invalid SSA form and
32322         split critical edges again.
32324 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32326         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
32328 2017-05-11  Richard Biener  <rguenther@suse.de>
32330         * passes.c (execute_function_todo): Verify loops if they are
32331         said to be up-to-date.
32332         * cfgexpand.c (pass_expand::execute): Discard loops for -dx.
32333         * trans-mem.c (pass_tm_edges::execute): Mark loops for fixup.
32335 2017-05-10  John David Anglin  <danglin@gcc.gnu.org>
32337         PR target/80090
32338         * config/pa/pa.c (pa_assemble_integer): When outputting a SYMBOL_REF,
32339         handle calling assemble_external ourself.
32341         PR target/79027
32342         * config/pa/pa.c (pa_cannot_change_mode_class): Reject changes to/from
32343         modes with zero size.  Enhance comment.
32345 2017-05-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32347         * config/rs6000/rs6000.c (altivec_init_builtins): Define POWER8
32348         built-ins for vec_xl and vec_xst with short and char pointer
32349         arguments.
32351 2017-05-10  Sebastian Peryt  <sebastian.peryt@intel.com>
32353         * config/i386/avx512fintrin.h (_mm_mask_max_round_sd)
32354         (_mm_maskz_max_round_sd, _mm_mask_max_round_ss)
32355         (_mm_maskz_max_round_ss, _mm_mask_min_round_sd)
32356         (_mm_maskz_min_round_sd, _mm_mask_min_round_ss)
32357         (_mm_maskz_min_round_ss): New intrinsics.
32358         * config/i386/i386-builtin-types.def (V2DF, V2DF, V2DF, V2DF, UQI, INT)
32359         (V4SF, V4SF, V4SF, V4SF, UQI, INT): New function type aliases.
32360         * config/i386/i386-builtin.def (__builtin_ia32_maxsd_mask_round)
32361         (__builtin_ia32_maxss_mask_round, __builtin_ia32_minsd_mask_round)
32362         (__builtin_ia32_minss_mask_round): New builtins.
32363         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32364         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
32365         * config/i386/sse.md (<sse>_vm<code><mode>3<round_saeonly_name>):
32366         Rename to ...
32367         (<sse>_vm<code><mode>3<mask_name><round_saeonly_name>): ... this.
32368         (v<maxmin_float><ssescalarmodesuffix>\t{<round_saeonly_op3>%2, %1, %0|%0, %1, %<iptr>2<round_saeonly_op3>}):
32369         Change to ...
32370         (v<maxmin_float><ssescalarmodesuffix>\t{<round_saeonly_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_saeonly_mask_op3>}):
32371         ... this.
32373 2017-05-10  Sebastian Peryt  <sebastian.peryt@intel.com>
32375         * config/i386/avx512fintrin.h (_mm_mask_mul_round_sd)
32376         (_mm_maskz_mul_round_sd, _mm_mask_mul_round_ss)
32377         (_mm_maskz_mul_round_ss, _mm_mask_div_round_sd)
32378         (_mm_maskz_div_round_sd, _mm_mask_div_round_ss)
32379         (_mm_maskz_div_round_ss, _mm_mask_mul_sd, _mm_maskz_mul_sd)
32380         (_mm_mask_mul_ss, _mm_maskz_mul_ss, _mm_mask_div_sd)
32381         (_mm_maskz_div_sd, _mm_mask_div_ss, _mm_maskz_div_ss): New intrinsics.
32382         * config/i386/i386-builtin-types.def (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32383         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): New function type aliases.
32384         * config/i386/i386-builtin.def (__builtin_ia32_divsd_mask_round)
32385         (__builtin_ia32_divss_mask_round, __builtin_ia32_mulsd_mask_round)
32386         (__builtin_ia32_mulss_mask_round): New builtins.
32387         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32388         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
32389         * config/i386/sse.md (<sse>_vm<multdiv_mnemonic><mode>3<round_name>):
32390         Rename to ...
32391         (<sse>_vm<multdiv_mnemonic><mode>3<mask_name><round_name>): ... this.
32392         (v<multdiv_mnemonic><ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|%0, %1, %<iptr>2<round_op3>}):
32393         Change to ...
32394         (v<multdiv_mnemonic><ssescalarmodesuffix>\t{<round_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}):
32395         ... this.
32397 2017-05-10  Julia Koval  <julia.koval@intel.com>
32399         * config/i386/avxintrin.h (_mm256_set_m128, _mm256_set_m128d)
32400         (_mm256_set_m128i, _mm256_setr_m128, _mm256_setr_m128d)
32401         (_mm256_setr_m128i): New intrinsics.
32403 2017-05-10  Julia Koval  <julia.koval@intel.com>
32405         * config/i386/avx512fintrin.h (_mm_mask_rcp14_sd)
32406         (_mm_maskz_rcp14_sd, _mm_mask_rcp14_ss)
32407         (_mm_maskz_rcp14_ss): New intrinsics.
32408         * config/i386/i386-builtin.def (__builtin_ia32_rcp14sd_mask)
32409         (__builtin_ia32_rcp14ss_mask): New builtins.
32410         * config/i386/sse.md (srcp14<mode>_mask): New pattern.
32412 2017-05-10  Peter Bergner  <bergner@vnet.ibm.com>
32414         PR tree-optimization/51513
32415         * tree-cfg.c (gimple_seq_unreachable_p): New function.
32416         (assert_unreachable_fallthru_edge_p): Use it.
32417         (group_case_labels_stmt): Likewise.
32418         * tree-cfg.h: Prototype it.
32419         * stmt.c: Include cfghooks.h and tree-cfg.h.
32420         (emit_case_dispatch_table) <gap_label>: New local variable.
32421         Use it to fill dispatch table gaps.
32422         Test for default_label before updating probabilities.
32423         (expand_case) <default_label>: Remove unneeded initialization.
32424         Test for unreachable default case statement and remove its edge.
32425         Set default_label accordingly.
32426         * tree-ssa-ccp.c (optimize_unreachable): Update comment.
32428 2017-05-10  Carl Love  <cel@us.ibm.com>
32430         * config/rs6000/rs6000-c: Add support for built-in functions
32431         vector signed char      vec_neg (vector signed char)
32432         vector signed short int vec_neg (vector short int)
32433         vector signed int       vec_neg (vector signed int)
32434         vector signed long long vec_neg (vector signed long long)
32435         vector float            vec_neg (vector float)
32436         vector double           vec_neg (vector double)
32437         * config/rs6000/rs6000-builtin.def: Add definitions for NEG function
32438         overload.
32439         * config/rs6000/altivec.h: Add define for vec_neg
32440         * doc/extend.texi: Update the built-in documentation for the
32441         new built-in functions.
32443 2017-05-10  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
32445         PR tree-optimization/77644
32446         * match.pd (sqrt(x) cmp sqrt(y) -> x cmp y): New pattern.
32448 2017-05-10  Nathan Sidwell  <nathan@acm.org>
32450         * dumpfile.h (TDI_lang_all): New.
32451         (TDF_KIND): New. Renumber others
32452         (TDF_LANG, TDF_TREE, TDF_RTL, TDF_IPA): Enumerate value, rather
32453         than bits.
32454         * dumpfile.c (dump_files): Mark language dumps as TDF_LANG.  add
32455         lang-all.
32456         (get_dump_file_name): Adjust suffix generation.
32457         (dump_enable_all): Use TDF_KIND.
32458         * doc/invoke.texi (-fdump-lang-all): Document.
32460         * dumpfile.h: Tabify.
32462 2017-05-10  Wilco Dijkstra  <wdijkstr@arm.com>
32464         PR target/80671
32465         * config/aarch64/cortex-a57-fma-steering.c (merge_forest):
32466         Move member access before delete.
32468 2017-05-10  Alexandre Oliva  <aoliva@redhat.com>
32470         * tree-inline.c (expand_call_inline): Split block at stmt
32471         before the call.
32473 2017-05-09  Michael Meissner  <meissner@linux.vnet.ibm.com>
32475         PR target/68163
32476         * config/rs6000/rs6000.md (f32_lr): Delete mode attributes that
32477         are now unused after splitting mov{sf,sd}_hardfloat.
32478         (f32_lr2): Likewise.
32479         (f32_lm): Likewise.
32480         (f32_lm2): Likewise.
32481         (f32_li): Likewise.
32482         (f32_li2): Likewise.
32483         (f32_lv): Likewise.
32484         (f32_sr): Likewise.
32485         (f32_sr2): Likewise.
32486         (f32_sm): Likewise.
32487         (f32_sm2): Likewise.
32488         (f32_si): Likewise.
32489         (f32_si2): Likewise.
32490         (f32_sv): Likewise.
32491         (f32_dm): Likewise.
32492         (f32_vsx): Likewise.
32493         (f32_av): Likewise.
32494         (mov<mode>_hardfloat): Split into separate movsf and movsd pieces.
32495         For movsf, order stores so the VSX stores occur before the GPR
32496         store which encourages the register allocator to use a traditional
32497         FPR instead of a GPR.  For movsd, order the stores so that the GPR
32498         store comes before the VSX stores to allow the power6 to work.
32499         This is due to the power6 not having a 32-bit integer store
32500         instruction from a FPR.
32501         (movsf_hardfloat): Likewise.
32502         (movsd_hardfloat): Likewise.
32504 2017-05-09  Martin Sebor  <msebor@redhat.com>
32506         PR translation/80280
32507         * config/sol2-c.c (cmn_err_flag_specs): Initialize new data member
32508         added in r247778.
32510         PR translation/80280
32511         * config/i386/msformat-c.c (ms_printf_flag_specs): Initialize new
32512         data member added in r247778.
32513         (ms_scanf_flag_specs, ms_strftime_flag_specs): Same.
32515 2017-05-09  Nathan Sidwell  <nathan@acm.org>
32517         * tree.h (tree_fits_shwi_p, tree_fits_uhwi_p): Unconditionally pure.
32519         * ipa-devirt.c (default_hash_traits<type_pair>): Skip struct-scope
32520         typedefs.
32522 2017-05-09  Marek Polacek  <polacek@redhat.com>
32524         * doc/invoke.texi: Fix typo.
32526 2017-05-09  Richard Biener  <rguenther@suse.de>
32528         * tree-vrp.c (vrp_val_is_max): Adjust comment.
32529         (vrp_val_is_min): Likewise.
32530         (set_value_range_to_value): Likewise.
32531         (set_value_range_to_nonnegative): Likewise.
32532         (gimple_assign_nonzero_p): Likewise.
32533         (gimple_stmt_nonzero_p): Likewise.
32534         (vrp_int_const_binop): Likewise.  Remove unreachable case.
32535         (adjust_range_with_scev): Adjust comments.
32536         (compare_range_with_value): Likewise.
32537         (extract_range_from_phi_node): Likewise.
32538         (test_for_singularity): Likewise.
32540 2017-05-09  Richard Biener  <rguenther@suse.de>
32542         * tree-vrp.c (get_single_symbol): Add assert that we don't
32543         get overflowed constants as invariant part.
32544         (compare_values_warnv): Add comment before the TREE_NO_WARNING
32545         checks.  Use wi::cmp instead of recursing for integer constants.
32546         (compare_values): Just ignore whether we assumed undefined
32547         overflow instead of failing the compare.
32548         (extract_range_for_var_from_comparison_expr): Add comment before the
32549         TREE_NO_WARNING sets.
32550         (test_for_singularity): Likewise.
32551         (extract_range_from_comparison): Do not disable optimization
32552         when we assumed undefined overflow.
32553         (extract_range_basic): Remove init of unused var.
32555 2017-05-09  Richard Biener  <rguenther@suse.de>
32557         * tree-vrp.c (vrp_int_const_binop): Use wide-ints and simplify.
32558         (extract_range_from_multiplicative_op_1): Adjust.
32559         (extract_range_from_binary_expr_1): Use int_const_binop.
32561 2017-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
32563         PR target/80101
32564         * config/rs6000/power6.md: Replace store_data_bypass_p calls with
32565         rs6000_store_data_bypass_p in seven define_bypass directives and
32566         in several comments.
32567         * config/rs6000/rs6000-protos.h: Add prototype for
32568         rs6000_store_data_bypass_p function.
32569         * config/rs6000/rs6000.c (rs6000_store_data_bypass_p): New
32570         function implements slightly different (rs6000-specific) semantics
32571         than store_data_bypass_p, returning false rather than aborting
32572         with assertion error when arguments do not satisfy the
32573         requirements of store data bypass.
32574         (rs6000_adjust_cost): Replace six calls of store_data_bypass_p with
32575         rs6000_store_data_bypass_p.
32577 2017-05-08  Max Filippov  <jcmvbkbc@gmail.com>
32579         * config/xtensa/xtensa-protos.h
32580         (xtensa_initial_elimination_offset): New declaration.
32581         * config/xtensa/xtensa.c (xtensa_initial_elimination_offset):
32582         New function. Move its body from the INITIAL_ELIMINATION_OFFSET
32583         macro definition, add case for FRAME_POINTER_REGNUM when
32584         FRAME_GROWS_DOWNWARD.
32585         * config/xtensa/xtensa.h (FRAME_GROWS_DOWNWARD): New macro definition.
32586         (INITIAL_ELIMINATION_OFFSET): Replace body with call to
32587         xtensa_initial_elimination_offset.
32589 2017-05-08  Nathan Sidwell  <nathan@acm.org>
32591         * doc/invoke.texi: Alphabetize -fdump options.
32593 2017-05-08  Martin Sebor  <msebor@redhat.com>
32595         PR translation/80280
32596         * config/sol2-c.c (solaris_pragma_align): Correct quoting.
32598 2017-05-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
32600         * target.def (compute_frame_layout): New optional target hook.
32601         * doc/tm.texi.in (TARGET_COMPUTE_FRAME_LAYOUT): Add hook.
32602         * doc/tm.texi (TARGET_COMPUTE_FRAME_LAYOUT): Add documentation.
32603         * lra-eliminations.c (update_reg_eliminate): Call compute_frame_layout
32604         target hook.
32605         * reload1.c (verify_initial_elim_offsets): Likewise.
32606         * config/arm/arm.c (TARGET_COMPUTE_FRAME_LAYOUT): Define.
32607         (use_simple_return_p): Call arm_compute_frame_layout if needed.
32608         (arm_get_frame_offsets): Split up into this ...
32609         (arm_compute_frame_layout): ... and this function.
32611 2017-05-08  Richard Sandiford  <richard.sandiford@arm.com>
32613         * config/aarch64/constraints.md (Usa): New constraint.
32614         * config/aarch64/aarch64.md (*movsi_aarch64, *movdi_aarch64): Use it.
32616 2017-05-08  Thomas Preud'homme  <thomas.preudhomme@arm.com>
32618         * config.gcc (arm*-*-*): Set TM_MULTILIB_CONFIG from
32619         with_multilib_list after it has been checked.
32621 2017-05-08  Richard Biener  <rguenther@suse.de>
32623         * tree-ssa-pre.c (bitmap_set_and): Avoid bitmap copy.
32624         (bitmap_set_subtract_values): Likewise.
32626 2017-05-08  Richard Biener  <rguenther@suse.de>
32628         * tree-vrp.c (gimple_assign_nonzero_warnv_p): Rename to ...
32629         (gimple_assign_nonzero): ... this and remove strict_overflow_p
32630         argument.
32631         (gimple_stmt_nonzero_warnv_p): Rename to ...
32632         (gimple_stmt_nonzero_p): ... this and remove strict_overflow_p
32633         argument.
32634         (vrp_stmt_computes_nonzero): Remove strict_overflow_p argument.
32635         (extract_range_basic): Adjust, do not disable propagation on
32636         strict overflow sensitive simplification.
32637         (vrp_visit_cond_stmt): Likewise.
32639 2017-05-05  Jan Hubicka  <hubicka@ucw.cz>
32641         * ipa-inline-analysis.c (estimate_function_body_sizes): Recompute
32642         body size unconditionally.
32644 2017-05-07  Jeff Law  <law@redhat.com>
32646         Revert:
32647         2017-05-06  Jeff Law  <law@redhat.com>
32648         PR tree-optimization/78496
32649         * tree-vrp.c (simplify_assert_expr_using_ranges): Remove debugging
32650         code.
32652         PR tree-optimization/78496
32653         * tree-vrp.c (simplify_assert_expr_using_ranges): New function.
32654         (simplify_stmt_using_ranges): Call it.
32655         (vrp_dom_walker::before_dom_children): Extract equivalences
32656         from an ASSERT_EXPR with an equality comparison against a
32657         constant.
32659 2017-05-06  Jeff Law  <law@redhat.com>
32661         PR tree-optimization/78496
32662         * tree-vrp.c (simplify_assert_expr_using_ranges): Remove debugging
32663         code.
32665         PR tree-optimization/78496
32666         * tree-vrp.c (simplify_assert_expr_using_ranges): New function.
32667         (simplify_stmt_using_ranges): Call it.
32668         (vrp_dom_walker::before_dom_children): Extract equivalences
32669         from an ASSERT_EXPR with an equality comparison against a
32670         constant.
32672 2017-05-06  Richard Sandiford  <richard.sandiford@linaro.org>
32674         * lra-constraints.c (lra_copy_reg_equiv): New function.
32675         (split_reg): Use it to copy equivalence information from the
32676         original register to the spill register.
32678 2017-05-06  Richard Sandiford  <richard.sandiford@linaro.org>
32680         PR rtl-optimization/75964
32681         * simplify-rtx.c (simplify_const_relational_operation): Remove
32682         invalid handling of comparisons of integer ABS.
32684 2017-05-06  Uros Bizjak  <ubizjak@gmail.com>
32686         * config/i386/i386.c (ext_80387_constant_init): Do not explicitly
32687         initialize to zero.
32688         (init_regs): Remove declaration.
32689         (function_arg_advance_32): Initialize error_p as boolean variable.
32691 2017-05-05  Nathan Sidwell  <nathan@acm.org>
32693         * store-motion.c (remove_reachable_equiv_notes): Reformat long
32694         lines.  Use for (;;).
32696 2017-05-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32698         * config/rs6000/rs6000.c (rs6000_vect_nonmem): New static var.
32699         (rs6000_init_cost): Initialize rs6000_vect_nonmem.
32700         (rs6000_add_stmt_cost): Update rs6000_vect_nonmem.
32701         (rs6000_finish_cost): Avoid vectorizing simple copy loops with
32702         VF=2 that require versioning.
32704 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32706         * diagnostic.h (CARET_LINE_MARGIN): Convert from macro to const
32707         int.
32709 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32711         * diagnostic.h (diagnostic_override_option_index): Convert from
32712         macro to inline function.
32714 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32716         * diagnostic.c (last_module_changed_p): New function.
32717         (set_last_module): New function.
32718         (diagnostic_report_current_module): Convert macro usage to
32719         the above functions.
32720         * diagnostic.h (diagnostic_context::last_module): Strengthen
32721         from const line_map * to const line_map_ordinary *.
32722         (diagnostic_last_module_changed): Delete macro.
32723         (diagnostic_set_last_module): Delete macro.
32725 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32727         * diagnostic.c (diagnostic_impl): Replace report_diagnostic
32728         with diagnostic_report_diagnostic.
32729         (diagnostic_n_impl_richloc): Likewise.
32730         * diagnostic.h (report_diagnostic): Delete macro.
32731         * rtl-error.c (diagnostic_for_asm): Replace report_diagnostic
32732         with diagnostic_report_diagnostic.
32733         * substring-locations.c (format_warning_va): Likewise.
32735 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32737         * diagnostic.c (diagnostic_report_diagnostic): Eliminate
32738         save/restor of format_spec.  Move option-printing code to...
32739         (print_option_information): ...this new function, and
32740         reimplement by simply printing to the pretty_printer,
32741         rather than appending to the format string.
32743 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32745         * diagnostic.c (diagnostic_report_diagnostic): Split out pragma
32746         handling logic into...
32747         (update_effective_level_from_pragmas): ...this new function.
32749 2017-05-04  Andrew Waterman  <andrew@sifive.com>
32751         * config/riscv/riscv.opt (mstrict-align): New option.
32752         * config/riscv/riscv.h (STRICT_ALIGNMENT): Use it.  Update comment.
32753         (SLOW_UNALIGNED_ACCESS): Define.
32754         (riscv_slow_unaligned_access): Declare.
32755         * config/riscv/riscv.c (riscv_tune_info): Add slow_unaligned_access
32756         field.
32757         (riscv_slow_unaligned_access): New variable.
32758         (rocket_tune_info): Set slow_unaligned_access to true.
32759         (optimize_size_tune_info): Set slow_unaligned_access to false.
32760         (riscv_cpu_info_table): Add entry for optimize_size_tune_info.
32761         (riscv_valid_lo_sum_p): Use TARGET_STRICT_ALIGN.
32762         (riscv_option_override): Set riscv_slow_unaligned_access.
32763         * doc/invoke.texi: Add -mstrict-align to RISC-V.
32765 2017-05-04  Kito Cheng  <kito.cheng@gmail.com>
32767         * config/riscv/riscv.md: Unify indentation.
32769 2017-05-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
32771         PR target/79038
32772         PR target/79202
32773         PR target/79203
32774         * config/rs6000/rs6000.md (u code attribute): Add FIX and
32775         UNSIGNED_FIX.
32776         (extendsi<mode>2): Add support for doing sign extension via
32777         VUPKHSW and XXPERMDI if the value is in Altivec registers and we
32778         don't have ISA 3.0 instructions.
32779         (extendsi<mode>2 splitter): Likewise.
32780         (fix_trunc<mode>si2): If we are at ISA 2.07 (VSX small integer),
32781         generate the normal insns since SImode can now go in vector
32782         registers.  Disallow the special UNSPECs needed for previous
32783         machines to hide SImode being used.  Add new insns
32784         fctiw{,w}_<mode>_smallint if SImode can go in vector registers.
32785         (fix_trunc<mode>si2_stfiwx): Likewise.
32786         (fix_trunc<mode>si2_internal): Likewise.
32787         (fixuns_trunc<mode>si2): Likewise.
32788         (fixuns_trunc<mode>si2_stfiwx): Likewise.
32789         (fctiw<u>z_<mode>_smallint): Likewise.
32790         (fctiw<u>z_<mode>_mem): New combiner pattern to prevent conversion
32791         of floating point to 32-bit integer from doing a direct move to
32792         the GPR registers to do a store.
32793         (fctiwz_<mode>): Break long line.
32795 2017-05-05  Bin Cheng  <bin.cheng@arm.com>
32797         * Makefile.in (GTFILES): Add tree-ssa-loop-ivopts.c.
32798         * tree-ssa-loop-ivopts.c (compute_max_addr_offset): Delete.
32799         (addr_list, addr_offset_valid_p): New.
32800         (split_address_groups): Check offset validity with above function.
32801         (gt-tree-ssa-loop-ivopts.h): Include header file.
32803 2017-05-05  Nathan Sidwell  <nathan@acm.org>
32805         * config.gcc (arm*-*-*): Add missing 'fi'.
32807 2017-05-05  Steve Ellcey  <sellcey@cavium.com>
32809         * doc/invoke.texi (-fopt-info): Explicitly say order of options
32810         included in -fopt-info does not matter.
32811         * doc/optinfo.texi (-fopt-info): Fix description of default
32812         behavour. Explicitly say order of options included in -fopt-info
32813         does not matter.
32815 2017-05-05  Thomas Preud'homme  <thomas.preudhomme@arm.com>
32817         * config.gcc: Allow combinations of aprofile and rmprofile values for
32818         --with-multilib-list.
32819         * config/arm/t-multilib: New file.
32820         * config/arm/t-aprofile: Remove initialization of MULTILIB_*
32821         variables.  Remove setting of ISA and floating-point ABI in
32822         MULTILIB_OPTIONS and MULTILIB_DIRNAMES.  Set architecture and FPU in
32823         MULTI_ARCH_OPTS_A and MULTI_ARCH_DIRS_A rather than MULTILIB_OPTIONS
32824         and MULTILIB_DIRNAMES respectively.  Add comment to introduce all
32825         matches.  Add architecture matches for marvel-pj4 and generic-armv7-a
32826         CPU options.
32827         * config/arm/t-rmprofile: Likewise except for the matches changes.
32828         * doc/install.texi (--with-multilib-list): Document the combination of
32829         aprofile and rmprofile values and warn about pitfalls in doing that.
32831 2017-05-05  Wilco Dijkstra  <wdijkstr@arm.com>
32833         * config/aarch64/aarch64.md (movsi_aarch64): Remove '*' from r=w.
32834         (movdi_aarch64): Likewise.
32836 2017-05-05  Jakub Jelinek  <jakub@redhat.com>
32838         PR tree-optimization/80632
32839         * tree-switch-conversion.c (struct switch_conv_info): Add target_vop
32840         field.
32841         (build_arrays): Initialize it for virtual phis.
32842         (fix_phi_nodes): Use it for virtual phis.
32844         PR tree-optimization/80558
32845         * tree-vrp.c (extract_range_from_binary_expr_1): Optimize
32846         [x, y] op z into [x op, y op z] for op & or | if conditions
32847         are met.
32849 2017-05-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
32850             Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
32852         PR target/71607
32853         * config/arm/arm.md (use_literal_pool): Remove.
32854         (64-bit immediate split): No longer takes cost into consideration
32855         if arm_disable_literal_pool is enabled.
32856         * config/arm/arm.c (arm_tls_referenced_p): Add diagnostic if TLS is
32857         used when arm_disable_literal_pool is enabled.
32858         (arm_max_const_double_inline_cost): Remove use of
32859         arm_disable_literal_pool.
32860         (push_minipool_fix): Add assert.
32861         (arm_reorg): Add return if arm_disable_literal_pool is enabled.
32862         * config/arm/vfp.md (no_literal_pool_df_immediate): New.
32863         (no_literal_pool_sf_immediate): New.
32865 2017-05-05  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
32867         PR tree-optimization/80613
32868         * tree-ssa-dce.c (propagate_necessity): Remove cases for
32869         BUILT_IN_STRDUP and BUILT_IN_STRNDUP.
32871 2017-05-05  Richard Biener  <rguenther@suse.de>
32873         * tree-ssa-pre.c (get_or_alloc_expr_for): Simplify.
32875 2017-05-05  Georg-Johann Lay  <avr@gjlay.de>
32877         * config/avr/avr.md [flag_strict_overflow]: Remove any occurence
32878         of this flag from insn conditions due to removal from r247495.
32880 2017-05-05  Wilco Dijkstra  <wdijkstr@arm.com>
32882         * config/arm/aarch-common.c (arm_early_load_addr_dep_ptr):
32883         New function.
32884         (arm_early_store_addr_dep_ptr): Likewise.
32885         * config/arm/aarch-common-protos.h
32886         (arm_early_load_addr_dep_ptr): Add prototype.
32887         (arm_early_store_addr_dep_ptr): Likewise.
32888         * config/arm/cortex-a53.md: Add new bypasses.
32890 2017-05-05  Jakub Jelinek  <jakub@redhat.com>
32892         * tree.c (next_type_uid): Change type to unsigned.
32893         (type_hash_canon): Decrement back next_type_uid if
32894         freeing a type node with the highest TYPE_UID.  For INTEGER_TYPEs
32895         also ggc_free TYPE_MIN_VALUE, TYPE_MAX_VALUE and TYPE_CACHED_VALUES
32896         if possible.
32898 2017-05-04  Martin Sebor  <msebor@redhat.com>
32900         * builtins.c: Fix a trivial typo in a comment.
32902         PR middle-end/79234
32903         * builtins.c (check_sizes): Adjust to handle reading past the end.
32904         Avoid printing excessive upper bound of ranges.  Use %E to print
32905         tree nodes instead of converting them to %wu.
32906         (expand_builtin_memchr): New function.
32907         (compute_dest_size): Rename...
32908         (compute_objsize): ...to this.
32909         (expand_builtin_memcpy): Adjust.
32910         (expand_builtin_mempcpy): Adjust.
32911         (expand_builtin_strcat): Adjust.
32912         (expand_builtin_strcpy): Adjust.
32913         (check_strncat_sizes): Adjust.
32914         (expand_builtin_strncat): Adjust.
32915         (expand_builtin_strncpy): Adjust and simplify.
32916         (expand_builtin_memset): Adjust.
32917         (expand_builtin_bzero): Adjust.
32918         (expand_builtin_memcmp): Adjust.
32919         (expand_builtin): Handle memcmp.
32920         (maybe_emit_chk_warning): Check strncat just once.
32922 2017-05-04  Martin Sebor  <msebor@redhat.com>
32924         PR preprocessor/79214
32925         PR middle-end/79222
32926         PR middle-end/79223
32927         * builtins.c (check_sizes): Add inlining context and issue
32928         warnings even when -Wno-system-headers is set.
32929         (check_strncat_sizes): Same.
32930         (expand_builtin_strncat): Same.
32931         (expand_builtin_memmove): New function.
32932         (expand_builtin_stpncpy): Same.
32933         (expand_builtin): Handle memmove and stpncpy.
32935 2017-05-04  Bin Cheng  <bin.cheng@arm.com>
32937         * tree-ssa-loop-ivopts.c (struct cost_pair): Remove field inv_expr
32938         which is not used any more.
32940 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32942         * config/aarch64/aarch64.c (generic_tunings): Update prefetch model.
32944 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32946         * config/aarch64/aarch64.c (cortexa35_tunings): Set jump alignment to 4.
32947         (cortexa53_tunings): Likewise.
32948         (cortexa57_tunings): Likewise.
32949         (cortexa72_tunings): Likewise.
32950         (cortexa73_tunings): Likewise.
32952 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32954         * config/aarch64/aarch64.c (generic_tunings): Set jump alignment to 4.
32955         Set loop alignment to 8.
32957 2017-05-04  Martin Sebor  <msebor@redhat.com>
32959         PR translation/80280
32960         * builtins.c (expand_builtin_object_size): Add missing quoting to
32961         %D and like directives.
32962         * hsa-gen.c (hsa_type_for_scalar_tree_type): Same.
32963         (hsa_type_for_tree_type): Same.
32964         (verify_function_arguments): Same.
32965         * symtab.c (symbol_table::change_decl_assembler_name): Same.
32966         * varasm.c (get_section): Same.
32967         (mark_weak): Same.
32969 2017-05-04  Martin Sebor  <msebor@redhat.com>
32971         PR translation/80280
32972         * config/i386/i386.c (ix86_function_versions): Quote a %D directive.
32974 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32976         * config/aarch64/aarch64.c (generic_addrcost_table):
32977         Change HI/TI mode setting.
32979 2017-05-04  Martin Jambor  <mjambor@suse.cz>
32981         PR tree-optimization/80622
32982         * tree-sra.c (comes_initialized_p): New function.
32983         (build_accesses_from_assign): Only set write lazily when
32984         comes_initialized_p is false.
32985         (analyze_access_subtree): Use comes_initialized_p.
32986         (propagate_subaccesses_across_link): Assert !comes_initialized_p
32987         instead of testing for PARM_DECL.
32989 2017-05-04  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
32991         * config/aarch64/aarch64.md (prefetch); Adjust predicate and
32992         constraint on operand 0 to allow more general addressing modes.
32993         Adjust output template.
32994         * config/aarch64/aarch64.c (aarch64_address_valid_for_prefetch_p):
32995         New function.
32996         * config/aarch64/aarch64-protos.h
32997         (aarch64_address_valid_for_prefetch_p): Declare prototype.
32998         * config/aarch64/constraints.md (Dp): New address constraint.
32999         * config/aarch64/predicates.md (aarch64_prefetch_operand): New
33000         predicate.
33002 2017-05-04  Jan Hubicka  <hubicka@ucw.cz>
33004         * ipa-cp.c (perform_estimation_of_a_value): Drop base_time parameter;
33005         update use of estimate_ipcp_clone_size_and_time.
33006         (estimate_local_effects): Update use of
33007         estimate_ipcp_clone_size_and_time and perform_estimation_of_a_value.
33008         * ipa-inline.h (estimate_ipcp_clone_size_and_time): Update prototype.
33009         * ipa-inline-analysis.c (estimate_ipcp_clone_size_and_time):
33010         Return nonspecialized time.
33012 2017-05-04  Richard Biener  <rguenther@suse.de>
33014         * tree-ssa-alias.c (get_continuation_for_phi): Improve looking
33015         for the last VUSE which def dominates the PHI.  Directly call
33016         maybe_skip_until.
33017         (get_continuation_for_phi_1): Remove.
33019 2017-05-04  Richard Sandiford  <richard.sandiford@linaro.org>
33021         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): Add commentary
33022         to explain the use of truncating division.  Cap the number of
33023         iterations to the maximum given by nb_iterations_upper_bound,
33024         if defined.
33026 2017-05-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
33028         * configure.ac (--enable-mingw-wildcard): Add new configurable feature.
33029         * configure: Regenerate.
33030         * config.in: Regenerate.
33031         * config/i386/driver-mingw32.c: new file.
33032         * config/i386/x-mingw32: Add rule to build driver-mingw32.o.
33033         * config.host: Link driver-mingw32.o on MinGW host.
33034         * doc/install.texi: Document new --enable-mingw-wildcard configure
33035         option.
33037 2017-05-04  Marek Polacek  <polacek@redhat.com>
33039         PR tree-optimization/80612
33040         * calls.c (get_size_range): Check for INTEGRAL_TYPE_P.
33042 2017-05-04  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
33043             Andre Simoes Dias Vieira  <andre.simoesdiasvieira@arm.com>
33045         * config/arm/arm.md (movsi): Change TARGET_32BIT to TARGET_HAVE_MOVT.
33046         (movt splitter): Likewise.
33047         * config/arm/arm.c (arm_option_check_internal): Change arm_arch_thumb2
33048         to TARGET_HAVE_MOVT, and merge with -mslow-flash-data check.
33049         (const_ok_for_arm): Change else to else if (TARGET_THUMB2) and add else
33050         block for Thumb-1 with MOVT.
33051         (thumb2_legitimate_address_p): Move code block ...
33052         (can_avoid_literal_pool_for_label_p): ... into this new function.
33053         (thumb1_legitimate_address_p): Add check for TARGET_HAVE_MOVT and
33054         literal pool.
33055         (thumb_legitimate_constant_p): Add conditional on TARGET_HAVE_MOVT
33056         * doc/invoke.texi (-mpure-code): Change "ARMv7-M targets" for
33057         "M-profile targets with the MOVT instruction".
33059 2017-05-04  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
33061         * config/arm/arm-builtins.c (arm_init_builtins): Rename
33062         __builtin_arm_ldfscr to __builtin_arm_get_fpscr, and rename
33063         __builtin_arm_stfscr to __builtin_arm_set_fpscr.
33065 2017-05-04  Martin Liska  <mliska@suse.cz>
33067         * tree-vrp.c (simplify_cond_using_ranges_2): Remove unused
33068         variable cond_code.
33070 2017-05-04  Richard Biener  <rguenther@suse.de>
33072         * tree.c (array_at_struct_end_p): Handle arrays at struct
33073         end with flexarrays more conservatively.  Refactor and treat
33074         arrays of arrays or aggregates more strict.  Fix
33075         VIEW_CONVERT_EXPR handling.  Remove allow_compref argument.
33076         * tree.c (array_at_struct_end_p): Adjust prototype.
33077         * emit-rtl.c (set_mem_attributes_minus_bitpos): Adjust.
33078         * gimple-fold.c (get_range_strlen): Likewise.
33079         * tree-chkp.c (chkp_may_narrow_to_field): Likewise.
33081 2017-05-04  Richard Biener  <rguenther@suse.de>
33083         PR tree-optimization/31130
33084         * tree-vrp.c (needs_overflow_infinity): Remove as always returning
33085         false.
33086         (supports_overflow_infinity): Likewise.
33087         (is_negative_overflow_infinity): Likewise.
33088         (is_positive_overflow_infinity): Likewise.
33089         (is_overflow_infinity): Likewise.
33090         (stmt_overflow_infinity): Likewise.
33091         (overflow_infinity_range_p): Likewise.
33092         (usable_range_p): Remove as always returning true.
33093         (make_overflow_infinity): Remove.
33094         (negative_overflow_infinity): Likewise.
33095         (positive_overflow_infinity): Likewise.
33096         (avoid_overflow_infinity): Likewise.
33097         (set_value_range): Adjust accordingly.
33098         (set_value_range_to_nonnegative): Likewise, remove now unused
33099         overflow_infinity arg.
33100         (vrp_operand_equal_p): Adjust.
33101         (update_value_range): Likewise.
33102         (range_int_cst_singleton_p): Likewise.
33103         (operand_less_p): Likewise.
33104         (compare_values_warnv): Likewise.
33105         (extract_range_for_var_from_comparison_expr): Likewise.
33106         (vrp_int_const_binop): Likewise.
33107         (zero_nonzero_bits_from_vr): Likewise.
33108         (extract_range_from_multiplicative_op_1): Likewise.
33109         (extract_range_from_binary_expr_1): Likewise.
33110         (extract_range_from_unary_expr): Likewise.
33111         (extract_range_from_comparison): Likewise.
33112         (extract_range_basic): Likewise.
33113         (adjust_range_with_scev): Likewise.
33114         (compare_ranges): Likewise.
33115         (compare_range_with_value): Likewise.
33116         (dump_value_range): Likewise.
33117         (test_for_singularity): Likewise, remove strict_overflow_p parameter
33118         never used.
33119         (simplify_cond_using_ranges): Adjust.
33121 2017-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
33123         * brig-builtins.def: Added a builtin for class_f64.
33124         * builtin-types.def: Added a builtin type needed by class_f64.
33126 2017-05-03  Jason Merrill  <jason@redhat.com>
33128         * timevar.def: Add TV_CONSTEXPR.
33130 2017-05-03  David Malcolm  <dmalcolm@redhat.com>
33132         * common.opt (fdiagnostics-parseable-fixits): Fix typo.
33134 2017-05-03  Martin Jambor  <mjambor@suse.cz>
33136         * ipa-prop.c (ipa_update_after_lto_read): Removed.
33137         * ipa-prop.h (ipa_update_after_lto_read): Remove declaration.
33138         * ipa-cp.c (ipcp_propagate_stage): Do not call
33139         ipa_update_after_lto_read.
33140         * ipa-inline.c (ipa_inline): Likewise.
33142 2017-05-03  Martin Jambor  <mjambor@suse.cz>
33144         * ipa-prop.h (ipa_edge_args): Make a class.  Mark with for_user GTY
33145         tag.  Added a default constructor and a destructor.
33146         (ipa_edge_args_sum_t): New class;
33147         (ipa_edge_args_sum): Declare.
33148         (ipa_edge_args_vector): Remove declaration.
33149         (IPA_EDGE_REF): Use ipa_edge_args_sum.
33150         (ipa_free_edge_args_substructures): Remove declaration.
33151         (ipa_check_create_edge_args): Use ipa_edge_args_sum.
33152         (ipa_edge_args_info_available_for_edge_p): Likewise.
33153         * ipa-prop.c (ipa_edge_args_vector): Removed.
33154         (edge_removal_hook_holder): Likewise.
33155         (edge_duplication_hook_holder): Likewise.
33156         (ipa_edge_args_sum): New variable.
33157         (ipa_propagate_indirect_call_infos): Test ipa_edge_args_sum instead of
33158         ipa_edge_args_vector.
33159         (ipa_free_edge_args_substructures): Likewise.
33160         (ipa_free_all_edge_args): Free ipa_edge_args_sum instead of
33161         ipa_edge_args_vector.
33162         (ipa_edge_removal_hook): Turned into method
33163         ipa_edge_args_sum_t::remove.
33164         (ipa_edge_duplication_hook): Turned into method
33165         ipa_edge_args_sum_t::duplicate.
33166         (ipa_register_cgraph_hooks): Create ipa_edge_args_sum instead of
33167         registering edge hooks.
33168         (ipa_unregister_cgraph_hooks): Do not unregister edge hooks.
33169         * ipa-inline-analysis.c (estimate_function_body_sizes): Test
33170         ipa_edge_args_sum instead of ipa_edge_args_vector.
33171         * ipa-profile.c (ipa_profile): Likewise.
33173 2017-05-03  Martin Jambor  <mjambor@suse.cz>
33175         * symbol-summary.h (function_summary): New method exists.
33176         (function_summary::symtab_removal): Deallocate through release.
33177         (call_summary): New class.
33178         (gt_ggc_mx): New overload.
33179         (gt_pch_nx): Likewise.
33180         (gt_pch_nx): Likewise.
33182 2017-05-03  Jeff Law  <law@redhat.com>
33184         PR tree-optimization/78496
33185         * tree-vrp.c (simplify_cond_using_ranges_1): Renamed
33186         from simplify_cond_using_ranges.  Split off code to walk
33187         backwards through casts into ...
33188         (simplify_cond_using_ranges_2): New function.
33189         (simplify_stmt_using_ranges): Call simplify_cond_using_ranges_1.
33190         (execute_vrp): After identifying jump threads, call
33191         simplify_cond_using_ranges_2.
33193 2017-05-03  Jan Hubicka  <hubicka@ucw.cz>
33195         PR bootstrap/80609
33196         * ipa-inline.h (inline_summary): Add ctor.
33197         (create_ggc): Do not use ggc_cleared_alloc.
33199 2017-05-03  Jeff Downs  <heydowns@somuchpressure.net>
33200             Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
33202         * gcc.c (handle_braces): Support escaping in switch matching
33203         text.
33204         * doc/invoke.texi (Spec Files): Document it.
33205         Remove superfluous @code markup in items.
33207 2017-05-03  David Malcolm  <dmalcolm@redhat.com>
33209         * diagnostic-show-locus.c (struct column_range): New struct.
33210         (get_affected_columns): New function.
33211         (get_printed_columns): New function.
33212         (struct correction): New struct.
33213         (correction::ensure_capacity): New function.
33214         (correction::ensure_terminated): New function.
33215         (struct line_corrections): New struct.
33216         (line_corrections::~line_corrections): New dtor.
33217         (line_corrections::add_hint): New function.
33218         (layout::print_trailing_fixits): Reimplement in terms of the new
33219         classes.
33220         (selftest::test_overlapped_fixit_printing): New function.
33221         (selftest::diagnostic_show_locus_c_tests): Call it.
33223 2017-05-03  Nathan Sidwell  <nathan@acm.org>
33225         Canonicalize canonical type hashing
33226         * tree.h (type_hash_canon_hash): Declare.
33227         * tree.c (type_hash_list, attribute_hash_list): Move into
33228         type_hash_canon_hash.
33229         (build_type_attribute_qual_variant): Break out hash code calc into
33230         type_hash_canon_hash.
33231         (type_hash_canon_hash): New.  Generic type hash computation.
33232         (build_range_type_1, build_array_type_1, build_function_type,
33233         build_method_type_directly, build_offset_type, build_complex_type,
33234         make_vector_type): Call it.
33236 2017-05-03  Richard Biener  <rguenther@suse.de>
33238         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
33239         When all DRs have unknown misaligned do not always peel
33240         when there is a store but apply the same costing model as if
33241         there were only loads.
33243 2017-05-03  Richard Biener  <rguenther@suse.de>
33245         Revert
33246         PR tree-optimization/80492
33247         * tree-ssa-alias.c (decl_refs_may_alias_p): Handle
33248         compare_base_decls returning dont-know properly.
33250 2017-05-03  Thomas Preud'homme  <thomas.preudhomme@arm.com>
33252         * config/arm/iterators.md (CCSI): New mode iterator.
33253         (arch): New mode attribute.
33254         * config/arm/sync.md (atomic_compare_and_swap<mode>_1): Rename into ...
33255         (atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1): This and ...
33256         (atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): This.  Use CCSI
33257         code iterator for success result mode.
33258         * config/arm/arm.c (arm_expand_compare_and_swap): Adapt code to use
33259         the corresponding new insn generators.
33261 2017-05-03  Bin Cheng  <bin.cheng@arm.com>
33263         Revert r247509
33264         2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33265         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
33267 2017-05-03  Richard Sandiford  <richard.sandiford@linaro.org>
33269         * tree-data-ref.h (SUB_CONFLICTS_IN_A): Wrap SUB argument in brackets.
33270         (SUB_CONFLICTS_IN_B, SUB_LAST_CONFLICT, SUB_DISTANCE): Likewise.
33271         (DDR_A): Wrap DDR argument in brackets.
33272         (DDR_B, DDR_AFFINE_P, DDR_ARE_DEPENDENT, DDR_SUBSCRIPTS): Likewise.
33273         (DDR_LOOP_NEST, DDR_INNER_LOOP, DDR_SELF_REFERENCE): Likewise.
33274         (DDR_REVERSED_P): Likewise.
33276 2017-05-03  Jakub Jelinek  <jakub@redhat.com>
33278         PR tree-optimization/79472
33279         * tree-switch-conversion.c (struct switch_conv_info): Add
33280         contiguous_range and default_case_nonstandard fields.
33281         (collect_switch_conv_info): Compute contiguous_range and
33282         default_case_nonstandard fields, don't clear final_bb if
33283         contiguous_range and only the default case doesn't have the required
33284         structure.
33285         (check_all_empty_except_final): Set default_case_nonstandard instead
33286         of failing if contiguous_range and the default case doesn't have empty
33287         block.
33288         (check_final_bb): Add SWTCH argument, don't fail if contiguous_range
33289         and only the default case doesn't have the required constants.  Skip
33290         virtual phis.
33291         (gather_default_values): Skip virtual phis.  Allow non-NULL CASE_LOW
33292         if default_case_nonstandard.
33293         (build_constructors): Build constant 1 just once.  Assert that default
33294         values aren't inserted in between cases if contiguous_range.  Skip
33295         virtual phis.
33296         (build_arrays): Skip virtual phis.
33297         (prune_bbs): Add DEFAULT_BB argument, don't remove that bb.
33298         (fix_phi_nodes): Don't add e2f phi arg if default_case_nonstandard.
33299         Handle virtual phis.
33300         (gen_inbound_check): Handle default_case_nonstandard case.
33301         (process_switch): Adjust check_final_bb caller.  Call
33302         gather_default_values with the first non-default case instead of
33303         default case if default_case_nonstandard.
33305 2017-05-02  Nathan Sidwell  <nathan@acm.org>
33307         * ggc-page.c (move_ptes_to_front): Replace unsigned >0 with i--
33308         check.  Fix formatting.
33310 2017-05-02  Jan Hubicka  <hubicka@ucw.cz>
33312         * ipa-inline-analysis.c (estimate_node_size_and_time): Allow roundoff
33313         errors when comparing specialized and unspecialized times.
33315 2017-05-02  David Malcolm  <dmalcolm@redhat.com>
33317         * diagnostic-show-locus.c
33318         (layout::should_print_annotation_line_p): Make private.
33319         (layout::print_annotation_line): Make private.
33320         (layout::annotation_line_showed_range_p): Make private.
33321         (layout::show_ruler): Make private.
33322         (layout::print_source_line): Make private.  Pass in line and
33323         line_width, rather than calling location_get_source_line.  Drop
33324         returned value.
33325         (layout::print_leading_fixits): New method.
33326         (layout::print_any_fixits): Rename to...
33327         (layout::print_trailing_fixits): ...this, and make private.
33328         Don't print newline fixits.
33329         (diagnostic_show_locus): Move logic for printing one row into...
33330         (layout::print_line): ...this new function.  Move the
33331         location_get_source_line call and error-handling from
33332         print_source_line to here.  Call print_leading_fixits, and rename
33333         print_any_fixits to print_trailing_fixits.
33334         (selftest::test_fixit_insert_containing_newline): Update now that
33335         newlines are partially supported.
33336         (selftest::test_fixit_insert_containing_newline_2): New test.
33337         (selftest::test_fixit_replace_containing_newline): Update comments.
33338         (selftest::diagnostic_show_locus_c_tests): Call the new test.
33339         * edit-context.c (class added_line): New class.
33340         (class edited_line): Describe newline handling in comment.
33341         (edited_line::actually_edited_p): New method.
33342         (edited_line::print_content): Delete redundant decl.
33343         (edited_line::m_predecessors): New field.
33344         (edited_file::print_content): Call edited_line::print_content.
33345         (edited_file::print_diff): Update to support newlines.
33346         (edited_file::print_diff_hunk): Likewise.
33347         (edited_file::print_run_of_changed_lines): New function.
33348         (edited_file::print_diff_line): Convert to...
33349         (print_diff_line): ...this.
33350         (edited_file::get_effective_line_count): New function.
33351         (edited_line::edited_line): Initialize new field m_predecessors.
33352         (edited_line::~edited_line): Clean up m_predecessors.
33353         (edited_line::apply_fixit): Handle newlines.
33354         (edited_line::get_effective_line_count): New function.
33355         (edited_line::print_content): New function.
33356         (edited_line::print_diff_lines): New function.
33357         (selftest::test_applying_fixits_insert_containing_newline): New
33358         test.
33359         (selftest::test_applying_fixits_replace_containing_newline): New
33360         test.
33361         (selftest::insert_line): New function.
33362         (selftest::test_applying_fixits_multiple_lines): Add example of
33363         inserting a line.
33364         (selftest::edit_context_c_tests): Call the new tests.
33366 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33368         * tree-ssa-loop-ivopts.c (get_scaled_computation_cost_at): Delete
33369         parameter cand.  Update dump information.
33370         (get_computation_cost): Update uses.
33372 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33374         * tree-ssa-loop-ivopts.c (get_computation_aff_1): New.
33375         (get_computation_aff): Reorder parameters.  Use get_computation_aff_1.
33376         (get_computation_at, rewrite_use_address): Update use of
33377         get_computation_aff.
33379 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33381         * tree-ssa-loop-ivopts.c (get_computation_at): Reorder parameters.
33382         (get_computation): Delete.
33383         (get_computation_cost): Implement like get_computation_cost_at.
33384         Use get_computation_at.
33385         (get_computation_cost_at): Delete.
33386         (rewrite_use_nonlinear_expr): Use get_computation_at.
33387         (rewrite_use_compare, remove_unused_ivs): Ditto.
33389 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33391         * tree-ssa-loop-ivopts.c (rewrite_use_address): Simple refactor.
33393 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33395         * tree-ssa-loop-ivopts.c (struct iv_ca): Rename n_regs to n_invs.
33396         (ivopts_global_cost_for_size): Rename parameter and update uses.
33397         (iv_ca_recount_cost): Update uses.
33398         (iv_ca_set_remove_invs, iv_ca_set_no_cp): Record invariants and
33399         candidates seperately in n_invs and n_cands.
33400         (iv_ca_set_add_invs, iv_ca_set_cp, iv_ca_new): Ditto.
33402 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33404         * tree-ssa-loop-ivopts.c (struct walk_tree_data): New.
33405         (find_inv_vars_cb): New.
33406         (find_depends): Renamed to ...
33407         (find_inv_vars): ... this.
33408         (add_candidate_1, force_var_cost): Call find_inv_vars.
33409         (split_address_cost, determine_group_iv_cost_cond): Ditto.
33411 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33413         * tree-ssa-loop-ivopts.c (struct cost_pair): Rename depends_on to
33414         inv_vars.  Add inv_exprs.
33415         (struct iv_cand): Rename depends_on to inv_vars.
33416         (struct ivopts_data): Rename max_inv_id/n_invariant_uses to
33417         max_inv_var_id/n_inv_var_uses.  Move max_inv_expr_id around.
33418         Refactor field used_inv_exprs from has_map to array n_inv_expr_uses.
33419         (dump_cand): Dump inv_vars.
33420         (tree_ssa_iv_optimize_init): Support inv_vars and inv_exprs.
33421         (record_invariant, find_depends, add_candidate_1): Ditto.
33422         (set_group_iv_cost, force_var_cost): Ditto.
33423         (split_address_cost, ptr_difference_cost, difference_cost): Ditto.
33424         (get_computation_cost_at, get_computation_cost): Ditto.
33425         (determine_group_iv_cost_generic): Ditto.
33426         (determine_group_iv_cost_address): Ditto.
33427         (determine_group_iv_cost_cond, autoinc_possible_for_pair): Ditto.
33428         (determine_group_iv_costs): Ditto.
33429         (iv_ca_recount_cost): Update call to ivopts_global_cost_for_size.
33430         (iv_ca_set_remove_invariants): Renamed to ...
33431         (iv_ca_set_remove_invs): ... this.  Support inv_vars and inv_exprs.
33432         (iv_ca_set_no_cp): Use iv_ca_set_remove_invs.
33433         (iv_ca_set_add_invariants):  Renamed to ...
33434         (iv_ca_set_add_invs): ... this.  Support inv_vars and inv_exprs.
33435         (iv_ca_set_cp): Use iv_ca_set_add_invs.
33436         (iv_ca_has_deps): Support inv_vars and inv_exprs.
33437         (iv_ca_new, iv_ca_free, iv_ca_dump, free_loop_data): Ditto.
33438         (create_new_ivs): Remove useless dump.
33440 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33442         * tree-ssa-loop-ivopts.c (get_computation_cost_at): Remove pseudo
33443         iv_cand code.
33444         (determine_group_iv_cost_cond, determine_iv_cost): Ditto.
33445         (iv_ca_set_no_cp, create_new_iv): Ditto.
33447 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33449         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
33451 2017-05-02  Xi Ruoyao  <ryxi@stu.xidian.edu.cn>
33453         * tree.h (EXPR_CILK_SPAWN): Use macro TREE_CHECK2 instead of
33454         function tree_check2.
33456 2017-05-02  Martin Liska  <mliska@suse.cz>
33458         * doc/gcov.texi: Add missing preposition.
33459         * gcov.c (function_info::function_info): Properly fill up
33460         all member variables.
33462 2017-05-02  Tamar Christina  <tamar.christina@arm.com>
33464         * expr.c (expand_expr_real_2): Re-cost if previous costs are the same.
33466 2017-05-02  Tamar Christina  <tamar.christina@arm.com>
33468         * simplify-rtx.c (simplify_binary_operation_1): Add LSHIFTRT case.
33470 2017-05-02  Martin Liska  <mliska@suse.cz>
33472         PR lto/77954.
33473         * lto-streamer-in.c (lto_read_tree_1): Remove
33474         LTO_STREAMER_DEBUG.
33475         * lto-streamer.c (struct tree_hash_entry): Likewise.
33476         (struct tree_entry_hasher): Likewise.
33477         (tree_entry_hasher::hash): Likewise.
33478         (tree_entry_hasher::equal): Likewise.
33479         (lto_streamer_init): Likewise.
33480         (lto_orig_address_map): Likewise.
33481         (lto_orig_address_get): Likewise.
33482         (lto_orig_address_remove): Likewise.
33483         * lto-streamer.h: Likewise.
33484         * tree-streamer-in.c (streamer_alloc_tree): Likewise.
33485         * tree-streamer-out.c (streamer_write_tree_header): Likewise.
33487 2017-05-02  Sebastian Peryt  <sebastian.peryt@intel.com>
33489         * config/i386/avx512fintrin.h (_mm_mask_add_round_sd)
33490         (_mm_maskz_add_round_sd, _mm_mask_add_round_ss)
33491         (mm_maskz_add_round_ss, _mm_mask_sub_round_sd)
33492         (mm_maskz_sub_round_sd, _mm_mask_sub_round_ss)
33493         (mm_maskz_sub_round_ss, _mm_mask_add_sd)
33494         (mm_maskz_add_sd, _mm_mask_add_ss, _mm_maskz_add_ss)
33495         (mm_mask_sub_sd, _mm_maskz_sub_sd, _mm_mask_sub_ss)
33496         (mm_maskz_sub_ss): New intrinsics.
33497         * config/i386/i386-builtin-types.def (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
33498         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): New function type aliases.
33499         * config/i386/i386-builtin.def (__builtin_ia32_addsd_mask_round)
33500         (__builtin_ia32_addss_mask_round, __builtin_ia32_subsd_mask_round)
33501         (__builtin_ia32_subss_mask_round): New builtins.
33502         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
33503         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
33504         * config/i386/sse.md (<sse>_vm<plusminus_insn><mode>3<round_name>):
33505         Renamed to ...
33506         (<sse>_vm<plusminus_insn><mode>3<mask_name><round_name>): ... this.
33507         (v<plusminus_mnemonic><ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|%0, %1, %<iptr>2<round_op3>}):
33508         Changed to ...
33509         (v<plusminus_mnemonic><ssescalarmodesuffix>\t{<round_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}):
33510         ... this.
33512 2017-05-02  Martin Jambor  <mjambor@suse.cz>
33514         PR tree-optimization/78687
33515         * tree-sra.c (access): New field parent.
33516         (process_subtree_disqualification): New function.
33517         (disqualify_candidate): Call it.
33518         (build_accesses_from_assign): Reset write flag if creating an
33519         assighnment link.
33520         (build_access_subtree): Fill in parent field and also prpagate
33521         down grp_write flag.
33522         (create_artificial_child_access): New parameter set_grp_write, set
33523         grp_write to its value.
33524         (propagate_subaccesses_across_link): Also propagate grp_write flag
33525         values.
33526         (propagate_all_subaccesses): Push the closest parent back to work
33527         queue if add_access_to_work_queue returned true.
33529 2017-05-02  Richard Biener  <rguenther@suse.de>
33531         * common.opt (fstrict-overflow): Alias negative to fwrapv.
33532         * doc/invoke.texi (fstrict-overflow): Remove all traces of
33533         -fstrict-overflow documentation.
33534         * tree.h (TYPE_OVERFLOW_UNDEFINED): Do not test flag_strict_overflow.
33535         (POINTER_TYPE_OVERFLOW_UNDEFINED): Test !flag_wrapv instead of
33536         flag_strict_overflow.
33537         * ipa-inline.c (can_inline_edge_p): Do not test flag_strict_overflow.
33538         * lto-opts.c (lto_write_options): Do not stream it.
33539         * lto-wrapper.c (merge_and_complain): Do not handle it.
33540         * opts.c (default_options_table): Do not set -fstrict-overflow.
33541         (finish_options): Likewise do not clear it when sanitizing.
33542         * simplify-rtx.c (simplify_const_relational_operation): Do not
33543         test flag_strict_overflow.
33545 2017-05-02  Uros Bizjak  <ubizjak@gmail.com>
33547         * config/alpha/alpha.md (*add<mode>3_ieee): Merge to add<mode>3
33548         using enabled attribute.
33549         (*sub<mode>3_ieee): Merge to sub<mode>3 using enabled attribute.
33550         (*mul<mode>3_ieee): Merge to mul<mode>3 using enabled attribute.
33551         (*div<mode>3_ieee): Merge to div<mode>3 using enabled attribute.
33552         (*sqrt<mode>2_ieee): Merge to sqrt<mode>2 using enabled attribute.
33553         (*fix_truncdfdi_ieee): Merge to *fix_truncdfdi2 using enabled attribute.
33554         (*fix_truncsfdi_ieee): Merge to *fix_truncsfdi2 using enabled attribute.
33555         (*floatdisf_ieee): Merge to floatdisf2 using enabled attribute.
33556         (*floatdidf_ieee): Merge to floatdidf2 using enabled attribute.
33557         (*truncdfsf2_ieee): Merge to truncdfsf2 using enabled attribute.
33558         (*cmpdf_ieee): Merge to *cmpdf_internal using enabled attribute.
33560 2017-05-02  Uros Bizjak  <ubizjak@gmail.com>
33562         * config/i386/i386.c (ix86_code_end): Use {FIRST,LAST}_INT_REG.
33564 2017-05-02  Richard Biener  <rguenther@suse.de>
33566         PR tree-optimization/80591
33567         Revert
33568         2017-04-10  Richard Biener  <rguenther@suse.de>
33570         * tree-ssa-structalias.c (find_func_aliases): Properly handle
33571         asm inputs.
33573 2017-05-02  Richard Biener  <rguenther@suse.de>
33575         PR tree-optimization/80549
33576         * tree-cfgcleanup.c (mfb_keep_latches): New helper.
33577         (cleanup_tree_cfg_noloop): Create forwarders to known loop
33578         headers if they do not have a preheader.
33580 2017-05-02  Martin Liska  <mliska@suse.cz>
33582         PR other/80589
33583         * common.opt: Fix typo.
33584         * doc/invoke.texi: Likewise.
33586 2017-05-01  Jan Beulich  <jbeulich@suse.com>
33588         * config/i386/sse.md (xop_vpermil2<mode>3): Do not allow operand
33589         swapping, add (x,x,m,x,n) alternative.
33591 2017-05-01  Nathan Sidwell  <nathan@acm.org>
33593         * calls.c (combine_pending_stack_adjustment_and_call): Remove
33594         unnecessary unadjusted_alignment check.
33596 2017-05-01  Xi Ruoyao  <ryxi@stu.xidian.edu.cn>
33598         PR c++/80038
33599         * cilk_common.c (expand_builtin_cilk_detach): Move pedigree
33600         operations here.
33601         * gimplify.c (gimplify_cilk_detach): New function.
33602         (gimplify_call_expr, gimplify_modify_expr): Call it as needed.
33603         * tree-core.h: Document EXPR_CILK_SPAWN.
33604         * tree.h (EXPR_CILK_SPAWN): Define.
33606 2017-05-01  David Malcolm  <dmalcolm@redhat.com>
33608         * diagnostic-show-locus.c (layout::get_expanded_location): Rewrite
33609         to use new fixit_hint representation, using the "replace" logic.
33610         (get_line_span_for_fixit_hint): Likewise.
33611         (layout::print_any_fixits): Likewise.
33612         (selftest::test_one_liner_many_fixits): Rename to...
33613         (selftest::test_one_liner_many_fixits_1): ...this, and update
33614         comment and expected output to reflect that the multiple fix-it
33615         hints are now consolidated into one insertion.
33616         (selftest::test_one_liner_many_fixits_2): New test.
33617         (selftest::test_diagnostic_show_locus_one_liner): Update for
33618         above.
33619         (selftest::test_fixit_consolidation): Update for fix-it API
33620         change.
33621         * diagnostic.c (print_parseable_fixits): Likewise.
33622         * edit-context.c (edited_line::m_line_events): Convert from
33623         auto_vec <line_event *> to auto_vec <line_event>.
33624         (class line_event): Convert from abstract base class to a concrete
33625         class, taking over the role of replace_event.
33626         (class insert_event): Delete.
33627         (class replace_event): Rename to class line_event.  Convert to
33628         half-open range.
33629         (edit_context::add_fixits): Reimplement.
33630         (edit_context::apply_insert): Delete.
33631         (edit_context::apply_replace): Rename to...
33632         (edit_context::apply_fixit): ...this.  Convert to half-open range.
33633         (edited_file::apply_insert): Delete.
33634         (edited_file::apply_replace): Rename to...
33635         (edited_file::apply_fixit): ...this.
33636         (edited_line::~edited_line): Drop deletion of events.
33637         (edited_line::apply_insert): Delete.
33638         (edited_line::apply_replace): Rename to...
33639         (edited_line::apply_fixit): ...this.  Convert to half-open range.
33640         Update for change to type of m_line_events.
33641         * edit-context.h (edit_context::apply_insert): Delete.
33642         (edit_context::apply_replace): Rename to...
33643         (edit_context::apply_fixit): ...this.
33645 2017-05-01  Martin Sebor  <msebor@redhat.com>
33647         * gimple-ssa-sprintf.c (format_integer): Set knownrange when it's
33648         known.
33650 2017-05-01  Uros Bizjak  <ubizjak@gmail.com>
33652         PR target/68491
33653         * config/i386/cpuid.h (__get_cpuid): Always return 0 when
33654         __get_cpuid_max returns 0.
33655         (__get_cpuid_count): Ditto.
33657 2017-05-01  Eric Botcazou  <ebotcazou@adacore.com>
33659         * tree.c (substitute_in_expr) <tcc_vl_exp>: Also inline a call if the
33660         replacement expression is another instance of one of its arguments.
33662 2017-05-01  Jakub Jelinek  <jakub@redhat.com>
33664         PR target/79430
33665         * rtlanal.c (reg_set_p): If reg is a stack_pointer_rtx, also
33666         check for stack push/pop autoinc.
33667         * config/i386/i386.c (ix86_agi_dependent): Return false
33668         if the only reason why modified_in_p returned true is that
33669         addr is SP based and set_insn is a push or pop.
33671 2017-04-29  Jan Hubicka  <hubicka@ucw.cz>
33673         * ipa-inline.c (compute_inlined_call_time): Remove now unnecesary
33674         overflow check.
33676 2017-04-29  Jan Hubicka  <hubicka@ucw.cz>
33678         PR ipa/79224
33679         * ipa-inline-analysis.c (dump_predicate): Add optional parameter NL.
33680         (account_size_time): Use two predicates - exec_pred and
33681         nonconst_pred_ptr.
33682         (evaluate_conditions_for_known_args): Compute both clause and
33683         nonspec_clause.
33684         (evaluate_properties_for_edge): Evaulate both clause and nonspec_clause.
33685         (inline_summary_t::duplicate): Update.
33686         (estimate_function_body_sizes): Caluculate exec and nonconst predicates
33687         separately.
33688         (compute_inline_parameters): Likewise.
33689         (estimate_edge_size_and_time): Update caluclation of time.
33690         (estimate_node_size_and_time): Compute both time and nonspecialized
33691         time.
33692         (estimate_ipcp_clone_size_and_time): Update.
33693         (inline_merge_summary): Update.
33694         (do_estimate_edge_time): Update.
33695         (do_estimate_edge_size): Update.
33696         (do_estimate_edge_hints): Update.
33697         (inline_read_section, inline_write_summary): Stream both new predicates.
33698         * ipa-inline.c (compute_uninlined_call_time): Take uninlined_call_time
33699         as argument.
33700         (compute_inlined_call_time): Cleanup.
33701         (big_speedup_p): Update.
33702         (edge_badness): Update.
33703         * ipa-inline.h (INLINE_TIME_SCALE): Remove.
33704         (size_time_entry): Replace predicate by exec_predicate and
33705         nonconst_predicate.
33706         (edge_growth_cache_entry): Cache both time nad nonspecialized time.
33707         (estimate_edge_time): Return also nonspec_time.
33708         (reset_edge_growth_cache): Update.
33710 2017-04-29  Jakub Jelinek  <jakub@redhat.com>
33712         PR rtl-optimization/80491
33713         * ifcvt.c (noce_process_if_block): When looking for x setter
33714         with missing else_bb, don't check only the insn right before
33715         cond_earliest, but look for the last insn that x is modified in
33716         within the same bb.
33718         PR rtl-optimization/80491
33719         * alias.c (memory_modified_in_insn_p): Return true for CALL_INSNs.
33721 2017-04-29  Marc Glisse  <marc.glisse@inria.fr>
33723         PR tree-optimization/80487
33724         * tree-ssa-alias.c (stmt_kills_ref_p): Handle stpncpy and strncpy.
33726 2017-04-29  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
33728         PR tree-optimization/79697
33729         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Check if callee
33730         is BUILT_IN_STRDUP, BUILT_IN_STRNDUP, BUILT_IN_REALLOC.
33731         (propagate_necessity): Check if def_callee is BUILT_IN_STRDUP or
33732         BUILT_IN_STRNDUP.
33733         * gimple-fold.c (gimple_fold_builtin_realloc): New function.
33734         (gimple_fold_builtin): Call gimple_fold_builtin_realloc.
33736 2017-04-28  Martin Sebor  <msebor@redhat.com>
33738         PR tree-optimization/80523
33739         * gimple-ssa-sprintf.c (target_to_host_charmap): New global variable.
33740         (init_target_to_host_charmap, target_to_host, target_strtol10): New
33741         functions.
33742         (maybe_warn, format_directive, parse_directive): Use new functions.
33743         (pass_sprintf_length::execute): Call init_target_to_host_charmap.
33745 2017-04-28  Marc Glisse  <marc.glisse@inria.fr>
33747         * match.pd (X+Z OP Y+Z, X-Z OP Y-Z, Z-X OP Z-Y): New transformations.
33749 2017-04-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
33751         * configure.ac (SYSTEM_HEADER_DIR, BUILD_SYSTEM_HEADER_DIR,
33752         target_header_dir): Set correctly.
33753         * configure: Regenerated.
33754         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
33755         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
33756         instead of SYSTEM_HEADER_DIR.
33758 2017-04-28  Jan Hubicka  <hubicka@ucw.cz>
33760         * ipa-cp.c (perform_estimation_of_a_value): Turn time to sreal.
33761         (estimate_local_effects): Likewise.
33762         * ipa-inline.c (compute_inlined_call_time, want_inline_small_function_p,
33763         edge_badness, inline_small_functions, dump_overall_stats): LIkewise.
33764         * ipa-inline.h (edge_growth_cache_entry, estimate_time_after_inlining,
33765         estimate_ipcp_clone_size_and_time, do_estimate_edge_time,
33766         do_estimate_edge_time, estimate_edge_time): Likewise.
33767         * ipa-inline-analysis.c (estimate_node_size_and_time,
33768         estimate_ipcp_clone_size_and_time, do_estimate_edge_time): Likewise.
33769         (estimate_time_after_inlining): Remove.
33771 2017-04-28  Martin Liska  <mliska@suse.cz>
33773         * doc/gcov.texi: Enhance documentation of gcov.
33775 2017-04-28  Martin Liska  <mliska@suse.cz>
33777         * doc/gcov.texi: Sort options in alphabetic order.
33778         * doc/gcov-dump.texi: Likewise.
33779         * doc/gcov-tool.texi: Likewise.
33780         * gcov.c (print_usage): Likewise.
33781         * gcov-dump.c (print_usage): Likewise.
33782         * gcov-tool.c (print_merge_usage_message): Likewise.
33783         (print_rewrite_usage_message): Likewise.
33784         (print_overlap_usage_message): Likewise.
33786 2017-04-28  Martin Liska  <mliska@suse.cz>
33788         PR gcov-profile/53915
33789         * gcov.c (format_gcov): Print 'NAN %' when top > bottom.
33791 2017-04-28  Martin Liska  <mliska@suse.cz>
33793         PR gcov-profile/79891
33794         * gcov.c (add_line_counts): Assign BBs to lines just if the BB
33795         is marked by compiler as living on a line.
33796         (get_cycles_count): Remove usage of the union.
33797         (output_intermediate_file): Likewise.
33798         (find_source): Fix GNU coding style.
33799         (accumulate_line_counts): Remove old non-all block mode.
33800         (output_lines): Remove usage of the union.
33801         * profile.c (output_location): Include all BBs, even if
33802         belonging to a same line (and file) as a previous BB.
33804 2017-04-28  Martin Liska  <mliska@suse.cz>
33806         * gcov.c (process_args): Handle new argument 'w'.
33807         (read_graph_file): Assign ID to BBs.
33808         (output_branch_count): Display BB # if verbose flag is set.
33809         (output_lines): Likewise for arcs.
33810         (print_usage): Add '--verbose' option help.
33811         * doc/gcov.texi: Document --verbose (-w) option.
33813 2017-04-28  Martin Liska  <mliska@suse.cz>
33815         * gcov.c (struct block_location_info): New struct.
33816         (process_file): Fill up the new structure.
33817         (read_graph_file): Replace usage of encoding by the newly added
33818         struct.
33819         (add_line_counts): Likewise.
33820         (accumulate_line_counts): Remove usage of the union.
33821         (function_info::function_info): New function.
33822         (function_info::~function_info): Likewise.
33823         (process_file): Call delete instead of release_function.
33824         (release_function): Release the function.
33825         (release_structures): Call delete instead of release_function.
33826         (solve_flow_graph): Replace usage of num_blocks.
33827         (find_exception_blocks): Likewise.
33828         (output_lines): Fix GNU coding style.
33830 2017-04-28  Martin Liska  <mliska@suse.cz>
33832         PR driver/56469
33833         * coverage.c (coverage_remove_note_file): New function.
33834         * coverage.h: Declare the function.
33835         * toplev.c (finalize): Clean if an error has been seen.
33837 2017-04-28  Martin Liska  <mliska@suse.cz>
33839         PR gcov-profile/80031
33840         * gcov-dump.c (tag_blocks): Just print number of basic blocks.
33841         * gcov-io.h (GCOV_TAG_BLOCKS_NUM): Remove unused macro.
33842         * gcov.c (read_graph_file): Read just number of blocks.
33843         * profile.c (branch_prob): Do not stream 0 flags per a basic
33844         block.
33846 2017-04-28  Martin Liska  <mliska@suse.cz>
33848         * gcov-dump.c (tag_*): Add new argument to declarations.
33849         (dump_gcov_file): Likewise.
33850         (tag_blocks): Add and use new argument depth.
33851         (tag_arcs): Likewise.
33852         (tag_lines): Likewise.
33853         (tag_counters): Likewise.
33854         (tag_summary): Likewise.
33855         (dump_working_sets): Use depth to do a proper indentation.
33857 2017-04-28  Jakub Jelinek  <jakub@redhat.com>
33859         PR bootstrap/80531
33860         * cgraph.h (symtab_node::debug_symtab): No longer inline.
33861         * symtab.c (symtab_node::debug_symtab): Move definition here.
33863 2017-04-28  Richard Biener  <rguenther@suse.de>
33865         * lto-streamer.h (LTO_major_version): Bump to 7.
33867 2017-04-28  Richard Biener  <rguenther@suse.de>
33869         * tree-vrp.c (assert_info): New struct.
33870         (add_assert_info): New helper.
33871         (register_edge_assert_for_2): Refactor to add asserts to a vector
33872         of assert_info.
33873         (register_edge_assert_for_1): Likewise.
33874         (register_edge_assert_for): Likewise.
33875         (finish_register_edge_assert_for): New helper actually registering
33876         asserts where live on edge.
33877         (find_conditional_asserts): Adjust.
33878         (find_switch_asserts): Likewise.
33879         (evrp_dom_walker::try_find_new_range): Generalize.
33880         (evrp_dom_walker::before_dom_children): Use register_edge_assert_for.
33882 2017-04-27  Marek Polacek  <polacek@redhat.com>
33884         PR sanitizer/80349
33885         * fold-const.c (fold_binary_loc) <case EQ_EXPR, NE_EXPR>: Convert
33886         arg10 and arg11 to itype.
33888 2017-04-27  Jonathan Wakely  <jwakely@redhat.com>
33890         * doc/extend.texi (Object Size Checking): Improve grammar.
33892 2017-04-27  Richard Earnshaw  <rearnsha@arm.com>
33894         PR target/80530
33895         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Ensure
33896         that the logic for permitting reciprocal estimates matches that
33897         in use_rsqrt_p.
33899 2017-04-27  Jakub Jelinek  <jakub@redhat.com>
33901         PR c++/80534
33902         * tree.c (type_cache_hasher::equal): Only compare
33903         TYPE_TYPELESS_STORAGE flag on non-aggregate element types.
33904         (build_array_type_1): Only hash TYPE_TYPELESS_STORAGE flag on
33905         non-aggregate element types.
33906         * tree.h (TYPE_TYPELESS_STORAGE): Fix comment typo, add more details
33907         about the flag on ARRAY_TYPEs in the comment, formatting fix.
33909 2017-04-27  Richard Biener  <rguenther@suse.de>
33911         PR middle-end/80533
33912         * emit-rtl.c (set_mem_attributes_minus_bitpos): When
33913         stripping ARRAY_REFs from MEM_EXPR make sure we're not
33914         keeping a reference to a trailing array.
33916 2017-04-27  Richard Biener  <rguenther@suse.de>
33918         PR middle-end/80539
33919         * tree-chrec.c (chrec_fold_plus_poly_poly): Deal with not
33920         being in loop-closed SSA form conservatively.
33921         (chrec_fold_multiply_poly_poly): Likewise.
33923 2017-04-27  Tamar Christina  <tamar.christina@arm.com>
33925         PR middle-end/79665
33926         * expr.c (expand_expr_real_2): Move TRUNC_MOD_EXPR, FLOOR_MOD_EXPR,
33927         CEIL_MOD_EXPR, ROUND_MOD_EXPR cases.
33929 2017-04-27  Jakub Jelinek  <jakub@redhat.com>
33931         PR target/77728
33932         * config/aarch64/aarch64.c (struct aarch64_fn_arg_alignment): Remove.
33933         (aarch64_function_arg_alignment): Return unsigned int again, but still
33934         ignore TYPE_FIELDS chain decls other than FIELD_DECLs.
33935         (aarch64_layout_arg): Adjust aarch64_function_arg_alignment caller.
33936         Don't emit -Wpsabi note.
33937         (aarch64_function_arg_boundary): Likewise.
33938         (aarch64_gimplify_va_arg_expr): Adjust aarch64_function_arg_alignment
33939         caller.
33941 2017-04-26  Nathan Sidwell  <nathan@acm.org>
33943         * tree.h (crc32_unsigned_n): Declare.
33944         (crc32_unsigned, crc32_unsigned): Make inline.
33945         * tree.c (crc32_unsigned_bits): Replace with ...
33946         (crc32_unsigned_n): ... this.
33947         (crc32_unsigned, crc32_byte): Remove.
33948         (crc32_string): Remove unnecessary braces.
33950 2017-04-25  Jan Hubicka  <hubicka@ucw.cz>
33952         * ipa-cp.c (estimate_local_effects): Convert sreal to int.
33953         * ipa-inline-analysis.c (MAX_TIME): Remove.
33954         (account_size_time): Use sreal for time.
33955         (dump_inline_summary): Update.
33956         (estimate_function_body_sizes): Update.
33957         (estimate_edge_size_and_time): Update.
33958         (estimate_calls_size_and_time): Update.
33959         (estimate_node_size_and_time): Update.
33960         (inline_merge_summary): Update.
33961         (inline_update_overall_summary): Update.
33962         (estimate_time_after_inlining): Update.
33963         (inline_read_section): Update.
33964         (inline_write_summary): Update.
33965         * ipa-inline.c (compute_uninlined_call_time): Update.
33966         (compute_inlined_call_time): Update.
33967         (recursive_inlining): Update.
33968         (inline_small_functions): Update.
33969         (dump_overall_stats): Update.
33970         * ipa-inline.h: Include sreal.h.
33971         (size_time_entry): Turn time to sreal.
33972         (inline_summary): Turn self_time nad time to sreal.
33974 2017-04-25  Jan Hubicka  <hubicka@ucw.cz>
33976         * sreal.c: Include backend.h, tree.h, gimple.h, cgraph.h and
33977         data-streamer.h
33978         (sreal::stream_out, sreal::stream_in): New.
33979         * sreal.h (sreal::stream_out, sreal::stream_in): Declare.
33981 2017-04-25  Jakub Jelinek  <jakub@redhat.com>
33983         * Makefile.in (s-options): Invoke opt-gather.awk with LC_ALL=C in the
33984         environment.
33986 2017-04-25  Uros Bizjak  <ubizjak@gmail.com>
33988         PR target/70799
33989         * config/i386/i386.c (dimode_scalar_to_vector_candidate_p):
33990         Handle ASHIFTRT.
33991         (dimode_scalar_chain::compute_convert_gain): Ditto.
33992         (dimode_scalar_chain::make_vector_copies): Ditto.
33993         (dimode_scalar_chain::convert_reg): Ditto.
33994         (dimode_scalar_chain::convert_insn): Ditto.
33995         * config/i386/sse.md (VI24_AVX512BW_1): Remove mode iterator.
33996         (VI248_AVX512BW_1): New mode iterator.
33997         (<mask_codefor>ashr<mode>3<mask_name>): Merge insn pattern with
33998         <mask_codefor>ashrv2di3<mask_name> insn using VI248_AVX512BW_1
33999         mode iterator.
34001 2017-04-25  Martin Sebor  <msebor@redhat.com>
34003         PR tree-optimization/80497
34004         * gimple-ssa-sprintf.c (get_int_range): Avoid assuming all integer
34005         constants are representable in HOST_WIDE_INT.
34006         (parse_directive): Ditto.
34008 2017-04-25  Martin Sebor  <msebor@redhat.com>
34010         PR bootstrap/80486
34011         * dominance.c (dom_info::m_n_basic_blocks): Change type to unsigned.
34012         (new_zero_array): Adjust signature.
34013         (dom_info::dom_init): Used unsigned rather that size_t.
34014         (dom_info::dom_info): Same.
34016 2017-04-25  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
34017             Jakub Jelinek  <jakub@redhat.com>
34019         PR target/77728
34020         * config/arm/arm.c: Include gimple.h.
34021         (aapcs_layout_arg): Emit -Wpsabi note if arm_needs_doubleword_align
34022         returns negative, increment ncrn only if it returned positive.
34023         (arm_needs_doubleword_align): Return int instead of bool,
34024         ignore DECL_ALIGN of non-FIELD_DECL TYPE_FIELDS chain
34025         members, but if there is any such non-FIELD_DECL
34026         > PARM_BOUNDARY aligned decl, return -1 instead of false.
34027         (arm_function_arg): Emit -Wpsabi note if arm_needs_doubleword_align
34028         returns negative, increment nregs only if it returned positive.
34029         (arm_setup_incoming_varargs): Likewise.
34030         (arm_function_arg_boundary): Emit -Wpsabi note if
34031         arm_needs_doubleword_align returns negative, return
34032         DOUBLEWORD_ALIGNMENT only if it returned positive.
34034 2017-04-25  Marek Polacek  <polacek@redhat.com>
34036         PR sanitizer/80349
34037         * fold-const.c (fold_binary_loc) <case BIT_IOR_EXPR>: Convert arg0's
34038         first argument to type.
34040 2017-04-25  Bill Seurer  <seurer@linux.vnet.ibm.com>
34042         PR target/80482
34043         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Change
34044         type checks to test for compatibility instead of equality.
34046 2017-04-25  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
34047             Jakub Jelinek  <jakub@redhat.com>
34049         PR target/77728
34050         * config/aarch64/aarch64.c (struct aarch64_fn_arg_alignment): New
34051         type.
34052         (aarch64_function_arg_alignment): Return aarch64_fn_arg_alignment
34053         struct.  Ignore DECL_ALIGN of decls other than FIELD_DECL for
34054         the alignment computation, but return their maximum in warn_alignment.
34055         (aarch64_layout_arg): Adjust aarch64_function_arg_alignment caller.
34056         Emit a -Wpsabi note if warn_alignment is 16 bytes, but alignment
34057         is smaller.
34058         (aarch64_function_arg_boundary): Likewise.  Simplify using MIN/MAX.
34059         (aarch64_gimplify_va_arg_expr): Adjust aarch64_function_arg_alignment
34060         caller.
34062 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34064         * config/arc/simdext.md (dmpyh): Fix typo.
34066 2017-04-25  Richard Biener  <rguenther@suse.de>
34068         PR tree-optimization/80492
34069         * alias.c (compare_base_decls): Handle registers with asm
34070         specification conservatively.
34071         * tree-ssa-alias.c (decl_refs_may_alias_p): Handle
34072         compare_base_decls returning dont-know properly.
34074 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34076         * config/arc/arc.c (LEGITIMATE_OFFSET_ADDRESS_P): Delete macro.
34077         (legitimate_offset_address_p): New function.
34078         (arc_legitimate_address_p): Use above function.
34080 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34082         * config/arc/arc.c (arc_output_mi_thunk): Emit PIC calls.
34084 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34086         * config/arc/arc.c (arc_conditional_register_usage): Use ACCL,
34087         ACCH registers whenever they are available.
34089 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34091         * config/arc/arc.c (arc_conditional_register_usage): Make D0, D1
34092         double regs fix when not used.
34094 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34096         * config/arc/arc.h (REGNO_OK_FOR_BASE_P): Consider also extension
34097         core registers.
34098         (REG_OK_FOR_INDEX_P_NONSTRICT): Likewise.
34099         (REG_OK_FOR_BASE_P_NONSTRICT): Likewise.
34101 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34103         * config/arc/arc.c (arc_output_addsi): Check for h-register class
34104         when emitting short ADD instructions.
34106 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34108         * config/arc/arc.md (cmpsi_cc_insn_mixed): Use 'h' register
34109         constraint.
34110         (cmpsi_cc_c_insn): Likewise.
34111         (cbranchsi4_scratch): Compute proper instruction length using
34112         compact_hreg_operand.
34113         * config/arc/predicates.md (compact_hreg_operand): New predicate.
34115 2017-04-25  Richard Biener  <rguenther@suse.de>
34117         PR middle-end/80509
34118         * passes.c (pass_manager::pass_manager): Initialize
34119         m_name_to_pass_map.
34121 2017-04-25  Richard Biener  <rguenther@suse.de>
34123         PR tree-optimization/79201
34124         * tree-ssa-sink.c (statement_sink_location): Handle calls.
34126 2017-04-25  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
34128         PR target/80464
34129         * config/s390/vector.md: Split MEM->GPR vector moves for
34130         non-s_operand addresses.
34132 2017-04-25  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
34134         PR target/79895
34135         * config/s390/predicates.md (reload_const_wide_int_operand): New
34136         predicate.
34137         * config/s390/s390.md ("movti"): Remove d/P alternative.
34138         ("movti_bigconst"): New pattern definition.
34140 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
34142         PR target/80080
34143         * s390-protos.h (s390_expand_cs_hqi): Removed.
34144         (s390_expand_cs, s390_expand_atomic_exchange_tdsi): New prototypes.
34145         * config/s390/s390.c (s390_emit_compare_and_swap): Handle all integer
34146         modes as well as CCZ1mode and CCZmode.
34147         (s390_expand_atomic_exchange_tdsi, s390_expand_atomic): Adapt to new
34148         signature of s390_emit_compare_and_swap.
34149         (s390_expand_cs_hqi): Likewise, make static.
34150         (s390_expand_cs_tdsi): Generate an explicit compare before trying
34151         compare-and-swap, in some cases.
34152         (s390_expand_cs): Wrapper function.
34153         (s390_expand_atomic_exchange_tdsi): New backend specific expander for
34154         atomic_exchange.
34155         (s390_match_ccmode_set): Allow CCZmode <-> CCZ1 mode.
34156         * config/s390/s390.md ("atomic_compare_and_swap<mode>"): Merge the
34157         patterns for small and large integers.  Forbid symref memory operands.
34158         Move expander to s390.c.  Require cc register.
34159         ("atomic_compare_and_swap<DGPR:mode><CCZZ1:mode>_internal")
34160         ("*atomic_compare_and_swap<TDI:mode><CCZZ1:mode>_1")
34161         ("*atomic_compare_and_swapdi<CCZZ1:mode>_2")
34162         ("*atomic_compare_and_swapsi<CCZZ1:mode>_3"): Use s_operand to forbid
34163         symref memory operands.  Remove CC mode and call s390_match_ccmode
34164         instead.
34165         ("atomic_exchange<mode>"): Allow and implement all integer modes.
34167 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
34169         * config/s390/s390.md (define_peephole2): New peephole to help
34170         combining the load-and-test pattern with volatile memory.
34172 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
34174         * config/s390/s390.md ("cstorecc4"): Use load-on-condition and deal
34175         with CCZmode for TARGET_Z196.
34177 2017-04-25  Jakub Jelinek  <jakub@redhat.com>
34179         PR rtl-optimization/80501
34180         * combine.c (make_compound_operation_int): Set subreg_code to SET
34181         even for AND with mask of the sign bit of mode.
34183         PR rtl-optimization/80500
34184         * loop-unroll.c (combine_var_copies_in_loop_exit): Call copy_rtx on
34185         sum's initial value.
34187 2017-04-25  Julian Brown  <julian@codesourcery.com>
34188             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
34190         * config/aarch64/thunderx2t99.md (thunderx2t99_crc): New Reservation.
34192 2017-04-25  Marc Glisse  <marc.glisse@inria.fr>
34194         * fold-const.c (tree_single_nonzero_warnv_p): Handle SSA_NAME.
34196 2017-04-25  Julian Brown  <julian@codesourcery.com>
34197             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
34199         * config/aarch64/thunderx2t99.md (thunderx2t99_aes): New Reservation.
34200         (thunderx2t99_sha): New Reservation.
34202 2017-04-25  Julian Brown  <julian@codesourcery.com>
34203             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
34205         * config/aarch64/aarch64-simd.md (aarch64_simd_vec_set<mode>): Fix
34206         type for 1-element load.
34208 2017-04-24  Marc Glisse  <marc.glisse@inria.fr>
34210         * match.pd (X/[ex]C CMP Y/[ex]C): New transformation.
34212 2017-04-24  Martin Jambor  <mjambor@suse.cz>
34214         PR tree-optimization/80293
34215         * tree-sra.c (scalarizable_type_p): New parameter const_decl, make
34216         char arrays not totally scalarizable if it is false.
34217         (analyze_all_variable_accesses): Pass correct value in the new
34218         parameter.  Add a statistics counter.
34220 2017-04-24  Jan Hubicka  <hubicka@ucw.cz>
34222         PR middle-end/79931
34223         * ipa-devirt.c (dump_possible_polymorphic_call_targets): Fix ICE.
34225 2017-04-24  Richard Biener  <rguenther@suse.de>
34227         PR tree-optimization/80494
34228         * tree-scalar-evolution.c (analyze_scalar_evolution_1): Bail
34229         out for complex types.
34231 2017-04-24  Richard Biener  <rguenther@suse.de>
34233         * tree-ssa-sccvn.h (run_scc_vn): Adjust prototype.
34234         * tree-ssa-sccvn.c (print_scc): Print SCC size.
34235         (extract_and_process_scc_for_name): Never fail but drop SCC to varying.
34236         (DFS): Adjust and never fail.
34237         (sccvn_dom_walker::fail): Remove.
34238         (sccvn_dom_walker::before_dom_children): Adjust.
34239         (run_scc_vn): Likewise and never fail.
34240         * tree-ssa-pre.c (pass_pre::execute): Adjust.
34241         (pass_fre::execute): Likewise.
34243 2017-04-24  Richard Biener  <rguenther@suse.de>
34245         PR tree-optimization/79725
34246         * tree-ssa-sink.c (statement_sink_location): Return whether
34247         failure reason was zero uses.  Move that check later.
34248         (sink_code_in_bb): Deal with zero uses by removing the stmt
34249         if possible.
34251 2017-04-24  Richard Biener  <rguenther@suse.de>
34253         PR c++/2972
34254         * tree-ssa-uninit.c (warn_uninitialized_vars): Handle some
34255         pointer-based references.
34257 2017-04-24  Richard Biener  <rguenther@suse.de>
34259         PR bootstrap/79814
34260         * pass_manager.h (pass_manager::operator new): Remove.
34261         (pass_manager::operator delete): Likewise.
34262         * passes.c (pass_manager::operator new): Remove.
34263         (pass_manager::operator delete): Likewise.
34264         (pass_manager::pass_manager): Zero individual pass members.
34266 2017-04-23  Uros Bizjak  <ubizjak@gmail.com>
34268         PR target/70799
34269         * config/i386/i386.c (dimode_scalar_to_vector_candidate_p)
34270         <case ASHIFT, case LSHIFTRT>: Also consider variable shifts.
34271         Check "XEXP (src, 1)" operand here.
34272         <case PLUS, case MINUS, case IOR, case XOR, case AND>:
34273         Check "XEXP (src, 1)" operand here.
34274         (dimode_scalar_chain::make_vector_copies): Detect count register
34275         of a shift instruction.  Zero extend count register from QImode
34276         to DImode to satisfy vector shift pattern count operand predicate.
34277         Substitute vector shift count operand with a DImode copy.
34278         (dimode_scalar_chain::convert_reg): Ditto, zero-extend from
34279         vector register.
34281 2017-04-21  Uros Bizjak  <ubizjak@gmail.com>
34283         * config/i386/i386.md (*extzvqi_mem_rex64): Move above *extzv<mode>.
34284         Remove UNSPEC_NOREX_MEM tag.  Update corresponding peephole2 pattern.
34285         (*insvqi_1_mem_rex64): Move above insv<mode>_1.  Remove
34286         UNSPEC_NOREX_MEM tag.  Update corresponding peephole2 pattern.
34287         (UNSPEC_NOREX_MEM): Remove definition.
34289 2017-04-21  Richard Biener  <rguenther@suse.de>
34291         PR tree-optimization/79547
34292         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
34293         Handle strlen, strcmp, strncmp, strcasecmp, strncasecmp, memcmp,
34294         bcmp, strspn, strcspn, __builtin_object_size and __builtin_constant_p
34295         without any constraints.
34297 2017-04-21  Richard Biener  <rguenther@suse.de>
34299         PR tree-optimization/78847
34300         * fold-const.c (split_tree): Handle POINTER_PLUS_EXPR.
34302 2017-04-21  Richard Biener  <rguenther@suse.de>
34304         * tree.h (build_qualified_type): Annotate with CXX_MEM_STAT_INFO.
34305         (build_distinct_type_copy): Likewise.
34306         (build_variant_type_copy): Likewise.
34307         * tree.c (build_qualified_type): Pass down mem-stat info.
34308         (build_distinct_type_copy): Likewise.
34309         (build_variant_type_copy): Likewise.
34311 2017-04-21  Richard Biener  <rguenther@suse.de>
34313         PR tree-optimization/80237
34314         * tree-ssa-pre.c (find_leader_in_sets): Add third set argument,
34315         defaulted to NULL.
34316         (phi_translate_1): Also allow a leader in AVAIL_OUT of pred
34317         for a simplified result.
34319 2016-04-21  Richard Biener  <rguenther@suse.de>
34321         * tree-ssa-loop-ivcanon.c (constant_after_peeling): Do not require
34322         sth as strict as a simple_iv but a chrec without symbols and an
34323         operand defined in the loop we are peeling (and not some subloop).
34324         (propagate_constants_for_unrolling): Propagate all constants.
34326 2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
34328         PR target/79804
34329         * config/i386/i386.c (print_reg): Remove assert for disalowed
34330         regno values, call output_operand_lossage instead.
34332 2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
34334         PR target/78090
34335         * config/i386/constraints.md (Yc): New register constraint.
34336         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
34337         Use Yc constraint for alternative 2 of operand 0.  Remove
34338         preferred_for_speed attribute.
34340 2017-04-20  Alexander Monakov  <amonakov@ispras.ru>
34342         * omp-low.c (lower_lastprivate_clauses): Correct handling of linear and
34343         lastprivate clauses in SIMT case.
34345 2017-04-20  Volker Reichelt  <v.reichelt@netcologne.de>
34347         * doc/invoke.texi (-Wextra-semi): Document new warning option.
34349 2017-04-20  Richard Biener  <rguenther@suse.de>
34351         PR tree-optimization/57796
34352         * tree-vect-stmts.c (vect_model_store_cost): Cost scatters
34353         as N scalar stores.
34354         (vect_model_load_cost): Cost gathers as N scalar loads.
34356 2017-04-20  Richard Biener  <rguenther@suse.de>
34358         * ggc-page.c (ggc_allocated_p): Rename to ...
34359         (safe_lookup_page_table_entry): ... this and return the lookup
34360         result.
34361         (gt_ggc_m_S): Use safe_lookup_page_table_entry.
34363 2017-04-20  Richard Biener  <rguenther@suse.de>
34365         PR tree-optimization/80453
34366         * tree-ssa-sccvn.h (struct vn_phi_s): Add cclhs and ccrhs members.
34367         * tree-ssa-sccvn.c (cond_stmts_equal_p): Use recorded lhs and rhs
34368         from the conditions.
34369         (vn_phi_eq): Pass them down.
34370         (vn_phi_lookup): Record them.
34371         (vn_phi_insert): Likewise.
34373 2017-04-20  Matthew Fortune  <matthew.fortune@imgtec.com>
34375         * config/mips/mips.c (mips_expand_vec_perm_const): Re-fix
34376         uninitialized variable warning to avoid buffer overrun.
34378 2017-04-20  Alexander Monakov  <amonakov@ispras.ru>
34380         PR other/71250
34381         * doc/invoke.texi (-Wmissing-field-initializers): Mention that warning
34382         is suppressed for '{ 0 }' in C.
34384 2017-04-20  Jakub Jelinek  <jakub@redhat.com>
34386         * BASE-VER: Set to 8.0.0.
34388 2017-04-20  Thomas Preud'homme  <thomas.preudhomme@arm.com>
34390         * config/arm/arm.c (arm_elf_asm_cdtor): Create non-default
34391         priority .init_array and .fini_array section with SECTION_NOTYPE
34392         flag.
34394 2017-04-20  Jakub Jelinek  <jakub@redhat.com>
34396         PR middle-end/80423
34397         * tree.h (build_array_type): Add typeless_storage default argument.
34398         * tree.c (type_cache_hasher::equal): Also compare
34399         TYPE_TYPELESS_STORAGE flag for ARRAY_TYPEs.
34400         (build_array_type): Add typeless_storage argument, set
34401         TYPE_TYPELESS_STORAGE to it, if shared also hash it, and pass to
34402         recursive call.
34403         (build_nonshared_array_type): Adjust build_array_type_1 caller.
34404         (build_array_type): Likewise.  Add typeless_storage argument.
34406 2017-04-19  Eric Botcazou  <ebotcazou@adacore.com>
34407             Jakub Jelinek  <jakub@redhat.com>
34409         PR tree-optimization/80426
34410         * tree-vrp.c (extract_range_from_binary_expr_1): For an additive
34411         operation on symbolic operands, also compute the overflow for the
34412         invariant part when the operation degenerates into a negation.
34414 2017-04-19  Jakub Jelinek  <jakub@redhat.com>
34416         PR debug/80461
34417         * dwarf2out.c (modified_type_die, gen_type_die_with_usage):
34418         Check for t with zero TYPE_QUALS_NO_ADDR_SPACE.
34420         PR debug/80436
34421         * tree-ssa-loop-manip.c (find_uses_to_rename_def): Ignore debug uses.
34423 2017-04-19  Georg-Johann Lay  <avr@gjlay.de>
34425         PR target/80462
34426         * config/avr/avr.c (tree.h): Include it.
34427         (cgraph.h): Include it.
34428         (avr_encode_section_info): Don't warn for uninitialized progmem
34429         variable if it's just an alias.
34431 2017-04-19  Richard Biener  <rguenther@suse.de>
34433         PR ipa/65972
34434         * auto-profile.c (afdo_vpt_for_early_inline): Update SSA
34435         when needed by AutoPGO.
34437 2017-04-19  Paulo J. Matos  <paulo@matos-sorge.com>
34439         PR lto/50345
34440         * doc/lto.texi: Remove an extra 'that'.
34442 2017-04-19  Segher Boessenkool  <segher@kernel.crashing.org>
34444         PR rtl-optimization/80429
34445         * ira.c (split_live_ranges_for_shrink_wrap): Don't split regs that
34446         are only used in debug insns.
34448 2017-04-19  Eric Botcazou  <ebotcazou@adacore.com>
34449             Vladimir Makarov  <vmakarov@redhat.com>
34451         * config/sparc/predicates.md (input_operand): Add comment.  Return
34452         true for any memory operand when LRA is in progress.
34453         * config/sparc/sparc.c (sparc_expand_move): Minor formatting fix.
34455 2017-04-18  Jeff Law  <law@redhat.com>
34457         PR target/74563
34458         * mips.md ({return,simple_return}_internal): Do not overwrite
34459         operands[0].
34461 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34463         PR tree-optimization/80443
34464         * tree-vrp.c (intersect_ranges): For signed 1-bit precision type,
34465         instead of adding 1, subtract -1 and similarly instead of subtracting
34466         1 add -1.
34468 2017-04-18  Richard Sandiford  <richard.sandiford@arm.com>
34470         PR rtl-optimization/80357
34471         * haifa-sched.c (tmp_bitmap): New variable.
34472         (model_recompute): Handle duplicate use records.
34473         (alloc_global_sched_pressure_data): Initialize tmp_bitmap.
34474         (free_global_sched_pressure_data): Free it.
34476 2017-04-18  Bernd Edlinger  <bernd.edlinger@hotmail.de>
34478         Revert:
34479         2017-02-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
34480         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
34481         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
34482         instead of SYSTEM_HEADER_DIR.
34484 2017-04-18  Jeff Law  <law@redhat.com>
34486         PR middle-end/80422
34487         * cfgcleanup.c (try_crossjump_to_edge): Verify SRC1 and SRC2 have
34488         predecessors after walking up the insn chain.
34490 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34492         PR debug/80263
34493         * dwarf2out.c (modified_type_die): Try harder not to emit internal
34494         sizetype type into debug info.
34496 2017-04-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
34498         PR target/80099
34499         * config/rs6000/rs6000.c (rs6000_expand_vector_extract): Eliminate
34500         unneeded test for TARGET_UPPER_REGS_SF.
34501         * config/rs6000/vsx.md (vsx_extract_v4sf_var): Likewise.
34503 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34505         PR sanitizer/80444
34506         * sancov.c (sancov_pass): Use gsi_start_nondebug_after_labels_bb
34507         instead of gsi_after_labels.
34509 2017-04-18  Jeff Law  <law@redhat.com>
34511         * regcprop.c (maybe_mode_change): Avoid creating copies of the
34512         stack pointer.
34514         Revert:
34515         2017-04-13  Jeff Law  <law@redhat.com>
34516         * config/mips.mips.md (zero_extendsidi2): Do not allow SP to appear
34517         in operands[1] if it is a MEM and TARGET_MIPS16 is active.
34519 2017-04-18  Georg-Johann Lay  <avr@gjlay.de>
34521         PR target/79453
34522         * config/avr/avr.c (intl.h): Include it.
34523         (avr_pgm_check_var_decl) [reason]: Wrap diagnostic snippets into _().
34525 2017-04-18  Martin Liska  <mliska@suse.cz>
34527         PR gcov-profile/78783
34528         * gcov-tool.c (gcov_output_files): Validate that destination
34529         file is either removed by the tool or by a user.
34531 2017-04-14  Andrew Burgess  <andrew.burgess@embecosm.com>
34532             Guy Benyei  <guybe@mellanox.com>
34534         * config/arc/arc.c (arc_reorg): Move loop_end_id into a more local
34535         block, and do not negate it, the stored id is already negative.
34537 2017-04-14  Andrew Burgess  <andrew.burgess@embecosm.com>
34539         * config/arc/arc.md (doloop_begin_i): Use @pcl assembler syntax.
34541 2017-04-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
34543         PR target/80098
34544         * config/rs6000/rs6000-cpus.def (OTHER_P9_VECTOR_MASKS): Define
34545         masks of options that should be turned off if the VSX vector
34546         options are turned off.
34547         (OTHER_P8_VECTOR_MASKS): Likewise.
34548         (OTHER_VSX_VECTOR_MASKS): Likewise.
34549         * config/rs6000/rs6000.c (rs6000_option_override_internal): Call
34550         rs6000_disable_incompatible_switches to validate no type switches
34551         like -mvsx.
34552         (rs6000_incompatible_switch): New function to disallow turning on
34553         other vector options if -mno-vsx, -mno-power8-vector, or
34554         -mno-power9-vector are specified.
34556 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34558         * config/arc/arc.h (CRT_CALL_STATIC_FUNCTION): Use long calls.
34560 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34562         * config/arc/arc-protos.h (arc_decl_pretend_args): Remove.
34563         * config/arc/arc.c (arc_decl_pretend_args): Likewise.
34564         * config/arc/arc.h (CFA_FRAME_BASE_OFFSET): Likewise.
34565         (ARG_POINTER_CFA_OFFSET): Likewise.
34567 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34569         * config/arc/arc.c (arc_mode_dependent_address_p): Relax
34570         conditions to take advantage of various optimizations.
34572 2017-04-13  Jeff Law  <law@redhat.com>
34574         * config/mips.mips.md (zero_extendsidi2): Do not allow SP to appear
34575         in operands[1] if it is a MEM and TARGET_MIPS16 is active.
34576         (zero_extendsidi2_dext): Likewise.
34578 2017-04-13  Jakub Jelinek  <jakub@redhat.com>
34580         PR sanitizer/80403
34581         * fold-const.c (fold_ternary_loc): Revert
34582         use op0 instead of fold_convert_loc (loc, type, arg0) part of
34583         2017-04-12 change.
34585 2017-04-13  Vladimir Makarov  <vmakarov@redhat.com>
34587         PR rtl-optimization/80343
34588         * lra-remat.c (update_scratch_ops): Assign original hard reg to
34589         new scratch pseudo.
34591 2017-04-13  Denis Khalikov <d.khalikov@partner.samsung.com>
34593         PR sanitizer/80414
34594         * ubsan.c (ubsan_expand_bounds_ifn): Pass original index
34595         to ubsan_encode_value.
34597 2017-04-13  Jeff Law  <law@redhat.com>
34599         * reload1.c (eliminate_regs_1): Call gen_rtx_raw_SUBREG for SUBREGs
34600         appearing in DEBUG_INSNs.
34602 2017-04-13  Martin Liska  <mliska@suse.cz>
34604         PR gcov-profile/80413
34605         * gcov-io.c (gcov_write_string): Copy to buffer just when
34606         allocated size is greater than zero.
34608 2017-04-13  Jakub Jelinek  <jakub@redhat.com>
34610         PR debug/80321
34611         * dwarf2out.c (decls_for_scope): Ignore declarations of
34612         current_function_decl in BLOCK_NONLOCALIZED_VARS.
34614 2017-04-12  Jan Hubicka  <hubicka@ucw.cz>
34616         PR lto/69953
34617         * ipa-visibility.c (non_local_p): Fix typos.
34618         (localize_node): When localizing symbol in same comdat group,
34619         dissolve the group only when we know external symbols are going
34620         to be privatized.
34621         (function_and_variable_visibility): Do not localize DECL_EXTERNAL.
34623 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34625         PR tree-optimization/79390
34626         * optabs.c (emit_conditional_move): If the preferred op2/op3 operand
34627         order does not result in usable sequence, retry with reversed operand
34628         order.
34630         PR sanitizer/80403
34631         PR sanitizer/80404
34632         PR sanitizer/80405
34633         * fold-const.c (fold_ternary_loc): Use op1 instead of arg1 as argument
34634         to fold_build2_loc.  Convert TREE_OPERAND (tem, 0) to type.  Use
34635         op0 instead of fold_convert_loc (loc, type, arg0).
34637 2017-04-12  Jeff Law  <law@redhat.com>
34639         * genattrtab.c (write_eligible_delay): Verify DELAY_INSN still
34640         has a delay slot in the generated code.
34642         * config/cris/cris.md (cris_preferred_reload_class): Return
34643         GENNONACR_REGS rather than GENERAL_REGS.
34645 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34647         PR c/80163
34648         * expr.c <CASE_CONVERT>: For EXPAND_INITIALIZER determine SIGN_EXTEND
34649         vs. ZERO_EXTEND based on signedness of treeop0's type rather than
34650         signedness of the result type.
34652 2017-04-12  Richard Biener  <rguenther@suse.de>
34653             Jeff Law  <law@redhat.com>
34655         PR tree-optimization/80359
34656         * tree-ssa-dse.c (maybe_trim_partially_dead_store): Do not
34657         trim stores to TARGET_MEM_REFs.
34659 2017-04-12  Richard Biener  <rguenther@suse.de>
34661         PR tree-optimization/79390
34662         * gimple-ssa-split-paths.c (is_feasible_trace): Restrict
34663         threading case even more.
34665 2017-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
34667         PR target/80382
34668         * config/rs6000/sync.md (atomic_load<mode>, atomic_store<mode): Test
34669         for quad_address_p for TImode, instead of just not indexed_address.
34671 2017-04-12  Richard Biener  <rguenther@suse.de>
34672             Bernd Edlinger  <bernd.edlinger@hotmail.de>
34674         PR middle-end/79671
34675         * alias.c (component_uses_parent_alias_set_from): Handle
34676         TYPE_TYPELESS_STORAGE.
34677         (get_alias_set): Likewise.
34678         * tree-core.h (tree_type_common): Add typeless_storage flag.
34679         * tree.h (TYPE_TYPELESS_STORAGE): New macro.
34680         * stor-layout.c (place_union_field): Set TYPE_TYPELESS_STORAGE
34681         for types containing members with TYPE_TYPELESS_STORAGE.
34682         (place_field): Likewise.
34683         (layout_type): Likewise for ARRAY_TYPE.
34684         * lto-streamer-out.c (hash_tree): Hash TYPE_TYPELESS_STORAGE.
34685         * tree-streamer-in.c (unpack_ts_type_common_value_fields): Stream
34686         TYPE_TYPELESS_STORAGE.
34687         * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
34689 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34691         PR sanitizer/80349
34692         * fold-const.c (fold_binary_loc) <case BIT_IOR_EXPR>: Convert arg0's
34693         first argument to type.
34695 2017-04-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
34697         PR target/80376
34698         PR target/80315
34699         * config/rs6000/rs6000.c (rs6000_expand_unop_builtin): Return
34700         CONST0_RTX (mode) rather than const0_rtx where appropriate.
34701         (rs6000_expand_binop_builtin): Likewise.
34702         (rs6000_expand_ternop_builtin): Likewise; also add missing
34703         vsx_xxpermdi_* variants; also fix typo (arg1 => arg2) for
34704         vshasigma built-ins.
34705         * doc/extend.texi: Document that vec_xxpermdi's third argument
34706         must be a constant.
34708 2017-04-11  Uros Bizjak  <ubizjak@gmail.com>
34710         * config/i386/i386.c (dimode_scalar_chain::compute_convert_gain):
34711         Use shift_const cost parameter when calculating gain of STV shifts.
34713 2017-04-11  Vladimir Makarov  <vmakarov@redhat.com>
34715         PR rtl-optimization/70478
34716         * lra-constraints.c (process_alt_operands): Check memory for
34717         disfavoring memory insn operand.
34719 2017-04-11  Jakub Jelinek  <jakub@redhat.com>
34721         PR middle-end/80100
34722         * simplify-rtx.c (simplify_binary_operation_1) <case IOR>: Perform
34723         left shift in unsigned HOST_WIDE_INT type.
34725         PR rtl-optimization/80385
34726         * simplify-rtx.c (simplify_unary_operation_1): Don't transform
34727         (not (neg X)) into (plus X -1) for complex or non-integral modes.
34729         PR libgomp/80394
34730         * omp-low.c (scan_omp_task): Don't optimize away empty tasks
34731         if they have any depend clauses.
34733 2017-04-11  Martin Liska  <mliska@suse.cz>
34735         PR ipa/80212
34736         * cgraph.c (cgraph_node::dump): Dump calls_comdat_local.
34737         * ipa-split.c (split_function): Create a local comdat symbol
34738         if caller is in a comdat group.
34740 2017-04-11  Martin Liska  <mliska@suse.cz>
34742         PR ipa/80212
34743         * ipa-cp.c (determine_versionability): Handle calls_comdat_local
34744         flags.
34746 2017-04-11  Martin Sebor  <msebor@redhat.com>
34748         PR middle-end/80364
34749         * gimple-ssa-sprintf.c (get_int_range): Remove second argument and
34750         always use the int type.  Use INTEGRAL_TYPE_P() rather than testing
34751         for INTEGER_TYPE.
34752         (directive::set_width, directive::set_precision, format_character):
34753         Adjust.
34754         (parse_directive): Use INTEGRAL_TYPE_P() rather than testing for
34755         INTEGER_TYPE.
34757 2017-04-11  Richard Earnshaw  <rearnsha@arm.com>
34759         PR target/80389
34760         * config/arm/arm.c (arm_configure_build_target): When -mcpu and -arch
34761         conflict, set target->arch_name instead of target->cpu_name.
34763 2017-04-11  Richard Biener  <rguenther@suse.de>
34765         PR tree-optimization/80374
34766         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Use
34767         build_zero_cst, remove fold_convertible_p check again.
34769 2017-04-11  Martin Liska  <mliska@suse.cz>
34771         PR sanitizer/70878
34772         * ubsan.c (instrument_object_size): Do not instrument register
34773         variables.
34775 2017-04-11  Jakub Jelinek  <jakub@redhat.com>
34777         PR target/80381
34778         * config/i386/i386-builtin-types.def
34779         (V16HI_FTYPE_V16HI_INT_V16HI_UHI_COUNT,
34780         V16HI_FTYPE_V16HI_V8HI_V16HI_UHI_COUNT,
34781         V16SI_FTYPE_V16SI_INT_V16SI_UHI_COUNT,
34782         V16SI_FTYPE_V16SI_V4SI_V16SI_UHI_COUNT,
34783         V2DI_FTYPE_V2DI_INT_V2DI_UQI_COUNT,
34784         V2DI_FTYPE_V2DI_V2DI_V2DI_UQI_COUNT,
34785         V32HI_FTYPE_V32HI_INT_V32HI_USI_COUNT,
34786         V32HI_FTYPE_V32HI_V8HI_V32HI_USI_COUNT,
34787         V4DI_FTYPE_V4DI_INT_V4DI_UQI_COUNT,
34788         V4DI_FTYPE_V4DI_V2DI_V4DI_UQI_COUNT,
34789         V4SI_FTYPE_V4SI_INT_V4SI_UQI_COUNT,
34790         V4SI_FTYPE_V4SI_V4SI_V4SI_UQI_COUNT,
34791         V8DI_FTYPE_V8DI_INT_V8DI_UQI_COUNT,
34792         V8DI_FTYPE_V8DI_V2DI_V8DI_UQI_COUNT,
34793         V8HI_FTYPE_V8HI_INT_V8HI_UQI_COUNT,
34794         V8HI_FTYPE_V8HI_V8HI_V8HI_UQI_COUNT,
34795         V8SI_FTYPE_V8SI_INT_V8SI_UQI_COUNT,
34796         V8SI_FTYPE_V8SI_V4SI_V8SI_UQI_COUNT): New function type aliases.
34797         * config/i386/i386-builtin.def (__builtin_ia32_pslld512_mask,
34798         __builtin_ia32_pslldi512_mask, __builtin_ia32_psllq512_mask,
34799         __builtin_ia32_psllqi512_mask, __builtin_ia32_psrad512_mask,
34800         __builtin_ia32_psradi512_mask, __builtin_ia32_psraq512_mask,
34801         __builtin_ia32_psraqi512_mask, __builtin_ia32_psrld512_mask,
34802         __builtin_ia32_psrldi512_mask, __builtin_ia32_psrlq512_mask,
34803         __builtin_ia32_psrlqi512_mask, __builtin_ia32_psllwi128_mask,
34804         __builtin_ia32_pslldi128_mask, __builtin_ia32_psllqi128_mask,
34805         __builtin_ia32_psllw128_mask, __builtin_ia32_pslld128_mask,
34806         __builtin_ia32_psllq128_mask, __builtin_ia32_psllwi256_mask,
34807         __builtin_ia32_psllw256_mask, __builtin_ia32_pslldi256_mask,
34808         __builtin_ia32_pslld256_mask, __builtin_ia32_psllqi256_mask,
34809         __builtin_ia32_psllq256_mask, __builtin_ia32_psradi128_mask,
34810         __builtin_ia32_psrad128_mask, __builtin_ia32_psradi256_mask,
34811         __builtin_ia32_psrad256_mask, __builtin_ia32_psraqi128_mask,
34812         __builtin_ia32_psraq128_mask, __builtin_ia32_psraqi256_mask,
34813         __builtin_ia32_psraq256_mask, __builtin_ia32_psrldi128_mask,
34814         __builtin_ia32_psrld128_mask, __builtin_ia32_psrldi256_mask,
34815         __builtin_ia32_psrld256_mask, __builtin_ia32_psrlqi128_mask,
34816         __builtin_ia32_psrlq128_mask, __builtin_ia32_psrlqi256_mask,
34817         __builtin_ia32_psrlq256_mask, __builtin_ia32_psrawi256_mask,
34818         __builtin_ia32_psraw256_mask, __builtin_ia32_psrawi128_mask,
34819         __builtin_ia32_psraw128_mask, __builtin_ia32_psrlwi256_mask,
34820         __builtin_ia32_psrlw256_mask, __builtin_ia32_psrlwi128_mask,
34821         __builtin_ia32_psrlw128_mask, __builtin_ia32_psllwi512_mask,
34822         __builtin_ia32_psllw512_mask, __builtin_ia32_psrawi512_mask,
34823         __builtin_ia32_psraw512_mask, __builtin_ia32_psrlwi512_mask,
34824         __builtin_ia32_psrlw512_mask): Use _COUNT suffixed function type
34825         aliases.
34826         * config/i386/i386.c (ix86_expand_args_builtin): Rename last_arg_count
34827         flag to second_arg_count, handle 4 argument function type _COUNT
34828         aliases, handle second_arg_count on second argument rather than last.
34830 2017-04-10  Jeff Law  <law@redhat.com>
34832         PR tree-optimization/80374
34833         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Do not try to
34834         record anything if we can not convert integer_zero_node to the
34835         desired type.
34837 2017-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
34839         PR target/80108
34840         * config/rs6000/rs6000.c (rs6000_option_override_internal):
34841         Enhance special handling given to the TARGET_P9_MINMAX option in
34842         relation to certain other options.
34844 2017-04-10  Bin Cheng  <bin.cheng@arm.com>
34846         PR tree-optimization/80153
34847         * tree-ssa-loop-ivopts.c (add_iv_candidate_for_use): Check and
34848         remove POINTER_PLUS_EXPR's base part directly, rather than through
34849         aff_tree.
34851 2017-04-10  Richard Biener  <rguenther@suse.de>
34852             Bin Cheng  <bin.cheng@arm.com>
34854         PR tree-optimization/80153
34855         * tree-affine.c (aff_combination_to_tree): Get base pointer from
34856         the first element of pointer type aff_tree.  Build result expr in
34857         aff_tree's type.
34858         (add_elt_to_tree): Convert to type unconditionally.  Remove other
34859         fold_convert calls.
34860         * tree-ssa-loop-ivopts.c (alloc_iv): Pass in consistent types.
34861         (rewrite_use_nonlinear_expr): Check invariant using iv information.
34863 2017-04-10  Richard Biener  <rguenther@suse.de>
34865         * tree-ssa-structalias.c (find_func_aliases): Properly handle
34866         asm inputs.
34868 2017-04-10  Vladimir Makarov  <vmakarov@redhat.com>
34870         PR rtl-optimization/70478
34871         * lra-constraints.c (curr_small_class_check): New.
34872         (update_and_check_small_class_inputs): New.
34873         (process_alt_operands): Update curr_small_class_check.  Disfavor
34874         alternative insn memory operands.  Check available regs for small
34875         class operands.
34877 2017-03-31  Matthew Fortune  <matthew.fortune@imgtec.com>
34879         PR target/80057
34880         * config/mips/mips.opt (-mvirt): Update description.
34881         * doc/invoke.texi (-mvirt): Likewise.
34883 2017-04-10  Richard Biener  <rguenther@suse.de>
34885         PR middle-end/80362
34886         * fold-const.c (fold_binary_loc): Look at unstripped ops when
34887         looking for NEGATE_EXPR in -A / -B to A / B folding.
34889 2017-04-10  Martin Liska  <mliska@suse.cz>
34891         PR gcov-profile/80224
34892         * gcov.c (print_usage): Fix usage string.
34893         (get_gcov_intermediate_filename): Remove.
34894         (output_gcov_file): Use both for normal and intermediate format.
34895         (generate_results): Do not initialize special file for
34896         intermediate format.
34898 2017-04-10  Richard Biener  <rguenther@suse.de>
34900         PR tree-optimization/80304
34901         * tree-ssa-loop-im.c (ref_indep_loop_p_1): Also recurse
34902         for safelen.
34904 2017-04-10  Nathan Sidwell  <nathan@acm.org>
34906         PR target/79905
34907         * config/rs6000/rs6000.c (rs6000_vector_type): New.
34908         (rs6000_init_builtins): Use it.
34910 2016-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
34912         * config/arm/arm.md (<mrc>): Add mode to SET source.
34913         (<mrrc>): Likewise.
34915 2017-04-10  Richard Biener  <rguenther@suse.de>
34917         PR middle-end/80344
34918         * gimplify.c (is_gimple_mem_rhs_or_call): Allow CLOBBERs.
34920 2017-04-10  Jakub Jelinek  <jakub@redhat.com>
34922         PR target/80324
34923         * config/i386/avx512fintrin.h (_mm512_reduce_add_epi32,
34924         _mm512_reduce_mul_epi32, _mm512_reduce_and_epi32,
34925         _mm512_reduce_or_epi32, _mm512_mask_reduce_add_epi32,
34926         _mm512_mask_reduce_mul_epi32, _mm512_mask_reduce_and_epi32,
34927         _mm512_mask_reduce_or_epi32, _mm512_reduce_min_epi32,
34928         _mm512_reduce_max_epi32, _mm512_reduce_min_epu32,
34929         _mm512_reduce_max_epu32, _mm512_mask_reduce_min_epi32,
34930         _mm512_mask_reduce_max_epi32, _mm512_mask_reduce_min_epu32,
34931         _mm512_mask_reduce_max_epu32, _mm512_reduce_add_ps,
34932         _mm512_reduce_mul_ps, _mm512_mask_reduce_add_ps,
34933         _mm512_mask_reduce_mul_ps, _mm512_reduce_min_ps, _mm512_reduce_max_ps,
34934         _mm512_mask_reduce_min_ps, _mm512_mask_reduce_max_ps,
34935         _mm512_reduce_add_epi64, _mm512_reduce_mul_epi64,
34936         _mm512_reduce_and_epi64, _mm512_reduce_or_epi64,
34937         _mm512_mask_reduce_add_epi64, _mm512_mask_reduce_mul_epi64,
34938         _mm512_mask_reduce_and_epi64, _mm512_mask_reduce_or_epi64,
34939         _mm512_reduce_min_epi64, _mm512_reduce_max_epi64,
34940         _mm512_mask_reduce_min_epi64, _mm512_mask_reduce_max_epi64,
34941         _mm512_reduce_min_epu64, _mm512_reduce_max_epu64,
34942         _mm512_mask_reduce_min_epu64, _mm512_mask_reduce_max_epu64,
34943         _mm512_reduce_add_pd, _mm512_reduce_mul_pd, _mm512_mask_reduce_add_pd,
34944         _mm512_mask_reduce_mul_pd, _mm512_reduce_min_pd, _mm512_reduce_max_pd,
34945         _mm512_mask_reduce_min_pd, _mm512_mask_reduce_max_pd): New intrinsics.
34947 2017-04-08  Vladimir Makarov  <vmakarov@redhat.com>
34949         PR rtl-optimization/70478
34950         * lra-constraints.c: Reverse the last patch.
34952 2017-04-08  Andreas Tobler  <andreast@gcc.gnu.org>
34954         * config/aarch64/aarch64-freebsd.h: Define MCOUNT_NAME.
34955         Add comment for WCHAR_T.
34957 2017-04-08  Martin Liska  <mliska@suse.cz>
34959         Revert:
34960         2017-04-07  Martin Liska  <mliska@suse.cz>
34962         PR ipa/80212
34963         * ipa-split.c (split_function): Add function part to a same comdat
34964         group.
34966 2017-04-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
34968         PR target/80358
34969         * config/rs6000/rs6000.c (expand_block_compare): Fix boundary check.
34971 2017-04-07  Pat Haugen  <pthaugen@us.ibm.com>
34973         * rs6000/rs6000.c (vec_load_pendulum): Rename...
34974         (vec_pairing): ...to this.
34975         (power9_sched_reorder2): Rewrite code for pairing vector/vecload insns.
34976         (rs6000_sched_init): Adjust for name change.
34977         (struct rs6000_sched_context): Likewise.
34978         (rs6000_init_sched_context): Likewise.
34979         (rs6000_set_sched_context): Likewise.
34981 2017-04-07  Jakub Jelinek  <jakub@redhat.com>
34983         PR target/80322
34984         PR target/80323
34985         PR target/80325
34986         PR target/80326
34987         * config/i386/avxintrin.h (_mm256_cvtsd_f64, _mm256_cvtss_f32): New
34988         intrinsics.
34989         * config/i386/avx512fintrin.h (_mm512_int2mask, _mm512_mask2int,
34990         _mm512_abs_ps, _mm512_mask_abs_ps, _mm512_abs_pd, _mm512_mask_abs_pd,
34991         _mm512_cvtsd_f64, _mm512_cvtss_f32): Likewise.
34993 2017-04-07  Andreas Tobler  <andreast@gcc.gnu.org>
34995         * config/aarch64/aarch64-freebsd.h: Define WCHAR_TYPE.
34997 2017-04-07  Vladimir Makarov  <vmakarov@redhat.com>
34999         PR rtl-optimization/70703
35000         * ira-color.c (update_conflict_hard_regno_costs): Use
35001         int64_t instead of HOST_WIDE_INT.
35003 2017-04-07  Vladimir Makarov  <vmakarov@redhat.com>
35005         PR rtl-optimization/70478
35006         * lra-constraints.c (process_alt_operands): Disfavor alternative
35007         insn memory operands.
35009 2017-04-07  Jeff Law  <law@redhat.com>
35011         * config/iq2000/iq2000.c (final_prescan_insn): Do not separate a
35012         CALL and NOTE_INSN_CALL_ARG_LOCATION.
35014 2017-04-07  Martin Liska  <mliska@suse.cz>
35016         PR target/79889
35017         * config/aarch64/aarch64.c (aarch64_process_target_attr):
35018         Show error message instead of an ICE.
35020 2017-04-07  Martin Liska  <mliska@suse.cz>
35022         PR ipa/80212
35023         * ipa-split.c (split_function): Add function part to a same comdat
35024         group.
35026 2017-04-07  Richard Biener  <rguenther@suse.de>
35028         PR middle-end/80341
35029         * tree.c (get_unwidened): Also handle ! for_type case for
35030         INTEGER_CSTs.
35031         * convert.c (do_narrow): Split out from ...
35032         (convert_to_integer_1): ... here.  Do not pass final truncation
35033         type to get_unwidened for TRUNC_DIV_EXPR.
35035 2017-04-07  Richard Biener  <rguenther@suse.de>
35037         * tree-affine.c (wide_int_ext_for_comb): Take type rather
35038         than aff_tree.
35039         (aff_combination_const): Adjust.
35040         (aff_combination_scale): Likewise.
35041         (aff_combination_add_elt): Likewise.
35042         (aff_combination_add_cst): Likewise.
35043         (aff_combination_convert): Likewise.
35044         (add_elt_to_tree): Likewise.  Remove unused argument.
35045         (aff_combination_to_tree): Adjust calls to add_elt_to_tree.
35047 2017-04-07  Sebastian Huber  <sebastian.huber@embedded-brains.de>
35049         * config/arm/arm.h (ARM_DEFAULT_SHORT_ENUMS): Provide default
35050         definition.
35051         * config/arm/arm.c (arm_default_short_enums): Use
35052         ARM_DEFAULT_SHORT_ENUMS.
35053         * config/arm/rtems.h (ARM_DEFAULT_SHORT_ENUMS): Define.
35055 2017-04-06  Jakub Jelinek  <jakub@redhat.com>
35057         PR debug/80234
35058         * dwarf2out.c (gen_member_die): Handle C++17 inline static data
35059         members with redundant out-of-class redeclaration.
35061 2017-04-06  Uros Bizjak  <ubizjak@gmail.com>
35063         PR target/80286
35064         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): New pattern.
35065         * config/i386/i386.md (*zero_extendsidi2):
35066         Add (?*x,*x) and (?*v,*v) alternatives.
35068 2017-04-06  Uros Bizjak  <ubizjak@gmail.com>
35070         PR target/79733
35071         * config/i386/i386.c (ix86_expand_builtin)
35072         <case IX86_BUILTIN_K{,OR}TEST{C,Z}{8,16,32,64}>: Determine insn operand
35073         mode from insn data. Convert operands to insn operand mode.
35074         Copy operands that don't satisfy insn predicate to a register.
35076 2017-04-06  Sam Thursfield  <sam.thursfield@codethink.co.uk>
35078         * config/rs6000/x-aix: Increase memory limit for genautomata on AIX.
35079         Update comments.
35081 2017-04-06  Richard Biener  <rguenther@suse.de>
35083         PR tree-optimization/80334
35084         * tree-ssa-loop-ivopts.c (rewrite_use_address): Properly
35085         preserve alignment of accesses.
35087 2017-04-06  Richard Biener  <rguenther@suse.de>
35089         PR tree-optimization/80262
35090         * tree-sra.c (build_ref_for_offset): Preserve address-space
35091         information.
35092         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address):
35093         Drop useless address-space information on MEM_REF offsets.
35095 2017-04-05  Andreas Schwab  <schwab@linux-m68k.org>
35097         * builtins.def (BUILT_IN_UPDATE_SETJMP_BUF): Fix type.
35099 2017-04-05  Vladimir Makarov  <vmakarov@redhat.com>
35101         PR rtl-optimization/70703
35102         * ira-color.c (update_conflict_hard_regno_costs): Use
35103         HOST_WIDE_INT instead of long.
35105 2017-04-05  Uros Bizjak  <ubizjak@gmail.com>
35107         PR target/80298
35108         * config/i386/mmintrin.h: Add -msse target option when __SSE__ is
35109         not defined for x86_64 target.  Add -mmmx target option when __SSE2__
35110         is not defined.
35111         * config/i386/mm3dnow.h: Add -msse target when __SSE__ is not defined
35112         for x86_64 target.  Handle -m3dnowa option.
35114 2017-04-05  Vladimir Makarov  <vmakarov@redhat.com>
35116         PR rtl-optimization/70703
35117         * ira-color.c (update_costs_from_allocno): Use the smallest mode.
35118         (update_conflict_hard_regno_costs): Use long instead of unsigned
35119         arithmetic for cost calculation.
35121 2017-04-05  Jakub Jelinek  <jakub@redhat.com>
35122             Bernd Edlinger  <bernd.edlinger@hotmail.de>
35124         PR sanitizer/80308
35125         * asan.c (asan_store_shadow_bytes): Fix location of last_chunk_value
35126         for big endian.
35128 2017-04-05  Eric Botcazou  <ebotcazou@adacore.com>
35130         PR target/78002
35131         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Replace
35132         ptr_mode with Pmode throughout.
35133         * config/aarch64/aarch64.md (probe_stack_range_<PTR:mode): Rename
35134         into probe_stack_range and use DImode.
35136 2017-04-05  Dominik Vogt  <vogt@linux.vnet.ibm.com>
35138         PR target/79890
35139         * config/s390/s390.c (s390_register_info_gprtofpr): Return if
35140         call_eh_return is true.
35142 2017-04-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35144         * config/s390/s390-c.c (s390_resolve_overloaded_builtin):
35145         Initialize last_match_fntype_index.
35147 2017-04-05  Jakub Jelinek  <jakub@redhat.com>
35149         PR target/80310
35150         * tree-nvr.c: Include internal-fn.h.
35151         (pass_return_slot::execute): Ignore internal calls without
35152         direct optab.
35154 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
35155             Richard Biener  <rguenther@suse.de>
35157         PR c++/80297
35158         * genmatch.c (capture::gen_transform): For GENERIC unshare_expr
35159         captures used multiple times, except for the last use.
35160         * generic-match-head.c: Include gimplify.h.
35162 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
35164         PR tree-optimization/79390
35165         * target.h (struct noce_if_info): Declare.
35166         * targhooks.h (default_noce_conversion_profitable_p): Declare.
35167         * target.def (noce_conversion_profitable_p): New target hook.
35168         * ifcvt.h (struct noce_if_info): New type, moved from ...
35169         * ifcvt.c (struct noce_if_info): ... here.
35170         (noce_conversion_profitable_p): Renamed to ...
35171         (default_noce_conversion_profitable_p): ... this.  No longer
35172         static nor inline.
35173         (noce_try_store_flag_constants, noce_try_addcc,
35174         noce_try_store_flag_mask, noce_try_cmove, noce_try_cmove_arith,
35175         noce_convert_multiple_sets): Use targetm.noce_conversion_profitable_p
35176         instead of noce_conversion_profitable_p.
35177         * config/i386/i386.c: Include ifcvt.h.
35178         (ix86_option_override_internal): Don't override
35179         PARAM_MAX_RTL_IF_CONVERSION_INSNS default.
35180         (ix86_noce_conversion_profitable_p): New function.
35181         (TARGET_NOCE_CONVERSION_PROFITABLE_P): Redefine.
35182         * config/i386/x86-tune.def (X86_TUNE_ONE_IF_CONV_INSN): Adjust comment.
35183         * doc/tm.texi.in (TARGET_NOCE_CONVERSION_PROFITABLE_P): Add.
35184         * doc/tm.texi: Regenerated.
35186 2017-04-04  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35188         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Grammar
35189         correction.
35191 2017-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
35193         PR target/80307
35194         * config/arm/arm.c (thumb1_rtx_costs): Give a cost of 32
35195         instructions for small multiply cores.
35197 2017-04-04  Jeff Law  <law@redhat.com>
35199         * config/mips/mips.c (mips_multi_add): Zero initialize the newly
35200         added member.
35201         (mips_expand_vec_perm_const): Initialize elements in orig_perm
35202         that are not set by the loop over the elements.
35204 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
35206         PR target/80286
35207         * config/i386/i386.c (ix86_expand_args_builtin): If op has scalar
35208         int mode, convert_modes it to mode as unsigned, otherwise use
35209         lowpart_subreg to mode rather than SImode.
35210         * config/i386/sse.md (<mask_codefor>ashr<mode>3<mask_name>,
35211         ashr<mode>3, ashr<mode>3<mask_name>, <shift_insn><mode>3<mask_name>):
35212         Use DImode instead of SImode for the shift count operand.
35213         * config/i386/mmx.md (mmx_ashr<mode>3, mmx_<shift_insn><mode>3):
35214         Likewise.
35216 2017-04-04  Richard Biener  <rguenther@suse.de>
35218         PR middle-end/80281
35219         * match.pd (A + (-B) -> A - B): Make sure to preserve unsigned
35220         arithmetic done for the negate or the plus.  Simplify.
35221         (A - (-B) -> A + B): Likewise.
35222         * fold-const.c (split_tree): Make sure to not negate pointers.
35224 2017-04-04  Segher Boessenkool  <segher@kernel.crashing.org>
35226         PR rtl-optimization/60818
35227         * simplify-rtx.c (simplify_binary_operation_1): Do not replace
35228         a compare of comparisons with the thing compared if this results
35229         in a different machine mode.
35231 2017-04-03  Jonathan Wakely  <jwakely@redhat.com>
35233         * alias.c (base_alias_check): Fix typo in comment.
35234         * cgraph.h (class ipa_polymorphic_call_context): Likewise.
35235         * cgraphunit.c (symbol_table::compile): Likewise.
35236         * collect2.c (maybe_run_lto_and_relink): Likewise.
35237         * config/arm/arm.c (arm_thumb1_mi_thunk): Likewise.
35238         * config/avr/avr-arch.h (avr_arch_info_t): Likewise.
35239         * config/avr/avr.c (avr_map_op_t): Likewise.
35240         * config/cr16/cr16.h (DATA_ALIGNMENT): Likewise.
35241         * config/epiphany/epiphany.c (TARGET_ARG_PARTIAL_BYTES): Likewise.
35242         * config/epiphany/epiphany.md (movcc): Likewise.
35243         * config/i386/i386.c (legitimize_pe_coff_extern_decl): Likewise.
35244         * config/m68k/m68k.c (struct _sched_ib, m68k_sched_variable_issue):
35245         Likewise.
35246         * config/mips/mips.c (mips_save_restore_reg): Likewise.
35247         * config/rx/rx.c (rx_is_restricted_memory_address): Likewise.
35248         * config/s390/s390.c (Z10_EARLYLOAD_DISTANCE): Likewise.
35249         * config/sh/sh.c (sh_rtx_costs): Likewise.
35250         * fold-const.c (fold_truth_andor): Likewise.
35251         * genautomata.c (collapse_flag): Likewise.
35252         * gengtype.h (struct type::u::s): Likewise.
35253         * gensupport.c (has_subst_attribute, add_mnemonic_string): Likewise.
35254         * input.c (FORMAT_AMOUNT): Likewise.
35255         * ipa-cp.c (class ipcp_lattice, agg_replacements_to_vector)
35256         (known_aggs_to_agg_replacement_list): Likewise.
35257         * ipa-inline-analysis.c: Likewise.
35258         * ipa-inline.h (estimate_edge_time, estimate_edge_hints): Likewise.
35259         * ipa-polymorphic-call.c
35260         (ipa_polymorphic_call_context::restrict_to_inner_class): Likewise.
35261         * loop-unroll.c (analyze_insn_to_expand_var): Likewise.
35262         * lra.c (lra_optional_reload_pseudos, lra_subreg_reload_pseudos):
35263         Likewise.
35264         * modulo-sched.c (apply_reg_moves): Likewise.
35265         * omp-expand.c (build_omp_regions_1): Likewise.
35266         * trans-mem.c (struct tm_wrapper_hasher): Likewise.
35267         * tree-ssa-loop-ivopts.c (may_eliminate_iv): Likewise.
35268         * tree-ssa-loop-niter.c (maybe_lower_iteration_bound): Likewise.
35269         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
35270         * value-prof.c: Likewise.
35271         * var-tracking.c (val_reset): Likewise.
35273 2017-04-03  Richard Biener  <rguenther@suse.de>
35275         PR tree-optimization/80275
35276         * fold-const.c (split_address_to_core_and_offset): Handle
35277         POINTER_PLUS_EXPR.
35279 2017-04-03  Eric Botcazou  <ebotcazou@adacore.com>
35281         * tree-nested.c (get_descriptor_type): Make sure that the alignment of
35282         descriptors is at least equal to that of functions.
35284 2017-04-02  Uros Bizjak  <ubizjak@gmail.com>
35286         * config/i386/sse.md (movdi_to_sse): Add missing DONE.
35288 2017-04-02  Uros Bizjak  <ubizjak@gmail.com>
35290         PR target/80250
35291         * config/i386/sse.md (mov<IMOD4:mode>): Remove insn pattern.
35292         (mov<IMOD4:mode>): New expander.
35293         (*mov<IMOD4:mode>_internal): New insn and split pattern.
35295 2017-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
35297         PR rtl-optimization/79405
35298         * fwprop.c (propagations_left): New variable.
35299         (forward_propagate_into): Decrement it.
35300         (fwprop_init): Initialize it.
35301         (fw_prop): If the variable has reached zero, stop propagating.
35302         (fwprop_addr): Ditto.
35304 2017-03-31  Jakub Jelinek  <jakub@redhat.com>
35306         PR debug/79255
35307         * dwarf2out.c (decls_for_scope): If BLOCK_NONLOCALIZED_VAR is
35308         a FUNCTION_DECL, pass it as decl instead of origin to
35309         process_scope_var.
35311 2017-03-31  Alexander Monakov  <amonakov@ispras.ru>
35313         * config/nvptx/nvptx.c (nvptx_output_softstack_switch): Correct format
35314         string.
35316 2017-03-31  Pat Haugen  <pthaugen@us.ibm.com>
35318         PR target/80107
35319         * config/rs6000/rs6000.md (extendhi<mode>2): Add test for
35320         TARGET_VSX_SMALL_INTEGER.
35322 2017-03-31  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35324         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
35325         reference to the OpenPOWER 64-Bit ELF V2 ABI Specification.
35327 2017-03-31  Matthew Fortune  <matthew.fortune@imgtec.com>
35329         * config/mips/mips-msa.md (msa_vec_extract_<msafmt_f>): Update
35330         extraction from odd-numbered MSA register.
35332 2017-03-31  Jakub Jelinek  <jakub@redhat.com>
35334         PR middle-end/80173
35335         * expmed.c (store_bit_field_1): Don't attempt to create
35336         a word subreg out of hard registers wider than word if they
35337         have HARD_REGNO_NREGS of 1 for their mode.
35339         PR middle-end/80163
35340         * varasm.c (initializer_constant_valid_p_1): Disallow sign-extending
35341         conversions to integer types wider than word and pointer.
35343         PR debug/80025
35344         * cselib.h (rtx_equal_for_cselib_1): Add depth argument.
35345         (rtx_equal_for_cselib_p): Pass 0 to it.
35346         * cselib.c (cselib_hasher::equal): Likewise.
35347         (rtx_equal_for_cselib_1): Add depth argument.  If depth
35348         is 128, don't look up VALUE locs and punt.  Increment
35349         depth in recursive calls when walking VALUE locs.
35351 2017-03-31  Bernd Edlinger  <bernd.edlinger@hotmail.de>
35353         * gcov.c (md5sum_to_hex): Fix output of MD5 hex bytes.
35354         (make_gcov_file_name): Use the canonical path name for generating
35355         the MD5 value.
35356         (read_line): Fix handling of files with ascii null bytes.
35358 2017-03-30  Matthew Fortune  <matthew.fortune@imgtec.com>
35360         * config/mips/mips.c (mips_expand_vector_init): Create a const_vector
35361         to initialise a vector register instead
35362         of using a const_int.
35364 2017-03-30  Jakub Jelinek  <jakub@redhat.com>
35366         PR translation/80189
35367         * gimplify.c (omp_default_clause): Use %qs instead of %s in
35368         diagnostic messages.
35370 2017-03-30  Peter Bergner  <bergner@vnet.ibm.com>
35372         PR target/80246
35373         * config/rs6000/dfp.md (dfp_dxex_<mode>): Update mode of operand 0.
35374         (dfp_diex_<mode>): Update mode of operand 1.
35375         * doc/extend.texi (dxex, dxexq): Document change to return type.
35376         (diex, diexq): Document change to argument type.
35378 2017-03-30  Martin Jambor  <mjambor@suse.cz>
35380         PR ipa/77333
35381         * cgraph.h (cgraph_build_function_type_skip_args): Declare.
35382         * cgraph.c (redirect_call_stmt_to_callee): Set gimple fntype so that
35383         it reflects the signature changes performed at the callee side.
35384         * cgraphclones.c (build_function_type_skip_args): Make public, renamed
35385         to cgraph_build_function_type_skip_args.
35386         (build_function_decl_skip_args): Adjust call to the above function.
35388 2017-03-30  Jakub Jelinek  <jakub@redhat.com>
35390         PR target/80206
35391         * config/i386/sse.md
35392         (<extract_type>_vextract<shuffletype><extract_suf>_mask): Use
35393         register as dest whenever it is a MEM not rtx_equal_p to the
35394         corresponding dup operand, and when forcing into reg move the
35395         reg into the memory afterwards.
35396         (<extract_type_2>_vextract<shuffletype><extract_suf_2>_mask):
35397         Likewise.  Use <ssehalfvecmode> instead of <ssequartermode>
35398         for the force_reg mode.
35399         (avx512vl_vextractf128<mode>): Use register as dest either
35400         always when a MEM, or when it is a MEM not rtx_equal_p to the
35401         corresponding dup operand, or even not when it is a CONST_VECTOR
35402         depending on the mode and lo vs. hi.
35403         (avx512dq_vextract<shuffletype>64x2_1_maskm): Remove extraneous
35404         parens.
35405         (avx512f_vextract<shuffletype>32x4_1_maskm): Likewise.
35406         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name>):
35407         Likewise.  Require that operands[2] is even.
35408         (<mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name>):
35409         Remove extraneous parens.  Require that operands[2] is a multiple
35410         of 4.
35411         (vec_extract_lo_<mode><mask_name>): Don't bother testing if
35412         operands[0] is a MEM if <mask_applied>, the predicates/constraints
35413         disallow memory then.
35415 2017-03-30  Richard Biener  <rguenther@suse.de>
35417         PR tree-optimization/77498
35418         * tree-ssa-pre.c (phi_translate_1): Do not allow simplifications
35419         to non-constants over backedges.
35421 2017-03-29  Segher Boessenkool  <segher@kernel.crashing.org>
35423         PR rtl-optimization/80233
35424         * combine.c (combine_instructions): Only take NONDEBUG_INSN_P insns
35425         as last_combined_insn.  Do not test for BARRIER_P separately.
35427 2017-03-29  Andreas Schwab  <schwab@suse.de>
35429         PR ada/80146
35430         * calls.c (prepare_call_address): Convert funexp to Pmode before
35431         copying to temp reg.
35433 2017-03-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35435         PR tree-optimization/80158
35436         * gimple-ssa-strength-reduction.c (replace_mult_candidate):
35437         Handle possible future case of more than one alternate
35438         interpretation.
35439         (replace_rhs_if_not_dup): Likewise.
35440         (replace_one_candidate): Likewise.
35442 2017-03-28  Vladimir Makarov  <vmakarov@redhat.com>
35444         PR rtl-optimization/80193
35445         * ira.c (ira): Do not check allocation for LRA.
35447 2017-03-28  Alexander Monakov  <amonakov@ispras.ru>
35449         * config/nvptx/nvptx-protos.h (nvptx_output_simt_enter): Declare.
35450         (nvptx_output_simt_exit): Declare.
35451         * config/nvptx/nvptx.c (nvptx_init_unisimt_predicate): Use
35452         cfun->machine->unisimt_location.  Handle NULL unisimt_predicate.
35453         (init_softstack_frame): Move initialization of crtl->is_leaf to...
35454         (nvptx_declare_function_name): ...here.  Emit declaration of local
35455         memory space buffer for omp_simt_enter insn.
35456         (nvptx_output_unisimt_switch): New.
35457         (nvptx_output_softstack_switch): New.
35458         (nvptx_output_simt_enter): New.
35459         (nvptx_output_simt_exit): New.
35460         * config/nvptx/nvptx.h (struct machine_function): New fields
35461         has_simtreg, unisimt_location, simt_stack_size, simt_stack_align.
35462         * config/nvptx/nvptx.md (UNSPECV_SIMT_ENTER): New unspec.
35463         (UNSPECV_SIMT_EXIT): Ditto.
35464         (omp_simt_enter_insn): New insn.
35465         (omp_simt_enter): New expansion.
35466         (omp_simt_exit): New insn.
35467         * config/nvptx/nvptx.opt (msoft-stack-reserve-local): New option.
35469         * internal-fn.c (expand_GOMP_SIMT_ENTER): New.
35470         (expand_GOMP_SIMT_ENTER_ALLOC): New.
35471         (expand_GOMP_SIMT_EXIT): New.
35472         * internal-fn.def (GOMP_SIMT_ENTER): New internal function.
35473         (GOMP_SIMT_ENTER_ALLOC): Ditto.
35474         (GOMP_SIMT_EXIT): Ditto.
35475         * target-insns.def (omp_simt_enter): New insn.
35476         (omp_simt_exit): Ditto.
35477         * omp-low.c (struct omplow_simd_context): New fields simt_eargs,
35478         simt_dlist.
35479         (lower_rec_simd_input_clauses): Implement SIMT privatization.
35480         (lower_rec_input_clauses): Likewise.
35481         (lower_lastprivate_clauses): Handle SIMT privatization.
35483         * omp-offload.c: Include langhooks.h, tree-nested.h, stor-layout.h.
35484         (ompdevlow_adjust_simt_enter): New.
35485         (find_simtpriv_var_op): New.
35486         (execute_omp_device_lower): Handle IFN_GOMP_SIMT_ENTER,
35487         IFN_GOMP_SIMT_ENTER_ALLOC, IFN_GOMP_SIMT_EXIT.
35489         * tree-inline.h (struct copy_body_data): New field dst_simt_vars.
35490         * tree-inline.c (expand_call_inline): Handle SIMT privatization.
35491         (copy_decl_for_dup_finish): Ditto.
35493         * tree-ssa.c (execute_update_addresses_taken): Handle GOMP_SIMT_ENTER.
35495 2017-03-28  Uros Bizjak  <ubizjak@gmail.com>
35497         PR target/53383
35498         * config/i386/i386.c (ix86_option_override_internal): Always
35499         allow -mpreferred-stack-boundary=3 for 64-bit targets.
35501 2017-03-28  Bin Cheng  <bin.cheng@arm.com>
35503         * tree-vect-loop.c (optimize_mask_stores): Add bb to the right loop.
35505 2017-03-28  Bin Cheng  <bin.cheng@arm.com>
35507         * tree-vect-loop-manip.c (slpeel_add_loop_guard): New param and
35508         mark new edge's irreducible flag accordign to it.
35509         (vect_do_peeling): Check loop preheader edge's irreducible flag
35510         and pass it to function slpeel_add_loop_guard.
35512 2017-03-28  Richard Sandiford  <richard.sandiford@arm.com>
35514         PR tree-optimization/80218
35515         * tree-call-cdce.c (shrink_wrap_one_built_in_call_with_conds):
35516         Update block frequencies and counts.
35518 2017-03-28  Richard Biener  <rguenther@suse.de>
35520         PR tree-optimization/78644
35521         * tree-ssa-ccp.c (evaluate_stmt): When we may not use the value
35522         of a simplification result we may not use it at all.
35524 2017-03-28  Richard Biener  <rguenther@suse.de>
35526         PR ipa/80205
35527         * tree-inline.c (copy_phis_for_bb): Do not create PHI node
35528         without arguments, generate default definition of a SSA name.
35530 2017-03-28  Richard Biener  <rguenther@suse.de>
35532         PR middle-end/80222
35533         * gimple-fold.c (gimple_fold_indirect_ref): Do not touch
35534         TYPE_REF_CAN_ALIAS_ALL references.
35535         * fold-const.c (fold_indirect_ref_1): Likewise.
35537 2017-03-28  Martin Liska  <mliska@suse.cz>
35539         PR ipa/80104
35540         * cgraphunit.c (cgraph_node::expand_thunk): Mark argument of a
35541         thunk call as DECL_GIMPLE_REG_P when vector or complex type.
35543 2017-03-28  Claudiu Zissulescu  <claziss@synopsys.com>
35544             Thomas Petazzoni  <thomas.petazzoni@free-electrons.com>
35546         * config/arc/arc.h (CPP_SPEC): Add subtarget_cpp_spec.
35547         (EXTRA_SPECS): Define.
35548         (SUBTARGET_EXTRA_SPECS): Likewise.
35549         (SUBTARGET_CPP_SPEC): Likewise.
35550         * config/arc/elf.h (EXTRA_SPECS): Renamed to
35551         SUBTARGET_EXTRA_SPECS.
35552         * config/arc/linux.h (SUBTARGET_CPP_SPEC): Define.
35554 2017-03-28  Claudiu Zissulescu  <claziss@synopsys.com>
35556         * config/arc/simdext.md (vst64_insn): Update pattern.
35557         (vld32wh_insn): Likewise.
35558         (vld32wl_insn): Likewise.
35559         (vld64_insn): Likewise.
35560         (vld32_insn): Likewise.
35562 2017-03-28  Marek Polacek  <polacek@redhat.com>
35564         PR sanitizer/80067
35565         * fold-const.c (fold_comparison): Use protected_set_expr_location
35566         instead of SET_EXPR_LOCATION.
35568 2017-03-28  Markus Trippelsdorf  <markus@trippelsdorf.de>
35570         * tree.c (add_expr): Avoid name lookup warning.
35572 2017-03-27  Jeff Law  <law@redhat.com>
35574         PR tree-optimization/80216
35575         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Fix typo in
35576         function name.  Limit recursion depth.
35577         (record_temporary_equivalences): Corresponding changes.
35579 2017-03-27  Jonathan Wakely  <jwakely@redhat.com>
35581         * doc/invoke.texi (-Wno-narrowing): Reorder so default behavior is
35582         covered first.
35584 2017-03-27  Jakub Jelinek  <jakub@redhat.com>
35586         PR target/80102
35587         * reg-notes.def (REG_CFA_NOTE): Define.  Use it for CFA related
35588         notes.
35589         * cfgcleanup.c (reg_note_cfa_p): New array.
35590         (insns_have_identical_cfa_notes): New function.
35591         (old_insns_match_p): Don't cross-jump in between /f
35592         and non-/f instructions.  If both i1 and i2 are frame related,
35593         verify all CFA notes, their order and content.
35595 2017-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
35597         PR target/78543
35598         * config/rs6000/rs6000.md (bswaphi2_extenddi): Combine bswap
35599         HImode and SImode with zero extend to DImode to one insn.
35600         (bswap<mode>2_extenddi): Likewise.
35601         (bswapsi2_extenddi): Likewise.
35602         (bswaphi2_extendsi): Likewise.
35603         (bswaphi2): Combine bswap HImode and SImode into one insn.
35604         Separate memory insns from swapping register.
35605         (bswapsi2): Likewise.
35606         (bswap<mode>2): Likewise.
35607         (bswaphi2_internal): Delete, no longer used.
35608         (bswapsi2_internal): Likewise.
35609         (bswap<mode>2_load): Split bswap HImode/SImode into separate load,
35610         store, and gpr<-gpr swap insns.
35611         (bswap<mode>2_store): Likewise.
35612         (bswaphi2_reg): Register only splitter, combine with the splitter.
35613         (bswaphi2 splitter): Likewise.
35614         (bswapsi2_reg): Likewise.
35615         (bswapsi2 splitter): Likewise.
35616         (bswapdi2): If we have the LDBRX and STDBRX instructions, split
35617         the insns into load, store, and register/register insns.
35618         (bswapdi2_ldbrx): Likewise.
35619         (bswapdi2_load): Likewise.
35620         (bswapdi2_store): Likewise.
35621         (bswapdi2_reg): Likewise.
35623 2017-03-27  Gunther Nikl  <gnikl@users.sourceforge.net>
35625         * system.h (HAVE_DESIGNATED_INITIALIZERS): Fix non C++ case.
35626         (HAVE_DESIGNATED_UNION_INITIALIZERS): Likewise.
35628 2017-03-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
35630         PR target/80103
35631         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Edit and
35632         add comments.
35633         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
35634         special handling for target option conflicts between dform
35635         options (-mpower9-dform, -mpower9-dform-vector,
35636         -mpower9-dform-scalar) and -mno-direct-move.
35638 2017-03-27  Richard Biener  <rguenther@suse.de>
35640         PR tree-optimization/80181
35641         * tree-ssa-ccp.c (likely_value): UNDEFINED ^ X is UNDEFINED.
35643 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35645         * config/arc/predicates.md (move_double_src_operand): Replace the
35646         call to move_double_src_operand with a call to address_operand.
35648 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35650         * config/arc/elf.h (ARGET_ARC_TP_REGNO_DEFAULT): Define.
35651         * config/arc/linux.h (ARGET_ARC_TP_REGNO_DEFAULT): Likewise.
35652         * config/arc/arc.opt (mtp-regno): Use ARGET_ARC_TP_REGNO_DEFAULT.
35654 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35656         * config/arc/predicates.md (long_immediate_loadstore_operand):
35657         Consider scaled addresses cases.
35659 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35661         * config/arc/arc.c (arc_epilogue_uses): BLINK should be also
35662         restored when in interrupt.
35663         * config/arc/arc.md (simple_return): ARCv2 rtie instruction
35664         doesn't have delay slot.
35666 2017-03-27  Richard Biener  <rguenther@suse.de>
35668         PR ipa/79776
35669         * tree-ssa-structalias.c (associate_varinfo_to_alias): Skip
35670         inlined thunk clones.
35672 2017-03-27  Jakub Jelinek  <jakub@redhat.com>
35674         PR sanitizer/80168
35675         * asan.c (instrument_derefs): Copy over last operand from
35676         original COMPONENT_REF to the new COMPONENT_REF with
35677         DECL_BIT_FIELD_REPRESENTATIVE.
35678         * ubsan.c (instrument_object_size): Likewise.
35680 2017-03-27  Richard Biener  <rguenther@suse.de>
35682         PR tree-optimization/80170
35683         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Make
35684         sure DR/SCEV didnt fold in constants we do not see when looking
35685         at the reference base alignment.
35687 2017-03-27  Richard Biener  <rguenther@suse.de>
35689         PR middle-end/80171
35690         * gimple-fold.c (fold_ctor_reference): Properly guard against
35691         NULL return value from canonicalize_constructor_val.
35693 2017-03-25  Uros Bizjak  <ubizjak@gmail.com>
35695         PR target/80180
35696         * config/i386/i386.c (ix86_expand_builtin)
35697         <IX86_BUILTIN_RDSEED{16,32,64}_STEP>: Do not expand arg0 between
35698         flags reg setting and flags reg using instructions.
35699         <IX86_BUILTIN_RDRAND{16,32,64}_STEP>: Ditto.  Use non-flags reg
35700         clobbering instructions to zero extend op2.
35702 2017-03-25  Gerald Pfeifer  <gerald@pfeifer.com>
35704         * doc/install.texi (Configuration) <--with-aix-soname>:
35705         Update link to AIX ld.
35707 2017-03-25  Bernd Schmidt  <bschmidt@redhat.com>
35709         PR rtl-optimization/80160
35710         PR rtl-optimization/80159
35711         * lra-assigns.c (must_not_spill_p): Tighten new test to also take
35712         reg_alternate_class into account.
35714 2017-03-24  Vladimir Makarov  <vmakarov@redhat.com>
35716         PR target/80148
35717         * lra-assigns.c (assign_by_spills): Add spilled non-reload pseudos
35718         to consider in curr_insn_transform.
35720 2017-03-24  Jakub Jelinek  <jakub@redhat.com>
35722         * genrecog.c (validate_pattern): Add VEC_SELECT validation.
35723         * genmodes.c (emit_min_insn_modes_c): Call emit_mode_nunits
35724         and emit_mode_inner.
35726 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35728         * config/s390/s390-builtins.def: Add VXE builtins.  Add a flags
35729         argument to the overloaded builtin variants.  Use the new flag to
35730         deprecate certain builtin variants.
35731         * config/s390/s390-builtin-types.def: Add new builtin types.
35732         * config/s390/s390-builtins.h: Support new flags field for
35733         overloaded builtins.
35734         * config/s390/s390-c.c (OB_DEF_VAR): New flags field.
35735         (s390_macro_to_expand): Enable vector float data type.
35736         (s390_cpu_cpp_builtins_internal): Indicate support of the new
35737         builtins by incrementing the __VEC__ version number.
35738         (s390_expand_overloaded_builtin): Support expansion of vec_xl and
35739         vec_xst.
35740         (s390_resolve_overloaded_builtin): Emit error messages depending
35741         on the builtin flags.
35742         * config/s390/s390.c (s390_expand_builtin): Support additional
35743         flags argument.  Change error message to match the messages
35744         emitted in s390-c.c.
35745         * config/s390/s390.md: New UNSPEC_* constants.
35746         (op_type): Add new instruction types.
35747         * config/s390/vecintrin.h: Add new builtins and test data class
35748         constants.
35749         * config/s390/vx-builtins.md (V_HW_32_64): Add V4SF.
35750         (V_HW_4, VEC_HW, VECF_HW): New mode iterators.
35751         (VEC_INEXACT, VEC_NOINEXACT): New constants.
35752         ("vec_splats<mode>", "vec_insert<mode>", "vec_promote<mode>")
35753         ("vec_insert_and_zero<mode>", "vec_mergeh<mode>")
35754         ("vec_mergel<mode>"): V_HW -> VEC_HW.
35756         ("vlrlrv16qi", "vstrlrv16qi", "vbpermv16qi", "vec_msumv2di")
35757         ("vmslg", "*vftci<mode>_cconly", "vftci<mode>_intcconly")
35758         ("*vftci<mode>", "vftci<mode>_intcc", "vec_double_s64")
35759         ("vec_double_u64", "vfmin<mode>", "vfmax<mode>"): New definition.
35761         ("and_av2df3", "and_cv2df3", "vec_andc_av2df3")
35762         ("vec_andc_cv2df3", "xor_av2df3", "xor_cv2df3", "vec_nor_av2df3")
35763         ("vec_nor_cv2df3", "ior_av2df3", "ior_cv2df3", "vec_nabs")
35764         ("*vftcidb", "*vftcidb_cconly", "vftcidb"): Remove definition.
35766         ("vec_all_<fpcmpcc:code>v2df", "vec_any_<fpcmpcc:code>v2df")
35767         ("vec_scatter_elementv4si_DI", "vec_cmp<fpcmp:code>v2df")
35768         ("vec_di_to_df_s64", "vec_di_to_df_u64", "vec_df_to_di_u64")
35769         ("vfidb", "*vldeb", "*vledb", "*vec_cmp<insn_cmp>v2df_cconly")
35770         ("vec_cmpeqv2df_cc", "vec_cmpeqv2df_cc", "vec_cmphv2df_cc")
35771         ("vec_cmphev2df_cc", "*vec_cmpeqv2df_cc")
35772         ("*vec_cmphv2df_cc", "*vec_cmphev2df_cc"): Enable new modes as ...
35774         ("vec_all_<fpcmpcc:code><mode>", "vec_any_<fpcmpcc:code><mode>")
35775         ("vec_scatter_element<V_HW_4:mode>_DI")
35776         ("vec_cmp<fpcmp:code><mode>", "vcdgb", "vcdlgb", "vclgdb")
35777         ("vec_fpint<mode>", "vflls")
35778         ("vflrd", "*vec_cmp<insn_cmp><mode>_cconly", "vec_cmpeq<mode>_cc")
35779         ("vec_cmpeq<mode>_cc", "vec_cmph<mode>_cc", "vec_cmphe<mode>_cc")
35780         ("*vec_cmpeq<mode>_cc", "*vec_cmph<mode>_cc")
35781         ("*vec_cmphe<mode>_cc"): ... these.
35783         ("vec_ctd_s64", "vec_ctsl", "vec_ctul", "vec_st2f"): Use rounding
35784         mode constant instead of magic value.
35786 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35788         * config/s390/s390.c (s390_expand_vec_compare): Support other
35789         vector floating point modes than just V2DF.
35790         (s390_expand_vcond): Likewise.
35791         (s390_hard_regno_mode_ok): Allow SFmode values in VRs.
35792         (s390_cannot_change_mode_class): Prevent mode changes between TF
35793         and V1TF in vector registers.
35794         * config/s390/s390.md (DF, SF): New mode attributes.
35795         ("*cmp<mode>_ccs", "add<mode>3", "sub<mode>3", "mul<mode>3")
35796         ("fma<mode>4", "fms<mode>4", "div<mode>3", "*neg<mode>2"): Add
35797         SFmode support for VRs.
35798         * config/s390/vector.md (V_HW, V_HW2, VT_HW, ti*, nonvec): Add new
35799         vector fp modes.
35800         (VFT, VF_HW): New mode iterators.
35801         (vw, sdx): New mode attributes.
35802         ("addv2df3", "subv2df3", "mulv2df3", "divv2df3", "sqrtv2df2")
35803         ("fmav2df4","fmsv2df4", "negv2df2", "absv2df2", "*negabsv2df2")
35804         ("smaxv2df3", "sminv2df3", "*vec_cmp<VFCMP_HW_OP:code>v2df_nocc")
35805         ("vec_cmpuneqv2df", "vec_cmpltgtv2df", "vec_orderedv2df")
35806         ("vec_unorderedv2df"): Adjust the v2df only patterns to support
35807         also the new vector floating point modes.  Renaming to ...
35809         ("add<mode>3", "sub<mode>3", "mul<mode>3", "div<mode>3")
35810         ("sqrt<mode>2", "fma<mode>4", "fms<mode>4", "neg<mode>2")
35811         ("abs<mode>2", "negabs<mode>2", "smax<mode>3")
35812         ("smin<mode>3", "*vec_cmp<VFCMP_HW_OP:code><mode>_nocc")
35813         ("vec_cmpuneq<mode>", "vec_cmpltgt<mode>", "vec_ordered<mode>")
35814         ("vec_unordered<mode>"): ... these.
35816         ("neg_fma<mode>4", "neg_fms<mode>4", "*smax<mode>3_vxe")
35817         ("*smin<mode>3_vxe", "*sminv2df3_vx", "*vec_extendv4sf")
35818         ("*vec_extendv2df"): New insn definitions.
35820 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35822         * config/s390/s390.md ("*adddi3_sign", "*subdi3_sign", "mulditi3")
35823         ("mulditi3_2", "*muldi3_sign"): New patterns.
35824         ("muldi3", "*muldi3", "mulsi3", "*mulsi3"): Add an expander and
35825         rename the pattern definition.
35827 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35829         * config/s390/s390.md ("indirect_jump"): Turn insn definition into
35830         expander.
35831         ("*indirect_jump", "*indirect2_jump"): New pattern definitions.
35833 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35835         * config/s390/s390.c (s390_expand_vec_init): Use vllezl
35836         instruction if possible.
35837         * config/s390/vector.md (vec_halfnumelts): New mode
35838         attribute.
35839         ("*vec_vllezlf<mode>"): New pattern.
35841 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35843         * config/s390/vector.md ("popcountv16qi2", "popcountv8hi2")
35844         ("popcountv4si2", "popcountv2di2"): Rename to ...
35845         ("popcount<mode>2", "popcountv8hi2_vx", "popcountv4si2_vx")
35846         ("popcountv2di2_vx"): ... these and add !TARGET_VXE to the
35847         condition.
35848         ("popcount<mode>2_vxe"): New pattern.
35850 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35852         * common/config/s390/s390-common.c (processor_flags_table): Add
35853         arch12.
35854         * config.gcc: Add arch12.
35855         * config/s390/driver-native.c (s390_host_detect_local_cpu):
35856         Default to arch12 for unknown CPU model numbers.
35857         * config/s390/s390-builtins.def: Add B_VXE builtin flag.
35858         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Adjust
35859         PROCESSOR_max sanity check.
35860         * config/s390/s390-opts.h (enum processor_type): Add
35861         PROCESSOR_ARCH12.
35862         * config/s390/s390.c (processor_table): Add arch12.
35863         (s390_expand_builtin): Add check for B_VXE flag.
35864         (s390_issue_rate): Add PROCESSOR_ARCH12.
35865         (s390_get_sched_attrmask): Likewise.
35866         (s390_get_unit_mask): Likewise.
35867         (s390_sched_score): Enable z13 scheduling for arch12.
35868         (s390_sched_reorder): Likewise.
35869         (s390_sched_variable_issue): Likewise.
35870         * config/s390/s390.h (enum processor_flags): Add PF_ARCH12 and
35871         PF_VXE.
35872         (s390_tune_attr): Use z13 scheduling also for arch12.
35873         (TARGET_CPU_ARCH12, TARGET_CPU_ARCH12_P, TARGET_CPU_VXE)
35874         (TARGET_CPU_VXE_P, TARGET_ARCH12, TARGET_ARCH12_P, TARGET_VXE)
35875         (TARGET_VXE_P): New macros.
35876         * config/s390/s390.md: Add arch12 to cpu attribute.  Add arch12
35877         and vxe to cpu_facility.  Add arch12 and vxe to enabled attribute.
35878         * config/s390/s390.opt: Add arch12 as processor_type.
35880 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35882         * config/s390/s390.md
35883         ("fixuns_truncdddi2", "fixuns_trunctddi2")
35884         ("fixuns_trunc<BFP:mode><GPR:mode>2"): Merge into ...
35885         ("fixuns_trunc<FP:mode><GPR:mode>2"): New expander.
35887         ("fixuns_trunc<BFP:mode><GPR:mode>2", "fixuns_trunc<mode>si2"):
35888         Rename expanders to ...
35890         ("fixuns_trunc<BFP:mode><GPR:mode>2_emu")
35891         ("fixuns_truncdddi2_emu"): ... these.
35893         ("fixuns_trunc<mode>si2_emu"): New expander.
35895         ("*fixuns_truncdfdi2_z13"): Rename to ...
35896         ("*fixuns_truncdfdi2_vx"): ... this.
35898 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35900         * config/s390/2964.md: Remove the single element vector compare
35901         instructions which are no longer used.
35902         * config/s390/s390.c (s390_select_ccmode): Remove handling of
35903         vector CCmodes.
35904         (s390_canonicalize_comparison): Remove handling of DFmode
35905         compares.
35906         (s390_expand_vec_compare_scalar): Remove function.
35907         (s390_emit_compare): Don't call s390_expand_vec_compare_scalar.
35908         * config/s390/s390.md ("*vec_cmp<insn_cmp>df_cconly"): Remove
35909         pattern.
35910         ("*cmp<mode>_ccs"): Add wfcdb instruction.
35912 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35914         * config/s390/s390.md ("mov<mode>_64dfp" DD_DF): Use vleig for loading a
35915         FP zero.
35916         ("*mov<mode>_64" DD_DF): Remove the vector instructions. These
35917         will anyway by matched by mov<mode>_64dfp.
35919 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35921         * config/s390/s390.md ("mov<mode>" SD_SF): Change vleg/vsteg to
35922         vlef/vstef.  Add missing operand to vleif.
35924 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35926         * config/s390/s390.c (s390_expand_vec_init): Enable vector load
35927         pair for all vector types with 64 bit elements.
35928         * config/s390/vx-builtins.md (V_HW_64): Move mode iterator to ...
35929         * config/s390/vector.md (V_HW_64): ... here.
35930         (V_128_NOSINGLE): New mode iterator.
35931         ("vec_init<V_HW:mode>"): Use V_128 as mode iterator.
35932         ("*vec_splat<mode>"): Use V_128_NOSINGLE mode iterator.
35933         ("*vec_tf_to_v1tf", "*vec_ti_to_v1ti"): New pattern definitions.
35934         ("*vec_load_pairv2di"): Change to ...
35935         ("*vec_load_pair<mode>"): ... this one.
35937 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35939         * config/s390/constraints.md: Add comments.
35940         (jKK): Reject element sizes > 8 bytes.
35941         * config/s390/s390.c (s390_split_ok_p): Enable splitting also for
35942         s_operands.
35943         * config/s390/s390.md: Add the s_operand checks formerly in
35944         s390_split_ok_p to various splitters where they are still
35945         required.
35946         * config/s390/vector.md ("mov<mode>" V_128): Add GPR alternatives
35947         for 128 bit vectors.  Plus two splitters.
35949 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35951         * config/s390/s390.md: Rename the cpu facilty vec to vx throughout
35952         the file.
35954 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35956         PR target/79893
35957         * config/s390/s390-c.c (s390_adjust_builtin_arglist): Issue an
35958         error if the boundary argument is not constant.
35960 2017-03-24  Jakub Jelinek  <jakub@redhat.com>
35962         PR rtl-optimization/80112
35963         * loop-doloop.c (doloop_condition_get): Don't check condition
35964         if cmp isn't SET with IF_THEN_ELSE src.
35966 2017-03-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35968         PR tree-optimization/80158
35969         * gimple-ssa-strength-reduction.c (replace_mult_candidate): When
35970         replacing a candidate statement, also replace it for the
35971         candidate's alternate interpretation.
35972         (replace_rhs_if_not_dup): Likewise.
35973         (replace_one_candidate): Likewise.
35975 2017-03-24  Richard Biener  <rguenther@suse.de>
35977         PR tree-optimization/80167
35978         * graphite-isl-ast-to-gimple.c
35979         (translate_isl_ast_to_gimple::is_valid_rename): Handle default-defs
35980         properly.
35981         (translate_isl_ast_to_gimple::get_rename): Likewise.
35983 2017-03-23  Kelvin Nilsen  <kelvin@gcc.gnu.org>
35985         * config/rs6000/rs6000.c (rs6000_option_override_internal): Change
35986         handling of certain combinations of target options, including the
35987         combinations -mpower8-vector vs. -mno-vsx, -mpower9-vector vs.
35988         -mno-power8-vector, and -mpower9_dform vs. -mno-power9-vector.
35990 2017-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
35992         PR target/71436
35993         * config/arm/arm.md (*load_multiple): Add reload_completed to
35994         matching condition.
35996 2017-03-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35997             Richard Biener  <rguenth@suse.de>
35999         PR tree-optimization/79908
36000         PR tree-optimization/80136
36001         * tree-stdarg.c (expand_ifn_va_arg_1): For a VA_ARG whose LHS has
36002         been cast away, gimplify_and_add suffices.
36004 2017-03-23  Markus Trippelsdorf  <markus@trippelsdorf.de>
36006         * tree-vrp.c (identify_jump_threads): Delete avail_exprs.
36008 2017-03-23  Richard Biener  <rguenther@suse.de>
36010         PR tree-optimization/80032
36011         * gimplify.c (gimple_push_cleanup): Forced unconditional
36012         cleanups still have to go to the conditional_cleanups
36013         sequence.
36015 2017-03-22  Jakub Jelinek  <jakub@redhat.com>
36017         PR tree-optimization/80072
36018         * tree-ssa-reassoc.c (struct operand_entry): Change id field type
36019         to unsigned int.
36020         (next_operand_entry_id): Change type to unsigned int.
36021         (sort_by_operand_rank): Make sure to return the right return value
36022         even if unsigned fields are bigger than INT_MAX.
36023         (struct oecount): Change cnt and id type to unsigned int.
36024         (oecount_hasher::equal): Formatting fix.
36025         (oecount_cmp): Make sure to return the right return value
36026         even if unsigned fields are bigger than INT_MAX.
36027         (undistribute_ops_list): Change next_oecount_id type to unsigned int.
36029         PR c++/80129
36030         * gimplify.c (gimplify_modify_expr_rhs) <case COND_EXPR>: Clear
36031         TREE_READONLY on result if writing it more than once.
36033         PR sanitizer/80110
36034         * doc/invoke.texi (-fsanitize=thread): Document that with
36035         -fnon-call-exceptions atomics are not able to throw
36036         exceptions.
36038         PR sanitizer/80110
36039         * tsan.c: Include tree-eh.h.
36040         (instrument_builtin_call): Call maybe_clean_eh_stmt or
36041         maybe_clean_or_replace_eh_stmt where needed.
36042         (instrument_memory_accesses): Add cfg_changed argument.
36043         Call gimple_purge_dead_eh_edges on each block and set *cfg_changed
36044         if it returned true.
36045         (tsan_pass): Adjust caller.  Return TODO_cleanup_cfg if cfg_changed.
36047         PR rtl-optimization/63191
36048         * config/i386/i386.c (ix86_delegitimize_address): Turn into small
36049         wrapper function, moved the whole old content into ...
36050         (ix86_delegitimize_address_1): ... this.  New inline function.
36051         (ix86_find_base_term): Use ix86_delegitimize_address_1 with
36052         true as last argument instead of ix86_delegitimize_address.
36054 2017-03-22  Wilco Dijkstra  <wdijkstr@arm.com>
36056         * config/aarch64/aarch64.c (generic_branch_cost): Copy
36057         cortexa57_branch_cost.
36059 2017-03-22  Wilco Dijkstra  <wdijkstr@arm.com>
36061         * config/aarch64/aarch64.c (generic_tunings): Add AES fusion.
36063 2017-03-21  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
36065         PR target/80123
36066         * doc/md.texi (Constraints): Document wA constraint.
36067         * config/rs6000/constraints.md (wA): New.
36068         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Add wA reg_class.
36069         (rs6000_init_hard_regno_mode_ok): Init wA constraint.
36070         * config/rs6000/rs6000.h (RS6000_CONSTRAINT_wA): New.
36071         * config/rs6000/vsx.md (vsx_splat_<mode>): Use wA constraint.
36073 2017-03-22  Cesar Philippidis  <cesar@codesourcery.com>
36075         PR c++/80029
36076         * gimplify.c (is_oacc_declared): New function.
36077         (oacc_default_clause): Use it to set default flags for acc declared
36078         variables inside parallel regions.
36079         (gimplify_scan_omp_clauses): Strip firstprivate pointers for acc
36080         declared variables.
36081         (gimplify_oacc_declare): Gimplify the declare clauses.  Add the
36082         declare attribute to any decl as necessary.
36084 2017-03-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
36086         PR target/80082
36087         * config/arm/arm-isa.h (isa_bit_lpae): New feature bit.
36088         (ISA_ARMv7ve): Add isa_bit_lpae to the definition.
36089         * config/arm/arm-protos.h (arm_arch7ve): Rename into ...
36090         (arm_arch_lpae): This.
36091         * config/arm/arm.c (arm_arch7ve): Rename into ...
36092         (arm_arch_lpae): This.  Define it in term of isa_bit_lpae.
36093         * config/arm/arm.h (TARGET_HAVE_LPAE): Redefine in term of
36094         arm_arch_lpae.
36096 2017-03-22  Martin Liska  <mliska@suse.cz>
36098         PR target/79906
36099         * config/rs6000/rs6000.c (rs6000_inner_target_options): Show
36100         error message instead of an ICE.
36102 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36104         * doc/extend.texi (6.11 Additional Floating Types): Revise.
36106 2017-03-21  Kelvin Nilsen  <kelvin@gcc.gnu.org>
36108         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Add
36109         comments.
36110         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
36111         comments.
36113 2017-03-21  Martin Sebor  <msebor@redhat.com>
36115         * doc/extend.texi: Use "cannot" instead of "can't."
36116         * doc/hostconfig.texi: Same.
36117         * doc/install.texi: Same.
36118         * doc/invoke.texi: Same.
36119         * doc/loop.texi: Same.
36120         * doc/md.texi: Same.
36121         * doc/objc.texi: Same.
36122         * doc/rtl.texi: Same.
36123         * doc/tm.texi: Same.
36124         * doc/tm.texi.in: Same.
36125         * doc/trouble.texi: Same.
36127 2017-03-21  Alexandre Oliva  <aoliva@redhat.com>
36129         PR debug/63238
36130         * dwarf2out.c (struct checksum_attributes): Add at_alignment.
36131         (collect_checksum_attributes): Set it.
36132         (die_checksum_ordered): Use it.
36134 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36136         PR tree-optimization/79908
36137         * tree-stdarg.c (expand_ifn_va_arg_1): Revert the following
36138         change: For a VA_ARG whose LHS has been cast away, use
36139         force_gimple_operand to construct the side effects.
36141 2017-03-21  David Malcolm  <dmalcolm@redhat.com>
36143         PR translation/80001
36144         * omp-offload.c (oacc_loop_fixed_partitions): Make diagnostics
36145         more amenable to translation.
36146         (oacc_loop_auto_partitions): Likewise.
36148 2017-03-21  Marek Polacek  <polacek@redhat.com>
36149             Martin Sebor  <msebor@redhat.com>
36151         PR tree-optimization/80109
36152         * gimple-ssa-warn-alloca.c (alloca_call_type): Only call get_range_info
36153         on INTEGRAL_TYPE_P.
36155 2017-03-21  Jakub Jelinek  <jakub@redhat.com>
36156             Segher Boessenkool  <segher@kernel.crashing.org>
36158         PR target/80125
36159         * combine.c (can_combine_p): Revert the 2017-03-20 change, only
36160         check reg_used_between_p between insn and one of succ or succ2
36161         depending on if succ is artificial insn not inserted into insn
36162         stream.
36164 2017-03-21  Martin Liska  <mliska@suse.cz>
36166         PR gcov-profile/80081
36167         * Makefile.in: Add gcov-dump and fix installation of gcov-tool.
36168         * doc/gcc.texi: Include gcov-dump stuff.
36169         * doc/gcov-dump.texi: New file.
36171 2017-03-21  Toma Tabacu  <toma.tabacu@imgtec.com>
36173         PR rtl-optimization/79150
36174         * config/mips/mips.c (mips_block_move_loop): Emit a NOP after the
36175         conditional jump, if the jump is the last insn of the loop.
36177 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36178             Richard Biener  <rguenth@suse.de>
36180         PR tree-optimization/79908
36181         * tree-stdarg.c (expand_ifn_va_arg_1): For a VA_ARG whose LHS has
36182         been cast away, use force_gimple_operand to construct the side
36183         effects.
36185 2017-03-21  Martin Liska  <mliska@suse.cz>
36187         PR libfortran/79956
36188         * simplify-rtx.c (simplify_immed_subreg): Initialize a variable
36189         to NULL.
36191 2017-03-21  Brad Spengler <spender@grsecurity.net>
36193         PR plugins/80094
36194         * plugin.c (htab_hash_plugin): New function.
36195         (add_new_plugin): Use it and adjust.
36196         (parse_plugin_arg_opt): Adjust.
36197         (init_one_plugin): Likewise.
36199 2017-03-21  Richard Biener  <rguenther@suse.de>
36201         PR tree-optimization/80032
36202         * gimplify.c (gimple_push_cleanup): Add force_uncond parameter,
36203         if set force the cleanup to happen unconditionally.
36204         (gimplify_target_expr): Push inserted clobbers with force_uncond
36205         to avoid them being removed by control-dependent DCE.
36207 2017-03-21  Richard Biener  <rguenther@suse.de>
36209         PR tree-optimization/80122
36210         * tree-inline.c (copy_bb): Do not expans va-arg packs or
36211         va_arg_pack_len when the inlined call stmt requires pack
36212         expansion itself.
36213         * tree-inline.h (struct copy_body_data): Make call_stmt a gcall *.
36215 2017-03-21  Jakub Jelinek  <jakub@redhat.com>
36217         PR sanitizer/78158
36218         * tsan.c (instrument_builtin_call): If the memory model argument
36219         is not a constant, assume it is valid.
36221         PR c/67338
36222         * fold-const.c (round_up_loc): Negate divisor in unsigned type to
36223         avoid UB.
36225 2017-03-20  Segher Boessenkool  <segher@kernel.crashing.org>
36227         PR rtl-optimization/79910
36228         * combine.c (can_combine_p): Do not allow combining an I0 or I1
36229         if its dest is used by an insn before I2 (other than the combined
36230         insns themselves, which are properly handled already).
36232 2017-03-20  Segher Boessenkool  <segher@kernel.crashing.org>
36234         Revert:
36235         2017-03-17  Bernd Schmidt  <bschmidt@redhat.com>
36237         * combine.c (record_used_regs): New static function.
36238         (try_combine): Handle situations where there is an additional
36239         instruction between I2 and I3 which needs to have a LOG_LINK
36240         updated.
36242         Revert:
36243         2017-03-17  Jim Wilson  <jim.wilson@linaro.org>
36245         * combine.c (try_combine): Delete redundant i1 test.  Call
36246         prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
36248 2017-03-20  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
36250         PR target/80083
36251         * config/rs6000/rs6000.md (*movsi_internal1): Fix constraints for
36252         alternatives 13/14.
36254 2017-03-20  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36256         PR tree-optimization/80054
36257         * gimple-ssa-strength-reduction.c (all_phi_incrs_profitable): Fail
36258         the optimization if a PHI or any of its arguments is not dominated
36259         by the candidate's basis.  Use gphi* rather than gimple* as
36260         appropriate.
36261         (replace_profitable_candidates): Clean up a gimple* variable that
36262         should be a gphi* variable.
36264 2017-03-20  Martin Sebor  <msebor@redhat.com>
36266         PR c++/52477
36267         * doc/extend.texi (attribute constructor): Document present limitation.
36269 2017-03-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
36271         PR target/79963
36272         * config/rs6000/altivec.h (vec_all_ne): Under __cplusplus__ and
36273         __POWER9_VECTOR__ #ifdef control, change template definition to
36274         use Power9-specific built-in function.
36275         (vec_any_eq): Likewise.
36276         * config/rs6000/vector.md (vector_ae_v2di_p): Change the flag used
36277         to control outcomes from this test.
36278         (vector_ae_<mode>p): For VEC_F modes, likewise.
36280 2017-03-20  Ian Lance Taylor  <iant@google.com>
36282         * config/i386/i386.c (ix86_function_regparm): Save an extra
36283         register for -fsplit-stack with DECL_STATIC_CHAIN.
36285 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com>
36287         PR target/79912
36288         * config/riscv/riscv.c (riscv_preferred_reload_class): Remove.
36289         (TARGET_PREFERRED_RELOAD_CLASS): Likewise.
36291 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com>
36293         * config/riscv/riscv.c (riscv_print_operand): Use "fence
36294         iorw,ow".
36295         * config/riscv/sync.mc (mem_thread_fence_1): Use "fence
36296         iorw,iorw".
36298 2017-03-20  Marek Polacek  <polacek@redhat.com>
36300         PR sanitizer/80063
36301         * asan.c (DEF_SANITIZER_BUILTIN): Use do { } while (0).
36303 2017-03-20  Richard Biener  <rguenther@suse.de>
36305         PR tree-optimization/80113
36306         * graphite-isl-ast-to-gimple.c (copy_loop_phi_nodes): Do not
36307         allocate extra SSA name for PHI def.
36308         (add_close_phis_to_outer_loops): Likewise.
36309         (add_close_phis_to_merge_points): Likewise.
36310         (copy_loop_close_phi_args): Likewise.
36311         (copy_cond_phi_nodes): Likewise.
36313 2017-03-20  Martin Liska  <mliska@suse.cz>
36315         PR middle-end/79753
36316         * tree-chkp.c (chkp_build_returned_bound): Do not build
36317         returned bounds for a LHS that's not a BOUNDED_P type.
36319 2017-03-20  Martin Liska  <mliska@suse.cz>
36321         PR target/79769
36322         PR target/79770
36323         * tree-chkp.c (chkp_find_bounds_1): Handle REAL_CST,
36324         COMPLEX_CST and VECTOR_CST.
36326 2017-03-20  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
36328         PR target/78857
36329         * config/s390/s390.md ("cmp<mode>_ccs_0"): Add a clobber of the
36330         target operand.  A new splitter adds the clobber statement in case
36331         the target operand is dead anyway.
36333 2017-03-19  Gerald Pfeifer  <gerald@pfeifer.com>
36335         * doc/install.texi (Specific) <sparc-*-linux*>: No longer refer
36336         to age-old versions of binutils and glibc.
36338 2017-03-18  Segher Boessenkool  <segher@kernel.crashing.org>
36340         * doc/contrib.texi (Contributors): Remove duplicate entry for myself.
36342 2017-03-18  Gerald Pfeifer  <gerald@pfeifer.com>
36344         * doc/contrib.texi (Contributors): Add Segher Boessenkool.
36346 2017-03-18  Gerald Pfeifer  <gerald@pfeifer.com>
36348         * doc/install.texi (Specific) <arm-*-eabi>: Remove old
36349         requirement for binutils 2.13.
36351 2017-03-17  Jim Wilson  <jim.wilson@linaro.org>
36353         * combine.c (try_combine): Delete redundant i1 test.  Call
36354         prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
36356 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com
36358         * doc/install.texi (Specific) <riscv32-*-elf>: Add riscv32-*-elf,
36359         riscv32-*-linux, riscv64-*-elf, riscv64-*-linux to the table of
36360         contents.
36361         <riscv64-*-elf>: Re-arrange section
36362         <riscv32-*-elf>: Add a note about requiring binutils 2.28.
36363         <riscv32-*-linux>: Likewise.
36364         <riscv64-*-elf>: Likewise
36365         <riscv64-*-linux>: Likewise.
36367 2017-03-17  Richard Earnshaw  <rearnsha@arm.com>
36369         PR target/80052
36370         * aarch64.opt(verbose-cost-dump): Fix typo.
36372 2017-03-17  Pat Haugen  <pthaugen@us.ibm.com>
36374         PR target/79951
36375         * config/rs6000/rs6000.md (copysign<mode>3_fcpsgn): Test
36376         for VECTOR_UNIT_VSX_P (<MODE>mode) too.
36378 2017-03-17  Bernd Schmidt  <bschmidt@redhat.com>
36380         * reload.c (find_reloads): When reloading a nonoffsettable address,
36381         use RELOAD_OTHER for it and its address reloads.
36383         PR rtl-optimization/79910
36384         * combine.c (record_used_regs): New static function.
36385         (try_combine): Handle situations where there is an additional
36386         instruction between I2 and I3 which needs to have a LOG_LINK
36387         updated.
36389 2017-03-17  Jeff Law  <law@redhat.com>
36391         PR tree-optimization/71437
36392         * tree-vrp.c (simplify_stmt_for_jump_threading): Lookup the
36393         conditional in the hash table first.
36394         (vrp_dom_walker::before_dom_children): Extract condition from
36395         ASSERT_EXPR.  Record condition, its inverion and any implied
36396         conditions as well.
36398 2017-03-17  Marek Polacek  <polacek@redhat.com>
36399             Markus Trippelsdorf  <markus@trippelsdorf.de>
36401         PR tree-optimization/80079
36402         * gimple-ssa-store-merging.c (class pass_store_merging): Initialize
36403         m_stores_head.
36405 2017-03-17  Richard Biener  <rguenther@suse.de>
36407         PR middle-end/80075
36408         * tree-eh.c (stmt_could_throw_1_p): Only handle gimple assigns.
36409         Properly verify the LHS before the RHS possibly claims to be
36410         handled.
36411         (stmt_could_throw_p): Hande gimple conds fully here.  Clobbers
36412         do not throw.
36414 2017-03-17  Martin Jambor  <mjambor@suse.cz>
36416         * doc/invoke.texi (Option Options): Include -fipa-vrp in the list.
36417         (List of -O2 options): Likewise.
36418         (-fipa-bit-cp): Replace "ipa" with "interprocedural."
36419         (-fipa-vrp) New.
36421 2017-03-17  Tom de Vries  <tom@codesourcery.com>
36423         * gcov-dump.c (print_usage): Print bug_report_url.
36425 2017-03-17  Richard Biener  <rguenther@suse.de>
36427         PR middle-end/80050
36428         * genmatch.c (parser::next): Remove pointless check for CPP_EOF.
36429         (parser::peek): Likewise.
36431 2017-03-17  Richard Biener  <rguenther@suse.de>
36433         PR tree-optimization/80048
36434         * sese.c (free_sese_info): Properly release rename_map and
36435         copied_bb_map elements.
36437 2017-03-16  Alexandre Oliva  <aoliva@redhat.com>
36439         * gimple-ssa-store-merging.c (struct imm_store_chain_info):
36440         Add linked-list forward and backlinks.  Insert on
36441         construction, remove on destruction.
36442         (class pass_store_merging): Add m_stores_head field.
36443         (pass_store_merging::terminate_and_process_all_chains):
36444         Iterate over m_stores_head list.
36445         (pass_store_merging::terminate_all_aliasing_chains):
36446         Likewise.
36447         (pass_store_merging::execute): Check for debug stmts first.
36448         Push new chains onto the m_stores_head stack.
36450 2017-03-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
36452         PR target/71294
36453         * config/rs6000/vsx.md (vsx_splat_<mode>, VSX_D iterator): Allow a
36454         SPLAT operation on ISA 2.07 64-bit systems that have direct move,
36455         but no MTVSRDD support, by doing MTVSRD and XXPERMDI.
36457 2017-03-16  Jeff Law  <law@redhat.com>
36459         PR tree-optimization/71437
36460         * tree-ssa-dom.c (dom_opt_dom_walker): Remove thread_across_edge
36461         member function.  Implementation moved into after_dom_children
36462         member function and into the threader's thread_outgoing_edges
36463         function.
36464         (dom_opt_dom_walker::after_dom_children): Simplify by moving
36465         some code into new thread_outgoing_edges.
36466         * tree-ssa-threadedge.c (thread_across_edge): Make static and simplify
36467         definition.  Simplify marker handling (do it here).   Assume we always
36468         have the available expression and the const/copies tables.
36469         (thread_outgoing_edges): New function extracted from tree-ssa-dom.c
36470         and tree-vrp.c
36471         * tree-ssa-threadedge.h (thread_outgoing_edges): Declare.
36472         * tree-vrp.c (equiv_stack): No longer file scoped.
36473         (vrp_dom_walker): New class.
36474         (vrp_dom_walker::before_dom_children): New member function.
36475         (vrp_dom_walker::after_dom_children): Likewise.
36476         (identify_jump_threads):  Setup domwalker.  Use it rather than
36477         walking edges in a random order by hand.  Simplify setup/finalization.
36478         (finalize_jump_threads): Remove.
36479         (vrp_finalize): Do not call identify_jump_threads here.
36480         (execute_vrp): Do it here instead and call thread_through_all_blocks
36481         here too.
36483         PR tree-optimization/71437
36484         * tree-ssa-dom.c (pfn_simplify): Add basic_block argument.  All
36485         callers changed.
36486         (simplify_stmt_for_jump_threading): Add basic_block argument.  All
36487         callers changed.
36488         (lhs_of_dominating_assert): Moved from here into tree-vrp.c.
36489         (dom_opt_dom_walker::thread_across_edge): Remove
36490         handle_dominating_asserts argument.  All callers changed.
36491         (record_temporary_equivalences_from_stmts_at_dest): Corresponding
36492         changes.  Remove calls to lhs_of_dominating_assert.  Other
36493         uses of handle_dominating_asserts turn into unconditional code
36494         (simplify_control_stmt_condition_1): Likewise.
36495         (simplify_control_stmt_condition): Likewise.
36496         (thread_through_normal_block, thread_across_edge): Likewise.
36497         * tree-ssa-threadedge.h (thread_across_edge): Corresponding changes.
36498         * tree-vrp.c (lhs_of_dominating_assert): Move here.  Return original
36499         object if it is not an SSA_NAME.
36500         (simplify_stmt_for_jump_threading): Call lhs_of_dominating_assert
36501         before calling into the VRP specific simplifiers.
36502         (identify_jump_threads): Remove handle_dominating_asserts
36503         argument.
36505 2017-03-16  Jakub Jelinek  <jakub@redhat.com>
36507         PR fortran/79886
36508         * tree-diagnostic.c (default_tree_printer): No longer static.
36509         * tree-diagnostic.h (default_tree_printer): New prototype.
36511 2017-03-16  Tamar Christina  <tamar.christina@arm.com>
36513         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>)
36514         Change ins into fmov.
36516 2017-03-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36518         * config/aarch64/iterators.md (h_con): Return "x" for V4HF and V8HF.
36519         * config/aarch64/aarch64-simd.md (*aarch64_fma4_elt_from_dup<mode>):
36520         Use h_con constraint for operand 1.
36521         (*aarch64_fnma4_elt_from_dup<mode>): Likewise.
36522         (*aarch64_mulx_elt_from_dup<mode>): Likewise for operand 2.
36524 2017-03-15  Jeff Law  <law@redhat.com>
36526         PR tree-optimization/71437
36527         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): New function.
36528         (record_temporary_equivalences): Use it.
36530         PR tree-optimization/71437
36531         * tree-ssa-dom.c (struct cond_equivalence): Moved from here into
36532         tree-ssa-scopedtables.
36533         (lookup_avail_expr, build_and_record_new_cond): Likewise.
36534         (record_conditions, record_cond, vuse_eq): Likewise.
36535         (record_edge_info): Adjust to API tweak of record_conditions.
36536         (simplify_stmt_for_jump_threading): Similarly for lookup_avail_expr.
36537         (record_temporary_equivalences, optimize_stmt): Likewise.
36538         (eliminate_redundant_computations): Likewise.
36539         (record_equivalences_from_stmt): Likewise.
36540         * tree-ssa-scopedtables.c: Include options.h and params.h.
36541         (vuse_eq): New function, moved from tree-ssa-dom.c
36542         (build_and_record_new_cond): Likewise.
36543         (record_conditions): Likewise.  Accept vector of conditions rather
36544         than edge_equivalence structure for first argument.
36545         for the first argument.
36546         (avail_exprs_stack::lookup_avail_expr): New member function, moved
36547         from tree-ssa-dom.c.
36548         (avail_exprs_stack::record_cond): Likewise.
36549         * tree-ssa-scopedtables.h (struct cond_equivalence): Moved here
36550         from tree-ssa-dom.c.
36551         (avail_exprs_stack): Add new member functions lookup_avail_expr
36552         and record_cond.
36553         (record_conditions): Declare.
36555 2017-03-15  Vladimir Makarov  <vmakarov@redhat.com>
36557         PR target/80017
36558         * lra-constraints.c (process_alt_operands): Increase reject for
36559         reloading an input/output operand.
36561 2017-03-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
36563         PR target/79038
36564         * config/rs6000/rs6000.md (float<QHI:mode><IEEE128:mode>2): Define
36565         insns to convert from signed/unsigned char/short to IEEE 128-bit
36566         floating point.
36567         (floatuns<QHI:mode><IEEE128:mode>2): Likewise.
36569 2017-03-15  Uros Bizjak  <ubizjak@gmail.com>
36571         PR target/80019
36572         * config/i386/i386.c (ix86_vector_duplicate_value): Create
36573         subreg of inner mode for values already in registers.
36575 2017-03-15  Bernd Schmidt  <bschmidt@redhat.com>
36577         * config/c6x/c6x.c (hwloop_optimize): Handle case where the old
36578         iteration reg is used after the loop.
36580 2017-03-14  Martin Sebor  <msebor@redhat.com>
36582         PR tree-optimization/79800
36583         * gimple-ssa-sprintf.c (format_floating: Add argument.  Handle
36584         precision in negative-positive range.
36585         (format_floating): Call non-const overload with adjusted precision.
36587 2017-03-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
36589         PR target/79947
36590         * config/rs6000/rs6000.h (TARGET_FRSQRTES): Add check for
36591         -mpowerpc-gfxopt.
36593 2017-03-14  Martin Sebor  <msebor@redhat.com>
36595         PR middle-end/80020
36596         * builtin-attrs.def (ATTR_ALLOC_SIZE_2_NOTHROW_LIST): New macro.
36597         * builtins.def (aligned_alloc): Use it.
36599         PR c/79936
36600         * Makefile.in (GTFILES): Add calls.c.
36601         * calls.c: Include "gt-calls.h".
36603 2017-03-14  Bernd Schmidt  <bschmidt@redhat.com>
36605         PR rtl-optimization/79728
36606         * regs.h (struct target_regs): New field
36607         x_contains_allocatable_regs_of_mode.
36608         (contains_allocatable_regs_of_mode): New macro.
36609         * reginfo.c (init_reg_sets_1): Initialize it, and change
36610         contains_reg_of_mode so it includes global regs as well.
36611         * reload.c (push_reload): Use contains_allocatable_regs_of_mode
36612         rather than contains_regs_of_mode.
36614 2017-03-14  Martin Liska  <mliska@suse.cz>
36616         * doc/invoke.texi: Document options that can't be combined with
36617         -fcheck-pointer-bounds.
36619 2017-03-14  Martin Liska  <mliska@suse.cz>
36621         PR middle-end/79831
36622         * doc/invoke.texi (-Wchkp): Document the option.
36624 2017-03-14  Martin Liska  <mliska@suse.cz>
36626         * Makefile.in: Install gcov-dump.
36628 2017-03-14  Martin Liska  <mliska@suse.cz>
36630         * multiple_target.c (expand_target_clones): Bail out for
36631         an invalid attribute.
36633 2017-03-14  Richard Biener  <rguenther@suse.de>
36635         * alias.c (struct alias_set_entry): Pack properly.
36636         * cfgloop.h (struct loop): Likewise.
36637         * cse.c (struct set): Likewise.
36638         * ipa-utils.c (struct searchc_env): Likewise.
36639         * loop-invariant.c (struct invariant): Likewise.
36640         * lra-remat.c (struct cand): Likewise.
36641         * recog.c (struct change_t): Likewise.
36642         * rtl.h (struct address_info): Likewise.
36643         * symbol-summary.h (function_summary): Likewise.
36644         * tree-loop-distribution.c (struct partition): Likewise.
36645         * tree-object-size.c (struct object_size_info): Likewise.
36646         * tree-ssa-loop-ivopts.c (struct cost_pair): Likewise.
36647         * tree-ssa-threadupdate.c (struct ssa_local_info_t): Likewise.
36648         * tree-vect-data-refs.c (struct _vect_peel_info): Likewise.
36649         * tree-vect-slp.c (struct _slp_oprnd_info): Likewise.
36650         * tree-vect-stmts.c (struct simd_call_arg_info): Likewise.
36651         * tree-vectorizer.h (struct _loop_vec_info): Likewise.
36652         (struct _stmt_vec_info): Likewise.
36654 2017-03-14  Martin Liska  <mliska@suse.cz>
36656         PR target/79892
36657         * multiple_target.c (create_dispatcher_calls): Check that
36658         a target can create a function dispatcher.
36660 2017-03-14  Martin Liska  <mliska@suse.cz>
36662         PR lto/66295
36663         * multiple_target.c (expand_target_clones): Drop local.local
36664         flag for default implementation.
36666 2017-03-14  Richard Biener  <rguenther@suse.de>
36668         PR tree-optimization/80030
36669         * tree-vect-stmts.c (vectorizable_store): Plug memleak.
36671 2017-03-13  Kito Cheng  <kito.cheng@gmail.com>
36673         * config/riscv/riscv.c (riscv_emit_float_compare>: Use
36674         gcc_fallthrough() instead of __attribute__((fallthrough));
36676 2017-03-13  Gerald Pfeifer  <gerald@pfeifer.com>
36678         * doc/gcc.texi: Remove "up" link to (DIR).
36679         * doc/gccint.texi: Ditto.
36681 2017-03-13  Gerald Pfeifer  <gerald@pfeifer.com>
36683         * doc/install.texi (Specific) <avr>: Remove reference to
36684         binutils 2.13.
36686 2017-03-13  Jeff Law  <law@redhat.com>
36688         * config/riscv/riscv.c (riscv_emit_float_compare): Use fallthru
36689         attribute rather than comments.
36691         * config/pdp11/pdp11.md (movmemhi): Adjust operand numbers to
36692         match_scratch operand is highest.
36694 2017-03-13  Martin Liska  <mliska@suse.cz>
36696         PR middle-end/78339
36697         * ipa-pure-const.c (warn_function_noreturn): If the declarations
36698         is a CHKP clone, use original declaration.
36700 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36702         * config/arc/arc.c (arc_init): Use multiplier whenever we have it.
36703         (arc_conditional_register_usage): Use a different allocation order
36704         when optimizing for size.
36705         * common/config/arc/arc-common.c (arc_option_optimization_table):
36706         Section anchors default on when optimizing for size.
36708 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36710         * config/arc/arc.md (*tst_bitfield_tst): Fix pattern.
36712 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36714         * config/arc/arc.c (arc_output_addsi): Emit code density adds.
36715         * config/arc/arc.md (cpu_facility): Add cd variant.
36716         (*movqi_insn): Add code density variant.
36717         (*movhi_insn): Likewise.
36718         (*movqi_insn): Likewise.
36719         (*addsi3_mixed): Likewise.
36720         (subsi3_insn): Likewise.
36722 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36724         * config/arc/arc.md (movsi_cond_exec): Update constraint.
36726 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36728         * config/arc/arc.c (arc_legitimize_pic_address): Handle PIC
36729         expressions with MINUS and UNARY ops.
36731 2017-03-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36733         PR target/79911
36734         * config/arm/neon.md (vec_sel_widen_ssum_lo<VQI:mode><VW:mode>3):
36735         Rename to...
36736         (vec_sel_widen_ssum_lo<mode><V_half>3): ... This. Avoid mismatch
36737         between vec_select and vector argument.
36738         (vec_sel_widen_ssum_hi<VQI:mode><VW:mode>3): Rename to...
36739         (vec_sel_widen_ssum_hi<mode><V_half>3): ... This. Likewise.
36740         (vec_sel_widen_usum_lo<VQI:mode><VW:mode>3): Rename to...
36741         (vec_sel_widen_usum_lo<mode><V_half>3): ... This.
36742         (vec_sel_widen_usum_hi<VQI:mode><VW:mode>3): Rename to...
36743         (vec_sel_widen_usum_hi<mode><V_half>3): ... This.
36745 2017-03-13  Richard Biener  <rguenther@suse.de>
36747         PR other/79991
36748         * params.def (vect-max-peeling-for-alignment): Fix typo.
36750 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36752         * doc/install.texi (Specific) <mips-*-*>: Remove description of
36753         issue that only occurred with binutils below 2.18.
36755 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36757         * doc/install.texi (Specific) <cris-axis-elf>: No longer
36758         refer to binutils 2.11/2.12 minimum.
36760 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36762         * doc/install.texi (Specific) <powerpc-*-*>: Remove link to
36763         ftp.kernel.org and simplify binutils requirement.
36765 2017-03-11  Gerald Pfeifer  <gerald@pfeifer.com>
36767         * doc/invoke.texi (Warning Options): Fix spelling of link-time
36768         optimization.
36769         (Optimize Options): Ditto.  Also remove redundancy.
36771 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36773         PR translation/79848
36774         * ipa-devirt.c (warn_types_mismatch): Simplify uses of "%<%s%>" to
36775         "%qs".
36776         * ipa-pure-const.c (suggest_attribute): Likewise.  Convert _
36777         to G_ to avoid double translation.
36779 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36781         PR translation/79923
36782         * auto-profile.c (get_combined_location): Convert leading
36783         character of diagnostics to lower case and remove trailing period.
36784         (read_profile): Likewise for various diagnostics.
36785         * config/arm/arm.c (arm_option_override): Remove trailing period
36786         from various diagnostics.
36787         * config/msp430/msp430.c (msp430_expand_delay_cycles): Likewise.
36788         (msp430_expand_delay_cycles): Likewise.
36790 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36792         PR target/79925
36793         * config/aarch64/aarch64.c (aarch64_validate_mcpu): Quote the
36794         full command-line argument, rather than just "str".
36795         (aarch64_validate_march): Likewise.
36796         (aarch64_validate_mtune): Likewise.
36798 2017-03-10  Bernd Schmidt  <bschmidt@redhat.com>
36800         PR rtl-optimization/78911
36801         * lra-assigns.c (must_not_spill_p): New function.
36802         (spill_for): Use it.
36804 2017-03-10  Jakub Jelinek  <jakub@redhat.com>
36806         PR tree-optimization/79981
36807         * tree-vrp.c (extract_range_basic): Handle IMAGPART_EXPR of
36808         ATOMIC_COMPARE_EXCHANGE ifn result.
36809         (stmt_interesting_for_vrp, vrp_visit_stmt): Handle
36810         IFN_ATOMIC_COMPARE_EXCHANGE.
36812 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36814         PR driver/79875
36815         * opts.c (parse_sanitizer_options): Add missing question mark to
36816         "did you mean" message.
36818 2017-03-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36820         * config/rs6000/rs6000-builtin.def (VMULEUB_UNS): Remove orphaned
36821         built-in.
36822         (VMULEUH_UNS): Likewise.
36823         (VMULOUB_UNS): Likewise.
36824         (VMULOUH_UNS): Likewise.
36825         * config/rs6000/rs6000.c (builtin_function_type): Remove
36826         references to ALTIVEC_BUILTIN_VMUL[EO]U[BH]_UNS.
36828 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36830         PR bootstrap/79952
36831         * read-rtl-function.c (function_reader::read_rtx_operand): Update
36832         x with result of extra_parsing_for_operand_code_0.
36833         (function_reader::extra_parsing_for_operand_code_0): Convert
36834         return type from void to rtx, returning x.  When reading
36835         SYMBOL_REF with SYMBOL_FLAG_HAS_BLOCK_INFO, reallocate x to the
36836         larger size containing struct block_symbol.
36838 2017-03-10  Segher Boessenkool  <segher@kernel.crashing.org>
36840         * config/rs6000/rs6000.c (rs6000_option_override_internal): Disallow
36841         -mfloat128-hardware without -m64.
36843 2017-03-10  Will Schmidt <will_schmidt@vnet.ibm.com>
36845         PR target/79941
36846         * config/rs6000/rs6000.c (builtin_function_type): Add VMUL*U[HB]
36847         entries to the case statement that marks unsigned arguments to
36848         overloaded functions.
36850 2017-03-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
36852         * config/rs6000/rs6000.c (rs6000_option_override_internal): Fix
36853         two typographic errors in the handling of TARGET_UPPER_REGS_DI.
36855 2017-03-10  Pat Haugen  <pthaugen@us.ibm.com>
36857         PR target/79907
36858         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Test
36859         TARGET_UPPER_REGS_DI when setting 'wi' constraint regclass.
36861 2017-03-10  Martin Liska  <mliska@suse.cz>
36863         PR target/65705
36864         PR target/69804
36865         * toplev.c (process_options): Enable MPX with LSAN and UBSAN.
36866         * tree-chkp.c (chkp_walk_pointer_assignments): Verify that
36867         FIELD != NULL.
36869 2017-03-10  Olivier Hainque  <hainque@adacore.com>
36871         * tree-switch-conversion (array_value_type): Start by resetting
36872         candidate type to it's main variant.
36874 2017-03-10  Jakub Jelinek  <jakub@redhat.com>
36876         PR rtl-optimization/79909
36877         * combine.c (try_combine): Use simplify_replace_rtx on individual
36878         CALL_INSN_FUNCTION_USAGE elements instead of replace_rtx on copy_rtx
36879         of the whole CALL_INSN_FUNCTION_USAGE.
36881         PR tree-optimization/79972
36882         * gimple-ssa-warn-alloca.c (alloca_call_type): Only call
36883         get_range_info on SSA_NAMEs.  Formatting fixes.
36885 2017-03-10  Richard Biener  <rguenther@suse.de>
36886             Jakub Jelinek  <jakub@redhat.com>
36888         PR tree-optimization/77975
36889         * tree-ssa-loop-niter.c (get_base_for): Allow phi argument from latch
36890         edge to be constant.
36891         (get_val_for): For constant x return it.  Formatting fix.
36892         (loop_niter_by_eval): Avoid pointless looping if the next iteration
36893         would use the same bases as the current one.
36895 2017-03-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36897         * config/rs6000/rs6000.c (rs6000_gen_le_vsx_permute): Use rotate
36898         instead of vec_select for V1TImode.
36899         * conifg/rs6000/vsx.md (VSX_LE): Remove mode iterator that is no
36900         longer needed.
36901         (VSX_LE_128): Add V1TI to this mode iterator.
36902         (*vsx_le_perm_load_<mode>): Change to use VSX_D mode iterator.
36903         (*vsx_le_perm_store_<mode>): Likewise.
36904         (pre-reload splitter for VSX stores): Likewise.
36905         (post-reload splitter for VSX stores): Likewise.
36906         (*vsx_xxpermdi2_le_<mode>): Likewise.
36907         (*vsx_lxvd2x2_le_<mode>): Likewise.
36908         (*vsx_stxvd2x2_le_<mode>): Likewise.
36910 2017-03-09  Michael Eager  <eager@eagercon.com>
36912         Correct failures with --enable-checking=yes,rtl.
36914         * config/microblaze/microblaze.c (microblaze_expand_shift):
36915         Replace GET_CODE test with CONST_INT_P and INTVAL test with
36916         test for const0_rtx.
36917         * config/microblaze/microblaze.md (ashlsi3_byone, ashrsi3_byone,
36918         lshrsi3_byone): Replace INTVAL with test for const1_rtx.
36920 2017-03-09  Richard Biener  <rguenther@suse.de>
36922         PR tree-optimization/79977
36923         * graphite-scop-detection.c (scop_detection::merge_sese):
36924         Handle the case of extra exits to blocks dominating the entry.
36926 2017-03-09  Toma Tabacu  <toma.tabacu@imgtec.com>
36928         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
36929         Document rdynamic.
36931 2017-03-09  Vladimir Makarov  <vmakarov@redhat.com>
36933         PR rtl-optimization/79949
36934         * lra-constraints.c (process_alt_operands): Check memory when
36935         trying to predict a cycle.  Print about the overall increase.
36937 2017-03-09  Richard Biener  <rguenther@suse.de>
36939         PR middle-end/79971
36940         * gimple-expr.c (useless_type_conversion_p): Preserve
36941         TYPE_SATURATING for fixed-point types.
36943 2017-03-09  Richard Biener  <rguenther@suse.de>
36945         PR ipa/79970
36946         * ipa-prop.c (ipa_modify_formal_parameters): Avoid changing
36947         alignment of BLKmode params.
36949 2017-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36951         PR target/79913
36952         * config/aarch64/iterators.md (VALL_F16_NO_V2Q): New mode iterator.
36953         (VALL_NO_V2Q): Likewise.
36954         (VDQF_DF): Delete.
36955         * config/aarch64/aarch64-simd.md
36956         (aarch64_dup_lane_<vswap_width_name><mode>): Use VALL_F16_NO_V2Q
36957         iterator.
36958         (*aarch64_simd_vec_copy_lane_<vswap_width_name><mode>): Use
36959         VALL_NO_V2Q mode iterator.
36960         (*aarch64_vgetfmulx<mode>): Use VDQF iterator.
36962 2017-03-09  Martin Liska  <mliska@suse.cz>
36964         PR tree-optimization/79631
36965         * tree-chkp-opt.c (chkp_is_constant_addr): Call
36966         tree_int_cst_sign_bit just for INTEGER constants.
36968 2017-03-09  Martin Liska  <mliska@suse.cz>
36970         PR target/65705
36971         PR target/69804
36972         * toplev.c (process_options): Disable -fcheck-pointer-bounds with
36973         sanitizers.
36975 2017-03-09  Marek Polacek  <polacek@redhat.com>
36977         PR c++/79672
36978         * tree.c (inchash::add_expr): Handle TREE_VEC.
36980 2017-03-09  Martin Liska  <mliska@suse.cz>
36982         PR ipa/79764
36983         (chkp_narrow_size_and_offset): New function.
36984         (chkp_parse_array_and_component_ref): Support BIT_FIELD_REF.
36985         (void chkp_parse_bit_field_ref): New function.
36986         (chkp_make_addressed_object_bounds): Add case for BIT_FIELD_REF.
36987         (chkp_process_stmt): Use chkp_parse_bit_field_ref.
36989 2017-03-09  Martin Liska  <mliska@suse.cz>
36991         PR ipa/79761
36992         * tree-chkp.c (chkp_get_bound_for_parm): Get bounds for a param.
36993         (chkp_find_bounds_1): Remove gcc_unreachable.
36995 2017-03-09  Jakub Jelinek  <jakub@redhat.com>
36997         PR sanitizer/79944
36998         * asan.c (get_mem_refs_of_builtin_call): For BUILT_IN_ATOMIC* and
36999         BUILT_IN_SYNC*, determine the access type from the size suffix and
37000         always build a MEM_REF with that type.  Handle forgotten
37001         BUILT_IN_SYNC_FETCH_AND_NAND_16 and BUILT_IN_SYNC_NAND_AND_FETCH_16.
37003         PR target/79932
37004         * config/i386/avx512vlintrin.h (_mm256_cmpge_epi32_mask,
37005         _mm256_cmpge_epi64_mask, _mm256_cmpge_epu32_mask,
37006         _mm256_cmpge_epu64_mask, _mm256_cmple_epi32_mask,
37007         _mm256_cmple_epi64_mask, _mm256_cmple_epu32_mask,
37008         _mm256_cmple_epu64_mask, _mm256_cmplt_epi32_mask,
37009         _mm256_cmplt_epi64_mask, _mm256_cmplt_epu32_mask,
37010         _mm256_cmplt_epu64_mask, _mm256_cmpneq_epi32_mask,
37011         _mm256_cmpneq_epi64_mask, _mm256_cmpneq_epu32_mask,
37012         _mm256_cmpneq_epu64_mask, _mm256_mask_cmpge_epi32_mask,
37013         _mm256_mask_cmpge_epi64_mask, _mm256_mask_cmpge_epu32_mask,
37014         _mm256_mask_cmpge_epu64_mask, _mm256_mask_cmple_epi32_mask,
37015         _mm256_mask_cmple_epi64_mask, _mm256_mask_cmple_epu32_mask,
37016         _mm256_mask_cmple_epu64_mask, _mm256_mask_cmplt_epi32_mask,
37017         _mm256_mask_cmplt_epi64_mask, _mm256_mask_cmplt_epu32_mask,
37018         _mm256_mask_cmplt_epu64_mask, _mm256_mask_cmpneq_epi32_mask,
37019         _mm256_mask_cmpneq_epi64_mask, _mm256_mask_cmpneq_epu32_mask,
37020         _mm256_mask_cmpneq_epu64_mask, _mm_cmpge_epi32_mask,
37021         _mm_cmpge_epi64_mask, _mm_cmpge_epu32_mask, _mm_cmpge_epu64_mask,
37022         _mm_cmple_epi32_mask, _mm_cmple_epi64_mask, _mm_cmple_epu32_mask,
37023         _mm_cmple_epu64_mask, _mm_cmplt_epi32_mask, _mm_cmplt_epi64_mask,
37024         _mm_cmplt_epu32_mask, _mm_cmplt_epu64_mask, _mm_cmpneq_epi32_mask,
37025         _mm_cmpneq_epi64_mask, _mm_cmpneq_epu32_mask, _mm_cmpneq_epu64_mask,
37026         _mm_mask_cmpge_epi32_mask, _mm_mask_cmpge_epi64_mask,
37027         _mm_mask_cmpge_epu32_mask, _mm_mask_cmpge_epu64_mask,
37028         _mm_mask_cmple_epi32_mask, _mm_mask_cmple_epi64_mask,
37029         _mm_mask_cmple_epu32_mask, _mm_mask_cmple_epu64_mask,
37030         _mm_mask_cmplt_epi32_mask, _mm_mask_cmplt_epi64_mask,
37031         _mm_mask_cmplt_epu32_mask, _mm_mask_cmplt_epu64_mask,
37032         _mm_mask_cmpneq_epi32_mask, _mm_mask_cmpneq_epi64_mask,
37033         _mm_mask_cmpneq_epu32_mask, _mm_mask_cmpneq_epu64_mask): Move
37034         definitions outside of __OPTIMIZE__ guarded section.
37036         PR target/79932
37037         * config/i386/avx512bwintrin.h (_mm512_packs_epi32,
37038         _mm512_maskz_packs_epi32, _mm512_mask_packs_epi32,
37039         _mm512_packus_epi32, _mm512_maskz_packus_epi32,
37040         _mm512_mask_packus_epi32): Move definitions outside of __OPTIMIZE__
37041         guarded section.
37043 2017-03-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
37045         * config/s390/vx-builtins.md ("vfee<mode>", "vfeez<mode>")
37046         ("vfenez<mode>"): Add missing constraints.
37048 2017-03-08  Martin Sebor  <msebor@redhat.com>
37050         PR target/79928
37051         * config/nds32/nds32.c (nds32_option_override):
37052         Fix misspelled diagnostic.
37054 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
37056         PR c/79940
37057         * gimplify.c (gimplify_omp_for): Replace index var in outer
37058         taskloop statement with an artificial variable and add
37059         OMP_CLAUSE_PRIVATE clause for it.
37061 2017-03-08  Richard Biener  <rguenther@suse.de>
37063         PR tree-optimization/79955
37064         * tree-ssa-uninit.c (warn_uninitialized_vars): Do not warn
37065         for accesses that are completely outside of the variable.
37067 2017-03-08  Andrew Haley  <aph@redhat.com>
37069         PR tree-optimization/79943
37070         * tree-ssa-loop-split.c (compute_new_first_bound): When
37071         calculating the new upper bound, (END-BEG) should be added, not
37072         subtracted.
37074 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
37076         * config/avr/avr.md (setmemhi): Make sure match_dup
37077         operand number comes before match_scratch.
37079 2017-03-08  Richard Biener  <rguenther@suse.de>
37081         PR tree-optimization/79920
37082         * tree-vect-slp.c (vect_create_mask_and_perm): Remove and inline
37083         with ncopies == 1 to ...
37084         (vect_transform_slp_perm_load): ... here.  Properly compute
37085         all element loads by iterating VF times over the group.  Do
37086         not handle ncopies (computed in a broken way) in
37087         vect_create_mask_and_perm.
37089 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
37091         PR sanitizer/79904
37092         * internal-fn.c (expand_vector_ubsan_overflow): If arg0 or arg1
37093         is a uniform vector, use uniform_vector_p return value instead of
37094         building ARRAY_REF on folded VIEW_CONVERT_EXPR to array type.
37096 2017-03-07  Marek Polacek  <polacek@redhat.com>
37098         PR middle-end/79809
37099         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Use HOST_WIDE_INT.
37100         (alloca_call_type): Likewise.
37102 2017-03-07  Martin Liska  <mliska@suse.cz>
37104         * gcov.c (process_args): Put comment to correct location.
37106 2017-03-07  Martin Liska  <mliska@suse.cz>
37108         PR middle-end/68270
37109         * tree-chkp.c (chkp_may_narrow_to_field): Add new argument ref.
37110         Use array_at_struct_end_p instead of DECL_CHAIN (field).
37111         (chkp_narrow_bounds_for_field): Likewise.
37112         (chkp_parse_array_and_component_ref): Pass one more argument to
37113         call.
37115 2017-03-07  Richard Biener  <rguenther@suse.de>
37117         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Preserve
37118         preheaders.
37120 2017-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
37122         * config/i386/i386.c (ix86_local_alignment): Align most aggregates
37123         of 16 bytes and more to 16 bytes, not those of 16 bits and more.
37125 2017-03-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
37127         PR c/79855
37128         * params.def (PARAM_STORE_MERGING_ALLOW_UNALIGNED): Add full stop
37129         to end of description.
37130         (PARAM_MAX_STORES_TO_MERGE): Likewise.
37132 2017-03-07  Jakub Jelinek  <jakub@redhat.com>
37134         PR rtl-optimization/79901
37135         * config/i386/sse.md (*avx512bw_<code><mode>3<mask_name>): Renamed to
37136         ...
37137         (*avx512f_<code><mode>3<mask_name>): ... this.
37138         (<code><mode>3 with maxmin code iterator): Use VI8_AVX2_AVX512F
37139         iterator instead of VI8_AVX2_AVX512BW.
37141         PR rtl-optimization/79901
37142         * expr.c (expand_expr_real_2): For vector MIN/MAX, if there is no
37143         min/max expander, expand it using expand_vec_cond_expr.
37145         PR sanitizer/79897
37146         * ubsan.c (ubsan_encode_value): Call mark_addressable on the
37147         temporary.
37149 2017-03-06  Jakub Jelinek  <jakub@redhat.com>
37151         PR c++/79821
37152         * dwarf2out.h (dw_vec_const): Change array type from unsigned char *
37153         to void * for PCH reasons.
37154         * dwarf2out.c (output_loc_operands, output_die): Cast
37155         v.val_vec.array to unsigned char *.
37157 2017-03-06  John David Anglin  <danglin@gcc.gnu.org>
37159         PR target/77850
37160         * config/pa/pa-64.h (PAD_VARARGS_DOWN): Don't pad down complex and
37161         vector types.
37163 2017-03-06  Vladimir Makarov  <vmakarov@redhat.com>
37165         PR rtl-optimization/79571
37166         * lra-constraints.c (process_alt_operands): Calculate static
37167         reject and subtract it from overall when only addresses will be
37168         reloaded.
37170 2017-03-06  Julia Koval  <julia.koval@intel.com>
37172         PR target/79793
37173         * config/i386/i386.c (ix86_minimum_incoming_stack_boundary): Set
37174         incoming stack boundary to 128 for 64-bit targets.
37176 2017-03-06  Richard Biener  <rguenther@suse.de>
37178         PR tree-optimization/79894
37179         * tree-vectorizer.c (vectorize_loops): Set loop_vectorized_call
37180         to NULL after folding it.
37182 2017-03-06  Richard Biener  <rguenther@suse.de>
37184         PR tree-optimization/79824
37185         * tree-vect-stmts.c (get_group_load_store_type): Fix alignment
37186         check disabling peeling for gaps.
37188 2017-03-06  Toma Tabacu  <toma.tabacu@imgtec.com>
37190         * doc/sourcebuild.texi (Effective-Target Keywords, Environment
37191         attributes): Document gettimeofday.
37193 2017-03-06  Robin Dapp  <rdapp@linux.vnet.ibm.com>
37195         * config/s390/s390.c (s390_option_override_internal): Set
37196         PARAM_MIN_VECT_LOOP_BOUND
37198 2017-03-06  Robin Dapp  <rdapp@linux.vnet.ibm.com>
37200         * config/s390/s390.c (s390_asm_output_function_label): Use nopr %r0.
37201         * config/s390/s390.md: Likewise.
37203 2017-03-06  Jakub Jelinek  <jakub@redhat.com>
37205         PR target/79812
37206         * config/i386/sse.md (VI8F_256_512): Remove mode iterator.
37207         (<avx2_avx512>_perm<mode>): Rename to ...
37208         (avx2_perm<mode>): ... this.  Use VI8F_256 iterator instead
37209         of VI8F_256_512.
37210         (<avx512>_perm<mode>_mask): Rename to ...
37211         (avx512vl_perm<mode>_mask): ... this.  Use VI8F_256 iterator instead
37212         of VI8F_256_512.
37213         (<avx2_avx512>_perm<mode>_1<mask_name>): Rename to ...
37214         (avx2_perm<mode>_1<mask_name): ... this.  Use VI8F_256 iterator
37215         instead of VI8F_256_512.
37216         (avx512f_perm<mode>): New define_expand.
37217         (avx512f_perm<mode>_mask): Likewise.
37218         (avx512f_perm<mode>_1<mask_name>): New define_insn.
37219         (<avx512>_vec_dup<mode>_1): Fix up vec_select mode.
37221 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37223         * config/mips/mips-msa.md (msa_fmax_a_<msafmt>, msa_fmin_a_<msafmt>,
37224         msa_max_a_<msafmt>, msa_min_a_<msafmt>): Introduce mode interator for
37225         if_then_else.
37226         (smin<mode>3, smax<mode>3): Change operand print code from 'B' to 'E'.
37228 2017-03-06  Martin Liska  <mliska@suse.cz>
37230         PR sanitize/79783
37231         * asan.c (asan_expand_poison_ifn): Do not expand ASAN_POISON
37232         when having a SSA NAME w/o VAR_DECL assigned to it.
37234 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37236         * config/mips/mips-msa.md (msa_dotp_<su>_d, msa_dpadd_<su>_d,
37237         msa_dpsub_<su>_d): Fix MODE for vec_select.
37239 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37241         * config/mips/mips.c (mips_gen_const_int_vector): Change type of last
37242         argument.
37243         * config/mips/mips-protos.h (mips_gen_const_int_vector): Likewise.
37245 2017-03-06  Richard Biener  <rguenther@suse.de>
37247         * lto-streamer.c (lto_check_version): Use %qs in diagnostics.
37248         * plugin.c (register_plugin_info): Likewise.
37249         * tree-chkp.c (chkp_make_static_const_bounds): Likewise.
37251 2017-03-05  Jakub Jelinek  <jakub@redhat.com>
37253         * config/i386/sse.md (sse_storehps, sse_storelps,
37254         avx_<castmode><avxsizesuffix>_<castmode>,
37255         avx512f_<castmode><avxsizesuffix>_<castmode>,
37256         avx512f_<castmode><avxsizesuffix>_256<castmode>): Require
37257         in condition that at least one operand is not a MEM.
37259 2017-03-03  Jakub Jelinek  <jakub@redhat.com>
37261         PR middle-end/79805
37262         * internal-fn.def (ATOMIC_BIT_TEST_AND_SET, ATOMIC_BIT_TEST_AND_RESET,
37263         ATOMIC_BIT_TEST_AND_COMPLEMENT, ATOMIC_COMPARE_EXCHANGE): Remove
37264         ECF_NOTHROW.
37265         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Set
37266         gimple_call_nothrow_p flag based on whether original builtin can throw.
37267         If it can, emit following stmts on the fallthrough edge.
37268         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Similarly, except
37269         don't create new bb if inserting just debug stmts on the edge, try to
37270         insert them on the fallthru bb or just reset debug stmts.
37272 2017-03-03  Segher Boesssenkool  <segher@kernel.crashing.org>
37274         PR target/43763
37275         * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Save and
37276         restore recog_data (including the operand rtxes inside it) around
37277         the call to get_insn_template.
37279 2017-03-03  Martin Sebor  <msebor@redhat.com>
37281         PR tree-optimization/79699
37282         * context.c (context::~context): Free MPFR caches to avoid
37283         a memory leak on program exit.
37285 2017-03-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
37287         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
37288         Use wide_int::ulow () instead of .elt (0).
37290 2017-03-03  Uros Bizjak  <ubizjak@gmail.com>
37292         * config/i386/i386.md (*pushtf): Change *roF constraint to *roC.
37293         (*pushxf): Limit oF constraint to 32bit targets and add oC
37294         constraint for 64bit targets.
37295         (pushxf splitter): Use PUSH_ROUNDING to calculate stack adjustment.
37296         (*pushdf): Change rmF constraint to rmC.
37298 2017-03-03  Martin Liska  <mliska@suse.cz>
37300         * tree-ssa-loop-prefetch.c (pass_loop_prefetch::execute):
37301         Remove unused variable.
37303 2017-03-03  Jakub Jelinek  <jakub@redhat.com>
37305         PR target/79807
37306         * config/i386/i386.c (ix86_expand_multi_arg_builtin): If target
37307         is a memory operand, increase num_memory.
37308         (ix86_expand_args_builtin): Likewise.
37310 2017-03-03  Jan Hubicka  <jh@suse.cz>
37312         PR lto/79760
37313         * ipa-devirt.c (maybe_record_node): Properly handle
37314         __cxa_pure_virtual visibility.
37316 2017-03-03  Martin Liska  <mliska@suse.cz>
37318         PR tree-optimization/79803
37319         * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Remove
37320         assert.
37321         (pass_loop_prefetch::execute): Disabled optimization if an
37322         assumption about L1 cache size is not met.
37324 2017-03-03  Martin Liska  <mliska@suse.cz>
37326         PR rtl-optimization/79574
37327         * gcse.c (struct gcse_expr): Use HOST_WIDE_INT instead of int.
37328         (hash_scan_set): Likewise.
37329         (dump_hash_table): Likewise.
37330         (hoist_code): Likewise.
37332 2017-03-03  Richard Biener  <rguenther@suse.de>
37334         * fixed-value.c (fixed_from_string): Restore use of elt (1)
37335         in place of uhigh ().
37336         (fixed_convert_from_real): Likewise.
37338 2017-03-03  Uros Bizjak  <ubizjak@gmail.com>
37340         PR target/79514
37341         * config/i386/i386.md (*pushxf_rounded): Use Pmode instead of DImode.
37343 2017-03-03  Richard Biener  <rguenther@suse.de>
37345         PR middle-end/79818
37346         * match.pd ( X +- C1 CMP C2 -> X CMP C2 -+ C1): Add missing
37347         TYPE_OVERFLOW_UNDEFINED check.
37349 2017-03-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
37351         * config/rs6000/vector.md (vector_ne_<mode>_p): Correct operand
37352         numbers.
37353         (vector_ae_<mode>_p): Likewise.
37354         (vector_nez_<mode>_p): Likewise.
37355         (vector_ne_v2di_p): Likewise.
37356         (vector_ae_v2di_p): Likewise.
37357         (vector_ne_<mode>_p): Likewise.
37358         * config/rs6000/vsx.md (vsx_tsqrt<mode>2_fg): Correct operand
37359         numbers.
37360         (vsx_tsqrt<mode>2_fe): Likewise.
37362 2017-03-02  Uros Bizjak  <ubizjak@gmail.com>
37364         PR target/79514
37365         * config/i386/i386.md (*pushxf_rounded): New insn_and_split pattern.
37367 2017-03-02  Jakub Jelinek  <jakub@redhat.com>
37369         PR rtl-optimization/79780
37370         * cprop.c (one_cprop_pass): When second and further conditional trap
37371         in a single basic block is turned into an unconditional trap, turn it
37372         into a deleted note to avoid RTL verification failures.
37374 2017-03-02  Richard Biener  <rguenther@suse.de>
37376         * fold-const.c (const_binop): Use ulow () instead of elt (0).
37378 2017-03-02  Richard Biener  <rguenther@suse.de>
37380         PR tree-optimization/79345
37381         PR c++/42000
37382         * tree-ssa-alias.c (walk_aliased_vdefs_1): Take a limit
37383         param and abort the walk, returning -1 if it is hit.
37384         (walk_aliased_vdefs): Take a limit param and pass it on.
37385         * tree-ssa-alias.h (walk_aliased_vdefs): Add a limit param,
37386         defaulting to 0 and return a signed int.
37387         * tree-ssa-uninit.c (struct check_defs_data): New struct.
37388         (check_defs): New helper.
37389         (warn_uninitialized_vars): Use walk_aliased_vdefs to warn
37390         about uninitialized memory.
37391         * fixed-value.c (fixed_from_string): Use ulow/uhigh to avoid
37392         bogus uninitialized warning.
37393         (fixed_convert_from_real): Likewise.
37395 2017-03-02  Bin Cheng  <bin.cheng@arm.com>
37397         PR tree-optimization/66768
37398         * tree-ssa-loop-ivopts.c (find_interesting_uses_address): Skip addr
37399         iv_use if base object can't be determined.
37401 2017-03-02  Jakub Jelinek  <jakub@redhat.com>
37403         PR tree-optimization/79345
37404         * gensupport.h (struct pattern_stats): Add min_scratch_opno field.
37405         * gensupport.c (get_pattern_stats_1) <case MATCH_SCRATCH>: Update it.
37406         (get_pattern_stats): Initialize it.
37407         * genemit.c (gen_expand): Verify match_scratch numbers come after
37408         match_operand/match_dup numbers.
37409         * config/i386/i386.md (<s>mul<mode>3_highpart): Swap match_dup and
37410         match_scratch numbers.
37411         * config/i386/sse.md (avx2_gathersi<mode>, avx2_gatherdi<mode>):
37412         Likewise.
37413         * config/s390/s390.md (trunctdsd2): Likewise.
37415 2017-03-02  Richard Biener  <rguenther@suse.de>
37417         * wide-int.h (wide_int_storage::operator=): Implement in terms
37418         of wi::copy.
37420 2017-03-02  Richard Biener  <rguenther@suse.de>
37422         PR tree-optimization/79777
37423         * tree-ssa-pre.c (eliminate_insert): Give up if we simplify
37424         the to insert expression to sth existing.
37426 2017-03-01  Martin Sebor  <msebor@redhat.com>
37428         PR middle-end/79692
37429         * gimple-ssa-sprintf.c
37430         (directive::known_width_and_precision): New function.
37431         (format_integer): Use it.
37432         (get_mpfr_format_length): Consider the full range of precision
37433         when computing %g output with the # flag.  Set the likely byte
37434         count to 3 rather than 1 when precision is indeterminate.
37435         (format_floating): Correct the lower bound of precision.
37437 2017-03-01  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
37439         * doc/invoke.texi: Document default code model for 64-bit Linux.
37441 2017-03-01  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
37443         PR target/79752
37444         * config/rs6000/rs6000.md (peephole2 for udiv/umod): Should emit
37445         udiv rather than div since input pattern is unsigned.
37447 2017-03-01  Uros Bizjak  <ubizjak@gmail.com>
37449         * config/i386/i386.c (print_reg): Warn for values of
37450         unsupported size in integer register.
37452 2017-03-01  Michael Meissner  <meissner@linux.vnet.ibm.com>
37454         PR target/79439
37455         * config/rs6000/predicates.md (current_file_function_operand): Do
37456         not allow self calls to be local if the function is replaceable.
37458 2017-03-01  Kelvin Nilsen  <kelvin@gcc.gnu.org>
37460         PR target/79395
37461         * config/rs6000/altivec.h (vec_ctz and others): Change the
37462         preprocessor macro that controls conditional compilation from
37463         _ARCH_PWR9 to __POWER9_VECTOR__.
37464         (vec_all_ne): Change parameterization of __altivec_scalar_pred
37465         macro expansion under preprocessor #ifdef __POWER9_VECTOR__
37466         control (instead of _ARCH_PWR9 control) so that template
37467         definition uses power9-specific function.
37468         (vec_any_eq): Likewise.
37469         (vec_all_ne): Change macro definition to use a power9-specific
37470         expansion under #ifdef __POWER9_VECTOR__ control (instead of
37471         _ARCH_PWR9 control).
37472         (vec_any_eq) Likewise.
37473         * config/rs6000/rs6000-builtin.def (CMPNEF): Remove BU_P9V_AV_2
37474         expansion for CMPNEF to remove support for xvcmpnesp instruction.
37475         (CMPNED): Remove BU_P9V_AV2 expansion for CMPNED to remove
37476         support for xvcmpnedp instruction.
37477         (VCMPNEB_P): Replace BU_P9V_AV_P macro expansion with BU_P9V_AV_2
37478         macro expansion so that Power9 implementation of vec_all_ne does
37479         not use the AltiVec predicate framework.
37480         (VCMPNEH_P): Likewise.
37481         (VCMPNEW_P): Likewise.
37482         (VCMPNED_P): Likewise.
37483         (VCMPNEFP_P): Likewise.
37484         (VCMPNEDP_P): Likewise.
37485         (VCMPAEB_P): Add BU_P9V_AV_2 macro expansion to change
37486         implementation of vec_any_eq to not use AltiVec predicate
37487         framework.
37488         (VCMPAEH_P): Likewise.
37489         (VCMPAEW_P): Likewise.
37490         (VCMPAED_P): Likewise.
37491         (VCMPAEFP_P): Likewise.
37492         (VCMPAEDP_P): Likewise.
37493         (VCMPNE_P): Replace BU_P9V_OVERLOAD_P macro expansion with
37494         BU_P9V_OVERLOAD_2 so that Power9 implementation of vec_all_ne does
37495         not use the AltiVec predicate framework.
37496         (VCMPAE_P): Add BU_P9V_OVERLOAD_2 macro to change implementation
37497         of vec_any_eq to not use AltiVec predicate framework.
37498         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Add
37499         support for predefined __POWER9_VECTOR__ macro to indicate that
37500         Power9 instruction selection is enabled.
37501         (altivec_overloaded_builtins): Remove extraneous
37502         ALTIVEC_BUILTIN_VEC_CMPNE entry for overloaded
37503         function argument types RS6000_BTI_bool_V16QI and
37504         RS6000_BTI_bool_V16QI.  Remove erroneous ALTIVEC_BUILTIN_VEC_CMPNE
37505         entry for overloaded function argument types RS6000_BTI_bool_V4SI
37506         andRS6000_BTI_bool_V4SI, mapping to P9V_BUILTIN_CMPNEB.  Remove
37507         two entries mapping to P9V_BUITIN_CMPNED and one entry mapping to
37508         P9V_BUILTIN_CMPNEF to force use of instructions not specific to
37509         Power9 for implementations of vec_cmpne.  Change the signature for
37510         all definitions of the overloaded P9V_BUILTIN_VEC_CMPNE_P function
37511         (representing vec_all_ne) to remove the previously described first
37512         argument of type RS6000_BTI_INTSI, as this was an artifact of
37513         reliance on the AltiVec predicate framework, which is no longer
37514         used in the implementation of these functions.  Add
37515         P9V_BUILTIN_VEC_VCMPAE_P entries (representing the vec_anyeq
37516         function) to match all of the P9V_BUILTIN_VEC_VCMNE_P entries
37517         since, unlike the AltiVec predicate framework implementation, we
37518         do not share function descriptors between vec_alle and vec_anyeq.
37519         (altivec_resolve_overloaded_builtin): Add SFmode and DFmode to the
37520         set of modes that receive special treatment even when
37521         TARGET_P9_VECTOR is true.  The special treatment emits code that
37522         does not depend on Power9 instructions.
37523         * config/rs6000/vector.md (vector_ne_<mode>_p): Change this
37524         define_expand to not rely on AltiVec predicate framework.
37525         (vector_ae_<mode>p): New define_expand to represent vec_any_eq
37526         function.
37527         (vector_ne_v2di_p): Change this define_expand to not rely on
37528         AltiVec predicate framework.
37529         (vector_ae_v2di_p): New define_expand to represent vec_any_eq
37530         function.
37531         (vector_ne_<mode>_p): Change this define_expand to not rely on
37532         AltiVec predicate framework.
37533         (vector_ae_<mode>p): New define_expand to represent vec_any_eq
37534         function.
37535         * config/rs6000/vsx.md (*vsx_ne_<mode>_p): For modes VSX_EXTRACT_I
37536         (V16QI, V8HI, V4SI), correct a typo in the code emitted for this
37537         define_insn pattern.
37538         (*vsx_ne_<mode>_p): For modes VSX_F (V4SF and V2DF), remove this
37539         define_insn pattern because the xvcmpne<VSs>. instruction is not
37540         supported.
37541         (vcmpne<VSs>): Remove this define_insn because xvcmpne<VSs>
37542         instruction is not supported.
37544 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37546         * config/nvptx/nvptx.c: Include intl.h.
37548 2017-03-01  Martin Jambor  <mjambor@suse.cz>
37550         PR lto/78140
37551         * ipa-prop.h (ipa_bits): Removed field known.
37552         (ipa_jump_func): Removed field vr_known.  Changed fields bits and m_vr
37553         to pointers.  Adjusted their comments to warn about their sharing.
37554         (ipcp_transformation_summary): Change bits to a vector of pointers.
37555         (ipa_check_create_edge_args): Moved to ipa-prop.c, declare.
37556         (ipa_get_ipa_bits_for_value): Declare.
37557         * tree-vrp.h (value_range): Mark as GTY((for_user)).
37558         * ipa-prop.c (ipa_bit_ggc_hash_traits): New.
37559         (ipa_bits_hash_table): Likewise.
37560         (ipa_vr_ggc_hash_traits): Likewise.
37561         (ipa_vr_hash_table): Likewise.
37562         (ipa_print_node_jump_functions_for_edge): Adjust for bits and m_vr
37563         being pointers and vr_known being removed.
37564         (ipa_set_jf_unknown): Likewise.
37565         (ipa_get_ipa_bits_for_value): New function.
37566         (ipa_set_jfunc_bits): Likewise.
37567         (ipa_get_value_range): New overloaded functions.
37568         (ipa_set_jfunc_vr): Likewise.
37569         (ipa_compute_jump_functions_for_edge): Use the above functions to
37570         construct bits and vr parts of jump functions.
37571         (ipa_check_create_edge_args): Move here from ipa-prop.h, also allocate
37572         ipa_bits_hash_table and ipa_vr_hash_table if they do not already
37573         exist.
37574         (ipcp_grow_transformations_if_necessary): Also allocate
37575         ipa_bits_hash_table and ipa_vr_hash_table if they do not already
37576         exist.
37577         (ipa_node_params_t::duplicate): Do not copy bits, just pointers to
37578         them.  Fix too long lines.
37579         (ipa_write_jump_function): Adjust for bits and m_vr being pointers and
37580         vr_known being removed.
37581         (ipa_read_jump_function): Use new setter functions to construct bits
37582         and vr parts of jump functions or set them to NULL.
37583         (write_ipcp_transformation_info): Adjust for bits being pointers.
37584         (read_ipcp_transformation_info): Likewise.
37585         (ipcp_update_bits): Likewise.  Fix excessively long lines a trailing
37586         space.
37587         Include gt-ipa-prop.h.
37588         * ipa-cp.c (propagate_bits_across_jump_function): Adjust for bits
37589         being pointers.
37590         (ipcp_store_bits_results): Likewise.
37591         (propagate_vr_across_jump_function): Adjust for m_vr being a pointer.
37592         Do not write to existing jump functions but use a temporary instead.
37594 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37596         PR c++/79681
37597         * fold-const.c (make_bit_field_ref): If orig_inner is COMPONENT_REF,
37598         attempt to use its first operand as BIT_FIELD_REF base.
37600 2017-03-01  Richard Biener  <rguenther@suse.de>
37602         PR middle-end/79721
37603         * tree-chrec.c (chrec_evaluate): Perform computation of Newtons
37604         interpolating formula in wrapping arithmetic.
37605         (chrec_apply): Convert chrec_evaluate return value to wanted type.
37607 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37609         PR tree-optimization/79734
37610         * tree-vect-generic.c (expand_vector_condition): Optimize
37611         AVX512 vector boolean VEC_COND_EXPRs into bitwise operations.
37612         Handle VEC_COND_EXPR where comparison has different inner width from
37613         type's inner width.
37615 2017-02-28  Sandra Loosemore  <sandra@codesourcery.com>
37617         * doc/invoke.texi (ARC Options): Copy-edit to fix punctuation,
37618         markup, and similar issues.  Remove @opindex entries for things
37619         that aren't options.  Add missing -mmpy-option entries.
37621 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37623         PR tree-optimization/79737
37624         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): If bitlen is
37625         a multiple of BITS_PER_UNIT and !BYTES_BIG_ENDIAN, clear
37626         tmpbuf[byte_size - 1].  Call natice_encode_expr with byte_size - 1
37627         instead of byte_size.  Formatting fix.
37628         (shift_bytes_in_array_right): Formatting fix.
37630 2017-02-28  Eric Botcazou  <ebotcazou@adacore.com>
37632         PR target/79749
37633         * config/sparc/sparc.c (sparc_frame_pointer_required): Add missing
37634         condition on optimize for the leaf function test.
37636 2017-02-28  Martin Liska  <mliska@suse.cz>
37638         PR lto/79625
37639         * read-rtl-function.c (function_reader::handle_unknown_directive):
37640         Bail out when one uses -flto.
37642 2017-02-28  Martin Liska  <mliska@suse.cz>
37644         * common.opt: Replace space with tabular for options of <number>
37645         type.
37646         * config/i386/i386.opt: Show <number> value for
37647         -mlarge-data-threshold.
37648         * opts.c (print_filtered_help): Do not display number in hexadecimal
37649         format.
37651 2017-02-28  Martin Liska  <mliska@suse.cz>
37653         * common.opt: Fix --help=option -Q for options which are of
37654         an enum type.
37656 2017-02-28  Uros Bizjak  <ubizjak@gmail.com>
37658         * config/i386/i386.c (print_reg): Error out for values
37659         of 8-bit size in invalid integer register.
37661 2017-02-28  Martin Sebor  <msebor@redhat.com>
37663         PR tree-optimization/79691
37664         * passes.def (pass_all_optimizations_g): Enable pass_sprintf_length.
37666 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37668         PR target/79729
37669         * config/i386/i386.c (ix86_print_operand) <case 'R'>: Replace
37670         gcc_unreachable with output_operand_lossage.
37672 2017-02-28  Richard Biener  <rguenther@suse.de>
37674         PR tree-optimization/79740
37675         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Allow redundant
37676         inserts.
37677         (visit_nary_op): Insert the nary into the hashtable if we
37678         pattern-matched sth.
37679         * tree-ssa-pre.c (eliminate_insert): Robustify.
37681 2017-02-28  Richard Biener  <rguenther@suse.de>
37683         PR middle-end/79731
37684         * fold-const.c (decode_field_reference): Reject out-of-bound
37685         accesses.
37687 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37689         * config/i386/i386.c: Include intl.h.
37690         (ix86_option_override_internal): Use cond ? G_("...") : G_("...")
37691         instead of just cond ? "..." : "...".
37692         * config/nvptx/nvptx.c (nvptx_goacc_validate_dims): Likewise.
37693         * coverage.c (read_counts_file): Likewise.
37694         * omp-offload.c: Include intl.h.
37695         (oacc_loop_fixed_partitions): Use cond ? G_("...") : G_("...") instead
37696         of just cond ? "..." : "...".
37697         * gcov.c (read_count_file): Use cond ? N_("...") : N_("...") instead
37698         of just cond ? "..." : "...".
37700 2017-02-28  Richard Earnshaw  <rearnsha@arm.com>
37702         PR target/79742
37703         * config/arm/parsecpu.awk (gen_data): Set tuning target to 'tune for'
37704         entry, if present.
37705         * config/arm/arm-cpus.in (cortex-m0plus.small-multiply): Correct
37706         'tune for' CPU name.
37707         * config/arm/arm-cpu-data.h: Regenerated.
37709 2017-02-28  Richard Biener  <rguenther@suse.de>
37711         PR tree-optimization/79732
37712         * tree-inline.c (expand_call_inline): Do not shadow var.
37714 2017-02-28  Richard Biener  <rguenther@suse.de>
37716         PR tree-optimization/79723
37717         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Preserve
37718         address-space properly.
37720 2017-02-28  Thomas Schwinge  <thomas@codesourcery.com>
37722         * doc/optinfo.texi (Optimization groups): Fix option used for
37723         OPTGROUP_ALL.
37724         * doc/invoke.texi (-fopt-info): Document "omp".
37725         * dumpfile.h: Sort OPTGROUP_OMP before OPTGROUP_VEC.
37726         (OPTGROUP_ALL): Add OPTGROUP_OMP.
37727         * hsa-gen.c (pass_data_gen_hsail): Use OPTGROUP_OMP.
37728         * ipa-hsa.c (pass_data_ipa_hsa): Likewise.
37729         * omp-simd-clone.c (pass_data_omp_simd_clone): Likewise.
37731         * dumpfile.h (OPTGROUP_OPENMP): Rename to OPTGROUP_OMP.  Adjust
37732         all users.
37733         * dumpfile.c (optgroup_options): Instead of "openmp", associate
37734         OPTGROUP_OMP with "omp".
37736 2017-02-27  Pat Haugen  <pthaugen@us.ibm.com>
37738         PR target/79544
37739         * config/rs6000/rs6000-c.c (struct altivec_builtin_types): Use VSRAD
37740         for arithmetic shift of unsigned V2DI.
37742 2017-02-27  Claudiu Zissulescu  <claziss@synopsys.com>
37744         * config.gcc (arc*-): Clean up, use arc/big.h, arc/elf.h, and
37745         arc/linux.h headers.
37746         * config/arc/arc.h (TARGET_OS_CPP_BUILTINS): Remove.
37747         (LINK_SPEC): Likewise.
37748         (ARC_TLS_EXTRA_START_SPEC): Likewise.
37749         (EXTRA_SPECS): Likewise.
37750         (STARTFILE_SPEC): Likewise.
37751         (ENDFILE_SPEC): Likewise.
37752         (LIB_SPEC): Likewise.
37753         (TARGET_SDATA_DEFAULT): Likewise.
37754         (TARGET_MMEDIUM_CALLS_DEFAULT): Likewise.
37755         (MULTILIB_DEFAULTS): Likewise.
37756         (DWARF2_UNWIND_INFO): Likewise.
37757         * config/arc/big.h: New file.
37758         * config/arc/elf.h: Likewise.
37759         * config/arc/linux.h: Likewise.
37760         * config/arc/t-uClibc: Remove.
37762 2017-02-27  Bin Cheng  <bin.cheng@arm.com>
37764         PR tree-optimization/77536
37765         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): New function.
37766         (tree_transform_and_unroll_loop): Use above function to compute the
37767         estimated niter of unrolled loop and use it when scaling profile.
37768         Also use count info rather than frequency if it's non-zero.
37769         * tree-ssa-loop-manip.h niter_for_unrolled_loop(): New declaration.
37770         * tree-vect-loop.c (scale_profile_for_vect_loop): New function.
37771         (vect_transform_loop): Call above function.
37773 2017-02-27  Richard Biener  <rguenther@suse.de>
37775         PR tree-optimization/45397
37776         * tree-ssa-pre.c (eliminate_insert): Handle BIT_AND_EXPR.
37777         * tree-ssa-sccvn.c (valueized_wider_op): New helper.
37778         (visit_nary_op): Add pattern matching for CSEing sign-changed
37779         or truncated operations with wider ones.
37781 2017-02-27  Richard Biener  <rguenther@suse.de>
37783         PR tree-optimization/79690
37784         * tree-vect-stmts.c (vectorizable_store): Use vector type
37785         built from the DR with address-space.
37787 2017-02-26  Gerald Pfeifer  <gerald@pfeifer.com>
37789         * doc/invoke.texi (Optimize Options): Refine the description
37790         of asan-use-after-return.
37792 2017-02-25  Alan Modra  <amodra@gmail.com>
37794         PR rtl-optimization/79584
37795         * lra-constraints.c (base_to_reg): Reload ad->base, the entire
37796         base, not ad->base_term, the reg within base.  Remove assertion
37797         that ad->base == ad->base_term.  Replace gen_int_mode using
37798         bogus mode with const0_rtx.
37800 2017-02-25  Jakub Jelinek  <jakub@redhat.com>
37802         PR middle-end/79396
37803         * tree-eh.c (operation_could_trap_p, stmt_could_throw_1_p): Handle
37804         FMA_EXPR like tcc_binary or tcc_unary.
37806         * tree-ssa-loop-niter.c (number_of_iterations_exit): Simplify warning.
37808         PR debug/77589
37809         * dwarf2out.c (struct dw_loc_list_struct): Add noted_variable_value
37810         bitfield.
37811         (size_of_loc_descr): Handle DW_OP_GNU_variable_value.
37812         (output_loc_operands): Handle DW_OP_call_ref and
37813         DW_OP_GNU_variable_value.
37814         (struct variable_value_struct): New type.
37815         (struct variable_value_hasher): Likewise.
37816         (variable_value_hash): New variable.
37817         (string_types): Remove.
37818         (copy_loc_descr): New function.
37819         (add_loc_descr_to_each): Clarify comment.  Use copy_loc_descr.
37820         (prepend_loc_descr_to_each): New function.
37821         (add_loc_list): Fix comment typo.  Use prepend_loc_descr_to_each
37822         instead of add_loc_descr_to_each if the first argument is single
37823         location list and the second has multiple.
37824         (resolve_args_picking_1): Handle DW_OP_GNU_variable_value.
37825         (loc_list_from_tree_1): For early_dwarf, emit DW_OP_GNU_variable_value
37826         when looking for variable value which doesn't have other location info.
37827         (loc_list_from_tree): Formatting fix.
37828         (gen_array_type_die): Simplify DW_AT_string_length handling.
37829         (adjust_string_types): Remove.
37830         (gen_subprogram_die): Don't call adjust_string_types nor test/set
37831         string_types.  Call resolve_variable_values.
37832         (prune_unused_types_walk_loc_descr): Handle DW_OP_GNU_variable_value.
37833         (resolve_addr_in_expr): Likewise.  Add A argument.
37834         (copy_deref_exprloc): Remove deref argument.  Adjust for the
37835         original expression being DW_OP_GNU_variable_value with optionally
37836         DW_OP_stack_value after it instead of DW_OP_call4 with DW_OP_deref
37837         optionally after it.
37838         (optimize_string_length): Rework for DW_OP_GNU_variable_value.
37839         (resolve_addr): Adjust optimize_string_length and resolve_addr_in_expr
37840         callers.  Set remove_AT_byte_size if removing DW_AT_string_length.
37841         (variable_value_hasher::hash, variable_value_hasher::equal): New
37842         methods.
37843         (resolve_variable_value_in_expr, resolve_variable_value,
37844         resolve_variable_values, note_variable_value_in_expr,
37845         note_variable_value): New functions.
37846         (dwarf2out_early_finish): Call note_variable_value on all toplevel
37847         DIEs.
37849 2017-02-24  Jakub Jelinek  <jakub@redhat.com>
37851         PR c/79677
37852         * opts.h (handle_generated_option): Add GENERATED_P argument.
37853         * opts-common.c (handle_option): Adjust function comment.
37854         (handle_generated_option): Add GENERATED_P argument, pass it to
37855         handle_option.
37856         (control_warning_option): Pass false to handle_generated_option
37857         GENERATED_P.
37858         * opts.c (maybe_default_option): Pass true to handle_generated_option
37859         GENERATED_P.
37860         * optc-gen.awk: Likewise.
37862 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37864         * config/sh/sh.md (tstsi_t): If operands[0] is a SUBREG instead of
37865         a REG, look at the REG it is a SUBREG of.
37866         (splitter for cmpeqsi_t): Ditto.
37868 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37870         * config/pa/pa.c (pa_combine_instructions): Do not share RTL.  Make
37871         the special USEs with the pattern of the insn, not the insn itself.
37873 2017-02-24  Matthew Fortune  <matthew.fortune@imgtec.com>
37875         PR target/79473
37876         * doc/invoke.texi: Document -mload-store-pairs.
37878 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37879             Sandra Loosemore  <sandra@codesourcery.com>
37881         * config/nios2/nios2.c (nios2_simple_const_p): Returns false if the
37882         argument isn't a CONST_INT.
37883         (nios2_alternate_compare_const): Assert op is a CONST_INT.
37884         (nios2_valid_compare_const_p): Assert op is a CONST_INT.
37885         (nios2_validate_compare): Bypass alternate compare logic if *op2
37886         is not a CONST_INT.
37887         (ldstwm_operation_p): Return false if first_base is not a REG or
37888         if first_offset is not a CONST_INT.
37890 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37892         * config/cris/cris.md: Use correct operand in a define_peephole2.
37894 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37896         * config/c6x/c6x.c (predicate_insn): Do not incorrectly share RTL.
37898 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37900         * config/arc/arc.c (arc_ccfsm_advance): Only take the PATTERN of
37901         this_insn if it is an INSN or JUMP_INSN.
37902         (force_offsettable): Look at base, not at addr.
37903         * config/arc/predicates.md (brcc_nolimm_operator): Don't call INTVAL
37904         on things that aren't necessarily CONST_INTs.
37906 2017-02-24  Uros Bizjak  <ubizjak@gmail.com>
37908         * doc/invoke.texi (x86 Options, -mfpmath=sse): Mention that
37909         -mfpmath=sse is the default also for x86-32 targets with SSE2
37910         instruction set when @option{-ffast-math} is enabled
37912 2017-02-24  Jeff Law  <law@redhat.com>
37914         PR rtl-optimizatoin/79286
37915         * ira.c (update_equiv_regs): Drop may_trap_p exception to
37916         dominance test.
37918 2017-02-24  Richard Biener  <rguenther@suse.de>
37920         PR tree-optimization/79389
37921         * gimple-ssa-split-paths.c (is_feasible_trace): Properly skip
37922         debug insns.
37924 2017-02-24  Aldy Hernandez  <aldyh@redhat.com>
37926         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update
37927         function comment to reflect reality.
37928         (loop_exits_before_overflow): Fix typo in function description.
37930 2017-02-24  Richard Biener  <rguenther@suse.de>
37932         PR tree-optimization/79389
37933         * gimple-ssa-split-paths.c (is_feasible_trace): Verify more
37934         properly that a threading opportunity exists.  Detect conditional
37935         copy/constant propagation opportunities.
37937 2017-02-23  Eric Botcazou  <ebotcazou@adacore.com>
37939         * config/visium/visium.md (type): Add trap.
37940         (b): New mode attribute.
37941         (*btst): Rename into...
37942         (*btst<mode>): ...this and adjust.
37943         (*cbranchsi4_btst_insn): Rename into...
37944         (*cbranch<mode>4_btst_insn): ...this and adjust.
37945         (trap): New define_insn.
37947 2017-02-23  Jakub Jelinek  <jakub@redhat.com>
37949         PR tree-optimization/79389
37950         * ifcvt.c (struct noce_if_info): Add rev_cond field.
37951         (noce_reversed_cond_code): New function.
37952         (noce_emit_store_flag): Use rev_cond if non-NULL instead of
37953         reversed_comparison_code.  Formatting fix.
37954         (noce_try_store_flag): Test rev_cond != NULL in addition to
37955         reversed_comparison_code.
37956         (noce_try_store_flag_constants): Likewise.
37957         (noce_try_store_flag_mask): Likewise.
37958         (noce_try_addcc): Use rev_cond if non-NULL instead of
37959         reversed_comparison_code.
37960         (noce_try_cmove_arith): Likewise.  Formatting fixes.
37961         (noce_try_minmax, noce_try_abs): Clear rev_cond.
37962         (noce_find_if_block): Initialize rev_cond.
37963         (find_cond_trap): Call noce_get_condition with then_bb == trap_bb
37964         instead of false as last argument never attempt to reverse it
37965         afterwards.
37967 2017-02-23  Bin Cheng  <bin.cheng@arm.com>
37969         PR tree-optimization/79663
37970         * tree-predcom.c (combine_chains): Process refs in reverse order
37971         only for ZERO length chains, and add explaining comment.
37973 2017-02-23  Jeff Law  <law@redhat.com>
37975         PR tree-optimization/79578
37976         * tree-ssa-dse.c (clear_bytes_written_by): Use OEP_ADDRESS_OF
37977         in call to operand_equal_p.
37979 2017-01-23  Dominique d'Humieres  <dominiq@lps.ens.fr>
37981         PR target/71017
37982         * config/i386/cpuid.h: Fix another undefined behavior.
37984 2017-02-23  Richard Biener  <rguenther@suse.de>
37986         PR tree-optimization/79683
37987         * tree-vect-stmts.c (vect_analyze_stmt): Do not overwrite
37988         vector types for data-refs.
37990 2017-02-23  Martin Liska  <mliska@suse.cz>
37992         * params.def (PARAM_MIN_NONDEBUG_INSN_UID): Change default to 0.
37994 2017-02-23  Jakub Jelinek  <jakub@redhat.com>
37996         PR middle-end/79665
37997         * internal-fn.c (get_range_pos_neg): Moved to ...
37998         * tree.c (get_range_pos_neg): ... here.  No longer static.
37999         * tree.h (get_range_pos_neg): New prototype.
38000         * expr.c (expand_expr_real_2) <case TRUNC_DIV_EXPR>: If both arguments
38001         are known to be in between 0 and signed maximum inclusive, try to
38002         expand both unsigned and signed divmod and use the cheaper one from
38003         those.
38005 2017-02-22  Jeff Law  <law@redhat.com>
38007         PR tree-optimization/79578
38008         * tree-ssa-dse.c (clear_bytes_written_by): Use operand_equal_p
38009         to compare base operands.
38011 2017-02-22  Segher Boessenkool  <segher@kernel.crashing.org>
38013         PR target/79211
38014         * config/rs6000/rs6000.md (*fsel<SFDF:mode><SFDF2:mode>4): Use
38015         gpc_reg_operand instead of fpr_reg_operand.
38017 2017-02-22  Sameera Deshpande  <sameera.deshpande@imgtec.com>
38019         * config/mips/mips.c (mips_return_in_memory): Force FP
38020         vector types to be returned in memory for o32 ABI.
38022 2017-02-22  Jakub Jelinek  <jakub@redhat.com>
38024         * dwarf2out.c (gen_variable_die): For -gdwarf-5, use DW_TAG_variable
38025         instead of DW_TAG_member for static data member declarations and don't
38026         set no_linkage_name for static inline data members.
38027         (gen_member_die): For -gdwarf-5 don't change DW_TAG_variable
38028         to DW_TAG_member.
38030 2017-02-22  Martin Liska  <mliska@suse.cz>
38032         * doc/invoke.texi: Replace inequality signs with square brackets
38033         for -Wnormalized.
38035 2017-02-22  Matthew Fortune  <matthew.fortune@imgtec.com>
38037         PR target/78660
38038         * lra-constraints.c (simplify_operand_subreg): Handle
38039         WORD_REGISTER_OPERATIONS targets.
38041 2017-02-22  Jakub Jelinek  <jakub@redhat.com>
38043         PR target/70465
38044         * reg-stack.c (emit_swap_insn): Treat (float_extend:?F (mem:?F))
38045         and (const_double:?F) like (mem:?F) for the purpose of fxch %st(1)
38046         elimination by swapping fld*.
38048 2017-02-22  Richard Biener  <rguenther@suse.de>
38050         PR tree-optimization/79673
38051         * tree-ssa-pre.c (compute_avail): Use wide_int_to_tree to
38052         convert the [TARGET_]MEM_REF offset INTEGER_CST, scrapping off
38053         irrelevant address-space qualifiers and avoiding a
38054         ADDR_SPACE_CONVERT_EXPR from fold_convert.
38056 2017-02-22  Richard Biener  <rguenther@suse.de>
38058         PR tree-optimization/79666
38059         * tree-vrp.c (extract_range_from_binary_expr_1): Make sure
38060         to not symbolically negate if that may introduce undefined
38061         overflow.
38063 2017-02-22  Martin Liska  <mliska@suse.cz>
38065         PR lto/79587
38066         * data-streamer-in.c (streamer_read_gcov_count): Remove assert.
38067         * data-streamer-out.c (streamer_write_gcov_count_stream):
38068         Likewise.
38069         * value-prof.c (stream_out_histogram_value): Make assert more
38070         precise based on type of counter.
38072 2017-02-21  Uros Bizjak  <ubizjak@gmail.com>
38074         PR target/79593
38075         * config/i386/i386.md (standard_x87sse_constant_load splitter):
38076         Use nonimmediate_operand instead of memory_operand for operand 1.
38077         (float-extend standard_x87sse_constant_load splitter): Ditto.
38079 2017-02-21  Jeff Law  <law@redhat.com>
38081         PR tree-optimization/79621
38082         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior): Ignore
38083         blocks with edges to themselves.
38085 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
38087         PR target/79633
38088         * tree-chkp-opt.c (chkp_optimize_string_function_calls): Use
38089         is_gimple_call instead of comparing gimple_code with GIMPLE_CALL.
38090         Use gimple_call_builtin_p.
38092         PR target/79570
38093         * sel-sched.c (moveup_expr_cached): Don't call sel_bb_head
38094         on temporarily removed DEBUG_INSNs.
38096         PR tree-optimization/79649
38097         * tree-loop-distribution.c (classify_partition): Give up on
38098         non-generic address space loads/stores.
38100 2017-02-21  Aldy Hernandez  <aldyh@redhat.com>
38102         * doc/loop.texi (Loop manipulation): Remove nonexistent
38103         tree_ssa_loop_version from the documentation.
38104         * cfgloopmanip.c (loop_version): Document CONDITION_BB argument.
38106 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
38108         PR target/79494
38109         * config/i386/i386.c (ix86_expand_split_stack_prologue): Call
38110         make_reg_eh_region_note_nothrow_nononlocal on call_insn.
38111         * config/rs6000/rs6000.c: Include except.h.
38112         (rs6000_expand_split_stack_prologue): Call
38113         make_reg_eh_region_note_nothrow_nononlocal on the call insn.
38115 2017-02-21  Martin Jambor  <mjambor@suse.cz>
38117         PR lto/79579
38118         * ipa-prop.c (ipa_prop_write_jump_functions): Bail out if no edges
38119         have been analyzed.
38121 2017-02-21  Martin Jambor  <mjambor@suse.cz>
38123         * common.opt (-fipa-cp-alignment): Mark as ignored and preserved
38124         for backward compatibility only.
38125         * doc/invoke.texi (Option Summary): Remove all references to
38126         -fipa-cp-alignment.
38128 2017-02-21  Matthew Fortune  <matthew.fortune@imgtec.com>
38130         PR target/78660
38131         Revert:
38132         2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38134         * lra-constraints.c (curr_insn_transform): Handle
38135         WORD_REGISTER_OPERATIONS requirements when reloading SUBREGs.
38137 2017-02-21  Martin Liska  <mliska@suse.cz>
38139         * config/i386/i386.opt: Replace -masm-dialect with -masm.
38141 2017-02-21  Thomas Schwinge  <thomas@codesourcery.com>
38143         PR translation/79638
38144         * config/nvptx/nvptx.c (ENTRY_TEMPLATE): Single out "%ntid.y".
38146 2017-02-21  Eric Botcazou  <ebotcazou@adacore.com>
38148         PR ada/67205
38149         * config/arm/arm.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): Define.
38150         (arm_function_ok_for_sibcall): Return false for an indirect call by
38151         descriptor if all the argument registers are used.
38152         (arm_relayout_function): Use FUNCTION_ALIGNMENT macro to adjust the
38153         alignment of the function.
38155 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
38157         PR tree-optimization/61441
38158         * simplify-rtx.c (simplify_const_unary_operation): For
38159         -fsignaling-nans and sNaN operand, return NULL_RTX rather than
38160         the sNaN unmodified.
38162 2017-02-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
38164         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
38165         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
38166         instead of SYSTEM_HEADER_DIR.
38168 2017-02-20  Gerald Pfeifer  <gerald@pfeifer.com>
38169             Martin LiÅ¡ka  <mliska@suse.cz>
38171         * doc/invoke.texi (use-after-scope-direct-emission-threshold):
38172         Fix typos and grammar, use active voice, and clarify.
38174 2017-02-20  Marek Polacek  <polacek@redhat.com>
38176         PR middle-end/79537
38177         * gimplify.c (gimplify_expr): Handle unused *&&L;.
38179         PR sanitizer/79558
38180         * ubsan.c (ubsan_type_descriptor): Check if TYPE_MAX_VALUE is null.
38182 2017-02-20  Jakub Jelinek  <jakub@redhat.com>
38184         PR target/79568
38185         * config/i386/i386.c (ix86_expand_builtin): Handle
38186         OPTION_MASK_ISA_AVX512VL and OPTION_MASK_ISA_64BIT in
38187         ix86_builtins_isa[fcode].isa as a requirement of those
38188         flags and any other flag in the bitmask.
38189         (ix86_init_mmx_sse_builtins): Use 0 instead of
38190         ~OPTION_MASK_ISA_64BIT as mask.
38191         * config/i386/i386-builtin.def (__builtin_ia32_rdtsc,
38192         __builtin_ia32_rdtscp, __builtin_ia32_pause, __builtin_ia32_bsrsi,
38193         __builtin_ia32_rdpmc, __builtin_ia32_rolqi, __builtin_ia32_rolhi,
38194         __builtin_ia32_rorqi, __builtin_ia32_rorhi): Likewise.
38196 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38198         PR target/78012
38199         * lra-constraints.c (split_reg): Check requested split mode
38200         is supported by the register.
38202 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38204         * lra-constraints.c (simplify_operand_subreg): Remove early
38205         return false.
38207 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38209         PR target/78660
38210         * lra-constraints.c (curr_insn_transform): Tighten condition
38211         for converting SUBREG reloads from OP_OUT to OP_INOUT.
38213 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38215         PR target/78660
38216         * lra-constraints.c (curr_insn_transform): Handle
38217         WORD_REGISTER_OPERATIONS requirements when reloading SUBREGs.
38219 2017-02-19  Uros Bizjak  <ubizjak@gmail.com>
38221         Revert:
38222         2016-05-30  Uros Bizjak  <ubizjak@gmail.com>
38224         * config/i386/sync.md (mfence_nosse): Use "lock orl $0, -4(%esp)".
38226 2017-02-19  Jonathan Wakely  <jwakely@redhat.com>
38228         PR c++/69523
38229         * doc/invoke.texi (C++ Dialect Options) [-Wliteral-suffix]: Update
38230         description.
38232 2017-02-19  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
38234         * gimple-pretty-print.c (dump_ternary_rhs): Adjust gimple dump format
38235         for FMA_EXPR.
38237 2017-02-18  Jakub Jelinek  <jakub@redhat.com>
38239         * final.c (last_columnnum, override_columnnum): New variables.
38240         (final_start_function): Set last_columnnum, pass it to begin_prologue
38241         hook and pass 0 to dwarf2out_begin_prologue.
38242         (final_scan_insn): Update override_columnnum.  Pass last_columnnum
38243         to source_line debug hook.
38244         (notice_source_line): Compute last_columnnum and for debug_column_info
38245         return true on column changes.
38246         * debug.h (struct gcc_debug_hooks): Add column argument to
38247         source_line and begin_prologue hooks.
38248         (debug_nothing_int_charstar_int_bool): Remove prototype.
38249         (debug_nothing_int_int_charstar,
38250         debug_nothing_int_int_charstar_int_bool): New prototypes.
38251         (dwarf2out_begin_prologue): Add column argument.
38252         * debug.c (do_nothing_debug_hooks): Adjust source_line and
38253         begin_prologue hooks.
38254         (debug_nothing_int_charstar_int_bool): Remove.
38255         (debug_nothing_int_int_charstar,
38256         debug_nothing_int_int_charstar_int_bool): New functions.
38257         * dwarf2out.c (dwarf2out_begin_prologue): Add column argument, pass it
38258         through to dwarf2out_source_line.
38259         (dwarf2_lineno_debug_hooks): Adjust begin_prologue hook.
38260         (dwarf2out_source_line): Add column argument, emit it if requested.
38261         * sdbout.c (sdbout_source_line, sdbout_begin_prologue): Add column
38262         arguments.
38263         * xcoffout.h (xcoffout_begin_prologue, xcoffout_source_line): Likewise.
38264         * xcoffout.c (xcoffout_begin_prologue, xcoffout_source_line): Likewise.
38265         * vmsdbgout.c (vmsdbgout_begin_prologue): Add column argument, pass it
38266         through to dwarf2out_begin_prologue.
38267         (vmsdbgout_source_line): Add column argument, pass it through to
38268         dwarf2out_source_line.
38269         * dbxout.c (dbxout_begin_prologue): Add column argument, adjust
38270         dbxout_source_line caller.
38271         (dbxout_source_line): Add column argument.
38273         * common.opt (gno-column-info, gcolumn-info): New options.
38274         * dwarf2out.c (dwarf2_lineno_debug_hooks): Formatting fix.
38275         (check_die): Also test for multiple DW_AT_decl_column attributes.
38276         (add_src_coords_attributes, dwarf2out_imported_module_or_decl_1): Add
38277         DW_AT_decl_column if requested.
38278         (gen_subprogram_die): Compare and/or add also DW_AT_decl_column
38279         if requested.
38280         (gen_variable_die): Likewise.
38281         (add_call_src_coords_attributes): Add DW_AT_call_column if requested.
38282         * doc/invoke.texi (-gcolumn-info, -gno-column-info): Document.
38284         PR target/79569
38285         * config/i386/i386.opt (m3dnowa): Replace Undocumented with Report.
38286         * common/config/i386/i386-common.c (OPTION_MASK_ISA_3DNOW_A_SET): Define.
38287         (ix86_handle_option): Handle OPT_m3dnowa.
38288         * doc/invoke.texi (-m3dnowa): Document.
38289         * doc/extend.texi (__builtin_ia32_pmulhuw, __builtin_ia32_pf2iw): Use
38290         -m3dnowa instead of -m3dnow -march=athlon.
38292         PR target/79559
38293         * config/i386/i386.c (ix86_print_operand): Use output_operand_lossage
38294         instead of gcc_assert for K, r and R code checks.  Formatting fixes.
38296 2017-02-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
38298         PR target/79261
38299         * config/rs6000/rs6000.c (rs6000_expand_ternop_builtin): Add
38300         support for CODE_FOR_vsx_xxpermdi_v2d[fi]_be.
38301         * config/rs6000/rs6000.md (reload_gpr_from_vsx<mode>): Call
38302         generator for vsx_xxpermdi_<mode>_be.
38303         * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Remove logic to
38304         force big-endian semantics.
38305         (vsx_xxpermdi_<mode>_be): New define_expand with same
38306         implementation as previous version of vsx_xxpermdi_<mode>.
38308 2017-02-17  Jakub Jelinek  <jakub@redhat.com>
38310         PR tree-optimization/79327
38311         * gimple-ssa-sprintf.c (format_integer): Remove likely_adjust
38312         variable, its initialization and use.
38314 2017-02-17  Julia Koval  <julia.koval@intel.com>
38316         * common/config/i386/i386-common.c (OPTION_MASK_ISA_RDPID_SET): New.
38317         (OPTION_MASK_ISA_PKU_UNSET): New.
38318         (ix86_handle_option): Handle -mrdpid.
38319         * config/i386/cpuid.h (bit_RDPID): New.
38320         * config/i386/driver-i386.c (host_detect_local_cpu):
38321         Detect RDPID feature.
38322         * config/i386/i386-builtin.def (__builtin_ia32_rdpid): New.
38323         * config/i386/i386-c.c (ix86_target_macros_internal):
38324         Handle RDPID flag.
38325         * config/i386/i386.c (ix86_target_string): Add -mrdpid to isa2_opts.
38326         (ix86_valid_target_attribute_inner_p): Add "rdpid".
38327         (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
38328         * config/i386/i386.h (TARGET_RDPID, TARGET_RDPID_P): New.
38329         * config/i386/i386.md (define_insn "rdpid"): New.
38330         * config/i386/i386.opt Add -mrdpid.
38331         * config/i386/immintrin.h (_rdpid_u32): New.
38333 2017-02-17  Vladimir Makarov  <vmakarov@redhat.com>
38335         PR rtl-optimization/79541
38336         * lra-constraints.c (curr_insn_transform): Remove wrong asm insn
38337         instead of transforming it into USE.
38339 2017-02-17  Segher Boessenkool  <segher@kernel.crashing.org>
38341         * config/rs6000/rs6000.md (extendsfdf2): Remove default arguments.
38342         If HONOR_SNANS (SFmode) force the input to a register.
38343         (*extendsfdf2_fpr): Add !HONOR_SNANS (SFmode) condition.
38344         (*extendsfdf2_snan): New pattern, used when using SNaNs; it generates
38345         an frsp or similar insn.
38347 2017-02-17  Martin Liska  <mliska@suse.cz>
38349         PR rtl-optimization/79577
38350         * params.def (selsched-max-sched-times): Increase minimum to 1.
38352 2017-02-17  Martin Liska  <mliska@suse.cz>
38354         PR rtl-optimization/79574
38355         * gcse.c (want_to_gcse_p): Prevent integer overflow.
38357 2017-02-17  Martin Liska  <mliska@suse.cz>
38359         PR tree-optimization/79529
38360         * tree-ssa-loop-unswitch.c (is_maybe_undefined): Use
38361         ssa_defined_default_def_p to handle cases which are implicitly
38362         defined.
38363         * tree-ssa.c (ssa_defined_default_def_p): New function.
38364         (ssa_undefined_value_p): Use ssa_defined_default_def_p to handle cases
38365         which are implicitly defined.
38366         * tree-ssa.h (ssa_defined_default_def_p): Declare.
38368 2017-02-17  Richard Biener  <rguenther@suse.de>
38370         PR middle-end/79576
38371         * params.def (max-ssa-name-query-depth): Limit to 10.
38373 2017-02-17  Richard Biener  <rguenther@suse.de>
38375         PR tree-optimization/79552
38376         * tree-ssa-structalias.c (visit_loadstore): Properly verify
38377         default defs.
38379 2017-02-17  Richard Biener  <rguenther@suse.de>
38381         PR bootstrap/79567
38382         * genmatch.c (output_line_directive): Handle DIR_SEPARATOR_2.
38384 2017-02-17  Marek Polacek  <polacek@redhat.com>
38386         PR middle-end/79536
38387         * fold-const.c (fold_negate_expr_1): Renamed from fold_negate_expr.
38388         (fold_negate_expr): New wrapper.
38390 2017-02-16  Sandra Loosemore  <sandra@codesourcery.com>
38392         * doc/invoke.texi (C++ Dialect Options) [-Wno-non-template-friend]: 
38393         Correct terminology and de-emphasize pre-standard behavior.
38395 2017-02-16  Alan Modra  <amodra@gmail.com>
38397         PR rtl-optimization/79286
38398         * ira.c (def_dominates_uses): New function.
38399         (update_equiv_regs): Don't create an equivalence for insns that
38400         may trap where the register def does not dominate the use.
38402 2017-02-16  Vladimir Makarov  <vmakarov@redhat.com>
38404         PR rtl-optimization/78127
38405         * lra.c (lra): Call lra_eliminate before finish the loop after
38406         lra_constraint.
38408 2017-02-16  Richard Biener  <rguenther@suse.de>
38410         * graphite.h: Do not include isl/isl_val_gmp.h, instead include
38411         isl/isl_val.h.
38412         * graphite-isl-ast-to-gimple.c (gmp_cst_to_tree): Remove.
38413         (gcc_expression_from_isl_expr_int): Use generic isl_val interface.
38414         * graphite-sese-to-poly.c: Do not include isl/isl_val_gmp.h.
38415         (isl_val_int_from_wi): New function.
38416         (extract_affine_gmp): Rename to ...
38417         (extract_affine_wi): ... this, take a widest_int.
38418         (extract_affine_int): Just wrap extract_affine_wi.
38419         (add_param_constraints): Use isl_val_int_from_wi.
38420         (add_loop_constraints): Likewise, and extract_affine_wi.
38422 2017-02-15  Jeff Law  <law@redhat.com>
38424         PR middle-end/79521
38425         * ira-costs.c (scan_one_insn): Check have_regs_of_mode before calling
38426         ira_init_register_move_cost_if_necessary.
38428 2017-02-15  Martin Sebor  <msebor@redhat.com>
38430         PR middle-end/32003
38431         * doc/invoke.texi (-fdump-final-insns): Replace option accidentally
38432         removed in a prior commit.
38434 2017-02-15  Bin Cheng  <bin.cheng@arm.com>
38436         PR tree-optimization/79347
38437         * tree-vect-loop-manip.c (vect_do_peeling): Maintain profile
38438         counters during peeling.
38440 2017-02-15  Thomas Schwinge  <thomas@codesourcery.com>
38442         * Makefile.in (site.exp): Remove "set ISLVER".
38444 2017-02-15  Jakub Jelinek  <jakub@redhat.com>
38446         PR target/79487
38447         * real.c (real_from_integer): Call real_convert even for decimal.
38449 2017-02-15  Dominik Vogt  <vogt@linux.vnet.ibm.com>
38451         PR target/79421
38452         * config/s390/s390.c: define TARGET_CUSTOM_FUNCTION_DESCRIPTORS.
38454 2017-02-14  Andrew Pinski  <apinski@cavium.com>
38456         * config/aarch64/aarch64-cores.def (thunderx2t99): Move to under 'C"
38457         cores and change the partno/implementer to be correct.
38458         (thunderx2t99p1): New core which replaces thunderx2t99 and still has
38459         the 'B" as the implementer.
38460         * config/aarch64/aarch64-tune.md: Regenerate.
38462 2017-02-14  Carl Love  <cel@us.ibm.com>
38464         * config/rs6000/rs6000.c: Add case statement entry to make the
38465         xvcvuxdsp built-in argument unsigned.
38466         * config/rs6000/vsx.md: Fix the source and return operand types so they
38467         match the instruction definitions from the ISA document.  Fix typo
38468         in the instruction generation for the (define_insn "vsx_xvcvuxdsp"
38469         statement.
38471 2017-02-14  Vladimir Makarov  <vmakarov@redhat.com>
38473         PR target/79282
38474         * lra-int.h (struct lra_operand_data, struct lra_insn_reg): Add
38475         member early_clobber_alts.
38476         * lra-lives.c (reg_early_clobber_p): New.
38477         (process_bb_lives): Use it.
38478         * lra.c (new_insn_reg): New arg early_clobber_alts.  Use it.
38479         (debug_operand_data): Initialize early_clobber_alts.
38480         (setup_operand_alternative): Set up early_clobber_alts.
38481         (collect_non_operand_hard_regs): Ditto.  Pass early clobber
38482         alternatives to new_insn_reg.
38483         (add_regs_to_insn_regno_info): Add arg early_clobber_alts.  Use
38484         it.
38485         (lra_update_insn_regno_info): Pass the new arg.
38487 2017-02-14  Jakub Jelinek  <jakub@redhat.com>
38489         PR middle-end/79505
38490         * omp-offload.c (free_oacc_loop): Release loop->ifns vector.
38491         (new_oacc_loop_raw): Don't clear already cleared fields.
38493         PR target/79481
38494         * config/i386/avx512pfintrin.h (_mm512_prefetch_i32gather_pd,
38495         _mm512_prefetch_i32gather_ps, _mm512_prefetch_i64gather_pd,
38496         _mm512_prefetch_i64gather_ps): New inline functions and macros.
38498 2017-02-14  Uros Bizjak  <ubizjak@gmail.com>
38500         PR target/79495
38501         * config/i386/i386.md (*movxf_internal): Add (o,rC) alternative.
38503 2017-02-14  H.J. Lu  <hongjiu.lu@intel.com>
38505         PR target/79498
38506         * config/i386/i386.c (timode_scalar_chain::convert_insn): Insert
38507         the extra instruction to the right place to store 128-bit constant
38508         when needed.
38510 2017-02-14  Martin Sebor  <msebor@redhat.com>
38512         PR middle-end/79448
38513         * gimple-ssa-sprintf.c (format_directive): Avoid issuing INT_MAX
38514           warning for strings of unknown length.
38516 2017-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
38518         * config.gcc (supported_defaults) [powerpc*-*-*]: Update.
38520 2017-02-14  Jeff Law  <law@redhat.com>
38522         PR target/79404
38523         * ira-costs.c (scan_one_insn): Initialize register move costs
38524         for pseudos seen in USE/CLOBBER insns.
38526         PR tree-optimization/79095
38527         * tree-vrp.c (extract_range_from_binary_expr_1): For EXACT_DIV_EXPR,
38528         if the numerator has the range ~[0,0] make the resultant range ~[0,0].
38529         (extract_range_from_binary_expr): For MINUS_EXPR with no derived range,
38530         if the operands are known to be not equal, then the resulting range
38531         is ~[0,0].
38532         (intersect_ranges): If the new range is ~[0,0] and the old range is
38533         wide, then prefer ~[0,0].
38534         * tree-vrp.c (overflow_comparison_p_1): New function.
38535         (overflow_comparison_p): New function.
38536         * tree-vrp.c (register_edge_assert_for_2): Register additional asserts
38537         if NAME is used in an overflow test.
38538         (vrp_evaluate_conditional_warnv_with_ops): If the ops represent an
38539         overflow check that can be expressed as an equality test, then adjust
38540         ops to be that equality test.
38542 2017-02-14  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
38544         * config/s390/s390-builtin-types.def: Remove flags argument.
38545         * config/s390/s390.c (s390_init_builtins): Likewise.
38547 2017-02-14  Martin Liska  <mliska@suse.cz>
38549         * tree-ssa-loop-unswitch.c (hoist_guard): Release get_loop_body
38550         vector.  Fix trailing white spaces.
38552 2017-02-14  James Greenhalgh  <james.greenhalgh@arm.com>
38554         * config/aarch64/aarch64.c (aarch64_simd_container_mode): Handle
38555         HFmode.
38557 2017-02-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38559         PR rtl-optimization/68664
38560         * config/arm/arm.c (arm_sched_can_speculate_insn):
38561         New function.  Declare prototype.
38562         (TARGET_SCHED_CAN_SPECULATE_INSN): Define.
38564 2017-02-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38566         PR rtl-optimization/68664
38567         * config/aarch64/aarch64.c (aarch64_sched_can_speculate_insn):
38568         New function.
38569         (TARGET_SCHED_CAN_SPECULATE_INSN): Define.
38571 2017-02-14  Amit Pawar  <amit.pawar@amd.com>
38573         * config/i386/i386.c (znver1_cost): Fix the alignment for function and
38574         max skip bytes for function, loop and jump.
38576 2017-02-14  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
38578         * gimple-pretty-print.c (dump_unary_rhs): Adjust dump format for
38579         ABS_EXPR for gimple dump.
38581 2017-02-14  Jakub Jelinek  <jakub@redhat.com>
38583         PR target/79462
38584         * config/sh/sh.c (expand_cbranchdi4): Don't clear operands[4].
38586         PR tree-optimization/79408
38587         * tree-vrp.c (simplify_div_or_mod_using_ranges): Handle also the
38588         case when on TRUNC_MOD_EXPR op0 is INTEGER_CST.
38589         (simplify_stmt_using_ranges): Call simplify_div_or_mod_using_ranges
38590         also if rhs1 is INTEGER_CST.
38592 2017-02-14  Richard Biener  <rguenther@suse.de>
38594         PR middle-end/79432
38595         * tree-into-ssa.c (insert_phi_nodes): When the function can
38596         have abnormal edges rewrite SSA names with broken use-def
38597         dominance out of SSA and register them for PHI insertion.
38599 2017-02-13  Martin Sebor  <msebor@redhat.com>
38601         PR middle-end/79496
38602         * gimple-ssa-sprintf.c (pass_sprintf_length::handle_gimple_call): Avoid
38603         clearing info.nowrite flag when snprintf size argument is a range.
38605 2017-02-13  Jakub Jelinek  <jakub@redhat.com>
38607         * cprop.c (cprop_jump): Add missing space in string literal.
38608         * tree-ssa-structalias.c (rewrite_constraints): Likewise.
38609         (get_constraint_for_component_ref): Likewise.
38610         * df-core.c (df_worklist_dataflow_doublequeue): Likewise.
38611         * tree-outof-ssa.c (insert_partition_copy_on_edge): Likewise.
38612         * lra-constraints.c (process_alt_operands): Likewise.
38613         * ipa-inline.c (inline_small_functions): Likewise.
38614         * tree-ssa-sccvn.c (visit_reference_op_store): Likewise.
38615         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
38616         * trans-mem.c (diagnose_tm_1_op): Likewise.
38617         * omp-grid.c (grid_find_single_omp_among_assignments): Likewise.
38618         (grid_parallel_clauses_gridifiable): Likewise.
38620         * config/nvptx/mkoffload.c (process): Add space in between
38621         , and %d.
38623         * config/i386/i386.h (REG_CLASS_NAMES): Add , in between
38624         "MOD4_SSE_REGS" and "ALL_REGS".
38626         * spellcheck.c (test_data): Add , in between "foo" and "food".
38628 2017-02-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
38630         PR target/79449
38631         * config/rs6000/rs6000.c (expand_block_compare): Make sure runtime
38632         boundary crossing check and subsequent code generation agree.
38634 2017-02-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38636         * config/aarch64/aarch64.c (has_memory_op): Delete.
38637         (aarch64_madd_needs_nop): Use contains_mem_rtx_p instead of
38638         has_memory_op.
38640 2017-02-13  Jakub Jelinek  <jakub@redhat.com>
38642         PR rtl-optimization/79388
38643         PR rtl-optimization/79450
38644         * combine.c (distribute_notes): When removing TEM_INSN for which
38645         corresponding dest has last value recorded, invalidate that last
38646         value.
38648 2017-02-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38650         * config/arm/arm.c (arm_print_tune_info): Use ASM_COMMENT_START instead
38651         of explicit '@'.  Add missing assembly comment marker on branch costs
38652         printout.
38654 2017-02-13  Nathan Sidwell  <nathan@acm.org>
38656         * gengtype-lex.l (<in_struct>): Add '/'.
38658 2017-02-13  Martin Liska  <mliska@suse.cz>
38660         PR c/79471
38661         * calls.c (expand_call): Replace XALLOCAVEC with XCNEWVEC.
38663 2017-02-13  Richard Biener  <rguenther@suse.de>
38665         * configure.ac (HAVE_ISL_OPTIONS_SET_SCHEDULE_SERIALIZE_SCCS):
38666         Remove.
38667         * configure: Re-generate.
38668         * config.in: Likewise.
38669         * graphite-dependences.c: Simplify as if
38670         HAVE_ISL_OPTIONS_SET_SCHEDULE_SERIALIZE_SCCS was defined.
38671         * graphite-isl-ast-to-gimple.c: Likewise.
38672         * graphite-optimize-isl.c: Likewise.
38673         * graphite-poly.c: Likewise.
38674         * graphite-sese-to-poly.c: Likewise.
38675         * graphite.h: Likewise.
38676         * toplev.c: Include isl/version.h and use isl_version () for
38677         printing the ISL version.
38678         * doc/install.texi: Update ISL requirement.
38680 2017-02-12  Gerald Pfeifer  <gerald@pfeifer.com>
38682         * doc/standards.texi (Standards): Update reference to
38683         Objective-C 2.0.
38685 2017-02-12  Gerald Pfeifer  <gerald@pfeifer.com>
38687         * doc/extend.texi (Named Address Spaces): sourceware.org now
38688         defaults to https.
38689         * doc/install.texi (Binaries): Ditto.
38690         (Specific): Ditto.
38692 2017-02-11  Sandra Loosemore  <sandra@codesourcery.com>
38694         * doc/cpp.texi: Replace "stringify"/"stringification" with C
38695         standard terminology "stringize"/"stringizing" throughout.
38696         * doc/cppinternals.texi: Likewise.
38698 2017-02-11  Sandra Loosemore  <sandra@codesourcery.com>
38700         * doc/extend.texi: Fix some spelling mistakes and typos.
38701         * doc/invoke.texi: Likewise.
38703 2017-02-11  Jan Hubicka  <hubicka@ucw.cz>
38705         PR ipa/79224
38706         * params.def (inline-min-speedup) Change from 10 to 8.
38708 2017-02-11  Jakub Jelinek  <jakub@redhat.com>
38710         * doc/invoke.texi (fopenmp): Bump OpenMP version from 4.0 to
38711         4.5.
38713 2017-02-11  Jan Hubicka  <hubicka@ucw.cz>
38715         PR ipa/79224
38716         * ipa-inline-analysis.c (get_minimal_bb): New function.
38717         (record_modified): Use it.
38718         (remap_edge_change_prob): Handle also ancestor functions.
38720 2017-02-11  Gerald Pfeifer  <gerald@pfeifer.com>
38722         * doc/contrib.texi (Contributors): Remove broken link into
38723         the Mauve CVS repository.
38725 2017-02-11  Jakub Jelinek  <jakub@redhat.com>
38727         PR middle-end/79454
38728         * internal-fn.c (expand_vector_ubsan_overflow): Use piece-wise
38729         result computation whenever lhs doesn't have vector mode, not
38730         just when it has BLKmode.
38732 2017-02-10  Gerald Pfeifer  <gerald@pfeifer.com>
38734         * doc/makefile.texi (profiledbootstrap): Refer to the
38735         installation instructions only in textual form.
38737 2017-02-10  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
38739         PR target/79295
38740         * config/rs6000/altivec.md (bcd<bcd_add_sub>): Fix constraints.
38742 2017-02-10  Gerald Pfeifer  <gerald@pfeifer.com>
38744         * doc/install.texi (Specific): Use https for blackfin.uclinux.org.
38745         (Specific): Update mingw-w64 reference.
38746         (Binaries): Ditto.
38747         (Specific): Remove broken link to Renesas RX processor.
38749 2017-02-10  Richard Biener  <rguenther@suse.de>
38751         * toplev.c (process_options): Do not mention obsolete graphite
38752         options when printing sorry message about missing graphite support.
38753         Mention -floop-nest-optimize.
38755 2017-02-10  Christophe Lyon  <christophe.lyon@linaro.org>
38757         * config/aarch64/arm_neon.h (vtst_p8): Rewrite without asm.
38758         (vtst_p16): Likewise.
38759         (vtstq_p8): Likewise.
38760         (vtstq_p16): Likewise.
38761         (vtst_p64): New.
38762         (vtstq_p64): Likewise.
38763         * config/arm/arm_neon.h (vgetq_lane_p64): New.
38764         (vset_lane_p64): New.
38765         (vsetq_lane_p64): New.
38767 2017-02-10  Jakub Jelinek  <jakub@redhat.com>
38769         PR tree-optimization/79411
38770         * tree-ssa-reassoc.c (is_reassociable_op): Return false if
38771         stmt operands are SSA_NAMEs used in abnormal phis.
38772         (can_reassociate_p): Return false if op is SSA_NAME used in abnormal
38773         phis.
38775 2017-02-09  Jan Hubicka  <hubicka@ucw.cz>
38777         PR ipa/70795
38778         * cgraphunit.c (cgraph_node::add_new_function): Set externally_visible
38779         flag if needed.
38781 2017-02-09  Jan Hubicka  <hubicka@ucw.cz>
38783         * tree-ssa-loop-unswitch.c (hoist_guard): Update profile.
38785 2017-02-09  Jakub Jelinek  <jakub@redhat.com>
38787         * omp-offload.c (oacc_loop_auto_partitions): Use || instead of |
38788         to avoid warning.
38790         PR c/79413
38791         * gimplify.h (is_gimple_sizepos): Only test for INTEGER_CST constants,
38792         not arbitrary TREE_CONSTANT.
38794         PR c/79431
38795         * gimplify.c (gimplify_adjust_omp_clauses): Ignore
38796         "omp declare target link" attribute unless is_global_var.
38797         * omp-offload.c (find_link_var_op): Likewise.
38799 2017-02-09  Nathan Sidwell  <nathan@codesourcery.com>
38800             Chung-Lin Tang  <cltang@codesourcery.com>
38802         * gimplify.c (gimplify_scan_omp_clauses): No special handling for
38803         OMP_CLAUSE_TILE.
38804         (gimplify_adjust_omp_clauses): Don't delete TILE.
38805         (gimplify_omp_for): Deal with TILE.
38806         * internal-fn.c (expand_GOACC_TILE): New function.
38807         * internal-fn.def (GOACC_DIM_POS): Comment may be overly conservative.
38808         (GOACC_TILE): New.
38809         * omp-expand.c (struct oacc_collapse): Add tile and outer fields.
38810         (expand_oacc_collapse_init): Add LOC paramter.  Initialize tile
38811         element fields.
38812         (expand_oacc_collapse_vars): Add INNER parm, adjust for tiling,
38813         avoid DIV for outermost collapse var.
38814         (expand_oacc_for): Insert tile element loop as needed.  Adjust.
38815         Remove out of date comments, fix whitespace.
38816         * omp-general.c (omp_extract_for_data): Deal with tiling.
38817         * omp-general.h (enum oacc_loop_flags): Add OLF_TILE flag,
38818         adjust OLF_DIM_BASE value.
38819         (struct omp_for_data): Add tiling field.
38820         * omp-low.c (scan_sharing_clauses): Allow OMP_CLAUSE_TILE.
38821         (lower_oacc_head_mark): Add OLF_TILE as appropriate.  Ensure 2 levels
38822         for auto loops.  Remove default auto determining, moved to
38823         oacc_loop_fixed_partitions.
38824         * omp-offload.c (struct oacc_loop): Change 'ifns' to vector of call
38825         stmts, add e_mask field.
38826         (oacc_dim_call): New function, abstracted out from oacc_thread_numbers.
38827         (oacc_thread_numbers): Use oacc_dim_call.
38828         (oacc_xform_tile): New.
38829         (new_oacc_loop_raw): Initialize e_mask, adjust for ifns vector.
38830         (finish_oacc_loop): Adjust for ifns vector.
38831         (oacc_loop_discover_walk): Append loop abstraction sites to list,
38832         add case for GOACC_TILE fns.
38833         (oacc_loop_xform_loop): Delete.
38834         (oacc_loop_process): Iterate over call list directly, and add
38835         handling for GOACC_TILE fns.
38836         (oacc_loop_fixed_partitions): Determine default auto, deal with TILE,
38837         dump partitioning.
38838         (oacc_loop_auto_partitions): Add outer_assign parm. Assign all but
38839         vector partitioning to outer loops.  Assign 2 partitions to loops
38840         when available. Add TILE handling.
38841         (oacc_loop_partition): Adjust oacc_loop_auto_partitions call.
38842         (execite_oacc_device_lower): Process GOACC_TILE fns,
38843         ignore unknown specs.
38844         * tree-nested.c (convert_nonlocal_omp_clauses): Allow OMP_CLAUSE_TILE.
38845         * tree.c (omp_clause_num_ops): Adjust TILE ops.
38846         * tree.h (OMP_CLAUSE_TILE_ITERVAR, OMP_CLAUSE_TILE_COUNT): New.
38848 2017-02-09  Gerald Pfeifer  <gerald@pfeifer.com>
38850         * configure.ac (ACX_BUGURL): Update.
38851         * configure: Regenerate.
38853 2017-02-09  Richard Biener  <rguenther@suse.de>
38855         PR tree-optimization/69823
38856         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
38857         Properly enumerate all BBs in the region.  Use auto_vec/auto_bitmap.
38859 2017-02-09  Andrew Burgess  <andrew.burgess@embecosm.com>
38861         * config/arc/arc-c.def: Add __NPS400__ definition.
38862         * config/arc/arc.h (CPP_SPEC): Don't define __NPS400__ here.
38863         (TARGET_NPS400): Define.
38865 2017-02-09  Andrew Burgess  <andrew.burgess@embecosm.com>
38867         * config/arc/arc-arch.h (arc_arch_t): Move unchanged to earlier in
38868         file.
38869         (arc_cpu_t): Change base_architecture field, arch, to a arc_arc_t
38870         pointer, arch_info.
38871         (arc_cpu_types): Fill the arch_info field with a pointer into the
38872         arc_arch_types table.
38873         (arc_selected_cpu): Declare.
38874         * config/arc/arc.c (arc_selected_cpu): Make global.
38875         (arc_selected_arch): Delete.
38876         (arc_base_cpu): Delete.
38877         (arc_override_options): Remove references to deleted variables,
38878         update access to arch information.
38879         (ARC_OPT): Update access to arch information.
38880         (ARC_OPTX): Likewise.
38881         * config/arc/arc.h (arc_base_cpu): Remove declaration.
38882         (TARGET_ARC600): Update access to arch information.
38883         (TARGET_ARC601): Likewise.
38884         (TARGET_ARC700): Likewise.
38885         (TARGET_EM): Likewise.
38886         (TARGET_HS): Likewise.
38887         * config/arc/driver-arc.c (arc_cpu_to_as): Update access to arch
38888         information.
38890 2017-02-08  Pat Haugen  <pthaugen@us.ibm.com>
38892         PR target/78604
38893         * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Invert
38894         condition/operands for integer GE/LE/GEU/LEU operations.
38896 2017-02-08  Segher Boessenkool  <segher@kernel.crashing.org>
38898         PR translation/79397
38899         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Fix spelling
38900         of AltiVec.
38902 2017-02-08  Martin Jambor  <mjambor@suse.cz>
38904         PR ipa/79375
38905         * ipa-prop.c (ipa_alloc_node_params): Make static, return bool
38906         whether allocation happened.
38907         (ipa_initialize_node_params): Do not call ipa_alloc_node_params if
38908         nothing was allocated.
38910 2017-02-08  Jakub Jelinek  <jakub@redhat.com>
38912         PR tree-optimization/79408
38913         * tree-vrp.c (simplify_div_or_mod_using_ranges): If op1 is not
38914         constant, but SSA_NAME with a known integer range, use the minimum
38915         of that range instead of op1 to determine if modulo can be replaced
38916         with its first operand.
38918 2016-02-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38920         * config/riscv/riscv.c (riscv_build_integer_1): Avoid use of INT16_MAX.
38922 2017-02-08  Richard Biener  <rguenther@suse.de>
38924         PR tree-optimization/71824
38925         * graphite-scop-detection.c (scop_detection::build_scop_breadth):
38926         Check all loops contained in the merged region.
38928 2017-02-07  Andrew Pinski  <apinski@cavium.com>
38930         * config/aarch64/aarch64.md (popcount<mode>2): New pattern.
38932 2017-02-07  Andrew Pinski  <apinski@cavium.com>
38934         * config/aarch64/aarch64-cores.def (thunderx): Disable LSE.
38935         (thunderxt88): Likewise.
38936         (thunderxt81): Disable LSE and change v8.1 to v8.
38937         (thunderxt83): Likewise.
38939 2017-02-07  Jakub Jelinek  <jakub@redhat.com>
38940             Richard Biener  <rguenther@suse.de>
38942         PR middle-end/79399
38943         * ira-int.h (struct target_ira_int): Change x_max_struct_costs_size
38944         type from int to size_t.
38945         * ira-costs.c (struct_costs_size): Change type from int to size_t.
38947 2017-02-07  Jakub Jelinek  <jakub@redhat.com>
38949         PR rtl-optimization/79386
38950         * cprop.c (bypass_conditional_jumps): Initialize
38951         bypass_last_basic_block already before splitting bbs after
38952         unconditional traps...
38953         (bypass_conditional_jumps): ... rather than here.
38955         PR target/79299
38956         * config/i386/sse.md (xtg_mode, gatherq_mode): New mode attrs.
38957         (*avx512f_gathersi<mode>, *avx512f_gathersi<mode>_2,
38958         *avx512f_gatherdi<mode>, *avx512f_gatherdi<mode>_2): Use them,
38959         fix -masm=intel patterns.
38961 2017-02-07  Richard Biener  <rguenther@suse.de>
38963         PR tree-optimization/79256
38964         PR middle-end/79278
38965         * builtins.c (get_object_alignment_2): Use min_align_of_type
38966         to extract alignment for MEM_REFs to honor BIGGEST_FIELD_ALIGNMENT
38967         and ADJUST_FIELD_ALIGN.
38969         * doc/tm.texi.in (ADJUST_FIELD_ALIGN): Adjust to take additional
38970         type parameter.
38971         * doc/tm.texi: Regenerate.
38972         * stor-layout.c (layout_decl): Adjust.
38973         (update_alignment_for_field): Likewise.
38974         (place_field): Likewise.
38975         (min_align_of_type): Likewise.
38976         * config/arc/arc.h (ADJUST_FIELD_ALIGN): Adjust.
38977         * config/epiphany/epiphany.h (ADJUST_FIELD_ALIGN): Likewise.
38978         * config/epiphany/epiphany.c (epiphany_adjust_field_align): Likewise.
38979         * config/frv/frv.h (ADJUST_FIELD_ALIGN): Likewise.
38980         * config/frv/frv.c (frv_adjust_field_align): Likewise.
38981         * config/i386/i386.h (ADJUST_FIELD_ALIGN): Likewise.
38982         * config/i386/i386.c (x86_field_alignment): Likewise.
38983         * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Likewise.
38984         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Likewise.
38985         * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Likewise.
38986         * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
38987         * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Likewise.
38988         * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
38989          Likewise.
38991         Revert
38992         2017-01-30  Richard Biener  <rguenther@suse.de>
38994         PR tree-optimization/79256
38995         * targhooks.c (default_builtin_vector_alignment_reachable): Honor
38996         BIGGEST_FIELD_ALIGNMENT and ADJUST_FIELD_ALIGN to fix up bogus
38997         alignment on TYPE.
38999 2017-02-07  Toma Tabacu  <toma.tabacu@imgtec.com>
39001         * config/mips/mips.c (mips_expand_builtin_insn): Convert the QImode
39002         argument of the pshufh, psllh, psllw, psrah, psraw, psrlh, psrlw
39003         builtins to SImode and emit a zero-extend, if necessary.
39005 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
39007         * docs/invoke.texi (RISC-V Options): Alphabetize.
39009 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
39011         * doc/invoke.texi (RISC-V Options): Use two spaces to separate
39012         options.
39014 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
39016         * config/riscv/riscv.c: New file.
39017         * common/config/riscv/riscv-common.c: Likewise.
39018         * config.gcc: Likewise.
39019         * config/riscv/constraints.md: Likewise.
39020         * config/riscv/elf.h: Likewise.
39021         * config/riscv/generic.md: Likewise.
39022         * config/riscv/linux.h: Likewise.
39023         * config/riscv/multilib-generator: Likewise.
39024         * config/riscv/peephole.md: Likewise.
39025         * config/riscv/pic.md: Likewise.
39026         * config/riscv/predicates.md: Likewise.
39027         * config/riscv/riscv-builtins.c: Likewise.
39028         * config/riscv/riscv-c.c: Likewise.
39029         * config/riscv/riscv-ftypes.def: Likewise.
39030         * config/riscv/riscv-modes.def: Likewise.
39031         * config/riscv/riscv-opts.h: Likewise.
39032         * config/riscv/riscv-protos.h: Likewise.
39033         * config/riscv/riscv.h: Likewise.
39034         * config/riscv/riscv.md: Likewise.
39035         * config/riscv/riscv.opt: Likewise.
39036         * config/riscv/sync.md: Likewise.
39037         * config/riscv/t-elf-multilib: Likewise.
39038         * config/riscv/t-linux: Likewise.
39039         * config/riscv/t-linux-multilib: Likewise.
39040         * config/riscv/t-riscv: Likewise.
39041         * configure.ac: Likewise.
39042         * doc/contrib.texi: Add Kito Cheng, Palmer Dabbelt, and Andrew
39043         Waterman as RISC-V maintainers.
39044         * doc/install.texi: Add RISC-V entries.
39045         * doc/invoke.texi: Add RISC-V options section.
39046         * doc/md.texi: Add RISC-V constraints section.
39047         * configure: Regenerated.
39049 2017-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
39051         PR target/66144
39052         * config/rs6000/vector.md (vcond<mode><mode>): Allow the true and
39053         false values to be constant vectors with all 0 or all 1 bits set.
39054         (vcondu<mode><mode>): Likewise.
39055         * config/rs6000/predicates.md (vector_int_reg_or_same_bit): New
39056         predicate.
39057         (fpmask_comparison_operator): Update comment.
39058         (vecint_comparison_operator): New predicate.
39059         * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Optimize
39060         vector conditionals when the true and false values are constant
39061         vectors with all 0 bits or all 1 bits set.
39063 2017-02-06  Martin Sebor  <msebor@redhat.com>
39065         PR  tree-optimization/79376
39066         * gimple-fold.c (get_range_strlen): Set the minimum length to zero.
39068 2017-02-06  Uros Bizjak  <ubizjak@gmail.com>
39070         * config/i386/sse.md (vector modes -> vec_extract* splitter): Use
39071         explicit subreg RTX with operand 1.  Use VECTOR_MODE_P predicate
39072         to simplify split condition.
39074 2017-02-06  Jakub Jelinek  <jakub@redhat.com>
39076         * omp-expand.c (oxpand_omp_atomic_fetch_op,
39077         expand_omp_atomic_pipeline): Return false if can_atomic_load_p is
39078         false.
39080 2017-02-06  Segher Boessenkool  <segher@kernel.crashing.org>
39082         PR rtl-optimization/68664
39083         * target.def (can_speculate_insn): New hook.
39084         * doc/tm.texi.in (TARGET_SCHED_CAN_SPECULATE_INSN): New hook.
39085         * doc/tm.texi: Regenerate.
39086         * sched-rgn.c (can_schedule_ready_p): Use the new hook.
39087         * config/rs6000/rs6000.c (TARGET_SCHED_CAN_SPECULATE_INSN): New macro.
39088         (rs6000_sched_can_speculate_insn): New function.
39090 2017-02-06  Jakub Jelinek  <jakub@redhat.com>
39092         PR tree-optimization/79284
39093         * tree-vectorizer.h (VECT_SCALAR_BOOLEAN_TYPE_P): Define.
39094         * tree-vect-stmts.c (vect_get_vec_def_for_operand,
39095         vectorizable_mask_load_store, vectorizable_operation,
39096         vect_is_simple_cond, get_same_sized_vectype): Use it instead
39097         of comparing TREE_CODE of a type against BOOLEAN_TYPE.
39098         * tree-vect-patterns.c (check_bool_pattern, search_type_for_mask_1,
39099         vect_recog_bool_pattern, vect_recog_mask_conversion_pattern): Likewise.
39100         * tree-vect-slp.c (vect_get_constant_vectors): Likewise.
39101         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
39102         Remove redundant gimple_code (stmt) == GIMPLE_ASSIGN test after
39103         is_gimple_assign (stmt).  Replace another such test with
39104         is_gimple_assign (stmt).
39106 2017-02-06  Georg-Johann Lay  <avr@gjlay.de>
39108         PR target/78883
39109         * config/avr/avr.c (rtl-iter.h): Include it.
39110         (TARGET_LEGITIMATE_COMBINED_INSN): New hook define...
39111         (avr_legitimate_combined_insn): ...and implementation.
39113 2017-02-06  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39115         * config/s390/predicates.md ("larl_operand"): Use macros from hwint.h.
39116         * config/s390/s390.c (s390_const_operand_ok)
39117         (s390_canonicalize_comparison, s390_extract_part)
39118         (s390_single_part, s390_contiguous_bitmask_nowrap_p)
39119         (s390_contiguous_bitmask_p, s390_rtx_costs)
39120         (legitimize_pic_address): Likewise.
39121         * config/s390/s390.md ("clzdi2", "clztidi2"): Likewise.
39122         * config/s390/vx-builtins.md ("vec_genbytemaskv16qi")
39123         ("vec_permi<mode>", "vfae<mode>", "*vfaes<mode>", "vstrc<mode>")
39124         ("*vstrcs<mode>"): Use UINTVAL() to set unsigned HOST_WIDE_INT.
39125         * config/s390/vector.md ("vec_vfenes<mode>"): Likewise.
39127 2017-02-06  Georg-Johann Lay  <avr@gjlay.de>
39129         * config/avr/avr.md (*addhi3_zero_extend): Add alternative where
39130         REGNO($0) == REGNO($1).
39132 2017-02-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39134         * config/s390/linux.h(SIZE_TYPE): Add comment.
39136 2017-02-06  Julian Brown  <julian@codesourcery.com>
39137             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
39138             Virendra Pathak  <virendra.pathak@broadcom.com>
39140         * config/aarch64/aarch64-cores.def: Change the scheduler
39141         to Thunderx2t99.
39142         * config/aarch64/aarch64.md: Include thunderx2t99.md.
39143         * config/aarch64/thunderx2t99.md: New file.
39145 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
39147         * doc/standards.texi (Go Language): Update link to language
39148         standard.
39150 2017-02-05  Jan Hubicka  <hubicka@ucw.cz>
39152         * tree-eh.c (lower_resx): Sanitize profile.
39153         (cleanup_empty_eh_move_lp): Likewise.
39155 2017-02-05  Jan Hubicka  <hubicka@ucw.cz>
39157         PR tree-ssa/79347
39158         * cfgloopmanip.c (lv_adjust_loop_entry_edge, loop_version): Add
39159         ELSE_PROB.
39160         * cfgloopmanip.h (loop_version): Update prototype.
39161         * modulo-sched.c (sms_schedule): Update call of loop_version.
39162         * tree-if-conv.c(version_loop_for_if_conversion): Likewise.
39163         * tree-parloops.c (gen_parallel_loop): Likewise.
39164         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Likewise.
39165         * tree-ssa-loop-split.c (split_loop): Likewise.
39166         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Likewise.
39167         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
39169 2017-02-05  Martin Liska  <mliska@suse.cz>
39171         PR bootstrap/78985
39172         * config/s390/s390.c (s390_gimplify_va_arg): Initialize local
39173         variable to NULL.
39174         (print_operand_address): Initialize a struct to zero.
39176 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
39178         * doc/contrib.texi (Contributors): Refer to Hans Boehm's
39179         garbage collector only in textual form.
39181 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
39183         * doc/extend.texi (x86 specific memory model extensions for
39184         transactional memory): Simplify a phrase.
39186 2017-02-05  Eric Botcazou  <ebotcazou@adacore.com>
39188         PR target/79353
39189         * config/sparc/sync.md (atomic_loaddi_1): Replace 'U' constraint with
39190         'r', 'm' constraint with 'T' and !TARGET_ARCH64 with TARGET_ARCH32.
39191         (atomic_storedi_1): Likewise.
39193 2017-02-04  Jakub Jelinek  <jakub@redhat.com>
39195         PR tree-optimization/79338
39196         * tree-parloops.c (gather_scalar_reductions): Don't call
39197         vect_analyze_loop_form for loop->inner before destroying loop's
39198         loop_vinfo.
39200 2017-02-03  Martin Sebor  <msebor@redhat.com>
39202         PR tree-optimization/79327
39203         * gimple-ssa-sprintf.c (tree_digits): Avoid adding the base prefix
39204         when precision has resulted in leading zeros.
39205         (format_integer): Adjust the likely counter to assume an unknown
39206         argument that may be zero is non-zero.
39208 2017-02-03  Jason Merrill  <jason@redhat.com>
39210         PR c++/78689
39211         * tree-inline.c (copy_tree_body_r) [COND_EXPR]: Revert change to
39212         avoid copying non-taken branch.
39214 2017-02-03  Jakub Jelinek  <jakub@redhat.com>
39216         PR tree-optimization/79340
39217         * tree-vect-loop.c (vectorizable_reduction): Release
39218         vec_defs elements after safe_splicing them into other vectors.
39219         Formatting fixes.
39221         PR tree-optimization/79327
39222         * gimple-ssa-sprintf.c (adjust_range_for_overflow): If returning
39223         true, always set *argmin and *argmax to TYPE_{MIN,MAX}_VALUE of
39224         dirtype.
39225         (format_integer): Use wide_int_to_tree instead of build_int_cst
39226         + to_?hwi.  If argmin is NULL, just set argmin and argmax to
39227         TYPE_{MIN,MAX}_VALUE of argtype.  Simplify and fix computation
39228         of shortest and longest sequence.
39230 2017-02-03  Uros Bizjak  <ubizjak@gmail.com>
39232         * config/i386/i386.c (dimode_scalar_chain::convert_reg):
39233         Use pextrd for TARGET_SSE4_1 when creating scalar copy.
39235 2017-02-03  Walter Lee  <walt@tilera.com>
39237         PR target/78862
39238         * config/tilegx/tilegx.md (tilegx_expand_prologue): Add blockage
39239         after initial stackframe link reg save.
39240         * config/tilepro/tilepro.md (tilepro_expand_prologue): Likewise.
39242 2017-02-03  Jakub Jelinek  <jakub@redhat.com>
39244         PR target/79354
39245         * config/rs6000/rs6000.md (movsi_from_sf): Use wb constraint instead of
39246         wu for stxssp alternative.
39248 2017-02-03  Martin Sebor  <msebor@redhat.com>
39250         PR tree-optimization/79352
39251         * gimple-fold.c (get_range_strlen): Add argument.
39252         (get_range_strlen): Change return type to bool.
39253         (get_maxval_strlen): Pass in a dummy argument.
39254         * gimple-fold.h (get_range_strlen): Change return type to bool.
39255         * gimple-ssa-sprintf.c (get_string_length): Set unlikely counter.
39256         * tree.h (array_at_struct_end_p): Add argument.
39257         * tree.c (array_at_struct_end_p): Handle it.
39259 2017-02-03  Martin Liska  <mliska@suse.cz>
39261         PR lto/66295
39262         * multiple_target.c (create_dispatcher_calls): Redirect edge
39263         from a caller of a dispatcher.
39264         (expand_target_clones): Make the clones local.
39265         (ipa_target_clone): Do both target clones and resolvers.
39266         (ipa_dispatcher_calls): Remove the pass.
39267         (pass_dispatcher_calls::gate): Likewise.
39268         (make_pass_dispatcher_calls): Likewise.
39269         * passes.def (pass_target_clone): Put as very first IPA early
39270         pass.
39272 2017-02-03  Martin Liska  <mliska@suse.cz>
39274         * symtab.c (symtab_node::binds_to_current_def_p): Bail out
39275         in case of a function with ifunc attribute.
39277 2017-02-03  Martin Liska  <mliska@suse.cz>
39279         * cgraph.c (cgraph_node::dump): Dump function version info.
39280         * symtab.c (symtab_node::dump_base): Add missing new line.
39282 2017-02-02  Jan Hubicka  <hubicka@ucw.cz>
39284         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): New function.
39285         (ifcombine_ifandif): Use it.
39287 2017-02-03  Martin Liska  <mliska@suse.cz>
39289         * doc/invoke.texi: Document default value for
39290         use-after-scope-direct-emission-threshold.
39292 2017-02-03  Martin Liska  <mliska@suse.cz>
39294         PR tree-optimization/79339
39295         * gimple-ssa-sprintf.c (format_floating_max): Call mpfr_clear.
39296         (format_floating): Likewise.
39298 2017-02-03  Martin Liska  <mliska@suse.cz>
39300         PR ipa/79337
39301         * ipa-prop.c (ipa_node_params_t::insert): Remove current
39302         implementation.
39303         (ipa_node_params_t::remove): Likewise.
39304         * ipa-prop.h (ipa_node_params::ipa_node_params): Make default
39305         initialization from removed ipa_node_params_t::insert.
39306         (ipa_node_params::~ipa_node_params): Move from removed
39307         ipa_node_params_t::release.
39308         * symbol-summary.h (symbol_summary::m_released): New member.
39309         Do not release a summary twice.  Do not allow to call finalizer
39310         for types of a summary that live in GGC memory.
39312 2017-02-02  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
39314         * config/aarch64/aarch64.c (thunderx2t99_tunings): Enable AES and
39315         cmp_branch fusion.
39317 2017-02-02  Martin Sebor  <msebor@redhat.com>
39319         PR middle-end/79275
39320         * gimple-ssa-sprintf.c (get_string_length): Set lower bound to zero.
39321         (format_string): Tighten up the range of output for non-constant
39322         strings and correct the expected range for wide non-constant strings.
39324 2017-02-02  Martin Sebor  <msebor@redhat.com>
39326         * doc/invoke.texi (-maccumulate-args): Fix bad grammar.
39328         PR middle-end/32003
39329         * doc/invoke.texi (-fdump-tree-): Remove pass-specific options from
39330         index.
39331         (-fdump-tree-@var): Add to index and document how to come up
39332         with pass-specific option and dump file names.
39333         (-fdump-passes): Clarify where to look for output.
39335 2017-02-02  Jan Hubicka  <hubicka@ucw.cz>
39337         PR middle-end/77445
39338         * tree-ssa-threadbackward.c (profitable_jump_thread_path): Dump
39339         statistics of the analyzed path; allow threading for speed when
39340         any of BBs along the path are optimized for speed.
39342 2017-02-02  Eric Botcazou  <ebotcazou@adacore.com>
39344         PR middle-end/78468
39345         * emit-rtl.c (init_emit): Add ??? comment for problematic alignment
39346         settings of the virtual registers.
39348         Revert again
39349         2016-08-23  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39351         * explow.c (get_dynamic_stack_size): Take known alignment of stack
39352         pointer + STACK_DYNAMIC_OFFSET into account when calculating the size
39353         needed.
39355 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39357         * config/s390/vx-builtins.md ("vec_ceil", "vec_floor")
39358         ("vec_trunc", "vec_roundc", "vec_round"): Remove expanders.
39360 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39362         * config/s390/s390.md: Add missing comments with the expanded
39363         mnemonics.
39364         * config/s390/vector.md: Likewise.
39365         * config/s390/vx-builtins.md: Likewise.
39367 2017-02-02  Jakub Jelinek  <jakub@redhat.com>
39369         PR target/79197
39370         * config/rs6000/rs6000.md (*fixuns_trunc<mode>di2_fctiduz): Rename to ...
39371         (fixuns_trunc<mode>di2): ... this, remove previous expander.  Put all
39372         conditions on a single line.
39374 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39376         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Rename
39377         __S390_VX__ to __VX__.
39379 2017-02-01  Andrew Pinski  <apinski@cavium.com>
39381         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost): Pass
39382         stmt_info to record_stmt_cost.
39383         (vect_get_known_peeling_cost): Pass stmt_info if known to
39384         record_stmt_cost.
39385         * config/aarch64/aarch64-protos.h (cpu_vector_cost): Split
39386         cpu_vector_cost field into
39387         scalar_int_stmt_cost and scalar_fp_stmt_cost.  Split vec_stmt_cost
39388         field into vec_int_stmt_cost and vec_fp_stmt_cost.
39389         * config/aarch64/aarch64.c (generic_vector_cost): Update for the
39390         splitting of scalar_stmt_cost and vec_stmt_cost.
39391         (thunderx_vector_cost): Likewise.
39392         (cortexa57_vector_cost): LIkewise.
39393         (exynosm1_vector_cost): Likewise.
39394         (xgene1_vector_cost): Likewise.
39395         (thunderx2t99_vector_cost): Improve after the splitting of the two
39396         fields.
39397         (aarch64_builtin_vectorization_cost): Update for the splitting of
39398         scalar_stmt_cost and vec_stmt_cost.
39400 2017-02-01  Torvald Riegel  <triegel@redhat.com>
39401             Richard Henderson  <rth@redhat.com>
39403         * builtins.c (fold_builtin_atomic_always_lock_free): Make "lock-free"
39404         conditional on existance of a fast atomic load.
39405         * optabs-query.c (can_atomic_load_p): New function.
39406         * optabs-query.h (can_atomic_load_p): Declare it.
39407         * optabs.c (expand_atomic_exchange): Always delegate to libatomic if
39408         no fast atomic load is available for the particular size of access.
39409         (expand_atomic_compare_and_swap): Likewise.
39410         (expand_atomic_load): Likewise.
39411         (expand_atomic_store): Likewise.
39412         (expand_atomic_fetch_op): Likewise.
39413         * testsuite/lib/target-supports.exp
39414         (check_effective_target_sync_int_128): Remove x86 because it provides
39415         no fast atomic load.
39416         (check_effective_target_sync_int_128_runtime): Likewise.
39418 2017-02-01  Richard Biener  <rguenther@suse.de>
39420         * graphite.c: Include tree-vectorizer.h for find_loop_location.
39421         (graphite_transform_loops): Provide opt-info for optimized nests.
39422         * tree-parloop.c (parallelize_loops): Provide opt-info for
39423         parallelized loops.
39425 2017-02-01  Richard Biener  <rguenther@suse.de>
39427         PR middle-end/79315
39428         * tree-cfg.c (move_stmt_op): Never set TREE_BLOCK when it
39429         was not set before.
39431 2017-02-01  Richard Biener  <rguenther@suse.de>
39433         PR tree-optimization/71824
39434         * graphite-scop-detection.c (scop_detection::build_scop_breadth):
39435         Verify the loops are valid in the merged SESE region.
39436         (scop_detection::can_represent_loop_1): Check analyzing the
39437         evolution of the number of iterations in the region succeeds.
39439 2017-01-31  Ian Lance Taylor  <iant@golang.org>
39441         * config/i386/i386.c (ix86_expand_split_stack_prologue): Add
39442         REG_ARGS_SIZE note to 32-bit push insns and call insn.
39444 2017-01-31  David Malcolm  <dmalcolm@redhat.com>
39446         PR preprocessor/79210
39447         * input.c (get_substring_ranges_for_loc): Replace line_width
39448         assertion with error-handling.
39450 2017-01-31  Richard Biener  <rguenther@suse.de>
39452         PR tree-optimization/77318
39453         * graphite-sese-to-poly.c (extract_affine): Fix assert.
39454         (create_pw_aff_from_tree): Take loop parameter.
39455         (add_condition_to_pbb): Pass loop of the condition to
39456         create_pw_aff_from_tree.
39458 2017-01-31  Jakub Jelinek  <jakub@redhat.com>
39460         * config/s390/s390.c (s390_asan_shadow_offset): New function.
39461         (TARGET_ASAN_SHADOW_OFFSET): Redefine.
39463 2017-01-31  Michael Meissner  <meissner@linux.vnet.ibm.com>
39465         PR target/78597
39466         PR target/79038
39467         * config/rs6000/rs6000-protos.h (convert_float128_to_int): Delete,
39468         no longer used.
39469         (convert_int_to_float128): Likewise.
39470         * config/rs6000/rs6000.c (convert_float128_to_int): Likewise.
39471         (convert_int_to_float128): Likewise.
39472         * config/rs6000/rs6000.md (UNSPEC_IEEE128_MOVE): Likewise.
39473         (UNSPEC_IEEE128_CONVERT): Likewise.
39474         (floatsi<mode>2, FLOAT128 iterator): Bypass calling
39475         rs6000_expand_float128_convert if we have IEEE 128-bit hardware.
39476         Use local variables for IBM extended format.
39477         (fix_trunc<mode>si2, FLOAT128 iterator): Likewise.
39478         (fix_trunc<mode>si2_fprs): Likewise.
39479         (fixuns_trunc<IEEE128:mode><SDI:mode>2): Likewise.
39480         (floatuns<IEEE128:mode>2, IEEE128 iterator): Likewise.
39481         (fix<uns>_<mode>si2_hw): Rework the IEEE 128-bt hardware support
39482         to know that we can now have integers of all sizes in vector
39483         registers.
39484         (fix<uns>_<mode>di2_hw): Likewise.
39485         (float<uns>_<mode>si2_hw): Likewise.
39486         (fix_<mode>si2_hw): Likewise.
39487         (fixuns_<mode>si2_hw): Likewise.
39488         (float<uns>_<mode>di2_hw): Likewise.
39489         (float_<mode>di2_hw): Likewise.
39490         (float_<mode>si2_hw): Likewise.
39491         (floatuns_<mode>di2_hw): Likewise.
39492         (floatuns_<mode>si2_hw): Likewise.
39493         (xscvqp<su>wz_<mode>): Delete, no longer used.
39494         (xscvqp<su>dz_<mode>): Likewise.
39495         (xscv<su>dqp_<mode>): Likewise.
39496         (ieee128_mfvsrd_64bit): Likewise.
39497         (ieee128_mfvsrd_32bit): Likewise.
39498         (ieee128_mfvsrwz): Likewise.
39499         (ieee128_mtvsrw): Likewise.
39500         (ieee128_mtvsrd_64bit): Likewise.
39501         (ieee128_mtvsrd_32bit): Likewise.
39503 2017-01-31  Martin Liska  <mliska@suse.cz>
39505         PR ipa/79285
39506         * ipa-prop.c (ipa_free_all_node_params): Call release method
39507         instead of ~sumbol_summary to not to trigger double times
39508         dtor of hash_map.
39510 2017-01-31  Aldy Hernandez  <aldyh@redhat.com>
39512         PR tree-optimization/71691
39513         * bitmap.h (class auto_bitmap): New.
39514         * tree-ssa-loop-unswitch.c (tree_may_unswitch_on): Call
39515         is_maybe_undefined instead of ssa_undefined_value_p.
39517 2017-01-31  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39519         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Rename
39520         __S390_ARCH_LEVEL__ to __ARCH__.
39522 2017-01-31  Jakub Jelinek  <jakub@redhat.com>
39524         PR tree-optimization/79267
39525         * value-prof.c (gimple_ic): Only drop lhs for noreturn calls
39526         if should_remove_lhs_p is true.
39528 2017-01-30  Alexandre Oliva  <aoliva@redhat.com>
39530         PR debug/63238
39531         * dwarf2out.c (clone_as_declaration): Drop DW_AT_alignment.
39532         (add_alignment_attribute): New.
39533         (base_type_die): Add alignment attribute.
39534         (subrange_type_die): Likewise.
39535         (modified_type_die): Likewise.
39536         (gen_array_type_die): Likewise.
39537         (gen_descr_array_type_die: Likewise.
39538         (gen_enumeration_type_die): Likewise.
39539         (gen_subprogram_die): Likewise.
39540         (gen_variable_die): Likewise.
39541         (gen_field_die): Likewise.
39542         (gen_ptr_to_mbr_type_die): Likewise.
39543         (gen_struct_or_union_type_die): Likewise.
39544         (gen_subroutine_type_die): Likewise.
39545         (gen_typedef_die): Likewise.
39546         (base_type_cmp): Compare alignment attribute.
39548 2017-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
39550         PR target/79170
39551         * config/rs6000/altivec.md (*setb_internal): Rename to setb_signed.
39552         (setb_unsigned) New pattern for setb with CCUNS.
39553         * config/rs6000/rs6000.c (expand_block_compare): Use a different
39554         subfc./subfe sequence to avoid overflow problems.  Generate a
39555         shorter sequence with cmpld/setb for power9.
39556         * config/rs6000/rs6000.md (subf<mode>3_carry_dot2): Add a new pattern
39557         for generating subfc. instruction.
39558         (cmpstrsi): Add TARGET_POPCNTD predicate as the generate sequence
39559         now uses this instruction.
39561 2017-01-30  Ian Lance Taylor  <iant@google.com>
39563         PR debug/79289
39564         * dwarf2out.c (gen_type_die_with_usage): When picking a variant
39565         for FUNCTION_TYPE/METHOD_TYPE, use the first matching one.
39567 2017-01-30  Martin Sebor  <msebor@redhat.com>
39569         * gimple-ssa-sprintf.c (fmtresult::adjust_for_width_or_precision):
39570         Move constant to the right of a relational operator.
39571         (get_mpfr_format_length, format_character, format_string): Ditto.
39572         (should_warn_p, maybe_warn): Same.
39574         * doc/invoke.texi (-Wformat-truncation=1): Fix typo.
39576 2017-01-30  Maxim Ostapenko  <m.ostapenko@samsung.com>
39578         PR lto/79061
39579         * asan.c (get_translation_unit_decl): Remove function.
39580         (asan_add_global): Force has_dynamic_init to zero in LTO mode.
39582 2017-01-30  Martin Liska  <mliska@suse.cz>
39584         PR gcov-profile/79259
39585         * opts.c (common_handle_option): Enable flag_ipa_bit_cp w/
39586         -fprofile-generate.
39588 2017-01-30  Martin Liska  <mliska@suse.cz>
39590         PR bootstrap/78985
39591         * config/aarch64/cortex-a57-fma-steering.c (func_fma_steering::analyze):
39592         Initialize variables with NULL value.
39594 2017-01-30  Richard Earnshaw  <rearnsh@arm.com>
39596         PR target/79260
39597         * config.gcc (arm*-*-*): Add arm/arm-flags.h and arm/arm-isa.h to
39598         tm_p_file.
39599         * arm/arm-protos.h: Don't directly include arm-flags.h and arm-isa.h.
39601 2017-01-30  Richard Biener  <rguenther@suse.de>
39603         PR tree-optimization/79276
39604         * tree-vrp.c (process_assert_insertions): Properly adjust common
39605         when removing a duplicate.
39607 2017-01-30  Richard Biener  <rguenther@suse.de>
39609         PR tree-optimization/79256
39610         * targhooks.c (default_builtin_vector_alignment_reachable): Honor
39611         BIGGEST_FIELD_ALIGNMENT and ADJUST_FIELD_ALIGN to fix up bogus
39612         alignment on TYPE.
39613         * tree.c (build_aligned_type): Set TYPE_USER_ALIGN.
39615 2017-01-30  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39617         PR target/79240
39618         * config/s390/s390.md ("*r<noxa>sbg_<mode>_srl_bitmask")
39619         ("*r<noxa>sbg_<mode>_sll_bitmask")
39620         ("*extzv_<mode>_srl<clobbercc_or_nocc>")
39621         ("*extzv_<mode>_sll<clobbercc_or_nocc>"):
39622         Use contiguous_bitmask_nowrap_operand.
39624 2017-01-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
39626         PR target/79268
39627         * config/rs6000/altivec.h (vec_xl): Revise #define.
39628         (vec_xst): Likewise.
39630 2017-01-27  Uros Bizjak  <ubizjak@gmail.com>
39632         * config/i386/i386.c (print_reg): Use REGNO instead of true_regnum.
39634 2017-01-27  Bernd Schmidt  <bschmidt@redhat.com>
39636         PR rtl-optimization/79194
39637         * cprop.c (one_cprop_pass): Move deletion of code after unconditional
39638         traps before call to bypass_conditional_jumps.
39640 2017-01-27  Vladimir Makarov  <vmakarov@redhat.com>
39642         PR tree-optimization/71374
39643         * lra-constraints.c (check_conflict_input_operands): New.
39644         (match_reload): Use it.
39646 2017-01-27  Vladimir Makarov  <vmakarov@redhat.com>
39648         PR target/79131
39649         * lra-assigns.c (find_hard_regno_for_1): Take endianess for into
39650         account to calculate conflict_set.
39652 2017-01-27  Bin Cheng  <bin.cheng@arm.com>
39654         PR rtl-optimization/78559
39655         * combine.c (try_combine): Discard REG_EQUAL and REG_EQUIV for
39656         other_insn in combine.
39658 2017-01-27  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
39660         * builtin-types.def: Use unsigned_char_type_node for BT_UINT8.  Use
39661         uint16_type_node for BT_UINT16.
39663 2017-01-27  David Malcolm  <dmalcolm@redhat.com>
39665         * doc/sourcebuild.texi (Testsuites): Add "GIMPLE Tests" and
39666         "RTL Tests" to menu.
39667         (GIMPLE Tests): New node.
39668         (RTL Tests): New node.
39670 2017-01-27  Richard Biener  <rguenther@suse.de>
39672         PR tree-optimization/79245
39673         * tree-loop-distribution.c (distribute_loop): Apply cost
39674         modeling also to detected patterns.
39676 2017-01-27  Richard Biener  <rguenther@suse.de>
39678         PR tree-optimization/71433
39679         * tree-vrp.c (register_new_assert_for): Revert earlier changes.
39680         (compare_assert_loc): New function.
39681         (process_assert_insertions): Sort and optimize assert locations
39682         to remove duplicates and push down identical assertions on
39683         edges to their destination block.
39685 2017-01-27  Richard Biener  <rguenther@suse.de>
39687         PR tree-optimization/79244
39688         * tree-vrp.c (remove_range_assertions): Forcefully propagate
39689         out SSA names even if abnormal.
39691 2017-01-27  Jakub Jelinek  <jakub@redhat.com>
39693         * realmpfr.h: Poison MPFR_RND{N,Z,U,D}.
39694         * gimple-ssa-sprintf.c (format_floating_max): Use GMP_RNDN
39695         instead of MPFR_RNDN.
39697 2017-01-27  Richard Earnshaw  <rearnsha@arm.com>
39699         PR target/79239
39700         * arm.c (arm_option_override): Don't call build_target_option_node
39701         until after doing all option overrides.
39702         (arm_valid_target_attribute_tree): Likewise.
39704 2017-01-27  Martin Liska  <mliska@suse.cz>
39706         * doc/invoke.texi (-fprofile-arcs): Document profiling support
39707         for {cd}tors and C++ {cd}tors.
39709 2017-01-27  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39711         * config/s390/s390.md ("*setmem_long_and")
39712         ("*setmem_long_and_31z"): Use zero_extend instead of and.
39714 2017-01-26  Martin Sebor  <msebor@redhat.com>
39716         * gimple-ssa-sprintf.c (format_floating): Simplify the computation
39717         of precision.
39719 2017-01-26  Martin Sebor  <msebor@redhat.com>
39721         * gimple-ssa-sprintf.c (format_floating): Test HAVE_XFmode and
39722         HAVE_DFmode before using XFmode or DFmode.
39723         (parse_directive): Avoid using the z length modifier to avoid
39724         the ISO C++98 does not support the â€˜z’ gnu_printf length modifier.
39726         PR middle-end/78703
39727         * gimple-ssa-sprintf.c (adjust_for_width_or_precision): Change
39728         to accept adjustment as an array.
39729         (get_int_range): New function.
39730         (struct directive): Make width and prec arrays.
39731         (directive::set_width, directive::set_precision): Call get_int_range.
39732         (format_integer, format_floating): Handle width and precision ranges.
39733         (format_string, parse_directive): Same.
39735 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39737         PR debug/79129
39738         * dwarf2out.c (generate_skeleton_bottom_up): For children with
39739         comdat_type_p set, just clone them, but keep the children in the
39740         original DIE.
39742         PR debug/78835
39743         * dwarf2out.c (prune_unused_types): Mark all functions with DIEs
39744         which have direct callers with -fvar-tracking-assignments enabled
39745         in the current TU.
39746         (resolve_addr): Avoid adding skeleton DIEs for DW_AT_call_origin
39747         inside of type units.
39749 2017-01-26  Martin Sebor  <msebor@redhat.com>
39751         PR middle-end/78703
39752         * gimple-ssa-sprintf.c (struct result_range): Add likely and
39753         unlikely counters.
39754         (struct format_result): Replace number_chars, number_chars_min,
39755         and number_chars_max with a single member of struct result_range.
39756         Remove bounded.
39757         (format_result::operator+=): Adjust.
39758         (struct fmtresult): Remove bounded.  Handle likely and unlikely
39759         counters.
39760         (fmtresult::adjust_for_width_or_precision): New function.
39761         (fmtresult:type_max_digits): New function.
39762         (bytes_remaining): Handle likely and unlikely counters.
39763         (min_bytes_remaining): Remove.
39764         (format_percent): Simplify.
39765         (format_integer, format_floating): Set likely and unlikely counters.
39766         (get_string_length, format_character, format_string): Same.
39767         (format_plain, should_warn_p): New function.
39768         (maybe_warn): Call should_warn_p.  Update diagnostic messages
39769         and handle those for all directives, including plain strings.
39770         (format_directive): Handle likely and unlikely counters.
39771         Remove unnecessary quoting from diagnostics.  Add an informational
39772         note.
39773         (add_bytes): Remove.
39774         (pass_sprintf_length::compute_format_length): Simplify.
39775         (try_substitute_return_value): Handle likely and unlikely counters.
39777 2017-01-26  Carl Love  <cel@us.ibm.com>
39779         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Remove
39780         bogus entries for the P8V_BUILTIN_VEC_VGBBD built-ins
39782 2017-01-26  Vladimir Makarov  <vmakarov@redhat.com>
39784         PR target/79131
39785         * lra-assigns.c (setup_live_pseudos_and_spill_after_risky): Take
39786         endianess for subregs into account.
39787         * lra-constraints.c (lra_constraints): Do risky transformations
39788         always on the first iteration.
39789         * lra-lives.c (check_pseudos_live_through_calls): Add arg
39790         last_call_used_reg_set.
39791         (process_bb_lives): Define and use last_call_used_reg_set.
39792         * lra.c (lra): Always continue after lra_constraints on the first
39793         iteration.
39795 2017-01-26  Kirill Yukhin  <kirill.yukhin@gmail.com>
39797         * gcc.target/i386/avx512bw-kshiftlq-2.c: Use unsigned long long
39798         constant.
39799         * gcc.target/i386/avx512bw-kshiftrq-2.c: Ditto.
39801 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39803         * config/i386/avx512fintrin.h (_ktest_mask16_u8,
39804         _ktestz_mask16_u8, _ktestc_mask16_u8, _kadd_mask16): Move to ...
39805         * config/i386/avx512dqintrin.h (_ktest_mask16_u8,
39806         _ktestz_mask16_u8, _ktestc_mask16_u8, _kadd_mask16): ... here.
39807         * config/i386/i386-builtin.def (__builtin_ia32_ktestchi,
39808         __builtin_ia32_ktestzhi, __builtin_ia32_kaddhi): Use
39809         OPTION_MASK_ISA_AVX512DQ instead of OPTION_MASK_ISA_AVX512F.
39810         * config/i386/sse.md (SWI1248_AVX512BWDQ2): New mode iterator.
39811         (kadd<mode>, ktest<mode>): Use it instead of SWI1248_AVX512BWDQ.
39813 2017-01-26  Marek Polacek  <polacek@redhat.com>
39815         PR c/79199
39816         * fold-const.c (operand_equal_p) [COND_EXPR]: Use OP_SAME_WITH_NULL
39817         for the third operand.
39819 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39821         PR middle-end/79236
39822         * omp-low.c (struct omp_context): Add simt_stmt field.
39823         (scan_omp_for): Return omp_context *.
39824         (scan_omp_simd): Set simt_stmt on the non-_simt_ SIMD
39825         context to the _simt_ SIMD stmt.
39826         (lower_omp_for): For combined SIMD with sibling _simt_
39827         SIMD, make sure to use the same decls in _looptemp_
39828         clauses as in the sibling.
39830 2017-01-26  David Sherwood  <david.sherwood@arm.com>
39832         PR middle-end/79212
39833         * gimplify.c (omp_notice_variable): Add GOVD_SEEN flag to variables in
39834         all contexts.
39836 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39838         PR target/70465
39839         * reg-stack.c (emit_swap_insn): Instead of fld a; fld b; fxchg %st(1);
39840         emit fld b; fld a; if possible.
39842         * brig-builtins.def: Update copyright years.
39843         * config/arm/arm_acle_builtins.def: Update copyright years.
39845 2017-01-25  Michael Meissner  <meissner@linux.vnet.ibm.com>
39847         PR target/79179
39848         * config/rs6000/vsx.md (vsx_extract_<mode>_store): Use wY
39849         constraint instead of o for the stxsd instruction.
39851 2017-01-25  Carl Love  <cel@us.ibm.com>
39853         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Fix order
39854         of entries for ALTIVEC_BUILTIN_VEC_PACKS and P8V_BUILTIN_VEC_VGBBD.
39856 2017-01-25  Jonathan Wakely  <jwakely@redhat.com>
39858         * doc/invoke.texi (C++ Dialect Options): Fix typo.
39860 2017-01-25  Richard Biener  <rguenther@suse.de>
39862         PR tree-optimization/69264
39863         * target.def (vector_alignment_reachable): Improve documentation.
39864         * doc/tm.texi: Regenerate.
39865         * targhooks.c (default_builtin_vector_alignment_reachable): Simplify
39866         and add a comment.
39867         * tree-vect-data-refs.c (vect_supportable_dr_alignment): Revert
39868         earlier changes with respect to TYPE_USER_ALIGN.
39869         (vector_alignment_reachable_p): Likewise.  Improve dumping.
39871 2016-01-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
39873         PR target/79145
39874         * config/arm/arm.md (xordi3): Force constant operand into a register
39875         for TARGET_IWMMXT.
39877 2016-01-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
39879         * doc/invoke.texi (-fstore-merging): Correct default optimization
39880         levels at which it is enabled.
39881         (-O): Move -fstore-merging from list to...
39882         (-O2): ... Here.
39884 2017-01-25  Richard Biener  <rguenther@suse.de>
39886         PR debug/78363
39887         * omp-expand.c: Include debug.h.
39888         (expand_omp_taskreg): Make sure to generate early debug before
39889         outlining anything from a function.
39890         (expand_omp_target): Likewise.
39891         (grid_expand_target_grid_body): Likewise.
39893 2017-01-25  Maxim Ostapenko  <m.ostapenko@samsung.com>
39895         PR lto/79061
39896         * asan.c (get_translation_unit_decl): New function.
39897         (asan_add_global): Extract modules file name from globals
39898         TRANSLATION_UNIT_DECL name.
39900 2017-01-24  Eric Botcazou  <ebotcazou@adacore.com>
39902         PR target/77439
39903         * config/arm/arm.c (arm_function_ok_for_sibcall): Add back restriction
39904         for long calls with APCS frame and VFP.
39906 2017-01-24  David Malcolm  <dmalcolm@redhat.com>
39908         * cfg.c (original_copy_tables_initialized_p): New function.
39909         * cfg.h (original_copy_tables_initialized_p): New decl.
39910         * cfgrtl.c (relink_block_chain): Guard the call to
39911         free_original_copy_tables with a call to
39912         original_copy_tables_initialized_p.
39913         * cgraph.h (symtab_node::native_rtl_p): New decl.
39914         * cgraphunit.c (symtab_node::native_rtl_p): New function.
39915         (symtab_node::needed_p): Don't assert for early assembly output
39916         for __RTL functions.
39917         (cgraph_node::finalize_function): Set "force_output" for __RTL
39918         functions.
39919         (cgraph_node::analyze): Bail out early for __RTL functions.
39920         (analyze_functions): Update assertion to support __RTL functions.
39921         (cgraph_node::expand): Bail out early for __RTL functions.
39922         * final.c (rest_of_clean_state): Don't call delete_tree_ssa for
39923         __RTL functions.
39924         * function.h (struct function): Update comment for field
39925         "pass_startwith".
39926         * gimple-expr.c: Include "tree-pass.h".
39927         (gimple_has_body_p): Return false for __RTL functions.
39928         * Makefile.in (OBJS): Add run-rtl-passes.o.
39929         * pass_manager.h (gcc::pass_manager::get_rest_of_compilation): New
39930         accessor.
39931         (gcc::pass_manager::get_clean_slate): New accessor.
39932         * passes.c: Include "insn-addr.h".
39933         (should_skip_pass_p): Add logging.  Update logic for running
39934         "expand" to be compatible with both __GIMPLE and __RTL.  Guard
39935         property-provider override so it is only done for gimple passes.
39936         Don't skip dfinit.
39937         (skip_pass): New function.
39938         (execute_one_pass): Call skip_pass when skipping passes.
39939         * read-md.c (md_reader::read_char): Support filtering
39940         the input to a subset of line numbers.
39941         (md_reader::md_reader): Initialize fields
39942         m_first_line and m_last_line.
39943         (md_reader::read_file_fragment): New function.
39944         * read-md.h (md_reader::read_file_fragment): New decl.
39945         (md_reader::m_first_line): New field.
39946         (md_reader::m_last_line): New field.
39947         * read-rtl-function.c (function_reader::create_function): Only
39948         create cfun if it doesn't already exist.  Set PROP_rtl on cfun's
39949         curr_properties.  Set DECL_INITIAL to a dummy block.
39950         (read_rtl_function_body_from_file_range): New function.
39951         * read-rtl-function.h (read_rtl_function_body_from_file_range):
39952         New decl.
39953         * run-rtl-passes.c: New file.
39954         * run-rtl-passes.h: New file.
39956 2017-01-24  Jeff Law  <law@redhat.com>
39958         * config/microblaze/microblaze.h (ASM_FORMAT_PRIVATE_NAME): Increase
39959         buffer size.
39961 2017-01-24  Bin Cheng  <bin.cheng@arm.com>
39963         PR tree-optimization/79159
39964         * tree-ssa-loop-niter.c (get_cst_init_from_scev): New function.
39965         (record_nonwrapping_iv): Improve boundary using above function if no
39966         value range information.
39968 2017-01-24  Pekka Jääskeläinen  <pekka@parmance.com>
39969             Martin Jambor  <mjambor@suse.cz>
39971         * brig-builtins.def: New file.
39972         * builtins.def (DEF_HSAIL_BUILTIN): New macro.
39973         (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
39974         (DEF_HSAIL_SAT_BUILTIN): Likewise.
39975         (DEF_HSAIL_INTR_BUILTIN): Likewise.
39976         (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
39977         * builtin-types.def (BT_INT8): New.
39978         (BT_INT16): Likewise.
39979         (BT_UINT8): Likewise.
39980         (BT_UINT16): Likewise.
39981         (BT_FN_ULONG): Likewise.
39982         (BT_FN_UINT_INT): Likewise.
39983         (BT_FN_UINT_ULONG): Likewise.
39984         (BT_FN_UINT_LONG): Likewise.
39985         (BT_FN_UINT_PTR): Likewise.
39986         (BT_FN_ULONG_PTR): Likewise.
39987         (BT_FN_INT8_FLOAT): Likewise.
39988         (BT_FN_INT16_FLOAT): Likewise.
39989         (BT_FN_UINT32_FLOAT): Likewise.
39990         (BT_FN_UINT16_FLOAT): Likewise.
39991         (BT_FN_UINT8_FLOAT): Likewise.
39992         (BT_FN_UINT64_FLOAT): Likewise.
39993         (BT_FN_UINT16_UINT32): Likewise.
39994         (BT_FN_UINT32_UINT16): Likewise.
39995         (BT_FN_UINT16_UINT16_UINT16): Likewise.
39996         (BT_FN_INT_PTR_INT): Likewise.
39997         (BT_FN_UINT_PTR_UINT): Likewise.
39998         (BT_FN_LONG_PTR_LONG): Likewise.
39999         (BT_FN_ULONG_PTR_ULONG): Likewise.
40000         (BT_FN_VOID_UINT64_UINT64): Likewise.
40001         (BT_FN_UINT8_UINT8_UINT8): Likewise.
40002         (BT_FN_INT8_INT8_INT8): Likewise.
40003         (BT_FN_INT16_INT16_INT16): Likewise.
40004         (BT_FN_INT_INT_INT): Likewise.
40005         (BT_FN_UINT_FLOAT_UINT): Likewise.
40006         (BT_FN_FLOAT_UINT_UINT): Likewise.
40007         (BT_FN_ULONG_UINT_UINT): Likewise.
40008         (BT_FN_ULONG_UINT_PTR): Likewise.
40009         (BT_FN_ULONG_ULONG_ULONG): Likewise.
40010         (BT_FN_UINT_UINT_UINT): Likewise.
40011         (BT_FN_VOID_UINT_PTR): Likewise.
40012         (BT_FN_UINT_UINT_PTR: Likewise.
40013         (BT_FN_UINT32_UINT64_PTR): Likewise.
40014         (BT_FN_INT_INT_UINT_UINT): Likewise.
40015         (BT_FN_UINT_UINT_UINT_UINT): Likewise.
40016         (BT_FN_UINT_UINT_UINT_PTR): Likewise.
40017         (BT_FN_UINT_ULONG_ULONG_UINT): Likewise.
40018         (BT_FN_ULONG_ULONG_ULONG_ULONG): Likewise.
40019         (BT_FN_LONG_LONG_UINT_UINT): Likewise.
40020         (BT_FN_ULONG_ULONG_UINT_UINT): Likewise.
40021         (BT_FN_VOID_UINT32_UINT64_PTR): Likewise.
40022         (BT_FN_VOID_UINT32_UINT32_PTR): Likewise.
40023         (BT_FN_UINT_UINT_UINT_UINT_UINT): Likewise.
40024         (BT_FN_UINT_FLOAT_FLOAT_FLOAT_FLOAT): Likewise.
40025         (BT_FN_ULONG_ULONG_ULONG_UINT_UINT): Likewise.
40026         * doc/frontends.texi: List BRIG FE.
40027         * doc/install.texi (Testing): Add BRIG tesring requirements.
40028         * doc/invoke.texi (Overall Options): Mention BRIG.
40029         * doc/standards.texi (Standards): Doucment BRIG HSA version.
40031 2017-01-24  Richard Biener  <rguenther@suse.de>
40033         PR translation/79208
40034         * ipa-devirt.c (odr_types_equivalent_p): Fix typo in diagnostic.
40036 2017-01-24  Martin Jambor  <mjambor@suse.cz>
40038         PR bootstrap/79198
40039         * ipa-prop.c (ipa_free_all_node_params): Call summary destructor.
40040         * ipa-prop.c (ipa_node_params_t::insert): Initialize fields known_csts
40041         and known_contexts.
40043 2017-01-24  Aldy Hernandez  <aldyh@redhat.com>
40045         PR middle-end/79123
40046         * gimple-ssa-warn-alloca.c (alloca_call_type): Make sure
40047         casts from signed to unsigned really don't have a range.
40049 2017-01-24  Markus Trippelsdorf  <markus@trippelsdorf.de>
40051         * gimple-ssa-sprintf.c (format_floating): Change MPFR_RNDx to
40052         GMP_RNDx for compatiblity.
40054 2017-01-24  Martin Liska  <mliska@suse.cz>
40056         PR bootstrap/79132
40057         * tree-ssa-reassoc.c (rewrite_expr_tree_parallel): Insert assert
40058         that would prevent us to call alloca with -1 as argument.
40060 2017-01-24  Jakub Jelinek  <jakub@redhat.com>
40062         * dwarf2out.c (output_compilation_unit_header, output_file_names):
40063         Avoid -Wformat-security warning.
40065 2017-01-23  Andrew Pinski  <apinski@cavium.com>
40067         * config/aarch64/aarch64.c (thunderx2t99_addrcost_table): Improve
40068         cost table.
40070 2017-01-23  Martin Sebor  <msebor@redhat.com>
40072         PR middle-end/78703
40073         * gimple-ssa-sprintf.c (warn_level): New global.
40074         (format_integer): Use it here and throughout the rest of the file.
40075         Use the same switch to compute sign as base.
40076         (maybe_warn): New function.
40077         (format_directive): Factor out warnings into maybe_warn.
40078         Add debugging output.  Use warn_level.
40079         (add_bytes): Use warn_level.
40080         (pass_sprintf_length::compute_format_length): Add debugging output.
40081         (try_substitute_return_value): Same.
40082         (pass_sprintf_length::handle_gimple_call): Set and use warn_level.
40084         PR middle-end/78703
40085         * gimple-ssa-sprintf.c (struct format_result): Remove constant member.
40086         (struct fmtresult, format_integer, format_floating): Adjust.
40087         (fmtresult::fmtresult): Set max correctly in two argument ctor.
40088         (get_string_length, format_string,format_directive): Same.
40089         (pass_sprintf_length::compute_format_length): Same.
40090         (try_substitute_return_value): Simplify slightly.
40092         PR middle-end/78703
40093         * gimple-ssa-sprintf.c (pass_sprintf_length::gate): Adjust formatting.
40094         (fmtresult::operator+=): Outlined.
40095         (struct fmtresult): Add ctors.
40096         (struct conversion_spec): Rename...
40097         (struct directive): ...to this.  Add and remove data members.
40098         (directive::set_width, directive::set_precision): New functions.
40099         (format_percent): Use fmtresult ctor.
40100         (get_width_and_precision): Remove.
40101         (format_integer): Make naming changes.  Avoid computing width and
40102         precision.
40103         (format_floating): Same.  Adjust indentation.
40104         (format_character, format_none): New functions.
40105         (format_string): Moved character handling to format_character.
40106         (format_directive): Remove arguments, change return type.
40107         (parse_directive): New function.
40108         (pass_sprintf_length::compute_format_length): Move directive
40109         parsing to parse_directive.
40111 2017-01-23  Jakub Jelinek  <jakub@redhat.com>
40113         * tree.h (assign_assembler_name_if_neeeded): Rename to ...
40114         (assign_assembler_name_if_needed): ... this.
40115         * tree.c (assign_assembler_name_if_neeeded): Rename to ...
40116         (assign_assembler_name_if_needed): ... this.
40117         (free_lang_data_in_cgraph): Adjust callers.
40118         * cgraphunit.c (cgraph_node::analyze): Likewise.
40119         * omp-expand.c (expand_omp_taskreg, expand_omp_target):
40120         Likewise.
40122 2017-01-23  Richard Biener  <rguenther@suse.de>
40124         PR tree-optimization/79088
40125         PR tree-optimization/79188
40126         * tree-ssa-threadupdate.c (mark_threaded_blocks): Move code
40127         resetting loop bounds after last path deletion.  Reset loop
40128         bounds of the target loop, make code match the comments.
40129         * tree-ssa-threadbackwards.c (pass_early_thread_jumps::execute):
40130         Make sure loops need no fixups.
40132 2017-01-23  Kelvin Nilsen  <kelvin@gcc.gnu.org>
40134         * config/rs6000/rs6000-builtin.def (VSIEDPF): Add scalar insert
40135         exponent support with double type for first argument.
40136         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Changed
40137         type returned by __builtin_vec_extract_sig,
40138         __builtin_vec_extract_sig_sp, and __builtin_vec_extract_sig_dp
40139         functions from "vector int" to "vector unsigned int" or from
40140         "vector long long int" to "vector unsigned long long int".
40141         Changed type returned by __builtin_vec_extract_exp,
40142         __builtin_vec_extract_exp_sp, and __builtin_vec_extract_exp_dp
40143         functions from "vector int" to "vector unsigned int" or from
40144         "vector long long int" to "vector unsigned long long int".
40145         Changed return type of __builtin_vec_test_data_class,
40146         __builtin_vec_test_data_class_sp, and
40147         __builtin_vec_test_data_class_dp from "vector int" to
40148         "vector bool int" or from "vector long long int" to "vector bool
40149         long long int" and changed second argument type from "unsigned
40150         int" to "int".  Added new overloaded function forms "vector float
40151         __builtin_vec_insert_exp (vector float, vector unsigned int)" and
40152         "vector float __builtin_vec_insert_exp_sp (vector float, vector
40153         unsigned int)" and "vector double __builtin_vec_insert_exp (vector
40154         double, vector unsigned long long int)" and "vector double
40155         __builtin_vec_insert_exp_dp (vector double, vector unsigned long
40156         long int)".  Changed return type of
40157         __builtin_scalar_test_data_class and
40158         __builtin_scalar_test_data_class_sp and
40159         __builtin_scalar_test_data_class_dp from "unsigned int" to "bool
40160         int" and changed second argument from "unsigned int" to "int".
40161         Changed type returned by __builtin_scalar_test_neg,
40162         __builtin_scalar_test_neg_sp, and __builtin_scalar_test_neg_dp
40163         from "int" to "bool int".  Added new overloaded function form
40164         "double __builtin_scalar_insert_exp (double, unsigned long long int)".
40165         * config/rs6000/vsx.md (xsiexpdpf): New insn for scalar insert
40166         exponent double-precision with floating point first argument.
40167         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
40168         documentation of scalar_test_data_class, scalar_test_neg,
40169         scalar_extract_sig, scalar_extract_exp, scalar_insert_exp,
40170         vector_extract_exp, vec_extract_sig, vec_insert_exp, and
40171         vec_test_data_class built-in functions to reflect refinements in
40172         their type signatures.
40174 2017-01-23  Andreas Tobler  <andreast@gcc.gnu.org>
40176         * config/aarch64/aarch64.c (aarch64_elf_asm_constructor): Increase
40177         size of buf.
40178         (aarch64_elf_asm_destructor): Likewise.
40180 2017-01-23  Bernd Schmidt  <bschmidt@redhat.com>
40182         PR rtl-optimization/78634
40183         * config/i386/i386.c (ix86_max_noce_ifcvt_seq_cost): New function.
40184         (TARGET_MAX_NOCE_IFCVT_SEQ_COST): Define.
40185         * ifcvt.c (noce_try_cmove): Add missing cost check.
40187         PR rtl-optimization/71724
40188         * combine.c (if_then_else_cond): Look for situations where it is
40189         beneficial to undo the work of one of the recursive calls.
40191 2017-01-23  Bin Cheng  <bin.cheng@arm.com>
40193         PR tree-optimization/70754
40194         * tree-predcom.c (stmt_combining_refs): New parameter INSERT_BEFORE.
40195         (reassociate_to_the_same_stmt): New parameter INSERT_BEFORE.  Insert
40196         combined stmt before it if not NULL.
40197         (combine_chains): Process refs reversely and compute dominance point
40198         for root ref.
40200 2017-01-23  Martin Liska  <mliska@suse.cz>
40202         PR tree-optimization/79196
40203         * tree-ssa-strlen.c (fold_strstr_to_memcmp): Rename to ...
40204         (fold_strstr_to_strncmp): ... this.  Fold the pattern to strncmp
40205         instead of memcmp.
40206         (strlen_optimize_stmt): Call the renamed function.
40208 2017-01-23  Michael Matz  <matz@suse.de>
40210         PR tree-optimization/78384
40211         * tree-ssa-loop-split.c (patch_loop_exit): Use correct edge.
40213 2017-01-23  Richard Biener  <rguenther@suse.de>
40215         PR tree-optimization/79186
40216         * tree-vrp.c (register_new_assert_for): Make sure we've seen
40217         both incoming edges before moving an assert.
40219 2017-01-23  Martin Jambor  <mjambor@suse.cz>
40221         * ipa-prop.c (load_from_param_1): Removed.
40222         (load_from_unmodified_param): Bits from load_from_param_1 put back
40223         here.
40224         (load_from_param): Removed.
40225         (compute_complex_assign_jump_func): Removed stmt2 and just replaced it
40226         with stmt.  Reverted back to use of load_from_unmodified_param.
40228 2017-01-23  Martin Jambor  <mjambor@suse.cz>
40230         PR ipa/79108
40231         * ipa-prop.h (ipa_param_descriptor): Anotate with with GTY(()).
40232         (ipa_node_params): Annotate with GTY((for_user)).  Make descriptors
40233         field a pointer to garbage collected vector, mark lattices and
40234         ipcp_orig_node with GTY((skip)).
40235         (ipa_get_param_count): Adjust to descriptors being a pointer.
40236         (ipa_get_param): Likewise.
40237         (ipa_get_type): Likewise.
40238         (ipa_get_param_move_cost): Likewise.
40239         (ipa_set_param_used): Likewise.
40240         (ipa_get_controlled_uses): Likewise.
40241         (ipa_set_controlled_uses): Likewise.
40242         (ipa_is_param_used): Likewise.
40243         (ipa_node_params_t): Move into garbage collector.  New methods insert
40244         and remove.
40245         (ipa_node_params_sum): Annotate wth GTY(()).
40246         (ipa_check_create_node_params): Adjust to ipa_node_params_sum being
40247         garbage collected.
40248         (ipa_load_from_parm_agg): Adjust declaration.
40249         * ipa-icf.c (param_used_p): Adjust to descriptors being a pointer.
40250         * ipa-profile.c (ipa_profile): Likewise.
40251         * ipa-prop.c (ipa_get_param_decl_index_1): Likewise.
40252         (ipa_populate_param_decls): Make descriptors parameter garbage
40253         collected.
40254         (ipa_dump_param): Adjust to descriptors being a pointer.
40255         (ipa_alloc_node_params): Likewise.
40256         (ipa_initialize_node_params): Likewise.
40257         (load_from_param_1): Make descriptors parameter garbage collected.
40258         (load_from_unmodified_param): Likewise.
40259         (load_from_param): Likewise.
40260         (ipa_load_from_parm_agg): Likewise.
40261         (ipa_node_params::~ipa_node_params): Removed.
40262         (ipa_free_all_node_params): Remove call to delete operator.
40263         (ipa_node_params_t::insert): New.
40264         (ipa_node_params_t::remove): Likewise.
40265         (ipa_node_params_t::duplicate): Adjust to descriptors being a pointer,
40266         copy known_csts and known_contexts vectors.
40267         (ipa_read_node_info): Adjust to descriptors being a pointer.
40268         (ipcp_modif_dom_walker): Make m_descriptors field garbage
40269         collected.
40270         (ipcp_transform_function): Make descriptors variable garbage
40271         collected.
40273 2017-01-23  Andrew Senkevich  <andrew.senkevich@intel.com>
40275         * config/i386/avx512bwintrin.h: Add k-mask test, kortest intrinsics.
40276         * config/i386/avx512dqintrin.h: Ditto.
40277         * config/i386/avx512fintrin.h: Ditto.
40278         * config/i386/i386.c: Handle new builtins.
40279         * config/i386/i386-builtin.def: Add new builtins.
40280         * config/i386/sse.md (ktest<mode>, kortest<mode>): New.
40281         (UNSPEC_KORTEST, UNSPEC_KTEST): New.
40283 2017-01-23  Jakub Jelinek  <jakub@redhat.com>
40284             Martin Liska  <mliska@suse.cz>
40286         * asan.h: Define ASAN_USE_AFTER_SCOPE_ATTRIBUTE.
40287         * asan.c (asan_expand_poison_ifn): Support stores and use
40288         appropriate ASAN report function.
40289         * internal-fn.c (expand_ASAN_POISON_USE): New function.
40290         * internal-fn.def (ASAN_POISON_USE): Declare.
40291         * tree-into-ssa.c (maybe_add_asan_poison_write): New function.
40292         (maybe_register_def): Create ASAN_POISON_USE when sanitizing.
40293         * tree-ssa-dce.c (eliminate_unnecessary_stmts): Remove
40294         ASAN_POISON calls w/o LHS.
40295         * tree-ssa.c (execute_update_addresses_taken): Create clobber
40296         for ASAN_MARK (UNPOISON, &x, ...) in order to prevent usage of a LHS
40297         from ASAN_MARK (POISON, &x, ...) coming to a PHI node.
40298         * gimplify.c (asan_poison_variables): Add attribute
40299         use_after_scope_memory to variables that really needs to live
40300         in memory.
40301         * tree-ssa.c (is_asan_mark_p): Do not rewrite into SSA when
40302         having the attribute.
40304 2017-01-23  Martin Liska  <mliska@suse.cz>
40306         * asan.c (create_asan_shadow_var): New function.
40307         (asan_expand_poison_ifn): Likewise.
40308         * asan.h (asan_expand_poison_ifn): New declaration.
40309         * internal-fn.c (expand_ASAN_POISON): Likewise.
40310         * internal-fn.def (ASAN_POISON): New builtin.
40311         * sanopt.c (pass_sanopt::execute): Expand
40312         asan_expand_poison_ifn.
40313         * tree-inline.c (copy_decl_for_dup_finish): Make function
40314         external.
40315         * tree-inline.h (copy_decl_for_dup_finish): Likewise.
40316         * tree-ssa.c (is_asan_mark_p): New function.
40317         (execute_update_addresses_taken): Rewrite local variables
40318         (identified just by use-after-scope as addressable) into SSA.
40320 2017-01-22  Gerald Pfeifer  <gerald@pfeifer.com>
40322         * doc/install.texi (Specific): opensource.apple.com uses https
40323         now. Remove trailing slash.
40325 2017-01-22  Gerald Pfeifer  <gerald@pfeifer.com>
40327         * README.Portability: Remove note on an Irix compatibility issue.
40329 2017-01-22  Dimitry Andric <dim@FreeBSD.org>
40331         * gcov.c (INCLUDE_ALGORITHM): Define.
40332         (INCLUDE_VECTOR): Define.
40333         No longer include <vector> and <algorithm> directly.
40335 2017-01-21  Gerald Pfeifer  <gerald@pfeifer.com>
40337         * doc/extend.texi (Thread-Local): Change www.akkadia.org reference
40338         to https.
40339         * doc/invoke.texi (Code Gen Options): Ditto.
40341 2017-01-21  Jan Hubicka  <hubicka@ucw.cz>
40343         PR lto/78407
40344         * cfg.c (update_bb_profile_for_threading): Fix updating of probablity.
40346 2017-01-21  Bernd Schmidt  <bschmidt@redhat.com>
40348         rtl-optimization/79125
40349         * cprop.c (local_cprop_pass): Handle cases where we make an
40350         unconditional trap.
40352 2017-01-20  Segher Boessenkool  <segher@kernel.crashing.org>
40354         PR target/61729
40355         PR target/77850
40356         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Adjust address to
40357         read from, for big endian.
40359 2017-01-20  Jiong Wang  <jiong.wang@arm.com>
40361         * config/aarch64/aarch64-builtins.c (aarch64_init_builtins): Register
40362         register pauth builtins for LP64 only.
40364 2017-01-20  Marek Polacek  <polacek@redhat.com>
40366         PR c/79152
40367         * gimplify.c (should_warn_for_implicit_fallthrough): Handle consecutive
40368         non-case labels.
40370 2017-01-20  Alexander Monakov  <amonakov@ispras.ru>
40372         * omp-expand.c (expand_omp_simd): Clear PROP_gimple_lomp_dev regardless
40373         of safelen status.
40374         * omp-offload.c (pass_omp_device_lower::gate): Use PROP_gimple_lomp_dev.
40375         * passes.c (dump_properties): Handle PROP_gimple_lomp_dev.
40376         * tree-inline.c (expand_call_inline): Propagate PROP_gimple_lomp_dev.
40378 2017-01-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
40380         PR target/71270
40381         * config/arm/arm.c (neon_valid_immediate): Reject vector constants
40382         in big-endian mode when they are not a single duplicated value.
40384 2017-01-20  Richard Biener  <rguenther@suse.de>
40386         * BASE-VER: Bump to 7.0.1.
40388 2017-01-20  Alexander Monakov  <amonakov@ispras.ru>
40390         * omp-low.c (omplow_simd_context): New struct.  Use it...
40391         (lower_rec_simd_input_clauses): ...here and...
40392         (lower_rec_input_clauses): ...here to hold common data.  Adjust all
40393         references to idx, lane, max_vf, is_simt.
40395 2017-01-20  Graham Markall  <graham.markall@embecosm.com>
40397         * config/arc/arc.h (LINK_SPEC): Use arclinux_nps emulation when
40398         mcpu=nps400.
40400 2017-01-20  Martin Jambor  <mjambor@suse.cz>
40402         * hsa.h: Renaed to hsa-common.h.  Adjusted a comment.
40403         * hsa.c: Renaed to hsa-common.c.  Change include of gt-hsa.h to
40404         gt-hsa-common.h.
40405         * Makefile.in (OBJS): Rename hsa.o to hsa-common.o.
40406         (GTFILES): Rename hsa.c to hsa-common.c.
40407         * hsa-brig.c: Change include of hsa.h to hsa-common.h.
40408         * hsa-dump.c: Likewise.
40409         * hsa-gen.c: Likewise.
40410         * hsa-regalloc.c: Likewise.
40411         * ipa-hsa.c: Likewise.
40412         * omp-expand.c: Likewise.
40413         * omp-low.c: Likewise.
40414         * toplev.c: Likewise.
40416 2017-01-20  Marek Polacek  <polacek@redhat.com>
40418         PR c/64279
40419         * doc/invoke.texi: Document -Wduplicated-branches.
40420         * fold-const.c (operand_equal_p): Handle MODIFY_EXPR, INIT_EXPR,
40421         COMPOUND_EXPR, PREDECREMENT_EXPR, PREINCREMENT_EXPR,
40422         POSTDECREMENT_EXPR, POSTINCREMENT_EXPR, CLEANUP_POINT_EXPR, EXPR_STMT,
40423         STATEMENT_LIST, and RETURN_EXPR.  For non-pure non-const functions
40424         return 0 only when not OEP_LEXICOGRAPHIC.
40425         (fold_build_cleanup_point_expr): Use the expression
40426         location when building CLEANUP_POINT_EXPR.
40427         * tree-core.h (enum operand_equal_flag): Add OEP_LEXICOGRAPHIC.
40428         * tree.c (add_expr): Handle error_mark_node.
40430 2017-01-20  Martin Liska  <mliska@suse.cz>
40432         PR lto/69188
40433         * tree-profile.c (init_ic_make_global_vars): Do not call
40434         finalize_decl.
40435         (gimple_init_gcov_profiler): Likewise.
40437 2017-01-20  Martin Liska  <mliska@suse.cz>
40439         PR ipa/71190
40440         * cgraph.h (maybe_create_reference): Remove argument and
40441         update comment.
40442         * cgraphclones.c (cgraph_node::create_virtual_clone): Remove one
40443         argument.
40444         * ipa-cp.c (create_specialized_node): Likewise.
40445         * symtab.c (symtab_node::maybe_create_reference): Handle
40446         VAR_DECLs and ADDR_EXPRs and select ipa_ref_use type.
40448 2017-01-20  Martin Liska  <mliska@suse.cz>
40450         * read-rtl-function.c (function_reader::create_function): Use
40451         build_decl instread of build_decl_stat.
40453 2017-01-20  Andrew Senkevich  <andrew.senkevich@intel.com>
40455         * config/i386/avx512bwintrin.h: Add k-mask registers shift intrinsics.
40456         * config/i386/avx512dqintrin.h: Ditto.
40457         * config/i386/avx512fintrin.h: Ditto.
40458         * config/i386/i386-builtin-types.def: Add new types.
40459         * config/i386/i386.c: Handle new types.
40460         * config/i386/i386-builtin.def (__builtin_ia32_kshiftliqi)
40461         (__builtin_ia32_kshiftlihi, __builtin_ia32_kshiftlisi)
40462         (__builtin_ia32_kshiftlidi, __builtin_ia32_kshiftriqi)
40463         (__builtin_ia32_kshiftrihi, __builtin_ia32_kshiftrisi)
40464         (__builtin_ia32_kshiftridi): New.
40465         * config/i386/sse.md (k<code><mode>): Rename *k<code><mode>.
40467 2017-01-19  Segher Boessenkool  <segher@kernel.crashing.org>
40469         PR target/78875
40470         PR target/79140
40471         * config/rs6000/rs6000.c (TARGET_STACK_PROTECT_GUARD): Unconditionally
40472         define to rs6000_init_stack_protect_guard.
40473         (rs6000_init_stack_protect_guard): New function.
40475 2017-01-19  Matthew Fortune  <matthew.fortune@imgtec.com>
40476             Yunqiang Su  <yunqiang.su@imgtec.com>
40478         * config.gcc (supported_defaults): Add madd4.
40479         (with_madd4): Add validation.
40480         (all_defaults): Add madd4.
40481         * config/mips/mips.opt (mmadd4): New option.
40482         * config/mips/mips.h (OPTION_DEFAULT_SPECS): Add a default for
40483         mmadd4.
40484         (TARGET_CPU_CPP_BUILTINS): Add builtin_define for
40485         __mips_no_madd4.
40486         (ISA_HAS_UNFUSED_MADD4): Gate with mips_madd4.
40487         (ISA_HAS_FUSED_MADD4): Likewise.
40488         * doc/invoke.texi (-mmadd4): Document the new option.
40489         * doc/install.texi (--with-madd4): Document the new option.
40491 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40493         * config/aarch64/aarch64-builtins.c (enum aarch64_builtins): New
40494         entries for AARCH64_PAUTH_BUILTIN_XPACLRI,
40495         AARCH64_PAUTH_BUILTIN_PACIA1716, AARCH64_PAUTH_BUILTIN_AUTIA1716.
40496         (aarch64_init_pauth_hint_builtins): New.
40497         (aarch64_init_builtins): Call aarch64_init_pauth_hint_builtins.
40498         (aarch64_expand_builtin): Expand new builtins.
40500 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40502         * reg-notes.def (CFA_TOGGLE_RA_MANGLE): New reg-note.
40503         * combine-stack-adj.c (no_unhandled_cfa): Handle
40504         REG_CFA_TOGGLE_RA_MANGLE.
40505         * dwarf2cfi.c (dwarf2out_frame_debug): Handle REG_CFA_TOGGLE_RA_MANGLE.
40506         * config/aarch64/aarch64.c (aarch64_expand_prologue): Generates DWARF
40507         info for return address signing.
40508         (aarch64_expand_epilogue): Likewise.
40510 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40512         * config/aarch64/aarch64-opts.h (aarch64_function_type): New enum.
40513         * config/aarch64/aarch64-protos.h
40514         (aarch64_return_address_signing_enabled): New declaration.
40515         * config/aarch64/aarch64.c (aarch64_return_address_signing_enabled):
40516         New function.
40517         (aarch64_expand_prologue): Sign return address before it's pushed onto
40518         stack.
40519         (aarch64_expand_epilogue): Authenticate return address fetched from
40520         stack.
40521         (aarch64_override_options): Sanity check for ILP32 and ISA level.
40522         (aarch64_attributes): New function attributes for "sign-return-address".
40523         * config/aarch64/aarch64.md (UNSPEC_AUTI1716, UNSPEC_AUTISP,
40524         UNSPEC_PACI1716, UNSPEC_PACISP, UNSPEC_XPACLRI): New unspecs.
40525         ("*do_return"): Generate combined instructions according to key index.
40526         ("<pauth_mnem_prefix>sp", "<pauth_mnem_prefix1716", "xpaclri"): New.
40527         * config/aarch64/iterators.md (PAUTH_LR_SP, PAUTH_17_16): New integer
40528         iterators.
40529         (pauth_mnem_prefix, pauth_hint_num_a): New integer attributes.
40530         * config/aarch64/aarch64.opt (msign-return-address=): New.
40531         * doc/extend.texi (AArch64 Function Attributes): Documents
40532         "sign-return-address=".
40533         * doc/invoke.texi (AArch64 Options): Documents "-msign-return-address=".
40535 2017-01-19  Matthew Fortune  <matthew.fortune@imgtec.com>
40537         * doc/invoke.texi: Add missing -mlxc1-sxc1 options to
40538         overall option summary.
40540 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40542         * config/aarch64/aarch64-arches.def: New entry for "armv8.3-a".
40543         * config/aarch64/aarch64.h (AARCH64_FL_V8_3, AARCH64_FL_FOR_ARCH8_3,
40544         AARCH64_ISA_V8_3, TARGET_ARMV8_3): New.
40545         * doc/invoke.texi (AArch64 Options): Document "armv8.3-a".
40547 2017-01-19  Michael Meissner  <meissner@linux.vnet.ibm.com>
40549         * config/rs6000/rs6000-cpus.def (ISA_3_0_MASKS_SERVER): Enable
40550         -mpower9-minmax by default for -mcpu=power9.
40551         (ISA_3_MASKS_IEEE): Require -mvsx-small-integer to enable IEEE
40552         128-bit floating point.
40554 2017-01-20  Alan Modra  <amodra@gmail.com>
40556         * config/rs6000/rs6000.md (cmpstrnsi, cmpstrsi): Fail if
40557         optimizing for size.
40559 2017-01-20  Alan Modra  <amodra@gmail.com>
40561         PR target/79144
40562         * config/rs6000/rs6000.c (expand_strn_compare): Get the asm name
40563         for strcmp and strncmp from corresponding builtin decl.
40565 2017-01-19  Uros Bizjak  <ubizjak@gmail.com>
40567         * config.gcc (x86_64-*-rtems*): Use i386/rtemself.h
40568         instead of i386/rtems-64.h.
40569         * config/i386/rtems-64.h: Remove.
40571 2017-01-19  Uros Bizjak  <ubizjak@gmail.com>
40573         PR target/78478
40574         Revert:
40575         2013-11-05  Uros Bizjak  <ubizjak@gmail.com>
40577         * config/i386/rtemself.h (LONG_DOUBLE_TYPE_SIZE): New define.
40579 2017-01-19  Tamar Christina  <tamar.christina@arm.com>
40581         * config/aarch64/aarch64.c (aarch64_simd_gen_const_vector_dup):
40582         Change int to HOST_WIDE_INT.
40583         * config/aarch64/aarch64-protos.h
40584         (aarch64_simd_gen_const_vector_dup): Likewise.
40585         * config/aarch64/aarch64-simd.md: Add copysign<mode>3.
40587 2017-01-19  David Malcolm  <dmalcolm@redhat.com>
40589         * langhooks-def.h (lhd_type_for_size): New decl.
40590         (LANG_HOOKS_TYPE_FOR_SIZE): Define as lhd_type_for_size.
40591         * langhooks.c (lhd_type_for_size): New function, taken from
40592         lto_type_for_size.
40594 2017-01-19  Pat Haugen  <pthaugen@us.ibm.com>
40596         * config/rs6000/power9.md (power9-alu): Remove 'cmp' type and add
40597         define_bypass for CR latency.
40598         (power9-cracked-alu): Update bypass latency and remove power9-branch.
40599         (power9-alu2): Add define_bypass for CR latency.
40600         (power9-cmp): New.
40601         (power9-mul): Update insn latency.
40602         (power9-mul-compare): Update insn latency, bypass latency and remove
40603         power9-branch.
40605 2016-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
40607         * config/aarch64/aarch64-protos.h (aarch64_nopcrelative_literal_loads):
40608         Delete.
40609         * config/aarch64/aarch64.md
40610         (aarch64_reload_movcp<GPF_TF:mode><P:mode>): Delete reference to
40611         aarch64_nopcrelative_literal_loads.
40612         (aarch64_reload_movcp<VALL:mode><P:mode>): Likewise.
40614 2017-01-19  Chenghua Xu  <paul.hua.gm@gmail.com>
40616         * config/mips/mips.h (ISA_HAS_FUSED_MADD4): Enable for
40617         TARGET_LOONGSON_3A.
40618         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_LOONGSON_3A.
40620 2017-01-19  Doug Gilmore  <doug.gilmore@imgtec.com>
40622         PR target/78176
40623         * config.gcc (supported_defaults): Add lxc1-sxc1.
40624         (with_lxc1_sxc1): Add validation.
40625         (all_defaults): Add lxc1-sxc1.
40626         * config/mips/mips.opt (mlxc1-sxc1): New option.
40627         * config/mips/mips.h (OPTION_DEFAULT_SPECS): Add a default for
40628         mlxc1-sxc1.
40629         (TARGET_CPU_CPP_BUILTINS): Add builtin_define for
40630         __mips_no_lxc1_sxc1.
40631         (ISA_HAS_LXC1_SXC1): Gate with mips_lxc1_sxc1.
40632         * doc/invoke.texi (-mlxc1-sxc1): Document the new option.
40633         * doc/install.texi (--with-lxc1-sxc1): Document the new option.
40635 2017-01-19  Richard Biener  <rguenther@suse.de>
40637         PR tree-optimization/72488
40638         * tree-ssa-sccvn.c (run_scc_vn): When we abort the VN make
40639         sure to restore SSA info.
40640         * tree-ssa.c (verify_ssa): Verify SSA info is not shared.
40642 2017-01-19  Richard Earnshaw  <rearnsha@arm.com>
40644         PR rtl-optimization/79121
40645         * expr.c (expand_expr_real_2, case LSHIFT_EXPR): Look at the signedness
40646         of the inner type when shifting an extended value.
40648 2017-01-17  Jan Hubicka  <hubicka@ucw.cz>
40650         PR lto/78407
40651         * symtab.c (symtab_node::equal_address_to): Fix comparing of
40652         interposable aliases.
40654 2017-01-18  Peter Bergner  <bergner@vnet.ibm.com>
40656         PR target/78516
40657         * config/rs6000/spe.md (mov_si<mode>_e500_subreg0): Fix constraints.
40658         Use the evmergelohi instruction.
40659         (mov_si<mode>_e500_subreg4_2_le): Likewise.
40660         (mov_sitf_e500_subreg8_2_be): Likewise.
40661         (mov_sitf_e500_subreg12_2_le): Likewise.
40662         (mov_si<mode>_e500_subreg0_2_le): Fix constraints.
40663         (mov_si<mode>_e500_subreg4_2_be): Likewise.
40664         (mov_sitf_e500_subreg8_2_le): Likewise.
40665         (mov_sitf_e500_subreg12_2_be): Likewise.
40667 2017-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40669         * config/rs6000/altivec.md (altivec_vbpermq): Change "type"
40670         attribute from vecsimple to vecperm.
40671         (altivec_vbpermq2): Likewise.
40673 2017-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40675         PR target/79040
40676         * config/rs6000/altivec.h: Fix typo of vec_cntlz to vec_cnttz.
40678 2017-01-18  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
40679         * config/rs6000/rs6000-protos.h (expand_strn_compare): Add arg.
40680         * config/rs6000/rs6000.c (expand_strn_compare): Add ability to expand
40681         strcmp. Fix bug where comparison didn't stop with zero byte. Fix
40682         case where N arg is SIZE_MAX.
40683         * config/rs6000/rs6000.md (cmpstrnsi): Args to expand_strn_compare.
40684         (cmpstrsi): Add pattern.
40686 2017-01-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
40688         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
40689         __builtin_vec_revb builtins.
40690         * config/rs6000/rs6000-builtins.def (P9V_BUILTIN_XXBRQ_V16QI): Add
40691         built-in functions to support generation of the ISA 3.0 XXBR<x>
40692         vector byte reverse instructions.
40693         (P9V_BUILTIN_XXBRQ_V1TI): Likewise.
40694         (P9V_BUILTIN_XXBRD_V2DI): Likewise.
40695         (P9V_BUILTIN_XXBRD_V2DF): Likewise.
40696         (P9V_BUILTIN_XXBGW_V4SI): Likewise.
40697         (P9V_BUILTIN_XXBGW_V4SF): Likewise.
40698         (P9V_BUILTIN_XXBGH_V8HI): Likewise.
40699         (P9V_BUILTIN_VEC_REVB): Likewise.
40700         * config/rs6000/vsx.md (p9_xxbrq_v1ti): New insns/expanders to
40701         generate the ISA 3.0 XXBR<x> vector byte reverse instructions.
40702         (p9_xxbrq_v16qi): Likewise.
40703         (p9_xxbrd_<mode>, VSX_D iterator): Likewise.
40704         (p9_xxbrw_<mode>, VSX_W iterator): Likewise.
40705         (p9_xxbrh_v8hi): Likewise.
40706         * config/rs6000/altivec.h (vec_revb): Define if ISA 3.0.
40707         * doc/extend.texi (RS/6000 Altivec Built-ins): Document the
40708         vec_revb built-in functions.
40710 2017-01-18  Uros Bizjak  <ubizjak@gmail.com>
40712         PR rtl-optimization/78952
40713         * config/i386/i386.md (any_extract): New code iterator.
40714         (*insvqi_2): Use any_extract for source operand.
40715         (*insvqi_3): Use any_shiftrt for source operand.
40717 2017-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
40719         * config/aarch64/aarch64.c (aarch64_sched_adjust_priority)
40720         New function.
40721         (TARGET_SCHED_ADJUST_PRIORITY): Define target hook.
40723 2017-01-18  Matthias Klose  <doko@ubuntu.com>
40725         * doc/install.texi: Allow default for --with-target-bdw-gc-include.
40727 2016-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40729         * config/rs6000/altivec.h (vec_bperm): Change #define.
40730         * config/rs6000/altivec.md (UNSPEC_VBPERMD): New enum constant.
40731         (altivec_vbpermq2): New define_insn.
40732         (altivec_vbpermd): Likewise.
40733         * config/rs6000/rs6000-builtin.def (VBPERMQ2): New monomorphic
40734         function interface.
40735         (VBPERMD): Likewise.
40736         (VBPERM): New polymorphic function interface.
40737         * config/rs6000/r6000-c.c (altivec_overloaded_builtins_table):
40738         Add entries for P9V_BUILTIN_VEC_VBPERM.
40739         * doc/extend.texi: Add interfaces for vec_bperm.
40741 2017-01-18  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
40743         * config/s390/s390-c.c (s390_expand_overloaded_builtin): Downcase
40744         first letter of error messages.
40745         (s390_resolve_overloaded_builtin): Likewise.
40746         * config/s390/s390.c (s390_expand_builtin): Likewise.
40747         (s390_invalid_arg_for_unprototyped_fn): Likewise.
40748         (s390_valid_target_attribute_inner_p): Likewise.
40749         * config/s390/s390.md ("tabort"): Likewise.
40751 2017-01-18  Toma Tabacu  <toma.tabacu@imgtec.com>
40753         * config/mips/mips.h (ISA_HAS_DIV3): Remove unused macro.
40754         (ISA_AVOID_DIV_HILO): New macro.
40755         (ISA_HAS_DIV): Use new ISA_AVOID_DIV_HILO macro.
40756         (ISA_HAS_DDIV): Likewise.
40758 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40760         * doc/invoke.texi (fabi-version): Correct number of occurrences.
40762 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40764         * doc/invoke.texi (fabi-version): Spelling fix.
40766 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40768         PR c++/70182
40769         * doc/invoke.texi (fabi-version): Mention mangling fix for
40770         operator names.
40772 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40774         PR c++/77489
40775         * doc/invoke.texi (fabi-version): Document discriminator mangling.
40777 2017-01-17  Segher Boessenkool  <segher@kernel.crashing.org>
40779         PR target/78875
40780         * config/rs6000/rs6000-opts.h (stack_protector_guard): New enum.
40781         * config/rs6000/rs6000.c (rs6000_option_override_internal): Handle
40782         the new options.
40783         * config/rs6000/rs6000.md (stack_protect_set): Handle the new more
40784         flexible settings.
40785         (stack_protect_test): Ditto.
40786         * config/rs6000/rs6000.opt (mstack-protector-guard=,
40787         mstack-protector-guard-reg=, mstack-protector-guard-offset=): New
40788         options.
40789         * doc/invoke.texi (Option Summary) [RS/6000 and PowerPC Options]:
40790         Add -mstack-protector-guard=, -mstack-protector-guard-reg=, and
40791         -mstack-protector-guard-offset=.
40792         (RS/6000 and PowerPC Options): Ditto.
40794 2017-01-17  Uros Bizjak  <ubizjak@gmail.com>
40796         * config/i386/i386.h (MASK_CLASS_P): New define.
40797         * config/i386/i386.c (inline_secondary_memory_needed): Ensure that
40798         there are no registers from different register sets also when
40799         mask registers are used.  Update function comment.
40800         * config/i386/i386.md (*movsi_internal): Split (*k/*krm) alternative
40801         to (*k/*r) and (*k/*km) alternatives.
40803 2017-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
40805         * config/aarch64/aarch64.md (eh_return): Remove pattern and splitter.
40806         * config/aarch64/aarch64.h (AARCH64_EH_STACKADJ_REGNUM): Remove.
40807         (EH_RETURN_HANDLER_RTX): New define.
40808         * config/aarch64/aarch64.c (aarch64_frame_pointer_required):
40809         Force frame pointer in EH return functions.
40810         (aarch64_expand_epilogue): Add barrier for eh_return.
40811         (aarch64_final_eh_return_addr): Remove.
40812         (aarch64_eh_return_handler_rtx): New function.
40813         * config/aarch64/aarch64-protos.h (aarch64_final_eh_return_addr):
40814         Remove.
40815         (aarch64_eh_return_handler_rtx): New prototype.
40817 2017-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40819         * config/rs6000/altivec.h (vec_rlmi): New #define.
40820         (vec_vrlnm): Likewise.
40821         (vec_rlnm): Likewise.
40822         * config/rs6000/altivec.md (UNSPEC_VRLMI): New UNSPEC enum value.
40823         (UNSPEC_VRLNM): Likewise.
40824         (VIlong): New mode iterator.
40825         (altivec_vrl<VI_char>mi): New define_insn.
40826         (altivec_vrl<VI_char>nm): Likewise.
40827         * config/rs6000/rs6000-builtin.def (VRLWNM): New monomorphic
40828         function entry.
40829         (VRLDNM): Likewise.
40830         (RLNM): New polymorphic function entry.
40831         (VRLWMI): New monomorphic function entry.
40832         (VRLDMI): Likewise.
40833         (RLMI): New polymorphic function entry.
40834         * config/rs6000/r6000-c.c (altivec_overloaded_builtin_table): Add
40835         new entries for P9V_BUILTIN_VEC_RLMI and P9V_BUILTIN_VEC_RLNM.
40836         * doc/extend.texi: Add description of vec_rlmi, vec_rlnm, and
40837         vec_vrlnm.
40839 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
40841         PR debug/78839
40842         * dwarf2out.c (field_byte_offset): Restore the
40843         PCC_BITFIELD_TYPE_MATTERS behavior for INTEGER_CST DECL_FIELD_OFFSET
40844         and DECL_FIELD_BIT_OFFSET.  Use fold_build2 instead of build2 + fold.
40845         (analyze_variants_discr, gen_variant_part): Use fold_build2 instead
40846         of build2 + fold.
40848 2017-01-17  Eric Botcazou  <ebotcazou@adacore.com>
40850         PR ada/67205
40851         * config/aarch64/aarch64.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): Define
40853 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
40855         PR debug/71669
40856         * dwarf2out.c (add_data_member_location_attribute): For constant
40857         offset bitfield emit for -gdwarf-5 DW_AT_data_bit_offset attribute
40858         instead of DW_AT_data_member_location, DW_AT_bit_offset and
40859         DW_AT_byte_size attributes.
40861 2017-01-17  Eric Botcazou  <ebotcazou@adacore.com>
40863         * config/rs6000/rs6000.c (rs6000_emit_move): Also use a TOC reference
40864         after forcing to constant memory when the code model is medium.
40866 2017-01-17  Julia Koval  <julia.koval@intel.com>
40868         PR target/76731
40869         * config/i386/avx512fintrin.h
40870         (_mm512_i32gather_ps): Change __addr type to void const*.
40871         (_mm512_mask_i32gather_ps): Ditto.
40872         (_mm512_i32gather_pd): Ditto.
40873         (_mm512_mask_i32gather_pd): Ditto.
40874         (_mm512_i64gather_ps): Ditto.
40875         (_mm512_mask_i64gather_ps): Ditto.
40876         (_mm512_i64gather_pd): Ditto.
40877         (_mm512_mask_i64gather_pd): Ditto.
40878         (_mm512_i32gather_epi32): Ditto.
40879         (_mm512_mask_i32gather_epi32): Ditto.
40880         (_mm512_i32gather_epi64): Ditto.
40881         (_mm512_mask_i32gather_epi64): Ditto.
40882         (_mm512_i64gather_epi32): Ditto.
40883         (_mm512_mask_i64gather_epi32): Ditto.
40884         (_mm512_i64gather_epi64): Ditto.
40885         (_mm512_mask_i64gather_epi64): Ditto.
40886         (_mm512_i32scatter_ps): Change __addr type to void*.
40887         (_mm512_mask_i32scatter_ps): Ditto.
40888         (_mm512_i32scatter_pd): Ditto.
40889         (_mm512_mask_i32scatter_pd): Ditto.
40890         (_mm512_i64scatter_ps): Ditto.
40891         (_mm512_mask_i64scatter_ps): Ditto.
40892         (_mm512_i64scatter_pd): Ditto.
40893         (_mm512_mask_i64scatter_pd): Ditto.
40894         (_mm512_i32scatter_epi32): Ditto.
40895         (_mm512_mask_i32scatter_epi32): Ditto.
40896         (_mm512_i32scatter_epi64): Ditto.
40897         (_mm512_mask_i32scatter_epi64): Ditto.
40898         (_mm512_i64scatter_epi32): Ditto.
40899         (_mm512_mask_i64scatter_epi32): Ditto.
40900         (_mm512_i64scatter_epi64): Ditto.
40901         (_mm512_mask_i64scatter_epi64): Ditto.
40902         * config/i386/avx512pfintrin.h
40903         (_mm512_mask_prefetch_i32gather_pd): Change __addr type to void const*.
40904         (_mm512_mask_prefetch_i32gather_ps): Ditto.
40905         (_mm512_mask_prefetch_i64gather_pd): Ditto.
40906         (_mm512_mask_prefetch_i64gather_ps): Ditto.
40907         (_mm512_prefetch_i32scatter_pd): Change __addr type to void*.
40908         (_mm512_prefetch_i32scatter_ps): Ditto.
40909         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
40910         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
40911         (_mm512_prefetch_i64scatter_pd): Ditto.
40912         (_mm512_prefetch_i64scatter_ps): Ditto.
40913         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
40914         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
40915         * config/i386/avx512vlintrin.h
40916         (_mm256_mmask_i32gather_ps): Change __addr type to void const*.
40917         (_mm_mmask_i32gather_ps): Ditto.
40918         (_mm256_mmask_i32gather_pd): Ditto.
40919         (_mm_mmask_i32gather_pd): Ditto.
40920         (_mm256_mmask_i64gather_ps): Ditto.
40921         (_mm_mmask_i64gather_ps): Ditto.
40922         (_mm256_mmask_i64gather_pd): Ditto.
40923         (_mm_mmask_i64gather_pd): Ditto.
40924         (_mm256_mmask_i32gather_epi32): Ditto.
40925         (_mm_mmask_i32gather_epi32): Ditto.
40926         (_mm256_mmask_i32gather_epi64): Ditto.
40927         (_mm_mmask_i32gather_epi64): Ditto.
40928         (_mm256_mmask_i64gather_epi32): Ditto.
40929         (_mm_mmask_i64gather_epi32): Ditto.
40930         (_mm256_mmask_i64gather_epi64): Ditto.
40931         (_mm_mmask_i64gather_epi64): Ditto.
40932         (_mm256_i32scatter_ps): Change __addr type to void*.
40933         (_mm256_mask_i32scatter_ps): Ditto.
40934         (_mm_i32scatter_ps): Ditto.
40935         (_mm_mask_i32scatter_ps): Ditto.
40936         (_mm256_i32scatter_pd): Ditto.
40937         (_mm256_mask_i32scatter_pd): Ditto.
40938         (_mm_i32scatter_pd): Ditto.
40939         (_mm_mask_i32scatter_pd): Ditto.
40940         (_mm256_i64scatter_ps): Ditto.
40941         (_mm256_mask_i64scatter_ps): Ditto.
40942         (_mm_i64scatter_ps): Ditto.
40943         (_mm_mask_i64scatter_ps): Ditto.
40944         (_mm256_i64scatter_pd): Ditto.
40945         (_mm256_mask_i64scatter_pd): Ditto.
40946         (_mm_i64scatter_pd): Ditto.
40947         (_mm_mask_i64scatter_pd): Ditto.
40948         (_mm256_i32scatter_epi32): Ditto.
40949         (_mm256_mask_i32scatter_epi32): Ditto.
40950         (_mm_i32scatter_epi32): Ditto.
40951         (_mm_mask_i32scatter_epi32): Ditto.
40952         (_mm256_i32scatter_epi64): Ditto.
40953         (_mm256_mask_i32scatter_epi64): Ditto.
40954         (_mm_i32scatter_epi64): Ditto.
40955         (_mm_mask_i32scatter_epi64): Ditto.
40956         (_mm256_i64scatter_epi32): Ditto.
40957         (_mm256_mask_i64scatter_epi32): Ditto.
40958         (_mm_i64scatter_epi32): Ditto.
40959         (_mm_mask_i64scatter_epi32): Ditto.
40960         (_mm256_i64scatter_epi64): Ditto.
40961         (_mm256_mask_i64scatter_epi64): Ditto.
40962         (_mm_i64scatter_epi64): Ditto.
40963         (_mm_mask_i64scatter_epi64): Ditto.
40964         * config/i386/i386-builtin-types.def (V16SF_V16SF_PCFLOAT_V16SI_HI_INT)
40965         (V8DF_V8DF_PCDOUBLE_V8SI_QI_INT, V8SF_V8SF_PCFLOAT_V8DI_QI_INT)
40966         (V8DF_V8DF_PCDOUBLE_V8DI_QI_INT, V16SI_V16SI_PCINT_V16SI_HI_INT)
40967         (V8DI_V8DI_PCINT64_V8SI_QI_INT, V8SI_V8SI_PCINT_V8DI_QI_INT)
40968         (V8DI_V8DI_PCINT64_V8DI_QI_INT, V2DF_V2DF_PCDOUBLE_V4SI_QI_INT)
40969         (V4DF_V4DF_PCDOUBLE_V4SI_QI_INT, V2DF_V2DF_PCDOUBLE_V2DI_QI_INT)
40970         (V4DF_V4DF_PCDOUBLE_V4DI_QI_INT, V4SF_V4SF_PCFLOAT_V4SI_QI_INT)
40971         (V8SF_V8SF_PCFLOAT_V8SI_QI_INT, V4SF_V4SF_PCFLOAT_V2DI_QI_INT)
40972         (V4SF_V4SF_PCFLOAT_V4DI_QI_INT, V2DI_V2DI_PCINT64_V4SI_QI_INT)
40973         (V4DI_V4DI_PCINT64_V4SI_QI_INT, V2DI_V2DI_PCINT64_V2DI_QI_INT)
40974         (V4DI_V4DI_PCINT64_V4DI_QI_INT, V4SI_V4SI_PCINT_V4SI_QI_INT)
40975         (V8SI_V8SI_PCINT_V8SI_QI_INT, V4SI_V4SI_PCINT_V2DI_QI_INT)
40976         (V4SI_V4SI_PCINT_V4DI_QI_INT, VOID_PFLOAT_HI_V16SI_V16SF_INT)
40977         (VOID_PFLOAT_QI_V8SI_V8SF_INT, VOID_PFLOAT_QI_V4SI_V4SF_INT)
40978         (VOID_PDOUBLE_QI_V8SI_V8DF_INT, VOID_PDOUBLE_QI_V4SI_V4DF_INT)
40979         (VOID_PDOUBLE_QI_V4SI_V2DF_INT, VOID_PFLOAT_QI_V8DI_V8SF_INT)
40980         (VOID_PFLOAT_QI_V4DI_V4SF_INT, VOID_PFLOAT_QI_V2DI_V4SF_INT)
40981         (VOID_PDOUBLE_QI_V8DI_V8DF_INT, VOID_PDOUBLE_QI_V4DI_V4DF_INT)
40982         (VOID_PDOUBLE_QI_V2DI_V2DF_INT, VOID_PINT_HI_V16SI_V16SI_INT)
40983         (VOID_PINT_QI_V8SI_V8SI_INT, VOID_PINT_QI_V4SI_V4SI_INT)
40984         (VOID_PLONGLONG_QI_V8SI_V8DI_INT, VOID_PLONGLONG_QI_V4SI_V4DI_INT)
40985         (VOID_PLONGLONG_QI_V4SI_V2DI_INT, VOID_PINT_QI_V8DI_V8SI_INT)
40986         (VOID_PINT_QI_V4DI_V4SI_INT, VOID_PINT_QI_V2DI_V4SI_INT)
40987         (VOID_PLONGLONG_QI_V8DI_V8DI_INT, VOID_QI_V8SI_PCINT64_INT_INT)
40988         (VOID_PLONGLONG_QI_V4DI_V4DI_INT, VOID_PLONGLONG_QI_V2DI_V2DI_INT)
40989         (VOID_HI_V16SI_PCINT_INT_INT, VOID_QI_V8DI_PCINT64_INT_INT)
40990         (VOID_QI_V8DI_PCINT_INT_INT): Remove.
40991         (V16SF_V16SF_PCVOID_V16SI_HI_INT,  V8DF_V8DF_PCVOID_V8SI_QI_INT)
40992         (V8SF_V8SF_PCVOID_V8DI_QI_INT, V8DF_V8DF_PCVOID_V8DI_QI_INT)
40993         (V16SI_V16SI_PCVOID_V16SI_HI_INT, V8DI_V8DI_PCVOID_V8SI_QI_INT)
40994         (V8SI_V8SI_PCVOID_V8DI_QI_INT, V8DI_V8DI_PCVOID_V8DI_QI_INT)
40995         (VOID_PVOID_HI_V16SI_V16SF_INT, VOID_PVOID_QI_V8SI_V8DF_INT)
40996         (VOID_PVOID_QI_V8DI_V8SF_INT, VOID_PVOID_QI_V8DI_V8DF_INT)
40997         (VOID_PVOID_HI_V16SI_V16SI_INT, VOID_PVOID_QI_V8SI_V8DI_INT)
40998         (VOID_PVOID_QI_V8DI_V8SI_INT, VOID_PVOID_QI_V8DI_V8DI_INT)
40999         (V2DF_V2DF_PCVOID_V4SI_QI_INT, V4DF_V4DF_PCVOID_V4SI_QI_INT)
41000         (V2DF_V2DF_PCVOID_V2DI_QI_INT, V4DF_V4DF_PCVOID_V4DI_QI_INT
41001         (V4SF_V4SF_PCVOID_V4SI_QI_INT, V8SF_V8SF_PCVOID_V8SI_QI_INT)
41002         (V4SF_V4SF_PCVOID_V2DI_QI_INT, V4SF_V4SF_PCVOID_V4DI_QI_INT)
41003         (V2DI_V2DI_PCVOID_V4SI_QI_INT, V4DI_V4DI_PCVOID_V4SI_QI_INT)
41004         (V2DI_V2DI_PCVOID_V2DI_QI_INT, V4DI_V4DI_PCVOID_V4DI_QI_INT)
41005         (V4SI_V4SI_PCVOID_V4SI_QI_INT, V8SI_V8SI_PCVOID_V8SI_QI_INT)
41006         (V4SI_V4SI_PCVOID_V2DI_QI_INT, V4SI_V4SI_PCVOID_V4DI_QI_INT)
41007         (VOID_PVOID_QI_V8SI_V8SF_INT, VOID_PVOID_QI_V4SI_V4SF_INT)
41008         (VOID_PVOID_QI_V4SI_V4DF_INT, VOID_PVOID_QI_V4SI_V2DF_INT)
41009         (VOID_PVOID_QI_V4DI_V4SF_INT, VOID_PVOID_QI_V2DI_V4SF_INT)
41010         (VOID_PVOID_QI_V4DI_V4DF_INT, VOID_PVOID_QI_V2DI_V2DF_INT)
41011         (VOID_PVOID_QI_V8SI_V8SI_INT, VOID_PVOID_QI_V4SI_V4SI_INT)
41012         (VOID_PVOID_QI_V4SI_V4DI_INT, VOID_PVOID_QI_V4SI_V2DI_INT)
41013         (VOID_PVOID_QI_V4DI_V4SI_INT, VOID_PVOID_QI_V2DI_V4SI_INT)
41014         (VOID_PVOID_QI_V4DI_V4DI_INT, VOID_PVOID_QI_V2DI_V2DI_INT)
41015         (VOID_QI_V8SI_PCVOID_INT_INT, VOID_HI_V16SI_PCVOID_INT_INT)
41016         (VOID_QI_V8DI_PCVOID_INT_INT): Add.
41017         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Adjust builtin
41018         definitions accordingly.
41020 2017-01-17  Kito Cheng  <kito.cheng@gmail.com>
41021             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
41023         PR target/79079
41024         * internal-fn.c (expand_mul_overflow): Use convert_modes instead of
41025         gen_lowpart.
41027 2017-01-17  Vladimir Makarov  <vmakarov@redhat.com>
41029         PR target/79058
41030         * ira-conflicts.c (ira_build_conflicts): Update total conflict
41031         hard regs for inner regno.
41033 2017-01-17  Martin Liska  <mliska@suse.cz>
41035         PR ipa/71207
41036         * ipa-polymorphic-call.c (contains_type_p): Fix wrong
41037         assumption and add comment.
41039 2017-01-17  Nathan Sidwell  <nathan@acm.org>
41041         * ipa-visibility.c (localize_node): New function, broken out of ...
41042         (function_and_variable_visibility): ... here. Call it.
41044 2017-01-17  Jan Hubicka  <hubicka@ucw.cz>
41046         PR middle-end/77445
41047         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges):
41048         correctly set frequency of oudgoing edge.
41049         (duplicate_thread_path): Fix profile updating.
41051 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
41053         PR other/79046
41054         * configure.ac: Add GCC_BASE_VER.
41055         * Makefile.in (version): Use @get_gcc_base_ver@ instead of cat to get
41056         version from BASE-VER file.
41057         (CFLAGS-gcc.o): Add -DBASEVER=$(BASEVER_s).
41058         (gcc.o): Depend on $(BASEVER).
41059         * common.opt (dumpfullversion): New option.
41060         * gcc.c (driver_handle_option): Handle OPT_dumpfullversion.
41061         * doc/invoke.texi: Document -dumpfullversion.
41062         * doc/install.texi: Document --with-gcc-major-version-only.
41063         * configure: Regenerated.
41065 2017-01-17  Richard Biener  <rguenther@suse.de>
41067         PR tree-optimization/71433
41068         * tree-vrp.c (register_new_assert_for): Merge same asserts
41069         on all incoming edges.
41070         (process_assert_insertions_for): Handle insertions at the
41071         beginning of BBs.
41073 2017-01-17  Gerald Pfeifer  <gerald@pfeifer.com>
41075         * config/i386/cygwin.h (LIBGCJ_SONAME): Remove.
41076         * config/i386/mingw32.h (LIBGCJ_SONAME): Remove.
41078 2017-01-17  Kaz Kojima  <kkojima@gcc.gnu.org>
41080         PR target/78633
41081         * config/sh/sh.md (cmpeqsi_t+1): Call copy_rtx to avoid invalid
41082         RTL sharing.
41084 2017-01-17  Alan Modra  <amodra@gmail.com>
41086         PR target/79066
41087         * config/rs6000/rs6000.md (elf_high, elf_low): Disable when pic.
41088         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack): Don't allow
41089         symbolic stack limit when pic.
41091 2017-01-16  Martin Sebor  <msebor@redhat.com>
41093         PR tree-optimization/78608
41094         * gimple-ssa-sprintf.c (tree_digits): Avoid negating TYPE_MIN.
41096 2017-01-16  Jeff Law  <law@redhat.com>
41098         Revert:
41099         2016-12-02  Tadek Kijkowski  <tkijkowski@gmail.com>
41100         * Makefile.in (PREPROCESSOR_DEFINES): Add a level of indirection
41101         for several include directories that may be relative to sysroot.
41102         * config/i386/x-mingw32 (gplus_includedir): Define.
41103         (gplus_tool_includedir, gplus_backward_include_dir): Likewise.
41104         (native_system_includedir): Likewise.
41105         * config/i386/mingw32.h (STANDARD_STARTFILE_PREFIX_1): Do not
41106         override if TARGET_SYSTEM_ROOT is defined.
41107         (NATIVE_SYSTEM_HEADER_DIR): Likewise.
41109         PR tree-optimization/79090
41110         PR tree-optimization/33562
41111         PR tree-optimization/61912
41112         PR tree-optimization/77485
41113         * tree-ssa-dse.c (compute_trims): Accept STMT argument.  Dump STMT
41114         and computed trims into the dump file.
41116 2017-01-17  Uros Bizjak  <ubizjak@gmail.com>
41118         * config/i386/i386.h (LIMIT_RELOAD_CLASS): Remove.
41120 2017-01-16  Jakub Jelinek  <jakub@redhat.com>
41122         PR c/79089
41123         * gimplify.c (gimplify_init_constructor): If want_value and
41124         object == lhs, unshare lhs to avoid invalid tree sharing.  Formatting
41125         fix.
41127         PR target/79080
41128         * loop-doloop.c (doloop_modify): Call unshare_all_rtl_in_chain on
41129         sequence.  Formatting fixes.
41130         (doloop_optimize): Formatting fixes.
41132         PR driver/49726
41133         * gcc.c (debug_level_greater_than_spec_func): New function.
41134         (static_spec_functions): Add debug-level-gt spec function.
41135         (ASM_DEBUG_SPEC, cpp_options): Use %:debug-level-gt(0) instead of
41136         !g0.
41137         * config/darwin.h (DSYMUTIL_SPEC, ASM_DEBUG_SPEC): Likewise.
41138         * config/darwin9.h (DSYMUTIL_SPEC, ASM_DEBUG_SPEC): Likewise.
41139         * common.opt (g, gcoff, gdwarf, gdwarf-, ggdb, gno-pubnames,
41140         gpubnames, ggnu-pubnames, gno-record-gcc-switches,
41141         grecord-gcc-switches, gno-strict-dwarf, gstrict-dwarf, gstabs,
41142         gstabs+, gtoggle, gvms, gxcoff, gxcoff+): Add Driver flag.
41144 2017-01-16  Uros Bizjak  <ubizjak@gmail.com>
41146         * config/i386/i386.h (HARD_REGNO_CALLER_SAVE_MODE): Apply HImode and
41147         QImode fixups to general and mask registers only.
41149 2017-01-16  Carl Love  <cel@us.ibm.com>
41151         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Add support
41152         for built-in functions
41153         vector signed char vec_nabs (vector signed char)
41154         vector signed short vec_nabs (vector signed short)
41155         vector signed int vec_nabs (vector signed int)
41156         vector signed long long vec_nabs (vector signed long long)
41157         vector float vec_nabs (vector float)
41158         vector double vec_nabs (vector double)
41159         * config/rs6000/rs6000-builtin.def: Add definitions for NABS functions
41160         and NABS overload.
41161         * config/rs6000/altivec.md: New define_expand nabs<mode>2 types
41162         * config/rs6000/altivec.h: New define for vec_nabs built-in function.
41163         * doc/extend.texi: Update the documentation file for the new built-in
41164         functions.
41166 2017-01-16  Martin Sebor  <msebor@redhat.com>
41168         * gimple-ssa-sprintf.c (format_directive): Correct a typo in a warning
41169         message.
41171 2017-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
41173         * config/rs6000/rs6000.c (rtx_is_swappable_p): Change
41174         UNSPEC_VSX__XXSPLTD to require special splat handling.
41176 2017-01-16  David Malcolm  <dmalcolm@redhat.com>
41178         PR bootstrap/78616
41179         * system.h: Poison strndup.
41181 2017-01-16  Alan Modra  <amodra@gmail.com>
41183         PR target/79098
41184         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Don't
41185         use a switch.
41187 2017-01-16  Georg-Johann Lay  <avr@gjlay.de>
41189         * config/avr/avr.h (BRANCH_COST) [reload_completed]: Increase by 4.
41191 2017-01-15  Uros Bizjak  <ubizjak@gmail.com>
41193         * config/i386/i386.c (ix86_legitimate_combined_insn): Do not
41194         call recog here.  Assert that INSN_CODE (insn) is non-negative.
41196 2017-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
41198         PR target/72749
41199         * cfgrtl.c (rtl_split_edge): Also patch jump insns that jump to the
41200         fallthrough.
41201         * haifa-sched.c (dump_insn_stream): Don't crash if there is a label
41202         in the currently scheduled RTL fragment.
41204 2017-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
41206         PR rtl-optimization/78751
41207         * ifcvt.c (find_cond_trap): If we generated a non-existing insn,
41208         give up.
41210 2017-01-14  Jeff Law  <law@redhat.com>
41212         PR tree-optimization/79090
41213         * tree-ssa-dse.c (valid_ao_ref_for_dse): Reject zero length and
41214         variable length stores.
41215         (compute_trims): Delete dead assignment to *trim_tail.
41216         (dse_dom_walker::dse_optimize_stmt): Optimize mem* calls with
41217         zero length.
41219 2017-01-14  Bernd Schmidt  <bschmidt@redhat.com>
41221         PR rtl-optimization/78626
41222         PR rtl-optimization/78727
41223         * cprop.c (one_cprop_pass): Collect unconditional traps in the middle
41224         of a block, and split such blocks after everything else is finished.
41226 2017-01-14  Alan Modra  <amodra@gmail.com>
41228         PR target/72749
41229         * combine.c (recog_for_combine_1): Set INSN_CODE before calling
41230         target legitimate_combined_insn.
41231         * config/rs6000/rs6000.c (TARGET_LEGITIMATE_COMBINED_INSN): Define.
41232         (rs6000_legitimate_combined_insn): New function.
41233         * config/rs6000/rs6000.md (UNSPEC_DOLOOP): Delete, and remove
41234         all uses.
41235         (ctr<mode>_internal3): Rename from *ctr<mode>_internal5.
41236         (ctr<mode>_internal4): Rename from *ctr<mode>_internal6.
41237         (ctr<mode>_internal1, ctr<mode>_internal2): Remove '*' from name.
41239 2017-01-14  Gerald Pfeifer  <gerald@pfeifer.com>
41241         * doc/frontends.texi (G++ and GCC): Remove references to Java.
41243 2017-01-13  Jeff Law  <law@redhat.com>
41245         PR tree-optimization/33562
41246         PR tree-optimization/61912
41247         PR tree-optimization/77485
41248         * tree-ssa-dse.c (delete_dead_call): Accept gsi rather than
41249         a statement.
41250         (delete_dead_assignment): Likewise.
41251         (dse_dom_walker::dse_optimize_stmt): Pass in the gsi rather than
41252         statement to delete_dead_call and delete_dead_assignment.
41254 2017-01-13  David Malcolm  <dmalcolm@redhat.com>
41256         PR c/78304
41257         * substring-locations.c (format_warning_va): Strengthen case 1 so
41258         that both endpoints of the substring must be within the format
41259         range for just the substring to be printed.
41261 2017-01-13  Uros Bizjak  <ubizjak@gmail.com>
41263         * config/i386/i386.opt (msgx): Use ix86_isa_flags2 variable.
41264         * config/i386/i386.c (ix86_target_string): Add missing options
41265         to isa_opts and reorder options by implied ISAs.  Rename isa_opts2 to
41266         isa2_opts, ix86_flag_opts to flag2_opts, ix86_target_other to
41267         flags_other and ix86_target_other to flags2_other.  Display unknown
41268         isa2 options.
41269         (ix86_valid_target_attribute_inner_p): Add missing options and
41270         reorder options by implied ISAs, as in ix86_target_string.
41272 2017-01-13  Richard Sandiford  <richard.sandiford@arm.com>
41274         * hash-table.h (hash_table::too_empty_p): New function.
41275         (hash_table::expand): Use it.
41276         (hash_table::traverse): Likewise.
41277         (hash_table::empty_slot): Use sizeof (value_type) instead of
41278         sizeof (PTR) to convert bytes to elements.  Shrink the table
41279         if the current size is excessive for the current number of
41280         elements.
41282 2017-01-13  Richard Sandiford  <richard.sandiford@arm.com>
41284         * ira-costs.c (record_reg_classes): Break from the inner loop
41285         early once alt_fail is known to be true.  Update outer loop
41286         handling accordingly.
41288 2017-01-13  Jeff Law  <law@redhat.com>
41290         * tree-ssa-dse.c (decrement_count): New function.
41291         (increment_start_addr, maybe_trim_memstar_call): Likewise.
41292         (dse_dom_walker::optimize_stmt): Call maybe_trim_memstar_call directly
41293         when we know the partially dead statement is a mem* function.
41295         PR tree-optimization/61912
41296         PR tree-optimization/77485
41297         * tree-ssa-dse.c: Include expr.h.
41298         (maybe_trim_constructor_store): New function.
41299         (maybe_trim_partially_dead_store): Call maybe_trim_constructor_store.
41301         PR tree-optimization/33562
41302         PR tree-optimization/61912
41303         PR tree-optimization/77485
41304         * doc/invoke.texi: Document new dse-max-object-size param.
41305         * params.def (PARM_DSE_MAX_OBJECT_SIZE): New PARAM.
41306         * tree-ssa-dse.c: Include params.h.
41307         (dse_store_status): New enum.
41308         (initialize_ao_ref_for_dse): New, partially extracted from
41309         dse_optimize_stmt.
41310         (valid_ao_ref_for_dse, normalize_ref): New.
41311         (setup_live_bytes_from_ref, compute_trims): Likewise.
41312         (clear_bytes_written_by, maybe_trim_complex_store): Likewise.
41313         (maybe_trim_partially_dead_store): Likewise.
41314         (maybe_trim_complex_store): Likewise.
41315         (dse_classify_store): Renamed from dse_possibly_dead_store_p.
41316         Track what bytes live from the original store.  Return tri-state
41317         for dead, partially dead or live.
41318         (dse_dom_walker): Add constructor, destructor and new private members.
41319         (delete_dead_call, delete_dead_assignment): New extracted from
41320         dse_optimize_stmt.
41321         (dse_optimize_stmt): Make a member of dse_dom_walker.
41322         Use initialize_ao_ref_for_dse.
41324         PR tree-optimization/33562
41325         PR tree-optimization/61912
41326         PR tree-optimization/77485
41327         * sbitmap.h (bitmap_count_bits): Prototype.
41328         (bitmap_clear_range, bitmap_set_range): Likewise.
41329         * sbitmap.c (bitmap_clear_range): New function.
41330         (bitmap_set_range, sbitmap_popcount, bitmap_count_bits): Likewise.
41332 2017-01-13  Martin Liska  <mliska@suse.cz>
41334         PR ipa/79043
41335         * function.c (set_cfun): Add new argument force.
41336         * function.h (set_cfun): Likewise.
41337         * ipa-inline-transform.c (inline_call): Use the function when
41338         strict alising from is dropped for function we inline to.
41340 2017-01-13  Richard Biener  <rguenther@suse.de>
41342         * tree-pretty-print.c (dump_generic_node): Fix inverted condition
41343         for dumping GIMPLE INTEGER_CSTs.
41345 2017-01-13  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
41347         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define __STDC_VERSION__
41348         to 201112L since C++17.
41350 2017-01-13  Maxim Ostapenko  <m.ostapenko@samsung.com>
41352         PR sanitizer/78887
41353         * asan.c (asan_needs_odr_indicator_p): Don't emit ODR indicators
41354         if -fsanitize=kernel-address is present.
41356 2017-01-13  Richard Biener  <rguenther@suse.de>
41358         * tree-pretty-print.c (dump_generic_node): Dump INTEGER_CSTs
41359         as _Literal ( type ) number in case usual suffixes do not
41360         preserve all information.
41362 2017-01-13  Richard Biener  <rguenther@suse.de>
41364         PR tree-optimization/77283
41365         * gimple-ssa-split-paths.c: Include gimple-ssa.h, tree-phinodes.h
41366         and ssa-iterators.h.
41367         (is_feasible_trace): Implement a cost model based on joiner
41368         PHI node uses.
41370 2017-01-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
41372         PR target/79004
41373         * config/rs6000/rs6000.md (FP_ISA3): Do not optimize converting
41374         char or short to __float128/_Float128 directly.
41376 2017-01-12  Martin Sebor  <msebor@redhat.com>
41378         to -Wformat-overflow.
41379         * gimple-ssa-sprintf.c (pass_sprintf_length::gate): Adjust.
41380         (min_bytes_remaining): Same.
41381         (get_string_length): Same.
41382         (format_string): Same.
41383         (format_directive): Same.
41384         (add_bytes): Same.
41385         (pass_sprintf_length::handle_gimple_call): Same.
41387 2017-01-12  Jakub Jelinek  <jakub@redhat.com>
41389         * gimple-ssa-sprintf.c (try_substitute_return_value): Remove
41390         info.nowrite calls with no lhs that can't throw.  Return bool
41391         whether gsi_remove has been called or not.
41392         (pass_sprintf_length::handle_gimple_call): Return bool whether
41393         try_substitute_return_value called gsi_remove.  Formatting fix.
41394         (pass_sprintf_length::execute): Don't use gsi_remove if
41395         handle_gimple_call returned true.
41397         PR bootstrap/79069
41398         * cfgrtl.c (rtl_tidy_fallthru_edge): For any_uncondjump_p that can't
41399         be removed due to side-effects, don't remove following barrier nor
41400         turn the successor edge into fallthru edge.
41402 2017-01-12  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
41404         PR target/79044
41405         * config/rs6000/rs6000.c (insn_is_swappable_p): Mark
41406         element-reversing loads and stores as not swappable.
41408 2017-01-12  Nathan Sidwell  <nathan@acm.org>
41409             Nicolai Stange  <nicstange@gmail.com>
41411         * combine.c (try_combine): Don't ignore result of overlap checking
41412         loop.  Combine overlap & asm check into single loop.
41414 2017-01-12  Richard Biener  <rguenther@suse.de>
41416         * tree-pretty-print.c (dump_generic_node): Provide -gimple
41417         variant for MEM_REF.  Sanitize INTEGER_CST for -gimple.
41419 2017-01-12  Richard Biener  <rguenther@suse.de>
41421         * tree.c (initialize_tree_contains_struct): Make TS_OPTIMIZATION
41422         and TS_TARGET_OPTION directly derive from TS_BASE.
41423         * tree-core.h (tree_optimization_option): Derive from tree_base.
41424         (tree_target_option): Likewise.
41426 2017-01-11  Uros Bizjak  <ubizjak@gmail.com>
41428         * config/i386/i386.c (memory_address_length): Increase len
41429         only when rip_relative_addr_p returns false.
41431 2017-01-11  Julia Koval  <julia.koval@intel.com>
41433         * common/config/i386/i386-common.c (OPTION_MASK_ISA_SGX_UNSET): New.
41434         (OPTION_MASK_ISA_SGX_SET): New.
41435         (ix86_handle_option): Handle OPT_msgx.
41436         * config.gcc: Added sgxintrin.h.
41437         * config/i386/driver-i386.c (host_detect_local_cpu): Detect sgx.
41438         * config/i386/i386-c.c (ix86_target_macros_internal): Define __SGX__.
41439         * config/i386/i386.c (ix86_target_string): Add -msgx.
41440         (PTA_SGX): New.
41441         (ix86_option_override_internal): Handle new options.
41442         (ix86_valid_target_attribute_inner_p): Add sgx.
41443         * config/i386/i386.h (TARGET_SGX, TARGET_SGX_P): New.
41444         * config/i386/i386.opt: Add msgx.
41445         * config/i386/sgxintrin.h: New file.
41446         * config/i386/x86intrin.h: Add sgxintrin.h.
41448 2017-01-11  Jakub Jelinek  <jakub@redhat.com>
41450         PR c++/71537
41451         * fold-const.c (maybe_nonzero_address): Return 1 for function
41452         local objects.
41453         (tree_single_nonzero_warnv_p): Don't handle function local objects
41454         here.
41456         PR c++/72813
41457         * gcc.c (default_compilers): Don't add -o %g.s for -S -save-temps
41458         of c-header.
41460 2017-01-11  David Malcolm  <dmalcolm@redhat.com>
41462         PR driver/78877
41463         * opts.c: Include "spellcheck.h"
41464         (struct string_fragment): New struct.
41465         (struct edit_distance_traits<const string_fragment &>): New
41466         struct.
41467         (get_closest_sanitizer_option): New function.
41468         (parse_sanitizer_options): Offer suggestions for unrecognized arguments.
41470 2017-01-11  Jakub Jelinek  <jakub@redhat.com>
41472         * dwarf2out.c (DWARF_COMPILE_UNIT_HEADER_SIZE): For DWARF5 decrease
41473         by 12.
41474         (DWARF_COMDAT_TYPE_UNIT_HEADER_SIZE): Always
41475         DWARF_COMPILE_UNIT_HEADER_SIZE plus 12.
41476         (DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE): Define.
41477         (calc_base_type_die_sizes): Use DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE
41478         for initial die_offset if dwarf_split_debug_info.
41479         (output_comp_unit): Use DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE for
41480         initial next_die_offset if dwo_id is non-NULL.  Don't emit padding
41481         fields.
41482         (output_skeleton_debug_sections): Formatting fix.  Use
41483         DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE instead of
41484         DWARF_COMPILE_UNIT_HEADER_SIZE.  Don't emit padding.
41486 2017-01-11  Wilco Dijkstra  <wdijkstr@arm.com>
41488         * config/arm/cortex-a53.md: Add bypasses for
41489         cortex_a53_r2f_cvt.
41490         (cortex_a53_r2f): Only use for transfers.
41491         (cortex_a53_f2r): Likewise.
41492         (cortex_a53_r2f_cvt): Add reservation for conversions.
41493         (cortex_a53_f2r_cvt): Likewise.
41495 2017-01-11  Tamar Christina  <tamar.christina@arm.com>
41497         * config/arm/arm_neon.h: Add __artificial__ and gnu_inline
41498         to all inlined functions, change static to extern.
41500 2017-01-11  Christophe Lyon  <christophe.lyon@linaro.org>
41502         PR target/78253
41503         * config/arm/arm.c (legitimize_pic_address): Handle reference to
41504         weak symbol.
41505         (arm_assemble_integer): Likewise.
41507 2017-01-11  Richard Earnshaw  <rearnsha@arm.com>
41509         * config.gcc: Use new awk script to check CPU, FPU and architecture
41510         parameters for --with-... options.
41511         * config/arm/parsecpu.awk: New file
41512         * config/arm/arm-cpus.in: New file.
41513         * config/arm/arm-opts.h: Include arm-cpu.h instead of processing .def
41514         files.
41515         * config/arm/arm.c: Include arm-cpu-data.h instead of processing .def
41516         files.
41517         * config/arm/t-arm: Update dependency rules.
41518         * common/config/arm/arm-common.c: Include arm-cpu-cdata.h instead
41519         of processing .def files.
41520         * config/arm/genopt.sh: Deleted.
41521         * config/arm/gentune.sh: Deleted.
41522         * config/arm/arm-cores.def: Deleted.
41523         * config/arm/arm-arches.def: Deleted.
41524         * config/arm/arm-fpus.def: Deleted.
41525         * config/arm/arm-tune.md: Regenerated.
41526         * config/arm/arm-tables.opt: Regenerated.
41527         * config/arm/arm-cpu.h: New generated file.
41528         * config/arm/arm-cpu-data.h: New generated file.
41529         * config/arm/arm-cpu-cdata.h: New generated file.
41531 2017-01-11  Maxim Ostapenko  <m.ostapenko@samsung.com>
41533         PR lto/79042
41534         * lto-cgraph.c (lto_output_varpool_node): Pack dynamically_initialized
41535         bit.
41536         (input_varpool_node): Unpack dynamically_initialized bit.
41538 2017-01-11  Eric Botcazou  <ebotcazou@adacore.com>
41540         PR rtl-optimization/79032
41541         * lra-constraints.c (simplify_operand_subreg): In the MEM case, test
41542         the alignment of the adjusted memory reference against that of MODE,
41543         instead of the alignment of the original memory reference.
41545 2017-01-11  Martin Jambor  <mjambor@suse.cz>
41547         * hsa.c (hsa_callable_function_p): Revert addition of DECL_ARTIFICIAL
41548         test.
41549         * ipa-hsa.c (process_hsa_functions): Only duplicate non-artificial
41550         decorated functions.
41552 2017-01-11  Richard Biener  <rguenther@suse.de>
41554         * tree-vrp.c (evrp_dom_walker::before_dom_children): Also
41555         set range/nonnull info for PHI results.  Do not set it on
41556         stmts marked for removal.
41558 2017-01-10  Eric Botcazou  <ebotcazou@adacore.com>
41560         * expr.c (store_field): In the bitfield case, fetch the return value
41561         from the registers before applying a single big-endian adjustment.
41562         Always do a final load for a BLKmode value not larger than a word.
41564 2017-01-10  David Malcolm  <dmalcolm@redhat.com>
41566         PR c++/77949
41567         * input.c (selftest::test_accessing_ordinary_linemaps): Verify
41568         that we correctly handle column numbers greater than
41569         LINE_MAP_MAX_COLUMN_NUMBER.
41571 2017-01-10  Martin Sebor  <msebor@redhat.com>
41573         PR middle-end/78245
41574         * gimple-ssa-sprintf.c (get_destination_size): Call
41575         {init,fini}object_sizes.
41576         * tree-object-size.c (addr_object_size): Adjust.
41577         (pass_through_call): Adjust.
41578         (pass_object_sizes::execute): Adjust.
41579         * tree-object-size.h (fini_object_sizes): Declare.
41581 2017-01-10  Martin Sebor  <msebor@redhat.com>
41583         PR tree-optimization/78775
41584         * builtins.c (get_size_range): Move...
41585         * calls.c: ...to here.
41586         (alloc_max_size): Accept zero argument.
41587         (operand_signed_p): Remove.
41588         (maybe_warn_alloc_args_overflow): Call get_size_range.
41589         * calls.h (get_size_range): Declare.
41591 2017-01-10  Joe Seymour  <joe.s@somniumtech.com>
41593         * config/msp430/driver-msp430.c (msp430_mcu_data): Sync with data
41594         from TI's devices.csv file as of September 2016.
41595         * config/msp430/msp430.c (msp430_mcu_data): Likewise.
41597 2017-01-10  Sandra Loosemore  <sandra@codesourcery.com>
41599         * doc/extend.texi: Tweak formatting to fix overfull hbox warnings.
41600         * doc/invoke.texi: Likewise.
41601         * doc/md.texi: Likewise.
41602         * doc/objc.texi: Likewise.
41604 2017-01-10  Joshua Conner  <joshconner@google.com>
41606         * config/arm/fuchsia-elf.h: New file.
41607         * config/fuchsia.h: New file.
41608         * config.gcc (*-*-fuchsia*): Set native_system_header_dir.
41609         (aarch64*-*-fuchsia*, arm*-*-fuchsia*, x86_64-*-fuchsia*): Add to
41610         targets.
41611         * config.host: (aarch64*-*-fuchsia*, arm*-*-fuchsia*): Add to hosts.
41613 2016-01-10  Richard Biener  <rguenther@suse.de>
41615         PR tree-optimization/79034
41616         * tree-call-cdce.c (shrink_wrap_one_built_in_call_with_conds):
41617         Propagate out degenerate PHIs in the joiner.
41619 2017-01-10  Martin Liska  <mliska@suse.cz>
41621         * ipa-icf.c (sort_sem_items_by_decl_uid): New function.
41622         (sort_congruence_classes_by_decl_uid): Likewise.
41623         (sort_congruence_class_groups_by_decl_uid): Likewise.
41624         (sem_item_optimizer::merge_classes): Sort class, groups in these
41625         classes and members in the groups by DECL_UID of declarations.
41626         This would make merge operations stable.
41628 2017-01-10  Martin Liska  <mliska@suse.cz>
41630         * ipa-icf.c (sem_item_optimizer::sem_item_optimizer): Remove
41631         usage of m_classes_vec.
41632         (sem_item_optimizer::~sem_item_optimizer):  Likewise.
41633         (sem_item_optimizer::get_group_by_hash): Likewise.
41634         (sem_item_optimizer::subdivide_classes_by_equality): Likewise.
41635         (sem_item_optimizer::subdivide_classes_by_sensitive_refs): Likewise.
41636         (sem_item_optimizer::verify_classes): Likewise.
41637         (sem_item_optimizer::process_cong_reduction): Likewise.
41638         (sem_item_optimizer::dump_cong_classes): Likewise.
41639         (sem_item_optimizer::merge_classes): Likewise.
41640         * ipa-icf.h (congruence_class_hash): Rename from
41641         congruence_class_group_hash.  Remove declaration of m_classes_vec.
41643 2017-01-10  Andrew Senkevich  <andrew.senkevich@intel.com>
41645         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VPOPCNTDQ_SET,
41646         OPTION_MASK_ISA_AVX512VPOPCNTDQ_UNSET): New.
41647         * config.gcc: Add avx512vpopcntdqintrin.h.
41648         * config/i386/avx512vpopcntdqintrin.h: New.
41649         * config/i386/cpuid.h (bit_AVX512VPOPCNTDQ): New.
41650         * config/i386/i386-builtin-types.def: Add new types.
41651         * config/i386/i386-builtin.def (__builtin_ia32_vpopcountd_v16si,
41652         __builtin_ia32_vpopcountd_v16si_mask, __builtin_ia32_vpopcountq_v8di,
41653         __builtin_ia32_vpopcountq_v8di_mask): New.
41654         * config/i386/i386-c.c (ix86_target_macros_internal): Define
41655         __AVX512VPOPCNTDQ__.
41656         * config/i386/i386.c (ix86_target_string): Add -mavx512vpopcntdq.
41657         (PTA_AVX512VPOPCNTDQ): Define.
41658         * config/i386/i386.h (TARGET_AVX512VPOPCNTDQ,
41659         TARGET_AVX512VPOPCNTDQ_P): Define.
41660         * config/i386/i386.opt: Add mavx512vpopcntdq.
41661         * config/i386/immintrin.h: Include avx512vpopcntdqintrin.h.
41662         * config/i386/sse.md (define_insn "vpopcount<mode><mask_name>"): New.
41664 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
41666         PR middle-end/77484
41667         * predict.def (PRED_CALL): Set to 67.
41669 2017-01-09  Eric Botcazou  <ebotcazou@adacore.com>
41671         * expr.c (store_field): In the bitfield case, if the value comes from
41672         a function call and is of an aggregate type returned in registers, do
41673         not modify the field mode; extract the value in all cases if the mode
41674         is BLKmode and the size is not larger than a word.
41676 2017-01-09  Dominique d'Humieres  <dominiq@lps.ens.fr>
41678         PR target/71017
41679         * config/i386/cpuid.h: Fix undefined behavior.
41681 2017-01-04  Jeff Law  <law@redhat.com>
41683         PR tree-optimization/79007
41684         PR tree-optimization/67955
41685         * tree-ssa-alias.c (same_addr_size_stores_p): Only need to be
41686         conservative for pt.null when flag_non_call_exceptions is on.
41688 2017-01-09  Jakub Jelinek  <jakub@redhat.com>
41690         PR translation/79019
41691         PR translation/79020
41692         * params.def (PARAM_INLINE_MIN_SPEEDUP,
41693         PARAM_IPA_CP_SINGLE_CALL_PENALTY,
41694         PARAM_USE_AFTER_SCOPE_DIRECT_EMISSION_THRESHOLD): Fix typos
41695         in descriptions.
41696         * config/avr/avr.opt (maccumulate-args): Likewise.
41697         * config/msp430/msp430.opt (mwarn-mcu): Likewise.
41698         * common.opt (freport-bug): Likewise.
41699         * cif-code.def (CIF_FINAL_ERROR): Likewise.
41700         * doc/invoke.texi (ipa-cp-single-call-penalty): Likewise.
41701         * config/s390/s390.c (s390_invalid_binary_op): Fix spelling in
41702         translatable string.
41703         * config/i386/i386.c (function_value_32): Likewise.
41704         * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Likewise.
41705         * config/msp430/msp430.c (msp430_option_override, msp430_attr):
41706         Likewise.
41707         * config/msp430/driver-msp430.c (msp430_select_hwmult_lib): Likewise.
41708         * common/config/msp430/msp430-common.c (msp430_handle_option):
41709         Likewise.
41710         * symtab.c (symtab_node::verify_base): Likewise.
41711         * opts.c (set_debug_level): Likewise.
41712         * tree.c (verify_type_variant): Likewise.  Fix typo in comment.
41713         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Add
41714         missing whitespace to translatable strings.
41715         * config/avr/avr.md (bswapsi2): Fix typo in comment.
41716         * config/sh/superh.h: Likewise.
41717         * config/i386/xopintrin.h: Likewise.
41718         * config/i386/znver1.md: Likewise.
41719         * config/rs6000/rs6000.c (struct rs6000_opt_mask): Likewise.
41720         * ipa-inline-analysis.c (compute_inline_parameters): Likewise.
41721         * double-int.h (struct double_int): Likewise.
41722         * double-int.c (div_and_round_double): Likewise.
41723         * wide-int.cc: Likewise.
41724         * tree-ssa.c (non_rewritable_mem_ref_base): Likewise.
41725         * tree-ssa-sccvn.c (vn_reference_lookup_3): Likewise.
41726         * cfgcleanup.c (crossjumps_occured): Renamed to ...
41727         (crossjumps_occurred): ... this.
41728         (try_crossjump_bb, try_head_merge_bb, try_optimize_cfg, cleanup_cfg):
41729         Adjust all uses.
41731         PR tree-optimization/78899
41732         * tree-if-conv.c (version_loop_for_if_conversion): Instead of
41733         returning bool return struct loop *, NULL for failure and the new
41734         loop on success.
41735         (versionable_outer_loop_p): Don't version outer loop if it has
41736         dont_vectorized bit set.
41737         (tree_if_conversion): When versioning outer loop, ensure
41738         tree_if_conversion is performed also on the inner loop of the
41739         non-vectorizable outer loop copy.
41740         * tree-vectorizer.c (set_uid_loop_bbs): Formatting fix.  Fold
41741         LOOP_VECTORIZED in inner loop of the scalar outer loop and
41742         prevent vectorization of it.
41743         (vectorize_loops): For outer + inner LOOP_VECTORIZED, ensure
41744         the outer loop vectorization of the non-scalar version is attempted
41745         before vectorization of the inner loop in scalar version.  If
41746         outer LOOP_VECTORIZED guarded loop is not vectorized, prevent
41747         vectorization of its inner loop.
41748         * tree-vect-loop-manip.c (rename_variables_in_bb): If outer_loop
41749         has 2 inner loops, rename also on edges from bb whose single pred
41750         is outer_loop->header.  Fix typo in function comment.
41752 2017-01-09  Martin Sebor  <msebor@redhat.com>
41754         PR bootstrap/79033
41755         * asan.c (asan_emit_stack_protection): Increase local buffer size
41756         to avoid snprintf truncation warning.
41758 2017-01-09  Andrew Pinski  <apinski@cavium.com>
41760         * config/aarch64/aarch64-cores.def: Add thunderx2t99.  Change vulcan
41761         to reference thunderx2t99 for the tuning structure
41762         * config/aarch64/aarch64-cost-tables.h (vulcan_extra_costs):
41763         Rename to ...
41764         (thunderx2t99_extra_costs): This.
41765         * config/aarch64/aarch64-tune.md: Regenerate.
41766         * config/aarch64/aarch64.c (vulcan_addrcost_table): Rename to ...
41767         (vulcan_addrcost_table): This.
41768         (vulcan_regmove_cost): Rename to ...
41769         (thunderx2t99_regmove_cost): This.
41770         (vulcan_vector_cost): Rename to ...
41771         (thunderx2t99_vector_cost): this.
41772         (vulcan_branch_cost): Rename to ...
41773         (thunderx2t99_branch_cost): This.
41774         (vulcan_tunings): Rename to ...
41775         (thunderx2t99_tunings): This and s/vulcan/thunderx2t99 .
41776         * doc/invoke.texi (AARCH64/mtune): Add thunderx2t99.
41778 2017-01-09  Martin Jambor  <mjambor@suse.cz>
41780         PR ipa/78365
41781         PR ipa/78599
41782         * ipa-prop.h (ipa_jump_func): Swap positions of vr_known and m_vr.
41783         * ipa-cp.c (ipa_vr_operation_and_type_effects): New function.
41784         (propagate_vr_accross_jump_function): Use the above function for all
41785         value range computations for pass-through jump functions and type
41786         converasion from explicit value range values.
41787         (ipcp_propagate_stage): Do not attempt to deduce types of formal
41788         parameters from TYPE_ARG_TYPES.
41789         * ipa-prop.c (ipa_write_jump_function): Remove trailing whitespace.
41790         (ipa_write_node_info): Stream type of the actual argument.
41791         (ipa_read_node_info): Likewise. Also remove trailing whitespace.
41793 2017-01-09  Martin Liska  <mliska@suse.cz>
41795         PR pch/78970
41796         * gcc.c (driver_handle_option): Handle OPT_E and set have_E.
41797         (lookup_compiler): Do not show error message with have_E.
41799 2017-01-09  Jakub Jelinek  <jakub@redhat.com>
41801         PR tree-optimization/78938
41802         * tree-vect-stmts.c (vectorizable_condition): For non-masked COND_EXPR
41803         where comp_vectype is VECTOR_BOOLEAN_TYPE_P, use
41804         BIT_{NOT,XOR,AND,IOR}_EXPR on the comparison operands instead of
41805         {EQ,NE,GE,GT,LE,LT}_EXPR directly inside of VEC_COND_EXPR.  Formatting
41806         fixes.
41808 2017-01-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
41810         * tree-ssa-address.c (gen_addr_rtx): Don't handle index if it
41811         is const0_rtx.
41813 2017-01-09  Richard Biener  <rguenther@suse.de>
41815         PR tree-optimization/78997
41816         * tree-vect-slp.c (vect_mask_constant_operand_p): Handle SSA
41817         name condition properly.
41819 2017-01-09  Richard Biener  <rguenther@suse.de>
41821         PR debug/79000
41822         * dwarf2out.c (is_cxx): New overload with context.
41823         (is_naming_typedef_decl): Use it.
41825 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41827         * invoke.texi (Option Summary): Correct spacing in option lists
41828         and add line breaks to fix over-long lines.
41830 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41832         PR middle-end/17660
41834         * extend.texi (Common Variable Attributes): Add xref to GCC
41835         Internals manual to explain mode attribute keywords.
41837 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41839         PR other/16519
41840         * doc/invoke.texi (Option Summary): Move -pthread to Linker Options
41841         and Preprocessor Options.
41842         (Options for Linking): Document -pthread here....
41843         (RS/6000 and PowerPC Options): ...not here.
41844         (Solaris 2 Options): ...or here.
41845         * doc/cppopts.texi: Document -pthread.
41847 2017-01-08  Martin Sebor  <msebor@redhat.com>
41849         PR middle-end/77708
41850         * doc/invoke.texi (Warning Options): Document -Wformat-truncation.
41851         * gimple-ssa-sprintf.c (call_info::reval_used, call_info::warnopt):
41852         New member functions.
41853         (format_directive): Used them.
41854         (add_bytes): Same.
41855         (pass_sprintf_length::handle_gimple_call): Same.
41856         * graphite-sese-to-poly.c (tree_int_to_gmp): Increase buffer size
41857         to avoid truncation for any argument.
41858         (extract_affine_mul): Same.
41859         * tree.c (get_file_function_name): Same.
41861 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
41863         PR middle-end/77484
41864         * predict.def (PRED_INDIR_CALL): Set to 86.
41866 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41868         PR preprocessor/54124
41869         * doc/cppopts.texi: Reformat -d subtable to list the full name
41870         of the options.  Add cross-reference to the docs for the general
41871         compiler -d options.
41872         * doc/invoke.texi (Developer Options): Add cross-reference to the
41873         preprocessor-specific -d option documentation.
41875 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41877         PR preprocessor/13498
41878         * doc/cpp.texi (Search Path): Rewrite to remove obsolete and
41879         redudant material, and reflect new command-line options.
41880         (System Headers): Likewise.
41882 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41884         * doc/cppdiropts.texi: Merge documentation of -I, -iquote,
41885         -isystem, and -idirafter.  Copy-edit.
41886         * doc/cppopts.texi: Copy-edit.  Remove contradiction about
41887         default for -ftrack-macro-expansion.  Delete obsolete and
41888         badly-formatted implementation details about -fdebug-cpp output.
41889         * doc/cppwarnopts.texi: Copy-edit.
41891 2017-01-07  David Malcolm  <dmalcolm@redhat.com>
41893         PR c++/72803
41894         * input.c (selftest::test_accessing_ordinary_linemaps): Verify
41895         that the transition from a max line width >= 1<<10 to narrower
41896         lines works correctly.
41898 2017-01-07  Alexandre Oliva  <aoliva@redhat.com>
41900         * doc/options.texi (PerFunction): New.
41901         * opt-functions.awk (switch_flags): Map both Optimization and
41902         PerFunction to CL_OPTIMIZATION.
41903         * opth-gen.awk: Test for PerFunction flag along with
41904         Optimization.
41905         * optc-save-gen.awk: Likewise.  Introduce var_opt_hash and set
41906         it only when the latter is present.  Skip those that don't in
41907         the hash function generator.
41908         * common.opt (fvar-tracking): Mark as PerFunction instead of
41909         Optimization.
41910         (fvar-tracking-assignments): Likewise.
41911         (fvar-tracking-assignments-toggle): Likewise.
41912         (fvar-tracking-uninit): Likewise.
41914 2017-01-07  Jakub Jelinek  <jakub@redhat.com>
41916         PR translation/79018
41917         * params.def (PARAM_MAX_STORES_TO_MERGE): Add missing space between
41918         the and store.
41920 2017-01-06  Mikael Pettersson  <mikpelinux@gmail.com>
41922         PR target/57583
41923         * config/m68k/m68k.opt (LONG_JUMP_TABLE_OFFSETS): New option.
41924         * config/m68k/linux.h (ASM_RETURN_CASE_JUMP): Handle
41925         TARGET_LONG_JUMP_TABLE_OFFSETS.
41926         * config/m68k/m68kelf.h (ASM_RETURN_CASE_JUMP): Likewise.
41927         * config/m68k/netbsd-elf.h (ASM_RETURN_CASE_JUMP): Likewise.
41928         * config/m68k/m68k.h (CASE_VECTOR_MODE): Likewise.
41929         (ASM_OUTPUT_ADDR_DIFF_ELF): Likewise.
41930         * config/m68k/m68k.md (tablejump expander): Likewise.
41931         (*tablejump_pcrel_hi): Renamed from unnamed insn, reject
41932         TARGET_LONG_JUMP_TABLE_OFFSETS.
41933         (*tablejump_pcrel_si): New insn, handle TARGET_LONG_JUMP_TABLE_OFFSETS.
41934         * doc/invoke.texi (M68K options): Add -mlong-jump-table-offsets.
41936 2017-01-06  Edgar E. Iglesias <edgar.iglesias@xilinx.com>
41937             David Holsgrove <david.holsgrove@xilinx.com>
41939         * common/config/microblaze/microblaze-common.c
41940         (TARGET_EXCEPT_UNWIND_INFO): Remove.
41941         * config/microblaze/microblaze-protos.h (microblaze_eh_return):
41942         New prototype.
41943         * config/microblaze/microblaze.c (microblaze_must_save_register)
41944         (microblaze_expand_epilogue, microblaze_return_addr): Handle
41945         calls_eh_return.
41946         (microblaze_eh_return): New function.
41947         * config/microblaze/microblaze.h (RETURN_ADDR_OFFSET)
41948         (EH_RETURN_DATA_REGNO, MB_EH_STACKADJ_REGNUM)
41949         (EH_RETURN_STACKADJ_RTX, ASM_PREFERRED_EH_DATA_FORMAT): New macros.
41950         * config/microblaze/microblaze.md (eh_return): New pattern.
41952 2017-01-06  Jakub Jelinek  <jakub@redhat.com>
41954         * system.h (GCC_DIAGNOSTIC_PUSH_IGNORED, GCC_DIAGNOSTIC_POP,
41955         GCC_DIAGNOSTIC_STRINGIFY): Define.
41957         * read-rtl.c (rtx_reader::read_rtx_code): Avoid -Wsign-compare warning.
41959 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41961         * config/arm/arm.md (<mcrr>): New.
41962         (<mrrc>): New.
41963         * config/arm/arm.c (arm_arch5te): New.
41964         (arm_option_override): Set arm_arch5te.
41965         (arm_coproc_builtin_available): Add support for mcrr, mcrr2, mrrc
41966         and mrrc2.
41967         * config/arm/arm-builtins.c (MCRR_QUALIFIERS): Define to...
41968         (arm_mcrr_qualifiers): ... this. New.
41969         (MRRC_QUALIFIERS): Define to...
41970         (arm_mrrc_qualifiers): ... this. New.
41971         * config/arm/arm_acle.h (__arm_mcrr, __arm_mcrr2, __arm_mrrc,
41972         __arm_mrrc2): New.
41973         * config/arm/arm_acle_builtins.def (mcrr, mcrr2, mrrc, mrrc2): New.
41974         * config/arm/iterators.md (MCRRI, mcrr, MCRR): New.
41975         (MRRCI, mrrc, MRRC): New.
41976         * config/arm/unspecs.md (VUNSPEC_MCRR, VUNSPEC_MCRR2, VUNSPEC_MRRC,
41977         VUNSPEC_MRRC2): New.
41979 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41981         * config/arm/arm.md (<mcr>): New.
41982         (<mrc>): New.
41983         * config/arm/arm.c (arm_coproc_builtin_available): Add
41984         support for mcr, mrc, mcr2 and mrc2.
41985         * config/arm/arm-builtins.c (MCR_QUALIFIERS): Define to...
41986         (arm_mcr_qualifiers): ... this. New.
41987         (MRC_QUALIFIERS): Define to ...
41988         (arm_mrc_qualifiers): ... this. New.
41989         (MCR_QUALIFIERS): Define to ...
41990         (arm_mcr_qualifiers): ... this. New.
41991         * config/arm/arm_acle.h (__arm_mcr, __arm_mrc, __arm_mcr2,
41992         __arm_mrc2): New.
41993         * config/arm/arm_acle_builtins.def (mcr, mcr2, mrc, mrc2): New.
41994         * config/arm/iterators.md (MCRI, mcr, MCR, MRCI, mrc, MRC): New.
41995         * config/arm/unspecs.md (VUNSPEC_MCR, VUNSPEC_MCR2, VUNSPEC_MRC,
41996         VUNSPEC_MRC2): New.
41998 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
42000         * config/arm/arm.md (*ldc): New.
42001         (*stc): New.
42002         (<ldc>): New.
42003         (<stc>): New.
42004         * config/arm/arm.c (arm_coproc_builtin_available): Add
42005         support for ldc,ldcl,stc,stcl,ldc2,ldc2l,stc2 and stc2l.
42006         (arm_coproc_ldc_stc_legitimate_address): New.
42007         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
42008         'qualifier_const_pointer'.
42009         (LDC_QUALIFIERS): Define to...
42010         (arm_ldc_qualifiers): ... this. New.
42011         (STC_QUALIFIERS): Define to...
42012         (arm_stc_qualifiers): ... this. New.
42013         * config/arm/arm-protos.h
42014         (arm_coproc_ldc_stc_legitimate_address): New.
42015         * config/arm/arm_acle.h (__arm_ldc, __arm_ldcl, __arm_stc,
42016         __arm_stcl, __arm_ldc2, __arm_ldc2l, __arm_stc2, __arm_stc2l): New.
42017         * config/arm/arm_acle_builtins.def (ldc, ldc2, ldcl, ldc2l, stc,
42018         stc2, stcl, stc2l): New.
42019         * config/arm/constraints.md (Uz): New.
42020         * config/arm/iterators.md (LDCI, STCI, ldc, stc, LDC STC): New.
42021         * config/arm/unspecs.md (VUNSPEC_LDC, VUNSPEC_LDC2, VUNSPEC_LDCL,
42022         VUNSPEC_LDC2L, VUNSPEC_STC, VUNSPEC_STC2, VUNSPEC_STCL,
42023         VUNSPEC_STC2L): New.
42025 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
42027         * config/arm/arm.md (<cdp>): New.
42028         * config/arm/arm.c (neon_const_bounds): Rename this ...
42029         (arm_const_bounds): ... this.
42030         (arm_coproc_builtin_available): New.
42031         * config/arm/arm-builtins.c (SIMD_MAX_BUILTIN_ARGS): Increase.
42032         (arm_type_qualifiers): Add 'qualifier_unsigned_immediate'.
42033         (CDP_QUALIFIERS): Define to...
42034         (arm_cdp_qualifiers): ... this. New.
42035         (void_UP): Define.
42036         (arm_expand_builtin_args): Add case for 6 arguments.
42037         * config/arm/arm-protos.h (neon_const_bounds): Rename this ...
42038         (arm_const_bounds): ... this.
42039         (arm_coproc_builtin_available): New.
42040         * config/arm/arm_acle.h (__arm_cdp): New.
42041         (__arm_cdp2): New.
42042         * config/arm/arm_acle_builtins.def (cdp): New.
42043         (cdp2): New.
42044         * config/arm/iterators.md (CDPI,CDP,cdp): New.
42045         * config/arm/neon.md: Rename all 'neon_const_bounds' to
42046         'arm_const_bounds'.
42047         * config/arm/types.md (coproc): New.
42048         * config/arm/unspecs.md (VUNSPEC_CDP, VUNSPEC_CDP2): New.
42049         * doc/extend.texi (ACLE): Add a mention of Coprocessor intrinsics.
42050         * doc/sourcebuild.texi (arm_coproc1_ok, arm_coproc2_ok,
42051         arm_coproc3_ok, arm_coproc4_ok): Document new effective targets.
42053 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
42055         * config/arm/arm-builtins.c (arm_unsigned_binop_qualifiers): New.
42056         (UBINOP_QUALIFIERS): New.
42057         (si_UP): Define.
42058         (acle_builtin_data): New. Change comment.
42059         (arm_builtins): Remove ARM_BUILTIN_CRC32B, ARM_BUILTIN_CRC32H,
42060         ARM_BUILTIN_CRC32W, ARM_BUILTIN_CRC32CB, ARM_BUILTIN_CRC32CH,
42061         ARM_BUILTIN_CRC32CW. Add ARM_BUILTIN_ACLE_BASE and include
42062         arm_acle_builtins.def.
42063         (ARM_BUILTIN_ACLE_PATTERN_START): Define.
42064         (arm_init_acle_builtins): New.
42065         (CRC32_BUILTIN): Remove.
42066         (bdesc_2arg): Remove entries for crc32b, crc32h, crc32w,
42067         crc32cb, crc32ch and crc32cw.
42068         (arm_init_crc32_builtins): Remove.
42069         (arm_init_builtins): Use arm_init_acle_builtins rather
42070         than arm_init_crc32_builtins.
42071         (arm_expand_acle_builtin): New.
42072         (arm_expand_builtin): Use 'arm_expand_acle_builtin'.
42073         * config/arm/arm_acle_builtins.def: New.
42075 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
42077         * config/arm/arm-builtins.c (neon_builtin_datum): Rename to ..
42078         (arm_builtin_datum): ... this.
42079         (arm_init_neon_builtin): Rename to ...
42080         (arm_init_builtin): ... this. Add a new parameters PREFIX
42081         and USE_SIG_IN_NAME.
42082         (arm_init_neon_builtins): Replace 'arm_init_neon_builtin' with
42083         'arm_init_builtin'. Replace type 'neon_builtin_datum' with
42084         'arm_builtin_datum'.
42085         (arm_init_vfp_builtins): Likewise.
42086         (builtin_arg): Rename enum's replacing 'NEON_ARG' with
42087         'ARG_BUILTIN' and add a 'ARG_BUILTIN_NEON_MEMORY.
42088         (arm_expand_neon_args): Rename to ...
42089         (arm_expand_builtin_args): ... this. Rename builtin_arg
42090         enum values and differentiate between ARG_BUILTIN_MEMORY
42091         and ARG_BUILTIN_NEON_MEMORY.
42092         (arm_expand_neon_builtin_1): Rename to ...
42093         (arm_expand_builtin_1): ... this. Rename builtin_arg enum
42094         values, arm_expand_builtin_args and add bool parameter NEON.
42095         (arm_expand_neon_builtin): Use arm_expand_builtin_1.
42096         (arm_expand_vfp_builtin): Likewise.
42097         (NEON_MAX_BUILTIN_ARGS): Remove, it was unused.
42099 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
42101         PR middle-end/77484
42102         * predict.def (PRED_POLYMORPHIC_CALL): Set to 59.
42103         * predict.c (tree_estimate_probability_bb): Reverse direction of
42104         polymorphic call predictor.
42106 2017-01-06  David Malcolm  <dmalcolm@redhat.com>
42108         * passes.c (execute_one_pass): Split out pass-skipping logic into...
42109         (determine_pass_name_match): ...this new function and...
42110         (should_skip_pass_p): ...this new function.
42112 2017-01-06  Nathan Sidwell  <nathan@acm.org>
42114         * ipa-visibility.c (function_and_variable_visibility): Reformat
42115         comments and long lines.  Remove extrneous if.
42116         * symtab.c (symtab_node::make_decl_local): Fix code format.
42117         (symtab_node::set_section_for_node): Fix comment typo.
42119 2017-01-06  Martin Liska  <mliska@suse.cz>
42121         PR bootstrap/79003
42122         * lra-constraints.c: Rename invariant to lra_invariant.
42123         * predict.c (set_even_probabilities): Initialize e to NULL.
42125 2017-01-05  Martin Sebor  <msebor@redhat.com>
42127         PR tree-optimization/78910
42128         * gimple-ssa-sprintf.c (tree_digits): Add an argument.
42129         (format_integer): Correct off-by-one error in the handling
42130         of precision with negative numbers in signed conversions..
42132 2017-01-05  Eric Botcazou  <ebotcazou@adacore.com>
42134         * doc/invoke.texi (C Dialect Options): Adjust -fsso-struct entry.
42136 2017-01-05  Jakub Jelinek  <jakub@redhat.com>
42138         PR tree-optimization/71016
42139         * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Pass cond_stmt to
42140         factor_out_conditional_conversion.  Formatting fix.
42141         (factor_out_conditional_conversion): Add cond_stmt argument.
42142         If arg1 is INTEGER_CST, punt if new_arg0 is not any operand of
42143         cond_stmt and if arg0_def_stmt is not the only stmt in its bb.
42144         Formatting fix.
42146 2017-01-05  David Malcolm  <dmalcolm@redhat.com>
42148         * Makefile.in (OBJS): Add read-md.o, read-rtl.o,
42149         read-rtl-function.o, and selftest-rtl.o.
42150         * config/aarch64/aarch64.c: Include selftest.h and selftest-rtl.h.
42151         (selftest::aarch64_test_loading_full_dump): New function.
42152         (selftest::aarch64_run_selftests): New function.
42153         (TARGET_RUN_TARGET_SELFTESTS): Wire it up to
42154         selftest::aarch64_run_selftests.
42155         * config/i386/i386.c
42156         (selftest::ix86_test_loading_dump_fragment_1): New function.
42157         (selftest::ix86_test_loading_call_insn): New function.
42158         (selftest::ix86_test_loading_full_dump): New function.
42159         (selftest::ix86_test_loading_unspec): New function.
42160         (selftest::ix86_run_selftests): Call the new functions.
42161         * emit-rtl.c (maybe_set_max_label_num): New function.
42162         * emit-rtl.h (maybe_set_max_label_num): New decl.
42163         * function.c (instantiate_decls): Guard call to
42164         instantiate_decls_1 with if (DECL_INITIAL (fndecl)).
42165         * function-tests.c (selftest::verify_three_block_rtl_cfg): Remove
42166         "static".
42167         * gensupport.c (gen_reader::gen_reader): Pass "false"
42168         for new "compact" param of rtx_reader.
42169         * print-rtl.c (rtx_writer::print_rtx_operand): Print "(nil)"
42170         rather than an empty string for NULL strings.
42171         * read-md.c: Potentially include config.h rather than bconfig.h.
42172         Wrap include of errors.h with #ifdef GENERATOR_FILE.
42173         (have_error): New global, copied from errors.c.
42174         (md_reader::read_name): Rename to...
42175         (md_reader::read_name_1): ...this, adding "out_loc" param,
42176         and converting "missing name or number" to returning false, rather
42177         than failing.
42178         (md_reader::read_name): Reimplement in terms of read_name_1.
42179         (md_reader::read_name_or_nil): New function.
42180         (md_reader::read_string): Handle "(nil)" by returning NULL.
42181         (md_reader::md_reader): Add new param "compact".
42182         (md_reader::read_md_files): Wrap with #ifdef GENERATOR_FILE.
42183         (md_reader::read_file): New method.
42184         * read-md.h (md_reader::md_reader): Add new param "compact".
42185         (md_reader::read_file): New method.
42186         (md_reader::is_compact): New accessor.
42187         (md_reader::read_name): Convert return type from void to file_location.
42188         (md_reader::read_name_or_nil): New decl.
42189         (md_reader::read_name_1): New decl.
42190         (md_reader::m_compact): New field.
42191         (noop_reader::noop_reader): Pass "false" for new "compact" param
42192         of rtx_reader.
42193         (rtx_reader::rtx_reader): Add new "compact" param.
42194         (rtx_reader::read_rtx_operand): Make virtual and convert return
42195         type from void to rtx.
42196         (rtx_reader::read_until): New decl.
42197         (rtx_reader::handle_any_trailing_information): New virtual function.
42198         (rtx_reader::postprocess): New virtual function.
42199         (rtx_reader::finalize_string): New virtual function.
42200         (rtx_reader::m_in_call_function_usage): New field.
42201         (rtx_reader::m_reuse_rtx_by_id): New field.
42202         * read-rtl-function.c: New file.
42203         * selftest-rtl.c (selftest::assert_rtx_ptr_eq_at): New function.
42204         * selftest-rtl.h (ASSERT_RTX_PTR_EQ): New macro.
42205         (selftest::verify_three_block_rtl_cfg): New decl.
42206         * read-rtl-function.h: New file.
42207         * read-rtl.c: Potentially include config.h rather than bconfig.h.
42208         For host, include function.h, memmodel.h, and emit-rtl.h.
42209         (one_time_initialization): New function.
42210         (struct compact_insn_name): New struct.
42211         (compact_insn_names): New array.
42212         (find_code): Handle insn codes in compact dumps.
42213         (apply_subst_iterator): Wrap with #ifdef GENERATOR_FILE.
42214         (bind_subst_iter_and_attr): Likewise.
42215         (add_condition_to_string): Likewise.
42216         (add_condition_to_rtx): Likewise.
42217         (apply_attribute_uses): Likewise.
42218         (add_current_iterators): Likewise.
42219         (apply_iterators): Likewise.
42220         (initialize_iterators): Guard usage of apply_subst_iterator with
42221         #ifdef GENERATOR_FILE.
42222         (read_conditions): Wrap with #ifdef GENERATOR_FILE.
42223         (md_reader::read_mapping): Likewise.
42224         (add_define_attr_for_define_subst): Likewise.
42225         (add_define_subst_attr): Likewise.
42226         (read_subst_mapping): Likewise.
42227         (check_code_iterator): Likewise.
42228         (rtx_reader::read_rtx): Likewise.  Move one-time initialization
42229         logic to...
42230         (one_time_initialization): New function.
42231         (rtx_reader::read_until): New method.
42232         (read_flags): New function.
42233         (parse_reg_note_name): New function.
42234         (rtx_reader::read_rtx_code): Initialize "iterator" to NULL.
42235         Handle reuse_rtx ids.
42236         Wrap iterator lookup within #ifdef GENERATOR_FILE.
42237         Add parsing support for RTL dumps, mirroring the special-cases in
42238         print_rtx, by calling read_flags, reading REG_NOTE names, INSN_UID
42239         values, and calling handle_any_trailing_information.
42240         (rtx_reader::read_rtx_operand): Convert return type from void
42241         to rtx, returning return_rtx.  Handle case 'e'.  Call
42242         finalize_string on XSTR and XTMPL fields.
42243         (rtx_reader::read_nested_rtx):  Handle dumps in which trailing
42244          "(nil)" values were omitted.  Call the postprocess vfunc on the
42245         return_rtx.
42246         (rtx_reader::rtx_reader): Add new "compact" param and pass to base
42247         class ctor.  Initialize m_in_call_function_usage.  Call
42248         one_time_initialization.
42249         * rtl-tests.c (selftest::test_uncond_jump): Call
42250         set_new_first_and_last_insn.
42251         * rtl.h (read_rtx): Wrap decl with #ifdef GENERATOR_FILE.
42252         * selftest-rtl.c: New file.
42253         * selftest-rtl.h (class selftest::rtl_dump_test): New class.
42254         (selftest::get_insn_by_uid): New decl.
42255         * selftest-run-tests.c (selftest::run_tests): Call
42256         read_rtl_function_c_tests.
42257         * selftest.h  (selftest::read_rtl_function_c_tests): New decl.
42258         * tree-dfa.c (ssa_default_def): Return NULL_TREE for rtl function
42259         dumps.
42261 2017-01-05  Uros Bizjak  <ubizjak@gmail.com>
42263         * config/i386/i386.md (*testqi_ext_3): No need to handle memory
42264         operands in a special way.  Assert that pos+len <= mode precision.
42266 2017-01-05  Jakub Jelinek  <jakub@redhat.com>
42268         * common.opt (fvect-cost-model): Remove RejectNegative flag, use
42269         3 argument Alias with unlimited for the negative form.
42270         (fno-vect-cost-model): Removed.
42272 2017-01-05  Martin Liska  <mliska@suse.cz>
42274         * hsa-gen.c (gen_hsa_divmod): New function.
42275         (gen_hsa_insn_for_internal_fn_call): Use the function for IFN_DIVMOD.
42277 2017-01-05  Martin Liska  <mliska@suse.cz>
42279         PR pch/78970
42280         * gcc.c (lookup_compiler): Reject '-' filename for a precompiled
42281         header.
42283 2017-01-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
42285         * config/s390/s390.c (s390_expand_setmem): Unroll the loop for
42286         small constant length operands.
42288 2017-01-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
42290         * config/s390/s390.c (s390_expand_setmem): Avoid overlapping bytes
42291         between loop iterations.
42293 2017-01-05  Martin Liska  <mliska@suse.cz>
42295         PR sanitizer/78815
42296         * gimplify.c (gimplify_decl_expr): Compare to
42297         asan_poisoned_variables instread of checking flags.
42298         (gimplify_target_expr): Likewise.
42299         (gimplify_expr): Likewise.
42300         (gimplify_function_tree): Conditionally initialize
42301         asan_poisoned_variables.
42303 2017-01-04  Jeff Law  <law@redhat.com>
42305         PR tree-optimizatin/78812
42306         * rtl.h (contains_mem_rtx_p): Prototype.
42307         * ifcvt.c (containts_mem_rtx_p): Move from here to...
42308         * rtlanal.c (contains_mem_rtx_p): Here and remove static linkage.
42309         * gcse.c (prune_expressions): Use contains_mem_rtx_p to discover
42310         and prune MEMs that are not at the toplevel of a SET_SRC rtx.  Look
42311         through ZERO_EXTEND and SIGN_EXTEND when trying to avoid pruning MEMs.
42313 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42315         * input.c (assert_char_at_range): Default-initialize actual_range.
42317 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42319         * df-scan.c (df_ref_create_structure): Make regno unsigned,
42320         to match the caller.
42322 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42324         * cfgexpand.c (expand_gimple_basic_block): Disregard debug
42325         insns after final jump in test to emit dummy move.
42327 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42329         * gimple-iterator.h (gsi_one_nondebug_before_end_p): New.
42330         * tree-eh.c (cleanup_empty_eh): Skip more debug stmts.
42332 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42334         * multiple_target.c (create_dispatcher_calls): Init e_next.
42335         * tree-ssa-loop-split.c (split_loop): Init border.
42336         * tree-vect-loop.c (vect_determine_vectorization_factor): Init
42337         scalar_type.
42339 2017-01-04  Michael Meissner  <meissner@linux.vnet.ibm.com>
42341         PR target/71977
42342         PR target/70568
42343         PR target/78823
42344         * config/rs6000/predicates.md (sf_subreg_operand): New predicate.
42345         (altivec_register_operand): Do not return true if the operand
42346         contains a SUBREG mixing SImode and SFmode.
42347         (vsx_register_operand): Likewise.
42348         (vsx_reg_sfsubreg_ok): New predicate.
42349         (vfloat_operand): Do not return true if the operand contains a
42350         SUBREG mixing SImode and SFmode.
42351         (vint_operand): Likewise.
42352         (vlogical_operand): Likewise.
42353         (gpc_reg_operand): Likewise.
42354         (int_reg_operand): Likewise.
42355         * config/rs6000/rs6000-protos.h (valid_sf_si_move): Add declaration.
42356         * config/rs6000/rs6000.c (valid_sf_si_move): New function to
42357         determine if a MOVSI or MOVSF operation contains SUBREGs that mix
42358         SImode and SFmode.
42359         (rs6000_emit_move_si_sf_subreg): New helper function.
42360         (rs6000_emit_move): Call rs6000_emit_move_si_sf_subreg to possbily
42361         fixup SUBREGs involving SImode and SFmode.
42362         * config/rs6000/vsx.md (SFBOOL_*): New constants that are operand
42363         numbers for the new peephole2 optimization.
42364         (peephole2 for SFmode unions): New peephole2 to optimize cases in
42365         the GLIBC math library that do AND/IOR/XOR operations on single
42366         precision floating point.
42367         * config/rs6000/rs6000.h (TARGET_NO_SF_SUBREG): New internal
42368         target macros to say whether we need to avoid SUBREGs mixing
42369         SImode and SFmode.
42370         (TARGET_ALLOW_SF_SUBREG): Likewise.
42371         * config/rs6000/rs6000.md (UNSPEC_SF_FROM_SI): New unspecs.
42372         (UNSPEC_SI_FROM_SF): Likewise.
42373         (iorxor): Change spacing.
42374         (and_ior_xor): New iterator for AND, IOR, and XOR.
42375         (movsi_from_sf): New insns for SImode/SFmode SUBREG support.
42376         (movdi_from_sf_zero_ext): Likewise.
42377         (mov<mode>_hardfloat, FMOVE32 iterator): Use register_operand
42378         instead of gpc_reg_operand.  Add SImode/SFmode SUBREG support.
42379         (movsf_from_si): New insn for SImode/SFmode SUBREG support.
42380         (fma<mode>4): Use gpc_reg_operand instead of register_operand.
42381         (fms<mode>4): Likewise.
42382         (fnma<mode>4): Likewise.
42383         (fnms<mode>4): Likewise.
42384         (nfma<mode>4): Likewise.
42385         (nfms<mode>4): Likewise.
42387 2017-01-04  Marek Polacek  <polacek@redhat.com>
42389         PR c++/64767
42390         * doc/invoke.texi: Document -Wpointer-compare.
42392 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42394         * optc-gen.awk: Emit #error for -W*/-f*/-m* Enum without
42395         RejectNegative.
42397         * dwarf2out.c (output_loc_list): Don't throw away 64K+ location
42398         descriptions for -gdwarf-5 and emit them as uleb128 instead of
42399         2-byte data.
42401 2017-01-04  Kelvin Nilsen  <kelvin@gcc.gnu.org>
42403         PR target/78056
42404         * doc/sourcebuild.texi (PowerPC-specific attributes): Add
42405         documentation of the powerpc_popcntb_ok attribute.
42406         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
42407         code to issue warning messages if a requested CPU configuration is
42408         not supported by the binary (assembler and loader) toolchain.
42409         (spe_init_builtins): Add two assertions to prevent ICE if attempt is
42410         made to define a built-in function that has been disabled.
42411         (paired_init_builtins): Add assertion to prevent ICE if attempt is
42412         made to define a built-in function that has been disabled.
42413         (altivec_init_builtins): Add comment explaining why definition
42414         of the DST built-in functions is not preceded by an assertion
42415         check.  Add assertions to prevent ICE if attempts are made to
42416         define an altivec predicate or an abs* built-in function that has
42417         been disabled.
42418         (htm_init_builtins): Add comment explaining why definition of the
42419         htm built-in functions is not preceded by an assertion check.
42421 2017-01-04  Jeff Law  <law@redhat.com>
42423         PR tree-optimizatin/67955
42424         * tree-ssa-alias.c (same_addr_size_stores_p): Check offsets first.
42425         Allow any SSA_VAR_P as the base objects.  Use integer_zerop.  Verify
42426         the points-to solution does not include pt_null.  Use DECL_PT_UID
42427         unconditionally.
42429 2017-01-04  Uros Bizjak  <ubizjak@gmail.com>
42431         * config/i386/i386.md (HI/SImode test with imm to QImode splitters):
42432         Use gen_int_mode instead of gen_lopwart for const_int operands.
42434 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42436         PR tree-optimization/71563
42437         * match.pd: Simplify X << Y into X if Y is known to be 0 or
42438         out of range value - has low bits known to be zero.
42440 2017-01-04  Alan Modra  <amodra@gmail.com>
42442         * Makefile.in (aclocal_deps): Update and order as per aclocal.m4.
42443         * configure: Regenerate.
42444         * config.in: Regenerate.
42446 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42448         PR bootstrap/77569
42449         * input.c (ebcdic_execution_charset::on_error): Don't use strstr for
42450         a substring of the message, but strcmp with the whole message.  Ifdef
42451         ENABLE_NLS, translate the message first using dgettext.
42453 2017-01-03  Jeff Law  <law@redhat.com>
42455         PR tree-optimizatin/78856
42456         * tree-ssa-threadupdate.c: Include tree-vectorizer.h.
42457         (mark_threaded_blocks): Remove code to truncate thread paths that
42458         cross multiple loop headers.  Instead invalidate the cached loop
42459         iteration information and handle case of a thread path walking
42460         into an irreducible region.
42462 2017-01-03  Michael Meissner  <meissner@linux.vnet.ibm.com>
42464         PR target/78900
42465         * config/rs6000/rs6000.c (rs6000_split_signbit): Change some
42466         assertions.  Add support for doing the signbit if the IEEE 128-bit
42467         floating point value is in a GPR.
42468         * config/rs6000/rs6000.md (Fsignbit): Delete.
42469         (signbit<mode>2_dm): Delete using <Fsignbit> and just use "wa".
42470         Update the length attribute if the value is in a GPR.
42471         (signbit<mode>2_dm_<su>ext): Add combiner pattern to eliminate
42472         the sign or zero extension instruction, since the value is always 0/1.
42473         (signbit<mode>2_dm2): Delete using <Fsignbit>.
42475         PR target/78953
42476         * config/rs6000/vsx.md (vsx_extract_<mode>_store_p9): If we are
42477         extracting SImode to a GPR register so that we can generate a
42478         store, limit the vector to be in a traditional Altivec register
42479         for the vextuwrx instruction.
42481 2017-01-03  Ian Lance Taylor  <iant@google.com>
42483         * godump.c (go_format_type): Treat ENUMERAL_TYPE like INTEGER_TYPE.
42485 2017-01-03  Martin Sebor  <msebor@redhat.com>
42487         PR tree-optimization/78696
42488         * gimple-ssa-sprintf.c (format_floating): Correct handling of
42489         precision.  Use MPFR for %f for greater fidelity.  Correct handling
42490         of %g.
42491         (pass_sprintf_length::compute_format_length): Set width and precision
42492         specified by asrerisk to void_node for vararg functions.
42493         (try_substitute_return_value): Adjust dump output.
42495 2017-01-03  David Edelsohn  <dje.gcc@gmail.com>
42497         * doc/invoke.texi (RS6000 options): LRA is enabled by default.
42499 2017-01-03  Eric Botcazou  <ebotcazou@adacore.com>
42501         * doc/invoke.texi (SPARC options): Document -mlra as the default.
42502         * config/sparc/sparc.c (sparc_option_override): Force LRA unless
42503         -mlra/-mno-lra was passed to the compiler.
42505 2017-01-03  James Cowgill  <James.Cowgill@imgtec.com>
42507         PR rtl-optimization/65618
42508         * emit-rtl.c (try_split): Move initialization of "before" and
42509         "after" to just before the call to emit_insn_after_setloc.
42511 2017-01-03  Gerald Pfeifer  <gerald@pfeifer.com>
42513         * doc/md.texi (Standard Names): Remove reference to Java frontend.
42515 2017-01-03  Pierre-Marie de Rodat  <derodat@adacore.com>
42517         * dwarf2out.c (gen_enumeration_type_die): When
42518         -gno-strict-dwarf, add a DW_AT_encoding attribute.
42520 2017-01-03  Jakub Jelinek  <jakub@redhat.com>
42522         PR tree-optimization/78965
42523         * gimple-ssa-sprintf.c (pass_sprintf_length::compute_format_length):
42524         Change first argument from const call_info & to call_info &.  For %n
42525         set info.nowrite to false.
42527         PR middle-end/78901
42528         * gimple-ssa-sprintf.c (try_substitute_return_value): Don't change
42529         possibly throwing calls.
42531         * genmatch.c (dt_node::gen_kids_1): If generic_exprs include SSA_NAME
42532         and exprs_len || fns_len, emit the code for SSA_NAME next to the exprs
42533         and fns handling, rather than in a separate case SSA_NAME.
42535 2017-01-02  Jeff Law  <law@redhat.com>
42537         * config/darwin-driver.c (darwin_driver_init): Const-correctness
42538         fixes for first_period and second_period variables.
42540 2017-01-02  Uros Bizjak  <ubizjak@gmail.com>
42542         PR target/78967
42543         * config/i386/i386.md (UNSPEC_NOREX_MEM): New unspec.
42544         (*insvqi_1): New insn pattern.
42545         (*insvqi_1_mem_rex64): Ditto.
42546         (*insvqi_2): Ditto.
42547         (*insvqi_3): Rename from *insvqi.
42549         (*extzvqi_mem_rex64): Add UNSPEC_NOREX_MEM tag.
42551 2017-01-02  Gerald Pfeifer  <gerald@pfeifer.com>
42553         * doc/cfg.texi (Edges): Remove reference to Java.
42554         (Maintaining the CFG): Ditto.
42556 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
42558         PR middle-end/77674
42559         * symtab.c (symtab_node::binds_to_current_def_p): Fix handling of
42560         transparent aliases.
42562 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
42564         PR middle-end/77484
42565         * predict.def (PRED_CALL): Update hitrate.
42566         (PRED_INDIR_CALL, PRED_POLYMORPHIC_CALL): New predictors.
42567         * predict.c (tree_estimate_probability_bb): Split CALL predictor
42568         into direct/indirect/polymorphic variants.
42570 2017-01-01  Jakub Jelinek  <jakub@redhat.com>
42572         Update copyright years.
42574         * gcc.c (process_command): Update copyright notice dates.
42575         * gcov-dump.c (print_version): Ditto.
42576         * gcov.c (print_version): Ditto.
42577         * gcov-tool.c (print_version): Ditto.
42578         * gengtype.c (create_file): Ditto.
42579         * doc/cpp.texi: Bump @copying's copyright year.
42580         * doc/cppinternals.texi: Ditto.
42581         * doc/gcc.texi: Ditto.
42582         * doc/gccint.texi: Ditto.
42583         * doc/gcov.texi: Ditto.
42584         * doc/install.texi: Ditto.
42585         * doc/invoke.texi: Ditto.
42587 Copyright (C) 2017 Free Software Foundation, Inc.
42589 Copying and distribution of this file, with or without modification,
42590 are permitted in any medium without royalty provided the copyright
42591 notice and this notice are preserved.