1 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
4 return stmt_vec_infos rather than gimple stmts. Do not accept
6 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
8 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
9 Update use of get_later_stmt.
10 (vect_get_constant_vectors): Update call accordingly.
11 (vect_schedule_slp_instance): Likewise
12 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
13 (vect_slp_analyze_instance_dependence): Likewise.
14 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
16 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
18 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
19 (stmt_info_for_cost::stmt_info): ...this new field.
20 (add_stmt_costs): Update accordingly.
21 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
22 (vect_get_known_peeling_cost): Likewise.
23 (vect_estimate_min_profitable_iters): Likewise.
24 * tree-vect-stmts.c (record_stmt_cost): Likewise.
26 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
28 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
29 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
30 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
32 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
34 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
36 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
37 a gimple stmt to a stmt_vec_info.
38 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
40 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
42 * tree-vectorizer.h (vec_info::grouped_stores): Change from
43 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
44 (_loop_vec_info::reduction_chains): Likewise.
45 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
47 * tree-vect-slp.c (vect_analyze_slp): Likewise.
49 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
51 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
52 a gimple stmt to a stmt_vec_info.
53 (_stmt_vec_info::next_element): Likewise.
54 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
55 (vect_slp_analyze_and_verify_node_alignment)
56 (vect_analyze_group_access_1, vect_analyze_group_access)
57 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
58 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
59 (vect_supportable_dr_alignment): Update accordingly.
60 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
61 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
62 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
63 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
64 * tree-vect-slp.c (vect_build_slp_tree_1)
65 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
66 (vect_split_slp_store_group, vect_analyze_slp_instance)
67 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
68 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
69 (get_group_load_store_type, get_load_store_type)
70 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
71 (vect_transform_stmt, vect_remove_stores): Likewise.
73 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
75 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
77 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
78 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
79 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
80 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
81 (vector_alignment_reachable_p, vect_get_data_access_cost)
82 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
83 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
84 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
85 (vect_analyze_group_access_1, vect_analyze_group_access)
86 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
87 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
88 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
89 result of vect_dr_stmt and use the stmt_vec_info instead of
90 the associated gimple stmt.
91 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
92 (vect_gen_prolog_loop_niters): Likewise.
93 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
95 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
97 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
98 to a vec<stmt_vec_info>.
99 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
100 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
102 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
103 to a vec<stmt_vec_info>.
104 (bst_traits::value_type, bst_traits::value_type): Likewise.
105 (bst_traits::hash): Update accordingly.
106 (vect_get_and_check_slp_defs): Change the stmts parameter from
107 a vec<gimple *> to a vec<stmt_vec_info>.
108 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
109 (vect_build_slp_tree): Likewise.
110 (vect_build_slp_tree_2): Likewise. Update uses of
111 SLP_TREE_SCALAR_STMTS.
112 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
113 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
114 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
115 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
116 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
117 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
118 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
119 (vect_get_constant_vectors, vect_get_slp_defs)
120 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
121 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
122 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
123 instead of gimple stmts.
124 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
125 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
126 (vect_slp_analyze_instance_dependence): Update uses of
127 SLP_TREE_SCALAR_STMTS.
128 (vect_slp_analyze_and_verify_node_alignment): Likewise.
129 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
130 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
131 (get_initial_defs_for_reduction): Likewise.
132 (vect_create_epilog_for_reduction): Likewise.
133 (vectorize_fold_left_reduction): Likewise.
134 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
135 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
136 (can_vectorize_live_stmts): Likewise.
138 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
140 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
141 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
142 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
144 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
146 (gather_scalar_reductions): Update after above interface changes.
147 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
148 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
150 (vect_force_simple_reduction): Likewise.
151 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
152 LOOP_VINFO_REDUCTIONS.
153 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
155 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
157 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
158 a gimple stmt to a stmt_vec_info.
159 * tree-vect-loop.c (vect_active_double_reduction_p)
160 (vect_force_simple_reduction, vectorizable_reduction): Update
163 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
165 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
166 vec<gimple *> to a vec<stmt_vec_info>.
167 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
168 the reduction_phis argument from a vec<gimple *> to a
170 (vectorizable_reduction): Likewise the phis local variable that
171 is passed to vect_create_epilog_for_reduction. Update for new type
172 of SLP_TREE_VEC_STMTS.
173 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
174 (vectorizable_live_operation): Likewise.
175 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
176 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
178 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
180 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
181 a gimple stmt to a stmt_vec_info.
182 (vectorizable_condition, vectorizable_live_operation)
183 (vectorizable_reduction, vectorizable_induction): Pass back the
184 vectorized statement as a stmt_vec_info.
185 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
186 use of STMT_VINFO_VEC_STMT.
187 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
188 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
189 as stmt_vec_infos rather than gimple stmts.
190 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
192 (vectorizable_live_operation): Likewise.
193 (vectorizable_reduction, vectorizable_induction): Likewise,
194 updating use of STMT_VINFO_VEC_STMT.
195 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
196 of STMT_VINFO_VEC_STMT.
197 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
198 (vectorizable_simd_clone_call, vectorizable_conversion)
199 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
200 (vectorizable_store, vectorizable_load, vectorizable_condition)
201 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
202 from a gimple stmt to a stmt_vec_info.
203 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
204 pointer to a stmt_vec_info to the vectorizable_* routines.
206 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
208 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
209 a gimple stmt to a stmt_vec_info.
210 (is_pattern_stmt_p): Update accordingly.
211 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
212 (vect_record_grouped_load_vectors): Likewise.
213 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
214 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
215 (vect_model_reduction_cost): Likewise.
216 (vect_create_epilog_for_reduction): Likewise.
217 (vectorizable_reduction, vectorizable_induction): Likewise.
218 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
219 Return the stmt_vec_info for the pattern statement.
220 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
221 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
222 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
223 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
224 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
225 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
226 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
227 (free_stmt_vec_info, vect_is_simple_use): Likewise.
229 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
231 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
232 (vect_finish_stmt_generation): Likewise.
233 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
234 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
235 (vect_build_gather_load_calls): Use the return value of the above
236 functions instead of a separate call to vinfo_for_stmt. Use narrow
237 scopes for the input gimple stmt and wider scopes for the associated
238 stmt_vec_info. Use vec_info::lookup_def when setting these
239 stmt_vec_infos from an SSA_NAME definition.
240 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
241 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
242 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
243 (vectorizable_store, vectorizable_load, vectorizable_condition)
244 (vectorizable_comparison): Likewise.
245 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
246 (vectorizable_reduction): Likewise.
248 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
250 * tree-vectorizer.h (vect_is_simple_use): Add an optional
251 stmt_vec_info * parameter before the optional gimple **.
252 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
253 (process_use, vect_get_vec_def_for_operand_1): Update callers.
254 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
255 * tree-vect-loop.c (vectorizable_reduction): Likewise.
256 (vectorizable_live_operation): Likewise.
257 * tree-vect-patterns.c (type_conversion_p): Likewise.
258 (vect_look_through_possible_promotion): Likewise.
259 (vect_recog_rotate_pattern): Likewise.
260 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
262 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
264 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
265 a typedef to a wrapper class.
266 (NULL_STMT_VEC_INFO): New macro.
267 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
268 (stmt_vec_info::operator*): New function.
269 (stmt_vec_info::operator gimple *): Likewise.
270 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
271 (add_stmt_costs): Likewise.
272 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
273 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
274 (vect_get_known_peeling_cost): Likewise.
275 (vect_estimate_min_profitable_iters): Likewise.
276 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
277 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
278 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
279 (vectorizable_store, free_stmt_vec_infos): Likewise.
280 (new_stmt_vec_info): Change return type of xcalloc to
283 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
285 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
286 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
287 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
288 a single_imm_use-based sequence.
289 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
291 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
293 * tree-vectorizer.h (vec_info::lookup_def): Declare.
294 * tree-vectorizer.c (vec_info::lookup_def): New function.
295 * tree-vect-patterns.c (vect_get_internal_def): Use it.
296 (vect_widened_op_tree): Likewise.
297 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
298 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
299 (vectorizable_reduction): Likewise.
300 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
302 (vect_is_slp_reduction): Update calls accordingly. Use
303 vec_info::lookup_def.
304 (vect_is_simple_reduction): Likewise
305 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
307 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
309 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
310 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
311 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
313 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
314 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
315 (vect_update_vf_for_slp, vect_analyze_loop_operations)
316 (vect_is_slp_reduction, vectorizable_induction)
317 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
318 * tree-vect-patterns.c (vect_init_pattern_stmt):
319 (vect_determine_min_output_precision_1, vect_determine_precisions)
320 (vect_pattern_recog): Likewise.
321 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
322 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
323 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
324 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
325 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
326 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
327 info field from a loop to a loop_vec_info.
329 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
331 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
332 (vec_info::add_stmt): Declare.
333 * tree-vectorizer.c (vec_info::add_stmt): New function.
334 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
335 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
336 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
337 (vectorizable_induction): Likewise.
338 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
339 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
340 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
341 (vectorizable_load): Likewise.
342 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
343 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
344 (vect_recog_gather_scatter_pattern): Likewise.
345 (append_pattern_def_seq): Likewise. Remove a check that is
346 performed by add_stmt itself.
348 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
350 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
351 which make_ssa_name was called with new_stmt before new_stmt
354 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
356 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
358 (vect_is_slp_reduction): ...here...
359 (vect_is_simple_reduction): ...and here. Remove repetition of tests
360 that are already known to be false.
362 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
364 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
365 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
366 STMT_VINFO_NUM_SLP_USES when it's true.
367 (vect_free_slp_instance): Add a final_p parameter and pass it to
369 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
370 (vect_analyze_slp_instance): Likewise.
371 (vect_slp_analyze_operations): Likewise.
372 (vect_slp_analyze_bb_1): Likewise.
373 * tree-vectorizer.c (vec_info): Likewise.
374 * tree-vect-loop.c (vect_transform_loop): Likewise.
376 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
378 * tree-vect-loop.c (vectorizable_reduction): Assert that the
379 function is not called for second and subsequent members of
382 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
384 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
385 cases for nested loops from here to ...
386 (vect_create_epilog_for_reduction): ...here. Only call
387 vect_is_simple_use for inner-loop reductions.
389 2018-07-31 Martin Liska <mliska@suse.cz>
391 PR gcov-profile/85338
392 PR gcov-profile/85350
393 PR gcov-profile/85372
394 * profile.c (struct location_triplet): New.
395 (struct location_triplet_hash): Likewise.
396 (output_location): Do not output a BB that
397 is already recorded for a line.
398 (branch_prob): Use streamed_locations.
400 2018-07-31 Martin Liska <mliska@suse.cz>
402 PR gcov-profile/85370
403 * coverage.c (coverage_begin_function): Do not mark target
404 clones as artificial functions.
406 2018-07-31 Martin Liska <mliska@suse.cz>
408 PR gcov-profile/83813
409 PR gcov-profile/84758
410 PR gcov-profile/85217
411 PR gcov-profile/85332
412 * profile.c (branch_prob): Do not record GOTO expressions
413 for GIMPLE statements which locations are already streamed.
415 2018-07-31 Olivier Hainque <hainque@adacore.com>
417 * gcc.c (handle_spec_function): Accept a soft_matched_part
418 argument, as do_spec_1. Pass it down to ...
419 (eval_spec_function): Accept a soft_matched_part argument,
420 and pass it down to ...
421 (do_spec_2): Accept a soft_matched_part argument, and pass
422 it down to do_spec_1.
423 (do_spec_1): Pass soft_matched_part to handle_spec_function.
424 (handle_braces): Update call to handle_spec_function.
425 (driver::set_up_specs): Update calls to do_spec_2.
426 (compare_debug_dump_opt_spec_function): Likewise.
427 (compare_debug_self_opt_spec_function): Likewise.
429 2018-07-31 Olivier Hainque <hainque@adacore.com>
431 * common.opt (nolibc): New option.
432 * doc/invoke.texi (Link Options): Document it.
433 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
434 * config/alpha/linux.h: Likewise.
435 * config/arc/elf.h: Likewise.
436 * config/arm/uclinux-elf.h: Likewise.
437 * config/arm/unknown-elf.h: Likewise.
438 * config/avr/avrlibc.h: Likewise.
439 * config/bfin/bfin.h: Likewise.
440 * config/bfin/linux.h: Likewise.
441 * config/bfin/uclinux.h: Likewise.
442 * config/darwin.h: Likewise.
443 * config/darwin10.h: Likewise.
444 * config/darwin12.h: Likewise.
445 * config/gnu-user.h: Likewise.
446 * config/lm32/uclinux-elf.h: Likewise.
447 * config/pa/pa-hpux11.h: Likewise.
448 * config/pa/pa64-hpux.h: Likewise.
449 * config/sparc/sparc.h: Likewise.
451 2018-07-31 Olivier Hainque <hainque@adacore.com>
453 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
456 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
459 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
462 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
464 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
465 terminated string literal.
467 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
469 PR rtl-optimization/85160
470 * combine.c (is_just_move): New function.
471 (try_combine): Allow combining two instructions into two if neither of
472 the original instructions was a move.
474 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
477 * doc/extend.texi (Global Register Variables): Discourage use of type
479 (Local Register Variables): Likewise.
481 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
483 PR tree-optimization/86506
484 * hwint.h (ceil_log2): Resync with hwint.c implementation.
486 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
489 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
490 hard_regno, make sure no insn between `from` and `to` clobbers it.
492 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
493 Tom de Vries <tdevries@suse.de>
495 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
496 (PTX_DEFAULT_RUNTIME_DIM): ... this.
497 (nvptx_goacc_validate_dims): Set default worker and gang dims to
498 PTX_DEFAULT_RUNTIME_DIM.
499 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
501 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
503 * config/pa/pa.c (pa_output_addr_vec): Align address table.
504 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
505 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
507 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
509 * config/rs6000/constraints.md (wG constraint): Delete, no longer
511 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
512 predicate to reflect toc fusion has been deleted.
513 (toc_fusion_mem_raw): Delete, no longer used.
514 (toc_fusion_mem_wrapped): Likewise.
515 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
517 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
518 Delete, no longer used.
519 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
520 meant to be used for toc fusion.
521 (rs6000_debug_print_mode): Delete toc fusion debugging.
522 (rs6000_debug_reg_global): Likewise.
523 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
524 fusion and secondary reload support that were never used.
525 (rs6000_option_override_internal): Delete TOC fusion, that was only
526 partially defined, and it did not work unless you also used the
528 (rs6000_legitimate_address_p): Delete TOC fusion support.
529 (rs6000_opt_masks): Likewise.
530 (fusion_wrap_memory_address): Delete function, no longer used.
531 (fusion_split_address); Delete TOC fusion support.
532 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
533 longer used with toc fusion being deleted.
534 (TARGET_TOC_FUSION_FP): Likewise.
535 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
537 (toc fusion spliter): Delete TOC fusion support.
538 (toc_fusionload_<mode>): Likewise.
539 (toc_fusionload_di): Likewise.
540 (fusion_gpr_load_<mode>): Delete generator function, this insn no
541 longer needs to be named. Rename predicate to delete TOC fusion.
542 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
543 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
544 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
545 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
546 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
548 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
550 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
551 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
552 __int128 in built-in function prototypes.
553 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
554 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
556 2018-07-27 Martin Sebor <msebor@redhat.com>
558 PR tree-optimization/86696
559 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
560 types, including enums.
561 (handle_char_store): Be prepared for the above function to fail.
563 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
565 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
566 when optimization level is lower than 2 or optimize for size.
568 2018-07-26 Martin Sebor <msebor@redhat.com>
570 PR tree-optimization/86043
571 PR tree-optimization/86042
572 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
573 (get_string_cst_length): Rename...
574 (get_min_string_length): ...to this. Add argument.
575 (handle_char_store): Extend to handle multi-character stores by
577 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
578 * tree.h (initializer_zerop): Add argument.
580 2018-07-26 Jakub Jelinek <jakub@redhat.com>
583 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
584 declare target to variables if they have always,{to,from,tofrom} map
587 2018-07-26 Martin Liska <mliska@suse.cz>
590 * lto-wrapper.c: Add linker_output as prefix
591 for ltrans_output_file.
593 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
595 PR rtl-optimization/85805
596 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
597 value for hard registers if that was written in the same mode.
599 2018-07-26 Martin Liska <mliska@suse.cz>
601 PR gcov-profile/86536
602 * gcov.c (format_gcov): Use printf format %.*f directly
603 and do not handle special values.
605 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
607 * common/config/arc/arc-common.c (arc_option_optimization_table):
608 Update default optimizations for size.
610 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
612 * config/arc/arc.md (movsf_insn): Add short instruction selection.
613 * config/arc/constraints.md (CfZ): New constraint.
614 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
615 (subsf3_fpu): Likewise.
616 (cmpsf_fpu): Likewise.
617 (cmpsf_fpu_uneq): Likewise.
619 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
621 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
623 (arc_is_uncached_mem_p): uncached applies to both the variable and
626 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
628 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
631 2018-07-25 David Malcolm <dmalcolm@redhat.com>
633 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
634 field "m_scopes" from vec to auto_vec.
636 2018-07-25 Martin Liska <mliska@suse.cz>
638 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
641 2018-07-25 Richard Biener <rguenther@suse.de>
644 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
645 special wrt context_die late.
646 (gen_subprogram_die): Re-use DIEs in local scope.
648 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
650 PR tree-optimization/86644
651 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
653 2018-07-25 Martin Liska <mliska@suse.cz>
656 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
657 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
659 2018-07-25 Martin Liska <mliska@suse.cz>
662 * params.def: Explain ASan abbreviation and provide
663 a documentation link.
665 2018-07-24 Martin Sebor <msebor@redhat.com>
667 PR tree-optimization/86622
668 PR tree-optimization/86532
669 * builtins.h (string_length): Declare.
670 * builtins.c (c_strlen): Correct handling of non-constant offsets.
671 (check_access): Be prepared for non-constant length ranges.
672 (string_length): Make extern.
673 * expr.c (string_constant): Only handle the minor non-constant
674 array index. Use string_constant to compute the length of
675 a generic string constant.
677 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
679 PR tree-optimization/86618
680 * tree-vect-stmts.c (vectorizable_call): Don't take the address
681 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
683 2018-07-24 David Malcolm <dmalcolm@redhat.com>
685 PR tree-optimization/86636
686 * json.cc (json::object::set): Fix comment. Add assertions.
687 (json::array::append): Move here from json.h. Add comment and an
689 (json::string::string): Likewise.
690 * json.h (json::array::append): Move to json.cc.
691 (json::string::string): Likewise.
692 * optinfo-emit-json.cc
693 (optrecord_json_writer::impl_location_to_json): Assert that we
694 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
695 wrapper around it. Expand the location once, rather than three
697 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
698 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
700 (optrecord_json_writer::optinfo_to_json): Likewise, in four
701 places. Fix some overlong lines.
703 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
705 * config/aarch64/aarch64-simd.md
706 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
707 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
708 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
709 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
710 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
711 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
712 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
713 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
714 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
716 2018-07-24 Jakub Jelinek <jakub@redhat.com>
719 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
720 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
721 and abs_d == d, do the power of two handling if profitable.
723 2018-07-24 Richard Biener <rguenther@suse.de>
725 * match.pd: Add BIT_FIELD_REF canonicalizations.
727 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
730 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
732 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
734 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
735 terminated STRING_CST object.
737 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
739 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
741 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
743 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
744 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
745 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
746 the elements into a register.
747 (rs6000_split_v4si_init_di_reg): Delete.
748 (rs6000_split_v4si_init): Delete.
749 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
750 (vsx_init_v4si): Rewrite as a define_expand.
752 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
754 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
755 zero_extend argument from memory): New.
757 2018-07-22 Martin Sebor <msebor@redhat.com>
760 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
761 diagnosing calls with unknown arguments unless -Walloca-larger-than
762 is restricted to less than PTRDIFF_MAX bytes.
764 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
766 * doc/gcov.texi (Invoking Gcov): Editorial changes.
768 2018-07-20 David Malcolm <dmalcolm@redhat.com>
770 * pretty-print.c (text_info::set_location): Remove redundant
771 "line_table" parameter from call to rich_location::set_range.
773 2018-07-20 Martin Sebor <msebor@redhat.com>
776 * builtins.c (expand_builtin_alloca): Adjust.
777 * calls.c (alloc_max_size): Simplify.
778 * cgraphunit.c (cgraph_node::expand): Adjust.
779 * common.opt (larger_than_size, warn_frame_larger_than): Remove
781 (frame_larger_than_size): Same.
782 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
783 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
785 * doc/invoke.texi (GCC Command Options): Document option arguments.
786 Explain byte-size arguments and suffixes.
787 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
788 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
789 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
790 * doc/options.texi (UInteger): Expand.
791 (Host_Wide_Int, ByteSize): Document new properties.
792 * final.c (final_start_function_1): Include sizes in an error message.
793 * function.c (frame_offset_overflow): Same.
794 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
795 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
796 Diagnose unbounded alloca calls only for limits of less than
798 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
799 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
801 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
802 only for limits of less than PTRDIFF_MAX.
803 * langhooks-def.h (lhd_handle_option): Change function argument
805 * langhooks.c (lhd_handle_option): Same.
806 * langhooks.h (handle_option): Same.
807 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
809 (var_type, var_type_struct): Same.
810 (var_set): Handle ByteSize flag.
811 * optc-gen.awk: Add comments to output to ease debugging. Make
812 use of HOST_WIDE_INT where appropriate.
813 * opts-gen-save.awk: Use %lx to format unsigned long.
814 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
815 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
816 arguments. Parse bytes-size suffixes.
817 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
818 (enum_value_to_arg): Same.
819 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
820 (handle_option): Adjust.
821 (generate_option): Change function argument to HOST_WIDE_INT.
822 (cmdline_handle_error): Adjust.
823 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
824 (set_option): Change function argument to HOST_WIDE_INT.
825 (option_enabled): Handle cl_host_wide_int.
826 (get_option_state): Handle CLVC_SIZE.
827 (control_warning_option): Same.
828 * opts.c (common_handle_option): Change function argument to
829 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
830 OPT_Wvla_larger_than_.
831 * opts.h (enum cl_var_type): Add an enumerator.
832 * stor-layout.c (layout_decl): Print a more meaningful warning.
833 * toplev.c (output_stack_usage): Adjust.
835 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
837 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
838 call to inline_expand_builtin_string_cmp.
839 (expand_builtin_strcmp): Likewise.
840 (expand_builtin_strncmp): Likewise.
841 (inline_string_cmp): Delete the last parameter, change char_type_node
842 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
844 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
845 the inlining expansion on target where the type of the call has same or
846 narrower precision than unsigned char.
848 2018-07-20 David Malcolm <dmalcolm@redhat.com>
850 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
851 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
852 * common.opt (fsave-optimization-record): New option.
853 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
854 * doc/invoke.texi (-fsave-optimization-record): New option.
855 * dumpfile.c: Include "optinfo-emit-json.h".
856 (struct kv_pair): Move to coretypes.h.
857 (optgroup_options): Make non-static.
858 (dump_context::end_scope): Call
859 optimization_records_maybe_pop_dump_scope.
860 * dumpfile.h (optgroup_options): New decl.
863 * optinfo-emit-json.cc: New file.
864 * optinfo-emit-json.h: New file.
865 * optinfo.cc: Include "optinfo-emit-json.h".
866 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
867 (optinfo_enabled_p): Check optimization_records_enabled_p.
868 (optinfo_wants_inlining_info_p): Likewise.
869 * optinfo.h: Update comment.
870 * profile-count.c (profile_quality_as_string): New function.
871 * profile-count.h (profile_quality_as_string): New decl.
872 (profile_count::quality): New accessor.
873 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
874 and optinfo_emit_json_cc_tests.
875 * selftest.h (selftest::json_cc_tests): New decl.
876 (selftest::optinfo_emit_json_cc_tests): New decl.
877 * toplev.c: Include "optinfo-emit-json.h".
878 (compile_file): Call optimization_records_finish.
879 (do_compile): Call optimization_records_start.
880 * tree-ssa-live.c: Include optinfo.h.
881 (remove_unused_scope_block_p): Retain inlining information if
882 optinfo_wants_inlining_info_p returns true.
884 2018-07-20 Richard Biener <rguenther@suse.de>
887 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
888 to cover -flto-partition=none.
890 2018-07-20 Martin Liska <mliska@suse.cz>
892 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
893 (get_decl_source_range): Remove unused function.
895 2018-07-20 Richard Biener <rguenther@suse.de>
897 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
898 (struct vn_phi_s): Likewise.
899 (struct vn_reference_s): Likewise.
900 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
901 for searching the slot of an entry known to be in the hash itself.
902 (vn_phi_hasher::equal): Likewise.
903 (vn_reference_hasher::equal): Likewise.
904 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
906 (optimistic_info, current_info): Remove, keeping only valid_info.
907 (vn_reference_lookup_1): Remove fallback lookup.
908 (vn_reference_lookup_2): Likewise.
909 (vn_nary_op_lookup_1): Likewise.
910 (vn_phi_lookup): Likewise.
911 (vn_nary_build_or_lookup_1): Make sure to not chain the built
913 (vn_reference_insert): Adjust, chain the inserted hash element
914 at last_inserted_ref.
915 (vn_reference_insert_pieces): Likewise.
916 (visit_reference_op_call): Likewise.
917 (vn_nary_op_insert_into): Chain the inserted hash element at
919 (vn_nary_op_insert_pieces): Adjust.
920 (vn_nary_op_insert): Likewise.
921 (vn_nary_op_insert_stmt): Likewise.
922 (vn_phi_insert): Adjust, chain the inserted hash element at
924 (process_scc): Remove clearing and copying the optimistic
925 table. Instead remove elements inserted during an optimistic
926 iteration from the single table we maintain.
927 (init_scc_vn): Adjust.
928 (free_scc_vn): Likewise.
929 (sccvn_dom_walker::record_cond): Likewise.
930 (sccvn_dom_walker::after_dom_children): Likewise.
932 2018-07-19 Martin Sebor <msebor@redhat.com>
934 PR tree-optimization/84047
935 PR tree-optimization/83776
936 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
937 (check_array_bounds): Call it.
939 2018-07-19 Martin Sebor <msebor@redhat.com>
941 * align.h (align_flags): Use member initialization.
943 2018-07-19 David Malcolm <dmalcolm@redhat.com>
945 * Makefile.in (OBJS): Add optinfo.o.
946 * coretypes.h (class symtab_node): New forward decl.
947 (struct cgraph_node): New forward decl.
948 (class varpool_node): New forward decl.
949 * dump-context.h: New file.
950 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
952 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
953 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
954 (set_alt_dump_file): Likewise.
955 (dump_context::~dump_context): New dtor.
956 (dump_gimple_stmt): Move implementation to...
957 (dump_context::dump_gimple_stmt): ...this new member function.
958 Add the stmt to any pending optinfo, creating one if need be.
959 (dump_gimple_stmt_loc): Move implementation to...
960 (dump_context::dump_gimple_stmt_loc): ...this new member function.
961 Start a new optinfo and add the stmt to it.
962 (dump_gimple_expr): Move implementation to...
963 (dump_context::dump_gimple_expr): ...this new member function.
964 Add the stmt to any pending optinfo, creating one if need be.
965 (dump_gimple_expr_loc): Move implementation to...
966 (dump_context::dump_gimple_expr_loc): ...this new member function.
967 Start a new optinfo and add the stmt to it.
968 (dump_generic_expr): Move implementation to...
969 (dump_context::dump_generic_expr): ...this new member function.
970 Add the tree to any pending optinfo, creating one if need be.
971 (dump_generic_expr_loc): Move implementation to...
972 (dump_context::dump_generic_expr_loc): ...this new member
973 function. Add the tree to any pending optinfo, creating one if
975 (dump_printf): Move implementation to...
976 (dump_context::dump_printf_va): ...this new member function. Add
977 the text to any pending optinfo, creating one if need be.
978 (dump_printf_loc): Move implementation to...
979 (dump_context::dump_printf_loc_va): ...this new member function.
980 Start a new optinfo and add the stmt to it.
981 (dump_dec): Move implementation to...
982 (dump_context::dump_dec): ...this new member function. Add the
983 value to any pending optinfo, creating one if need be.
984 (dump_context::dump_symtab_node): New member function.
985 (dump_context::get_scope_depth): New member function.
986 (dump_context::begin_scope): New member function.
987 (dump_context::end_scope): New member function.
988 (dump_context::ensure_pending_optinfo): New member function.
989 (dump_context::begin_next_optinfo): New member function.
990 (dump_context::end_any_optinfo): New member function.
991 (dump_context::s_current): New global.
992 (dump_context::s_default): New global.
993 (dump_scope_depth): Delete global.
994 (dumpfile_ensure_any_optinfo_are_flushed): New function.
995 (dump_symtab_node): New function.
996 (get_dump_scope_depth): Reimplement in terms of dump_context.
997 (dump_begin_scope): Likewise.
998 (dump_end_scope): Likewise.
999 (selftest::temp_dump_context::temp_dump_context): New ctor.
1000 (selftest::temp_dump_context::~temp_dump_context): New dtor.
1001 (selftest::verify_item): New function.
1002 (ASSERT_IS_TEXT): New macro.
1003 (ASSERT_IS_TREE): New macro.
1004 (ASSERT_IS_GIMPLE): New macro.
1005 (selftest::test_capture_of_dump_calls): New test.
1006 (selftest::dumpfile_c_tests): Call it.
1007 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
1008 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
1009 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
1010 descriptive comment.
1011 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
1012 (dump_node, dump_bb): Move these unrelated decls.
1013 (class dump_manager): Add leading comment.
1014 * optinfo.cc: New file.
1015 * optinfo.h: New file.
1017 2018-07-19 Michael Collison <michael.collison@arm.com>
1018 Richard Henderson <rth@redhat.com>
1020 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
1021 (subti): Handle op1 zero.
1022 (subvti4, usub4ti4): New.
1023 (*sub<GPI>3_compare1_imm): New.
1024 (sub<GPI>3_carryinCV): New.
1025 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
1026 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
1028 2018-07-19 Michael Collison <michael.collison@arm.com>
1029 Richard Henderson <rth@redhat.com>
1031 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
1032 (addti3): Create simpler code if low part is already known to be 0.
1033 (addvti4, uaddvti4): New.
1034 (*add<GPI>3_compareC_cconly_imm): New.
1035 (*add<GPI>3_compareC_cconly): New.
1036 (*add<GPI>3_compareC_imm): New.
1037 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
1038 handle constants within this pattern..
1039 (*add<GPI>3_compareV_cconly_imm): New.
1040 (*add<GPI>3_compareV_cconly): New.
1041 (*add<GPI>3_compareV_imm): New.
1042 (add<GPI>3_compareV): New.
1043 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
1044 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
1045 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
1046 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
1047 with 'comparison' operator.
1048 (*add<GPI>3_compareV_cconly_imm): Ditto.
1049 (*add<GPI>3_compareV_cconly): Ditto.
1050 (*add<GPI>3_compareV_imm): Ditto.
1051 (add<GPI>3_compareV): Ditto.
1052 (add<mode>3_carryinC): Ditto.
1053 (*add<mode>3_carryinC_zero): Ditto.
1054 (*add<mode>3_carryinC): Ditto.
1055 (add<mode>3_carryinV): Ditto.
1056 (*add<mode>3_carryinV_zero): Ditto.
1057 (*add<mode>3_carryinV): Ditto.
1059 2018-07-19 Michael Collison <michael.collison@arm.com>
1060 Richard Henderson <rth@redhat.com>
1062 * config/aarch64/aarch64-modes.def (CC_V): New.
1063 * config/aarch64/aarch64-protos.h
1064 (aarch64_addti_scratch_regs): Declare
1065 (aarch64_subvti_scratch_regs): Declare.
1066 (aarch64_expand_subvti): Declare.
1067 (aarch64_gen_unlikely_cbranch): Declare
1068 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
1069 for signed overflow using CC_Vmode.
1070 (aarch64_get_condition_code_1): Handle CC_Vmode.
1071 (aarch64_gen_unlikely_cbranch): New function.
1072 (aarch64_addti_scratch_regs): New function.
1073 (aarch64_subvti_scratch_regs): New function.
1074 (aarch64_expand_subvti): New function.
1076 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
1078 * config/aarch64/aarch64-option-extensions.def: New entry for profile
1080 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
1081 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
1084 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
1087 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1088 address check not strict.
1090 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
1092 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
1094 (store_pair_lanes<mode>): Likewise.
1095 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
1096 enum value 'ADDR_QUERY_LDP_STP_N'.
1097 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
1098 (aarch64_print_address_internal): Add declaration.
1099 (aarch64_print_ldpstp_address): Remove.
1100 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
1101 (aarch64_print_operand): Change printing of 'y'.
1102 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
1103 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
1104 'true' rather than '1'.
1105 * config/aarch64/constraints.md (Uml): Likewise.
1106 (Uml): Rename to Umn.
1109 2018-07-19 Richard Biener <rguenther@suse.de>
1111 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
1113 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
1114 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
1115 (vn_reference_hasher): Likewise.
1116 (struct vn_tables_s): Remove obstack and alloc-pool members.
1117 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
1118 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
1119 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
1120 (vn_reference_insert_pieces): Likewise.
1121 (alloc_vn_nary_op_noinit): Adjust.
1122 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
1123 (vn_phi_eq): Adjust.
1124 (shared_lookup_phiargs): Remove.
1125 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
1126 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
1127 (visit_reference_op_call): Likewise.
1128 (copy_nary, copy_phi, copy_reference): Remove.
1129 (process_scc): Rewind the obstack when iterating. Do not
1130 copy the elements to valid_info but just move them from one
1131 hashtable to the other.
1132 (allocate_vn_table): Adjust.
1133 (free_vn_table): Likewise.
1134 (init_scc_vn): Likewise.
1135 (free_scc_vn): Likewise.
1137 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
1140 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
1141 indirect_return as function type attribute.
1142 (ix86_attribute_table): Change indirect_return to function
1144 * doc/extend.texi: Update indirect_return attribute.
1146 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
1148 * wide-int.h (widest2_int): New.
1149 * gimple-fold.c (arith_overflowed_p): Use it.
1150 * tree.h (widest2_int_cst): New.
1151 * tree-vrp.c (wide_int_binop_overflow): Rename from
1152 vrp_int_const_binop.
1153 Rewrite to work on trees.
1154 (extract_range_from_multiplicative_op_1): Abstract code to...
1155 (wide_int_range_min_max): ...here.
1156 (wide_int_range_cross_product): ...and here.
1157 (extract_range_from_binary_expr_1): Abstract overflow code to...
1158 (wide_int_range_mult_wrapping): ...here.
1159 * tree-vrp.h (wide_int_range_cross_product): New.
1160 (wide_int_range_mult_wrapping): New.
1162 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
1163 Julia Koval <julia.koval@intel.com>
1165 * config/i386/x86-tune-costs.h (skylake_memcpy,
1166 skylake_memset): Replace rep_prefix with unrolling for size 512.
1168 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
1171 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle comparision with EQ_EXPR
1174 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
1176 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
1177 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
1178 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
1179 previously known as "PowerPC AltiVec Built-in Functions". Move
1180 some material to new subsubsections "PowerPC AltiVec Built-in
1181 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
1183 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
1184 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
1185 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
1186 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
1188 2018-07-18 Richard Biener <rguenther@suse.de>
1190 PR tree-optimization/86557
1191 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
1194 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
1196 * config/s390/s390.c (s390_function_profiler): Generate CFI.
1198 2018-07-17 Jeff Law <law@redhat.com>
1200 * config/arm/arm.c (get_label_padding): Update for recent
1201 changes to label_to_alignment.
1203 PR tree-optimization/86010
1204 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
1206 * config/mips/mips.c (vr4130_align_insns): Update for recent
1207 changes to label_to_alignment.
1209 * config/frv/frv.c (frv_label_align): Update for recent changes
1210 to label_to_alignment.
1212 * config/nios2/nios2.c (nios2_label_align): Update for recent
1213 changes which dropped ALIGN_LABELS_LOG.
1215 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
1217 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
1218 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
1220 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
1222 * config/arc/arc.c (arc_label_align): Use align_labels instead of
1223 deprecated align_labels_log.
1225 2018-07-17 Richard Biener <rguenther@suse.de>
1228 * dwarf2out.c (init_sections_and_labels): Always generate
1229 a debug_line_str_section for early LTO debug.
1230 (dwarf2out_finish): Reset debug_line_str_hash output early.
1231 Bump counter for extra dwarf5 .debug_loc labels to not conflict
1233 (dwarf2out_early_finish): Output debug_line_str.
1235 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
1237 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
1238 index register on z196 or later.
1240 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
1242 * config/s390/s390.c (s390_default_align): Set default function
1244 (s390_override_options_after_change): Call s390_default align.
1245 (s390_option_override_internal): Call s390_default_align.
1246 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
1248 2018-07-17 Jakub Jelinek <jakub@redhat.com>
1251 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
1252 to _looptemp_ clauses, other than the first two.
1254 2018-07-17 Martin Liska <mliska@suse.cz>
1256 * opts.c: Do not enable OPT_falign_* for -Os.
1258 2018-07-17 Martin Liska <mliska@suse.cz>
1260 * align.h (MAX_CODE_ALIGN): New.
1261 (MAX_CODE_ALIGN_VALUE): New.
1262 * common/config/i386/i386-common.c (ix86_handle_option):
1263 (MAX_CODE_ALIGN): Moved to align.h.
1264 * final.c (MAX_CODE_ALIGN): Likewise.
1265 * opts.c (parse_and_check_align_values):
1266 (MAX_CODE_ALIGN): Likewise.
1267 (MAX_CODE_ALIGN_VALUE): Likewise.
1269 2018-07-17 Martin Liska <mliska@suse.cz>
1271 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
1272 in order to fulfil coding style.
1273 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
1274 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1275 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
1276 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
1277 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
1278 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
1279 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
1280 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
1281 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1283 2018-07-17 Martin Liska <mliska@suse.cz>
1285 * align.h: New file.
1286 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions directly.
1287 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
1288 align_flags of label_to_alignment.
1289 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into align_flags
1291 * config/m68k/m68k.c: Do not use removed align_labels_value and
1293 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
1294 (LOOP_ALIGN): Likewise.
1295 (LABEL_ALIGN): Likewise.
1296 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
1297 Remove not used macro.
1298 (rs6000_loop_align): Change return type to align_flags.
1299 (rs6000_loop_align_max_skip): Remove.
1300 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
1301 Change return type to align_flags.
1302 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
1303 Remove not used macro.
1304 (rs6000_loop_align): Change return type to align_flags.
1305 (rs6000_loop_align_max_skip): Remove.
1306 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
1307 * config/rx/rx-protos.h (rx_align_for_label): Make it
1309 * config/rx/rx.c (rx_align_for_label): Change return type
1311 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
1313 into align_flags class.
1314 (LABEL_ALIGN): Likewise.
1315 (LOOP_ALIGN): Likewise.
1316 * config/s390/s390.c (s390_label_align): Use align_flags
1318 (s390_asm_output_function_label): Likewise.
1319 * config/sh/sh.c (sh_override_options_after_change):
1320 Use align_flags class directly without macros.
1321 (find_barrier): Likewise.
1322 (barrier_align): Likewise.
1323 (sh_loop_align): Likewise.
1324 * config/spu/spu.c (spu_option_override):
1325 Use align_flags_tuple::get_value instead of removed macros.
1326 (spu_sched_init): Likewise.
1327 * config/spu/spu.h (GTY): Likewise.
1328 * config/visium/visium.c (visium_option_override):
1329 Set "8" as default secondary alignment.
1330 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
1331 in order to guarantee secondary alignment of 8.
1332 * coretypes.h: Include align.h header file.
1333 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
1334 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
1335 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
1336 * doc/tm.texi.in: Likewise.
1337 * final.c (struct label_alignment): Remove not used structure.
1338 (LABEL_ALIGN): Change type to align_flags.
1339 (LOOP_ALIGN): Likewise.
1340 (JUMP_ALIGN): Likewise.
1341 (default_loop_align_max_skip): Remove.
1342 (default_label_align_max_skip): Likewise.
1343 (default_jump_align_max_skip): Likewise.
1344 (default_label_align_after_barrier_max_skip):
1345 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
1346 (LABEL_TO_MAX_SKIP): Remove.
1347 (label_to_alignment): Return align_flags type instead of integer.
1348 (label_to_max_skip): Remove.
1349 (align_fuzz): Use align_flags type.
1350 (compute_alignments): Use align_flags type and use align_flags::max
1351 to combine multiple alignments.
1352 (grow_label_align): Grow vec instead of C array.
1353 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
1354 (shorten_branches): Use align_flags type and use align_flags::max
1355 to combine multiple alignments.
1356 (final_scan_insn_1): Remove usage of secondary alignment that comes
1357 from label alignment, but instead use proper secondary alignment
1358 which is computed in grow_label_align.
1359 * flags.h (struct align_flags_tuple): Move to align.h.
1360 (struct align_flags): Likewise.
1361 (state_align_loops): Rename to align_loops.
1362 (state_align_jumps): Rename to align_jumps.
1363 (state_align_labels): Rename to align_labels.
1364 (state_align_functions): Rename to align_functions.
1365 (align_loops_log): Remove.
1366 (align_jumps_log): Remove.
1367 (align_labels_log): Remove.
1368 (align_functions_log): Remove.
1369 (align_loops_max_skip): Remove.
1370 (align_jumps_max_skip): Remove.
1371 (align_labels_max_skip): Remove.
1372 (align_functions_max_skip): Remove.
1373 (align_loops_value): Remove.
1374 (align_jumps_value): Remove.
1375 (align_labels_value): Remove.
1376 (align_functions_value): Remove.
1377 * output.h (label_to_alignment): Change return type to align_flags.
1378 (label_to_max_skip): Remove.
1379 * target.def: Remove loop_align_max_skip, label_align_max_skip,
1380 jump_align_max_skip macros.
1381 * targhooks.h (default_loop_align_max_skip): Remove.
1382 (default_label_align_max_skip): Likewise.
1383 (default_jump_align_max_skip): Likewise.
1384 (default_label_align_after_barrier_max_skip): Remove.
1385 * toplev.c (read_log_maxskip): Use ::normalize function.
1386 (parse_N_M): Remove not used argument and also call ::normalize.
1387 (parse_alignment_opts): Do not pass unused arguments.
1388 * varasm.c (assemble_start_function): Use directly align_functions
1389 instead of removed macros.
1390 * system.h: Do not poison removed macros.
1392 2018-07-17 Jakub Jelinek <jakub@redhat.com>
1395 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
1396 and cond temporaries don't have reference type if iterator has
1397 pointer type. For init use &for_pre_body instead of pre_p if
1398 for_pre_body is non-empty.
1400 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
1402 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
1403 double-double modes to SFmode directly directly.
1404 (trunc<mode>sf2_fprs): Delete.
1406 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
1408 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
1409 for conversions between IFmode and the decimal floating point modes.
1410 (init_float128_ieee): Use the correct names for conversions between
1411 KFmode and the decimal floating point modes.
1413 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
1415 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
1416 for the conversions between TDmode and IFmode.
1417 (init_float128_ieee): Use more correct names for the conversions
1418 between TDmode and KFmode.
1420 2018-07-16 Jakub Jelinek <jakub@redhat.com>
1422 PR tree-optimization/86526
1423 * builtins.c (expand_builtin_memcmp): Formatting fixes.
1424 (inline_expand_builtin_string_cmp): Likewise.
1425 (inline_string_cmp): Likewise. Use c_readstr instead of
1426 builtin_memcpy_read_str. Add unit_mode temporary.
1428 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
1431 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
1432 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
1434 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
1436 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
1437 Alphabetize prototypes of built-in functions, separating out
1438 built-in functions that are listed in this section but should be
1439 described elsewhere.
1441 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
1444 * expmed.c (emit_store_flag): Do not emit setcc followed by a
1445 conditional move when trapping comparison was split to a
1446 non-trapping one (and vice versa).
1448 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
1450 * config/s390/s390.c (s390_function_profiler): Generate nops
1451 instead of profiler call sequences.
1452 * config/s390/s390.opt: Add the new option.
1454 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
1456 * config/s390/s390.c (s390_function_profiler): Generate
1457 __mcount_loc section.
1458 * config/s390/s390.opt: Add the new option.
1460 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
1462 * common.opt: Add the new warning.
1463 * config/s390/s390.c (s390_function_profiler): Emit "brasl
1464 %r0,__fentry__" when -mfentry is specified.
1465 (s390_option_override_internal): Disallow -mfentry for 31-bit
1467 * config/s390/s390.opt: Add the new option.
1469 2018-07-16 Richard Biener <rguenther@suse.de>
1472 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
1473 for function-local FUNCTION_DECL and RESULT_DECL immediately.
1475 2018-07-16 Martin Liska <mliska@suse.cz>
1478 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
1481 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
1483 * config/arc/arcHS.md: Update ARCHS scheduling rules.
1485 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
1487 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
1489 * config/arc/arc-cpus.def (hs4x): New cpu.
1491 * config/arc/arc-tables.opt: Regenerate.
1492 * config/arc/arc.c (arc_sched_issue_rate): New function.
1493 (TARGET_SCHED_ISSUE_RATE): Define.
1494 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
1495 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
1497 (attr tune): Add ARCHS4x tune values.
1498 (attr tune_dspmpy): Define.
1499 (*tst): Correct instruction type.
1500 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
1501 * config/arc/arcHS4x.md: New file.
1502 * config/arc/fpu.md: Update instruction type attributes.
1503 * config/arc/t-multilib: Regenerate.
1505 2018-07-16 Tom de Vries <tdevries@suse.de>
1508 * var-tracking.c (vt_initialize): Fix pre_dec handling.
1510 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1512 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
1515 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
1517 PR tree-optimization/86514
1518 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
1519 conversion to a boolean type from a type with greater precision.
1521 2018-07-16 Tom de Vries <tdevries@suse.de>
1523 * var-tracking.c (vt_initialize): Print adjusted insn slim if
1524 dump_flags request TDF_SLIM.
1526 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
1528 * fold-const.c (int_const_binop_1): Abstract...
1529 (wide_int_binop): ...wide int code here.
1530 (poly_int_binop): ...poly int code here.
1531 Abstract the rest of int_const_binop_1 into int_const_binop.
1532 * fold-const.h (wide_int_binop): New.
1533 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
1534 Remove useless PLUS/MINUS_EXPR case.
1535 (zero_nonzero_bits_from_vr): Move wide int code...
1536 (zero_nonzero_bits_from_bounds): ...here.
1537 (extract_range_from_binary_expr_1): Move mask optimization code...
1538 (range_easy_mask_min_max): ...here.
1539 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
1540 (range_easy_mask_min_max): New.
1542 2018-07-15 Jeff Law <law@redhat.com>
1545 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
1548 2018-07-14 Jim Wilson <jimw@sifive.com>
1550 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
1552 2018-07-14 Paul Koning <ni1d@arrl.net>
1554 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
1556 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
1558 * lto-streamer-out.c (copy_function_or_variable): Dump info about
1561 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
1562 Steve Munroe <munroesj52@gmail.com>
1564 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
1565 (_mm_andnot_si128): Likewise.
1566 (_mm_or_si128): Likewise.
1567 (_mm_xor_si128): Likewise.
1569 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
1572 * builtins.c (expand_builtin_memcmp): Inline the calls first
1573 when result_eq is false.
1574 (expand_builtin_strcmp): Inline the calls first.
1575 (expand_builtin_strncmp): Likewise.
1576 (inline_string_cmp): New routine. Expand a string compare
1577 call by using a sequence of char comparison.
1578 (inline_expand_builtin_string_cmp): New routine. Inline expansion
1579 a call to str(n)cmp/memcmp.
1580 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
1582 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
1584 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
1586 * config/arm/driver-arm.c: Include arm-native.h.
1587 (host_detect_local_cpu): Use auto-generated data tables.
1588 (vendors, arm_cpu_table): Delete. Move part information to ...
1589 * config/arm/arm-cpus.in: ... here.
1590 * config/arm/parsecpu.awk (gen_native): New function.
1591 (vendor, part): New CPU fields.
1592 (END): Add support for building the native CPU detection tables.
1593 * config/arm/t-arm (arm-native.h): Add build rule.
1594 (driver-arm.o): Add dependency on arm-native.h.
1596 2018-07-13 Richard Biener <rguenther@suse.de>
1599 * match.pd (addr1 - addr2): Allow either of the operand to
1602 2018-07-13 Tom de Vries <tdevries@suse.de>
1604 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
1607 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
1609 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
1610 arrays instead of numbered variables.
1612 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
1614 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
1615 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
1617 2018-07-13 Richard Biener <rguenther@suse.de>
1620 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
1621 instead of get_context_die.
1623 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
1624 Richard Biener <rguenther@suse.de>
1627 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
1628 that the loop latch destination where phi is defined.
1630 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
1632 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
1633 (riscv_expand_epilogue): Add assertion to check interrupt mode.
1634 (riscv_set_current_function): Extract getting interrupt type to new
1636 (riscv_get_interrupt_type): New function.
1637 (riscv_merge_decl_attributes): New function, checking interrupt type is
1639 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
1641 2018-07-12 Paul Koning <ni1d@arrl.net>
1643 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
1646 2018-07-12 Paul Koning <ni1d@arrl.net>
1648 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
1649 zero reference, add doloop_end instead.
1650 * doc/md.texi (decrement_and_branch_until_zero): Remove.
1651 (Looping patterns): Remove decrement_and_branch_until_zero. Add
1652 detail for doloop_end.
1654 2018-07-12 Martin Sebor <msebor@redhat.com>
1657 * attribs.c (decl_attributes): Reject conflicting attributes before
1658 calling attribute handlers.
1660 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
1662 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
1664 (gcc::dump_manager::get_dump_file_name): likewise.
1665 (dump_begin): Likewise.
1666 * dumpfile.h (dump_begin): Update prototype.
1667 (gcc::dump_manager::get_dump_file_name,
1668 gcc::dump_manager::get_dump_file_name): Update prototype.
1670 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
1672 * internal-fn.h (vectorizable_internal_fn_p): New function.
1673 * tree-vect-slp.c (compatible_calls_p): Likewise.
1674 (vect_build_slp_tree_1): Remove nops argument. Handle calls
1675 to internal functions.
1676 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
1678 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
1680 * fold-const.h (inverse_conditions_p): Declare.
1681 * fold-const.c (inverse_conditions_p): New function.
1682 * match.pd: Use inverse_conditions_p. Add folds of view_converts
1683 that test the inverse condition of a conditional internal function.
1684 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
1685 * internal-fn.c (internal_fn_mask_index): Handle conditional
1687 (vectorized_internal_fn_supported_p): New function.
1688 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
1689 (any_pred_load_store): Replace with...
1690 (need_to_predicate): ...this new variable.
1691 (redundant_ssa_names): New variable.
1692 (ifcvt_can_use_mask_load_store): Move initial checks to...
1693 (ifcvt_can_predicate): ...this new function. Handle tree codes
1694 for which a conditional internal function exists.
1695 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
1696 instead of ifcvt_can_use_mask_load_store. Update after variable
1698 (predicate_load_or_store): New function, split out from
1699 predicate_mem_writes.
1700 (check_redundant_cond_expr): New function.
1701 (value_available_p): Likewise.
1702 (predicate_rhs_code): Likewise.
1703 (predicate_mem_writes): Rename to...
1704 (predicate_statements): ...this. Use predicate_load_or_store
1705 and predicate_rhs_code.
1706 (combine_blocks, tree_if_conversion): Update after above name changes.
1707 (ifcvt_local_dce): Handle redundant_ssa_names.
1708 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
1709 general conditional functions.
1710 * tree-vect-stmts.c (vectorizable_call): Likewise.
1712 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
1713 Alan Hayward <alan.hayward@arm.com>
1714 David Sherwood <david.sherwood@arm.com>
1716 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
1717 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
1718 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
1719 plus and minus and convert them into IFN_COND_FMA-based sequences.
1720 (convert_mult_to_fma): Handle conditional plus and minus.
1722 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
1724 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
1725 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
1726 (cond_fnms_optab): New optabs.
1727 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
1729 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
1730 * internal-fn.h (get_conditional_internal_fn): Declare.
1731 (get_unconditional_internal_fn): Likewise.
1732 * internal-fn.c (cond_ternary_direct): New macro.
1733 (expand_cond_ternary_optab_fn): Likewise.
1734 (direct_cond_ternary_optab_supported_p): Likewise.
1735 (FOR_EACH_COND_FN_PAIR): Likewise.
1736 (get_conditional_internal_fn): New function.
1737 (get_unconditional_internal_fn): Likewise.
1738 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
1739 (gimple_match_op::gimple_match_op): Add a new overload for 5
1741 (gimple_match_op::set_op): Likewise.
1742 (gimple_resimplify5): Declare.
1743 * genmatch.c (decision_tree::gen): Generate simplifications for
1745 * gimple-match-head.c (gimple_simplify): Define an overload for
1746 5 operands. Handle calls with 5 arguments in the top-level overload.
1747 (convert_conditional_op): Handle conversions from unconditional
1748 internal functions to conditional ones.
1749 (gimple_resimplify5): New function.
1750 (build_call_internal): Pass a fifth operand.
1751 (maybe_push_res_to_seq): Likewise.
1752 (try_conditional_simplification): Try converting conditional
1753 internal functions to unconditional internal functions.
1754 Handle 3-operand unconditional forms.
1755 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
1756 Define ternary equivalents of the current rules for binary conditional
1758 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
1760 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
1761 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
1762 (optab): Handle them.
1763 (SVE_COND_FP_TERNARY): New int iterator.
1764 (sve_fmla_op, sve_fmad_op): New int attributes.
1765 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
1766 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
1767 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
1769 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
1771 * target.def (preferred_else_value): New target hook.
1772 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
1773 * doc/tm.texi: Regenerate.
1774 * targhooks.h (default_preferred_else_value): Declare.
1775 * targhooks.c (default_preferred_else_value): New function.
1776 * internal-fn.h (conditional_internal_fn_code): Declare.
1777 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
1778 (get_conditional_internal_fn): Use it.
1779 (conditional_internal_fn_code): New function.
1780 * gimple-match.h (gimple_match_cond): New struct.
1781 (gimple_match_op): Add a cond member function.
1782 (gimple_match_op::gimple_match_op): Update all forms to take a
1784 * genmatch.c (expr::gen_transform): Use the same condition as res_op
1785 for the suboperation, but don't specify a particular else_value.
1786 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
1787 (visit_nary_op, visit_reference_op_load): Pass
1788 gimple_match_cond::UNCOND to the gimple_match_op constructor.
1789 * gimple-match-head.c: Include tree-eh.h
1790 (convert_conditional_op): New function.
1791 (maybe_resimplify_conditional_op): Likewise.
1792 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
1793 (gimple_resimplify2): Likewise.
1794 (gimple_resimplify3): Likewise.
1795 (gimple_resimplify4): Likewise.
1796 (maybe_push_res_to_seq): Return null for conditional operations.
1797 (try_conditional_simplification): New function.
1798 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
1800 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
1802 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
1804 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
1806 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
1808 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
1810 (hash_tree): Do not hash DECL_FCONTEXT
1811 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
1812 Do not stream DECL_FCONTEXT.
1813 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
1814 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
1816 2018-07-12 Richard Biener <rguenther@suse.de>
1819 * dwarf2out.c (gen_block_die): Only output blocks when they have
1820 at least one !DECL_IGNORED_P variable.
1822 2018-07-12 Richard Biener <rguenther@suse.de>
1825 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
1826 Remove -mieee-fp handling.
1828 2018-07-12 Richard Biener <rguenther@suse.de>
1830 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
1831 left-over from last patch.
1833 2018-07-12 Jakub Jelinek <jakub@redhat.com>
1835 PR tree-optimization/86492
1836 * gimple-ssa-store-merging.c
1837 (imm_store_chain_info::coalesce_immediate_stores): Call
1838 check_no_overlap even for the merge_overlapping case. Formatting fix.
1840 2018-07-12 Richard Biener <rguenther@suse.de>
1843 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
1844 move possibly trapping operations into the conditional.
1846 2018-07-12 Richard Biener <rguenther@suse.de>
1848 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
1849 (vn_lookup_simplify_result): Remove recursion limit applied
1851 (vn_nary_build_or_lookup_1): Adjust.
1852 (try_to_simplify): Likewise.
1853 * gimple-match-head.c (gimple_resimplify1): Instead apply one
1855 (gimple_resimplify2): Likewise.
1856 (gimple_resimplify3): Likewise.
1857 (gimple_resimplify4): Likewise.
1859 2018-07-11 Jakub Jelinek <jakub@redhat.com>
1861 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
1862 Use __mmask64 type instead of __mmask8 for __M argument.
1863 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
1864 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
1866 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
1867 __mmask16 for __M argument.
1868 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
1869 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
1870 to __mmask16 instead of __mmask8.
1871 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
1872 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
1873 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
1874 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
1875 instead of __mmask16 for __U argument.
1876 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
1877 __mmask16 instead of __mmask8 for __U argument.
1878 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
1880 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
1882 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
1884 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
1886 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
1888 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
1890 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
1892 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
1894 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
1896 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
1898 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
1899 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
1900 return type as well as __M argument type and all casts from __mmask8
1902 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
1903 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
1904 return type as well as __M argument type and all casts from __mmask8
1906 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
1907 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
1908 return type as well as __M argument type and all casts from __mmask8
1910 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
1911 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
1912 return type as well as __M argument type and all casts from __mmask8
1914 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
1915 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
1918 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
1920 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
1921 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
1924 2018-07-11 Paul Koning <ni1d@arrl.net>
1926 * doc/md.texi (define_subst): Document how multiple occurrences of
1927 the same argument in the replacement pattern are handled.
1929 2018-07-11 Paul Koning <ni1d@arrl.net>
1931 * doc/extend.texi (Common Variable Attributes): Move "mode" into
1933 (Common Type Attributes): Add "mode" attribute.
1935 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
1937 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
1938 stream DECL_ORIGINAL_TYPE.
1939 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
1940 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
1941 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
1942 Do not walk original type.
1943 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
1945 (write_ts_decl_non_common_tree_pointers): Do not stream
1947 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
1948 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
1950 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
1952 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
1953 thread twice from the same starting edge.
1955 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
1957 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
1958 * gimple.c (gimple_call_nonnull_result_p): ...here...
1959 (gimple_call_nonnull_arg): ...and here.
1960 * gimple.h (gimple_call_nonnull_result_p): New.
1961 (gimple_call_nonnull_arg): New.
1963 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
1965 * config/arm/arm-cpus.in: Move information from fpu field of each
1966 cpu definition to the isa field.
1967 * config/arm/parsecpu.awk (fpu): Delete match rule.
1968 (gen_comm_data): Don't add bits from the CPU's FPU entry.
1970 2018-07-11 Richard Biener <rguenther@suse.de>
1973 * dwarf2out.c (init_sections_and_labels): Use
1974 output_asm_line_debug_info consistently.
1975 (dwarf2out_early_finish): Likewise.
1976 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
1979 2018-07-11 Richard Biener <rguenther@suse.de>
1981 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
1982 Rework father_bb setting in a way to avoid propagating constants
1983 multiple times on a loop body.
1985 2018-07-10 Mark Wielaard <mark@klomp.org>
1988 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
1990 2018-07-10 Richard Biener <rguenther@suse.de>
1992 * hash-map.h (hash_map::iterator::operator*): Return
1993 references to key and value.
1995 2018-07-10 Jakub Jelinek <jakub@redhat.com>
1998 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
1999 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
2000 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
2001 (gimplify_omp_for): For composite loops, move outer
2002 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
2003 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
2004 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
2005 TREE_LIST for both the original class iterator and the "last" helper
2006 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
2007 loop, remember has_decl_expr from outer composite loops for the
2008 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
2010 2018-07-09 Martin Sebor <msebor@redhat.com>
2014 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
2015 accessing implicitly initialized array elements.
2016 * expr.c (string_constant): Handle string initializers of
2017 character arrays within aggregates.
2018 * gimple-fold.c (fold_array_ctor_reference): Add argument.
2019 Store element offset. As a special case, handle zero size.
2020 (fold_nonarray_ctor_reference): Same.
2021 (fold_ctor_reference): Add argument. Store subobject offset.
2022 * gimple-fold.h (fold_ctor_reference): Add argument.
2024 2018-07-09 Paul Koning <ni1d@arrl.net>
2026 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
2027 (pdp11_insn_cost): New function.
2028 (pdp11_md_asm_adjust): New function.
2029 (TARGET_INVALID_WITHIN_DOLOOP): Define.
2030 (pdp11_rtx_costs): Update to match machine better.
2031 (output_addr_const_pdp11): Correct format mismatch warnings.
2032 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
2033 * config/pdp11/pdp11.md: General change to add base_cost and/or
2034 length attributes for use by new pdp11_insn_cost function.
2035 (MIN_BRANCH): Correct definition.
2037 (doloop_end): Use standard pattern name for looping pattern.
2038 (doloop_end_nocc): New.
2039 (movsf): Add another constraint alternative.
2040 (zero_extendqihi2): Add constraint alternatives for not in place
2042 (zero_extendhisi2): Remove.
2043 (shift patterns): Add CC handling variants.
2047 (define_peephole2): New peephole to recognize mov that sets CC for
2050 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2052 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
2053 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
2056 2018-07-09 Jakub Jelinek <jakub@redhat.com>
2059 * real.c (real_nextafter): Return true if result is denormal.
2061 2018-07-09 Martin Liska <mliska@suse.cz>
2063 * common.opt: Add back wrongly removed attribute.
2065 2018-07-09 Richard Biener <rguenther@suse.de>
2068 * dwarf2out.c (gen_block_die): For an early generated DIE
2069 always output high/low PC attributes.
2071 2018-07-09 Tom de Vries <tdevries@suse.de>
2073 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
2074 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
2075 onto VAR_DECL with abstract origin.
2077 2018-07-07 Jim Wilson <jimw@sifive.com>
2079 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
2081 2018-07-07 Tom de Vries <tdevries@suse.de>
2083 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
2085 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
2087 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
2089 (combine_bound): Use wide-int overflow calculation instead of
2091 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
2093 * fold-const.c (int_const_binop_2): Same.
2094 (extract_muldiv_1): Same.
2095 (fold_div_compare): Same.
2096 (fold_abs_const): Same.
2098 * poly-int.h (add): Same.
2102 * predict.c (predict_iv_comparison): Same.
2103 * profile-count.c (slow_safe_scale_64bit): Same.
2104 * simplify-rtx.c (simplify_const_binary_operation): Same.
2105 * tree-chrec.c (tree_fold_binomial): Same.
2106 * tree-data-ref.c (split_constant_offset_1): Same.
2107 * tree-if-conv.c (idx_within_array_bound): Same.
2108 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
2109 * tree-ssa-phiopt.c (minmax_replacement): Same.
2110 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
2111 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
2112 * vr-values.c (vr_values::adjust_range_with_scev): Same.
2113 * wide-int.cc (wi::add_large): Same.
2114 (wi::mul_internal): Same.
2115 (wi::sub_large): Same.
2116 (wi::divmod_internal): Same.
2117 * wide-int.h: Change overflow type to overflow_type for neg, add,
2118 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
2119 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
2120 mul_internal, divmod_internal.
2121 (overflow_type): New enum.
2122 (accumulate_overflow): New.
2124 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
2126 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
2127 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
2129 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
2131 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
2132 argument is checked for zero before entering loop, avoid checking again.
2134 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
2136 * gimplify.h (generic_expr_could_trap_p): Set as global function.
2137 * gimplify.h (generic_expr_could_trap_p): Likwise.
2138 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
2140 2018-07-06 Jakub Jelinek <jakub@redhat.com>
2142 PR tree-optimization/86401
2143 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
2144 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
2145 (fold_bit_and_mask): ... here. New helper function for match.pd.
2146 * fold-const.h (fold_bit_and_mask): Declare.
2147 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
2149 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
2152 * target.def (translate_mode_attribute): New hook.
2153 * targhooks.h (default_translate_mode_attribute): Declare.
2154 * targhooks.c (default_translate_mode_attribute): New function.
2155 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
2156 * doc/tm.texi: Regenerate.
2157 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
2158 (rs6000_translate_mode_attribute): New function.
2160 2018-07-06 Paul Koning <ni1d@arrl.net>
2162 * doc/md.texi (define_split): Document DONE and FAIL.
2163 (define_peephole2): Ditto.
2165 2018-07-05 Jeff Law <law@redhat.com>
2167 PR tree-optimization/86010
2168 * tree-ssa-dse.c (compute_trims): More aggressively trim at
2169 both the head and tail of mem* and str* calls.
2171 2018-07-05 Jim Wilson <jimw@sifive.com>
2173 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
2175 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
2177 * config/aarch64/aarch64-simd.md: correct flags text for
2178 MIN_EXPR replacement.
2180 2018-07-05 James Clarke <jrtc27@jrtc27.com>
2182 * configure: Regenerated.
2184 2018-07-05 Carl Love <cel@us.ibm.com>
2186 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
2187 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
2188 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
2189 VSX_BUILTIN_DOUBLEL_V4SF.
2191 2018-07-05 Martin Sebor <msebor@redhat.com>
2194 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
2195 than its domain to compute its the upper bound of a char array.
2197 2018-07-05 Nathan Sidwell <nathan@acm.org>
2199 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
2200 * doc/cpp.texi: Update comment.
2201 * doc/tm.texi: Rebuilt.
2202 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
2203 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
2204 * doc/extend.texi (Backwards Compatibility): Clarify it is system
2205 headers affected by extern "C".
2206 * system.h: Poison NO_IMPLICIT_EXTERN_C.
2207 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
2208 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
2209 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
2210 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
2211 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
2212 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
2213 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
2214 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
2215 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
2216 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
2217 config/sparc/sp64-elf.h, config/spu/spu.h,
2218 config/stormy16/stormy16.h, config/v850/v850.h,
2219 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
2220 define NO_IMPLICIT_EXTERN_C.
2221 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
2223 2018-07-05 Tamar Christina <tamar.christina@arm.com>
2226 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
2227 instead of GET_MODE_SIZE when comparing Units.
2229 2018-07-05 Tamar Christina <tamar.christina@arm.com>
2232 * rtlanal.c (set_noop_p): Constrain on mode change,
2233 include hard-reg-set.h
2235 2018-07-05 Tamar Christina <tamar.christina@arm.com>
2237 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
2239 2018-07-05 Jakub Jelinek <jakub@redhat.com>
2242 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
2245 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
2247 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
2249 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
2252 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
2254 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
2256 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
2259 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
2261 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
2263 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
2264 neg_*_op* variables.
2266 2018-07-04 Martin Liska <mliska@suse.cz>
2268 * tree-switch-conversion.c: Define
2269 max_ratio_for_speed and max_ratio_for_size constants.
2271 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
2272 Martin Liska <mliska@suse.cz>
2277 * common.opt: Rename align options with 'str_' prefix.
2278 * common/config/i386/i386-common.c (set_malign_value): New
2280 (ix86_handle_option): Use it to set -falign-* options/
2281 * config/aarch64/aarch64-protos.h (struct tune_params): Change
2282 type from int to string.
2283 * config/aarch64/aarch64.c: Update default values from int
2285 * config/alpha/alpha.c (alpha_override_options_after_change):
2287 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
2288 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
2289 max skip conditionally.
2290 * config/i386/freebsd.h (SUBALIGN_LOG): New.
2291 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
2292 max skip conditionally.
2293 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
2294 max skip conditionally.
2295 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
2296 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
2297 max skip conditionally.
2298 * config/i386/i386.c (struct ptt): Change type from int to
2300 (ix86_default_align): Set default values.
2301 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
2302 max skip conditionally.
2303 * config/i386/iamcu.h (SUBALIGN_LOG): New.
2304 (ASM_OUTPUT_MAX_SKIP_ALIGN):
2305 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
2306 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
2307 max skip conditionally.
2308 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
2309 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
2310 * config/i386/x86-64.h (SUBALIGN_LOG): New.
2311 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
2312 max skip conditionally.
2313 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
2314 * config/ia64/ia64.c (ia64_option_override): Set default values
2315 for alignment options.
2316 * config/m68k/m68k.c: Handle new str_align_* options.
2317 * config/mips/mips.c (mips_set_compression_mode): Change
2319 (mips_option_override): Set default values for options.
2320 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
2322 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2324 * config/rx/rx.c (rx_option_override): Likewise.
2325 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
2326 (LABEL_ALIGN): Use align_labels_log.
2327 (LOOP_ALIGN): Use align_loops_align.
2328 * config/s390/s390.c (s390_asm_output_function_label): Use new
2330 * config/sh/sh.c (sh_override_options_after_change):
2331 Change type of constants.
2332 * config/spu/spu.c (spu_sched_init): Likewise.
2333 * config/sparc/sparc.c (sparc_option_override): Set default
2335 * config/visium/visium.c (visium_option_override): Likewise.
2336 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
2337 emit p2align format with last argument if it's not needed.
2338 * doc/invoke.texi: Document extended format of -falign-*.
2339 * final.c: Use align_labels alignment.
2340 * flags.h (struct target_flag_state): Change type to use
2342 (struct align_flags_tuple): New.
2343 (struct align_flags): Likewise.
2344 (align_loops_log): Redefine macro to use new types.
2345 (align_loops_max_skip): Redefine macro to use new types.
2346 (align_jumps_log): Redefine macro to use new types.
2347 (align_jumps_max_skip): Redefine macro to use new types.
2348 (align_labels_log): Redefine macro to use new types.
2349 (align_labels_max_skip): Redefine macro to use new types.
2350 (align_functions_log): Redefine macro to use new types.
2351 (align_loops): Redefine macro to use new types.
2352 (align_jumps): Redefine macro to use new types.
2353 (align_labels): Redefine macro to use new types.
2354 (align_functions): Redefine macro to use new types.
2355 (align_functions_max_skip): Redefine macro to use new types.
2356 (align_loops_value): New macro.
2357 (align_jumps_value): New macro.
2358 (align_labels_value): New macro.
2359 (align_functions_value): New macro.
2360 * function.c (invoke_set_current_function_hook): Propagate
2361 alignment values from flags to global variables default in
2363 * ipa-icf.c (sem_function::equals_wpa): Use
2364 cl_optimization_option_eq instead of memcmp.
2365 * lto-streamer.h (cl_optimization_stream_out): Support streaming
2367 (cl_optimization_stream_in): Likewise.
2368 * optc-save-gen.awk: Support strings in cl_optimization.
2369 * opth-gen.awk: Likewise.
2370 * opts.c (finish_options): Remove error checking of invalid
2372 (MAX_CODE_ALIGN): Remove.
2373 (MAX_CODE_ALIGN_VALUE): Likewise.
2374 (parse_and_check_align_values): New function.
2375 (check_alignment_argument): Likewise.
2376 (common_handle_option): Use check_alignment_argument.
2377 * opts.h (parse_and_check_align_values): Declare.
2378 * toplev.c (init_alignments): Remove.
2379 (read_log_maxskip): New.
2380 (parse_N_M): Likewise.
2381 (parse_alignment_opts): Likewise.
2382 (backend_init_target): Remove usage of init_alignments.
2383 * toplev.h (parse_alignment_opts): Declare.
2384 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
2386 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
2387 * tree.c (cl_option_hasher::equal): New.
2388 * varasm.c: Use new global macros.
2390 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
2392 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
2393 Use a simpler align directive also if MAXSKIP = ALIGN-1.
2394 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2395 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2396 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2397 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
2398 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
2399 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
2400 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
2401 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2402 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2403 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2404 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2406 2018-07-04 Martin Liska <mliska@suse.cz>
2407 Jonathan Wakely <jwakely@redhat.com>
2409 * coverage.c: Use correct type.
2410 * doc/invoke.texi: Language correction.
2412 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
2415 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
2416 ENDBRANCH for non-tail call which may return via indirect branch.
2417 * doc/extend.texi: Document indirect_return attribute.
2419 2018-07-03 Martin Sebor <msebor@redhat.com>
2421 PR tree-optimization/86274
2422 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
2424 (format_floating): Correct handling of infinities and NaNs.
2426 2018-07-03 Martin Sebor <msebor@redhat.com>
2428 * print-tree.c (print_real_cst): New function.
2429 (print_node_brief): Call it.
2430 (print_node): Ditto.
2432 2018-07-03 Jeff Law <law@redhat.com>
2434 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
2435 into a single pattern.
2437 * config/h8300/h8300.md (ors code_iterator): New.
2438 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
2439 a single pattern and single splitter.
2440 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
2441 (iorqi3_1, xorqi3_1): Likewise.
2442 (iorqi3, xorqi3 expanders): Similarly.
2444 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
2445 (movmd_internal) into a single pattern using the P mode iterator.
2446 (movmd splitters): Similarly.
2447 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
2448 (movsd splitters): Similarly.
2450 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
2451 ADDB, ADDW and ADDL into a single ADD attribute which selects the
2452 right table based on the size of the operand.
2453 * config/h8300/h8300.md (length_table): Corresponding changes. All
2454 references to "addb", "addw" and "addl" changed to "add".
2455 (btst patterns): Merge two variants into a single pattern.
2456 (tstqi, tsthi): Likewise.
2457 (addhi3_incdec, addsi3_incdec): Likewise.
2458 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
2459 (mulhi3, mulsi3): Likewise.
2460 (udivhi3, udivsi3): Likewise.
2461 (divhi3, divsi3): Likewise.
2462 (andorqi3, andorhi3, andorsi3): Likewise.
2464 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
2467 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
2468 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
2470 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2472 PR tree-optimization/85694
2473 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
2474 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
2476 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
2477 (<u>avg<mode>3_ceil): New patterns.
2479 2018-07-03 David Malcolm <dmalcolm@redhat.com>
2481 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
2482 scan-tree-dump directive.
2483 * gcc.dg/vect/slp-perm-2.c: Likewise.
2484 * gcc.dg/vect/slp-perm-3.c: Likewise.
2485 * gcc.dg/vect/slp-perm-5.c: Likewise.
2486 * gcc.dg/vect/slp-perm-6.c: Likewise.
2487 * gcc.dg/vect/slp-perm-7.c: Likewise.
2488 * gcc.dg/vect/slp-perm-8.c: Likewise.
2490 2018-07-03 Marek Polacek <polacek@redhat.com>
2493 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
2495 2018-07-03 Richard Biener <rguenther@suse.de>
2498 * tree-ssa-structalias.c (find_func_clobbers): Properly
2499 handle indirect calls.
2501 2018-07-03 Jeff Law <law@redhat.com>
2503 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
2504 (shifts): New code iterator.
2505 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
2506 expander. Fix HImode handling on H8/SX.
2507 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
2508 (subqi3, subhi3, subsi3 expanders): Likewise.
2509 (andqi3, andhi3, andsi3 expanders): Likewise.
2510 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
2511 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
2512 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
2513 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
2514 (zero_extendqihi2, zero_extendqisi2): Likewise.
2515 (extendqihi2, extendqisi2): Likewise.
2516 (rotlqi3, rotlhi3, rotlsi3): Likewise.
2517 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
2518 (rotlqi3_1, rotlhi3_1): Likewise.
2519 (logicalhi3_sn, logicalsi3_sn): Likewise.
2520 (logicalhi3, logicalsi3): Likewise.
2522 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2524 * tree-vect-patterns.c (vect_recog_rotate_pattern)
2525 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
2526 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
2527 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
2528 type to append_pattern_def_seq instead of creating a stmt_vec_info
2530 (build_mask_conversion): Likewise. Remove vinfo argument.
2531 (vect_add_conversion_to_patterm): Likewise, renaming to...
2532 (vect_add_conversion_to_pattern): ...this.
2533 (vect_recog_mask_conversion_pattern): Update call to
2534 build_mask_conversion. Pass the vector type to
2535 append_pattern_def_seq here too.
2536 (vect_recog_gather_scatter_pattern): Update call to
2537 vect_add_conversion_to_pattern.
2539 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2541 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
2542 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
2543 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
2544 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
2545 STMT_VINFO_PATTERN_DEF_SEQ to null here.
2546 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
2547 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
2548 append_pattern_def_seq instead of new_pattern_def_seq.
2549 (vect_recog_divmod_pattern): Do both of the above.
2550 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
2553 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2555 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
2556 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
2557 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
2558 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
2559 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
2560 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
2561 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
2562 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
2563 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
2564 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
2565 parameter with a single stmt_vec_info.
2566 (vect_recog_func_ptr): Likewise.
2567 (vect_recog_gather_scatter_pattern): Likewise, folding in...
2568 (vect_try_gather_scatter_pattern): ...this.
2569 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
2570 the stmt_vec_info of the statement to be matched. Don't clear
2571 STMT_VINFO_RELATED_STMT.
2572 (vect_pattern_recog): Update call accordingly.
2574 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2576 PR tree-optimization/85694
2577 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
2578 (uavgM3_ceil): Document new optabs.
2579 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
2580 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
2582 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
2583 (savg_ceil_optab): New optabs.
2584 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
2585 (vect_vect_recog_func_ptrs): Add it.
2586 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
2587 constant directly from the associated lhs.
2589 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2591 * tree-vect-patterns.c (vect_split_statement): New function.
2592 (vect_convert_input): Use it to try to split an existing cast.
2594 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2596 * poly-int.h (print_hex): New function.
2597 * dumpfile.h (dump_dec, dump_hex): Declare.
2598 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
2599 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
2600 min_input_precision, operation_precision and operation_sign.
2601 * tree-vect-patterns.c (vect_get_range_info): New function.
2602 (vect_same_loop_or_bb_p, vect_single_imm_use)
2603 (vect_operation_fits_smaller_type): Delete.
2604 (vect_look_through_possible_promotion): Add an optional
2605 single_use_p parameter.
2606 (vect_recog_over_widening_pattern): Rewrite to use new
2607 stmt_vec_info infomration. Handle one operation at a time.
2608 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
2609 (vect_truncatable_operation_p, vect_set_operation_type)
2610 (vect_set_min_input_precision): New functions.
2611 (vect_determine_min_output_precision_1): Likewise.
2612 (vect_determine_min_output_precision): Likewise.
2613 (vect_determine_precisions_from_range): Likewise.
2614 (vect_determine_precisions_from_users): Likewise.
2615 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
2616 (vect_vect_recog_func_ptrs): Put over_widening first.
2618 (vect_pattern_recog): Call vect_determine_precisions.
2620 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2622 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
2623 statements that have been replaced by further pattern statements.
2624 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
2626 2018-07-03 Richard Biener <rguenther@suse.de>
2628 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
2629 always set *dt. Dump vectype in vectype overload.
2630 * dumpfile.h (dump_gimple_expr): New function.
2631 (dump_gimple_expr_loc): Likewise.
2632 * dumpfile.c (dump_gimple_expr): New function.
2633 (dump_gimple_expr_loc): Likewise.
2635 2018-07-02 Jeff Law <law@redhat.com>
2637 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
2638 the H8/300, H8/300H and H8/S variants into a single pattern.
2639 (movhi_h8300, movqi_h8300hs): Similarly.
2640 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
2641 (QHI mode iterator): New.
2643 * config/h8300/h8300.md: Remove trailing whitespace.
2645 2018-07-02 Jim Wilson <jimw@sifive.com>
2647 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
2648 instead of emit_insn for interrupt returns.
2649 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
2650 (riscv_sret, riscv_uret): Likewise.
2652 2018-07-02 David Malcolm <dmalcolm@redhat.com>
2654 * pretty-print.c (selftest::test_pp_format): Move save and restore
2655 of quotes to class auto_fix_quotes, and add an instance.
2656 * selftest.c: Include "intl.h".
2657 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
2658 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
2659 * selftest.h (selftest::auto_fix_quotes): New class.
2661 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
2663 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
2664 (aarch64_sve_prepare_conditional_op): Remove.
2665 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
2666 Allow aarch64_simd_reg_or_zero as select operand; remove
2667 the aarch64_sve_prepare_conditional_op call.
2668 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
2669 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
2670 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
2671 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
2672 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
2673 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
2674 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
2675 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
2676 and a splitters to match all of the *_any patterns.
2677 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
2679 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
2680 (SVE_COND_FP_BINARY_REV): Remove.
2681 (sve_int_op_rev, sve_fp_op_rev): New.
2682 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
2683 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
2684 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
2685 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
2686 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
2687 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
2688 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
2689 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
2690 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
2692 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
2693 Remove match_dup 1 from the inner unspec.
2694 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
2696 * config/aarch64/aarch64.md (movprfx): New attr.
2697 (length): Default movprfx to 8.
2698 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
2699 (*madd<SVE_I>, *msub<SVE_I): Likewise.
2700 (*<su>mul<SVE_I>3_highpart): Likewise.
2701 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
2702 (*v<ASHIFT><SVE_I>3): Likewise.
2703 (*<su><MAXMIN><SVE_I>3): Likewise.
2704 (*<su><MAXMIN><SVE_F>3): Likewise.
2705 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
2706 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
2707 (*div<SVE_F>4): Likewise.
2709 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
2711 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
2714 2018-07-02 Richard Biener <rguenther@suse.de>
2716 PR tree-optimization/86363
2717 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
2718 memset argument refers to a non-variable address.
2720 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
2722 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
2723 {PLUS,MINUS}_EXPR code to...
2724 (adjust_symbolic_bound): ...here,
2725 (combine_bound): ...here,
2726 (set_value_range_with_overflow): ...and here.
2728 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
2730 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
2732 (extract_range_from_abs_expr): ...here.
2734 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
2736 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
2737 -fno-omit-frame-pointer when not optimizing.
2739 2018-07-02 Martin Liska <mliska@suse.cz>
2742 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
2743 (propagate_nothrow): Likewise.
2745 2018-07-02 Martin Liska <mliska@suse.cz>
2748 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
2750 2018-07-02 David Malcolm <dmalcolm@redhat.com>
2752 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
2753 function in r262149, changing "loc" param from source_location to
2754 const dump_location_t &.
2755 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
2756 declaration, as above.
2758 2018-07-01 Paul Koning <ni1d@arrl.net>
2760 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
2761 -munit-asm, -mgnu-asm, -mdec-asm.
2762 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
2763 (pdp11_output_labelref): New.
2764 (pdp11_output_def): New.
2765 (pdp11_output_addr_vec_elt): New.
2766 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
2767 %# and %@ format codes.
2768 (pdp11_option_override): New.
2769 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
2770 (pdp11_output_ident): New.
2771 (pdp11_asm_named_section): New.
2772 (pdp11_asm_init_sections): New.
2773 (pdp11_file_start): New.
2774 (pdp11_file_end): New.
2775 (output_ascii): Use .ascii/.asciz for -mdec-asm.
2776 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
2777 %o, like %c but octal.
2778 (pdp11_option_override): New.
2779 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
2781 (DATA_SECTION_ASM_OP): Ditto.
2782 (READONLY_DATA_SECTION_ASM_OP): New.
2783 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
2784 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
2785 (ASM_OUTPUT_LABELREF): Ditto.
2786 (ASM_OUTPUT_DEF): Ditto.
2787 (ASM_OUTPUT_EXTERNAL): New.
2788 (ASM_OUTPUT_SOURCE_FILENAME): New.
2789 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
2790 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
2791 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
2792 %# and %@ format codes.
2793 * config/pdp11/pdp11.opt (mgnu-asm): New.
2794 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
2795 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
2796 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
2798 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
2800 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
2801 dereferencing path[] beyond its length.
2803 (debug_all_paths): New.
2804 (rewire_first_differing_edge): New.
2805 (adjust_paths_after_duplication): New.
2806 (duplicate_thread_path): Call adjust_paths_after_duplication.
2808 (thread_through_all_blocks): Add new argument to
2809 duplicate_thread_path.
2811 2018-06-30 Jim Wilson <jimw@sifive.com>
2813 * config/riscv/predicates.md (p2m1_shift_operand): New.
2814 (high_mask_shift_operand): New.
2815 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
2816 pattern using p2m1_shift_operand.
2817 (lshsi3_zero_extend_3+2): New combiner pattern using
2818 high_mask_shift_operand.
2820 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2822 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
2824 (vect_recog_rotate_pattern): ...here.
2825 (vect_convert_input): Try to insert casts of invariants in the
2827 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
2828 preheader to be empty.
2830 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2832 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
2833 vector type. If given, install it in the new statement's
2835 (vect_element_precision): New function.
2836 (vect_unpromoted_value): New struct.
2837 (vect_unpromoted_value::vect_unpromoted_value): New function.
2838 (vect_unpromoted_value::set_op): Likewise.
2839 (vect_look_through_possible_promotion): Likewise.
2840 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
2841 (vect_widened_op_tree, vect_convert_input): Likewise.
2842 (vect_convert_inputs, vect_convert_output): Likewise.
2843 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
2844 to handle the optional cast of the multiplication result and
2845 vect_widened_op_tree to detect the widened multiplication itself.
2846 Do not require the input and output of promotion casts to have
2847 the same sign, but base the signedness of the operation on the
2848 input rather than the result. If the pattern includes two
2849 promotions, check that those promotions have the same sign.
2850 Do not restrict the MULT_EXPR handling to a double-width result;
2851 handle quadruple-width results and wider. Use vect_convert_inputs
2852 to convert the inputs to the common type.
2853 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
2854 to handle the optional cast of the ABS result. Also allow a sign
2855 change or a sign extension between the ABS and MINUS.
2856 Use vect_widened_op_tree to detect the widened subtraction and use
2857 vect_convert_inputs to convert the inputs to the common type.
2858 (vect_handle_widen_op_by_const): Delete.
2859 (vect_recog_widen_op_pattern): New function.
2860 (vect_recog_widen_mult_pattern): Use it.
2861 (vect_recog_widen_shift_pattern): Likewise.
2862 (vect_recog_widen_sum_pattern): Use
2863 vect_look_through_possible_promotion to handle the promoted
2866 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2868 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
2869 the containing gimple_seq *.
2870 * gimple-iterator.h (gsi_for_stmt): Declare it.
2871 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
2872 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
2873 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
2874 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
2875 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
2877 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
2879 (vect_mark_pattern_stmts): ...here. Handle cases in which the
2880 statement being replaced is part of an existing pattern
2881 definition sequence, inserting the new pattern statements before
2883 (vect_pattern_recog_1): Don't return a bool. If the statement
2884 is already part of a pattern, instead apply pattern matching
2885 to the pattern definition statements. Don't clear the
2886 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
2887 (vect_pattern_recog): Don't break after the first match;
2888 continue processing the pattern definition statements instead.
2889 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
2891 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2893 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
2894 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
2895 (vect_recog_widen_sum_pattern): Use it.
2897 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2899 * tree-vect-loop.c (vectorizable_reduction): Assert that the
2900 phi is not a pattern statement and has not been replaced by
2901 a pattern statement.
2902 * tree-vect-patterns.c (type_conversion_p): Don't check
2903 STMT_VINFO_IN_PATTERN_P.
2904 (vect_recog_vector_vector_shift_pattern): Likewise.
2905 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
2906 the pattern statement rather than the original statement; check
2907 directly for a WIDEN_MULT_EXPR here.
2908 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
2909 vect_is_simple_use to return the pattern statement rather
2910 than the original statement; use is_pattern_stmt_p to check
2911 for such a pattern statement.
2912 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
2913 to return the pattern statement rather than the original statement;
2914 don't do the same transformation here.
2915 (vect_is_simple_use): If the defining statement has been replaced
2916 by a pattern statement, return the pattern statement instead.
2917 Remove the corresponding (local) transformation from the vectype
2920 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2922 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
2923 end and default to null.
2924 * tree-vect-loop.c (vect_create_epilog_for_reduction)
2925 (vectorizable_reduction): Update calls accordingly, dropping the
2926 gimple ** argument if the passed-back statement isn't needed.
2927 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
2928 (vect_recog_rotate_pattern): Likewise.
2929 (vect_recog_mask_conversion_pattern): Likewise.
2930 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
2931 (vect_mask_constant_operand_p): Likewise.
2932 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
2933 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
2934 (get_group_load_store_type, get_load_store_type): Likewise.
2935 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
2936 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
2937 (vectorizable_conversion, vectorizable_assignment): Likewise.
2938 (vectorizable_shift, vectorizable_operation): Likewise.
2939 (vectorizable_store, vect_is_simple_cond): Likewise.
2940 (vectorizable_condition, vectorizable_comparison): Likewise.
2941 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
2942 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
2943 and move it to the end. Cope with null def_stmt_outs.
2945 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
2947 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
2949 2018-06-29 Jeff Law <law@redhat.com>
2951 * config/v850/v850.c (v850_legitimate_address_p): Handle large
2952 displacements for TARGET_V850E2V3 and newer.
2953 (TARGET_LRA_P): Remove. Defaults to LRA now.
2954 * config/v850/v850.md (sign23byte_load): Remove.
2955 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
2956 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
2958 2018-06-29 Martin Liska <mliska@suse.cz>
2961 * coverage.c (coverage_init): Mangle full path name.
2962 * doc/invoke.texi: Document the change.
2963 * gcov-io.c (mangle_path): New.
2964 * gcov-io.h (mangle_path): Likewise.
2965 * gcov.c (mangle_name): Use mangle_path for path mangling.
2967 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2969 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
2970 if starting source register is not even.
2972 2018-06-29 Martin Liska <mliska@suse.cz>
2974 PR tree-optimization/86263
2975 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
2976 Make edge redirection.
2978 2018-06-29 David Malcolm <dmalcolm@redhat.com>
2980 * dumpfile.c (dump_loc): Add indentation based on scope depth.
2981 (dump_scope_depth): New variable.
2982 (get_dump_scope_depth): New function.
2983 (dump_begin_scope): New function.
2984 (dump_end_scope): New function.
2985 * dumpfile.h (get_dump_scope_depth): New declaration.
2986 (dump_begin_scope): New declaration.
2987 (dump_end_scope): New declaration.
2988 (class auto_dump_scope): New class.
2989 (AUTO_DUMP_SCOPE): New macro.
2990 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
2993 2018-06-29 Richard Biener <rguenther@suse.de>
2995 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
2996 compute_all_dependences succeeds.
2997 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
2998 exceed --param loop-max-datarefs-for-datadeps.
3000 2018-06-29 Jakub Jelinek <jakub@redhat.com>
3002 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
3004 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
3007 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
3008 alternative 0 in preferred_for_speed attribute.
3010 2018-06-28 Paul Koning <ni1d@arrl.net>
3012 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
3013 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
3014 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
3015 * config/pdp11/pdp11.md: Correct "length" attribute calculation
3016 for shift insn patterns.
3018 2018-06-28 David Malcolm <dmalcolm@redhat.com>
3020 * cgraph.c (cgraph_node::get_body): Replace assignments to
3021 "dump_file" with calls to set_dump_file.
3022 * dumpfile.c (alt_dump_file): Make static, and group with...
3023 (alt_flags): ...this definition.
3024 (dumps_are_enabled): New variable.
3025 (refresh_dumps_are_enabled): New function.
3026 (set_dump_file): New function.
3027 (set_alt_dump_file): New function.
3028 (gcc::dump_manager::dump_start): Replace assignments to
3029 "dump_file" and "alt_dump_file" with calls to set_dump_file and
3031 (gcc::dump_manager::dump_finish): Likewise.
3032 * dumpfile.h (alt_dump_file): Delete decl.
3033 (dumps_are_enabled): New variable decl.
3034 (set_dump_file): New function decl.
3035 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
3037 * tree-nested.c (lower_nested_functions): Replace assignments to
3038 "dump_file" with calls to set_dump_file.
3040 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
3042 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
3043 goto_locus of each outgoing edge of each basic block.
3045 2018-06-28 Richard Biener <rguenther@suse.de>
3047 * dwarf2out.c (decl_scope_table): Remove.
3048 (push_decl_scope): Likewise.
3049 (pop_decl_scope): Likewise.
3050 (gen_type_die_for_member): Do not call push/pop_decl_scope.
3051 (gen_struct_or_union_type_die): Likewise.
3052 (gen_tagged_type_die): Likewise.
3053 (dwarf2out_init): Do not initialize decl_scope_table.
3054 (dwarf2out_c_finalize): Do not free it.
3056 2018-06-28 Richard Biener <rguenther@suse.de>
3058 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
3059 deciding whether to not re-use a DIE.
3061 2018-06-28 Richard Biener <rguenther@suse.de>
3063 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
3064 DW_AT_abstract_origin attribute.
3066 2018-06-28 Martin Liska <mliska@suse.cz>
3068 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
3069 Use newly introduced constants.
3070 * tree-switch-conversion.h (struct jump_table_cluster):
3071 Define max_ratio_for_size and max_ratio_for_speed.
3073 2018-06-28 Martin Liska <mliska@suse.cz>
3075 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
3076 Add new checking assert to catch invalid state.
3077 (jump_table_cluster::can_be_handled): Handle single case
3079 (jump_table_cluster::is_beneficial): Bail out for such case.
3080 (bit_test_cluster::find_bit_tests):
3081 Add new checking assert to catch invalid state.
3082 (bit_test_cluster::can_be_handled): Handle single case
3084 (bit_test_cluster::is_beneficial): Bail out for such case.
3085 (switch_decision_tree::analyze_switch_statement):
3088 2018-06-28 Martin Liska <mliska@suse.cz>
3090 * common.opt: Introduce -completion option.
3091 * gcc.c (driver_handle_option): Handle it.
3092 (driver::main): Print completions if completion
3094 * opt-suggestions.c (option_proposer::get_completions):
3096 (option_proposer::suggest_completion): Likewise.
3097 (option_proposer::find_param_completions): Likewise.
3098 (verify_autocompletions): Likewise.
3099 (test_completion_valid_options): Likewise.
3100 (test_completion_valid_params): Likewise.
3101 (in_completion_p): Likewise.
3102 (empty_completion_p): Likewise.
3103 (test_completion_partial_match): Likewise.
3104 (test_completion_garbage): Likewise.
3105 (opt_proposer_c_tests): Likewise.
3106 * opt-suggestions.h: Declare new functions.
3107 * opts.c (common_handle_option): Handle OPT__completion_.
3108 * selftest-run-tests.c (selftest::run_tests): Add
3109 opt_proposer_c_tests.
3110 * selftest.c (assert_str_startswith): New.
3111 * selftest.h (assert_str_startswith): Likewise.
3112 (opt_proposer_c_tests): New.
3113 (ASSERT_STR_STARTSWITH): Likewise.
3115 2018-06-28 Martin Liska <mliska@suse.cz>
3117 * Makefile.in: Add opt-suggestions.o.
3118 * gcc-main.c: Include opt-suggestions.h.
3119 * gcc.c (driver::driver): Likewise.
3120 (driver::~driver): Remove m_option_suggestions.
3121 (driver::build_option_suggestions): Moved to option_proposer.
3122 (driver::suggest_option): Likewise.
3123 (driver::handle_unrecognized_options): Use option_proposer.
3124 * gcc.h (class driver): Add new memver m_option_proposer.
3125 * opt-suggestions.c: New file.
3126 * opt-suggestions.h: New file.
3128 2018-06-28 Martin Liska <mliska@suse.cz>
3130 * vec.h (class auto_string_vec): New (moved from auto_argvec).
3131 (auto_string_vec::~auto_string_vec): Likewise.
3133 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
3135 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
3136 prevent_decl_creation_for_types fields up and add reset_location field.
3137 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
3138 statement if id->reset_location is true.
3139 (copy_edges_for_bb): Do not set goto_locus on the new edges if
3140 id->reset_location is true.
3141 (copy_phis_for_bb): Force input_location on the arguments if
3142 id->reset_location is true.
3143 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
3144 is set on the function to be inlined.
3146 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
3148 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
3150 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
3152 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
3153 registers for Pmode.
3154 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
3155 hard registers for the clobbered pseudo.
3157 2018-06-27 Paul Koning <ni1d@arrl.net>
3159 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
3160 mutually exclusive options.
3161 * config/pdp11/constraints.md (h): New constraint.
3162 (O): Update definition to match shift code generation.
3163 (D): New constraint.
3164 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
3166 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
3168 (output_jump): Change arguments.
3169 (pdp11_fixed_cc_regs): New function.
3170 (pdp11_cc_mode): Ditto.
3171 (pdp11_expand_shift): Ditto.
3172 (pdp11_assemble_shift): Ditto.
3173 (pdp11_small_shift): Ditto.
3174 (pdp11_branch_cost): Remove.
3175 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
3177 (pdp11_register_move_cost): Update for CC registers.
3178 (pdp11_rtx_costs): Add case for LSHIFTRT.
3179 (pdp11_output_jump): Add CCNZ mode conditional branches.
3180 (notice_update_cc_on_set): Remove.
3181 (pdp11_cc_mode): New function.
3182 (simple_memory_operand): Correct pre/post decrement case.
3183 (no_side_effect_operand): New function.
3184 (pdp11_regno_reg_class): Add CC_REGS class.
3185 (pdp11_fixed_cc_regs): New function.
3186 (pdp11_small_shift): New function.
3187 (pdp11_expand_shift): New function to expand shift insns.
3188 (pdp11_assemble_shift): New function to output shifts.
3189 (pdp11_branch_cost): Remove.
3190 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
3191 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
3192 (WCHAR_TYPE): Ditto.
3193 (PTRDIFF_TYPE): Ditto.
3194 (ADJUST_INSN_LENGTH): New macro.
3195 (FIXED_REGISTERS): Add CC registers.
3196 (CALL_USED_REGISTERS): Ditto.
3198 (REG_CLASS_NAMES): Ditto.
3199 (REG_CLASS_CONTENTS): Ditto.
3200 (SELECT_CC_MODE): Use new function.
3201 (TARGET_FLAGS_REGNUM): New macro.
3202 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
3203 (cc0_reg_rtx): Remove.
3204 (CC_STATUS_MDEP): Remove.
3205 (CC_STATUS_MDEFP_INIT): Remove.
3206 (CC_IN_FPU): Remove.
3207 (NOTICE_UPDATE_CC): Remove.
3208 (REGISTER_NAMES): Add CC registers.
3209 (BRANCH_COST): Change to constant 1.
3210 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
3212 * config/pdp11/pdp11.opt (mbcopy): Remove.
3213 (mbcopy-builtin): Remove.
3214 (mbranch-cheap): Remove.
3215 (mbranch-expensive): Remove.
3216 * config/pdp11/predicates.md (expand_shift_operand): Update to
3217 match shift code generation.
3218 (ccnz_operator): New predicate.
3219 * doc/invoke.texi (PDP-11 Options): Remove deleted options
3220 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
3221 Remove non-existent option -mabshi, -mno-abshi. Document mutually
3223 * doc/md.texi (PDP-11): Document new D and h constraints. Update
3224 description of O constraint.
3226 2018-06-27 Jeff Law <law@redhat.com>
3227 Austin Law <austinklaw@gmail.com>
3229 * config/v850/v850.md (addsi3_set_flags): New pattern.
3230 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
3231 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
3232 (zero_extendhisi2_v850_set_flags): Likewise.
3233 (zero_extendqisi2_v850_set_flags): Likewise.
3234 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
3235 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
3236 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
3238 * config/v850/v850-protos.h (notice_update_cc): Remove.
3239 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
3240 (v850_print_operand): Handle 'D' and "d".
3241 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
3242 Add handling of arithmetic/logical operations compared against zero.
3243 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
3244 Do not look at v850_compare_op, instead get mode from last argument.
3245 (v850_gen_compare): Remove
3246 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
3247 after reload for prologue insns.
3248 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
3250 (construct_save_jarl): Likewise.
3251 (TARGET_FLAGS_REGNUM): Define.
3252 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
3253 (NOTICE_UPDATE_CC): Remove.
3254 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
3255 than cc0. Conditionalize on reload_completed.
3256 (cmpsi_insn, setfcc_insn): Likewise.
3257 (tst1 splitter): Turn into define_and_split which sets the flags
3259 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
3260 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
3261 (cstoresf4, cstoredf4): Clobber the flags.
3262 (cmpsi, cmpsf, cmpdf): Remove expanders.
3263 (setf_insn): Remove pattern.
3264 (addsi3): Turn into define_and_split which clobbers the flags after
3265 reload and a suitable pattern (addsi3_clobber_flags) for use after
3267 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
3268 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
3269 (ashrsi3, ashrsi3_v850e2): Likewise.
3270 (bins): Clobber the flags.
3271 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
3272 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
3273 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
3274 (call_value_internal_short, call_value_internal_long): Likewise.
3275 (callt_save_interrupt, callt_return_interrupt): Likewise.
3276 (save_interrupt, return_interrupt): Likewise.
3277 (callt_save_all_interrupt, save_all_interrupt): Likewise.
3278 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
3279 (restore_all_interrupt, _restore_all_interrupt): Likewise.
3280 (All FP comparisons): Only allow after reload has completed.
3282 (divh, divhu): Tweak output template.
3283 (branch_z_normal, branch_z_invert): Remove
3284 (branch_nz_normal, branch_nz_invert): Likewise.
3285 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
3287 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
3288 * config/v850/v850.c (notice_update_cc): Remove.
3289 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
3290 (CC_NO_CARRY): Likewise.
3291 (NOTICE_UPDATE_CC): Define to nothing.
3292 * config/v850/v850.md: Remove block comment on cc0 handling
3293 Remove "cc" attribute from all patterns. Remove cc_status handling
3294 from all patterns. Minor formatting fixes.
3296 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3298 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
3299 (cortex-a76.cortex-a55): Likewise.
3300 * config/aarch64/aarch64-tune.md: Regenerate.
3301 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
3302 cortex-a76.cortex-a55.
3304 2018-06-27 Jeff Law <law@redhat.com>
3306 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
3307 (MULTILIB_DIRNAMES): Similarly.
3309 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
3311 * gimple.h (gimple_return_retbnd): Delete.
3312 (gimple_return_set_retbnd): Likewise.
3313 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
3314 gimple_return_set_retbnd.
3315 * gimple-pretty-print.c (dump_gimple_return): Remove call to
3316 gimple_return_retbnd and adjust.
3317 * tree-inline.h (struct copy_body_data): Remove retbnd field.
3318 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
3319 Explicitly return NULL in a couple more cases. Move assertion
3320 on debug statements and remove unreachable code.
3321 (reset_debug_binding): Do not test id->retbnd.
3322 (expand_call_inline): Do not set it.
3324 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
3326 * configure.ac: Add --disable-gcov option.
3327 * configure: Regenerate.
3328 * Makefile.in: Honour @enable_gcov@.
3329 * doc/install.texi: Document --disable-gcov.
3331 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3333 * config/arm/arm-cpus.in (cortex-a76): New entry.
3334 (cortex-a76.cortex-a55): Likewise.
3335 * config/arm/arm-tables.opt: Regenerate.
3336 * config/arm/arm-tune.md: Likewise.
3337 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
3338 * doc/invoke.texi (ARM Options): Document cortex-a76 and
3339 cortex-a76.cortex-a55.
3341 2018-06-27 Tamar Christina <tamar.christina@arm.com>
3344 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
3346 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
3348 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
3350 (EPILOGUE_USES): Likewise.
3352 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
3354 * tree-inline.c (remap_location): New function extracted from...
3355 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
3356 (copy_phis_for_bb): ...here. Call remap_location.
3357 (copy_cfg_body): Adjust call to copy_edges_for_bb.
3359 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
3361 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
3362 unaligned vsx for 16B memset.
3364 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
3367 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
3368 ieee128_float_type_node to long_double_type_node unless
3369 TARGET_LONG_DOUBLE_128 is set.
3371 2018-06-26 David Malcolm <dmalcolm@redhat.com>
3373 * cfgloop.c (get_loop_location): Convert return type from
3374 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
3375 by implicit construction from rtx_insn *, and using
3376 dump_user_location_t::from_function_decl for the fallback case.
3377 * cfgloop.h (get_loop_location): Convert return type from
3378 location_t to dump_user_location_t.
3379 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
3380 dump_printf_loc to pass in a dump_location_t rather than a
3381 location_t, via the gimple stmt.
3382 * coverage.c (get_coverage_counts): Update calls to
3383 dump_printf_loc to pass in dump_location_t rather than a
3385 * doc/optinfo.texi (Dump types): Convert example of
3386 dump_printf_loc from taking "locus" to taking "insn". Update
3387 description of the "_loc" calls to cover dump_location_t.
3388 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
3390 (dump_user_location_t::dump_user_location_t): New constructors,
3391 from gimple *stmt and rtx_insn *.
3392 (dump_user_location_t::from_function_decl): New function.
3393 (dump_loc): Make static.
3394 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
3395 const dump_location_t &.
3396 (dump_generic_expr_loc): Delete.
3397 (dump_printf_loc): Convert param "loc" from location_t to
3398 const dump_location_t &.
3399 (selftest::test_impl_location): New function.
3400 (selftest::dumpfile_c_tests): New function.
3401 * dumpfile.h: Include "profile-count.h".
3402 (class dump_user_location_t): New class.
3403 (struct dump_impl_location_t): New struct.
3404 (class dump_location_t): New class.
3405 (dump_printf_loc): Convert 2nd param from source_location to
3406 const dump_location_t &.
3407 (dump_generic_expr_loc): Delete.
3408 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
3409 const dump_location_t &.
3410 * gimple-fold.c (fold_gimple_assign): Update call to
3411 dump_printf_loc to pass in a dump_location_t rather than a
3412 location_t, via the gimple stmt.
3413 (gimple_fold_call): Likewise.
3414 * gimple-loop-interchange.cc
3415 (loop_cand::analyze_iloop_reduction_var): Update for change to
3416 check_reduction_path.
3417 (tree_loop_interchange::interchange): Update for change to
3419 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
3420 change in return-type of find_loop_location.
3421 (graphite_regenerate_ast_isl): Likewise.
3422 * graphite-optimize-isl.c (optimize_isl): Likewise.
3423 * graphite.c (graphite_transform_loops): Likewise.
3424 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
3425 pass in a dump_location_t rather than a location_t, via the
3427 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3428 * ipa.c (walk_polymorphic_call_targets): Likewise.
3429 * loop-unroll.c (report_unroll): Convert "locus" param from
3430 location_t to dump_location_t.
3431 (decide_unrolling): Update for change to get_loop_location's
3433 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
3434 location_t to dump_user_location_t.
3435 (grid_find_single_omp_among_assignments_1): Updates calls to
3436 dump_printf_loc to pass in a dump_location_t rather than a
3437 location_t, via the gimple stmt.
3438 (grid_parallel_clauses_gridifiable): Convert "tloc" from
3439 location_t to dump_location_t. Updates calls to dump_printf_loc
3440 to pass in a dump_location_t rather than a location_t, via the
3442 (grid_inner_loop_gridifiable_p): Likewise.
3443 (grid_dist_follows_simple_pattern): Likewise.
3444 (grid_gfor_follows_tiling_pattern): Likewise.
3445 (grid_target_follows_gridifiable_pattern): Likewise.
3446 (grid_attempt_target_gridification): Convert initialization
3447 of local "grid" from memset to zero-initialization; FIXME: does
3448 this require C++11? Update call to dump_printf_loc to pass in a
3449 optinfo_location rather than a location_t, via the gimple stmt.
3450 * profile.c (read_profile_edge_counts): Updates call to
3451 dump_printf_loc to pass in a dump_location_t rather than a
3453 (compute_branch_probabilities): Likewise.
3454 * selftest-run-tests.c (selftest::run_tests): Call
3456 * selftest.h (dumpfile_c_tests): New decl.
3457 * tree-loop-distribution.c (pass_loop_distribution::execute):
3458 Update for change in return type of find_loop_location.
3459 * tree-parloops.c (parallelize_loops): Likewise.
3460 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
3461 "locus" from location_t to dump_user_location_t.
3462 (canonicalize_loop_induction_variables): Likewise.
3463 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
3464 for change in return type of find_loop_location.
3465 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
3466 to dump_printf_loc to pass in a dump_location_t rather than a
3467 location_t, via the stmt.
3468 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
3470 * tree-vect-loop-manip.c (find_loop_location): Convert return
3471 type from source_location to dump_user_location_t.
3472 (vect_do_peeling): Update for above change.
3473 (vect_loop_versioning): Update for change in type of
3475 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
3476 from location_t to dump_user_location_t.
3477 (vect_estimate_min_profitable_iters): Update for change in type
3479 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
3480 location_t to dump_location_t.
3481 (vect_slp_bb): Update for change in type of vect_location.
3482 * tree-vectorizer.c (vect_location): Convert from source_location
3483 to dump_user_location_t.
3484 (try_vectorize_loop_1): Update for change in vect_location's type.
3485 (vectorize_loops): Likewise.
3486 (increase_alignment): Likewise.
3487 * tree-vectorizer.h (vect_location): Convert from source_location
3488 to dump_user_location_t.
3489 (find_loop_location): Convert return type from source_location to
3490 dump_user_location_t.
3491 (check_reduction_path): Convert 1st param from location_t to
3492 dump_user_location_t.
3493 * value-prof.c (check_counter): Update call to dump_printf_loc to
3494 pass in a dump_user_location_t rather than a location_t; update
3495 call to error_at for change in type of "locus".
3496 (check_ic_target): Update call to dump_printf_loc to
3497 pass in a dump_user_location_t rather than a location_t, via the
3500 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
3502 * config/s390/s390.h (enum processor_flags): Do not use
3503 default tune parameter when -march was specified.
3505 2018-06-26 Jakub Jelinek <jakub@redhat.com>
3508 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
3509 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
3512 2018-06-26 Richard Biener <rguenther@suse.de>
3514 PR tree-optimization/86287
3516 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
3517 (vect_analyze_loop): Initialize n_stmts.
3519 2018-06-26 Richard Biener <rguenther@suse.de>
3522 * fold-const.c (fold_convertible_p): Pointer extension
3525 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
3528 * dwarf2out.c (loc_list_has_views): Adjust comments.
3529 (dw_loc_list): Split single cross-partition range with
3532 2018-06-25 Jeff Law <law@redhat.com>
3534 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
3535 on -mbig-switch by default.
3537 * config/v850/predicates.md (const_float_1_operand): Fix match_code
3539 (const_float_0_operand): Remove unused predicate.
3540 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
3541 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
3542 (recipsf2): New expander. Original pattern now called
3544 (recipdf2, recipdf2_insn): Similarly.
3545 (rsqrtsf2, rsqrtsf2_insn): Similarly
3546 (rsqrtdf2, rsqrtdf2_insn): Similarly
3548 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
3550 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
3551 Simplify logic for FreeBSD (twice).
3553 2018-06-25 Martin Sebor <msebor@redhat.com>
3555 PR tree-optimization/86204
3556 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
3557 a strnlen result if it's less than the length of the string.
3559 2018-06-25 Martin Sebor <msebor@redhat.com>
3561 PR tree-optimization/85700
3562 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
3563 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
3564 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
3566 2018-06-25 Martin Sebor <msebor@redhat.com>
3568 * doc/extend.texi (Zero-length arrays): Update and clarify.
3570 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
3572 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
3573 added IEEE/IBM long double multilib support on PowerPC little
3574 endian Linux systems.
3575 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
3576 (MULTILIB_DEFAULTS): Likewise.
3577 * config/rs6000/rs6000.c (rs6000_option_override_internal):
3579 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
3580 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
3581 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
3583 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
3586 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
3587 (REORDER_45): Likewise.
3589 2018-06-25 Jeff Law <law@redhat.com>
3591 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
3592 dividend to 32 bits. Adjust length.
3593 (udivmodhi4): Cleanup output template. Fix length.
3595 2018-06-25 Carl Love <cel@us.ibm.com>
3597 * config/rs6000/vsx.md: Change word selector to prefered location.
3599 2018-06-25 Richard Biener <rguenther@suse.de>
3601 PR tree-optimization/86304
3602 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
3603 epilogue-if-converted loops as well.
3605 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
3607 * lto-section-out.c (lto_begin_section): Do not print section
3608 name for noaddr and unnumbered dumps.
3610 2018-06-25 Richard Biener <rguenther@suse.de>
3612 * tree-vectorizer.h (struct vec_info_shared): New structure
3613 with parts split out from struct vec_info and loop_nest from
3614 struct _loop_vec_info.
3615 (struct vec_info): Adjust accordingly.
3616 (struct _loop_vec_info): Likewise.
3617 (LOOP_VINFO_LOOP_NEST): Adjust.
3618 (LOOP_VINFO_DATAREFS): Likewise.
3619 (LOOP_VINFO_DDRS): Likewise.
3620 (struct _bb_vec_info): Likewise.
3621 (BB_VINFO_DATAREFS): Likewise.
3622 (BB_VINFO_DDRS): Likewise.
3623 (struct _stmt_vec_info): Add dr_aux member.
3624 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
3625 (DR_MISALIGNMENT_UNINITIALIZED): New.
3626 (set_dr_misalignment): Adjust.
3627 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
3628 (vect_analyze_loop): Adjust prototype.
3629 (vect_analyze_loop_form): Likewise.
3630 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
3631 Compute dependences lazily.
3632 (vect_record_base_alignments): Use shared datarefs/ddrs.
3633 (vect_verify_datarefs_alignment): Likewise.
3634 (vect_analyze_data_refs_alignment): Likewise.
3635 (vect_analyze_data_ref_accesses): Likewise.
3636 (vect_analyze_data_refs): Likewise.
3637 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
3638 constructor parameter for shared part.
3639 (vect_analyze_loop_form): Pass in shared part and adjust.
3640 (vect_analyze_loop_2): Pass in storage for the number of
3641 stmts. Move loop nest finding to the caller. Compute
3643 (vect_analyze_loop): Pass in shared part.
3644 (vect_transform_loop): Verify shared datarefs are unchanged.
3645 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
3646 constructor parameter for shared part.
3647 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
3648 (vect_slp_bb): Verify shared datarefs are unchanged before
3650 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
3652 (new_stmt_vec_info): Initialize DR_AUX misalignment to
3653 DR_MISALIGNMENT_UNINITIALIZED.
3654 * tree-vectorizer.c (vec_info::vec_info): Add constructor
3655 parameter for shared part.
3656 (vec_info::~vec_info): Adjust.
3657 (vec_info_shared::vec_info_shared): New.
3658 (vec_info_shared::~vec_info_shared): Likewise.
3659 (vec_info_shared::save_datarefs): Likewise.
3660 (vec_info_shared::check_datarefs): Likewise.
3661 (try_vectorize_loop_1): Construct shared part live for analyses
3662 of a single loop for multiple vector sizes.
3663 * tree-parloops.c (gather_scalar_reductions): Adjust.
3665 2018-06-25 Richard Biener <rguenther@suse.de>
3667 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
3668 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
3669 (vect_analyze_data_refs): Remove similar code from here and
3670 simplify accordingly.
3672 2018-06-25 Richard Biener <rguenther@suse.de>
3674 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
3675 for reverse storage order accesses rather than asserting
3676 they cannot happen here.
3678 2018-06-25 Tom de Vries <tdevries@suse.de>
3681 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
3682 Use data16 instead of .byte for insn prefix.
3684 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
3687 * parser.c (make_char_string_pack): Pass this literal chars
3688 through cpp_interpret_string.
3689 (cp_parser_userdef_numeric_literal): Check the result of
3690 make_char_string_pack.
3692 2018-06-24 Maya Rashish <coypu@sdf.org>
3694 * ginclude/stddef.h: Simplify conditions around avoiding
3695 re-definition of __size_t.
3697 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
3699 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
3700 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
3702 2018-06-22 Maya Rashish <coypu@sdf.org>
3704 * doc/invoke.texi (mno-fancy-math-387): Update for changes
3705 made to OpenBSD and NetBSD through the years.
3707 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3709 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
3710 behavior of vec_pack (vector double, vector double) to match
3711 behavior of vec_float2 (vector double, vector double).
3713 2018-06-22 Olivier Hainque <hainque@adacore.com>
3715 * gimplify.c (gimplify_function_tree): Prevent creation
3716 of a trampoline for the address of the current function
3717 passed to entry/exit instrumentation hooks.
3719 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
3722 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
3725 2018-06-22 Martin Liska <mliska@suse.cz>
3727 PR tree-optimization/86263
3728 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
3729 Bail out if is_enabled is false.
3730 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
3732 (jump_table_cluster::is_enabled): New function.
3734 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
3736 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
3737 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
3738 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
3739 (lto_input_ts_binfo_tree_pointers): Likewise.
3740 * tree-streamer-out.c (streamer_write_tree_bitfields,
3741 write_ts_binfo_tree_pointers): Likewise.
3742 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
3744 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
3746 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
3748 2018-06-22 Martin Liska <mliska@suse.cz>
3750 * symbol-summary.h (get): Make it pure and inline move
3751 functionality from ::get function.
3752 (get): Remove and inline into ::get and ::get_create.
3753 (get_create): Move code from ::get function.
3755 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3758 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
3759 -x assembler-with-cpp.
3761 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3763 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
3764 _FILE_OFFSET_BITS=64 for C++.
3766 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
3768 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
3769 conversion insn that shows up when pr85657-3.c is compiled using
3770 IEEE 128-bit long double.
3771 (neg<mode>2_internal): Use the correct mode to check whether the
3772 mode is IBM extended.
3773 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
3774 multiply and divide external functions from being created more
3777 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
3779 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
3781 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
3782 the edge can be forwarded.
3783 (cfg_layout_merge_blocks): Likewise.
3785 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
3787 * except.c (finish_eh_generation): Commit edge insertions only after
3788 the EH edges have been redirected from post-landing to landing pads.
3790 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
3792 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
3793 create_tmp_var_for to create the FRAME decl.
3794 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
3796 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
3798 * tree-inline.c (copy_edges_for_bb): Minor tweak.
3799 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
3800 debug statement when resetting its value.
3801 (expand_call_inline): Copy the locus of the call onto the assignment
3802 of the return value, if any. Use local variable in more cases.
3804 2018-06-21 Martin Liska <mliska@suse.cz>
3806 * ipa-pure-const.c (propagate_nothrow): Use
3807 funct_state_summaries->get.
3808 (dump_malloc_lattice): Likewise.
3809 (propagate_malloc): Likewise.
3811 2018-06-21 Richard Biener <rguenther@suse.de>
3813 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
3814 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
3815 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
3817 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
3818 BLOCK_ABSTRACT_ORIGIN unconditionally.
3820 2018-06-21 David Malcolm <dmalcolm@redhat.com>
3822 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
3824 * ipa-reference.c (ipa_reference_c_finalize): Delete
3825 ipa_ref_opt_sum_summaries and set it to NULL.
3827 2018-06-21 Tom de Vries <tdevries@suse.de>
3829 PR tree-optimization/85859
3830 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
3831 test with comment from bb_no_side_effects_p.
3833 2018-06-21 Richard Biener <rguenther@suse.de>
3835 PR tree-optimization/86232
3836 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
3837 max for constant niter.
3839 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
3841 * config/aarch64/aarch64-simd.md
3842 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
3844 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
3846 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
3847 Make opernads of the unspec commutative.
3849 2018-06-21 Richard Biener <rguenther@suse.de>
3851 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
3852 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
3853 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
3854 (vect_analyze_data_ref_dependence): Re-order checks to deal with
3856 (vect_record_base_alignments): Do not record base alignment
3857 for gathers or scatters.
3858 (vect_compute_data_ref_alignment): Drop return value that is always
3859 true. Bail out early for gathers or scatters.
3860 (vect_enhance_data_refs_alignment): Bail out early for gathers
3862 (vect_find_same_alignment_drs): Likewise.
3863 (vect_analyze_data_refs_alignment): Remove dead code.
3864 (vect_slp_analyze_and_verify_node_alignment): Likewise.
3865 (vect_analyze_data_refs): For possible gathers or scatters do
3866 not create an alternate DR, just check their possible validity
3867 and mark them. Adjust DECL_NONALIASED handling to not rely
3869 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
3870 update inits of gathers or scatters.
3871 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
3872 Also copy gather/scatter flag to pattern vinfo.
3874 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
3876 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
3877 behavior of vec_packsu (vector unsigned long long, vector unsigned
3878 long long) to match behavior of vec_packs with same signature.
3880 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
3881 Thomas Schwinge <thomas@codesourcery.com>
3882 Cesar Philippidis <cesar@codesourcery.com>
3884 * gimplify.c (gimplify_scan_omp_clauses): Add support for
3885 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
3886 (gimplify_adjust_omp_clauses): Likewise.
3887 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
3888 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
3889 (gimplify_omp_target_update): Update handling of acc update and
3891 * omp-low.c (install_var_field): Remove unused parameter
3892 base_pointers_restrict.
3893 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
3894 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
3896 (omp_target_base_pointers_restrict_p): Delete.
3897 (scan_omp_target): Update call to scan_sharing_clauses.
3898 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
3900 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
3901 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
3902 (convert_local_omp_clauses): Likewise.
3903 * tree-pretty-print.c (dump_omp_clause): Likewise.
3904 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
3906 (omp_clause_code_name): Likewise.
3908 2018-06-20 Jakub Jelinek <jakub@redhat.com>
3911 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
3914 PR tree-optimization/86231
3915 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
3916 anti-range don't overwrite *vr0min before using it to compute *vr0max.
3918 2018-06-20 Tom de Vries <tdevries@suse.de>
3920 PR tree-optimization/86097
3921 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
3922 iv type if signedness of iv type is not the same as that of *nit.
3924 2018-06-20 Jakub Jelinek <jakub@redhat.com>
3926 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
3927 EDGE_EH edges, verify they are all EDGE_EH.
3929 2018-06-20 Maya Rashish <coypu@sdf.org>
3931 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
3933 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3935 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
3936 * config/aarch64/aarch64.c (xgene1_tunings): Add
3937 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
3938 (aarch64_mode_valid_for_sched_fusion_p):
3939 Allow 16-byte modes.
3940 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
3941 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
3943 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
3945 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
3946 * config/aarch64/iterators.md (VQ2): New mode iterator.
3948 2018-06-20 Martin Liska <mliska@suse.cz>
3950 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
3951 Change default ratio from 10 to 8.
3953 2018-06-20 Martin Liska <mliska@suse.cz>
3955 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
3957 (bit_test_cluster::find_bit_tests): Likewise.
3958 (switch_decision_tree::analyze_switch_statement): Find clusters.
3959 * tree-switch-conversion.h (struct jump_table_cluster): Document
3962 2018-06-20 Martin Liska <mliska@suse.cz>
3964 * tree-switch-conversion.c (switch_conversion::collect):
3965 Record m_uniq property.
3966 (switch_conversion::expand): Bail out for special conditions.
3967 (group_cluster::~group_cluster): New.
3968 (group_cluster::group_cluster): Likewise.
3969 (group_cluster::dump): Likewise.
3970 (jump_table_cluster::emit): New.
3971 (switch_decision_tree::fix_phi_operands_for_edges): New.
3972 (struct case_node): Remove struct.
3973 (jump_table_cluster::can_be_handled): New.
3974 (case_values_threshold): Moved to header.
3975 (reset_out_edges_aux): Likewise.
3976 (jump_table_cluster::is_beneficial): New.
3977 (bit_test_cluster::can_be_handled): Likewise.
3978 (add_case_node): Remove.
3979 (bit_test_cluster::is_beneficial): New.
3980 (case_bit_test::cmp): New.
3981 (bit_test_cluster::emit): New.
3982 (expand_switch_as_decision_tree_p): Remove.
3983 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
3984 (fix_phi_operands_for_edge): Likewise.
3985 (switch_decision_tree::analyze_switch_statement): New.
3986 (compute_cases_per_edge): Move ...
3987 (switch_decision_tree::compute_cases_per_edge): ... here.
3988 (try_switch_expansion): Likewise.
3989 (switch_decision_tree::try_switch_expansion): Likewise.
3990 (record_phi_operand_mapping): Likewise.
3991 (switch_decision_tree::record_phi_operand_mapping): Likewise.
3992 (emit_case_decision_tree): Likewise.
3993 (switch_decision_tree::emit): Likewise.
3994 (balance_case_nodes): Likewise.
3995 (switch_decision_tree::balance_case_nodes): Likewise.
3996 (dump_case_nodes): Likewise.
3997 (switch_decision_tree::dump_case_nodes): Likewise.
3998 (emit_jump): Likewise.
3999 (switch_decision_tree::emit_jump): Likewise.
4000 (emit_cmp_and_jump_insns): Likewise.
4001 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
4002 (emit_case_nodes): Likewise.
4003 (switch_decision_tree::emit_case_nodes): Likewise.
4004 (conditional_probability): Remove.
4005 * tree-switch-conversion.h (enum cluster_type): New.
4007 (struct cluster): Likewise.
4008 (cluster::cluster): Likewise.
4009 (struct simple_cluster): Likewise.
4010 (simple_cluster::simple_cluster): Likewise.
4011 (struct group_cluster): Likewise.
4012 (struct jump_table_cluster): Likewise.
4013 (struct bit_test_cluster): Likewise.
4014 (struct min_cluster_item): Likewise.
4015 (struct case_tree_node): Likewise.
4016 (case_tree_node::case_tree_node): Likewise.
4017 (jump_table_cluster::case_values_threshold): Likewise.
4018 (struct case_bit_test): Likewise.
4019 (struct switch_decision_tree): Likewise.
4020 (struct switch_conversion): Likewise.
4021 (switch_decision_tree::reset_out_edges_aux): Likewise.
4023 2018-06-20 Martin Liska <mliska@suse.cz>
4025 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
4026 (hoist_edge_and_branch_if_true): Likewise.
4027 (expand_switch_using_bit_tests_p): Likewise.
4028 (struct case_bit_test): Likewise.
4029 (case_bit_test_cmp): Likewise.
4030 (emit_case_bit_tests): Likewise.
4031 (switch_conversion::switch_conversion): New class.
4032 (struct switch_conv_info): Remove old struct.
4033 (collect_switch_conv_info): More to ...
4034 (switch_conversion::collect): ... this.
4035 (check_range): Likewise.
4036 (switch_conversion::check_range): Likewise.
4037 (check_all_empty_except_final): Likewise.
4038 (switch_conversion::check_all_empty_except_final): Likewise.
4039 (check_final_bb): Likewise.
4040 (switch_conversion::check_final_bb): Likewise.
4041 (create_temp_arrays): Likewise.
4042 (switch_conversion::create_temp_arrays): Likewise.
4043 (free_temp_arrays): Likewise.
4044 (gather_default_values): Likewise.
4045 (switch_conversion::gather_default_values): Likewise.
4046 (build_constructors): Likewise.
4047 (switch_conversion::build_constructors): Likewise.
4048 (constructor_contains_same_values_p): Likewise.
4049 (switch_conversion::contains_same_values_p): Likewise.
4050 (array_value_type): Likewise.
4051 (switch_conversion::array_value_type): Likewise.
4052 (build_one_array): Likewise.
4053 (switch_conversion::build_one_array): Likewise.
4054 (build_arrays): Likewise.
4055 (switch_conversion::build_arrays): Likewise.
4056 (gen_def_assigns): Likewise.
4057 (switch_conversion::gen_def_assigns): Likewise.
4058 (prune_bbs): Likewise.
4059 (switch_conversion::prune_bbs): Likewise.
4060 (fix_phi_nodes): Likewise.
4061 (switch_conversion::fix_phi_nodes): Likewise.
4062 (gen_inbound_check): Likewise.
4063 (switch_conversion::gen_inbound_check): Likewise.
4064 (process_switch): Use the newly created class.
4065 (switch_conversion::expand): New.
4066 (switch_conversion::~switch_conversion): New.
4067 * tree-switch-conversion.h: New file.
4069 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4071 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
4072 tree-vect-patterns.c.
4073 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
4074 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
4075 (vect_recog_sad_pattern): Likewise.
4076 (vect_recog_widen_sum_pattern): Likewise.
4077 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
4078 (vect_recog_widen_shift_pattern): Remove the type_in argument.
4079 (vect_recog_rotate_pattern): Likewise.
4080 (vect_recog_mult_pattern): Likewise.
4081 (vect_recog_vector_vector_shift_pattern): Likewise.
4082 (vect_recog_divmod_pattern): Likewise.
4083 (vect_recog_mixed_size_cond_pattern): Likewise.
4084 (vect_recog_bool_pattern): Likewise.
4085 (vect_recog_mask_conversion_pattern): Likewise.
4086 (vect_try_gather_scatter_pattern): Likewise.
4087 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
4088 (vect_recog_over_widening_pattern): Likewise.
4089 (vect_recog_gather_scatter_pattern): Likewise.
4090 (vect_recog_func_ptr): Move from tree-vectorizer.h
4091 (vect_vect_recog_func_ptrs): Move further down the file.
4092 (vect_recog_func): Likewise. Remove the third argument.
4093 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
4094 (vect_pattern_recog_1): Expect the pattern function to do any
4095 necessary target tests. Also expect it to provide a vector type.
4096 Remove the type_in handling.
4098 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4100 * tree-vect-patterns.c (vect_pattern_detected): New function.
4101 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
4102 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
4103 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
4104 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
4105 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
4106 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
4107 (vect_recog_mask_conversion_pattern)
4108 (vect_try_gather_scatter_pattern): Likewise.
4110 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4112 * tree-vect-patterns.c (vect_get_internal_def): New function.
4113 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
4114 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
4115 (search_type_for_mask_1): Use it.
4117 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4119 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
4120 redundant WIDEN_SUM_EXPR handling.
4121 (vect_recog_sad_pattern): Likewise.
4123 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4125 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
4126 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
4127 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
4128 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
4129 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
4131 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4133 * tree-vect-stmts.c (vectorizable_call): Make sure that we
4134 use the stmt_vec_info of the original bb statement for the
4135 new zero assignment, even if the call is part of a pattern.
4137 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4139 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
4140 that the sequence is attached to the original statement rather
4141 than the pattern statement.
4142 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
4143 PATTERN_DEF_SEQ from the original statement rather than
4144 the main pattern statement.
4145 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
4146 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
4147 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
4149 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4151 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
4152 definition statements before the early exit for statements that aren't
4154 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
4156 (vect_transform_loop): ...here. Process pattern definition
4157 statements without first checking whether the main pattern
4158 statement is live or relevant.
4160 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
4162 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
4163 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
4165 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
4167 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
4168 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
4169 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
4170 (expand_block_compare): Change select_block_compare_mode call.
4171 (expand_strncmp_align_check): Use new functions, fix comment.
4172 (emit_final_str_compare_gpr): New function.
4173 (expand_strn_compare): Refactor and clean up code.
4174 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
4176 2018-06-19 Tony Reix <tony.reix@atos.com>
4177 Damien Bergamini <damien.bergamini@atos.com>
4178 David Edelsohn <dje.gcc@gmail.com>
4180 * collect2.c (static_obj): New variable.
4181 (static_libs): New variable.
4182 (is_in_list): Uncomment declaration.
4183 (main): Track AIX libraries linked statically.
4184 (is_in_list): Uncomment definition.
4185 (scan_prog_file): Don't add AIX shared libraries initializer
4186 to constructor list if linking statically.
4188 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
4190 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
4192 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
4194 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
4196 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
4199 2018-06-19 Martin Liska <mliska@suse.cz>
4201 * config/i386/i386.c (ix86_can_inline_p): Do not use
4202 ipa_fn_summaries::get_create.
4203 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
4205 (devirtualization_time_bonus): Likewise.
4206 (ipcp_propagate_stage): Likewise.
4207 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
4208 (edge_set_predicate): Likewise.
4209 (evaluate_conditions_for_known_args): Likewise.
4210 (evaluate_properties_for_edge): Likewise.
4211 (ipa_call_summary::reset): Tranform to ...
4212 (ipa_call_summary::~ipa_call_summary): ... this.
4213 (ipa_fn_summary::reset): Transform to ...
4214 (ipa_fn_summary::~ipa_fn_summary): ... this.
4215 (ipa_fn_summary_t::remove): Rename to ...
4216 (ipa_fn_summary_t::remove_callees): ... this.
4217 (ipa_fn_summary_t::duplicate): Use placement new
4218 instead of memory copy.
4219 (ipa_call_summary_t::duplicate): Likewise.
4220 (ipa_call_summary_t::remove): Remove.
4221 (dump_ipa_call_summary): Change get_create to get.
4222 (ipa_dump_fn_summary): Dump only when summary exists.
4223 (analyze_function_body): Use symbol_summary::get instead
4225 (compute_fn_summary): Likewise.
4226 (estimate_edge_devirt_benefit): Likewise.
4227 (estimate_edge_size_and_time): Likewise.
4228 (inline_update_callee_summaries): Likewise.
4229 (remap_edge_change_prob): Likewise.
4230 (remap_edge_summaries): Likewise.
4231 (ipa_merge_fn_summary_after_inlining): Likewise.
4232 (write_ipa_call_summary): Likewise.
4233 (ipa_fn_summary_write): Likewise.
4234 (ipa_free_fn_summary): Likewise.
4235 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
4236 (struct ipa_call_summary): Likewise.
4237 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
4239 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
4240 (estimate_size_after_inlining): Likewise.
4241 (estimate_growth): Likewise.
4242 (growth_likely_positive): Likewise.
4243 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
4244 (inline_call): Likewise.
4245 * ipa-inline.c (caller_growth_limits): Likewise.
4246 (can_inline_edge_p): Likewise.
4247 (can_inline_edge_by_limits_p): Likewise.
4248 (compute_uninlined_call_time): Likewise.
4249 (compute_inlined_call_time): Likewise.
4250 (want_inline_small_function_p): Likewise.
4251 (edge_badness): Likewise.
4252 (update_caller_keys): Likewise.
4253 (update_callee_keys): Likewise.
4254 (inline_small_functions): Likewise.
4255 (inline_to_all_callers_1): Likewise.
4256 (dump_overall_stats): Likewise.
4257 (early_inline_small_functions): Likewise.
4258 (early_inliner): Likewise.
4259 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
4260 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4261 * ipa-pure-const.c (malloc_candidate_p): Likewise.
4262 * ipa-split.c (execute_split_functions): Likewise.
4263 * symbol-summary.h: Likewise.
4264 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
4266 2018-06-19 Richard Biener <rguenther@suse.de>
4268 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
4269 (vectorize_loops): ... here. Fix dbgcnt handling.
4270 (try_vectorize_loop): Wrap try_vectorize_loop_1.
4272 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
4275 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
4276 ieee128 argument takes up only one (vector) register, not two (floating
4279 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
4281 * gimplify.c (gimplify_init_constructor): Really never clear for an
4282 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
4284 2018-06-19 Richard Biener <rguenther@suse.de>
4286 PR tree-optimization/86179
4287 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
4288 after failed recognition.
4290 2018-06-18 Martin Sebor <msebor@redhat.com>
4293 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
4294 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
4295 Handle integer subtraction.
4296 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
4297 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
4299 2018-06-18 David Malcolm <dmalcolm@redhat.com>
4301 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
4302 param from rtx to rtx_insn *.
4303 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
4305 (frv_ifcvt_modify_insn): Likwise.
4306 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
4307 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
4308 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
4309 as_a <rtx_insn *> cast to local "unprotected_region" once
4310 it's been established that it's not NULL or pc_rtx.
4311 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
4312 param "sethi" from rtx to rtx_insn *.
4313 (nds32_group_float_insns): Likewise for param "insn".
4314 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
4316 (vax_output_int_subtract): Likewise.
4317 * config/vax/vax.c (vax_output_int_add): Likewise for param
4319 (vax_output_int_subtract): Likewise.
4320 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
4321 (emit_pattern_after): Likewise for param "after".
4322 (emit_insn_after): Likewise.
4323 (emit_jump_insn_after): Likewise.
4324 (emit_call_insn_after): Likewise.
4325 (emit_debug_insn_after): Likewise.
4326 (emit_pattern_before): Likewise for param "before".
4327 (emit_insn_before): Likewise.
4328 (emit_jump_insn_before): Likewise.
4329 * final.c (get_insn_template): Likewise for param "insn", removing
4331 * output.h (get_insn_template): Likewise for 2nd param.
4332 * rtl.h (emit_insn_before): Likewise.
4333 (emit_jump_insn_before): Likewise.
4334 (emit_debug_insn_before_noloc): Likewise.
4335 (emit_insn_after): Likewise.
4336 (emit_jump_insn_after): Likewise.
4337 (emit_call_insn_after): Likewise.
4338 (emit_debug_insn_after): Likewise.
4339 (set_insn_deleted): Likewise for param.
4341 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
4344 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
4345 floating point modes, so that IFmode is numerically greater than
4346 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
4347 to declare the ordering. This prevents IFmode from being
4348 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
4349 machine. Include rs6000-modes.h to share the fractional values
4350 between genmodes* and the rest of the compiler.
4354 * config/rs6000/rs6000-modes.h: New file.
4355 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
4356 meaning of rs6000_long_double_size so that 126..128 selects an
4357 appropriate 128-bit floating point type.
4358 (rs6000_option_override_internal): Likewise.
4359 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
4360 (TARGET_LONG_DOUBLE_128): Change the meaning of
4361 rs6000_long_double_size so that 126..128 selects an appropriate
4362 128-bit floating point type.
4363 (LONG_DOUBLE_TYPE_SIZE): Update comment.
4364 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
4365 source and destination to match the standard usage.
4366 (truncifkf2): Likewise.
4367 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
4368 ISA 2.07 to use an explicit clobber, instead of passing in a
4370 (copysign<mode>3_soft): Likewise.
4372 2018-06-18 David Malcolm <dmalcolm@redhat.com>
4374 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
4375 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
4376 (vect_slp_analyze_instance_dependence): Likewise.
4377 (vect_enhance_data_refs_alignment): Likewise.
4378 (vect_analyze_data_refs_alignment): Likewise.
4379 (vect_slp_analyze_and_verify_instance_alignment
4380 (vect_analyze_data_ref_accesses): Likewise.
4381 (vect_prune_runtime_alias_test_list): Likewise.
4382 (vect_analyze_data_refs): Likewise.
4383 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
4384 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
4385 (vect_analyze_scalar_cycles_1): Likewise.
4386 (vect_get_loop_niters): Likewise.
4387 (vect_analyze_loop_form_1): Likewise.
4388 (vect_update_vf_for_slp): Likewise.
4389 (vect_analyze_loop_operations): Likewise.
4390 (vect_analyze_loop): Likewise.
4391 (vectorizable_induction): Likewise.
4392 (vect_transform_loop): Likewise.
4393 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
4394 * tree-vect-slp.c (vect_analyze_slp): Likewise.
4395 (vect_make_slp_decision): Likewise.
4396 (vect_detect_hybrid_slp): Likewise.
4397 (vect_slp_analyze_operations): Likewise.
4398 (vect_slp_bb): Likewise.
4399 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
4400 (vectorizable_bswap): Likewise.
4401 (vectorizable_call): Likewise.
4402 (vectorizable_simd_clone_call): Likewise.
4403 (vectorizable_conversion): Likewise.
4404 (vectorizable_assignment): Likewise.
4405 (vectorizable_shift): Likewise.
4406 (vectorizable_operation): Likewise.
4407 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
4409 2018-06-18 Martin Sebor <msebor@redhat.com>
4411 PR tree-optimization/81384
4412 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
4413 * builtins.c (expand_builtin_strnlen): New function.
4414 (expand_builtin): Call it.
4415 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
4416 * builtins.def (BUILT_IN_STRNLEN): New.
4417 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
4418 Warn for bounds in excess of maximum object size.
4419 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
4420 single-value ranges. Handle strnlen.
4421 (handle_builtin_strlen): Handle strnlen.
4422 (strlen_check_and_optimize_stmt): Same.
4423 * doc/extend.texi (Other Builtins): Document strnlen.
4425 2018-06-18 Maya Rashish <coypu@sdf.org>
4427 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
4428 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
4429 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
4431 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
4433 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
4435 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4437 * tree.c (escaped_string::escape): Replace cast to char * by
4438 const_cast<char *> (unescaped).
4440 2018-06-18 Nick Clifton <nickc@redhat.com>
4443 * tree.c (escaped_string): New class. Converts an unescaped
4444 string into its escaped equivalent.
4445 (warn_deprecated_use): Use the new class to convert the
4446 deprecation message, if present.
4447 (test_escaped_strings): New self test.
4448 (test_c_tests): Add test_escaped_strings.
4449 * doc/extend.texi (deprecated): Add a note that the
4450 deprecation message is affected by the -fmessage-length
4451 option, and that control characters will be escaped.
4452 (#pragma GCC error): Document this pragma.
4453 (#pragma GCC warning): Likewise.
4454 * doc/invoke.texi (-fmessage-length): Document this option's
4455 effect on the #warning and #error preprocessor directives and
4456 the deprecated attribute.
4458 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
4460 * tree.c (decl_value_expr_lookup): Revert latest change.
4461 (decl_value_expr_insert): Likewise.
4463 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
4465 * gimplify.c (nonlocal_vlas): Delete.
4466 (nonlocal_vla_vars): Likewise.
4467 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
4469 (gimplify_body): Do not create and destroy nonlocal_vlas.
4470 * tree-nested.c: Include diagnostic.h.
4471 (use_pointer_in_frame): Tweak.
4472 (lookup_field_for_decl): Add assertion and declare the transformation.
4473 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
4474 internal error when the reference is in a wrong context. Do not
4475 create a debug decl by default.
4476 (note_nonlocal_block_vlas): Delete.
4477 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
4478 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
4479 create a debug decl by default.
4480 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
4481 call is in a wrong context.
4482 (fixup_vla_decls): New function.
4483 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
4484 debug variables were created.
4485 * tree.c (decl_value_expr_lookup): Add checking assertion.
4486 (decl_value_expr_insert): Likewise.
4488 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
4491 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
4492 * tree-scalar-evolution.c (interpret_expr): Likewise.
4493 (expression_expensive_p): Likewise.
4494 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
4495 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
4496 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
4497 (ssa_defined_by_minus_one_stmt_p): New.
4499 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
4502 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
4503 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
4504 * dojump.c (do_jump): Likewise.
4505 * expr.c (expand_expr_real_2): Check operand type's sign.
4506 * fold-const.c (const_unop): Handle ABSU_EXPR.
4507 (fold_abs_const): Likewise.
4508 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
4509 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
4510 (strip_sign_op_1): Likesise.
4511 * match.pd: Add new pattern to generate ABSU_EXPR.
4512 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
4513 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
4514 * tree-eh.c (operation_could_trap_helper_p): Likewise.
4515 * tree-inline.c (estimate_operator_cost): Likewise.
4516 * tree-pretty-print.c (dump_generic_node): Likewise.
4517 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
4518 * tree.def (ABSU_EXPR): New.
4520 2018-06-16 Jakub Jelinek <jakub@redhat.com>
4523 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
4524 documented as preserved for backward compatibility only.
4525 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
4527 PR rtl-optimization/86108
4528 * bb-reorder.c (create_forwarder_block): Renamed to ...
4529 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
4530 jump from new landing pad to the second part.
4531 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
4534 2018-06-15 Jakub Jelinek <jakub@redhat.com>
4537 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
4538 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
4539 Only call store_expr for halves if the mode is the same.
4542 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
4543 Fix up comment formatting.
4545 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
4547 * typed-splay-tree.h (typed_splay_tree::remove): New function.
4548 (typed_splay_tree::closure,
4549 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
4550 (typed_splay_tree::typed_splay_tree,
4551 typed_splay_tree::operator =): Declared private.
4552 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
4553 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
4554 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
4555 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
4556 typed_splay_tree::splay_tree_splay,
4557 typed_splay_tree::splay_tree_foreach_helper,
4558 typed_splay_tree::splay_tree_insert,
4559 typed_splay_tree::splay_tree_remove,
4560 typed_splay_tree::splay_tree_lookup,
4561 typed_splay_tree::splay_tree_predecessor,
4562 typed_splay_tree::splay_tree_successor,
4563 typed_splay_tree::splay_tree_min,
4564 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
4565 (typed_splay_tree::root, typed_splay_tree::comp,
4566 typed_splay_tree::delete_key,
4567 typed_splay_tree::delete_value): New data members.
4568 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
4569 typed_splay_tree::remove.
4571 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
4573 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
4574 -mginv and -mno-ginv to the assembler.
4575 * config/mips/mips.opt (-mcrc): New option.
4577 * doc/invoke.text (-mcrc): Document.
4580 2018-06-15 Nick Clifton <nickc@redhat.com>
4583 * tree.c (escaped_string): New class. Converts an unescaped
4584 string into its escaped equivalent.
4585 (warn_deprecated_use): Use the new class to convert the
4586 deprecation message, if present.
4587 (test_escaped_strings): New self test.
4588 (test_c_tests): Add test_escaped_strings.
4589 * doc/extend.texi (deprecated): Add a note that the
4590 deprecation message is affected by the -fmessage-length
4591 option, and that control characters will be escaped.
4592 (#pragma GCC error): Document this pragma.
4593 (#pragma GCC warning): Likewise.
4594 * doc/invoke.texi (-fmessage-length): Document this option's
4595 effect on the #warning and #error preprocessor directives and
4596 the deprecated attribute.
4598 2018-06-15 Richard Biener <rguenther@suse.de>
4600 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
4601 here, also noting vector size used.
4602 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
4603 size used in MSG_OPTIMIZED_LOCATIONS dump.
4604 (pass_slp_vectorize::execute): Adjust.
4606 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
4609 * config/arc/arc.c (arc_return_address_register): Fix
4612 2018-06-15 Richard Biener <rguenther@suse.de>
4615 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
4616 leave useless conversion stripping to force_gimple_operand_gsi.
4617 (gimplify_build2): Likewise.
4618 (gimplify_build1): Likewise.
4620 2018-06-15 Richard Biener <rguenther@suse.de>
4623 * tree-cfg.c (move_stmt_op): unshare invariant addresses
4624 before adjusting their block.
4626 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
4628 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
4629 multilibs for *-*-rtems*.
4630 * config/riscv/t-rtems: New file.
4632 2018-06-14 Jakub Jelinek <jakub@redhat.com>
4635 * match.pd ((A +- CST1) +- CST2): Punt if last resort
4636 unsigned_type_for returns NULL.
4639 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
4640 subregs of multi-word pseudos unless the float mode has word size.
4642 2018-06-14 Richard Biener <rguenther@suse.de>
4645 * tree-vect-generic.c (build_word_mode_vector_type): Remove
4646 duplicate and harmful type_hash_canon.
4647 * tree.c (type_hash_canon): Assert we didn't find ourselves.
4649 2018-06-14 Richard Biener <rguenther@suse.de>
4652 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
4655 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
4657 * config/rtems.h (STDINT_LONG32): Define.
4659 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
4660 Prachi Godbole <prachi.godbole@imgtec.com>
4662 * config/mips/mips-cpus.def: Define P6600.
4663 * config/mips/mips-tables.opt: Regenerate.
4664 * config/mips/mips.c (mips_ucbranch_type): New enum.
4665 (mips_rtx_cost_data): Add support for P6600.
4666 (mips_issue_rate): Likewise.
4667 (mips_multipass_dfa_lookahead): Likewise.
4668 (mips_avoid_hazard): Likewise.
4669 (mips_reorg_process_insns): Likewise.
4670 (mips_classify_branch_p6600): New function.
4671 * config/mips/mips.h (TUNE_P6600): New define.
4672 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
4673 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
4674 * config/mips/mips.md: Include p6600.md.
4675 (processor): Add p6600.
4676 * config/mips/p6600.md: New file.
4677 * doc/invoke.texi: Add p6600 to supported architectures.
4679 2018-06-13 Martin Sebor <msebor@redhat.com>
4681 PR tree-optimization/86114
4682 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
4684 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
4686 2018-06-13 Richard Biener <rguenther@suse.de>
4688 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
4689 Properly set vector type of the intermediate stmt.
4690 * tree-vect-stmts.c (vectorizable_operation): The destination
4691 var always has vectype_out type.
4693 2018-06-13 Jeff Law <law@redhat.com>
4695 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
4696 integer 0 for argument to print_rtl_with_bb.
4697 (rl78_reorg): Likewise.
4699 2018-06-13 David Malcolm <dmalcolm@redhat.com>
4701 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
4702 from rtx to rtx_insn *.
4703 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
4705 (add_sched_insns_for_speculation): Likewise for local "target",
4706 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
4707 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
4708 from rtx_insn ** to rtx_code_label **.
4709 (reorg_emit_nops): Likewise.
4710 (c6x_reorg): Likewise for local "call_labels".
4711 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
4713 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
4714 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
4715 the loops over LABEL_REFS.
4716 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
4718 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
4719 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
4720 (split_branches): Strengthen local "olabel" from rtx to
4721 rtx_insn *, adding a safe_as_a cast.
4722 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
4724 (add_insn_after): Likewise for first two params.
4725 (add_insn_before): Likewise.
4726 (remove_insn): Likewise for param.
4727 (emit_pattern_before_noloc): Likewise for second and third params.
4728 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
4729 (emit_call_insn_before_noloc): Likewise.
4730 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
4732 (emit_barrier_before): Likewise.
4733 (emit_label_before): Strengthen "label" param from "rtx" to
4734 "rtx_code_label *". Strengthen "before" param from "rtx" to
4736 (emit_insn_after_1): Strengthen "after" param from "rtx" to
4738 (emit_pattern_after_noloc): Likewise.
4739 (emit_insn_after_noloc): Likewise.
4740 (emit_jump_insn_after_noloc): Likewise.
4741 (emit_call_insn_after_noloc): Likewise.
4742 (emit_debug_insn_after_noloc): Likewise.
4743 (emit_barrier_after): Likewise.
4744 (emit_label_after): Likewise for both params.
4745 (emit_pattern_after_setloc): Likewise for "after" param. Convert
4746 "loc" param from "int" to "location_t".
4747 (emit_insn_after_setloc): Likewise.
4748 (emit_jump_insn_after_setloc): Likewise.
4749 (emit_call_insn_after_setloc): Likewise.
4750 (emit_debug_insn_after_setloc): Likewise.
4751 (emit_pattern_before_setloc): Likewise for "before" param. Convert
4752 "loc" param from "int" to "location_t".
4753 (emit_pattern_before): Convert NULL_RTX to NULL.
4754 (emit_insn_before_setloc): Convert "loc" param from "int" to
4756 (emit_jump_insn_before_setloc): Likewise.
4757 (emit_call_insn_before_setloc): Likewise.
4758 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
4759 rtx_insn *. Convert "loc" param from "int" to "location_t".
4760 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
4761 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
4762 Convert 3rd param from "int" to "location_t".
4763 (emit_barrier_before, emit_barrier_after, next_real_insn):
4764 Strengthen param from rtx to rtx_insn *.
4765 (emit_label_before): Strengthen 1st param from "rtx" to
4766 "rtx_code_label *". Strengthen 2nd param from "rtx" to
4768 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
4769 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
4770 Strengthen 2nd param from "rtx" to "rtx_insn *".
4771 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
4772 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
4773 Likewise. Convert 3rd param from "int" to "location_t".
4774 (emit_label_after): Strengthen 1st param from "rtx" to
4776 (next_real_insn, remove_insn): Strengthen param from "rtx" to
4778 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
4779 from "rtx" to "rtx_insn *".
4781 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
4783 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
4784 bodies streamed in with -Q.
4785 * dumpfile.c (dump_files): Add lto-stream-out dump file.
4786 * dumpfile.h (tree_dump_index): Add lto_stream_out.
4787 * gimple-streamer-out.c: Include gimple-pretty-print.h
4788 (output_bb): Dump stmts streamed.
4789 * lto-section-out.c: Include print-tree.h
4790 (lto_begin_section): Dump sections created.
4791 (lto_output_decl_index): Dump decl encoded.
4792 * lto-streamer-out.c: Include print-tree.h
4793 (create_output_block): Dump output block created.
4794 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
4795 (output_function): Dump function output.
4796 (output_constructor): Dump constructor streamed.
4797 (write_global_stream): Output indexes encoded.
4798 (produce_asm_for_decls): Dump streams encoded.
4799 * lto-streamer.c (streamer_dump_file): New global var.
4800 * lto-streamer.h (streamer_dump_file): Declare.
4801 * passes.c (ipa_write_summaries): Initialize streamer dump.
4802 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
4805 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
4808 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
4809 offsets for register save directives. Emit a second batch of save
4810 directives, if need be, when the function accesses prior frames.
4812 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4814 * config/arc/fpu.md (fmasf4): Force operand to register.
4815 (fnmasf4): Likewise.
4817 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4819 * config/arc/arc-protos.h (arc_pad_return): Remove.
4820 * config/arc/arc.c (machine_function): Remove force_short_suffix
4822 (arc_print_operand): Adjust printing of '&'.
4823 (arc_verify_short): Remove conditional printing of short suffix.
4824 (arc_final_prescan_insn): Remove reference to size_reason.
4825 (pad_return): New function.
4826 (arc_reorg): Call pad_return.
4827 (arc_pad_return): Remove.
4828 (arc_init_machine_status): Remove reference to force_short_suffix.
4829 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
4830 (attr length): When attribute iscompact is true force to 2
4831 regardless; in the case of maybe check if we want to force the
4832 instruction to have 4 bytes length.
4833 (nopv): Change it to generate 4 byte long nop as well.
4834 (blockage): New pattern.
4835 (simple_return): Remove call to arc_pad_return.
4836 (p_return_i): Likewise.
4838 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4840 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
4842 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4844 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
4847 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4849 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
4850 for ARC700 and ARCv2.
4852 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
4855 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
4856 operands[2] instead of operands[1].
4859 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
4861 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
4862 case, check whether the outer register overlaps an unallocatable
4863 register, not just whether it fits the required class.
4865 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
4867 * poly-int.h (can_div_trunc_p): Add new overload in which all values
4869 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
4870 (memrefs_conflict_p): Likewise.
4871 (init_alias_analysis): Likewise.
4872 * cfgexpand.c (expand_debug_expr): Likewise.
4873 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
4874 * cse.c (fold_rtx): Likewise.
4875 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
4876 * expr.c (emit_block_move_hints): Likewise.
4877 (clear_storage_hints, push_block, emit_push_insn): Likewise.
4878 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
4879 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
4880 (emit_group_store): Likewise.
4881 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
4882 to read the PRE/POST_MODIFY increment.
4883 * calls.c (store_one_arg): Use strip_offset.
4884 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
4886 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
4888 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
4889 (simplify_binary_operation_1): Extend CONST_INT handling to
4891 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
4892 than a HOST_WIDE_INT.
4893 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
4895 (adjust_mems, add_stores): Update accodingly.
4896 (vt_canonicalize_addr): Track polynomial offsets.
4897 (emit_note_insn_var_location): Likewise.
4898 (vt_add_function_parameter): Likewise.
4899 (vt_initialize): Likewise.
4901 2018-06-12 Jeff Law <law@redhat.com>
4903 * config.gcc (alpha*-*-freebsd*): Remove.
4904 * config/alpha/freebsd.h: Remove.
4906 2018-06-12 David Malcolm <dmalcolm@redhat.com>
4909 * spellcheck-tree.c (levenshtein_distance): Rename to...
4910 (get_edit_distance): ...this, and update for underlying renaming.
4911 * spellcheck-tree.h (levenshtein_distance): Rename to...
4912 (get_edit_distance): ...this.
4913 * spellcheck.c (levenshtein_distance): Rename to...
4914 (get_edit_distance): ...this. Convert from Levenshtein distance
4915 to Damerau-Levenshtein distance by supporting transpositions of
4916 adjacent characters. Rename "v1" to "v_next" and "v0" to
4918 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
4919 (selftest::test_edit_distance_unit_test_oneway): ...this, and
4920 update for underlying renaming.
4921 (selftest::levenshtein_distance_unit_test): Rename to...
4922 (selftest::test_get_edit_distance_unit): ...this, and update for
4923 underlying renaming.
4924 (selftest::test_find_closest_string): Add example from PR 69968
4925 where transposition helps
4926 (selftest::test_metric_conditions): Update for renaming.
4927 (selftest::test_metric_conditions): Likewise.
4928 (selftest::spellcheck_c_tests): Likewise.
4929 * spellcheck.h (levenshtein_distance): Rename both overloads to...
4930 (get_edit_distance): ...this.
4931 (best_match::consider): Update for renaming.
4933 2018-06-12 Martin Sebor <msebor@redhat.com>
4935 PR tree-optimization/85259
4936 * builtins.c (compute_objsize): Handle constant offsets.
4937 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
4938 true iff a warning has been issued.
4939 * gimple.h (gimple_nonartificial_location): New function.
4940 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
4941 gimple_nonartificial_location and handle -Wno-system-headers.
4942 (handle_builtin_stxncpy): Same.
4944 2018-06-12 Martin Sebor <msebor@redhat.com>
4947 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
4949 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
4951 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
4952 BUILTIN_VEC_XST entries for pointer to double and long long.
4954 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
4957 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
4959 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
4962 2018-06-12 Martin Liska <mliska@suse.cz>
4964 * doc/options.texi: Document IntegerRange.
4966 2018-06-12 Martin Liska <mliska@suse.cz>
4968 * config/i386/i386.opt: Make MPX-related options as Deprecated.
4969 * opt-functions.awk: Handle Deprecated flag.
4970 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
4972 (read_cmdline_option): Report warning for a deprecated option.
4973 * opts.h (struct cl_option): Add new field cl_deprecated.
4974 (CL_ERR_DEPRECATED): New.
4976 2018-06-12 Martin Liska <mliska@suse.cz>
4978 * doc/options.texi: Document Deprecated option flag.
4980 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4982 * config/arc/arc-arch.h (arc_extras): New enum.
4983 (arc_cpu_t):Add field extra.
4984 (arc_cpu_types): Consider the extras.
4985 * config/arc/arc-cpus.def: Add extras info.
4986 * config/arc/arc-opts.h (processor_type): Consider extra field.
4987 * config/arc/arc.c (arc_override_options): Handle extra field.
4989 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4991 * config/arc/arc-arch.h: Update ARC_OPTX macro.
4992 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
4994 * config/arc/arc.c (arc_init): Update pic warning.
4995 (irq_range): Update irq range parsing warnings.
4996 (arc_override_options): Update various warning messages.
4997 (arc_handle_aux_attribute): Likewise.
4999 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
5001 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
5003 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5005 * doc/sourcebuild.texi: Document usage of line number 0 in verify
5006 compiler messages directives.
5008 2018-06-12 Matthew Fortune <mfortune@gmail.com>
5010 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
5011 * config/mips/mips-tables.opt: Regenerate.
5012 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
5014 * doc/invoke.texi: Document -march=i6500.
5016 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
5018 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
5019 (i6400_gpmul): Add cpu_unit.
5020 (i6400_gpdiv): Likewise.
5021 (i6400_msa_add_d): Update reservations.
5022 (i6400_msa_int_add) Likewise.
5023 (i6400_msa_short_logic3) Likewise.
5024 (i6400_msa_short_logic2) Likewise.
5025 (i6400_msa_short_logic) Likewise.
5026 (i6400_msa_move) Likewise.
5027 (i6400_msa_cmp) Likewise.
5028 (i6400_msa_short_float2) Likewise.
5029 (i6400_msa_div_d) Likewise.
5030 (i6400_msa_long_logic1) Likewise.
5031 (i6400_msa_long_logic2) Likewise.
5032 (i6400_msa_mult) Likewise.
5033 (i6400_msa_long_float2) Likewise.
5034 (i6400_msa_long_float4) Likewise.
5035 (i6400_msa_long_float5) Likewise.
5036 (i6400_msa_long_float8) Likewise.
5037 (i6400_fpu_fadd): Include frint type.
5038 (i6400_fpu_store): New define_insn_reservation.
5039 (i6400_fpu_load): Likewise.
5040 (i6400_fpu_move): Likewise.
5041 (i6400_fpu_fcmp): Likewise.
5042 (i6400_fpu_fmadd): Likewise.
5043 (i6400_int_mult): Include imul3nc type and update reservation.
5044 (i6400_int_div): Include idiv3 type and update reservation.
5045 (i6400_int_load): Update to check type not move_type.
5046 (i6400_int_store): Likewise.
5047 (i6400_int_prefetch): Set zero latency.
5049 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
5051 * gcc.c: Document new %@{...} sequence.
5052 (LINK_COMMAND_SPEC): Use it for the -L switches.
5053 (cpp_unique_options): Use it for the -I switches.
5054 (at_file_argbuf): New global variable.
5055 (in_at_file): Likewise.
5056 (alloc_args): Create at_file_argbuf.
5057 (clear_args): Truncate at_file_argbuf.
5058 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
5059 (open_at_file): New function.
5060 (close_at_file): Likewise.
5061 (create_at_file): Delete.
5062 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
5065 (validate_switches_from_spec): Deal with %@{...} sequence.
5066 (validate_switches): Likewise.
5067 (driver::finalize): Call clear_args.
5069 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
5071 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
5073 2018-06-11 Martin Sebor <msebor@redhat.com>
5075 * doc/invoke.texi (-Wall): List -Wc++17-compat.
5076 (Wno-class-memaccess): Add @opindex.
5077 (Wno-templates, Wno-multiple-inheritance): Same.
5078 (Wno-virtual-inheritance, Wno-namespaces): Same.
5079 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
5080 (Wno-format-overflow, Wno-format-truncation): Same.
5081 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
5082 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
5083 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
5084 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
5085 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
5086 (Wno-misspelled-isr): Same.
5088 2018-06-11 Martin Sebor <msebor@redhat.com>
5090 * PR tree-optimization/86083
5091 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
5093 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
5095 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
5097 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
5100 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
5101 on the correct operand.
5102 (*movdi_internal64): Ditto.
5104 2018-06-11 Martin Liska <mliska@suse.cz>
5106 PR tree-optimization/86089
5107 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
5109 2018-06-11 Julia Koval <julia.koval@intel.com>
5111 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
5112 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
5113 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
5115 2018-06-11 Olivier Hainque <hainque@adacore.com>
5117 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
5118 for Ada with strict dwarf2.
5120 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
5123 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
5126 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
5128 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
5130 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
5132 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
5135 2018-06-08 Martin Liska <mliska@suse.cz>
5137 * tree-cfg.h (debug_function): Fix argument type to match
5140 2018-06-08 Martin Liska <mliska@suse.cz>
5142 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
5143 Remove usage of MPX-related (and removed) fields.
5144 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
5146 2018-06-08 David Malcolm <dmalcolm@redhat.com>
5148 * cfg.c (debug): Use TDF_NONE rather than 0.
5149 * cfghooks.c (debug): Likewise.
5150 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
5151 (struct dump_option_value_info): Convert to...
5152 (struct kv_pair): ...this template type.
5153 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
5155 (optinfo_verbosity_options): Likewise.
5156 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
5158 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
5159 than int for "optgroup_flags" param.
5160 (dump_generic_expr_loc): Use dump_flags_t rather than int for
5162 (dump_dec): Likewise.
5163 (dump_finish): Use TDF_NONE rather than 0.
5164 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
5165 rather than int for "optgroup_flags" param. Use TDF_NONE rather
5166 than 0. Update for change to option_ptr.
5167 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
5168 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
5169 0. Update for changes to optinfo_verbosity_options and
5171 (opt_info_switch_p): Convert optgroup_flags from int to
5173 (dump_basic_block): Use dump_flags_t rather than int
5174 for "dump_kind" param.
5175 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
5176 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
5177 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
5178 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
5179 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
5180 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
5181 TDF_NONE): Convert from macros to...
5182 (enum dump_flag): ...this new enum.
5183 (dump_flags_t): Update to use enum.
5184 (operator|, operator&, operator~, operator|=, operator&=):
5185 Implement for dump_flags_t.
5186 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
5187 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
5188 Convert from macros to...
5189 (enum optgroup_flag): ...this new enum.
5190 (optgroup_flags_t): New typedef.
5191 (operator|, operator|=): Implement for optgroup_flags_t.
5192 (struct dump_file_info): Convert field "alt_flags" to
5193 dump_flags_t. Convert field "optgroup_flags" to
5195 (dump_basic_block): Use dump_flags_t rather than int for param.
5196 (dump_generic_expr_loc): Likewise.
5197 (dump_dec): Likewise.
5198 (dump_register): Convert param "optgroup_flags" to
5200 (opt_info_enable_passes): Likewise.
5201 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
5203 * gimple-pretty-print.c (debug): Likewise.
5204 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
5205 (merged_store_group::apply_stores): Likewise.
5206 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
5207 * gimple.c (verify_gimple_pp): Likewise.
5208 * graphite-poly.c (print_pbb_body): Likewise.
5209 * passes.c (pass_manager::register_one_dump_file): Convert
5210 local "optgroup_flags" to optgroup_flags_t.
5211 * print-tree.c (print_node): Use TDF_NONE rather than 0.
5213 (debug_body): Likewise.
5214 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
5215 to optgroup_flags_t.
5216 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
5218 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
5219 (convert_mult_to_fma): Likewise.
5220 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
5221 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
5222 * tree-vect-data-refs.c (dump_lower_bound): Convert param
5223 "dump_kind" to dump_flags_t.
5225 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
5227 * config/rs6000/rs6000.c (min, max): Delete.
5229 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
5231 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
5232 -mabi=spe and -mabi=no-spe.
5234 2018-06-08 Martin Liska <mliska@suse.cz>
5236 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
5237 where we expect an existing summary.
5239 2018-06-08 Martin Liska <mliska@suse.cz>
5241 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
5242 * ipa-inline.h (estimate_edge_growth): Likewise.
5244 2018-06-08 Martin Liska <mliska@suse.cz>
5246 * cgraph.c (function_version_hasher::hash): Use
5247 cgraph_node::get_uid ().
5248 (function_version_hasher::equal):
5249 * cgraph.h (cgraph_node::get_uid): New method.
5250 * ipa-inline.c (update_caller_keys): Use
5251 cgraph_node::get_uid ().
5252 (update_callee_keys): Likewise.
5253 * ipa-utils.c (searchc): Likewise.
5254 (ipa_reduced_postorder): Likewise.
5255 * lto-cgraph.c (input_node): Likewise.
5256 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
5257 * symbol-summary.h (symtab_insertion): Likewise.
5258 (symtab_removal): Likewise.
5259 (symtab_duplication): Likewise.
5260 * tree-pretty-print.c (dump_function_header): Likewise.
5261 * tree-sra.c (convert_callers_for_node): Likewise.
5263 2018-06-08 Martin Liska <mliska@suse.cz>
5265 * cgraph.c (symbol_table::create_edge): Always assign a new
5267 (symbol_table::free_edge): Do not recycle numbers.
5268 * cgraph.h (cgraph_edge::get): New method.
5269 * symbol-summary.h (symtab_removal): Use it.
5270 (symtab_duplication): Likewise.
5271 (call_summary::hashable_uid): Remove.
5273 2018-06-08 Martin Liska <mliska@suse.cz>
5275 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
5276 (initialize_growth_caches): Remove.
5277 (free_growth_caches): Likewise.
5278 (do_estimate_edge_time): Use edge_growth_cache.
5279 (do_estimate_edge_size): Likewise.
5280 (do_estimate_edge_hints): Likewise.
5281 * ipa-inline.c (reset_edge_caches): Likewise.
5282 (recursive_inlining): Likewise.
5283 (inline_small_functions): Likewise.
5284 * ipa-inline.h (initialize_growth_caches): Remove.
5285 (estimate_edge_size): Likewise.
5286 (estimate_edge_time): Likewise.
5287 (estimate_edge_hints): Likewise.
5288 (reset_edge_growth_cache): Likewise.
5289 * symbol-summary.h (call_summary::remove): New method.
5291 2018-06-08 Martin Liska <mliska@suse.cz>
5293 * ipa-cp.c (class edge_clone_summary): New summary.
5294 (grow_edge_clone_vectors): Remove.
5295 (ipcp_edge_duplication_hook): Remove.
5296 (class edge_clone_summary_t): New call_summary class.
5297 (ipcp_edge_removal_hook): Remove.
5298 (edge_clone_summary_t::duplicate): New function.
5299 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
5300 (create_specialized_node): Likewise.
5301 (ipcp_driver): Initialize edge_clone_summaries and do not
5304 2018-06-08 Martin Liska <mliska@suse.cz>
5306 * symbol-summary.h (get): New function.
5307 (call_summary::m_initialize_when_cloning): New class member.
5309 2018-06-08 Martin Liska <mliska@suse.cz>
5311 * cgraph.c (cgraph_node::remove): Do not recycle uid.
5312 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
5313 (symbol_table::allocate_cgraph_symbol): Do not set uid.
5314 * passes.c (uid_hash_t): Record removed_nodes by their uids.
5315 (remove_cgraph_node_from_order): Use the removed_nodes set.
5316 (do_per_function_toporder): Likwise.
5317 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
5318 instead of summary_uid.
5319 (symtab_removal): Likewise.
5320 (symtab_duplication): Likewise.
5322 2018-06-08 Martin Liska <mliska@suse.cz>
5324 * ipa-cp.c (ipcp_store_bits_results): Use
5325 ipcp_transformation_sum.
5326 (ipcp_store_vr_results): Likewise.
5327 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
5329 (ipcp_transformation_initialize): ... this.
5330 (ipa_set_node_agg_value_chain):
5331 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
5332 (write_ipcp_transformation_info): Likewise.
5333 (read_ipcp_transformation_info): Likewise.
5334 (ipcp_update_bits): Likewise.
5335 (ipcp_update_vr): Likewise.
5336 (ipcp_transform_function): Likewise.
5337 * ipa-prop.h: Rename ipcp_transformation_summary to
5338 ipcp_transformation.
5339 (class ipcp_transformation_t): New function summary.
5340 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
5341 (ipa_get_agg_replacements_for_node): Likewise.
5343 2018-06-08 Martin Liska <mliska@suse.cz>
5345 * ipa-pure-const.c (struct funct_state_d): Do it class instead
5347 (class funct_state_summary_t): New function_summary class.
5348 (has_function_state): Remove.
5349 (get_function_state): Likewise.
5350 (set_function_state): Likewise.
5351 (add_new_function): Likewise.
5352 (funct_state_summary_t::insert): New function.
5353 (duplicate_node_data): Remove.
5354 (remove_node_data): Remove.
5355 (funct_state_summary_t::duplicate): New function.
5356 (register_hooks): Create new funct_state_summaries.
5357 (pure_const_generate_summary): Use it.
5358 (pure_const_write_summary): Likewise.
5359 (pure_const_read_summary): Likewise.
5360 (propagate_pure_const): Likewise.
5361 (propagate_nothrow): Likewise.
5362 (dump_malloc_lattice): Likewise.
5363 (propagate_malloc): Likewise.
5364 (execute): Do not register hooks, just remove summary
5366 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
5369 2018-06-08 Martin Liska <mliska@suse.cz>
5371 * ipa-reference.c (remove_node_data): Remove.
5372 (duplicate_node_data): Likewise.
5373 (class ipa_ref_var_info_summary_t): New class.
5374 (class ipa_ref_opt_summary_t): Likewise.
5375 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
5376 (get_reference_optimization_summary): Use
5377 ipa_ref_opt_sum_summaries.
5378 (set_reference_vars_info): Remove.
5379 (set_reference_optimization_summary): Likewise.
5380 (ipa_init): Create summaries.
5381 (init_function_info): Use function summary.
5382 (ipa_ref_opt_summary_t::duplicate): New function.
5383 (ipa_ref_opt_summary_t::remove): New function.
5384 (get_read_write_all_from_node): Fix GNU coding style.
5385 (propagate): Use function summary.
5386 (write_node_summary_p): Fix GNU coding style.
5387 (stream_out_bitmap): Likewise.
5388 (ipa_reference_read_optimization_summary): Use function summary.
5389 (ipa_reference_c_finalize): Do not release hooks.
5391 2018-06-08 Martin Liska <mliska@suse.cz>
5393 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
5394 (analyze_function_body): Extract multiple calls of get_create.
5395 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
5396 * ipa-inline.c (recursive_inlining): Use ::get method.
5397 * ipa-inline.h (estimate_edge_growth): Likewise.
5399 2018-06-08 Martin Liska <mliska@suse.cz>
5401 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
5403 (hsa_function_summary::hsa_function_summary): Use the new enum
5405 (hsa_gpu_implementation_p): Use hsa_summaries::get.
5406 * hsa-gen.c (hsa_get_host_function): Likewise.
5407 (get_brig_function_name): Likewise.
5408 * ipa-hsa.c (process_hsa_functions): Likewise.
5409 (ipa_hsa_write_summary): Likewise.
5410 * symbol-summary.h (symtab_duplication): Use ::get function/
5411 (get): New function.
5413 2018-06-08 Martin Liska <mliska@suse.cz>
5415 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
5417 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
5418 (hsa_register_kernel): Likewise.
5419 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
5420 * hsa-gen.c (hsa_get_host_function): Likewise.
5421 (get_brig_function_name): Likewise.
5422 (generate_hsa): Likewise.
5423 (pass_gen_hsail::execute): Likewise.
5424 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
5425 (devirtualization_time_bonus): Likewise.
5426 (ipcp_propagate_stage): Likewise.
5427 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
5428 (edge_set_predicate): Likewise.
5429 (evaluate_conditions_for_known_args): Likewise.
5430 (evaluate_properties_for_edge): Likewise.
5431 (ipa_fn_summary::reset): Likewise.
5432 (ipa_fn_summary_t::duplicate): Likewise.
5433 (dump_ipa_call_summary): Likewise.
5434 (ipa_dump_fn_summary): Likewise.
5435 (analyze_function_body): Likewise.
5436 (compute_fn_summary): Likewise.
5437 (estimate_edge_devirt_benefit): Likewise.
5438 (estimate_edge_size_and_time): Likewise.
5439 (estimate_calls_size_and_time): Likewise.
5440 (estimate_node_size_and_time): Likewise.
5441 (inline_update_callee_summaries): Likewise.
5442 (remap_edge_change_prob): Likewise.
5443 (remap_edge_summaries): Likewise.
5444 (ipa_merge_fn_summary_after_inlining): Likewise.
5445 (ipa_update_overall_fn_summary): Likewise.
5446 (read_ipa_call_summary): Likewise.
5447 (inline_read_section): Likewise.
5448 (write_ipa_call_summary): Likewise.
5449 (ipa_fn_summary_write): Likewise.
5450 (ipa_free_fn_summary): Likewise.
5451 * ipa-hsa.c (process_hsa_functions): Likewise.
5452 (ipa_hsa_write_summary): Likewise.
5453 (ipa_hsa_read_section): Likewise.
5454 * ipa-icf.c (sem_function::merge): Likewise.
5455 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
5456 (do_estimate_edge_time): Likewise.
5457 (estimate_size_after_inlining): Likewise.
5458 (estimate_growth): Likewise.
5459 (growth_likely_positive): Likewise.
5460 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
5461 (inline_call): Likewise.
5462 * ipa-inline.c (caller_growth_limits): Likewise.
5463 (can_inline_edge_p): Likewise.
5464 (can_inline_edge_by_limits_p): Likewise.
5465 (compute_uninlined_call_time): Likewise.
5466 (compute_inlined_call_time): Likewise.
5467 (want_inline_small_function_p): Likewise.
5468 (edge_badness): Likewise.
5469 (update_caller_keys): Likewise.
5470 (update_callee_keys): Likewise.
5471 (recursive_inlining): Likewise.
5472 (inline_small_functions): Likewise.
5473 (inline_to_all_callers_1): Likewise.
5474 (dump_overall_stats): Likewise.
5475 (early_inline_small_functions): Likewise.
5476 (early_inliner): Likewise.
5477 * ipa-inline.h (estimate_edge_growth): Likewise.
5478 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
5479 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
5480 * ipa-prop.h (IPA_NODE_REF): Likewise.
5481 (IPA_EDGE_REF): Likewise.
5482 * ipa-pure-const.c (malloc_candidate_p): Likewise.
5483 (propagate_malloc): Likewise.
5484 * ipa-split.c (execute_split_functions): Likewise.
5485 * symbol-summary.h: Rename get to get_create.
5487 (get_create): Likewise.
5488 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
5490 2018-06-08 Martin Liska <mliska@suse.cz>
5492 * symbol-summary.h (release): Move definition out of class
5494 (symtab_removal): Likewise.
5495 (symtab_duplication): Likewise.
5497 2018-06-08 Martin Liska <mliska@suse.cz>
5499 * symbol-summary.h (function_summary): Move constructor
5500 implementation out of class declaration.
5501 (release): Likewise.
5502 (symtab_insertion): Likewise.
5503 (symtab_removal): Likewise.
5504 (symtab_duplication): Likewise.
5507 2018-06-08 Martin Liska <mliska@suse.cz>
5509 * Makefile.in: Remove support for MPX (macros, related functions,
5510 fields in cgraph_node, ...).
5511 * builtin-types.def (BT_BND): Likewise.
5512 (BT_FN_BND_CONST_PTR): Likewise.
5513 (BT_FN_CONST_PTR_BND): Likewise.
5514 (BT_FN_VOID_PTR_BND): Likewise.
5515 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
5516 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
5517 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
5518 (expand_builtin_mempcpy_with_bounds): Likewise.
5519 (expand_builtin_memset_with_bounds): Likewise.
5520 (expand_builtin_memset_args): Likewise.
5521 (std_expand_builtin_va_start): Likewise.
5522 (expand_builtin): Likewise.
5523 (expand_builtin_with_bounds): Likewise.
5524 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
5525 (DEF_LIB_BUILTIN_CHKP): Likewise.
5526 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
5527 (DEF_CHKP_BUILTIN): Likewise.
5528 (BUILT_IN_MEMCPY): Likewise.
5529 (BUILT_IN_MEMMOVE): Likewise.
5530 (BUILT_IN_MEMPCPY): Likewise.
5531 (BUILT_IN_MEMSET): Likewise.
5532 (BUILT_IN_STPCPY): Likewise.
5533 (BUILT_IN_STRCAT): Likewise.
5534 (BUILT_IN_STRCHR): Likewise.
5535 (BUILT_IN_STRCPY): Likewise.
5536 (BUILT_IN_STRLEN): Likewise.
5537 (BUILT_IN_MEMCPY_CHK): Likewise.
5538 (BUILT_IN_MEMMOVE_CHK): Likewise.
5539 (BUILT_IN_MEMPCPY_CHK): Likewise.
5540 (BUILT_IN_MEMSET_CHK): Likewise.
5541 (BUILT_IN_STPCPY_CHK): Likewise.
5542 (BUILT_IN_STRCAT_CHK): Likewise.
5543 (BUILT_IN_STRCPY_CHK): Likewise.
5544 * calls.c (store_bounds): Likewise.
5545 (emit_call_1): Likewise.
5546 (special_function_p): Likewise.
5547 (maybe_warn_nonstring_arg): Likewise.
5548 (initialize_argument_information): Likewise.
5549 (finalize_must_preallocate): Likewise.
5550 (compute_argument_addresses): Likewise.
5551 (expand_call): Likewise.
5552 * cfgexpand.c (expand_call_stmt): Likewise.
5553 (expand_return): Likewise.
5554 (expand_gimple_stmt_1): Likewise.
5555 (pass_expand::execute): Likewise.
5556 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
5557 (cgraph_node::remove): Likewise.
5558 (cgraph_node::dump): Likewise.
5559 (cgraph_node::verify_node): Likewise.
5560 * cgraph.h (chkp_function_instrumented_p): Likewise.
5561 (symtab_node::get_alias_target): Likewise.
5562 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
5563 (cgraph_local_p): Likewise.
5564 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
5565 (cgraph_edge::rebuild_references): Likewise.
5566 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
5567 (walk_polymorphic_call_targets): Likewise.
5568 (cgraph_node::expand_thunk): Likewise.
5569 (symbol_table::output_weakrefs): Likewise.
5570 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
5571 (ix86_handle_option): Likewise.
5572 * config/i386/constraints.md: Likewise.
5573 * config/i386/i386-builtin-types.def (BND): Likewise.
5577 * config/i386/i386-builtin.def (BDESC_END): Likewise.
5578 (BDESC_FIRST): Likewise.
5580 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
5581 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
5582 * config/i386/i386.c (enum reg_class): Likewise.
5583 (ix86_target_string): Likewise.
5584 (ix86_option_override_internal): Likewise.
5585 (ix86_conditional_register_usage): Likewise.
5586 (ix86_valid_target_attribute_inner_p): Likewise.
5587 (ix86_set_indirect_branch_type): Likewise.
5588 (ix86_set_current_function): Likewise.
5589 (ix86_function_arg_regno_p): Likewise.
5590 (init_cumulative_args): Likewise.
5591 (ix86_function_arg_advance): Likewise.
5592 (ix86_function_arg): Likewise.
5593 (ix86_pass_by_reference): Likewise.
5594 (ix86_function_value_regno_p): Likewise.
5595 (ix86_function_value_1): Likewise.
5596 (ix86_function_value_bounds): Likewise.
5597 (ix86_return_in_memory): Likewise.
5598 (ix86_setup_incoming_vararg_bounds): Likewise.
5599 (ix86_va_start): Likewise.
5600 (indirect_thunk_need_prefix): Likewise.
5601 (print_reg): Likewise.
5602 (ix86_print_operand): Likewise.
5603 (ix86_expand_call): Likewise.
5604 (ix86_output_function_return): Likewise.
5605 (reg_encoded_number): Likewise.
5606 (BDESC_VERIFYS): Likewise.
5607 (ix86_init_mpx_builtins): Likewise.
5608 (ix86_init_builtins): Likewise.
5609 (ix86_emit_cmove): Likewise.
5610 (ix86_emit_move_max): Likewise.
5611 (ix86_expand_builtin): Likewise.
5612 (ix86_builtin_mpx_function): Likewise.
5613 (ix86_get_arg_address_for_bt): Likewise.
5614 (ix86_load_bounds): Likewise.
5615 (ix86_store_bounds): Likewise.
5616 (ix86_load_returned_bounds): Likewise.
5617 (ix86_store_returned_bounds): Likewise.
5618 (ix86_class_likely_spilled_p): Likewise.
5619 (ix86_hard_regno_mode_ok): Likewise.
5620 (x86_order_regs_for_local_alloc): Likewise.
5621 (ix86_mitigate_rop): Likewise.
5622 (ix86_bnd_prefixed_insn_p): Likewise.
5623 (ix86_mpx_bound_mode): Likewise.
5624 (ix86_make_bounds_constant): Likewise.
5625 (ix86_initialize_bounds): Likewise.
5626 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
5627 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
5628 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
5629 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
5630 (TARGET_CHKP_BOUND_MODE): Likewise.
5631 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
5632 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
5633 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
5634 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
5635 * config/i386/i386.h (TARGET_MPX): Likewise.
5636 (TARGET_MPX_P): Likewise.
5637 (VALID_BND_REG_MODE): Likewise.
5638 (FIRST_BND_REG): Likewise.
5639 (LAST_BND_REG): Likewise.
5640 (enum reg_class): Likewise.
5641 (BND_REG_P): Likewise.
5642 (BND_REGNO_P): Likewise.
5643 (BNDmode): Likewise.
5644 (ADJUST_INSN_LENGTH): Likewise.
5645 * config/i386/i386.md: Likewise.
5646 * config/i386/i386.opt: Likewise.
5647 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
5648 (defined): Likewise.
5649 (LINK_MPX): Likewise.
5650 (MPX_SPEC): Likewise.
5651 (LIBMPX_SPEC): Likewise.
5652 (LIBMPXWRAPPERS_SPEC): Likewise.
5653 (CHKP_SPEC): Likewise.
5654 * config/i386/predicates.md: Likewise.
5655 * dbxout.c (dbxout_type): Likewise.
5656 * doc/extend.texi: Likewise.
5657 * doc/invoke.texi: Likewise.
5658 * doc/md.texi: Likewise.
5659 * doc/tm.texi: Likewise.
5660 * doc/tm.texi.in: Likewise.
5661 * dwarf2out.c (is_base_type): Likewise.
5662 (gen_formal_types_die): Likewise.
5663 (gen_subprogram_die): Likewise.
5664 (gen_type_die_with_usage): Likewise.
5665 (gen_decl_die): Likewise.
5666 (dwarf2out_late_global_decl): Likewise.
5667 * expr.c (expand_assignment): Likewise.
5668 (emit_storent_insn): Likewise.
5669 (store_expr_with_bounds): Likewise.
5670 (store_expr): Likewise.
5671 (expand_expr_real_1): Likewise.
5672 * expr.h (store_expr_with_bounds): Likewise.
5673 * function.c (use_register_for_decl): Likewise.
5674 (struct bounds_parm_data): Likewise.
5675 (assign_parms_augmented_arg_list): Likewise.
5676 (assign_parm_find_entry_rtl): Likewise.
5677 (assign_parm_is_stack_parm): Likewise.
5678 (assign_parm_load_bounds): Likewise.
5679 (assign_bounds): Likewise.
5680 (assign_parms): Likewise.
5681 (expand_function_start): Likewise.
5682 * gcc.c (CHKP_SPEC): Likewise.
5683 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
5684 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
5685 (wrestrict_dom_walker::check_call): Likewise.
5686 * gimple.c (gimple_build_call_from_tree): Likewise.
5687 * gimple.h (enum gf_mask): Likewise.
5688 (gimple_call_with_bounds_p): Likewise.
5689 (gimple_call_set_with_bounds): Likewise.
5690 * gimplify.c (gimplify_init_constructor): Likewise.
5691 * ipa-cp.c (initialize_node_lattices): Likewise.
5692 (propagate_constants_across_call): Likewise.
5693 (find_more_scalar_values_for_callers_subset): Likewise.
5694 * ipa-hsa.c (process_hsa_functions): Likewise.
5695 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
5696 * ipa-icf.c (sem_function::merge): Likewise.
5697 * ipa-inline.c (early_inliner): Likewise.
5698 * ipa-pure-const.c (warn_function_noreturn): Likewise.
5699 (warn_function_cold): Likewise.
5700 (propagate_pure_const): Likewise.
5701 * ipa-ref.h (enum GTY): Likewise.
5702 * ipa-split.c (find_retbnd): Likewise.
5703 (consider_split): Likewise.
5704 (split_function): Likewise.
5705 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
5706 * ipa.c (walk_polymorphic_call_targets): Likewise.
5707 (symbol_table::remove_unreachable_nodes): Likewise.
5708 (process_references): Likewise.
5709 (cgraph_build_static_cdtor_1): Likewise.
5710 * lto-cgraph.c (lto_output_node): Likewise.
5711 (output_refs): Likewise.
5712 (compute_ltrans_boundary): Likewise.
5713 (input_overwrite_node): Likewise.
5714 (input_node): Likewise.
5715 (input_cgraph_1): Likewise.
5716 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
5717 * passes.c (pass_manager::execute_early_local_passes): Likewise.
5718 (class pass_chkp_instrumentation_passes): Likewise.
5719 (make_pass_chkp_instrumentation_passes): Likewise.
5720 * passes.def: Likewise.
5721 * rtl.h (struct GTY): Likewise.
5722 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
5723 * stor-layout.c (layout_type): Likewise.
5724 * symtab.c: Likewise.
5725 * target.def: Likewise.
5726 * targhooks.c (default_chkp_bound_type): Likewise.
5727 (default_chkp_bound_mode): Likewise.
5728 (default_builtin_chkp_function): Likewise.
5729 (default_chkp_function_value_bounds): Likewise.
5730 (default_chkp_make_bounds_constant): Likewise.
5731 (default_chkp_initialize_bounds): Likewise.
5732 * targhooks.h (default_chkp_bound_type): Likewise.
5733 (default_chkp_bound_mode): Likewise.
5734 (default_builtin_chkp_function): Likewise.
5735 (default_chkp_function_value_bounds): Likewise.
5736 (default_chkp_make_bounds_constant): Likewise.
5737 (default_chkp_initialize_bounds): Likewise.
5738 * toplev.c (compile_file): Likewise.
5739 (process_options): Likewise.
5740 * tree-core.h (DEF_BUILTIN): Likewise.
5741 (DEF_BUILTIN_CHKP): Likewise.
5742 * tree-inline.c (declare_return_variable): Likewise.
5743 (remap_gimple_stmt): Likewise.
5744 (copy_bb): Likewise.
5745 (initialize_inlined_parameters): Likewise.
5746 (expand_call_inline): Likewise.
5747 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
5748 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
5749 (make_pass_ipa_chkp_produce_thunks): Likewise.
5750 (make_pass_chkp): Likewise.
5751 (make_pass_chkp_opt): Likewise.
5752 (make_pass_chkp_instrumentation_passes): Likewise.
5753 * tree-pretty-print.c (dump_generic_node): Likewise.
5754 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
5755 * tree-ssa-dce.c (propagate_necessity): Likewise.
5756 (eliminate_unnecessary_stmts): Likewise.
5757 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
5758 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
5759 * tree-ssa-sccvn.h: Likewise.
5760 * tree-ssa-strlen.c (get_string_length): Likewise.
5761 (valid_builtin_call): Likewise.
5762 (adjust_last_stmt): Likewise.
5763 (handle_builtin_strchr): Likewise.
5764 (handle_builtin_strcpy): Likewise.
5765 (handle_builtin_stxncpy): Likewise.
5766 (handle_builtin_memcpy): Likewise.
5767 (handle_builtin_strcat): Likewise.
5768 (strlen_check_and_optimize_stmt): Likewise.
5769 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
5770 * tree-streamer-in.c: Likewise.
5771 * tree-streamer.c (record_common_node): Likewise.
5772 * tree.c (tree_code_size): Likewise.
5773 (wide_int_to_tree_1): Likewise.
5774 (type_contains_placeholder_1): Likewise.
5775 (build_common_tree_nodes): Likewise.
5776 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
5777 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
5778 (POINTER_BOUNDS_P): Likewise.
5779 (BOUNDED_TYPE_P): Likewise.
5780 (BOUNDED_P): Likewise.
5781 (CALL_WITH_BOUNDS_P): Likewise.
5782 (pointer_bounds_type_node): Likewise.
5783 * value-prof.c (gimple_ic): Likewise.
5784 * var-tracking.c (vt_add_function_parameters): Likewise.
5785 * varasm.c (make_decl_rtl): Likewise.
5786 (assemble_start_function): Likewise.
5787 (output_constant): Likewise.
5788 (maybe_assemble_visibility): Likewise.
5789 * varpool.c (ctor_for_folding): Likewise.
5790 * chkp-builtins.def: Remove.
5791 * ipa-chkp.c: Remove.
5792 * ipa-chkp.h: Remove.
5793 * rtl-chkp.c: Remove.
5794 * rtl-chkp.h: Remove.
5795 * tree-chkp-opt.c: Remove.
5796 * tree-chkp.c: Remove.
5797 * tree-chkp.h: Remove.
5799 2018-06-07 Carl Love <cel@us.ibm.com>
5801 * config/rs6000/vsx.md (vextract_fp_from_shorth,
5802 vextract_fp_from_shortl): Add BE support.
5804 2018-06-07 Paul Koning <ni1d@arrl.net>
5806 * compare-elim.c (try_merge_compare): Don't merge compare if
5807 address contains a side effect.
5808 (try_eliminate_compare): Likewise.
5810 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
5812 * config.gcc: Support "tremont".
5813 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
5814 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5816 * config/i386/i386.c (m_TREMONT): Define.
5817 (processor_target_table): Add "tremont".
5818 (PTA_TREMONT): Define.
5819 (ix86_lea_outperforms): Add TARGET_TREMONT.
5820 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
5821 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
5822 and M_INTEL_GOLDMONT_PLUS.
5823 (fold_builtin_cpu): Add "tremont".
5824 (ix86_add_stmt_cost): Add TARGET_TREMONT.
5825 (ix86_option_override_internal): Add "tremont".
5826 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
5827 (processor_type): Add PROCESSOR_TREMONT.
5828 * config/i386/x86-tune.def: Add m_TREMONT.
5829 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
5831 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5833 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
5834 symbol defined for msp430i* devices to be lower case.
5836 2018-06-07 Richard Biener <rguenther@suse.de>
5838 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
5839 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
5840 Properly wrap signed arithmetic if overflow wraps.
5842 2018-06-07 Jakub Jelinek <jakub@redhat.com>
5844 PR tree-optimization/69615
5845 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
5846 of a cast from a same precision integral SSA_NAME in a bb dominated
5847 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
5848 cast to utype if rhs2 has already a compatible type.
5850 2018-06-07 Richard Biener <rguenther@suse.de>
5852 PR tree-optimization/85935
5853 * graphite-scop-detection.c (find_params_in_bb): Analyze
5854 condition operands with respect to the correct loop. Assert
5855 the analysis doesn't fail.
5857 2018-06-04 Carl Love <cel@us.ibm.com>
5859 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
5860 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
5861 as it is slightly cheaper.
5862 (first_match_or_eos_index_<mode>):
5863 Calculate index using natural element order.
5864 (first_match_index_<mode>):
5865 Calculate index using natural element order.
5866 (first_match_or_eos_index_<mode>):
5867 Calculate index using natural order.
5868 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
5869 for BE and LE modes.
5870 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
5871 P9V_BUILTIN_VCLZLSBB_V16QI.
5872 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
5875 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
5877 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
5878 indentation and line wrap for many prototypes. Add missing
5879 @smallexample directives around block of prototypes for vec_xl and
5882 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
5884 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
5885 track if we pass or return IEEE 128-bit floating point.
5886 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
5887 C++ mangling that is compatible with GCC 8.1.
5888 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
5889 (init_cumulative_args): Note if we pass or return IEEE 128-bit
5890 floating point types.
5891 (rs6000_function_arg_advance_1): Likewise.
5892 (rs6000_mangle_type): Optionally generate mangled names that match
5893 what GCC 8.1 generated for IEEE 128-bit floating point types.
5894 (rs6000_globalize_decl_name): If we have an external function that
5895 passes or returns IEEE 128-bit types, generate a weak reference
5896 from the mangled name used in GCC 8.1 to the current mangled
5898 (rs6000_init_builtins): Make __ibm128 use the long double type if
5899 long double is IBM extended double. Make __float128 use the long
5900 double type if long double is IEEE 128-bit.
5903 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
5904 macro for __ibm128 built-in functions.
5905 (PACK_IF): Add __ibm128 pack/unpack functions.
5906 (UNPACK_IF): Likewise.
5907 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
5908 enable long double built-in functions if long double is IEEE
5909 128-bit floating point.
5910 (rs6000_invalid_builtin): Update long double built-in function
5912 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
5913 functions, adjust the built-in function to use the long double
5914 built-in function if __ibm128 and long double are the same type.
5915 * doc/extend.texi (PowerPC builtins): Update documention for
5916 __builtin_{,un}pack_longdouble. Add documentation for
5917 __builtin_{,un}pack_ibm128.
5919 2018-06-06 Jim Wilson <jimw@sifive.com>
5921 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
5922 (struct machine_function): New field interrupt_mode.
5923 (riscv_handle_type_attribute): New function. Add forward declaration.
5924 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
5925 (riscv_expand_epilogue): Check interrupt_mode field.
5926 (riscv_set_current_function): Check interrupt attribute args and
5927 set interrupt_mode field.
5928 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
5929 (riscv_sret, riscv_uret): New.
5930 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
5931 new arguments to interrupt attribute.
5933 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
5936 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
5937 Don't handle -mcpu=power8 if -mpower9-vector is also used.
5939 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
5941 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
5942 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
5943 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
5944 several redundant entries.
5946 2018-06-06 David Malcolm <dmalcolm@redhat.com>
5948 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
5949 type from "rtx" to "rtx_insn *".
5950 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
5951 for local "call_insn", removing cast.
5952 (ix86_expand_call): Likewise, introducing a "call_insn" local.
5954 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
5956 PR tree-optimization/86066
5957 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
5958 for BIT_INSERT_EXPR stores.
5960 2018-06-06 Richard Biener <rguenther@suse.de>
5962 PR tree-optimization/86062
5963 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
5964 component refs ontop
5965 of to be offsetted base.
5967 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5969 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
5970 to be static and remove check on interrupt attribute name.
5972 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
5974 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
5975 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
5977 2018-06-05 Steve Ellcey <sellcey@cavium.com>
5980 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
5982 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
5983 Remove second argument, change how error is called.
5984 (aarch64_layout_arg): Remove second argument from
5985 aarch64_err_no_fpadvsimd call.
5986 (aarch64_init_cumulative_args): Ditto.
5987 (aarch64_gimplify_va_arg_expr): Ditto.
5988 * config/aarch64/aarch64.md (mov<mode>): Ditto.
5990 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
5992 * config/i386/i386.md (simple_return_indirect_internal): New expander.
5993 (*simple_return_indirect_internal<mode>): Rename from
5994 simple_return_indirect_internal. Use W mode iterator.
5995 (rstorssp): New expander.
5996 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
5997 (clrssbsy): New expander.
5998 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
6000 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
6002 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
6004 (cmse_check_pointed_object): Likewise.
6006 2018-06-05 Martin Liska <mliska@suse.cz>
6008 PR gcov-profile/47618
6009 * doc/invoke.texi: Document how -fprofile-dir format
6012 2018-06-05 Richard Biener <rguenther@suse.de>
6014 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
6015 removal pretend DOM info isn't available so we do not update
6016 it and only remove edges, not dominated blocks. Actually free
6017 DOM info in case we removed something. Remove unreachable blocks.
6018 (mfb_keep_latches): Work with either DOM info or marked backedges.
6019 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
6020 first. Mark backedges if DOM info isn't available.
6021 (Re-)compute DOM info after cleanup_control_flow_pre.
6023 2018-06-05 Richard Biener <rguenther@suse.de>
6025 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
6026 (locus_discrim_hasher::hash): Adjust.
6027 (locus_discrim_hasher::equal): Likewise.
6028 (next_discriminator_for_locus): Work on line directly.
6029 (same_line_p): Pass in expanded locus1 as well.
6030 (assign_discriminators): Avoid redundant location expansions.
6032 2018-06-05 Richard Biener <rguenther@suse.de>
6034 PR tree-optimization/86046
6035 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
6036 if required after clearing TREE_ADDRESSABLE.
6038 2018-06-05 Richard Biener <rguenther@suse.de>
6040 PR tree-optimization/86047
6041 * tree-ssa-loop.c (for_each_index): Glob handling of all
6042 decls and constants and really handle all of them.
6044 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6047 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
6048 qualifier_void_pointer and qualifier_const_void_pointer.
6049 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
6050 (arm_init_builtins): Handle the above.
6051 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
6052 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
6053 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
6056 2018-06-05 Martin Liska <mliska@suse.cz>
6058 * auto-profile.c (read_autofdo_file): Do not use
6059 gcov_ctr_summary struct.
6060 (afdo_callsite_hot_enough_for_early_inline): Likewise.
6061 * coverage.c (struct counts_entry): Likewise.
6062 (read_counts_file): Read just single summary entry.
6063 (get_coverage_counts): Use gcov_summary struct.
6064 * coverage.h (get_coverage_counts): Likewise.
6065 * gcov-dump.c (dump_working_sets): Likewise.
6066 (tag_summary): Dump just single summary.
6067 * gcov-io.c (gcov_write_summary): Write just histogram
6069 (gcov_read_summary): Read just single summary.
6070 (compute_working_sets): Use gcov_summary struct.
6071 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
6072 of GCOV_COUNTERS_SUMMABLE.
6073 (GCOV_COUNTERS_SUMMABLE): Remove.
6074 (GCOV_FIRST_VALUE_COUNTER): Replace with
6075 GCOV_COUNTER_V_INTERVAL.
6076 (struct gcov_ctr_summary): Remove.
6077 (struct gcov_summary): Directly use fields of former
6079 (compute_working_sets): Use gcov_summary struct.
6080 * gcov.c (read_count_file): Do not use ctrs fields.
6081 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
6083 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
6085 * profile.c: Likewise.
6086 * profile.h: Likewise.
6088 2018-06-05 Martin Liska <mliska@suse.cz>
6090 PR gcov-profile/84846
6091 * gcov.c (output_lines): Print working directory only
6092 in intermediate format.
6094 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
6096 * config/s390/s390-builtin-types.def: Add void function type.
6097 * config/s390/s390-builtins.def: Use the function type for the
6100 2018-06-04 Jim Wilson <jimw@sifive.com>
6102 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
6104 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
6105 and maybe_eh_return. Change regno to unsigned int. Use new args to
6106 handle EH_RETURN_DATA_REGNO registers properly.
6107 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
6108 (riscv_expand_epilogue): Update comment. Change argument name and
6109 type. Update code to use new name and type. Pass new args to
6110 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
6112 * config/riscv/riscv.md (NORMAL_RETURN): New.
6113 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
6114 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
6115 (eh_return): Call gen_eh_return_internal and emit barrier.
6116 (eh_return_internal): Call riscv_expand_epilogue.
6118 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
6120 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
6121 bit_insertion field and declare can_be_merged_into method.
6122 (merged_store_group::can_be_merged_into): New method.
6123 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
6124 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
6125 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
6127 2018-06-04 Richard Biener <rguenther@suse.de>
6129 PR tree-optimization/85955
6130 * builtins.c (fold_builtin_sincos): Convert pointers to
6131 destination to appropriate type before dereferencing.
6133 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
6135 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
6137 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
6139 * expr.c (expand_expr_real_1): Force the operand into memory if
6140 its TYPE_MODE is BLKmode and if there is no integer mode for
6141 the number of bits being extracted.
6143 2018-06-04 Jakub Jelinek <jakub@redhat.com>
6147 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
6148 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
6150 2018-06-04 Richard Biener <rguenther@suse.de>
6152 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
6153 (cleanup_tree_cfg_noloop): ... single caller. Do
6154 start_recording_case_labels later.
6156 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
6158 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
6159 to _IMMINTRIN_H_INCLUDED.
6160 * config/i386/pconfigintrin.h: Ditto.
6161 * config/i386/waitpkgintrin.h: Ditto.
6162 * config/i386/immintrin.h: Add includes for sgxintrin.h,
6163 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
6164 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
6165 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
6166 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
6167 waitpkgintrin.h and cldemoteintrin.h.
6169 2018-06-04 Richard Biener <rguenther@suse.de>
6171 PR tree-optimization/86038
6172 * tracer.c (find_best_successor): Check probability for
6173 being initialized, bail out if not.
6175 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
6178 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
6179 of bits to ignore when comparing architectures.
6181 2018-06-04 Jakub Jelinek <jakub@redhat.com>
6183 PR tree-optimization/69615
6184 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
6185 maximum or minimum of the type, try to merge it also as if
6186 range1 is + [-, x - 1] or + [x + 1, -].
6189 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
6191 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
6193 PR tree-optimization/86034
6194 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
6195 the unsigned bitfield type in a bit insertion sequence if it does not
6196 have a larger precision than the bitfield size.
6197 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
6199 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
6201 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
6203 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
6205 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
6206 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
6207 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
6208 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
6210 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
6212 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
6213 Disable -fdelete-null-pointer-checks for ELF toolchain.
6215 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
6216 Kito Cheng <kito.cheng@gmail.com>
6218 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
6219 (nds32le-*-*, nds32be-*-*): Integrate checking process.
6220 (nds32*-*-*): Add glibc and uclibc conditions.
6221 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
6222 (TARGET_EXCEPT_UNWIND_INFO): Define.
6223 * config/nds32/elf.h: New file.
6224 * config/nds32/linux.h: New file.
6225 * config/nds32/nds32-elf.opt: New file.
6226 * config/nds32/nds32-linux.opt: New file.
6227 * config/nds32/nds32-fp-as-gp.c
6228 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
6229 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
6231 (nds32_asm_file_end): Ditto.
6232 (nds32_print_operand): Ditto.
6233 (nds32_insert_attributes): Ditto.
6234 (nds32_init_libfuncs): New function.
6235 (TARGET_HAVE_TLS): Define.
6236 (TARGET_INIT_LIBFUNCS): Define.
6237 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
6239 (TARGET_ELF): Apply different mcmodel setting.
6240 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
6241 been migrated into elf.h and linux.h files.
6242 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
6243 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
6244 (mcmodel): The content has been migrated into nds32-elf.opt and
6245 nds32-linux.opt files.
6246 * config/nds32/t-elf: New file.
6247 * config/nds32/t-linux: New file.
6249 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
6250 Shiva Chen <shiva0217@gmail.com>
6252 * config/nds32/constants.md (unspec_volatile_element): Add
6253 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
6254 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
6256 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
6257 (make_pass_nds32_fp_as_gp): Declare.
6258 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
6260 (nds32_asm_function_end_prologue): Remove unused asm output.
6261 (nds32_asm_function_begin_epilogue): Remove unused asm output.
6262 (nds32_asm_file_start): Output necessary fp_as_gp information.
6263 (nds32_option_override): Adjust register usage.
6264 (nds32_expand_prologue): Consider fp_as_gp situation.
6265 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
6266 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
6269 (simple_return): Ditto.
6270 (omit_fp_begin): Output special directive for fp_as_gp.
6271 (omit_fp_end): Output special directive for fp_as_gp.
6272 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
6273 mforbid-fp-as-gp): New options.
6275 2018-06-01 Mark Wielaard <mark@klomp.org>
6277 * dwarf2out.c (dwarf2out_finish): Remove generation of
6278 DW_AT_loclists_base.
6280 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
6282 * gimple-ssa-store-merging.c: Include gimple-fold.h.
6283 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
6284 (struct merged_store_group): Add bit_insertion field.
6285 (dump_char_array): Use standard hexadecimal format.
6286 (merged_store_group::merged_store_group): Set bit_insertion to false.
6287 (merged_store_group::apply_stores): Use optimal buffer size. Deal
6288 with BIT_INSERT_EXPR stores. Move up code updating the mask and
6289 also print the mask in the dump file.
6290 (pass_store_merging::gate): Minor tweak.
6291 (imm_store_chain_info::coalesce_immediate): Fix wrong association
6292 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
6293 stores with INTEGER_CST stores.
6294 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
6295 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
6296 and use it throughout. Generate bit insertion sequences if need be.
6297 (pass_store_merging::process_store): Remove redundant condition.
6298 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
6300 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
6302 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
6303 the 128-bit floating point types. Fix function comment.
6305 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6307 * config/aarch64/aarch64-simd.md
6308 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
6310 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
6313 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
6315 PR tree-optimization/85989
6316 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
6318 (backprop::intersect_uses): Check it when deciding whether this
6319 is a backedge reference.
6320 (backprop::process_block): Add each phi to m_visited_phis
6321 after visiting it, then clear it at the end.
6323 2018-06-01 Richard Biener <rguenther@suse.de>
6325 * tree-vectorizer.h (vect_dr_stmt): New function.
6326 (vect_get_load_cost): Adjust.
6327 (vect_get_store_cost): Likewise.
6328 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
6329 Use vect_dr_stmt instead of DR_SMTT.
6330 (vect_record_base_alignments): Likewise.
6331 (vect_calculate_target_alignment): Likewise.
6332 (vect_compute_data_ref_alignment): Likewise and make static.
6333 (vect_update_misalignment_for_peel): Likewise.
6334 (vect_verify_datarefs_alignment): Likewise.
6335 (vector_alignment_reachable_p): Likewise.
6336 (vect_get_data_access_cost): Likewise. Pass down
6337 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
6338 (vect_get_peeling_costs_all_drs): Likewise.
6339 (vect_peeling_hash_get_lowest_cost): Likewise.
6340 (vect_enhance_data_refs_alignment): Likewise.
6341 (vect_find_same_alignment_drs): Likewise.
6342 (vect_analyze_data_refs_alignment): Likewise.
6343 (vect_analyze_group_access_1): Likewise.
6344 (vect_analyze_group_access): Likewise.
6345 (vect_analyze_data_ref_access): Likewise.
6346 (vect_analyze_data_ref_accesses): Likewise.
6347 (vect_vfa_segment_size): Likewise.
6348 (vect_small_gap_p): Likewise.
6349 (vectorizable_with_step_bound_p): Likewise.
6350 (vect_prune_runtime_alias_test_list): Likewise.
6351 (vect_analyze_data_refs): Likewise.
6352 (vect_supportable_dr_alignment): Likewise.
6353 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
6354 (vect_gen_prolog_loop_niters): Likewise.
6355 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
6356 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
6358 (vect_recog_mask_conversion_pattern): Likewise.
6359 (vect_try_gather_scatter_pattern): Likewise.
6360 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
6361 to vect_get_store_cost.
6362 (vect_get_store_cost): Get stmt_info instead of DR.
6363 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
6364 (vect_get_load_cost): Get stmt_info instead of DR.
6366 2018-06-01 Richard Biener <rguenther@suse.de>
6369 * gimple-fold.c (var_decl_component_p): Also allow offsetted
6370 vars wrapped in MEM_REFs.
6372 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
6374 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
6375 Fix subreg tests so that we only return a choice between
6376 GENERAL_REGS and FP_REGS if the original classes included both.
6378 2018-06-01 Richard Biener <rguenther@suse.de>
6381 * tree-ssa-structalias.c (get_function_part_constraint):
6382 Handle NULL fi->decl.
6383 (find_func_aliases_for_call): Properly handle indirect
6384 fi from direct call.
6385 (find_func_clobbers): Likewise.
6386 (ipa_pta_execute): Likewise.
6387 (create_variable_info_for): For functions that are ifunc_resolver
6388 resolve to a varinfo that contains the result of the resolver call.
6389 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
6392 2018-05-31 Michael Collison <michael.collison@arm.com>
6394 * config/aarch64/aarch64.md:
6395 (*fix_to_zero_extenddfdi2): New pattern.
6396 * gcc.target/aarch64/fix_extend1.c: New testcase.
6398 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
6402 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
6403 and BUILT_IN_STRNCMP_EQ.
6404 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
6405 BUILT_IN_STRNCMP_EQ.
6406 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
6407 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
6408 (gimple_fold_builtin): Likewise.
6409 * tree-ssa-strlen.c (compute_string_length): New function.
6410 (determine_min_obsize): New function.
6411 (handle_builtin_string_cmp): New function to handle calls to
6412 string compare functions.
6413 (strlen_optimize_stmt): Add handling to builtin string compare
6415 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
6416 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
6417 * tree.c (build_common_builtin_nodes): Add new defines of
6418 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
6420 2018-05-31 Jakub Jelinek <jakub@redhat.com>
6423 * bb-reorder.c (pass_partition_blocks::gate): Return false for
6424 functions with naked attribute.
6426 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
6428 * config/i386/sse.md (avx_vec_concat<mode>):
6429 Substitute concat_tg_mode mode attribute with xtg_mode.
6430 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
6431 (concat_tg_mode): Remove mode attribute.
6433 2018-05-31 Martin Sebor <msebor@redhat.com>
6436 * calls.c (alloc_max_size): Correct a logic error/typo.
6437 Treat excessive arguments as infinite. Warn for invalid arguments.
6438 * doc/invoke.texi (-Walloc-size-larger-than): Update.
6440 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
6443 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
6444 and movx for Haswell.
6446 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
6447 Cesar Philippidis <cesar@codesourcery.com>
6450 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
6451 when emitting error on private/firstprivate reductions.
6452 * omp-low.c (lower_omp_target): Avoid reference-type processing
6453 on pointers for firstprivate clause.
6455 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
6457 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
6460 * config/aarch64/aarch64-simd.md
6461 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
6462 (aarch64_ld1_x3_<mode>): Likewise
6463 (aarch64_st1x2<VALLDIF:mode>): Likewise
6464 (aarch64_st1_x2_<mode>): Likewise
6465 (aarch64_st1x3<VALLDIF:mode>): Likewise
6466 (aarch64_st1_x3_<mode>): Likewise
6467 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
6468 (vld1_s8_x3): Likewise.
6469 (vld1_u16_x3): Likewise.
6470 (vld1_s16_x3): Likewise.
6471 (vld1_u32_x3): Likewise.
6472 (vld1_s32_x3): Likewise.
6473 (vld1_u64_x3): Likewise.
6474 (vld1_s64_x3): Likewise.
6475 (vld1_f16_x3): Likewise.
6476 (vld1_f32_x3): Likewise.
6477 (vld1_f64_x3): Likewise.
6478 (vld1_p8_x3): Likewise.
6479 (vld1_p16_x3): Likewise.
6480 (vld1_p64_x3): Likewise.
6481 (vld1q_u8_x3): Likewise.
6482 (vld1q_s8_x3): Likewise.
6483 (vld1q_u16_x3): Likewise.
6484 (vld1q_s16_x3): Likewise.
6485 (vld1q_u32_x3): Likewise.
6486 (vld1q_s32_x3): Likewise.
6487 (vld1q_u64_x3): Likewise.
6488 (vld1q_s64_x3): Likewise.
6489 (vld1q_f16_x3): Likewise.
6490 (vld1q_f32_x3): Likewise.
6491 (vld1q_f64_x3): Likewise.
6492 (vld1q_p8_x3): Likewise.
6493 (vld1q_p16_x3): Likewise.
6494 (vld1q_p64_x3): Likewise.
6495 (vst1_s64_x2): Likewise.
6496 (vst1_u64_x2): Likewise.
6497 (vst1_f64_x2): Likewise.
6498 (vst1_s8_x2): Likewise.
6499 (vst1_p8_x2): Likewise.
6500 (vst1_s16_x2): Likewise.
6501 (vst1_p16_x2): Likewise.
6502 (vst1_s32_x2): Likewise.
6503 (vst1_u8_x2): Likewise.
6504 (vst1_u16_x2): Likewise.
6505 (vst1_u32_x2): Likewise.
6506 (vst1_f16_x2): Likewise.
6507 (vst1_f32_x2): Likewise.
6508 (vst1_p64_x2): Likewise.
6509 (vst1q_s8_x2): Likewise.
6510 (vst1q_p8_x2): Likewise.
6511 (vst1q_s16_x2): Likewise.
6512 (vst1q_p16_x2): Likewise.
6513 (vst1q_s32_x2): Likewise.
6514 (vst1q_s64_x2): Likewise.
6515 (vst1q_u8_x2): Likewise.
6516 (vst1q_u16_x2): Likewise.
6517 (vst1q_u32_x2): Likewise.
6518 (vst1q_u64_x2): Likewise.
6519 (vst1q_f16_x2): Likewise.
6520 (vst1q_f32_x2): Likewise.
6521 (vst1q_f64_x2): Likewise.
6522 (vst1q_p64_x2): Likewise.
6523 (vst1_s64_x3): Likewise.
6524 (vst1_u64_x3): Likewise.
6525 (vst1_f64_x3): Likewise.
6526 (vst1_s8_x3): Likewise.
6527 (vst1_p8_x3): Likewise.
6528 (vst1_s16_x3): Likewise.
6529 (vst1_p16_x3): Likewise.
6530 (vst1_s32_x3): Likewise.
6531 (vst1_u8_x3): Likewise.
6532 (vst1_u16_x3): Likewise.
6533 (vst1_u32_x3): Likewise.
6534 (vst1_f16_x3): Likewise.
6535 (vst1_f32_x3): Likewise.
6536 (vst1_p64_x3): Likewise.
6537 (vst1q_s8_x3): Likewise.
6538 (vst1q_p8_x3): Likewise.
6539 (vst1q_s16_x3): Likewise.
6540 (vst1q_p16_x3): Likewise.
6541 (vst1q_s32_x3): Likewise.
6542 (vst1q_s64_x3): Likewise.
6543 (vst1q_u8_x3): Likewise.
6544 (vst1q_u16_x3): Likewise.
6545 (vst1q_u32_x3): Likewise.
6546 (vst1q_u64_x3): Likewise.
6547 (vst1q_f16_x3): Likewise.
6548 (vst1q_f32_x3): Likewise.
6549 (vst1q_f64_x3): Likewise.
6550 (vst1q_p64_x3): Likewise.
6552 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
6554 * config/msp430/msp430.c (msp430_output_labelref): Prepend
6555 user_label_prefix to name.
6557 * tree-core.h: Update comment about the format of NAME string
6558 passed to handler in attribute_spec.
6560 * config/msp430/msp430.md: Remove erroneous subreg expression from
6561 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
6562 zero_extend{q,h}isi2.
6564 2018-05-30 Borislav Petkov <bp@suse.de>
6566 * doc/extend.texi: Document some architecture specific
6567 constraints and sort entries.
6569 2018-05-30 Martin Sebor <msebor@redhat.com>
6572 * builtins.c (expand_builtin_stpcpy_1): New function.
6573 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
6574 only if the former succeeds.
6576 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
6578 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
6581 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
6583 * doc/invoke.texi (-flinker-output): Document
6585 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
6587 * passes.c (ipa_write_summaries): Only modify statements if body
6589 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
6590 incrementally linking.
6591 (ipa_passes): Likewise.
6592 * lto-cgraph.c (lto_output_node): When incrementally linking do not
6593 pass down resolution info.
6594 * common.opt (flag_incremental_link): Update info.
6595 * gcc.c (plugin specs): Turn flinker-output=* to
6596 -plugin-opt=-linker-output-known
6597 * toplev.c (compile_file): Also cut compilation when doing incremental
6599 * flag-types. (enum lto_partition_model): Add
6600 LTO_LINKER_OUTPUT_NOLTOREL.
6601 (invoke.texi): Add -flinker-output docs.
6602 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
6603 link same way as WPA; do not stream in dead initializers.
6605 * dwarf2out.c (dwarf2out_die_ref_for_decl,
6606 darf2out_register_external_decl): Support incremental link.
6608 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
6610 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
6612 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
6614 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
6615 it down to simple_object_copy_lto_debug_sections.
6616 (run_gcc): Determine incremental LTO link time and configure
6617 lto1 into non-wpa mode, disable renaming of debug sections.
6619 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
6621 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
6622 descriptions of various incorrectly documented functions.
6624 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
6627 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
6628 address check not strict.
6630 2018-05-30 Richard Biener <rguenther@suse.de>
6632 PR tree-optimization/85964
6633 * tracer.c (better_p): Drop initialized count check, we only
6634 call the function with initialized counts now.
6635 (find_best_successor): Do find a best edge if one
6636 has uninitialized count.
6637 (find_best_predecessor): Likewise. Do BB frequency check only
6638 if count is initialized.
6640 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
6642 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
6643 (aarch64_ldrstr_offset_compare): New.
6644 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
6645 load/store orderings.
6646 (aarch64_gen_adjusted_ldpstp): Likewise.
6648 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
6650 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
6651 Check for subset of GENERAL_REGS and FP_REGS.
6652 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
6655 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
6657 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
6658 and wi::to_poly_offset. Add the current offset and then check
6659 whether the sum fits, rather than using an unchecked addition of
6660 a checked term. Check for a shwi rather than a uhwi.
6661 * expr.c (get_bit_range): Use tree_to_poly_uint64.
6662 (store_constructor): Use poly_int_tree_p.
6663 (expand_expr_real_1): Likewise.
6664 * function.c (assign_temp): Likewise.
6665 * fold-const.c (const_binop): Use poly_int_tree_p and
6667 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
6669 * ipa-icf-gimple.c (func_checker::compare_operand): Use
6670 to_poly_offset for MEM offsets.
6671 * ipa-icf.c (sem_variable::equals): Likewise.
6672 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
6673 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
6674 wi::to_poly_offset for BIT_FIELD_REF offsets.
6675 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
6677 * var-tracking.c (emit_note_insn_var_location): Use
6678 tree_to_poly_uint64.
6680 2018-05-29 Jim Wilson <jimw@sifive.com>
6682 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
6684 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
6687 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
6688 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
6690 (sse4_1_round<mode>2): Use nonimmediate_operand
6691 for operand 1 predicate.
6693 2018-05-29 Martin Sebor <msebor@redhat.com>
6694 Richard Biener <rguenther@suse.de>
6697 * calls.c (get_size_range): Call determine_value_range instead
6698 of get_value_range..
6699 * tree-vrp.h (determine_value_range): Declared new function.
6700 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
6702 2018-05-29 Richard Biener <rguenther@suse.de>
6704 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
6705 sure to use non-pattern stmts for get_earlier_stmt arguments.
6706 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
6707 called on pattern stmts.
6708 (get_later_stmt): Likewise.
6710 2018-05-29 Martin Liska <mliska@suse.cz>
6712 PR gcov-profile/85759
6713 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
6716 2018-05-29 Jakub Jelinek <jakub@redhat.com>
6718 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
6720 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
6724 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
6725 VEC_PACK_FLOAT_EXPR): New tree codes.
6726 * tree-pretty-print.c (op_code_prio): Handle
6727 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
6728 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
6729 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
6730 * tree-inline.c (estimate_operator_cost): Likewise.
6731 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
6732 * fold-const.c (const_binop): Likewise.
6733 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
6734 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
6735 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
6736 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
6737 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
6738 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
6739 * expr.c (expand_expr_real_2): Likewise.
6740 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
6741 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
6742 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
6744 * optabs.c (expand_widen_pattern_expr): For
6745 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
6746 sign from result type rather than operand's type.
6747 (expand_binop_directly): For vec_packu_float_optab and
6748 vec_packs_float_optab allow result type to be different from operand's
6750 * optabs-tree.c (optab_for_tree_code): Handle
6751 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
6752 VEC_PACK_FLOAT_EXPR. Formatting fixes.
6753 * tree-vect-generic.c (expand_vector_operations_1): Handle
6754 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
6755 VEC_PACK_FLOAT_EXPR.
6756 * tree-vect-stmts.c (supportable_widening_operation): Handle
6758 (supportable_narrowing_operation): Handle FLOAT_EXPR.
6759 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
6760 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
6761 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
6762 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
6764 (vec_pack<floatprefix>_float_<mode>): New expander.
6765 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
6767 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
6768 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
6769 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
6770 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
6771 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
6773 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
6774 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
6775 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
6776 VEC_PACK_FLOAT_EXPR): Document.
6778 2018-05-29 Richard Biener <rguenther@suse.de>
6780 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
6782 (stmt_vec_info_vec): Make pointer.
6783 (init_stmt_vec_info_vec): Remove.
6784 (free_stmt_vec_info_vec): Likewise.
6785 (set_stmt_vec_info_vec): New function.
6786 (free_stmt_vec_infos): Likewise.
6787 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
6788 (set_vinfo_for_stmt): Likewise.
6789 (get_earlier_stmt): Likewise.
6790 (get_later_stmt): Likewise.
6791 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
6792 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
6793 (vec_info::~vec_info): Free stmt_vec_infos.
6794 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
6795 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
6796 (pass_slp_vectorize::execute): Likewise.
6797 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
6798 (free_stmt_vec_info_vec): Likewise.
6799 (set_stmt_vec_info_vec): New function.
6800 (free_stmt_vec_infos): Likewise.
6801 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
6802 the global stmt_vec_info_vec.
6803 * tree-parloops.c (gather_scalar_reductions): Use
6804 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
6807 2018-05-29 Richard Biener <rguenther@suse.de>
6809 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
6811 2018-05-29 Martin Liska <mliska@suse.cz>
6812 David Malcolm <dmalcolm@redhat.com>
6814 * vec.c (test_reverse): New.
6815 (vec_c_tests): Add new test.
6816 * vec.h (vl_ptr>::reverse): New function.
6818 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
6820 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
6822 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
6825 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
6827 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
6829 2018-05-28 Richard Biener <rguenther@suse.de>
6831 PR tree-optimization/85933
6832 * tree-vect-data-refs.c (vect_record_base_alignments): Only
6833 look at stmts marked as vectorizable.
6835 2018-05-28 Richard Biener <rguenther@suse.de>
6837 PR tree-optimization/85934
6838 * tree-vect-generic.c (expand_vector_operations_1): Hoist
6839 vector boolean check before scalar optimization.
6841 2018-05-28 Jakub Jelinek <jakub@redhat.com>
6843 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
6846 2018-05-28 Mark Wielaard <mark@klomp.org>
6848 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
6849 if it is an expression containing a minus sign.
6851 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
6853 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
6855 2018-05-27 Paul Koning <ni1d@arrl.net>
6857 * config/pdp11/pdp11.md (truncsihi2): Remove.
6859 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
6860 Chung-Ju Wu <jasonwucj@gmail.com>
6862 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
6864 (unaligned_store_dw): Ditto.
6865 * config/nds32/nds32-memory-manipulation.c
6866 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
6867 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
6868 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
6869 (emit_setmem_word_loop): Rename to ...
6870 (emit_setmem_doubleword_loop): ... this.
6871 (nds32_gen_dup_4_byte_to_word_value): New function.
6872 (nds32_gen_dup_8_byte_to_double_word_value): New function.
6873 (nds32_expand_setmem_loop): Refine implementation.
6874 (nds32_expand_setmem_loop_v3m): Ditto.
6875 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
6878 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
6880 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
6882 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
6884 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
6885 (nds32_init_machine_status): Initialize machine->attr_naked_p and
6886 machine->attr_no_prologue_p.
6887 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
6888 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
6889 (nds32_expand_epilogue): Consider attr_naked_p.
6890 (nds32_expand_epilogue_v3pop): Likewise.
6891 (nds32_can_use_return_insn): Likewise.
6892 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
6893 attr_no_prologue_p fields.
6894 * config/nds32/nds32.opt (mret-in-naked-func): New option.
6896 2018-05-27 Jakub Jelinek <jakub@redhat.com>
6899 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
6901 * config/i386/sse.md
6902 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
6904 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
6906 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
6908 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
6910 (*<floatsuffix>floatv2div2sf2): Rename to ...
6911 (*float<floatunssuffix>v2div2sf2): ... this.
6912 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
6913 (float<floatunssuffix>v2div2sf2_mask): ... this.
6914 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
6915 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
6916 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
6918 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
6920 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
6922 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
6924 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
6926 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
6928 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
6929 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
6930 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
6931 gen_ufix_truncv8dfv8si2.
6932 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
6933 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
6934 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
6935 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
6936 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
6937 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
6938 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
6939 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
6941 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
6945 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
6947 2018-05-25 Jim Wilson <jimw@sifive.com>
6949 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
6950 * config/riscv/riscv.c (struct machine_function): Add
6951 interrupt_handler_p and attribute_checked_p fields.
6952 (riscv_attribute_table): Add interrupt.
6953 (riscv_interrupt_type_p): New.
6954 (riscv_save_reg_p): Save extra regs for interrupt handler.
6955 (riscv_use_save_libcall): Return false for interrupt handler.
6956 (riscv_first_stack_step): Add forward declaration.
6957 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
6958 for interrupt handler with large frame. Use it for saved reg list.
6959 (riscv_expand_prologue): Move flag_stack_usage_info support to
6960 eliminate duplication.
6961 (riscv_expand_epilogue): Generate mret for interrupt handler.
6962 (riscv_epilogue_uses): New.
6963 (riscv_can_use_return_insn): Return false for interrupt handler.
6964 (riscv_function_ok_for_sibcall): Likewise.
6965 (riscv_set_current_function): Add interrupt handler support.
6966 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
6967 * config/riscv/riscv.md (UNSPECV_MRET): New.
6969 (riscv_frflags, riscv_fsflags): Use tab after opcode.
6971 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
6973 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
6975 PR tree-optimization/85712
6976 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
6977 this candidate has already been replaced in-situ by a copy.
6979 2018-05-25 Jason Merrill <jason@redhat.com>
6981 PR c++/80485 - inline function non-zero address.
6982 * symtab.c (nonzero_address): Check DECL_COMDAT.
6984 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
6987 * config/alpha/alpha.md (ashlsi3): New insn pattern.
6988 (*ashlsi_se): Rename from *ashldi_se. Define as sign
6989 extension of SImode operation. Use const123_operand predicate.
6990 (*saddsi_1): Remove.
6991 (*saddl_se_1): Ditto.
6993 (*ssubl_se_1): Ditto.
6994 * config/alpha/predicates.md (const123_operand): New predicate.
6995 * config/alpha/constraints.md (P): Use IN_RANGE.
6997 2018-05-25 Richard Biener <rguenther@suse.de>
6999 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
7001 (ref_maybe_used_by_stmt_p): Likewise.
7002 (stmt_may_clobber_ref_p): Likewise.
7003 (stmt_may_clobber_ref_p_1): Likewise.
7004 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
7006 (ref_maybe_used_by_stmt_p): Likewise.
7007 (stmt_may_clobber_ref_p): Likewise.
7008 (stmt_may_clobber_ref_p_1): Likewise.
7009 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
7010 the alias oracle to disambiguate DRs with stmts DR analysis
7012 (vect_analyze_data_refs): Do not give up on not analyzable
7013 DRs for BB vectorization. Remove code truncating the dataref
7016 2018-05-25 Jakub Jelinek <jakub@redhat.com>
7019 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
7020 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
7021 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
7023 2018-05-25 Richard Biener <rguenther@suse.de>
7025 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
7026 function, combining stmt data ref gathering and fatal analysis
7028 (vect_analyze_data_refs): Remove now redudnant code and simplify.
7029 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
7030 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
7031 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
7032 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
7034 2018-05-25 Bin Cheng <bin.cheng@arm.com>
7036 PR tree-optimization/85720
7037 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
7038 SCC if all partitions are builtins.
7039 (version_loop_by_alias_check): New parameter. Generate cancelable
7040 runtime alias check if all partitions are builtins.
7041 (distribute_loop): Update call to above function.
7043 2018-05-25 Bin Cheng <bin.cheng@arm.com>
7045 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
7046 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
7047 (parm_default_def_partition_arg): Ditto.
7048 (set_parm_default_def_partition): Ditto.
7049 (get_parm_default_def_partitions): Ditto and make it static.
7050 (get_undefined_value_partitions): Ditto and make it static.
7051 (remove_ssa_form): Refactor call to init_var_map here.
7052 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
7053 computation for loop region.
7054 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
7055 (register_default_def): Delete.
7056 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
7057 (parm_default_def_partition_arg): Ditto.
7058 (set_parm_default_def_partition): Ditto.
7059 (get_parm_default_def_partitions): Ditto and make it static.
7060 (get_undefined_value_partitions): Ditto and make it static.
7061 (coalesce_with_default, coalesce_with_default): Update comment.
7062 (create_coalesce_list_for_region): New func factored out from
7063 create_outofssa_var_map.
7064 (populate_coalesce_list_for_outofssa): New func factored out from
7065 create_outofssa_var_map and coalesce_ssa_name.
7066 (create_outofssa_var_map): Delete.
7067 (coalesce_ssa_name): Refactor to support live range computation.
7068 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
7069 (get_parm_default_def_partitions): Delete.
7070 (get_undefined_value_partitions): Ditto.
7071 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
7072 computation for loop region.
7073 (new_tree_live_info, loe_visit_block): Ditto.
7074 (live_worklist, set_var_live_on_entry): Ditto.
7075 (calculate_live_on_exit, verify_live_on_entry): Ditto.
7076 * tree-ssa-live.h (struct _var_map): New fields.
7077 (init_var_map): Change decl.
7078 (region_contains_p): New.
7080 2018-05-25 Bin Cheng <bin.cheng@arm.com>
7082 * tree-ssa-live.h (live_merge_and_clear): Delete.
7084 2018-05-25 Richard Biener <rguenther@suse.de>
7087 * tree-dump.c (dequeue_and_dump): Remove access to removed
7088 operand 2 of a SWITCH_EXPR.
7090 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
7092 * doc/sourcebuild.texi (vect_double_cond_arith): Include
7093 multiplication and division.
7094 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
7095 (cond_udiv@var{m}, cond_umod@var{m}): Document.
7096 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
7097 (cond_udiv_optab, cond_umod_optab): New optabs.
7098 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
7099 (IFN_COND_RDIV): New internal functions.
7100 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
7101 TRUNC_MOD_EXPR and RDIV_EXPR.
7102 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
7103 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
7105 (SVE_INT_BINARY): Include mult.
7106 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
7107 (optab, sve_int_op): Handle mult.
7108 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
7110 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
7111 for SVE_INT_BINARY_SD.
7113 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
7115 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
7116 (optab, sve_int_op): Handle div and udiv.
7117 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
7118 for SVE_INT_BINARY_SD.
7119 (*<optab><mode>3): New insn for the same.
7121 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
7123 * tree-vect-patterns.c: Include predict.h.
7124 (vect_recog_divmod_pattern): Restrict check for division support
7125 to when optimizing for size.
7127 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
7129 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
7130 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
7131 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
7132 (gimple_match_op::set_op): Likewise.
7133 (gimple_resimplify4): Declare.
7134 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
7135 (expr::gen_transform): Likewise.
7136 (decision_tree::gen): Generate a simplification routine for 4 operands.
7137 * gimple-match-head.c (gimple_simplify): Add an overload for
7138 4 operands. In the top-level function, handle up to 4 call
7139 arguments and call gimple_resimplify4.
7140 (gimple_resimplify4): New function.
7141 (build_call_internal): Pass a fourth operand.
7142 (maybe_push_to_seq): Likewise.
7143 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
7144 Fold VEC_COND_EXPRs of an operation and a default value into
7145 an IFN_COND_* function if possible.
7146 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
7148 (SVE_COND_FP_BINARY): Include them.
7149 (optab, sve_fp_op): Handle them.
7150 (SVE_INT_BINARY_REV): New code iterator.
7151 (SVE_COND_FP_BINARY_REV): New int iterator.
7152 (commutative): New int attribute.
7153 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
7155 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
7157 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
7158 (*cond_<optab><mode>): New patterns for reversed operands.
7160 2018-05-25 Richard Biener <rguenther@suse.de>
7162 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
7163 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
7164 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
7165 (STMT_VINFO_GROUPED_ACCESS): Adjust.
7166 * tree-vect-data-refs.c (everywhere): Adjust users.
7167 * tree-vect-loop.c (everywhere): Likewise.
7168 * tree-vect-slp.c (everywhere): Likewise.
7169 * tree-vect-stmts.c (everywhere): Likewise.
7170 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
7172 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7174 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
7176 (gcc_cv_as_section_exclude): ... this.
7177 Try Solaris as #exclude syntax.
7178 * configure: Regenerate.
7179 * config.in: Regenerate.
7180 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
7182 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
7183 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
7185 * varasm.c (default_elf_asm_named_section): Don't check if
7186 HAVE_GAS_SECTION_EXCLUDE is defined.
7188 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
7190 * doc/md.texi: Update the documentation of the cond_* optabs
7191 to mention the new final operand. Fix GET_MODE_NUNITS call.
7192 Describe the scalar case too.
7193 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
7194 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
7196 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
7197 (get_conditional_internal_fn): Update comment.
7198 * tree-vect-loop.c (vectorizable_reduction): Pass the original
7199 accumulator value as a final argument to conditional functions.
7200 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
7201 a define_expand and add an "else" operand. Assert for now that
7202 the else operand is equal to operand 2. Use SVE_INT_BINARY and
7203 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
7204 (*cond_<optab><mode>): New patterns.
7205 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
7206 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
7207 (UNSPEC_COND_EOR): Delete.
7208 (optab): Remove associated mappings.
7209 (SVE_INT_BINARY): New code iterator.
7210 (sve_int_op): Remove int attribute and add "minus" to the code
7212 (SVE_COND_INT_OP): Delete.
7213 (SVE_COND_FP_OP): Rename to...
7214 (SVE_COND_FP_BINARY): ...this.
7216 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
7218 * optabs.c (can_reuse_operands_p): New function.
7219 (maybe_legitimize_operands): Try to reuse the results for
7222 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
7224 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
7225 Add {q} suffix to insn mnemonic.
7227 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
7229 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
7230 (msp430_warn_func_return): New.
7232 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
7234 * fold-const.c (tree_nonzero_bits): New function.
7235 * fold-const.h (tree_nonzero_bits): Likewise.
7236 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
7237 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
7239 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
7243 * varasm.c (assemble_alias): Check ifunc_resolver only on
7246 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
7249 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
7250 when memory input operand is handled.
7252 2018-05-24 Luis Machado <luis.machado@linaro.org>
7254 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
7256 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
7258 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
7260 * match.pd: Delay FMA folds until after vectorization.
7262 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
7265 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
7266 address check not strict.
7268 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
7270 * gimple-match.h (gimple_match_op): New class.
7271 (mprts_hook): Replace parameters with a gimple_match_op *.
7272 (maybe_build_generic_op): Likewise.
7273 (gimple_simplified_result_is_gimple_val): Replace parameters with
7274 a const gimple_match_op *.
7275 (gimple_simplify): Replace code_helper * and tree * parameters with
7276 a gimple_match_op * parameter.
7277 (gimple_resimplify1): Replace code_helper *, tree and tree *
7278 parameters with a gimple_match_op * parameter.
7279 (gimple_resimplify2): Likewise.
7280 (gimple_resimplify3): Likewise.
7281 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
7282 parameters with a gimple_match_op * parameter.
7283 * gimple-match-head.c (gimple_simplify): Change prototypes of
7284 auto-generated functions to take a gimple_match_op * instead of
7285 separate code_helper * and tree * parameters. Make the same
7286 change in the top-level overload and update calls to the
7287 gimple_resimplify routines. Update calls to the auto-generated
7288 functions and to maybe_push_res_to_seq in the publicly-facing
7289 operation-specific gimple_simplify overloads.
7290 (gimple_match_op::MAX_NUM_OPS): Define.
7291 (gimple_resimplify1): Replace rcode and ops with a single res_op
7292 parameter. Update call to gimple_simplify.
7293 (gimple_resimplify2): Likewise.
7294 (gimple_resimplify3): Likewise.
7295 (mprts_hook): Replace parameters with a gimple_match_op *.
7296 (maybe_build_generic_op): Likewise.
7297 (build_call_internal): Replace type, nargs and ops with
7298 a gimple_match_op *.
7299 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
7300 with a single gimple_match_op *. Update calls to mprts_hook,
7301 build_call_internal and gimple_simplified_result_is_gimple_val.
7302 Factor out code that is common to the tree_code and combined_fn cases.
7303 * genmatch.c (expr::gen_transform): Replace tem_code and
7304 tem_ops with a gimple_match_op called tem_op. Update calls
7305 to the gimple_resimplify functions and maybe_push_res_to_seq.
7306 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
7307 res_ops. Update call to the gimple_resimplify functions.
7308 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
7309 (decision_tree::gen): Make the functions take a gimple_match_op *
7310 called res_op instead of separate res_code and res_ops parameters.
7311 Update call accordingly.
7312 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
7313 and ops with a single res_op parameter. Update calls to
7314 maybe_build_generic_op and maybe_push_res_to_seq.
7315 (fold_stmt_1): Update calls to gimple_simplify and
7316 replace_stmt_with_simplification.
7317 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
7318 and gimple_simplified_result_is_gimple_val.
7319 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
7321 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
7322 with a gimple_match_op *.
7323 (vn_nary_build_or_lookup): Likewise. Update call to
7324 vn_nary_build_or_lookup_1.
7325 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
7326 gimple_match_op *. Update calls to the gimple_resimplify routines
7327 and to gimple_simplified_result_is_gimple_val.
7328 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
7329 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
7330 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
7331 (visit_nary_op): Likewise.
7332 (visit_reference_op_load): Likewise.
7334 2018-05-23 Luis Machado <luis.machado@linaro.org>
7336 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
7337 modifier for printing the step amount.
7339 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
7342 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
7345 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
7347 * doc/sourcebuild.texi (Endianness): New subsubsection.
7349 2018-05-23 Luis Machado <luis.machado@linaro.org>
7351 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7352 <prefetch_dynamic_strides>: New const bool field.
7353 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7354 prefetch_dynamic_strides.
7355 (exynosm1_prefetch_tune): Likewise.
7356 (thunderxt88_prefetch_tune): Likewise.
7357 (thunderx_prefetch_tune): Likewise.
7358 (thunderx2t99_prefetch_tune): Likewise.
7359 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
7361 (aarch64_override_options_internal): Update to set
7362 PARAM_PREFETCH_DYNAMIC_STRIDES.
7363 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7364 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7365 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7366 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7367 prefetch-dynamic-strides setting.
7369 2018-05-23 Luis Machado <luis.machado@linaro.org>
7371 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7372 <minimum_stride>: New const int field.
7373 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7374 minimum_stride field defaulting to -1.
7375 (exynosm1_prefetch_tune): Likewise.
7376 (thunderxt88_prefetch_tune): Likewise.
7377 (thunderx_prefetch_tune): Likewise.
7378 (thunderx2t99_prefetch_tune): Likewise.
7379 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
7380 <default_opt_level>: Set to 3.
7381 (aarch64_override_options_internal): Update to set
7382 PARAM_PREFETCH_MINIMUM_STRIDE.
7383 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7384 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7385 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7386 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7387 stride is constant and is below the minimum stride threshold.
7389 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7391 * config/arm/arm-cpus.in (mode26): Delete.
7392 (armv4): Delete mode26 reference.
7393 * config/arm/arm.c (arm_configure_build_target): Delete use of
7396 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
7398 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
7400 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
7401 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
7403 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
7404 (floatunsdidf2): Ditto.
7406 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
7408 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
7409 (fixuns_trunc<mode>si2_avx512f): Ditto.
7410 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
7411 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
7412 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
7414 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
7416 PR rtl-optimization/79985
7417 * df-scan.c (df_insn_refs_collect): Remove special case for
7418 global registers and asm statements.
7420 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
7422 * extend.texi (Global Register Variables): Rewrite the bullet list.
7423 Note that the register is available for allocation. Note that access
7424 via inline asm must use constraints. Add note about async-signal
7425 handlers. Remove paragraph about automagic register selection.
7427 2018-05-23 Richard Biener <rguenther@suse.de>
7429 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
7430 of fixed offset from memset VN.
7432 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
7434 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
7436 (alloc_cand_and_find_basis): Initialize first_interp field.
7437 (slsr_process_mul): Modify first_interp field.
7438 (slsr_process_add): Likewise.
7439 (slsr_process_cast): Modify first_interp field for each new
7441 (slsr_process_copy): Likewise.
7442 (dump_candidate): Dump first_interp field.
7443 (replace_mult_candidate): Process all interpretations, not just
7445 (replace_rhs_if_not_dup): Likewise.
7446 (replace_one_candidate): Likewise.
7448 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
7450 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
7452 (aarch64_needs_frame_chain): New function.
7453 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
7455 2018-05-23 Sudakshina Das <sudi.das@arm.com>
7458 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
7459 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
7460 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
7461 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
7462 as true for strict-align.
7463 (aarch64_can_inline_p): Perform checks even when callee has no
7464 attributes to check for strict alignment.
7465 * doc/extend.texi (AArch64 Function Attributes): Document
7467 * doc/invoke.texi: (AArch64 Options): Likewise.
7469 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
7471 PR tree-optimization/85853
7472 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
7473 the handling of the root of the node to...
7474 (vect_slp_analyze_node_operations_1): ...this new function,
7475 and run the whole thing with the child nodes' def types
7476 set according to their SLP node's def type.
7478 2018-05-23 Richard Biener <rguenther@suse.de>
7481 * tree-data-ref.c (create_runtime_alias_checks): Defer
7482 and ignore overflow warnings.
7484 2018-05-23 Yury Gribov <tetra2005@gmail.com>
7486 PR tree-optimization/85822
7487 * tree-vrp.c (is_masked_range_test): Fix handling of negative
7490 2018-05-23 Richard Biener <rguenther@suse.de>
7492 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
7493 memset constants via native_interpret_expr.
7495 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
7498 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
7500 (cgraph_node::create_alias): Likewise.
7501 (cgraph_node::get_availability): Check ifunc_resolver instead
7502 of looking up ifunc attribute.
7503 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
7504 * varasm.c (do_assemble_alias): Likewise.
7505 (assemble_alias): Likewise.
7506 (default_binds_local_p_3): Likewise.
7507 * cgraph.h (cgraph_node): Add ifunc_resolver.
7508 (cgraph_node::only_called_directly_or_aliased_p): Return false
7510 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
7512 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
7513 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
7514 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
7515 instead of looking up ifunc attribute.
7517 2018-05-22 Luis Machado <luis.machado@linaro.org>
7519 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
7521 2018-05-22 Martin Sebor <msebor@redhat.com>
7524 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
7525 only when expasion succeeds.
7526 (expand_builtin_strcmp): Same.
7527 (expand_builtin_strncmp): Same.
7529 2018-05-22 Martin Sebor <msebor@redhat.com>
7531 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
7533 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
7534 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7536 * config/aarch64/aarch64-ldpstp.md: Replace uses of
7537 aarch64_mem_pair_operand with memory_operand and delete operand swapping
7539 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
7540 Add check for legitimate_address.
7541 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
7542 (aarch64_swap_ldrstr_operands): New.
7543 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
7546 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
7547 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7549 * config/aarch64/aarch64.md: New patterns to generate stp
7551 (store_pair_sw, store_pair_dw): New patterns to generate stp for
7552 single words and double words.
7553 (load_pair_sw, load_pair_dw): Likewise.
7554 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
7556 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
7558 * config/aarch64/aarch64-ldpstp.md: Modify peephole
7559 for different mode ldpstp and add peephole for merged zero stores.
7561 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
7563 (aarch64_gen_store_pair): Rename calls to match new patterns.
7564 (aarch64_gen_load_pair): Rename calls to match new patterns.
7565 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
7566 (load_pair<DREG:mode><DREG2:mode>): ... This.
7567 (store_pair<mode>): Rename to...
7568 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
7569 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
7571 (V_INT_EQUIV): Handle SImode.
7572 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
7575 2018-05-22 Martin Sebor <msebor@redhat.com>
7578 * calls.c (maybe_warn_nonstring_arg): Use string length to set
7579 or ajust the presumed bound on an operation to avoid unnecessary
7582 2018-05-22 Martin Sebor <msebor@redhat.com>
7584 PR tree-optimization/85826
7585 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
7586 assuming that a DECL necesarily has a constant size.
7588 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
7591 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
7593 2018-05-22 Richard Biener <rguenther@suse.de>
7595 PR tree-optimization/85834
7596 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
7597 non-constant and non-zero memset arguments.
7599 2018-05-22 Martin Liska <mliska@suse.cz>
7602 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
7604 2018-05-22 Richard Biener <rguenther@suse.de>
7606 PR tree-optimization/85863
7607 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
7608 comparisons when vectype is specified.
7609 (vectorizable_condition): Do not specify vectype for
7610 vect_is_simple_cond when SLP vectorizing.
7612 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
7615 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
7616 define __ibm128 as long double.
7617 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
7618 as a distinct type when IEEE 128-bit support is enabled.
7619 (init_float128_ieee): Fix up conversions between IFmode and IEEE
7620 128-bit types to use the correct functions.
7621 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
7622 convert between 128-bit floating point types that have different
7623 modes but the same representation, instead of using gen_lowpart to
7625 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
7627 (IFKF_reg): New attributes to give the register constraints for
7629 (extend<mode>tf2_internal): New insns to mark an explicit
7630 conversion between 128-bit floating point types that have a
7631 different mode but share the same representation.
7633 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
7635 PR tree-optimization/85814
7636 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
7637 a null return from get_strinfo when unsharing the next
7638 strinfo in the chain.
7640 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
7643 * varasm.c (weak_finish): Clean up weak_decls.
7645 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7647 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
7648 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
7649 UNSPEC_UADALP values.
7650 * config/aarch64/iterators.md (ABAL): New int iterator.
7653 (sur): Add mappings for the above.
7654 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
7656 (aarch64_<sur>abal<mode>_4): Likewise.
7657 (aarch64_<sur>adalp<mode>_3): Likewise.
7658 (<sur>sadv16qi): New define_expand.
7660 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
7662 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
7663 (*movdf_internal): Ditto.
7664 (*rcpsf2_sse): Ditto.
7665 (*rsqrtsf2_sse): Ditto.
7666 (*sqrt<mode>2_sse): Ditto.
7668 2018-05-21 Tamar Christina <tamar.christina@arm.com>
7670 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
7672 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
7673 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
7674 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
7675 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
7677 * config/aarch64/arm_neon.h: Likewise.
7678 * config/aarch64/iterators.md (VQ_I): New.
7680 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
7682 * config.gcc: Add arc/t-multilib-linux to tmake_file for
7684 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
7687 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
7689 * config/nds32/constraints.md (S): New constraint.
7690 * config/nds32/nds32.md (call_internal): Use constraint S.
7691 (call_value_internal): Likewise.
7692 (sibcall_internal): Likewise.
7693 (sibcall_value_internal): Likewise.
7695 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
7696 Chung-Ju Wu <jasonwucj@gmail.com>
7698 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
7701 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
7702 Chung-Ju Wu <jasonwucj@gmail.com>
7704 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
7705 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
7706 (nds32_rtx_costs_impl): Simplify.
7707 (nds32_address_cost_impl): Simplify.
7708 (nds32_init_rtx_costs): New function.
7709 (nds32_rtx_costs_speed_prefer): Likewise.
7710 (nds32_rtx_costs_size_prefer): Likewise.
7711 (nds32_address_cost_speed_prefer): Likewise.
7712 (nds32_address_cost_speed_fwprop): Likewise.
7713 (nds32_address_cost_size_prefer): Likewise.
7714 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
7715 * config/nds32/nds32.c (nds32_option_override): Use
7716 nds32_init_rtx_costs function.
7718 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
7720 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
7721 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
7722 (TARGET_PIPELINE_N8): Likewise.
7723 (TARGET_PIPELINE_N10): Likewise.
7724 (TARGET_PIPELINE_N13): Likewise.
7725 (TARGET_PIPELINE_GRAYWOLF): Likewise.
7727 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
7729 * config/nds32/nds32-fpu.md: Update copyright year.
7731 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7733 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
7735 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7737 * config/nds32/nds32.c
7738 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
7739 * config/nds32/nds32.opt (minline-asm-r15): New option.
7741 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7743 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
7745 * config/nds32/nds32.md (abssi2): New pattern.
7747 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
7749 * config/i386/i386.md (rex64namesuffix): New mode attribute.
7750 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
7751 Merge insn pattern from sse_cvtsi2ss<round_name> and
7752 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
7753 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
7754 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
7755 using SWI48 mode iterator.
7756 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
7757 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
7758 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
7759 pattern from sse_cvttss2si<round_saeonly_name>
7760 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
7761 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
7762 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
7763 using SWI48 mode iterator.
7764 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
7765 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
7766 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
7767 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
7768 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
7769 using SWI48 mode iterator.
7770 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
7771 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
7772 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
7773 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
7774 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
7775 SWI48 mode iterator.
7776 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
7777 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
7778 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
7779 pattern from sse_cvttsd2si<round_saeonly_name>
7780 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
7782 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7784 * config/nds32/nds32-md-auxiliary.c
7785 (nds32_valid_smw_lwm_base_p): Refine.
7786 (nds32_output_smw_single_word): Refine.
7787 (nds32_output_smw_double_word): New.
7788 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
7790 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7792 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
7793 (nds32_output_stack_pop): Refine.
7794 (nds32_expand_unaligned_load): Refine.
7795 (nds32_expand_unaligned_store): Refine.
7797 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7798 Chung-Ju Wu <jasonwucj@gmail.com>
7800 * config/nds32/constants.md: Add TP_REGNUM constant.
7801 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
7802 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
7804 * config/nds32/nds32-doubleword.md: Consider flag_pic.
7805 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
7806 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
7807 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
7808 and PIC code generation.
7809 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
7811 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
7813 * config/nds32/nds32.md: Support TLS and PIC.
7814 * config/nds32/nds32.c: Support TLS and PIC.
7815 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
7816 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
7819 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7821 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
7822 mode with E_ prefix.
7824 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7825 Chung-Ju Wu <jasonwucj@gmail.com>
7827 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
7828 * config/nds32/nds32-md-auxiliary.c
7829 (symbolic_reference_mentioned_p): New.
7830 (nds32_legitimize_ict_address): New.
7831 (nds32_expand_ict_move): New.
7832 (nds32_indirect_call_referenced_p): New.
7833 (nds32_symbol_binds_local_p): Delete.
7834 (nds32_long_call_p): Modify.
7835 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
7836 * config/nds32/nds32-protos.h
7837 (symbolic_reference_mentioned_p): Declare.
7838 (nds32_legitimize_ict_address): Declare.
7839 (nds32_expand_ict_move): Declare.
7840 (nds32_indirect_call_referenced_p): Declare.
7841 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
7842 (nds32_relax_group): Use nds32_ict_const_p as condition.
7843 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
7844 (nds32_asm_file_start): Output ict_model directive in asm code.
7845 (nds32_legitimate_address_p): Consider indirect call.
7846 (nds32_print_operand): Consider indirect call.
7847 (nds32_print_operand_address): Consider indirect call.
7848 (nds32_insert_attributes): Handle "indirect_call" attribute.
7849 (TARGET_LEGITIMATE_ADDRESS_P): Define.
7850 (TARGET_LEGITIMATE_CONSTANT_P): Define.
7851 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
7852 (TARGET_DELEGITIMIZE_ADDRESS): Define.
7853 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
7854 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
7855 (TARGET_ICT_MODEL_SMALL): Define.
7856 (TARGET_ICT_MODEL_LARGE): Define.
7857 * config/nds32/nds32.md (movsi): Consider ict model.
7858 (call, call_value): Consider ict model.
7859 (sibcall, sibcall_value): Consider ict model.
7860 * config/nds32/nds32.opt (mict-model): New option.
7861 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
7864 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
7865 Monk Chiang <sh.chiang04@gmail.com>
7866 Jim Wilson <jimw@sifive.com>
7868 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
7869 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
7870 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
7871 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
7872 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
7873 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
7874 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
7875 compute save_libcall_adjustment properly.
7876 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
7877 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
7878 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
7879 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
7880 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
7881 (ABI_SPEC): Handle mabi=ilp32e.
7882 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
7883 (RVE): Add RVE mask.
7884 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
7885 <-march>: Add rv32e as an example.
7887 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
7890 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
7891 (intra_create_variable_infos): Handle C++ constructors.
7893 2018-05-18 Martin Liska <mliska@suse.cz>
7895 * passes.def: Remove a redundant pass.
7897 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
7900 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
7902 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7904 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
7906 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
7908 (armv2, armv2a, armv3, armv3m): Delete architectures.
7909 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
7910 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
7911 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
7913 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
7914 (*mulsidi3adddi): Likewise.
7915 (mulsidi3): Likewise.
7916 (*mulsidi3_nov6): Likewise.
7917 (umulsidi3): Likewise.
7918 (umulsidi3_nov6): Likewise.
7919 (umaddsidi4): Likewise.
7920 (*umulsidi3adddi): Likewise.
7921 (smulsi3_highpart): Likewise.
7922 (*smulsi3_highpart_nov6): Likewise.
7923 (umulsi3_highpart): Likewise.
7924 (*umulsi3_highpart_nov6): Likewise.
7925 * config/arm/arm.h (arm_arch3m): Delete.
7926 * config/arm/arm.c (arm_arch3m): Delete.
7927 (arm_option_override_internal): Update armv3-related comment.
7928 (arm_configure_build_target): Delete use of isa_bit_mode32.
7929 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
7930 (arm_rtx_costs_internal): Delete check of arm_arch3m.
7931 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
7933 (mulusa3): Likewise.
7934 * config/arm/arm-protos.h (arm_arch3m): Delete.
7935 * config/arm/arm-tables.opt: Regenerate.
7936 * config/arm/arm-tune.md: Likewise.
7937 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
7938 deleted architectures.
7940 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7942 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
7943 (armv5t, armv5te): New features.
7944 (ARMv5, ARMv5e): Delete fgroups.
7945 (ARMv5t, ARMv5te): Adjust for above changes.
7947 (armv5, armv5e): Delete arches.
7948 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
7950 (*call_reg_arm): Likewise.
7951 (*call_value_reg_armv5): Likewise.
7952 (*call_value_reg_arm): Likewise.
7953 (*call_symbol): Likewise.
7954 (*call_value_symbol): Likewise.
7955 (*sibcall_insn): Likewise.
7956 (*sibcall_value_insn): Likewise.
7958 (prefetch): Likewise.
7959 (define_split and define_peephole2 dependent on arm_arch5):
7961 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
7963 (TARGET_ARM_QBIT): Likewise.
7964 (TARGET_DSP_MULTIPLY): Likewise.
7965 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
7966 (arm_arch5, arm_arch5e): Delete.
7967 (arm_arch5t, arm_arch5te): Declare.
7968 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
7969 (arm_arch5t): Declare.
7970 (arm_option_reconfigure_globals): Update for the above.
7971 (arm_options_perform_arch_sanity_checks): Update comment, replace
7972 use of arm_arch5 with arm_arch5t.
7973 (use_return_insn): Likewise.
7974 (arm_emit_call_insn): Likewise.
7975 (output_return_instruction): Likewise.
7976 (arm_final_prescan_insn): Likewise.
7977 (arm_coproc_builtin_available): Likewise.
7978 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
7979 arm_arch5e with arm_arch5t and arm_arch5te.
7980 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
7981 (arm_arch5t, arm_arch5te): Declare.
7982 * config/arm/arm-tables.opt: Regenerate.
7983 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
7984 * config/arm/t-multilib: Likewise.
7985 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
7986 instead of arm_arch5.
7987 (*call_reg_thumb1): Likewise.
7988 (*call_value_reg_thumb1_v5): Likewise.
7989 (*call_value_reg_thumb1): Likewise.
7990 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
7992 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
7994 2018-05-18 Martin Liska <mliska@suse.cz>
7996 PR gcov-profile/84846
7997 * doc/gcov.texi: Document -t option of gcov tool.
7999 2018-05-18 Martin Liska <mliska@suse.cz>
8001 PR gcov-profile/84846
8002 * gcov.c (print_usage): Add new -t option.
8003 (process_args): Handle the option.
8004 (generate_results): Use stdout as output when requested by
8007 2018-05-18 Martin Liska <mliska@suse.cz>
8009 PR gcov-profile/84846
8010 * coverage.c (coverage_init): Write PWD to .gcno file.
8011 * doc/gcov.texi: Document how working directory is printed.
8012 * gcov-dump.c (dump_gcov_file): Print PWD.
8013 * gcov.c (output_intermediate_file): Likewise.
8014 (read_graph_file): Read PWD string.
8015 (output_lines): Print PWD.
8017 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8020 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
8021 for retval and return false if all args to phi are zero.
8023 2018-05-18 Richard Biener <rguenther@suse.de>
8025 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
8027 (evrp_dom_walker::before_dom_children): Call it.
8029 2018-05-18 Richard Biener <rguenther@suse.de>
8031 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
8032 results when processing array refs with variable index.
8034 2018-05-18 Toon Moene <toon@moene.org>
8036 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
8037 directly after that of -floop-interchange. Indicate that both
8038 options are enabled by default when specifying -O3.
8040 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8042 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
8043 iterator. Delete separate integer-mode vec_set<mode> expander.
8044 (aarch64_simd_vec_setv2di): Delete.
8045 (vec_setv2di): Delete.
8046 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
8047 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
8048 the "w, r" alternative.
8050 2018-05-18 Martin Liska <mliska@suse.cz>
8052 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
8053 * tree-pass.h (make_pass_lower_switch_O0): New function.
8054 * tree-switch-conversion.c (node_has_low_bound): Remove.
8055 (node_has_high_bound): Likewise.
8056 (node_is_bounded): Likewise.
8057 (class pass_lower_switch): Make it a template type and create
8059 (pass_lower_switch::execute): Add template argument.
8060 (make_pass_lower_switch): New function.
8061 (make_pass_lower_switch_O0): New function.
8062 (do_jump_if_equal): Remove.
8063 (emit_case_nodes): Simplify to just handle all 3 cases and leave
8064 all the hard work to tree optimization passes.
8066 2018-05-18 Martin Liska <mliska@suse.cz>
8068 * dbgcnt.c (limit_low): Renamed from limit.
8069 (limit_high): New variable.
8070 (dbg_cnt_is_enabled): Check for upper limit.
8071 (dbg_cnt): Adjust dumping.
8072 (dbg_cnt_set_limit_by_index): Add new argument for high
8074 (dbg_cnt_set_limit_by_name): Likewise.
8075 (dbg_cnt_process_single_pair): Parse new format.
8076 (dbg_cnt_process_opt): Use strtok.
8077 (dbg_cnt_list_all_counters): Remove 'value' and add
8079 * doc/invoke.texi: Document changes.
8081 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
8083 * doc/sourcebuild.texi (scalar_all_fma): Document.
8084 * tree.def (FMA_EXPR): Delete.
8085 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
8086 * internal-fn.c (ternary_direct): New macro.
8087 (expand_ternary_optab_fn): Likewise.
8088 (direct_ternary_optab_supported_p): Likewise.
8089 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
8090 * builtins.c (fold_builtin_fma): Delete.
8091 (fold_builtin_3): Don't call it.
8092 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
8093 * expr.c (expand_expr_real_2): Likewise.
8094 * fold-const.c (operand_equal_p): Likewise.
8095 (fold_ternary_loc): Likewise.
8096 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
8097 * gimple.c (DEFTREECODE): Likewise.
8098 * gimplify.c (gimplify_expr): Likewise.
8099 * optabs-tree.c (optab_for_tree_code): Likewise.
8100 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
8101 * tree-eh.c (operation_could_trap_p): Likewise.
8102 (stmt_could_throw_1_p): Likewise.
8103 * tree-inline.c (estimate_operator_cost): Likewise.
8104 * tree-pretty-print.c (dump_generic_node): Likewise.
8105 (op_code_prio): Likewise.
8106 * tree-ssa-loop-im.c (stmt_cost): Likewise.
8107 * tree-ssa-operands.c (get_expr_operands): Likewise.
8108 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
8109 * fold-const-call.h (fold_fma): Delete.
8110 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
8111 CFN_FNMA and CFN_FNMS.
8113 * genmatch.c (combined_fn): New enum.
8114 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
8115 (commutative_op): New function.
8116 (commutate): Use it. Handle more than 2 operands.
8117 (dt_operand::gen_gimple_expr): Use commutative_op.
8118 (parser::parse_expr): Allow :c to be used with non-binary
8119 operators if the commutative operand is known.
8120 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
8121 CFN_FMS, CFN_FNMA and CFN_FNMS.
8122 (backprop::process_assign_use): Remove FMA_EXPR handling.
8123 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
8124 (gen_hsa_fma): New function.
8125 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
8126 IFN_FNMA and IFN_FNMS.
8127 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
8128 * gimple-fold.h (follow_all_ssa_edges): Declare.
8129 * gimple-fold.c (follow_all_ssa_edges): New function.
8130 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
8131 gimple_build interface and use follow_all_ssa_edges to fold the result.
8132 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
8133 instead of checking for optabs directly.
8134 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
8135 rather than FMA_EXPRs.
8136 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
8137 call to IFN_FMA instead of an FMA_EXPR.
8139 2018-05-17 Jim Wilson <jimw@sifive.com>
8141 * expr.c (do_tablejump): When converting index to Pmode, if we have a
8142 sign extended promoted subreg, and the range does not have the sign bit
8143 set, then do a sign extend.
8145 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
8146 test, check for sign extended subreg and/or constant operands, and
8147 do a sign extend in that case.
8149 2018-05-17 Steve Ellcey <sellcey@cavium.com>
8151 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
8152 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
8154 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
8155 Change logics_shift_reg to logics_shift_imm.
8156 (thunderx2t99_fp_loadpair_basic): Delete.
8157 (thunderx2t99_fp_storepair_basic): Delete.
8158 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
8159 (thunderx2t99_asimd_polynomial): Delete.
8160 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
8161 and neon_fp_mul_d_scalar_q.
8162 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
8163 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
8164 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
8165 (thunderx2t99_asimd_lut): Add missing tbl types.
8166 (thunderx2t99_asimd_ext): Delete.
8167 (thunderx2t99_asimd_load1_1_mult): Delete.
8168 (thunderx2t99_asimd_load1_2_mult): Delete.
8169 (thunderx2t99_asimd_load1_ldp): New.
8170 (thunderx2t99_asimd_load1): New.
8171 (thunderx2t99_asimd_load2): Add missing *load2* types.
8172 (thunderx2t99_asimd_load3): New.
8173 (thunderx2t99_asimd_load4): New.
8174 (thunderx2t99_asimd_store1_1_mult): Delete.
8175 (thunderx2t99_asimd_store1_2_mult): Delete.
8176 (thunderx2t99_asimd_store2_mult): Delete.
8177 (thunderx2t99_asimd_store2_onelane): Delete.
8178 (thunderx2t99_asimd_store_stp): New.
8179 (thunderx2t99_asimd_store1): New.
8180 (thunderx2t99_asimd_store2): New.
8181 (thunderx2t99_asimd_store3): New.
8182 (thunderx2t99_asimd_store4): New.
8184 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
8186 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
8187 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
8189 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
8190 Segher Boessenkool <segher@kernel.crashing.org>
8193 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
8196 2018-05-17 Richard Biener <rguenther@suse.de>
8198 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
8199 for pruning loop and prune defs feeding only already visited PHIs.
8201 2018-05-17 Richard Biener <rguenther@suse.de>
8203 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
8205 2018-05-17 Bin Cheng <bin.cheng@arm.com>
8206 Richard Biener <rguenther@suse.de>
8208 PR tree-optimization/85793
8209 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
8210 for VMAT_ELEMENTWISE.
8212 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
8214 * internal-fn.h (lookup_internal_fn): Declare
8215 * internal-fn.c (lookup_internal_fn): New function.
8216 * gimple.c (gimple_build_call_from_tree): Handle calls to
8218 * gimple-pretty-print.c (dump_gimple_call): Print "." before
8219 internal function names.
8220 * tree-pretty-print.c (dump_generic_node): Likewise.
8221 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
8223 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
8225 * gimple-fold.h (gimple_build): Make the function forms take
8226 combined_fn rather than built_in_function.
8227 (gimple_simplify): Likewise.
8228 * gimple-match-head.c (gimple_simplify): Likewise.
8229 * gimple-fold.c (gimple_build): Likewise.
8230 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
8231 rather than gimple_build_call_internal.
8232 (get_initial_defs_for_reduction): Likewise.
8233 (vect_create_epilog_for_reduction): Likewise.
8234 (vectorizable_live_operation): Likewise.
8236 2018-05-17 Martin Liska <mliska@suse.cz>
8238 * gimple-ssa-sprintf.c (format_directive): Do not use
8239 space in between 'G_' and '('.
8241 2018-05-17 Jakub Jelinek <jakub@redhat.com>
8244 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
8245 even if the mask is not all ones.
8248 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
8250 (ix86_gimple_fold_builtin): Likewise.
8253 * config/i386/i386.c: Include tree-vector-builder.h.
8254 (ix86_vector_shift_count): New function.
8255 (ix86_fold_builtin): Fold shift builtins by scalar count.
8256 (ix86_gimple_fold_builtin): Likewise.
8258 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
8259 _mm512_setzero): New intrinsics.
8261 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
8262 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8264 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
8265 code generation for cases where splatting a value is not useful.
8266 * simplify-rtx.c (simplify_ternary_operation): Simplify
8267 vec_merge across a vec_duplicate and a paradoxical subreg forming
8268 a vector mode to a vec_concat.
8270 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
8272 * config.gcc: Support "goldmont-plus".
8273 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
8275 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
8276 PROCESSOR_GOLDMONT_PLUS.
8277 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
8278 (processor_target_table): Add "goldmont-plus".
8279 (PTA_GOLDMONT_PLUS): Define.
8280 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
8281 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
8282 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
8283 (fold_builtin_cpu): Add "goldmont-plus".
8284 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
8285 (ix86_option_override_internal): Add "goldmont-plus".
8286 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
8287 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
8288 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
8289 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
8291 2018-05-17 Richard Biener <rguenther@suse.de>
8293 PR tree-optimization/85757
8294 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
8295 remove defs that only feed that PHI from further processing.
8297 2018-05-16 Jim Wilson <jimw@sifive.com>
8299 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
8301 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
8303 2018-05-16 Mark Wielaard <mark@klomp.org>
8305 * dwarf2out.c (count_index_strings): New function.
8306 (output_indirect_strings): Call count_index_strings and generate
8307 header for dwarf_version >= 5.
8309 2018-05-16 Mark Wielaard <mark@klomp.org>
8311 * dwarf2out.c (dwarf_FORM): New function.
8312 (set_indirect_string): Use dwarf_FORM.
8313 (reset_indirect_string): Likewise.
8314 (size_of_die): Likewise.
8315 (value_format): Likewise.
8316 (output_die): Likewise.
8317 (add_skeleton_AT_string): Likewise.
8318 (output_macinfo_op): Likewise.
8319 (index_string): Likewise.
8320 (output_index_string_offset): Likewise.
8321 (output_index_string): Likewise.
8322 (count_index_strings): Likewise.
8324 2018-05-16 Carl Love <cel@us.ibm.com>
8326 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
8327 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
8329 2018-05-16 Martin Jambor <mjambor@suse.cz>
8331 * ipa-prop.c (ipa_free_all_edge_args): Remove.
8332 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
8334 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
8336 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
8337 (fnma<mode>4): Likewise.
8338 (fms<mode>4): Likewise.
8339 (fnms<mode>4): Likewise.
8340 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
8341 (aarch64_fnma<mode>4): Likewise.
8342 (aarch64_fms<mode>4): Likewise.
8343 (aarch64_fnms<mode>4): Likewise.
8344 (aarch64_fnmadd<mode>4): Likewise.
8346 2018-05-16 Jason Merrill <jason@redhat.com>
8348 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
8350 2018-05-16 Richard Biener <rguenther@suse.de>
8352 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
8353 (dump_stmt_cost): Declare.
8354 (add_stmt_cost): Dump cost we add.
8355 (add_stmt_costs): New function.
8356 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
8358 (vect_analyze_stmt): Adjust prototype.
8359 (vectorizable_condition): Likewise.
8360 (vectorizable_live_operation): Likewise.
8361 (vectorizable_reduction): Likewise.
8362 (vectorizable_induction): Likewise.
8363 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
8364 cost vector to pass to vectorizable_ and record afterwards.
8365 (vect_model_reduction_cost): Take cost vector argument and adjust.
8366 (vect_model_induction_cost): Likewise.
8367 (vectorizable_reduction): Likewise.
8368 (vectorizable_induction): Likewise.
8369 (vectorizable_live_operation): Likewise.
8370 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
8371 SLP_TREE_NUMBER_OF_VEC_STMTS.
8372 (vect_analyze_slp_cost_1): Remove.
8373 (vect_analyze_slp_cost): Likewise.
8374 (vect_slp_analyze_node_operations): Take visited args and
8375 a target cost vector. Avoid processing already visited stmt sets.
8376 (vect_slp_analyze_operations): Use a local cost vector to gather
8377 costs and register those of non-discarded instances.
8378 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
8379 (vect_schedule_slp_instance): Remove copying of
8380 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
8382 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
8383 adding cost. Record cost entry location.
8384 (vect_prologue_cost_for_slp_op): Function to compute cost of
8385 a constant or invariant generated for SLP vect in the prologue,
8386 split out from vect_analyze_slp_cost_1.
8387 (vect_model_simple_cost): Make static. Adjust for SLP costing.
8388 (vect_model_promotion_demotion_cost): Likewise.
8389 (vect_model_store_cost): Likewise, make static.
8390 (vect_model_load_cost): Likewise.
8391 (vectorizable_bswap): Add cost vector arg and adjust.
8392 (vectorizable_call): Likewise.
8393 (vectorizable_simd_clone_call): Likewise.
8394 (vectorizable_conversion): Likewise.
8395 (vectorizable_assignment): Likewise.
8396 (vectorizable_shift): Likewise.
8397 (vectorizable_operation): Likewise.
8398 (vectorizable_store): Likewise.
8399 (vectorizable_load): Likewise.
8400 (vectorizable_condition): Likewise.
8401 (vectorizable_comparison): Likewise.
8402 (can_vectorize_live_stmts): Likewise.
8403 (vect_analyze_stmt): Likewise.
8404 (vect_transform_stmt): Adjust calls to vectorizable_*.
8405 * tree-vectorizer.c: Include gimple-pretty-print.h.
8406 (dump_stmt_cost): New function.
8408 2018-05-16 Richard Biener <rguenther@suse.de>
8410 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
8411 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
8412 * tree-ssa-dse.c: Include tree-ssa-loop.h.
8413 (check_name): New callback.
8414 (dse_classify_store): Track cycles via a visited bitmap of PHI
8415 defs and simplify handling of in-loop and across loop dead stores
8416 and properly fail for loop-variant refs. Handle byte-tracking with
8417 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
8420 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
8422 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
8423 (vect_get_mask_type_for_stmt): Likewise.
8424 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
8426 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
8427 to determine the statement's vector type and the vector type that
8428 should be used for calculating nunits. Deal with cases in which
8429 the type has to be deferred.
8430 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
8431 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
8432 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
8433 (vect_determine_vf_for_stmt): New functions, split out from...
8434 (vect_determine_vectorization_factor): ...here.
8435 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
8436 (vect_get_mask_type_for_stmt): New functions, split out from
8437 vect_determine_vectorization_factor.
8439 2018-05-16 Richard Biener <rguenther@suse.de>
8441 * tree-cfg.c (verify_gimple_assign_ternary): Properly
8442 verify the [VEC_]COND_EXPR embedded comparison.
8444 2018-05-15 Martin Sebor <msebor@redhat.com>
8446 PR tree-optimization/85753
8447 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
8448 RECORD_TYPE in addition to ARRAY_TYPE.
8450 2018-05-15 Martin Sebor <msebor@redhat.com>
8453 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
8455 2018-05-15 Richard Biener <rguenther@suse.de>
8457 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
8458 add by_clobber_p one. Change algorithm to collect all defs
8459 representing uses we need to walk and try reducing them to
8460 a single one before failing.
8461 (dse_dom_walker::dse_optimize_stmt): Adjust.
8463 2018-05-13 Mark Wielaard <mark@klomp.org>
8465 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
8466 (size_of_loc_descr): Likewise.
8467 (output_loc_operands): Likewise.
8468 (output_loc_operands_raw): Likewise.
8469 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
8470 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
8471 (hash_loc_operands): Likewise.
8472 (compare_loc_operands): Likewise.
8474 2018-05-14 Mark Wielaard <mark@klomp.org>
8476 * dwarf2out.c (count_index_addrs): New function.
8477 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
8479 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8481 PR tree-optimization/83648
8482 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
8483 return value as malloc candidate.
8485 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8488 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
8489 param as true in call to suggest_attribute.
8491 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
8493 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
8494 -mreadonly-in-sdata.
8496 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8498 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
8500 (aarch64_crypto_aesd_fused): Likewise.
8502 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
8504 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
8505 (movsi_aarch64): Likewise.
8506 (load_pairsi): Likewise.
8507 (load_pairdi): Likewise.
8508 (store_pairsi): Likewise.
8509 (store_pairdi): Likewise.
8510 (load_pairsf): Likewise.
8511 (load_pairdf): Likewise.
8512 (store_pairsf): Likewise.
8513 (store_pairdf): Likewise.
8514 (zero_extend): Likewise.
8515 (trunc): Swap alternatives.
8516 (fcvt_target): Add '?' to prefer w over r.
8518 2018-05-14 Jakub Jelinek <jakub@redhat.com>
8521 * config/i386/i386.md: Disallow non-commutative arithmetics in
8522 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
8523 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
8524 in the peephole2 before it.
8526 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
8528 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
8529 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
8530 (ix86_handle_option): Handle -mcldemote.
8531 * config.gcc: New header.
8532 * config/i386/cldemoteintrin.h: New file.
8533 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
8534 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
8536 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
8537 OPTION_MASK_ISA_CLDEMOTE.
8538 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
8539 (ix86_valid_target_attribute_inner_p): Ditto.
8540 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
8541 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
8542 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
8543 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
8544 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
8546 * config/i386/i386.opt: Add -mcldemote.
8547 * config/i386/x86intrin.h: New header.
8548 * doc/invoke.texi: Add -mcldemote.
8550 2018-05-14 Richard Biener <rguenther@suse.de>
8552 * doc/match-and-simplify.texi: Adjust :s documentation.
8554 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
8556 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
8557 intended memcpy size.
8558 (REORDER_45): Likewise.
8560 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
8562 * sort.cc: New file.
8563 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
8564 * vec.c (qsort_chk): Use gcc_qsort.
8565 * Makefile.in (OBJS-libcommon): Add sort.o.
8566 (build/sort.o): New target. Use it...
8567 (BUILD_RTL): ... here, and...
8568 (build/gencfn-macros): ... here, and...
8569 (build/genmatch): ... here.
8571 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
8572 Chung-Ju Wu <jasonwucj@gmail.com>
8574 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
8575 * config/nds32/nds32-graywolf.md: New file.
8576 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
8577 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
8579 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
8580 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
8581 * config/nds32/nds32.md (pipeline_model): Add graywolf.
8582 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
8583 * config/nds32/pipelines.md: Include n15 settings.
8585 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
8586 Chung-Ju Wu <jasonwucj@gmail.com>
8588 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
8589 * config/nds32/nds32-n13.md: New file.
8590 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
8591 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
8593 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
8594 * config/nds32/nds32.md (pipeline_model): Add n13.
8595 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
8596 * config/nds32/pipelines.md: Include n13 settings.
8598 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
8599 Chung-Ju Wu <jasonwucj@gmail.com>
8601 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
8602 * config/nds32/nds32-n10.md: New file.
8603 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
8604 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
8606 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
8607 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
8608 * config/nds32/nds32.md (pipeline_model): Add n10.
8609 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
8610 * config/nds32/pipelines.md: Include n10 settings.
8612 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
8613 Kito Cheng <kito.cheng@gmail.com>
8614 Chung-Ju Wu <jasonwucj@gmail.com>
8616 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
8617 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
8618 Add enum values for DSP extension instructions.
8619 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
8621 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
8622 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
8624 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
8625 * config/nds32/nds32-dspext.md: New file for DSP implementation.
8626 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
8627 * config/nds32/nds32-intrinsic.md: Likewise.
8628 * config/nds32/nds32_intrinsic.h: Likewise.
8629 * config/nds32/nds32-md-auxiliary.c: Likewise.
8630 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
8631 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
8632 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
8633 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
8634 * config/nds32/nds32-protos.h: New declarations for DSP extension.
8635 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
8636 TYPE_DMAC in switch statement.
8637 * config/nds32/nds32.c: New checking and implementation for DSP
8638 extension instructions.
8639 * config/nds32/nds32.h: Likewise.
8640 * config/nds32/nds32.md: Likewise.
8641 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
8642 * config/nds32/predicates.md: Implement new predicates for DSP
8645 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
8647 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
8648 Reformat alternatives and attributes so it is easier to identify
8649 which constraints/attributes go with which instruction.
8650 (mov<mode>_hardfloat32, FMOVE64): Likewise.
8651 (mov<mode>_softfloat32, FMOVE64): Likewise.
8652 (mov<mode>_hardfloat64, FMOVE64): Likewise.
8653 (mov<mode>_softfloat64, FMOVE64): Likewise.
8655 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
8657 * doc/extend.texi (PowerPC Built-in Functions): Rename this
8659 (Basic PowerPC Built-in Functions): The new name of the
8660 subsection previously known as "PowerPC Built-in Functions".
8661 (Basic PowerPC Built-in Functions Available on all Configurations):
8663 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
8664 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
8665 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
8666 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
8668 2018-05-11 Martin Jambor <mjambor@suse.cz>
8671 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
8674 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
8677 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
8679 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
8681 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
8682 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
8683 (ix86_handle_option): Handle -mwaitpkg.
8684 * config.gcc: New header.
8685 * config/i386/cpuid.h (bit_WAITPKG): New bit.
8686 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
8687 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
8689 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
8690 OPTION_MASK_ISA_WAITPKG.
8691 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
8692 (ix86_option_override_internal): Add PTA_WAITPKG.
8693 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
8694 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
8695 IX86_BUILTIN_TPAUSE.
8696 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
8697 __builtin_ia32_umwait and __builtin_ia32_tpause.
8698 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
8699 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
8700 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
8701 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
8702 UNSPECV_TPAUSE): New.
8703 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
8704 * config/i386/i386.opt: Add -mwaitpkg.
8705 * config/i386/waitpkgintrin.h: New file.
8706 * config/i386/x86intrin.h: New header.
8707 * doc/invoke.texi: Add -mwaitpkg.
8709 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
8712 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
8714 (cortex-m0): Use armv6s-m isa.
8715 (cortex-m0plus): Likewise.
8716 (cortex-m1): Likewise.
8717 (cortex-m0.small-multiply): Likewise.
8718 (cortex-m0plus.small-multiply): Likewise.
8719 (cortex-m1.small-multiply): Likewise.
8721 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
8722 Jakub Jelinek <jakub@redhat.com>
8724 PR tree-optimization/85692
8725 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
8726 source permute as well.
8728 2018-05-11 Martin Liska <mliska@suse.cz>
8731 * doc/extend.texi: Document LLVM style format for no_sanitize
8734 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
8736 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
8737 mode_supports_vsx_dform_quad to mode_supports_dq_form.
8738 (mode_supports_vsx_dform_quad): Likewise.
8739 (mode_supports_vmx_dform): Move these functions to be next to the
8740 other mode_supports functions.
8741 (mode_supports_dq_form): Likewise.
8742 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
8743 mode_supports_dq_form.
8744 (reg_offset_addressing_ok_p): Likewise.
8745 (offsettable_ok_by_alignment): Likewise.
8746 (rs6000_legitimate_offset_address_p): Likewise.
8747 (legitimate_lo_sum_address_p): Likewise.
8748 (rs6000_legitimize_address): Likewise.
8749 (rs6000_legitimize_reload_address): Likewise.
8750 (rs6000_secondary_reload_inner): Likewise.
8751 (rs6000_preferred_reload_class): Likewise.
8752 (rs6000_output_move_128bit): Likewise.
8754 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
8756 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
8757 Generate SImode target register for null target.
8758 <case IX86_BUILTIN_XGETBV>: Ditto.
8759 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
8760 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
8762 2018-05-10 Carl Love <cel@us.ibm.com>
8764 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
8765 dcbtt and dcbtstt if operands[2] is 0.
8767 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
8770 * config/i386/sse.md (usadv64qi): New expander.
8772 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
8774 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
8775 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
8776 -maltivec=be support.
8777 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
8778 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
8779 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
8780 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
8781 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
8782 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
8783 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
8784 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
8785 altivec_vsumsws): Adjust.
8786 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
8787 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
8788 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
8790 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
8791 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
8792 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
8793 (altivec_lve<VI_char>x): Delete expand.
8794 (*altivec_lve<VI_char>x_internal): Rename to...
8795 (altivec_lve<VI_char>x): ... this.
8796 (altivec_lvxl_<mode>): Delete expand.
8797 (*altivec_lvxl_<mode>_internal): Rename to ...
8798 (altivec_lvxl_<mode>): ... this.
8799 (altivec_stvxl_<mode>): Delete expand.
8800 (*altivec_stvxl_<mode>_internal): Rename to ...
8801 (altivec_stvxl_<mode>): ... this.
8802 (altivec_stve<VI_char>x): Delete expand.
8803 (*altivec_stve<VI_char>x_internal): Rename to ...
8804 (altivec_stve<VI_char>x): ... this.
8805 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
8806 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
8807 reduc_plus_scal_<mode>): Adjust.
8808 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
8810 (rs6000_cpu_cpp_builtins): Adjust.
8811 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
8812 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
8813 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
8814 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
8815 -maltivec=be support.
8816 (rs6000_split_vec_extract_var): Adjust.
8817 (rs6000_split_v4si_init): Adjust.
8818 (swap_selector_for_mode): Delete.
8819 (altivec_expand_lvx_be, altivec_expand_stvx_be,
8820 altivec_expand_stvex_be): Delete.
8821 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
8822 -maltivec=be support.
8823 (rs6000_gimple_fold_builtin): Ditto.
8824 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
8826 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
8827 (TARGET_DIRECT_MOVE_64BIT): Adjust.
8828 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
8829 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
8830 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
8831 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
8832 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
8833 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
8834 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
8835 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
8836 anonymous split): Adjust.
8837 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
8838 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
8840 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
8842 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
8843 when --with-gxx-include-dir is also specified.
8844 * configure: Regenerate.
8846 2018-05-09 Jim Wilson <jimw@sifive.com>
8849 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
8850 * config/riscv/t-withmultilib: New.
8851 * config/riscv/withmultilib.h: New.
8852 * doc/install.texi: Document RISC-V --with-multilib-list support.
8854 2018-05-09 Richard Biener <rguenther@suse.de>
8856 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
8858 (vect_bb_vectorization_profitable_p): Adjust. Compute
8859 actual scalar cost using the cost vector and the add_stmt_cost
8862 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
8864 PR rtl-optimization/85645
8865 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
8866 in the REG_CFA_REGISTER note for LR, don't leave it empty.
8868 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
8870 PR rtl-optimization/85645
8871 * shrink-wrap.c (spread_components): Return a boolean saying if
8872 anything was changed.
8873 (try_shrink_wrapping_separate): Iterate spread_components until
8874 nothing changes anymore.
8876 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
8878 PR rtl-optimization/85645
8879 * regrename.c (build_def_use): Also kill the chains that include the
8880 destination of a REG_CFA_REGISTER note.
8882 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
8884 PR rtl-optimization/85645
8885 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
8886 insn that has a REG_CFA_REGISTER note.
8888 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
8890 * cfgexpand.c (expand_clobber): New function.
8891 (expand_gimple_stmt_1): Use it.
8892 * tree-vect-stmts.c (vect_clobber_variable): New function,
8894 (vectorizable_simd_clone_call): ...here.
8895 (vectorizable_store): Emit a clobber either side of an
8896 IFN_STORE_LANES sequence.
8897 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
8899 2018-05-09 Tom de Vries <tom@codesourcery.com>
8902 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
8903 (define_insn "trap_if_false"): Add exit after trap.
8905 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
8907 PR rtl-optimization/85638
8908 * bb-reorder.c: Include common/common-target.h.
8909 (create_forwarder_block): New function extracted from...
8910 (fix_up_crossing_landing_pad): ...here. Rename into...
8911 (dw2_fix_up_crossing_landing_pad): ...this.
8912 (sjlj_fix_up_crossing_landing_pad): New function.
8913 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
8914 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
8915 from both partitions and exit the loop after one iteration.
8917 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
8920 * doc/extend.texi (PowerPC Built-in Functions): Rename this
8922 (Basic PowerPC Built-in Functions): The new name of the
8923 subsection previously known as "PowerPC Built-in Functions".
8924 (Basic PowerPC Built-in Functions Available on all Configurations):
8926 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
8928 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
8930 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
8932 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
8935 2018-05-08 Jim Wilson <jimw@sifive.com>
8937 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
8938 (LD_EMUL_SUFFIX): New.
8939 (LINK_SPEC): Use it.
8941 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
8943 * doc/extend.texi (PowerPC Built-in Functions): Rename this
8945 (Basic PowerPC Built-in Functions): The new name of the
8946 subsection previously known as "PowerPC Built-in Functions".
8947 (Basic PowerPC Built-in Functions Available on all Configurations):
8949 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
8951 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
8953 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
8955 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
8958 2018-05-08 Jakub Jelinek <jakub@redhat.com>
8961 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
8962 after cmpelim optimization.
8964 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
8966 * config.gcc: Support "goldmont".
8967 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
8968 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
8970 * config/i386/i386.c (m_GOLDMONT): Define.
8971 (processor_target_table): Add "goldmont".
8972 (PTA_GOLDMONT): Define.
8973 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
8974 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
8975 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
8976 (fold_builtin_cpu): Add "goldmont".
8977 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
8978 (ix86_option_override_internal): Add "goldmont".
8979 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
8980 (processor_type): Add PROCESSOR_GOLDMONT.
8981 * config/i386/i386.md: Add CPU "glm".
8982 * config/i386/glm.md: New file.
8983 * config/i386/x86-tune.def: Add m_GOLDMONT.
8984 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
8986 2018-05-08 Jakub Jelinek <jakub@redhat.com>
8989 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
8991 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
8992 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
8993 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
8996 * config/i386/i386.c (ix86_fold_builtin): Handle
8997 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
9000 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
9001 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
9003 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
9006 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
9007 (check_arch): Likewise.
9008 (check_fpu): Return the result rather than printing it.
9009 (end arch): Fix operator precedence.
9010 (end cpu): Likewise.
9011 (END): Print the result from check_fpu.
9013 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
9014 Alan Hayward <alan.hayward@arm.com>
9015 David Sherwood <david.sherwood@arm.com>
9017 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
9018 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
9019 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
9020 (*fcmuo<mode>_and): New patterns.
9022 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
9024 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
9025 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
9026 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
9027 (cmp_op, sve_imm_con): New code attributes.
9028 (SVE_COND_INT_CMP, imm_con): Delete.
9029 (cmp_op): Remove above unspecs from int attribute.
9030 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
9032 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
9033 comparison-specific unspecs.
9034 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
9035 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
9036 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
9037 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
9038 (*vec_fcm<cmp_op><mode>): Rename to...
9039 (*fcm<cmp_op><mode>): ...this and adjust likewise.
9040 (*vec_fcmuo<mode>): Rename to...
9041 (*fcmuo<mode>): ...this and adjust likewise.
9042 (*pred_fcm<cmp_op><mode>): New pattern.
9043 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
9044 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
9046 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
9048 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
9049 (aarch64_emit_sve_predicated_cond): New function.
9050 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
9051 (aarch64_emit_unspec_cond_or): Replace with...
9052 (aarch64_emit_sve_or_conds): ...this new function. Use
9053 aarch64_emit_sve_ptrue_op for the individual comparisons and
9054 aarch64_emit_binop to OR them together.
9055 (aarch64_emit_inverted_unspec_cond): Replace with...
9056 (aarch64_emit_sve_inverted_cond): ...this new function. Use
9057 aarch64_emit_sve_ptrue_op for the comparison and
9058 aarch64_emit_unop to invert the result.
9059 (aarch64_expand_sve_vec_cmp_float): Update after the above
9060 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
9062 2018-05-07 Nathan Sidwell <nathan@acm.org>
9064 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
9065 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
9066 (Backwards Compatibility): Likewise.
9068 2018-05-07 Luis Machado <luis.machado@linaro.org>
9072 2018-05-07 Luis Machado <luis.machado@linaro.org>
9074 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9075 <prefetch_dynamic_strides>: New const bool field.
9076 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9077 prefetch_dynamic_strides.
9078 (exynosm1_prefetch_tune): Likewise.
9079 (thunderxt88_prefetch_tune): Likewise.
9080 (thunderx_prefetch_tune): Likewise.
9081 (thunderx2t99_prefetch_tune): Likewise.
9082 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
9084 (aarch64_override_options_internal): Update to set
9085 PARAM_PREFETCH_DYNAMIC_STRIDES.
9086 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
9087 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
9088 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
9089 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
9090 prefetch-dynamic-strides setting.
9092 2018-05-07 Luis Machado <luis.machado@linaro.org>
9094 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9095 <minimum_stride>: New const int field.
9096 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9097 minimum_stride field.
9098 (exynosm1_prefetch_tune): Likewise.
9099 (thunderxt88_prefetch_tune): Likewise.
9100 (thunderx_prefetch_tune): Likewise.
9101 (thunderx2t99_prefetch_tune): Likewise.
9102 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
9103 (aarch64_override_options_internal): Update to set
9104 PARAM_PREFETCH_MINIMUM_STRIDE.
9105 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
9106 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
9107 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
9108 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
9109 stride is constant and is below the minimum stride threshold.
9111 2018-05-07 Luis Machado <luis.machado@linaro.org>
9113 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
9116 2018-05-07 Luis Machado <luis.machado@linaro.org>
9118 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9119 <prefetch_dynamic_strides>: New const bool field.
9120 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9121 prefetch_dynamic_strides.
9122 (exynosm1_prefetch_tune): Likewise.
9123 (thunderxt88_prefetch_tune): Likewise.
9124 (thunderx_prefetch_tune): Likewise.
9125 (thunderx2t99_prefetch_tune): Likewise.
9126 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
9128 (aarch64_override_options_internal): Update to set
9129 PARAM_PREFETCH_DYNAMIC_STRIDES.
9130 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
9131 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
9132 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
9133 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
9134 prefetch-dynamic-strides setting.
9136 2018-05-07 Luis Machado <luis.machado@linaro.org>
9138 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9139 <minimum_stride>: New const int field.
9140 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9141 minimum_stride field.
9142 (exynosm1_prefetch_tune): Likewise.
9143 (thunderxt88_prefetch_tune): Likewise.
9144 (thunderx_prefetch_tune): Likewise.
9145 (thunderx2t99_prefetch_tune): Likewise.
9146 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
9147 (aarch64_override_options_internal): Update to set
9148 PARAM_PREFETCH_MINIMUM_STRIDE.
9149 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
9150 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
9151 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
9152 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
9153 stride is constant and is below the minimum stride threshold.
9155 2018-05-06 Jakub Jelinek <jakub@redhat.com>
9158 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
9159 the type is addressable. Don't force op into register if it has
9162 2018-05-05 Roland McGrath <mcgrathr@google.com>
9165 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
9166 any section for which we don't know a specific type it should have,
9167 regardless of name. Previously this was done only for the exact
9168 names ".init_array", ".fini_array", and ".preinit_array".
9169 (default_elf_asm_named_section): Add comment about
9170 relationship with default_section_type_flags and SECTION_NOTYPE.
9171 (get_section): Don't consider it a type conflict if one side has
9172 SECTION_NOTYPE and the other doesn't, as long as neither has the
9173 SECTION_BSS et al used in the default_section_type_flags logic.
9175 2018-05-05 Tom de Vries <tom@codesourcery.com>
9178 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
9179 (workaround_barsyncs): New function.
9180 (nvptx_reorg): Use workaround_barsyncs.
9181 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
9182 (define_expand "nvptx_membar_cta"): New define_expand.
9183 (define_insn "*nvptx_membar_cta"): New insn.
9185 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
9187 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
9188 To improve optimization opportunities.
9189 * builtin-types.def: The new needed builtin types for the above.
9191 2018-05-04 Richard Biener <rguenther@suse.de>
9193 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
9194 * gimple-ssa-store-merging.c
9195 (imm_store_chain_info::output_merged_store): Remove redundant create,
9196 release split_store vector contents on failure.
9197 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
9198 scalar stmt vector on cache hit.
9200 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
9202 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
9204 * config.gcc (powerpc-xilinx-eabi*): Remove.
9205 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
9207 (fusion_addis_mem_combo_load): Ditto.
9208 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
9210 (rs6000_cpu_cpp_builtins): Ditto.
9211 * config/rs6000/rs6000-linux.c
9212 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
9213 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
9214 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
9216 (rs6000_setup_reg_addr_masks): Ditto.
9217 (rs6000_init_hard_regno_mode_ok): Ditto.
9218 (rs6000_option_override_internal): Ditto.
9219 (legitimate_lo_sum_address_p): Ditto.
9220 (rs6000_legitimize_address): Ditto.
9221 (rs6000_legitimize_reload_address): Ditto.
9222 (rs6000_legitimate_address_p): Ditto.
9223 (abi_v4_pass_in_fpr): Ditto.
9224 (setup_incoming_varargs): Ditto.
9225 (rs6000_gimplify_va_arg): Ditto.
9226 (rs6000_split_multireg_move): Ditto.
9227 (rs6000_savres_strategy): Ditto.
9228 (rs6000_emit_prologue_components): Ditto.
9229 (rs6000_emit_epilogue_components): Ditto.
9230 (rs6000_emit_prologue): Ditto.
9231 (rs6000_emit_epilogue): Ditto.
9232 (rs6000_elf_file_end): Ditto.
9233 (rs6000_function_value): Ditto.
9234 (rs6000_libcall_value): Ditto.
9235 * config/rs6000/rs6000.h: Ditto.
9236 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
9237 (TARGET_MINMAX): ... this. New.
9238 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
9239 * config/rs6000/rs6000.md: Remove Xilinx FP support.
9240 (*movsi_internal1_single): Delete.
9241 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
9242 mfpu=, mxilinx-fpu): Delete.
9243 * config/rs6000/singlefp.h: Delete.
9244 * config/rs6000/sysv4.h: Remove Xilinx FP support.
9245 * config/rs6000/t-rs6000: Ditto.
9246 * config/rs6000/t-xilinx: Delete.
9247 * config/rs6000/titan.md: Adjust for fp_type removal.
9248 * config/rs6000/vsx.md: Remove Xilinx FP support.
9249 (VStype_simple): Delete.
9250 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
9251 * config/rs6000/xfpu.h: Delete.
9252 * config/rs6000/xfpu.md: Delete.
9253 * config/rs6000/xilinx.h: Delete.
9254 * config/rs6000/xilinx.opt: Delete.
9255 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
9256 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
9258 2018-05-04 Tom de Vries <tom@codesourcery.com>
9261 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
9264 2018-05-04 Richard Biener <rguenther@suse.de>
9267 * tree-complex.c (update_complex_assignment): We are always in SSA form.
9268 (expand_complex_div_wide): Likewise.
9269 (expand_complex_operations_1): Likewise.
9270 (expand_complex_libcall): Preserve EH info of the original stmt.
9271 (tree_lower_complex): Handle removed blocks.
9272 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
9273 on complex multiplication and division libcall builtins.
9275 2018-05-04 Richard Biener <rguenther@suse.de>
9278 * fold-const.c (negate_expr_p): Restrict negation of operand
9279 zero of a division to when we know that can happen without
9281 (fold_negate_expr_1): Likewise.
9283 2018-05-04 Jakub Jelinek <jakub@redhat.com>
9286 * real.h (real_nextafter): Declare.
9287 * real.c (real_nextafter): New function.
9288 * fold-const-call.c (fold_const_nextafter): New function.
9289 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
9290 CASE_CFN_NEXTTOWARD.
9291 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
9292 even when arg1_mode is different from arg0_mode.
9294 2018-05-03 Nathan Sidwell <nathan@acm.org>
9296 * doc/extend.texi (Deprecated Features): Remove
9298 (Backwards Compatibility): Likewise.
9299 * doc/invoke.texi (C++ Language Options): Likewise.
9300 (C++ Dialect Options): Likewise.
9302 2018-05-03 Jakub Jelinek <jakub@redhat.com>
9305 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
9306 _mm512_mask_mullox_epi64): New intrinsics.
9308 2018-05-03 Tom de Vries <tom@codesourcery.com>
9311 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
9312 dump files): Add offload-tree.
9314 2018-05-03 Richard Biener <rguenther@suse.de>
9316 PR tree-optimization/85615
9317 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
9318 to loops not nested in BBs loop father to avoid creating multi-entry
9321 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9323 PR tree-optimization/70291
9324 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
9325 arguments. Change return type to tree. Emit libcall as a new
9326 statement rather than replacing existing one when inplace_p is true.
9327 (expand_complex_multiplication_components): New function.
9328 (expand_complex_multiplication): Expand floating-point complex
9329 multiplication using the above.
9330 (expand_complex_division): Rename inner_type parameter to type.
9331 Update expand_complex_libcall call-site.
9332 (expand_complex_operations_1): Update expand_complex_multiplication
9333 and expand_complex_division call-sites.
9335 2018-05-02 Jakub Jelinek <jakub@redhat.com>
9338 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
9339 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
9340 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
9341 the highest significant bit of the shift count mask is clear. In
9342 check whether and[sq]i3 is needed verify that all significant bits
9343 of the shift count other than the highest are set.
9345 2018-05-02 Tom de Vries <tom@codesourcery.com>
9348 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
9349 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
9350 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
9351 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
9352 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
9353 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
9354 __builtin_goacc_parlevel_size.
9356 2018-05-02 Richard Biener <rguenther@suse.de>
9358 PR tree-optimization/85597
9359 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
9360 do not use split vect_get_vec_defs call but call vect_get_slp_defs
9363 2018-05-02 Tom de Vries <tom@codesourcery.com>
9366 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
9367 dump files): Add ltrans-tree.
9369 2018-05-02 Tom de Vries <tom@codesourcery.com>
9372 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
9373 dump files): Add wpa-ipa.
9375 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
9377 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
9378 powerpc*-*-linux*paired* target.
9379 * config/rs6000/750cl.h: Delete.
9380 * config/rs6000/paired.h: Delete.
9381 * config/rs6000/paired.md: Delete.
9382 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
9384 * config/rs6000/rs6000-builtin.def: Remove paired float support.
9385 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
9386 comment. Remove paired float support.
9387 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
9388 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
9390 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
9391 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
9393 * config/rs6000/rs6000.c: Remove paired float support.
9394 (paired_expand_vector_init, paired_expand_vector_move,
9395 paired_emit_vector_compare, paired_emit_vector_cond_expr,
9396 (paired_expand_lv_builtin, paired_expand_stv_builtin,
9397 paired_expand_builtin, paired_expand_predicate_builtin,
9398 paired_init_builtins): Delete.
9399 * config/rs6000/rs6000.h: Remove paired float support.
9400 * config/rs6000/rs6000.md: Remove paired float support.
9401 (move_from_CR_ov_bit): Delete.
9402 * config/rs6000/rs6000.opt (mpaired): Delete.
9403 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
9404 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
9406 2018-05-02 Richard Biener <rguenther@suse.de>
9409 * gimplify.c (gimplify_save_expr): When in SSA form allow
9410 SAVE_EXPRs to compute to SSA vars.
9412 2018-05-02 Jakub Jelinek <jakub@redhat.com>
9415 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
9416 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
9417 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
9418 clobber operands[2], instead use a new pseudo. Formatting fixes.
9420 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
9422 PR tree-optimization/85586
9423 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
9424 exit early for statements in the same group if the accesses are
9427 2018-05-02 Tom de Vries <tom@codesourcery.com>
9430 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
9433 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
9435 PR tree-optimization/85143
9436 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
9438 2018-05-01 Tom de Vries <tom@codesourcery.com>
9441 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
9442 not found" error message.
9444 2018-05-01 Tom de Vries <tom@codesourcery.com>
9447 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
9448 * vec.c (test_ordered_remove_if): New function.
9449 (vec_c_tests): Call test_ordered_remove_if.
9450 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
9451 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
9452 * tree-vect-patterns.c (vect_pattern_recog_1): Use
9453 VEC_ORDERED_REMOVE_IF.
9455 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9457 PR tree-optimization/82665
9458 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
9459 pointer subtraction where arguments come from a memchr call.
9461 2018-05-01 Jakub Jelinek <jakub@redhat.com>
9463 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
9464 --push-state --as-needed and --pop-state instead of --as-needed and
9465 --no-as-needed if ld supports it.
9466 * configure: Regenerated.
9469 * doc/install.texi2html: Replace _002d with - and _002a with * in
9470 generated html files using sed.
9472 2018-04-30 David Malcolm <dmalcolm@redhat.com>
9475 * gcc-rich-location.c (blank_line_before_p): New function.
9476 (use_new_line): New function.
9477 (gcc_rich_location::add_fixit_insert_formatted): New function.
9478 * gcc-rich-location.h
9479 (gcc_rich_location::add_fixit_insert_formatted): New function.
9481 2018-04-30 David Malcolm <dmalcolm@redhat.com>
9483 * selftest.c (assert_streq): Rename "expected" and "actual" to
9484 "val1" and "val2". Extend NULL-handling to cover both inputs
9485 symmetrically, while still requiring both to be non-NULL for a pass.
9486 * selftest.h (assert_streq): Rename "expected" and "actual" to
9488 (ASSERT_EQ): Likewise.
9489 (ASSERT_EQ_AT): Likewise.
9490 (ASSERT_KNOWN_EQ): Likewise.
9491 (ASSERT_KNOWN_EQ_AT): Likewise.
9492 (ASSERT_NE): Likewise.
9493 (ASSERT_MAYBE_NE): Likewise.
9494 (ASSERT_MAYBE_NE_AT): Likewise.
9495 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
9496 the assertion to pass.
9497 (ASSERT_STREQ_AT): Likewise.
9499 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
9501 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
9502 interaction with -pie.
9504 2018-04-30 David Malcolm <dmalcolm@redhat.com>
9506 * selftest.h: Fix alphabetization of per-source-file selftest
9509 2018-04-30 Jason Merrill <jason@redhat.com>
9511 PR c++/61982 - dead stores to destroyed objects.
9512 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
9515 2018-04-30 Jason Merrill <jason@redhat.com>
9517 * tree.c (build_clobber): New.
9518 * tree.h: Declare it.
9519 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
9521 2018-04-30 David Malcolm <dmalcolm@redhat.com>
9523 * diagnostic-show-locus.c (layout::layout): Update for
9524 location_get_source_line returning a char_span.
9525 (struct char_span): Move to input.h.
9526 (struct correction): Update for fields in char_span becoming
9528 (struct source_line): Update for location_get_source_line
9529 returning a char_span.
9530 (layout::print_line): Likewise.
9531 * edit-context.c (edited_file::print_content): Likewise.
9532 (edited_file::print_diff_hunk): Likewise.
9533 (edited_file::print_run_of_changed_lines): Likewise.
9534 (edited_file::get_num_lines): Likewise.
9535 (edited_line::edited_line): Likewise.
9536 * final.c (asm_show_source): Likewise.
9537 * input.c (location_get_source_line): Convert return type
9538 from const char * to char_span, losing the final "line_len"
9540 (dump_location_info): Update for the above.
9541 (get_substring_ranges_for_loc): Likewise. Use a char_span
9542 when handling the literal within the line.
9543 (test_reading_source_line): Update for location_get_source_line
9544 returning a char_span.
9545 * input.h (class char_span): Move here from
9546 diagnostic-show-locus.c, converting from a struct to a class.
9547 Make data members private.
9548 (char_span::operator bool): New.
9549 (char_span::length): New.
9550 (char_span::get_buffer): New.
9551 (char_span::operator[]): New.
9552 (char_span::subspan): Make const.
9553 (char_span::xstrdup): New.
9554 (location_get_source_line): Convert return type from const char *
9555 to char_span, losing the final "line_size" param.
9557 2018-04-30 Jan Hubicka <jh@suse.cz>
9559 * lto-wrapper.c (ltrans_priorities): New static var.
9560 (cmp_priority): New.
9561 (run_gcc): Read priorities and if doing parallel build order
9562 the Makefile by them.
9564 2018-04-30 David Malcolm <dmalcolm@redhat.com>
9566 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
9568 2018-04-30 Richard Biener <rguenther@suse.de>
9570 * tree-cfg.c (verify_address): Remove base argument, add
9571 flag whether to check TREE_ADDRESSABLE and do that.
9572 (verify_expr): Remove.
9573 (verify_types_in_gimple_reference): Add pieces from verify_expr.
9574 (verify_gimple_assign_single): Likewise.
9575 (verify_gimple_switch): Likewise.
9576 (verify_expr_location_1): Dereference tp once. Add (disabled)
9577 piece from verify_expr.
9578 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
9580 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
9582 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
9584 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
9586 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
9587 (small_data_pattern): Likewise.
9588 (arc_rewrite_small_data): Likewise.
9589 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
9590 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
9591 (get_symbol_alignment): New function.
9592 (legitimate_small_data_address_p): Likewise.
9593 (legitimate_scaled_address): Update, call
9594 legitimate_small_data_address_p.
9595 (output_sdata): New static variable.
9596 (arc_print_operand): Update how we handle small data operands.
9597 (arc_print_operand_address): Likewise.
9598 (arc_legitimate_address_p): Update, use
9599 legitimate_small_data_address_p.
9600 (arc_rewrite_small_data_p): Remove.
9601 (arc_rewrite_small_data_1): Likewise.
9602 (arc_rewrite_small_data): Likewise.
9603 (small_data_pattern): Likewise.
9604 (compact_sda_memory_operand): Update to use
9605 legitimate_small_data_address_p and get_symbol_alignment.
9606 (prepare_move_operands): Don't rewite sdata pattern.
9607 (prepare_extend_operands): Remove.
9608 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
9610 (zero_extendqisi2): Likewise.
9611 (zero_extendhisi2): Likewise.
9612 (extendqihi2): Likewise.
9613 (extendqisi2): Likewise.
9614 (extendhisi2): Likewise.
9618 * config/arc/constraints.md (Usd): Change it to memory constraint.
9620 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
9622 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
9623 as source of std instructions.
9624 * config/arc/arc.md (movsi_insn): Update pattern predicate to
9625 allow 6-bit constants as source for store instructions.
9626 (movdi_insn): Update instruction pattern to allow 6-bit constants
9627 as source for store instructions.
9629 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
9631 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
9633 2018-04-30 Nathan Sidwell <nathan@acm.org>
9634 Sandra Loosemore <sandra@codesourcery.com>
9636 * dumpfile.c (dump_open): Allow '-' for stdout.
9637 * doc/invoke.texi (Developer Options): Document dump filename
9638 determination early. Document stdin/stdout selection.
9640 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
9642 Microblaze Target: PIC data text relative
9644 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
9645 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
9647 * config/microblaze/microblaze.h (microblaze_constant_address_p):
9648 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
9649 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
9650 New addressing mode for data-text relative position indepenedent code.
9651 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
9652 'ADDRESS_SYMBOLIC_TXT_REL'.
9653 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
9654 (microblaze_legitimate_pic_operand): Exclude function calls from
9655 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
9656 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
9658 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
9659 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
9660 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
9661 for 'address + offset'.
9662 (microblaze_expand_prologue): Add new function prologue call for
9664 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
9665 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
9666 table in case of TARGET_PIC_DATA_TEXT_REL.
9667 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
9668 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
9669 Add new macros 'UNSPEC_TEXT',
9670 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
9671 + exclude function calls from 'UNSPEC_PLT' in case of data text
9673 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
9674 new target hook for generating address diff vector tables in case of
9676 * doc/tm.texi : Regenerate.
9677 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
9678 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
9679 of addr diff vector generation.
9680 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
9681 target hook definition.
9682 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
9683 Add default function for generate_pic_addr_diff_vec -> flag_pic.
9684 * doc/invoke.texi (Add new pic option): Add new microblaze pic
9685 option for data text relative.
9687 2018-04-30 Richard Biener <rguenther@suse.de>
9689 * tree-chrec.h (evolution_function_is_constant_p): Remove
9691 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
9693 2018-04-30 Richard Biener <rguenther@suse.de>
9696 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
9698 2018-04-30 Richard Biener <rguenther@suse.de>
9700 PR tree-optimization/28364
9701 PR tree-optimization/85275
9702 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
9703 copying first exit test.
9705 2018-04-28 Mark Wielaard <mark@klomp.org>
9707 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
9709 (dwarf_AT): Handle DW_AT_addr_base.
9710 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
9712 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
9715 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
9716 (*ashl<dwi>3_doubleword_mask_1): Ditto.
9717 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
9718 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
9720 2018-04-28 Richard Biener <rguenther@suse.de>
9722 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
9723 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
9724 to reflect use. Only add interesting stmts.
9726 2018-04-27 Martin Jambor <mjambor@suse.cz>
9729 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
9730 the jump function allows for passing through aggregate values.
9732 2018-04-27 David Malcolm <dmalcolm@redhat.com>
9734 * input.h (in_system_header_at): Convert from macro to inline
9736 (from_macro_expansion_at): Likewise.
9737 (from_macro_definition_at): Likewise.
9739 2018-04-27 Jeff Law <law@redhat.com>
9741 * config.gcc: Mark tile* targets as deprecated/obsolete.
9743 2018-04-27 Richard Biener <rguenther@suse.de>
9745 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
9748 2018-04-27 Richard Biener <rguenther@suse.de>
9750 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
9752 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
9754 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
9755 with Yd constraint. Set "preferred_for_speed" attribute from
9756 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
9758 (*movdi_internal): Ditto.
9759 (movti_interunit splitters): Remove
9760 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
9761 (movdi_interunit splitters): Ditto.
9762 * config/i386/constraints.md (Ye): Remove.
9763 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
9765 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9768 * config/aarch64/constraints.md (Usg): Limit to 31.
9771 2018-04-27 Jakub Jelinek <jakub@redhat.com>
9773 PR tree-optimization/85529
9774 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
9775 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
9776 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
9777 zero extension or masking of the MSB bit.
9778 (optimize_range_tests): Add FIRST_BB argument, pass it through
9779 to optimize_range_tests_var_bound.
9780 (maybe_optimize_range_tests, reassociate_bb): Adjust
9781 optimize_range_tests callers.
9783 2018-04-26 Richard Biener <rguenther@suse.de>
9784 Jakub Jelinek <jakub@redhat.com>
9786 * cgraph.h (symbol_table): Just declare debug method here.
9787 * symtab.c (symbol_table::debug): Define.
9789 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
9791 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
9793 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
9795 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
9796 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
9797 (*movdi_internal): Substitute Yi and Yj constraint with x
9798 and Ym and Yn constraint with y constraint. Update "isa"
9799 attribute and set "preferred_for_speed" attribute from
9800 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
9801 (*movsi_internal): Ditto.
9802 (*movdf_internal): Ditto.
9803 (*movsf_internal): Ditto.
9804 (*zero_extendsidi2): Ditto.
9805 * config/i386/sse.md (vec_set<mode>_0): Ditto.
9806 (sse2_loadld): Ditto.
9807 (*vec_extract<ssevecmodelower>_0): Ditto.
9808 (*vec_extractv4si_0_zext_sse4): Ditto.
9809 (vec_concatv2di): Ditto.
9810 (*vec_dup<mode>): Ditto.
9811 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
9812 * config/i386/constraints.md (Yi): Remove.
9817 2018-04-26 Nathan Sidwell <nathan@acm.org>
9819 * dumpfile.c (dump_open): New.
9820 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
9821 (dump_finish): Detect stdio/stderr by value not name.
9823 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
9825 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
9827 2018-04-26 Tom de Vries <tom@codesourcery.com>
9830 * config/nvptx/nvptx.c (verify_neutering_jumps)
9831 (verify_neutering_labels): New function
9832 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
9834 2018-04-26 Tom de Vries <tom@codesourcery.com>
9837 * config/nvptx/nvptx.c (needs_neutering_p): New function.
9838 (nvptx_single): Use needs_neutering_p to skip over insns that do not
9841 2018-04-26 Richard Biener <rguenther@suse.de>
9842 Tom de Vries <tom@codesourcery.com>
9845 * lto-streamer-out.c (output_function): Fixup loops if required to match
9846 discovery done in the reader.
9848 2018-04-26 Richard Biener <rguenther@suse.de>
9850 PR tree-optimization/85116
9851 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
9852 have a loop exit from the single latch predecessor. Remove
9853 case of header with just condition.
9854 (ch_base::copy_headers): Exclude infinite loops from any
9856 (pass_ch::execute): Record exits.
9858 2018-04-26 Richard Biener <rguenther@suse.de>
9860 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
9861 prologue cost vector and pass it to vect_get_load_cost.
9862 (vect_get_peeling_costs_all_drs): Likewise.
9863 (vect_peeling_hash_get_lowest_cost): Likewise.
9864 (vect_enhance_data_refs_alignment): Likewise.
9866 2018-04-26 Richard Biener <rguenther@suse.de>
9869 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
9870 checking of integer<->pointer conversions.
9871 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
9872 sign-/zero-extending pointer types.
9873 (expand_omp_for_static_chunk): Likewise.
9875 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
9876 Jean Lee <xiaoyur347@gmail.com>
9878 * config/mips/mips.c (mips_asan_shadow_offset): New function.
9879 (TARGET_ASAN_SHADOW_OFFSET): Define.
9880 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
9881 true for -fsanitize=address.
9883 2018-04-25 Mark Wielaard <mark@klomp.org>
9885 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
9888 2018-04-25 Jakub Jelinek <jakub@redhat.com>
9890 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
9891 than "alu", remove explicit "memory" and "imm_disp" attributes.
9892 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
9895 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
9896 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
9897 gen_lowpart_no_emit.
9899 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
9902 * config/i386/i386.c (ix86_expand_builtin): Change memory
9903 operand to XI, extend p0 to Pmode.
9904 * config/i386/i386.md: Change unspec volatile and operand
9905 1 mode to XI, change operand 0 mode to P.
9907 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
9909 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
9910 GET_MODE_MASK before any checking.
9911 (nds32_can_use_bset_p): Likewise.
9912 (nds32_can_use_btgl_p): Likewise.
9914 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
9916 * config/nds32/nds32-doubleword.md: New define_split pattern for
9917 illegal register number.
9919 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
9921 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
9923 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
9925 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
9927 2018-04-25 Richard Biener <rguenther@suse.de>
9929 * lto-streamer.h (LTO_major_version): Bump to 8.
9931 2018-04-25 Jakub Jelinek <jakub@redhat.com>
9933 * BASE-VER: Set to 9.0.0.
9935 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
9937 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
9938 in __abskf2 and __powikf2.
9940 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9943 * config/aarch64/constraints.md (Usg, Usj): New constraints.
9944 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
9945 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
9946 Use the above on operand 2. Reindent.
9947 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
9949 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
9952 * common/config/i386/i386-common.c (ix86_handle_option): Don't
9954 * config/i386/i386.opt (mcet): Removed.
9955 * doc/install.texi: Remove -mcet documentation.
9956 * doc/invoke.texi: Likewise.
9958 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
9961 * doc/install.texi: Remove -mcet from bootstrap-cet.
9963 2018-04-24 Jakub Jelinek <jakub@redhat.com>
9966 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
9967 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
9971 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
9972 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
9973 containing a CONST_VECTOR.
9975 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
9977 * doc/install.texi: Update newlib dependency for nvptx.
9979 2018-04-24 Jakub Jelinek <jakub@redhat.com>
9982 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
9983 instead of INTVAL when shifting x left.
9985 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
9987 PR tree-optimization/85478
9988 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
9989 vect_grouped_store_supported for single element vectors.
9991 2018-04-24 Richard Biener <rguenther@suse.de>
9994 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
9995 load cost increase to the case of non-constant step.
9997 2018-04-24 Jakub Jelinek <jakub@redhat.com>
10000 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
10001 destination if any_malformed_asm.
10003 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
10005 PR middle-end/85496
10006 * expr.c (store_field): In the bitfield case, if the value comes from
10007 a function call and is returned in registers by means of a PARALLEL,
10008 do not change the mode of the temporary unless BLKmode and VOIDmode.
10010 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
10012 PR rtl-optimization/85423
10013 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
10014 dependencies to debug insns when the previous insn is non-debug.
10016 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
10018 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
10019 enums into a single definition.
10020 (fls): Fix predicates and printing.
10023 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
10025 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
10026 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
10027 and short u6 immediate.
10028 (check_if_valid_sleep_operand): Remove.
10029 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
10031 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
10033 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
10034 flag_always_save_lp condition.
10035 * config/nds32/nds32.opt (malways-save-lp): New option.
10037 2018-04-22 Shiva Chen <shiva0217@gmail.com>
10039 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
10040 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
10041 * config/nds32/nds32.h
10042 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
10043 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
10045 2018-04-22 Shiva Chen <shiva0217@gmail.com>
10047 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
10048 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
10050 2018-04-22 Shiva Chen <shiva0217@gmail.com>
10051 Chung-Ju Wu <jasonwucj@gmail.com>
10053 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
10055 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
10056 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
10058 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
10060 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
10062 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
10064 * config/nds32/nds32-protos.h (nds32_data_alignment,
10065 nds32_local_alignment): Declare.
10066 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
10067 nds32_local_alignment): New functions.
10068 (TARGET_CONSTANT_ALIGNMENT): Define.
10069 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
10071 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
10073 * config/nds32/nds32.c
10074 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
10075 (TARGET_MODES_TIEABLE_P): Likewise.
10077 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
10079 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
10080 level Ofast and Og.
10082 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
10083 Chung-Ju Wu <jasonwucj@gmail.com>
10085 * config/nds32/constants.md (unspec_volatile_element): Add enum values
10086 for unaligned access.
10087 * config/nds32/nds32-intrinsic.c: Implementation of expanding
10089 * config/nds32/nds32-intrinsic.md: Likewise.
10090 * config/nds32/nds32_intrinsic.h: Likewise.
10091 * config/nds32/nds32.h (nds32_builtins): Likewise.
10092 * config/nds32/nds32.opt (munaligned-access): New option.
10093 * config/nds32/nds32.c (nds32_asm_file_start): Display
10094 flag_unaligned_access status.
10096 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
10098 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
10099 -mno-relax is present.
10100 * config/riscv/linux.h (LINK_SPEC): Ditto.
10102 2018-04-20 Martin Sebor <msebor@redhat.com>
10105 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
10107 (gimple_fold_builtin_stxcpy_chk): Same.
10108 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
10110 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
10113 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
10114 __powikf2 when long double is IEEE 128-bit.
10116 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
10118 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
10119 step to make sure stack always aligned.
10121 2018-04-20 Carl Love <cel@us.ibm.com>
10124 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
10125 size check for arg0.
10127 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
10128 Tom de Vries <tom@codesourcery.com>
10131 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
10132 Emit insns for calls too.
10133 (nvptx_find_par): Always look for worker-level predecessor insn.
10134 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
10136 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
10137 (nvptx_process_pars): Propagate frames for calls.
10139 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
10142 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
10144 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
10145 (ix86_handle_option): Don't handle OPT_mibt.
10146 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
10148 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
10150 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
10151 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
10152 (ix86_target_macros): Define __CET__ with flag_cf_protection
10153 for -fcf-protection.
10154 * config/i386/i386.c (isa2_opts): Remove -mibt.
10155 * config/i386/i386.h (TARGET_IBT): Removed.
10156 (TARGET_IBT_P): Likewise.
10157 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
10158 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
10159 * config/i386/i386.opt (mcet): Update help message.
10160 (mshstk): Likewise.
10162 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
10163 -mcet as an alias for -mshstk.
10165 2018-04-20 Richard Biener <rguenther@suse.de>
10167 PR middle-end/85475
10168 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
10169 complexity by forcing a single use of the multiply operand.
10171 2018-04-20 Martin Jambor <mjambor@suse.cz>
10174 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
10175 recursion dependency to only apply to non-clones.
10177 2018-04-20 Martin Jambor <mjambor@suse.cz>
10180 * ipa-cp.c (create_specialized_node): Check that clones of
10181 self-recursive edges exist during IPA-CP.
10183 2018-04-19 Toon Moene <toon@moene.org>
10185 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
10188 2018-04-19 Jakub Jelinek <jakub@redhat.com>
10190 PR tree-optimization/85467
10191 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
10192 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
10193 VECTOR_CST element to type.
10195 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
10198 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
10199 * config/i386/i386.md (builtin_setjmp_setup): Removed.
10200 (builtin_longjmp): Likewise.
10201 (save_stack_nonlocal): New pattern.
10202 (restore_stack_nonlocal): Likewise.
10204 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
10207 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
10208 Replace ASM_OUTPUT_LABEL with fprintf.
10210 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
10213 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
10214 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
10215 * config/i386/i386-c.c (ix86_target_macros_internal): Also
10216 define __IBT__ and __SHSTK__ for -fcf-protection.
10217 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
10219 (ix86_trampoline_init): Likewise.
10220 (x86_output_mi_thunk): Likewise.
10221 (ix86_notrack_prefixed_insn_p): Likewise.
10222 (ix86_option_override_internal): Don't disallow -fcf-protection.
10223 * config/i386/i386.md (rdssp<mode>): Also enable for
10225 (incssp<mode>): Likewise.
10226 (nop_endbr): Likewise.
10227 * config/i386/i386.opt (mcet): Change help message to built-in
10230 (mshstk): Likewise.
10231 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
10232 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
10233 enable CET built-in functions.
10235 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
10237 * common/config/i386/i386-common.c
10238 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
10239 OPTION_MASK_ISA_MOVDIRI_UNSET,
10240 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
10241 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
10242 * config.gcc (movdirintrin.h): New header.
10243 * config/i386/cpuid.h (bit_MOVDIRI,
10244 bit_MOVDIR64B): New bits.
10245 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
10247 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
10248 (VOID, PVOID, PCVOID)): New function types.
10249 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
10250 __builtin_ia32_directstoreu_u64,
10251 __builtin_ia32_movdir64b): New builtins.
10252 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
10253 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
10255 (ix86_valid_target_attribute_inner_p): Ditto.
10256 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
10257 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
10258 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
10259 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
10260 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
10261 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
10262 (movdiri<mode>, movdir64b_<mode>): New.
10263 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
10264 * config/i386/immintrin.h: Include movdirintrin.h.
10265 * config/i386/movdirintrin.h: New file.
10266 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
10268 2018-04-19 Richard Biener <rguenther@suse.de>
10270 PR middle-end/85455
10271 * cfg.c (clear_bb_flags): When loop state says we have
10272 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
10274 2018-04-19 Richard Biener <rguenther@suse.de>
10276 PR tree-optimization/84737
10277 * tree-vect-data-refs.c (vect_copy_ref_info): New function
10278 copying restrict info.
10279 (vect_setup_realignment): Use it.
10280 * tree-vectorizer.h (vect_copy_ref_info): Declare.
10281 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
10282 the first DR to all generated stores.
10283 (vectorizable_load): Likewise for loads.
10285 2018-04-19 Jakub Jelinek <jakub@redhat.com>
10287 PR tree-optimization/85446
10288 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
10289 the integral and pointer types to have the same precision.
10291 * doc/install.texi: Document --disable-cet being the default and
10294 2018-04-18 Martin Liska <mliska@suse.cz>
10296 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
10299 2018-04-18 Martin Liska <mliska@suse.cz>
10302 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
10305 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
10306 arguments if they are comparable.
10308 2018-04-18 Martin Liska <mliska@suse.cz>
10311 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
10314 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
10317 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
10320 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
10321 ENDBR after calling __morestack.
10323 2018-04-18 David Malcolm <dmalcolm@redhat.com>
10326 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
10327 by using gcc_base_ver to generate a gcc_driver_version, and use
10328 it when generating GCC_DRIVER_NAME.
10329 * configure: Regenerate.
10331 2018-04-18 Jakub Jelinek <jakub@redhat.com>
10334 * config.gcc: Obsolete powerpc*-*-*spe*.
10336 2018-04-17 Jakub Jelinek <jakub@redhat.com>
10339 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
10340 (stabstr_D): Change type of unum from unsigned int to
10341 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
10344 2018-04-17 Jim Wilson <jimw@sifive.com>
10347 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
10348 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
10349 Set arg_pointer_offset after using pretend_args_size.
10351 2018-04-17 Jakub Jelinek <jakub@redhat.com>
10353 PR rtl-optimization/85431
10354 * dse.c (record_store): Ignore zero width stores.
10357 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
10358 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
10359 __builtin_stack_restore rather than after it.
10360 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
10361 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
10362 argument instead of virtual_dynamic_stack_rtx.
10364 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
10366 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
10368 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
10369 Add note to error message to explain internal mapping of overloaded
10370 built-in function name to non-overloaded built-in function name.
10371 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
10374 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
10377 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
10378 where the inputs overlap with the output.
10380 2018-04-17 Jakub Jelinek <jakub@redhat.com>
10382 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
10383 (=v, v) alternative and explicit "memory" attribute.
10384 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
10385 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
10387 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
10388 "sselog1" type instead of "sselog".
10389 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
10390 "sselog". Remove explicit "memory" attribute.
10391 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
10392 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
10394 (vec_extract_hi_v32hi): Merge all alternatives into one, use
10395 "sselog1" type instead of "sselog". Remove explicit "memory"
10397 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
10398 use "sselog1" type instead of "sselog". Remove explicit "memory"
10400 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
10401 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
10403 (vec_extract_hi_v64qi): Merge all alternatives into one, use
10404 "sselog1" type instead of "sselog". Remove explicit "memory"
10406 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
10407 use "sselog1" type instead of "sselog". Remove explicit "memory"
10411 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
10413 PR middle-end/85414
10414 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
10417 2018-04-17 Martin Jambor <mjambor@suse.cz>
10420 * ipa-cp.c (create_specialized_node): Call
10421 expand_all_artificial_thunks if necessary.
10423 2018-04-17 Martin Liska <mliska@suse.cz>
10426 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
10427 in message, remote space in between '_G' and '('.
10429 2018-04-17 Jakub Jelinek <jakub@redhat.com>
10432 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
10433 avx512f_vmcmp<mode>3<round_saeonly_name>,
10434 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
10435 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
10436 avx512f_rndscale<mode><round_saeonly_name>,
10437 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
10438 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
10439 Use %<iptr>2 instead of %2 for -masm=intel.
10440 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
10441 avx512f_vcvttss2usi<round_saeonly_name>,
10442 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
10444 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
10445 avx512f_vcvttsd2usi<round_saeonly_name>,
10446 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
10447 Use %q1 instead of %1 for -masm=intel.
10448 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
10449 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
10450 of %3 for -masm=intel.
10451 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
10453 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
10455 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
10457 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
10458 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
10460 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
10462 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
10463 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
10464 %g1 and one with %0 and %1.
10465 (avx512er_vmrcp28<mode><round_saeonly_name>,
10466 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
10467 %1 for -masm=intel.
10468 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
10469 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
10470 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
10471 of %0 and %{%4%} for -masm=intel.
10472 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
10473 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
10474 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
10475 order of %0 and %{%5%}%{z%} for -masm=intel.
10477 2018-04-17 Jan Hubicka <jh@suse.cz>
10480 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
10482 2018-04-17 Martin Liska <mliska@suse.cz>
10485 * multiple_target.c (create_dispatcher_calls): Set apostrophes
10486 for target_clone error message. Make default implementation
10487 clone to be a local declaration.
10488 (separate_attrs): Add new argument and check for an empty
10490 (expand_target_clones): Handle it.
10491 (ipa_target_clone): Make redirection just for target_clones
10494 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
10495 Tom de Vries <tom@codesourcery.com>
10497 PR middle-end/84955
10498 * omp-expand.c (expand_oacc_for): Add dummy false branch for
10499 tiled basic blocks without omp continue statements.
10501 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
10504 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
10505 vec_extract expression as having side effects to make sure it gets
10508 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
10511 * config/i386/i386.c (get_builtin_code_for_version): Check
10514 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
10517 * config.gcc: Support "skylake".
10518 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10520 * config/i386/i386.c (m_SKYLAKE): Define.
10521 (processor_target_table): Add "skylake".
10522 (ix86_option_override_internal): Add "skylake".
10523 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
10524 PROCESSOR_CANNONLAKE.
10525 (get_builtin_code_for_version): Fix priority for
10526 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
10527 PROCESSOR_SKYLAKE-AVX512.
10528 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
10529 (processor_type): Add PROCESSOR_SKYLAKE.
10531 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
10532 Jason Merrill <jason@redhat.com>
10535 * convert.c (convert_to_integer_1): Use direct recursion for
10536 enumeral types and types with a precision less than the number
10537 of bits in their mode.
10539 2018-04-16 Julia Koval <julia.koval@intel.com>
10542 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
10543 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
10545 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
10548 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
10549 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
10550 and -mno-direct-move.
10552 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
10555 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
10556 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
10557 Ensure negative shifts result in {0}.
10559 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
10561 PR rtl-optimization/79916
10562 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
10563 regs (if any) to define how to gnerate SD moves when LRA is in
10566 2018-04-13 Jakub Jelinek <jakub@redhat.com>
10568 PR rtl-optimization/85393
10569 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
10570 * except.c (expand_dw2_landing_pad_for_region): Make static.
10571 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
10572 a label and unconditional jump to old_bb, rather than
10573 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
10576 PR rtl-optimization/85376
10577 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
10578 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
10579 instead of a specific value.
10581 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
10582 Bin Cheng <bin.cheng@arm.com>
10584 PR tree-optimization/82965
10585 PR tree-optimization/83991
10586 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
10587 by_profile_only parameter.
10588 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
10589 information if the loop was predicted to iterate too many times.
10590 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
10592 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
10595 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
10598 2018-04-13 Martin Liska <mliska@suse.cz>
10599 Jakub Jelinek <jakub@redhat.com>
10601 PR middle-end/81657
10602 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
10603 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
10604 * builtins.c (expand_builtin_memory_copy_args): Use
10605 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
10606 handle dest_addr == pc_rtx.
10608 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
10611 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
10612 asked to not generate direct moves.
10613 (fix_trunc<mode>si2_stfiwx): Similar.
10614 (fix_trunc<mode>si2_internal): Similar.
10616 2018-04-12 Jakub Jelinek <jakub@redhat.com>
10619 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
10620 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
10621 lookup if dest in some wider mode is known to be const0_rtx and
10622 if so, record permanent equivalence for it to be ZERO_EXTEND of
10623 the narrower mode destination.
10625 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
10627 * lto-streamer-out.c (output_function): Revert 259346.
10628 * omp-expand.c (expand_oacc_for): Likewise.
10630 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
10632 PR rtl-optimization/85354
10633 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
10634 * sel-sched.c (sel_global_init): ... here.
10636 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
10639 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
10640 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
10641 mode for PE-COFF targets.
10642 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
10643 (i386_pe_asm_lto_end): Likewise.
10644 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
10645 (TARGET_ASM_LTO_END): Likewise.
10646 * config/i386/winnt.c (saved_debug_info_level): New static variable.
10647 (i386_pe_asm_lto_start): New function.
10648 (i386_pe_asm_lto_end): Likewise.
10650 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
10651 Richard Biener <rguenther@suse.de>
10653 PR middle-end/84955
10654 * lto-streamer-out.c (output_function): Fix CFG loop state before
10656 * omp-expand.c (expand_oacc_for): Handle calls to internal
10657 functions like regular functions.
10659 2018-04-12 Richard Biener <rguenther@suse.de>
10662 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
10663 for the early LTO debug to properly generate references to it
10664 during DIE emission. Do not re-use that for the skeleton for
10666 (dwarf2out_early_finish): Likewise.
10668 2018-04-12 Jakub Jelinek <jakub@redhat.com>
10671 * config/i386/sse.md
10672 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
10673 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
10674 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
10675 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
10676 and output is a reg, avoid creating invalid lowpart subreg, but
10677 instead split into a 512-bit move. Don't split if not AVX512VL,
10678 input is xmm16+ reg and output is a mem.
10679 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
10680 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
10681 xmm16+ reg and output is a mem.
10683 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10685 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
10686 also for flag_dwarf2_cfi_asm.
10688 2018-04-12 Jakub Jelinek <jakub@redhat.com>
10690 PR rtl-optimization/85342
10691 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
10692 a bool scalar var inside of the loop instead. Don't try to update
10693 recog_data.operand after failed apply_change_group.
10695 2018-04-12 Tom de Vries <tom@codesourcery.com>
10698 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
10699 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
10700 array with flexible array member as array without given dimension.
10701 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
10702 argument for undefined param to true.
10704 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
10707 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
10708 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
10709 from PowerPC section.
10710 * config/rs6000/sysv4.opt (mcall-): Improve help text.
10711 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
10712 help text that is too long.
10713 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
10714 help text that is too long.
10715 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
10716 help text that is too long.
10718 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
10720 * config/alpha/alpha.md (stack_probe_internal): Rename
10721 from "probe_stack". Update all callers.
10723 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
10725 PR rtl-optimization/84566
10726 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
10727 sched_macro_fuse_insns.
10729 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
10732 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
10733 (compute_block_dependences): ... from here.
10735 2018-04-11 Jakub Jelinek <jakub@redhat.com>
10737 PR tree-optimization/85331
10738 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
10739 from int to HOST_WIDE_INT.
10741 2018-04-11 Martin Jambor <mjambor@suse.cz>
10744 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
10745 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
10746 not the same as the source val.
10747 (cgraph_edge_brings_value_p): New parameter.
10748 (gather_edges_for_value): Pass destination value to
10749 cgraph_edge_brings_value_p.
10750 (perhaps_add_new_callers): Likewise.
10751 (get_info_about_necessary_edges): Likewise and exclude values brought
10752 only by self-recursive edges.
10753 (create_specialized_node): Redirect only clones of self-calling edges.
10754 (+self_recursive_pass_through_p): New function.
10755 (find_more_scalar_values_for_callers_subset): Use it.
10756 (find_aggregate_values_for_callers_subset): Likewise.
10757 (known_aggs_to_agg_replacement_list): Removed.
10758 (decide_whether_version_node): Re-calculate known constants for all
10759 remaining context clones.
10761 2018-04-11 Richard Biener <rguenther@suse.de>
10764 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
10765 from early DWARF output.
10766 (dwarf2out_early_finish): Output line info unconditionally into
10767 early DWARF and add reference to it.
10769 2018-04-11 Jakub Jelinek <jakub@redhat.com>
10772 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
10773 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
10774 other than V2DFmode using iptr mode attribute.
10775 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
10777 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
10779 PR rtl-optimization/84659
10780 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
10782 2018-04-11 Jakub Jelinek <jakub@redhat.com>
10785 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
10787 (output_loc_list): Pass address of a dummy size variable even in the
10788 locview handling loop.
10789 (index_location_lists): Add comment on why skip_loc_list_entry can't
10792 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
10795 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
10798 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
10801 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
10802 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
10803 and -mstring-compare-inline-limit.
10805 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
10808 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
10809 for stack clash protection in a register whenever we need it to be in
10812 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
10814 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
10815 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
10817 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
10820 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
10822 (mlong-double-): Ditto.
10823 * config/rs6000/sysv4.opt (msdata=): Ditto.
10824 (mtls-size=): Ditto.
10826 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
10828 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
10829 erroneous entries for
10830 "vector int vec_ldl (int, long int *)", and
10831 "vector unsigned int vec_ldl (int, unsigned long int *)".
10832 Add comments and entries for
10833 "vector bool char vec_ldl (int, bool char *)",
10834 "vector bool short vec_ldl (int, bool short *)",
10835 "vector bool int vec_ldl (int, bool int *)",
10836 "vector bool long long vec_ldl (int, bool long long *)",
10837 "vector pixel vec_ldl (int, pixel *)",
10838 "vector long long vec_ldl (int, long long *)",
10839 "vector unsigned long long vec_ldl (int, unsigned long long *)".
10840 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
10841 type tree bool_long_long_type_node and correct definition of
10842 bool_V2DI_type_node to make reference to this new type tree.
10843 (rs6000_mangle_type): Replace erroneous reference to
10844 bool_long_type_node with bool_long_long_type_node.
10845 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
10846 comments to emphasize sign distinctions for char and int types and
10847 replace RS6000_BTI_bool_long constant with
10848 RS6000_BTI_bool_long_long constant. Also add comment to restrict
10849 use of RS6000_BTI_pixel.
10850 (bool_long_type_node): Remove this macro definition.
10851 (bool_long_long_type_node): New macro definition
10853 2018-04-10 Jakub Jelinek <jakub@redhat.com>
10855 PR rtl-optimization/85300
10856 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
10857 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
10858 simplify_unary_operation fails.
10860 2018-04-10 Martin Liska <mliska@suse.cz>
10862 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
10863 cgraph_edge and ipa_ref.
10865 2018-04-10 Jakub Jelinek <jakub@redhat.com>
10869 * config/i386/sse.md
10870 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
10871 computation of the VEC_MERGE selector from mask.
10872 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
10873 Fix decoding of the VEC_MERGE selector into mask.
10875 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
10877 PR tree-optimization/85286
10878 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
10880 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
10882 * final.c (final_1): Set insn_last_address as well as
10883 insn_current_address.
10885 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10888 * explow.c (emit_stack_probe): Call validize_mem on memory location
10889 before passing it to gen_probe_stack. Create address operand and
10890 legitimize it for the probe_stack_address case.
10892 2018-04-09 Jan Hubicka <jh@suse.cz>
10895 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
10896 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
10897 * tree.c (free_lang_data_in_type): Fix handling of binfos;
10899 (free_lang_data): Rebuild type inheritance graph.
10901 2018-04-09 Martin Sebor <msebor@redhat.com>
10903 * invoke.texi (-finline-small-functions): Mention other optimization
10905 (-findirect-inlining, -fpartial-inlining): Same.
10906 (-finline-functions-called-once): Same.
10907 (-freorder-blocks-and-partition): Same.
10909 2018-04-09 Jan Hubicka <jh@suse.cz>
10912 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
10913 jumps; choose last target that matches the criteria (i.e.
10914 no partition changes for non-crossing jumps).
10915 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
10916 support for redirecting crossing jumps to non-crossing.
10918 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
10920 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
10921 also for naked functions.
10923 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
10925 * config/arc/arc.md (add_shift): New pattern.
10926 (add_shift2): Likewise.
10927 (sub_shift): Likewise.
10928 (sub_shift_cmp0_noout): Likewise.
10929 (compare_si_ashiftsi): Likewise.
10930 (xbfu_cmp0_noout): New combine pattern.
10931 (xbfu_cmp0"): Likewise.
10932 (movsi_set_cc_insn): Place the predicable variant first.
10933 (commutative_binary_cmp0_noout): Remove clobber.
10934 (commutative_binary_cmp0): New pattern.
10935 (noncommutative_binary_cmp0): Likewise.
10936 (noncommutative_binary_cmp0_noout): Likewise.
10937 (noncommutative_binary_comparison_result_used): Removed.
10938 (rsub_cmp0): New pattern.
10939 (rsub_cmp0_noout): Likewise.
10940 (extzvsi): Changed, keep only meaningful variants.
10941 (SQH, SEZ): New iterators.
10942 (SQH_postfix): New mode attribute.
10943 (SEZ_prefix): New code attribute.
10944 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
10945 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
10946 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
10947 of numerical value.
10948 (noncommutative_operator): Check the availability of barrel
10951 2018-04-09 Richard Biener <rguenther@suse.de>
10953 PR tree-optimization/85284
10954 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
10955 Only use the niter constraining form of simple_iv when the exit
10956 is always executed.
10958 2018-04-09 Tom de Vries <tom@codesourcery.com>
10961 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
10962 (define_expand "*memory_barrier"): New define_expand.
10963 (define_insn "memory_barrier"): New insn.
10965 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
10967 PR rtl-optimization/80463
10968 PR rtl-optimization/83972
10969 PR rtl-optimization/83480
10971 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
10972 correct producer for the insn.
10973 (tidy_control_flow): Fixup seqnos in case of debug insns.
10975 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
10977 PR rtl-optimization/83913
10979 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
10980 different sched-times when merging exprs.
10982 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
10984 PR rtl-optimization/83962
10986 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
10987 tidy_fallthru_edge and tidy_control_flow.
10989 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
10991 PR rtl-optimization/83530
10993 * sel-sched.c (force_next_insn): New global variable.
10994 (remove_insn_for_debug): When force_next_insn is true, also leave only
10995 next insn in the ready list.
10996 (sel_sched_region): When the region wasn't scheduled, make another pass
10997 over it with force_next_insn set to 1.
10999 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
11001 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
11003 * config/nds32/constants.md (unspec_volatile_element): Add enum values
11004 for interrupt control.
11005 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
11006 functions for interrupt control.
11007 * config/nds32/nds32-intrinsic.md: Likewise.
11008 * config/nds32/nds32_intrinsic.h: Likewise.
11009 * config/nds32/nds32.h (nds32_builtins): Likewise.
11011 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
11013 * config/nds32/nds32.c (nds32_init_machine_status,
11014 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
11015 strict_aligned_p field.
11016 (nds32_expand_to_rtl_hook): New function.
11017 (TARGET_EXPAND_TO_RTL_HOOK): Define.
11018 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
11020 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
11021 Chung-Ju Wu <jasonwucj@gmail.com>
11023 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
11024 * config/nds32/nds32-n7.md: New file.
11025 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
11026 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
11028 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
11029 * config/nds32/nds32.md (pipeline_model): Add n7.
11030 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
11031 * config/nds32/pipelines.md: Include n7 settings.
11033 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
11034 Chung-Ju Wu <jasonwucj@gmail.com>
11036 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
11037 * config/nds32/nds32-e8.md: New file.
11038 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
11039 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
11041 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
11042 * config/nds32/nds32.md (pipeline_model): Add e8.
11043 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
11044 * config/nds32/pipelines.md: Include e8 settings.
11046 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
11047 Chung-Ju Wu <jasonwucj@gmail.com>
11049 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
11050 * config/nds32/nds32-n8.md: New file.
11051 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
11052 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
11054 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
11055 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
11056 * config/nds32/nds32.md (pipeline_model): Add n8.
11057 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
11058 * config/nds32/pipelines.md: Include n8 settings.
11060 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
11061 Chung-Ju Wu <jasonwucj@gmail.com>
11063 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
11064 * config/nds32/nds32-n9-2r1w.md: New file.
11065 * config/nds32/nds32-n9-3r2w.md: New file.
11066 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
11067 nds32_register_ports): New or modify for cpu n9.
11068 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
11070 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
11071 * config/nds32/nds32-utils.c: New file.
11072 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
11073 TARGET_MUL_SLOW): Define.
11074 * config/nds32/nds32.md (pipeline_model): New attribute.
11075 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
11076 New options that support cpu n9.
11077 * config/nds32/pipelines.md: Include n9 settings.
11078 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
11080 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
11082 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
11083 information if necessary.
11084 (output_cond_branch_compare_zero): Likewise.
11085 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
11086 (nds32_target_alignment): Refine for alignment.
11087 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
11088 (FUNCTION_BOUNDARY): Modify.
11089 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
11091 * config/nds32/nds32.opt (malways-align, malign-functions): New.
11093 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
11095 * config/nds32/constants.md (unspec_volatile_element): Add values for
11096 TLB operation and data prefetch.
11097 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
11098 functions for TLB operation and data prefetch.
11099 * config/nds32/nds32-intrinsic.md: Likewise.
11100 * config/nds32/nds32_intrinsic.h: Likewise.
11101 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
11102 (nds32_print_operand): Likewise.
11103 * config/nds32/nds32.h (nds32_builtins): Likewise.
11105 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
11106 Andrew Pinski <pinsika@gcc.gnu.org>
11108 PR middle-end/82976
11109 * match.pd: Use constant_boolean_node of correct type instead of
11110 boolean_true_node or boolean_false_node for simplifying
11111 pointer comparisons to zero.
11113 2018-04-07 Jakub Jelinek <jakub@redhat.com>
11115 PR tree-optimization/80021
11116 * tree.c (verify_type_variant): Make error call in verify_variant_match
11117 translatable and remove final full stop.
11119 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
11121 * config/nds32/constants.md (unspec_volatile_element): Add
11122 UNSPEC_VOLATILE_EH_RETURN.
11123 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
11124 nds32_output_stack_pop): Support dwarf exception handling process.
11125 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
11126 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
11127 exception handling process.
11128 (nds32_compute_stack_frame): Likewise.
11129 (nds32_return_addr_rtx): Likewise.
11130 (nds32_initial_elimination_offset): Likewise.
11131 (nds32_expand_prologue): Likewise.
11132 (nds32_expand_epilogue): Likewise.
11133 (nds32_dynamic_chain_address): New function.
11134 * config/nds32/nds32.h (machine_function): Add fields for dwarf
11135 exception handling.
11136 (DYNAMIC_CHAIN_ADDRESS): Define.
11137 (EH_RETURN_DATA_REGNO): Define.
11138 (EH_RETURN_STACKADJ_RTX): Define.
11139 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
11140 patterns for dwarf exception handling.
11142 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
11144 * config/nds32/nds32.h: Clean up obsolete macros.
11146 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
11148 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
11149 Add enum values for particular instructions.
11150 * config/nds32/nds32-intrinsic.c: Implementation of expanding
11151 particular intrinsic functions.
11152 * config/nds32/nds32-intrinsic.md: Likewise.
11153 * config/nds32/nds32_intrinsic.h: Likewise.
11154 * config/nds32/nds32.h (nds32_builtins): Likewise.
11155 * config/nds32/nds32.md (type): Add pbsad and pbsada.
11156 (btst, ave): New patterns for particular instructions.
11158 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
11160 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
11161 Add enum values for atomic load/store and memory sync.
11162 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
11164 * config/nds32/nds32-intrinsic.md: Likewise.
11165 * config/nds32/nds32_intrinsic.h: Likewise.
11166 * config/nds32/nds32.h (nds32_builtins): Likewise.
11168 2018-04-07 Jakub Jelinek <jakub@redhat.com>
11170 PR tree-optimization/85257
11171 * fold-const.c (native_encode_vector): If not all elts could fit
11172 and off is -1, return 0 rather than offset.
11173 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
11174 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
11175 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
11176 adjust buffer in native_interpret_expr call.
11178 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
11180 * config/nds32/constants.md (unspec_volatile_element): Add cache
11181 control enum values.
11182 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
11183 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
11184 * config/nds32/nds32.c (nds32_cctl_names): New.
11185 (nds32_print_operand): Handle cache control register names.
11186 * config/nds32/nds32.h (nds32_builtins): New enum values.
11187 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
11189 * config/nds32/nds32.md (type): Add mmu.
11190 * config/nds32/pipelines.md (simple_insn): Add mmu.
11192 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
11194 * config/nds32/nds32.md (type): Remove call.
11195 * config/nds32/pipelines.md (simple_insn): Likewise.
11197 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
11199 * config/nds32/constants.md (unspec_volatile_element): Add
11200 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
11201 UNSPEC_VOLATILE_FMFCFG.
11202 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
11203 description for fmfcfg and fmfcsr.
11204 (bdesc_1arg): Add fmtcsr.
11205 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
11206 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
11207 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
11208 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
11209 unspec_fmfcfg): New patterns.
11210 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
11211 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
11212 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
11213 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
11214 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
11215 __nds32__fmfcfg): Define.
11217 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
11219 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
11220 intrinsic register names.
11221 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
11222 intrinsic register enum values and macros.
11224 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
11226 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
11227 for load/store addressing form.
11228 (nds32_print_operand_address): Likewise.
11230 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
11233 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
11234 based on LABEL_REF. Remove useless assertion.
11235 (pic_address_needs_scratch): Fix formatting.
11236 (sparc_legitimize_pic_address): Minor tweaks.
11237 (sparc_delegitimize_address): Adjust assertion accordingly.
11238 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
11239 into symbolic_operand.
11240 (movsi_high_pic_label_ref): Likewise.
11241 (movsi_lo_sum_pic_label_ref): Likewise.
11242 (movdi_pic_label_ref): Likewise.
11243 (movdi_high_pic_label_ref): Likewise.
11244 (movdi_lo_sum_pic_label_ref): Likewise.
11246 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
11248 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
11249 custom LIB_SPEC setup.
11251 2018-04-06 Ruslan Bukin <br@bsdpad.com>
11252 Kito Cheng <kito.cheng@gmail.com>
11254 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
11255 * config/riscv/freebsd.h: New.
11257 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
11259 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
11260 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
11263 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
11264 Kito Cheng <kito.cheng@gmail.com>
11266 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
11267 nds32_output_call, nds32_symbol_binds_local_p): New functions.
11268 * config/nds32/nds32-protos.h (nds32_output_call,
11269 nds32_output_return): Declare.
11270 * config/nds32/nds32.md: Refine all the call and return patterns.
11272 2018-04-06 Jakub Jelinek <jakub@redhat.com>
11275 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
11276 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
11278 PR rtl-optimization/84872
11279 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
11280 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
11281 EDGE_CROSSING edge.
11283 2018-04-06 Tamar Christina <tamar.christina@arm.com>
11285 * expr.c (copy_blkmode_to_reg): Revert 254862.
11286 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
11288 2018-04-06 Richard Biener <rguenther@suse.de>
11290 PR middle-end/85244
11291 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
11292 after seeing a component reference with an adjacent field. Treat
11293 refs to arrays at struct end of external decls similar to
11294 refs to unconstrained commons.
11296 2018-04-06 Jakub Jelinek <jakub@redhat.com>
11299 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
11300 look through SAVE_EXPRs with non-side-effects argument. Adjust
11302 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
11305 2018-04-06 Richard Biener <rguenther@suse.de>
11307 PR middle-end/85180
11308 * alias.c (find_base_term): New wrapper around find_base_term
11309 unwinding CSELIB_VAL_PTR changes.
11310 (find_base_term): Do not restore CSELIB_VAL_PTR during the
11313 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11315 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
11317 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
11318 constant definitions.
11319 ("nop"): lr 0,0 -> nopr r0
11320 ("nop_lr0", "nop_lr1"): New insn definitions.
11322 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
11324 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
11325 NDS32_V3PUSH_AVAILABLE_P macro.
11327 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
11328 Chung-Ju Wu <jasonwucj@gmail.com>
11330 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
11331 (nds32*-*-*): Add float and fpu_config into supported_defaults.
11332 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
11333 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
11334 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
11335 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
11336 * config/nds32/constraints.md: New constraints and checking for hard
11337 float configuration.
11338 * config/nds32/iterators.md: New mode iterator and attribute for hard
11339 float configuration.
11340 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
11342 * config/nds32/nds32-fpu.md: New file.
11343 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
11344 deal with hard float code generation.
11345 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
11347 (abi_type, float_reg_number): New enum type.
11348 * config/nds32/nds32-predicates.c: New predicates for hard float.
11349 * config/nds32/nds32-protos.h: Declare functions for hard float.
11350 * config/nds32/nds32.c: Implementation for hard float configuration.
11351 * config/nds32/nds32.h: Definitions for hard float configuration.
11352 * config/nds32/nds32.md: Include hard float machine description and
11353 modify patterns for hard float configuration.
11354 * config/nds32/nds32.opt: New options for hard float configuration.
11355 * config/nds32/predicates.md: New predicates for hard float
11358 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
11360 * common/config/nds32/nds32-common.c
11361 (nds32_option_optimization_table): Enable -mreleax-hint by default.
11363 2018-04-05 Jakub Jelinek <jakub@redhat.com>
11365 PR middle-end/85195
11366 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
11367 CONSTRUCTOR_ELT (ctor, ...)->value.
11369 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
11372 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
11374 2018-04-05 Tom de Vries <tom@codesourcery.com>
11377 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
11380 2018-04-05 Shiva Chen <shiva0217@gmail.com>
11381 Kito Cheng <kito.cheng@gmail.com>
11383 * config/nds32/constraints.md (U33): Fine-tune checking condition.
11384 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
11385 * config/nds32/nds32.h (nds32_16bit_address_type): Add
11386 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
11388 2018-04-05 Shiva Chen <shiva0217@gmail.com>
11389 Kito Cheng <kito.cheng@gmail.com>
11391 * config/nds32/constraints.md (Ufe): New memory constraint.
11392 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
11393 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
11394 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
11396 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
11397 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
11399 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11401 * config/nds32/nds32.md: Use optimize_size in the condition for
11402 alu-shift instructions.
11404 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11406 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
11408 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11410 * config/nds32/nds32.md (negsi2): Refine pattern.
11412 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
11413 Chung-Ju Wu <jasonwucj@gmail.com>
11415 * config/nds32/iterators.md (shift_rotate): New code iterator.
11416 (shift): New code attribute.
11417 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
11418 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
11419 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
11420 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
11421 bit-wise operations.
11422 (andsi3, *andsi3): Ditto.
11423 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
11424 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
11425 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
11426 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
11427 nds32_ior_operand, nds32_xor_operand): New predicates.
11429 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11431 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
11432 (addsi3, subsi3): ... this.
11434 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11436 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
11438 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11440 * config/nds32/nds32.md: Adjust indention.
11442 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
11444 * config/nds32/nds32.md (feature): New attribute.
11446 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11448 * config/nds32/nds32.md (subtype): New attribute.
11450 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
11453 * config/arm/arm-builtins.c (arm_expand_builtin): Change
11454 expansion to perform a bitwise AND of the argument followed by a
11455 boolean negation of the result.
11457 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
11459 PR rtl-optimization/84878
11460 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
11461 the basic block. Assert the use reference is not artificial and that
11462 it has an associated insn.
11464 2018-04-04 Michael Matz <matz@suse.de>
11466 * builtins.c (compute_objsize): Pass correct operand
11467 to array_at_struct_end_p.
11469 2018-04-04 Richard Biener <rguenther@suse.de>
11472 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
11473 from contexts for DINFO_LEVEL_TERSE and below.
11475 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
11477 * config/nds32/nds32-doubleword.md (move_<mode>): Require
11478 resiter_operand condition.
11479 * config/nds32/nds32.md (*move<mode>): Ditto.
11481 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
11482 Monk Chiang <sh.chiang04@gmail.com>
11484 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
11486 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
11488 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
11490 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
11491 Kito Cheng <kito.cheng@gmail.com>
11493 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
11494 nds32_cond_code_str, output_cond_branch,
11495 output_cond_branch_compare_zero, nds32_expand_cbranch,
11496 nds32_expand_cstore, nds32_expand_movcc,
11497 nds32_output_cbranchsi4_equality_zero,
11498 nds32_output_cbranchsi4_equality_reg,
11499 nds32_output_cbranchsi4_equality_reg_or_const_int,
11500 nds32_output_cbranchsi4_greater_less_zero: New functions.
11501 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
11502 nds32_expand_cstore, nds32_expand_movcc,
11503 nds32_output_cbranchsi4_equality_zero,
11504 nds32_output_cbranchsi4_equality_reg,
11505 nds32_output_cbranchsi4_equality_reg_or_const_int,
11506 nds32_output_cbranchsi4_greater_less_zero): Declare.
11507 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
11508 nds32_rimm11s_operand): New predicates.
11509 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
11510 * config/nds32/nds32.md: Rewrite all the branch and conditional move
11513 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
11515 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
11516 * config/nds32/nds32.md: Ditto.
11517 * config/nds32/pipelines.md: Ditto.
11519 2018-04-04 Richard Biener <rguenther@suse.de>
11521 PR tree-optimization/85168
11522 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
11523 propagating abnormals.
11525 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
11527 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
11529 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
11530 Kito Cheng <kito.cheng@gmail.com>
11532 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
11533 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
11534 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
11535 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
11536 * config/nds32/nds32.md (sibcall_internal): New.
11537 (sibcall_register): Remove.
11538 (sibcall_immediate): Remove.
11539 (sibcall_value_internal): New.
11540 (sibcall_value_register): Remove.
11541 (sibcall_value_immediate): Remove.
11542 * config/nds32/predicates.md (nds32_general_register_operand): New.
11543 (nds32_call_address_operand): New.
11545 2018-04-03 Jakub Jelinek <jakub@redhat.com>
11547 PR rtl-optimization/85167
11548 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
11549 bb_defs if *split_p, instead preinitialize it to NULL.
11551 PR tree-optimization/85156
11552 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
11553 evaluating the argument multiple times.
11555 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
11557 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
11559 (_mm_cvtpd_ps): Likewise.
11560 (_mm_cvttpd_epi32): Likewise.
11561 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
11562 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
11563 vector, pixel, and bool following altivec.h include.
11565 2018-04-03 Martin Sebor <msebor@redhat.com>
11567 * doc/extend.texi (Common Function Attributes): Clarify.
11568 (const attribute): Likewise.
11569 (pure attribute): Likewise.
11571 2018-04-03 Jakub Jelinek <jakub@redhat.com>
11574 * config/i386/i386.c (ix86_expand_vector_set): Use
11575 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
11577 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
11579 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
11580 instructions when changing rounding bits to preserve precision bits
11581 in the x87 control word.
11583 2018-04-03 Martin Liska <mliska@suse.cz>
11585 PR tree-optimization/82491
11586 * rtl.h (strip_offset_and_add): Replace += suboffset with
11587 poly_uint64 () + suboffset.
11589 2018-03-29 Martin Liska <mliska@suse.cz>
11590 Martin Jambor <mjambor@suse.cz>
11593 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
11594 param_type is not an integral or pointer type.
11596 2018-04-03 Richard Biener <rguenther@suse.de>
11598 * sese.h (recompute_all_dominators): Remove.
11600 2018-04-02 Martin Sebor <msebor@redhat.com>
11602 * doc/invoke.texi (-Wrestrict): Fix typos.
11604 2018-04-02 Jim Wilson <jimw@sifive.com>
11606 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
11607 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
11608 (<optab>di3, <optab>si3_extend): Likewise.
11609 (<optab>si3_mask, <optab>si3_mask_1): New.
11610 (<optab>di3_mask, <optab>di3_mask_1): New.
11611 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
11612 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
11613 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
11615 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
11617 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
11620 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
11622 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
11623 (nds32_canonicalize_comparison): New function.
11625 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
11626 Kito Cheng <kito.cheng@gmail.com>
11627 Kuan-Lin Chen <kuanlinchentw@gmail.com>
11629 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
11630 * config/nds32/constants.md (unspec_volatile_element): Add
11631 UNSPEC_VOLATILE_RELAX_GROUP.
11632 * config/nds32/nds32-relax-opt.c: New file.
11633 * config/nds32/nds32-predicates.c
11634 (nds32_symbol_load_store_p): New function.
11635 * config/nds32/nds32-protos.h
11636 (nds32_symbol_load_store_p): Declare function.
11637 (make_pass_nds32_relax_opt): Declare new rtl pass function.
11638 * config/nds32/nds32.c
11639 (nds32_register_pass): New function to register pass.
11640 (nds32_register_passes): New function to register passes.
11641 * config/nds32/nds32.md (relax_group): New pattern.
11642 * config/nds32/nds32.opt (mrelax-hint): New option.
11643 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
11645 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
11647 * config/nds32/t-nds32: Modify files dependency.
11649 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
11651 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
11652 (PROFILE_HOOK): Define its implementation.
11654 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
11656 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
11657 type and 32-bit size.
11659 2018-04-01 Jakub Jelinek <jakub@redhat.com>
11661 PR middle-end/85090
11662 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
11663 (V_128_256): New mode iterator.
11664 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
11665 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
11666 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
11668 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
11669 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
11671 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
11674 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
11675 NaN inputs correctly.
11677 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
11680 * config/rs6000/vsx.md (??r): New mode attribute.
11681 (*vsx_mov<mode>_64bit): Use it.
11682 (*vsx_mov<mode>_32bit): Likewise.
11684 2018-03-30 Martin Sebor <msebor@redhat.com>
11686 PR tree-optimization/84818
11687 * builtins.c (check_access): Use warning_n.
11689 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11692 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
11694 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
11697 2018-03-30 Julia Koval <julia.koval@intel.com>
11700 * x86-tune.def (movx, partial_reg_dependency): Enable for
11703 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
11705 PR inline-asm/84985
11706 * lra-constraints.c (process_alt_operands): Move setting
11707 this_alternative_matches below.
11709 2018-03-29 Martin Liska <mliska@suse.cz>
11712 * doc/invoke.texi: Document how LTO works with debug info.
11713 Describe auto-load support of binutils. Mention 'x86-64'
11714 as valid option value of -march option.
11716 2018-03-29 Jakub Jelinek <jakub@redhat.com>
11718 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
11721 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
11722 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
11723 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
11726 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
11729 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
11730 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
11731 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
11732 for RS6000_BTM_POWERPC64.
11733 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
11734 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
11735 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
11738 (DIVDEU): Likewise.
11740 2018-03-28 Carl Love <cel@us.ibm.com>
11743 2017-09-27 Carl Love <cel@us.ibm.com>
11745 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
11746 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
11747 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
11750 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11752 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
11753 instead of __vector bool.
11754 (_mm_max_pu8): Likewise.
11755 (_mm_min_pi16): Likewise.
11757 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
11760 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
11761 (DIVWEUO): Likewise.
11762 (DIVDEO): Likewise.
11763 (DIVDEUO): Likewise.
11764 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
11765 DIVWEUO and DIVDEUO.
11766 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
11767 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
11768 (div_extend): Likewise.
11769 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
11771 (__builtin_divweuo): Likewise.
11772 (__builtin_divdeo): Likewise.
11773 (__builtin_divdeuo): Likewise.
11775 2018-03-28 Jakub Jelinek <jakub@redhat.com>
11778 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
11779 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
11781 PR tree-optimization/82004
11782 * gimple-match-head.c (optimize_pow_to_exp): New function.
11783 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
11784 Don't fold to exp if optimize_pow_to_exp is false.
11786 2018-03-28 Martin Liska <mliska@suse.cz>
11789 * calls.c (initialize_argument_information): Fix trailing space.
11790 * common.opt: Fix typo and provide better explanation for
11791 -fsanitize-coverage option.
11792 * config/i386/i386.opt: Fix typo.
11794 2018-03-28 Jakub Jelinek <jakub@redhat.com>
11795 Martin Liska <mliska@suse.cz>
11798 * gimplify.c (asan_poison_variable): Don't do the check for
11799 gimplify_omp_ctxp here.
11800 (gimplify_decl_expr): Do it here.
11801 (gimplify_target_expr): Likewise.
11803 2018-03-28 Martin Liska <mliska@suse.cz>
11806 * config/i386/i386.c (ix86_function_arg_advance): Do not call
11807 chkp_type_bounds_count if MPX is not enabled.
11809 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
11811 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
11813 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
11816 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
11817 function to create the function decl for complex long double
11818 multiply and divide for -mabi=ieeelongdouble.
11819 (init_float128_ieee): Call it.
11821 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
11824 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
11825 -fcf-protection=branch -mibt.
11826 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
11828 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
11831 * config/arm/arm.c (arm_valid_symbolic_address): Handle
11832 arm_word_relocations.
11834 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
11837 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
11838 extern array declarations.
11840 2018-03-27 Richard Biener <rguenther@suse.de>
11842 PR middle-end/84067
11843 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
11844 explicit single_use checks.
11846 2018-03-27 Richard Biener <rguenther@suse.de>
11848 PR tree-optimization/85082
11849 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
11852 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
11854 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
11855 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
11856 Turn on fasynchronous-unwind-tables and funwind-tables.
11858 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
11861 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
11862 (*bmi_blsr_<mode>_ccz): Ditto.
11864 2018-03-26 Tom de Vries <tom@codesourcery.com>
11866 PR tree-optimization/85063
11867 * omp-general.c (offloading_function_p): New function. Factor out
11869 * omp-offload.c (pass_omp_target_link::gate): ... here.
11870 * omp-general.h (offloading_function_p): Declare.
11871 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
11872 with attribute omp declare target for offloading functions.
11874 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
11876 PR tree-optimization/84005
11877 * tree-data-ref.h (get_base_for_alignment): Declare.
11878 * tree-data-ref.c (get_base_for_alignment_1): New function.
11879 (get_base_for_alignment): Likewise.
11880 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
11881 get_base_for_alignment to find a suitable base object, instead
11882 of always using drb->base_address.
11884 2018-03-23 Jakub Jelinek <jakub@redhat.com>
11886 PR inline-asm/85022
11887 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
11888 known size by default.
11890 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
11892 PR inline-asm/85030
11893 * lra-constraints.c (process_alt_operands): Don't match BLKmode
11894 and non BLKmode operands.
11896 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11899 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
11900 Clean up attributes.
11902 2018-03-23 Richard Biener <rguenther@suse.de>
11905 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
11906 we are going to emit early debug for LTO.
11908 2018-03-23 Jakub Jelinek <jakub@redhat.com>
11910 PR inline-asm/85034
11911 * function.c (match_asm_constraints_1): Don't optimize if input
11912 doesn't satisfy general_operand predicate for output's mode.
11914 PR inline-asm/85022
11915 * alias.c (write_dependence_p): Don't require for x_canonicalized
11916 non-VOIDmode if x has VOIDmode.
11919 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
11920 just don't try to optimize it rather than assert it never happens.
11922 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
11924 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
11925 macro expansions for definition of ST_INTERNAL_<mode> and
11926 LD_INTERNAL_<mode> builtins.
11927 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
11929 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
11931 (altivec_expand_st_builtin): Likewise.
11932 (altivec_expand_builtin): Remove calls to deleted functions.
11933 (rs6000_address_for_altivec): Delete this function.
11934 * config/rs6000/vector.md: Remove expands for
11935 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
11937 2018-03-22 Sudakshina Das <sudi.das@arm.com>
11940 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
11941 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
11942 re-computing once computed.
11943 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
11944 (arm_init_machine_status): Initialize
11945 machine->static_chain_stack_bytes.
11947 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
11950 * doc/extend.texi: Add four new prototypes for vec_ld.
11951 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
11952 definitions for more logical presentation.
11953 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
11954 entries for V1TI variants of __builtin_altivec_ld builtin.
11955 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
11956 handling of V1TI variant of LVX icode pattern.
11957 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
11958 (rs6000_gimple_fold_builtin): Likewise.
11959 (altivec_init_builtins): Add code to define
11960 __builtin_altivec_lvx_v1ti function.
11962 2018-03-22 Jakub Jelinek <jakub@redhat.com>
11964 PR inline-asm/84941
11965 * function.c (match_asm_constraints_1): Don't do the optimization
11966 if input isn't a REG, SUBREG, MEM or constant.
11968 2018-03-22 Tom de Vries <tom@codesourcery.com>
11970 PR tree-optimization/84956
11971 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
11972 bb_has_abnormal_pred.
11974 2018-03-22 Jakub Jelinek <jakub@redhat.com>
11977 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
11978 DECL_INITIAL (decl) to decl at the end.
11979 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
11980 adjust the comment.
11982 2018-03-21 Joseph Myers <joseph@codesourcery.com>
11984 * doc/extend.texi (__builtin_tgmath): Document when complex
11985 integer types are treated as _Complex _Float64.
11987 2018-03-21 Tom de Vries <tom@codesourcery.com>
11989 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
11991 2018-03-21 Jakub Jelinek <jakub@redhat.com>
11993 PR tree-optimization/84960
11994 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
11995 if it is ENTRY block, move them into single succ of ENTRY in that case.
11997 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
11999 PR tree-optimization/84811
12000 * poly-int.h (poly_span_traits): Remove the T3 parameter and
12001 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
12002 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
12003 (known_subrange_p): Update accordingly. Cast each value involved
12004 in the size comparison, rather than casting the result of the
12007 2018-03-21 Jakub Jelinek <jakub@redhat.com>
12009 PR tree-optimization/84982
12010 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
12011 by flipping the least significant bit rather than all bits from
12012 bitpos to bitpos + bitsize - 1.
12014 2018-03-21 Nathan Sidwell <nathan@acm.org>
12016 * doc/extend.texi (Deprecated Features): Remove mention of
12017 long-deleted deprecations.
12019 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12022 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
12023 * configure: Regenerate.
12025 2018-03-21 Tom de Vries <tom@codesourcery.com>
12027 PR tree-optimization/83126
12028 * tree-parloops.c (num_phis): New function.
12029 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
12031 2018-03-21 Nathan Sidwell <nathan@acm.org>
12033 * doc/extend.texi (Deprecated Features): Update deprecated flags,
12034 mention anon-struct/union members and trailing attributes.
12036 2018-03-21 Bin Cheng <bin.cheng@arm.com>
12038 PR tree-optimization/84969
12039 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
12040 builtin memset partitions if they set different rhs values.
12042 2018-03-21 Jakub Jelinek <jakub@redhat.com>
12044 PR rtl-optimization/84989
12045 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
12046 VEC_DUPLICATE with scalar result mode.
12048 2018-03-21 Martin Liska <mliska@suse.cz>
12051 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
12052 not intended return statement.
12054 2018-03-21 Martin Liska <mliska@suse.cz>
12057 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
12058 (chkp_find_bound_slots_1): Limit number of iterations.
12060 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
12063 * Minor grammar fixes for x86 options.
12065 2018-03-20 Jakub Jelinek <jakub@redhat.com>
12068 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
12069 holding REG_CFA_RESTORE notes, instead turn them into a USE.
12071 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
12074 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
12075 (altivec_lvx_<mode>_1op): Likewise.
12076 (altivec_stvx_<mode>_2op): Likewise.
12077 (altivec_stvx_<mode>_1op): Likewise.
12078 (altivec_lvx_<VM2:mode>): New define_expand.
12079 (altivec_stvx_<VM2:mode>): Likewise.
12080 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
12081 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
12082 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
12083 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
12084 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
12085 (rs6000_gen_lvx): Likewise.
12086 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
12087 (altivec_expand_stv_builtin): Likewise.
12088 (altivec_expand_builtin): Likewise.
12089 * config/rs6000/vector.md: Likewise.
12091 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12094 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
12097 2018-03-20 Richard Biener <rguenther@suse.de>
12100 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
12101 sign-conversions as zero, fall back to standard scalar_stmt
12104 2018-03-20 Martin Liska <mliska@suse.cz>
12107 * predict.c (rebuild_frequencies): Handle case when we have
12108 PROFILE_ABSENT, but flag_guess_branch_prob is false.
12110 2018-03-20 Jakub Jelinek <jakub@redhat.com>
12113 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
12114 flag_section_anchors.
12115 * varasm.c (use_blocks_for_decl_p): Remove hack for
12116 dw2_force_const_mem.
12119 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
12121 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
12122 be created, use lowpart_subreg of operands[0] rather than operands[0]
12124 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
12125 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
12126 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
12127 and n constraint instead of aarch64_shift_imm_di and Usd.
12128 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
12129 (*aarch64_<optab>_reg_minus<mode>3): ... this.
12131 2018-03-20 Sudakshina Das <sudi.das@arm.com>
12134 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
12135 to favor GPR over NEON registers.
12136 (<shift>di3_neon): Likewise.
12138 2018-03-20 Tom de Vries <tom@codesourcery.com>
12141 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
12142 (nvptx_process_pars): Emit bar.sync asap and alap.
12144 2018-03-20 Tom de Vries <tom@codesourcery.com>
12147 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
12148 seen_label if seen_label is already set.
12150 2018-03-20 Jakub Jelinek <jakub@redhat.com>
12153 * config/i386/i386.c (fold_builtin_cpu): For features above 31
12154 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
12155 Use 1U instead of 1. Formatting fixes.
12158 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
12159 instead of TREE_TYPE (s1) for the return value.
12161 2018-03-19 Jakub Jelinek <jakub@redhat.com>
12163 PR tree-optimization/84946
12164 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
12165 bitsize + bitsize in poly_uint64 rather than poly_int64.
12168 * dwarf2asm.c: Include fold-const.c.
12169 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
12170 of decl rather than decl itself.
12172 PR rtl-optimization/84643
12173 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
12175 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
12178 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
12179 calling assemble_variable.
12181 2018-03-19 Sudakshina Das <sudi.das@arm.com>
12184 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
12185 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
12187 2018-03-19 Jim Wilson <jimw@sifive.com>
12190 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
12191 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
12192 (riscv_first_stack_step): Likewise.
12193 (riscv_option_override): Use STACK_BOUNDARY instead of
12194 MIN_STACK_BOUNDARY.
12195 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
12196 MIN_STACK_BOUNDARY.
12197 (BIGGEST_ALIGNMENT): Set to 128.
12198 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
12199 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
12202 2018-03-19 Richard Biener <rguenther@suse.de>
12204 PR tree-optimization/84933
12205 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
12206 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
12208 2018-03-19 Richard Biener <rguenther@suse.de>
12210 PR tree-optimization/84859
12211 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
12212 (cond_if_else_store_replacement): Perform sinking operation on
12213 single-store BBs regardless of MAX_STORES_TO_SINK setting.
12214 Generalize what a BB with a single eligible store is.
12216 2018-03-19 Richard Biener <rguenther@suse.de>
12218 PR tree-optimization/84929
12219 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
12220 chrec_is_positive against non-chrec arg.
12222 2018-03-19 Tamar Christina <tamar.christina@arm.com>
12225 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
12227 2018-03-18 Martin Liska <mliska@suse.cz>
12229 PR rtl-optimization/84635
12230 * regrename.c (build_def_use): Use matches_mode only when
12233 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
12235 PR tree-optimization/84913
12236 * tree-vect-loop.c (vectorizable_reduction): Don't try to
12237 vectorize chains of COND_EXPRs.
12239 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
12241 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
12243 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
12245 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
12247 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
12249 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
12251 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
12252 Kito Cheng <kito.cheng@gmail.com>
12254 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
12255 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
12256 (nds32_adjust_reg_alloc_order): New function.
12257 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
12259 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
12261 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
12262 nds32_print_operand, nds32_print_operand_address): Use
12263 HOST_WIDE_INT_PRINT_DEC instead.
12265 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
12267 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
12269 2018-03-17 Jakub Jelinek <jakub@redhat.com>
12272 * config/i386/i386.c (initial_ix86_tune_features,
12273 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
12274 unsigned long long.
12275 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
12276 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
12277 rather than 1u << ix86_tune. Formatting fix.
12278 (ix86_option_override_internal): Change ix86_arch_mask from
12279 unsigned int to unsigned HOST_WIDE_INT, initialize to
12280 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
12281 (ix86_function_specific_restore): Likewise.
12283 2018-03-16 Jakub Jelinek <jakub@redhat.com>
12286 * postreload.c (reload_combine_recognize_pattern): Perform
12287 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
12288 truncate_int_for_mode the result for the destination's mode.
12291 * hsa-gen.c (mem_type_for_type): Fix comment typo.
12292 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
12294 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
12297 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
12300 * lra-assigns.c (lra_split_hard_reg_for): Don't use
12301 regno_allocno_class_array and sorted_pseudos.
12302 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
12303 insns where regno is used.
12305 2018-03-16 Martin Liska <mliska@suse.cz>
12308 * multiple_target.c (create_dispatcher_calls): Redirect
12309 reference in the symbol table.
12311 2018-03-16 Martin Liska <mliska@suse.cz>
12314 * multiple_target.c (create_dispatcher_calls): Redirect also
12317 2018-03-16 Jakub Jelinek <jakub@redhat.com>
12321 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
12322 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
12323 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
12325 2018-03-16 Julia Koval <julia.koval@intel.com>
12327 * doc/invoke.texi (Skylake Server): Add CLWB.
12328 Cannonlake): Remove CLWB.
12330 2018-03-16 Jakub Jelinek <jakub@redhat.com>
12332 PR tree-optimization/84841
12333 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
12335 (FLOAT_ONE_CONST_TYPE): Define.
12336 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
12337 (sort_by_operand_rank): Put entries with higher constant_type last
12338 rather than first to match comments.
12340 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
12342 * config/nios2/nios2.md (movsi_internal): Fix thinko in
12345 2018-03-15 Jakub Jelinek <jakub@redhat.com>
12348 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
12349 check and use address of target always.
12351 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
12354 * config/i386/i386.c (indirect_thunk_needed): Update comments.
12355 (indirect_thunk_bnd_needed): Likewise.
12356 (indirect_thunks_used): Likewise.
12357 (indirect_thunks_bnd_used): Likewise.
12358 (indirect_return_needed): New.
12359 (indirect_return_bnd_needed): Likewise.
12360 (output_indirect_thunk_function): Add a bool argument for
12362 (output_indirect_thunk_function): Don't generate alias for
12363 function return thunk.
12364 (ix86_code_end): Call output_indirect_thunk_function to generate
12365 function return thunks.
12366 (ix86_output_function_return): Set indirect_return_bnd_needed
12367 and indirect_return_needed instead of indirect_thunk_bnd_needed
12368 and indirect_thunk_needed.
12370 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
12372 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
12373 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
12374 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
12376 2018-03-15 David Malcolm <dmalcolm@redhat.com>
12377 Paul Hua <paul.hua.gm@gmail.com>
12380 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
12382 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
12384 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
12385 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
12386 resp. SFmode cases.
12388 2018-03-15 Tamar Christina <tamar.christina@arm.com>
12391 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
12392 instead of GET_MODE_SIZE when comparing Units.
12394 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
12397 * varasm.c (hash_section): Return an unchangeble hash value
12398 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
12399 Return !aarch64_can_use_per_function_literal_pools_p ().
12401 2018-03-15 Jakub Jelinek <jakub@redhat.com>
12404 * optabs.c (emit_conditional_move): Pass address of cmode's copy
12405 rather than address of cmode as last argument to prepare_cmp_insn.
12407 2018-03-15 Julia Koval <julia.koval@intel.com>
12409 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
12410 F_AVX512VNNI, F_AVX512BITALG): New.
12412 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
12415 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
12416 insn for floating-point loads and stores.
12418 2018-03-14 Carl Love <cel@us.ibm.com>
12420 * config/rs6000/rs6000-c.c: Add macro definitions for
12421 ALTIVEC_BUILTIN_VEC_PERMXOR.
12422 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
12423 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
12424 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
12425 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
12427 * config/doc/extend.texi: Add prototypes for vec_permxor.
12429 2018-03-14 David Malcolm <dmalcolm@redhat.com>
12432 * diagnostic-show-locus.c (class layout_point): Convert m_line
12433 from int to linenum_type.
12434 (line_span::comparator): Use linenum "compare" function when
12435 comparing line numbers.
12436 (test_line_span): New function.
12437 (layout_range::contains_point): Convert param "row" from int to
12439 (layout_range::intersects_line_p): Likewise.
12440 (layout::will_show_line_p): Likewise.
12441 (layout::print_source_line): Likewise.
12442 (layout::should_print_annotation_line_p): Likewise.
12443 (layout::print_annotation_line): Likewise.
12444 (layout::print_leading_fixits): Likewise.
12445 (layout::annotation_line_showed_range_p): Likewise.
12446 (struct line_corrections): Likewise for field m_row.
12447 (line_corrections::line_corrections): Likewise for param "row".
12448 (layout::print_trailing_fixits): Likewise.
12449 (layout::get_state_at_point): Likewise.
12450 (layout::get_x_bound_for_row): Likewise.
12451 (layout::print_line): Likewise.
12452 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
12453 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
12454 * input.c (selftest::test_linenum_comparisons): New function.
12455 (selftest::input_c_tests): Call it.
12456 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
12457 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
12458 * selftest.h (ASSERT_GT): New macro.
12459 (ASSERT_GT_AT): New macro.
12460 (ASSERT_LT): New macro.
12461 (ASSERT_LT_AT): New macro.
12463 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
12465 PR rtl-optimization/84780
12466 * combine.c (distribute_links): Don't make a link based on pc_rtx.
12468 2018-03-14 Martin Liska <mliska@suse.cz>
12470 * tree.c (record_node_allocation_statistics): Use
12471 get_stats_node_kind.
12472 (get_stats_node_kind): New function extracted from
12473 record_node_allocation_statistics.
12474 (free_node): Use get_stats_node_kind.
12476 2018-03-14 Richard Biener <rguenther@suse.de>
12478 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
12479 that the value-set of ANTIC_IN doesn't grow.
12482 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
12484 (BB_VISITED_WITH_VISITED_SUCCS): New define.
12485 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
12487 2018-03-14 Julia Koval <julia.koval@intel.com>
12489 * config.gcc (icelake-client, icelake-server): New.
12491 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
12492 (initial_ix86_arch_features): Ditto.
12493 (PTA_SKYLAKE): Add SGX.
12494 (PTA_ICELAKE): Remove.
12495 (PTA_ICELAKE_CLIENT): New.
12496 (PTA_ICELAKE_SERVER): New.
12497 (ix86_option_override_internal): Split up icelake on icelake client and
12499 (get_builtin_code_for_version): Ditto.
12500 (fold_builtin_cpu): Ditto.
12501 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
12502 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
12503 * config/i386/i386.h (processor_type): Ditto.
12504 * doc/invoke.texi: Ditto.
12506 2018-03-14 Jakub Jelinek <jakub@redhat.com>
12509 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
12510 INTEGER_CST offset, add it together with bitpos / 8 and
12511 sign extend based on POINTER_SIZE.
12515 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
12518 * config/i386/constraints.md (Yc): New register constraint.
12519 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
12520 Use Yc constraint for alternative 2 of operand 0. Remove
12521 preferred_for_speed attribute.
12523 2018-03-14 Richard Biener <rguenther@suse.de>
12525 PR tree-optimization/84830
12526 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
12527 with the old one to avoid oscillations.
12529 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
12532 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
12534 (assign_by_spills): Return a flag of reload assignment failure.
12535 Do not process the reload assignment failures. Do not spill other
12536 reload pseudos if they has the same reg class. Update n if
12538 (lra_assign): Add a return arg. Set up from the result of
12539 assign_by_spills call.
12540 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
12541 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
12542 usage_insns if it is not NULL.
12543 (spill_hard_reg_in_range): New function.
12544 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
12545 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
12546 function prototypes.
12547 (lra_assign): Change prototype.
12548 * lra.c (lra): Add code to deal with fails by splitting hard reg
12551 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
12553 * config/riscv/riscv.opt (mrelax): New option.
12554 * config/riscv/riscv.c (riscv_file_start): Emit ".option
12555 "norelax" when riscv_mrelax is disabled.
12556 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
12558 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12561 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
12562 reassociation for int modes.
12564 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
12566 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
12567 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
12569 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
12570 * config/aarch64/aarch64-sve.md
12571 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
12572 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
12573 (*extend<mode><Vwide>2): Rename to...
12574 (aarch64_sve_extend<mode><Vwide>2): ...this.
12575 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
12576 renaming the old pattern to...
12577 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
12579 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
12580 define_expand, renaming the old pattern to...
12581 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
12582 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
12583 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
12584 account when deciding which SVE instruction the optab should use.
12585 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
12587 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
12589 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
12590 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
12591 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
12592 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
12593 (tlsdesc_small_<mode>): Turn a define_expand and use
12594 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
12595 (tlsdesc_small_advsimd_<mode>): ...this.
12596 (tlsdesc_small_sve_<mode>): New pattern.
12598 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
12600 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
12601 (UNSPEC_UMUL_HIGHPART): New constants.
12602 (MUL_HIGHPART): New int iteraor.
12603 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
12604 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
12606 (*<su>mul<mode>3_highpart): New define_insn.
12608 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
12611 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
12614 2018-03-13 Martin Liska <mliska@suse.cz>
12617 * (sem_item_optimizer::sem_item_optimizer): Initialize new
12619 (sem_item_optimizer::~sem_item_optimizer): Release it.
12620 (sem_item_optimizer::merge_classes): Register variable aliases.
12621 (sem_item_optimizer::fixup_pt_set): New function.
12622 (sem_item_optimizer::fixup_points_to_sets): Likewise.
12623 * ipa-icf.h: Declare new variables and functions.
12625 2018-03-13 Jakub Jelinek <jakub@redhat.com>
12627 PR middle-end/84834
12628 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
12629 integer_pow2p@2 and test integer_pow2p in condition.
12630 (A < 0 ? C : 0): Similarly for @1.
12632 PR middle-end/84831
12633 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
12634 characters starting at p contain '\0' character, don't look beyond
12638 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
12639 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
12642 * reg-stack.c (change_stack): Change update_end var from int to
12643 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
12644 also call set_block_for_insn on the newly added insns and rescan.
12647 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
12648 on the last operand.
12651 * tree.c (stabilize_reference_1): Return save_expr (e) for
12652 STATEMENT_LIST even if it doesn't have side-effects.
12654 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
12656 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
12658 2018-03-12 Renlin Li <renlin.li@arm.com>
12660 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
12661 aarch64_output_scalar_simd_mov_immediate.
12663 2018-03-12 Martin Sebor <msebor@redhat.com>
12665 PR tree-optimization/83456
12666 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
12667 for perfectly overlapping calls to memcpy.
12668 (gimple_fold_builtin_memory_chk): Same.
12669 (gimple_fold_builtin_strcpy): Handle no-warning.
12670 (gimple_fold_builtin_stxcpy_chk): Same.
12671 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
12673 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
12675 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
12676 parameter. Use it for SFmode.
12677 (rs6000_function_arg_advance_1): Adjust.
12678 (rs6000_function_arg): Adjust.
12679 (rs6000_gimplify_va_arg): Pass false for that new parameter.
12681 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
12683 PR rtl-optimization/84169
12684 PR rtl-optimization/84780
12685 * combine.c (can_combine_p): Check for a 2-insn combination whether
12686 the destination register is used between the two insns, too.
12688 2018-03-12 Richard Biener <rguenther@suse.de>
12690 PR tree-optimization/84803
12691 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
12692 for refs DR analysis didn't process.
12694 2018-03-12 Richard Biener <rguenther@suse.de>
12696 PR tree-optimization/84777
12697 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
12698 force-vectorize loops ignore whether we are optimizing for size.
12700 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
12702 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
12703 (TARGET_MD_ASM_ADJUST): Define.
12705 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
12706 Kito Cheng <kito.cheng@gmail.com>
12707 Chung-Ju Wu <jasonwucj@gmail.com>
12709 * config/nds32/nds32.c (nds32_compute_stack_frame,
12710 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
12711 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
12712 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
12713 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
12714 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
12715 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
12716 * config/nds32/nds32.md (prologue, epilogue): Use macro
12717 NDS32_V3PUSH_AVAILABLE_P to do checking.
12719 2018-03-11 Jakub Jelinek <jakub@redhat.com>
12722 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
12723 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
12724 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
12725 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
12726 addition of most attributes on !orig_type_die or the attribute not
12727 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
12729 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
12730 Chung-Ju Wu <jasonwucj@gmail.com>
12732 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
12733 __NDS32_VH__ macro.
12734 * config/nds32/nds32.opt (mvh): New option.
12736 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
12737 Chung-Ju Wu <jasonwucj@gmail.com>
12739 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
12741 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
12742 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
12745 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
12746 Chung-Ju Wu <jasonwucj@gmail.com>
12748 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
12750 * config/nds32/nds32-multiple.md (strlensi): New pattern.
12751 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
12753 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
12754 Kito Cheng <kito.cheng@gmail.com>
12755 Chung-Ju Wu <jasonwucj@gmail.com>
12757 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
12758 UNSPEC_FFMISM and UNSPEC_FLMISM.
12759 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
12760 for ffb, ffmism and flmism.
12761 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
12762 (unspec_ffmism): Ditto.
12763 (unspec_flmism): Ditto.
12764 (nds32_expand_builtin_impl): Check if string extension is available.
12765 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
12766 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
12768 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
12771 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
12774 * lra-assigns.c (assign_by_spills): Return a flag of reload
12775 assignment failure. Do not process the reload assignment
12776 failures. Do not spill other reload pseudos if they has the same
12778 (lra_assign): Add a return arg. Set up from the result of
12779 assign_by_spills call.
12780 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
12781 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
12782 usage_insns if it is not NULL.
12783 (spill_hard_reg_in_range): New function.
12784 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
12785 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
12786 function prototypes.
12787 (lra_assign): Change prototype.
12788 * lra.c (lra): Add code to deal with fails by splitting hard reg
12791 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
12794 * config/i386/i386.opt: Replace Enforcment with Enforcement.
12796 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
12799 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
12800 (dw_val_node): Add val_symbolic_view.
12801 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
12802 (symview_upper_bound): New.
12803 (new_line_info_table): Initialize symviews_since_reset.
12804 (dwarf2out_source_line): Count symviews_since_reset and set
12805 symview_upper_bound.
12806 (dw_val_equal_p): Handle symview.
12807 (add_AT_symview): New.
12808 (print_dw_val): Handle symview.
12809 (attr_checksum, attr_checksum_ordered): Likewise.
12810 (same_dw_val_p, size_of_die): Likewise.
12811 (value_format, output_die): Likewise.
12812 (add_high_low_attributes): Use add_AT_symview for entry_view.
12813 (dwarf2out_finish): Reset symview_upper_bound, clear
12816 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
12819 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
12820 Add strict argument and use it.
12821 (rs6000_split_multireg_move): Update for new strict argument.
12822 (mem_operand_gpr): Disallow all non-offsettable addresses.
12823 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
12825 2018-03-09 Jakub Jelinek <jakub@redhat.com>
12828 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
12829 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
12830 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
12833 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
12834 decl, use remap_type if we want to use the type.
12836 2018-03-09 Martin Sebor <msebor@redhat.com>
12838 PR tree-optimization/84526
12839 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
12841 (builtin_access::generic_overlap): Be prepared to handle non-array
12844 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
12846 PR rtl-optimization/84682
12847 * lra-constraints.c (process_address_1): Check is_address flag
12848 for address constraints.
12849 (process_alt_operands): Likewise.
12850 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
12851 preprocess_constraints.
12852 * recog.h (preprocess_constraints): Add oploc parameter.
12854 * recog.c (preprocess_constraints): Test address_operand for
12855 CT_ADDRESS constraints.
12857 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
12860 * lra-assigns.c (assign_by_spills): Return a flag of reload
12861 assignment failure. Do not process the reload assignment
12862 failures. Do not spill other reload pseudos if they has the same
12864 (lra_assign): Add a return arg. Set up from the result of
12865 assign_by_spills call.
12866 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
12867 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
12868 usage_insns if it is not NULL.
12869 (spill_hard_reg_in_range): New function.
12870 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
12871 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
12872 function prototypes.
12873 (lra_assign): Change prototype.
12874 * lra.c (lra): Add code to deal with fails by splitting hard reg
12877 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12880 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
12881 Accept complain bool parameter. Only emit errors if it is true.
12882 (arm_parse_cpu_option_name): Likewise.
12883 (arm_target_thumb_only): Adjust callers of the above.
12884 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
12885 prototype to take a default true bool parameter.
12886 (arm_parse_arch_option_name): Likewise.
12888 2018-03-09 David Malcolm <dmalcolm@redhat.com>
12889 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
12893 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
12894 * configure: Regenerate.
12895 * configure.ac ("linker --version-script option"): New.
12896 ("linker soname option"): New.
12898 2018-03-09 Richard Biener <rguenther@suse.de>
12900 PR tree-optimization/84775
12901 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
12902 immediate uses of predicate stmts and mark them modified.
12905 PR tree-optimization/84178
12906 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
12908 (version_loop_for_if_conversion): Delay update_ssa call.
12909 (tree_if_conversion): Delay update_ssa until after predicate
12912 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
12915 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
12916 when the function accesses prior frames.
12918 2018-03-08 Jakub Jelinek <jakub@redhat.com>
12921 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
12922 gen_llsym, otherwise call maybe_gen_llsym.
12924 PR inline-asm/84742
12925 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
12926 has ',' character inside of it.
12928 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12931 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
12932 as clobbering CC_REGNUM.
12934 2018-03-08 Richard Biener <rguenther@suse.de>
12936 PR middle-end/84552
12937 * tree-scalar-evolution.c: Include tree-into-ssa.h.
12938 (follow_copies_to_constant): Do not follow SSA names registered
12941 2018-03-08 Richard Biener <rguenther@suse.de>
12943 PR tree-optimization/84178
12944 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
12946 (version_loop_for_if_conversion): Delay update_ssa call.
12947 (tree_if_conversion): Delay update_ssa until after predicate
12950 2018-03-08 David Malcolm <dmalcolm@redhat.com>
12952 PR tree-optimization/84178
12953 * tree-if-conv.c (release_bb_predicate): Remove the
12954 the assertion that the stmts have NULL use_ops.
12955 Discard the statements, asserting that they haven't
12956 yet been added to a BB.
12958 2018-03-08 Richard Biener <rguenther@suse.de>
12960 PR tree-optimization/84746
12961 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
12962 (phi_translate): Pass in destination ANTIC_OUT set.
12963 (phi_translate_1): Likewise. For a simplified result lookup
12964 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
12965 (phi_translate_set): Adjust.
12966 (do_pre_regular_insertion): Likewise.
12967 (do_pre_partial_partial_insertion): Likewise.
12969 2018-03-08 Martin Liska <mliska@suse.cz>
12971 PR gcov-profile/84735
12972 * doc/gcov.texi: Document usage of profile files.
12973 * gcov-io.h: Document changes in the format.
12975 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
12979 * dwarf2out.c (struct dw_line_info_table): Update comments for
12981 (FORCE_RESET_NEXT_VIEW): New.
12982 (FORCE_RESETTING_VIEW_P): New.
12983 (RESETTING_VIEW_P): Check for -1 too.
12984 (ZERO_VIEW_P): Likewise.
12985 (new_line_info_table): Force-reset next view.
12986 (dwarf2out_begin_function): Likewise.
12987 (dwarf2out_source_line): Simplify zero_view_p initialization.
12988 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
12989 view directly. Omit view when omitting .loc at line 0.
12991 2018-03-08 Jakub Jelinek <jakub@redhat.com>
12993 PR tree-optimization/84740
12994 * tree-switch-conversion.c (process_switch): Call build_constructors
12995 only if info.phi_count is non-zero.
12997 PR tree-optimization/84739
12998 * tree-tailcall.c (find_tail_calls): Check call arguments against
12999 DECL_ARGUMENTS (current_function_decl) rather than
13000 DECL_ARGUMENTS (func) when checking for tail recursion.
13002 2018-03-07 Jakub Jelinek <jakub@redhat.com>
13004 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
13005 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
13006 Volker Reichelt's entry and add entries for people that perform
13007 GCC fuzzy testing and report numerous bugs.
13009 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
13012 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
13013 readonly data in sdata, if that is disabled.
13014 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
13015 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
13016 -mreadonly-in-sdata option.
13018 2018-03-07 Martin Sebor <msebor@redhat.com>
13020 PR tree-optimization/84468
13021 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
13022 basic block when looking for nul assignment.
13024 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
13027 * except.h (output_function_exception_table): Adjust prototype.
13028 * except.c (output_function_exception_table): Remove FNNAME parameter
13029 and add SECTION parameter. Ouput one part of the table at a time.
13030 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
13031 the first part of the exception table and emit unwind directives.
13032 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
13033 (i386_pe_seh_cold_init): Likewise.
13034 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
13035 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
13036 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
13037 (ix86_output_call_insn): Emit a nop in one more case for SEH.
13038 * config/i386/winnt.c: Include except.h.
13039 (struct seh_frame_state): Add reg_offset, after_prologue and
13040 in_cold_section fields.
13041 (i386_pe_seh_end_prologue): Set seh->after_prologue.
13042 (i386_pe_seh_cold_init): New function.
13043 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
13044 to seh->in_cold_section.
13045 (seh_emit_push): Record the offset of the push.
13046 (seh_emit_save): Record the offet of the save.
13047 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
13048 Test seh->after_prologue to disregard the epilogue.
13049 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
13050 (i386_pe_end_cold_function): New function.
13052 2018-03-07 Jakub Jelinek <jakub@redhat.com>
13055 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
13056 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
13059 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
13061 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
13062 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
13064 PR middle-end/84723
13065 * multiple_target.c: Include tree-inline.h and intl.h.
13066 (expand_target_clones): Diagnose and fail if node->definition and
13067 !tree_versionable_function_p (node->decl).
13069 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
13071 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
13073 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
13074 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
13075 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
13077 2018-03-06 Jakub Jelinek <jakub@redhat.com>
13080 * combine.c (try_combine): Use reg_or_subregno instead of handling
13081 just paradoxical SUBREGs and REGs.
13083 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
13085 * config/arc/arc.c (arc_finalize_pic): Remove function.
13086 (arc_must_save_register): We use single base PIC register, remove
13087 checks to save/restore the PIC register.
13088 (arc_expand_prologue): Likewise.
13089 * config/arc/arc-protos.h (arc_set_default_type_attributes):
13091 (arc_verify_short): Likewise.
13092 (arc_attr_type): Likewise.
13093 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
13094 (walk_stores): Likewise.
13095 (arc_address_cost): Make it static.
13096 (arc_verify_short): Likewise.
13097 (branch_dest): Likewise.
13098 (arc_attr_type): Likewise.
13099 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
13100 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
13101 (arc_final_prescan_insn): Remove inserting the nops due to
13102 hardware hazards. It is done in reorg step.
13103 (insn_length_variant_t): Remove.
13104 (insn_length_parameters_t): Likewise.
13105 (arc_insn_length_parameters): Likewise.
13106 (arc_get_insn_variants): Likewise.
13107 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
13109 2018-03-06 Jakub Jelinek <jakub@redhat.com>
13111 PR inline-asm/84683
13112 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
13115 PR tree-optimization/84687
13116 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
13118 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
13120 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13122 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
13123 Rename to ppc_speculation_barrier.
13124 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
13125 __builtin_ppc_speculation_barrier.
13127 2018-03-05 Jakub Jelinek <jakub@redhat.com>
13130 * combine.c (combine_simplify_rtx): Don't try to simplify if
13131 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
13134 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
13136 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
13137 to 32 bytes when compiling for POWER9.
13139 2018-03-05 Jakub Jelinek <jakub@redhat.com>
13142 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
13143 regparm >= 3 with no arg reg available also for calls with
13144 flag_force_indirect_call. Pass decl to ix86_function_regparm.
13147 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
13149 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
13151 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
13154 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
13156 2018-03-05 Richard Biener <rguenther@suse.de>
13158 PR tree-optimization/84486
13159 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
13160 When inserting a __builtin_assume_aligned call set the LHS
13161 SSA name alignment info accordingly.
13163 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
13165 PR tree-optimization/84114
13166 * config/aarch64/aarch64.c (aarch64_reassociation_width)
13167 Avoid reassociation of FLOAT_MODE addition.
13169 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
13171 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
13172 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
13173 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
13174 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
13175 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
13176 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
13177 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
13179 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
13180 __builtin_ia32_wbinvd): New builtins.
13181 (SPECIAL_ARGS2): New.
13182 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
13183 (SPECIAL_ARGS2): New.
13184 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
13185 (ix86_valid_target_attribute_inner_p): Ditto.
13186 (ix86_init_mmx_sse_builtins): Add special_args2.
13187 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
13188 TARGET_WBNOINVD_P): New.
13189 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
13190 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
13191 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
13192 * config/i386/immintrin.h (_wbinvd): New intrinsic.
13193 * config/i386/pconfigintrin.h: New file.
13194 * config/i386/wbnoinvdintrin.h: Ditto.
13195 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
13197 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
13199 2018-03-05 Richard Biener <rguenther@suse.de>
13201 PR tree-optimization/84670
13202 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
13204 (BB_VISITED_WITH_VISITED_SUCCS): New define.
13205 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
13206 (compute_antic_aux): Only assert the number of values in ANTIC_IN
13207 doesn't grow if all successors (recursively) were visited at least
13210 2018-03-05 Richard Biener <rguenther@suse.de>
13212 PR tree-optimization/84650
13213 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
13214 if executed in the loop pipeline.
13216 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
13218 * doc/configfiles.texi (Configuration Files): Move info about
13219 conditionalizing $target-protos.h to...
13220 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
13221 differs from $target-protos.h.
13223 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
13224 Chung-Ju Wu <jasonwucj@gmail.com>
13226 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
13227 * config/nds32/nds32-multiple.md (setmemsi): Define.
13228 * config/nds32/nds32-memory-manipulation.c
13229 (nds32_gen_dup_4_byte_to_word_value): New.
13230 (emit_setmem_word_loop): New.
13231 (emit_setmem_byte_loop): New.
13232 (nds32_expand_setmem_loop): New.
13233 (nds32_expand_setmem_loop_v3m): New.
13234 (nds32_expand_setmem_unroll): New.
13235 (nds32_expand_setmem): New.
13237 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
13238 Chung-Ju Wu <jasonwucj@gmail.com>
13240 * config/nds32/nds32-memory-manipulation.c
13241 (nds32_emit_load_store): New.
13242 (nds32_emit_post_inc_load_store): New.
13243 (nds32_emit_mem_move): New.
13244 (nds32_emit_mem_move_block): New.
13245 (nds32_expand_movmemsi_loop_unknown_size): New.
13246 (nds32_expand_movmemsi_loop_known_size): New.
13247 (nds32_expand_movmemsi_loop): New.
13248 (nds32_expand_movmemsi_unroll): New.
13249 (nds32_expand_movmemqi): Rename ...
13250 (nds32_expand_movmemsi): ... to this.
13251 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
13252 (movmemsi): ... to this.
13253 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
13254 (nds32_expand_movmemsi): ... to this.
13256 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
13257 Monk Chiang <sh.chiang04@gmail.com>
13258 Chung-Ju Wu <jasonwucj@gmail.com>
13260 * config/nds32/nds32-protos.h
13261 (nds32_expand_load_multiple): New arguments.
13262 (nds32_expand_store_multiple): Ditto.
13263 (nds32_valid_multiple_load_store): Rename ...
13264 (nds32_valid_multiple_load_store_p): ... to this.
13265 * config/nds32/nds32-memory-manipulation.c
13266 (nds32_expand_load_multiple): Refine implementation.
13267 (nds32_expand_store_multiple): Ditto.
13268 * config/nds32/nds32-multiple.md
13269 (load_multiple): Update nds32_expand_load_multiple interface.
13270 (store_multiple): Update nds32_expand_store_multiple interface.
13271 * config/nds32/nds32-predicates.c
13272 (nds32_valid_multiple_load_store): Rename ...
13273 (nds32_valid_multiple_load_store_p): ... to this and refine
13275 * config/nds32/predicates.md
13276 (nds32_load_multiple_and_update_address_operation): New predicate.
13277 (nds32_store_multiple_and_update_address_operation): New predicate.
13279 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
13280 Chung-Ju Wu <jasonwucj@gmail.com>
13282 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
13283 (combo): New attribute.
13284 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
13286 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
13288 * config/nds32/nds32.opt: Change -mcmodel= default value.
13290 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
13291 Monk Chiang <sh.chiang04@gmail.com>
13292 Chung-Ju Wu <jasonwucj@gmail.com>
13294 * config/nds32/constants.md (unspec_element): New enum.
13295 * config/nds32/constraints.md (Umw): New constraint.
13296 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
13297 * config/nds32/nds32-intrinsic.md: Likewise.
13298 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
13299 (nds32_valid_smw_lwm_base_p): New.
13300 (nds32_output_smw_single_word): New.
13301 (nds32_output_lmw_single_word): New.
13302 (nds32_expand_unaligned_load): New.
13303 (nds32_expand_unaligned_store): New.
13304 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
13305 (nds32_output_smw_single_word): Declare.
13306 (nds32_output_lmw_single_word): Declare.
13307 (nds32_expand_unaligned_load): Declare.
13308 (nds32_expand_unaligned_store): Declare.
13309 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
13310 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
13311 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
13312 NDS32_BUILTIN_UASTORE_DW.
13313 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
13316 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
13317 Kito Cheng <kito.cheng@gmail.com>
13318 Chung-Ju Wu <jasonwucj@gmail.com>
13320 * config/nds32/nds32-intrinsic.c
13321 (nds32_expand_builtin_null_ftype_reg): Delete.
13322 (nds32_expand_builtin_reg_ftype_imm): Ditto.
13323 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
13324 (nds32_read_argument): New.
13325 (nds32_legitimize_target): Ditto.
13326 (nds32_legitimize_argument): Ditto.
13327 (nds32_check_constant_argument): Ditto.
13328 (nds32_expand_unop_builtin): Ditto.
13329 (nds32_expand_unopimm_builtin): Ditto.
13330 (nds32_expand_binop_builtin): Ditto.
13331 (nds32_builtin_decl_impl): Ditto.
13332 (builtin_description): Ditto.
13333 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
13334 (nds32_init_builtins_impl): Ditto.
13335 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
13336 (nds32_builtin_decl): New.
13337 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
13338 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
13340 2018-03-02 Jeff Law <law@redhat.com>
13342 * reorg.c (stop_search_p): Handle DEBUG_INSN.
13343 (redundant_insn, fill_simple_delay_slots): Likewise.
13344 (fill_slots_from_thread): Likewise.
13345 * resource.c (mark_referenced_resources): Likewise.
13346 (mark_set_resources, find_dead_or_set_registers): Likewise.
13348 2018-03-02 Jakub Jelinek <jakub@redhat.com>
13350 * substring-locations.h (format_warning_va): Formatting fix for
13351 ATTRIBUTE_GCC_DIAG.
13352 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
13354 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
13355 * substring-locations.c: Include intl.h.
13356 (format_warning_va): Turned into small wrapper around
13357 format_warning_n_va, renamed to ...
13358 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
13359 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
13361 (format_warning_at_substring_n): New function.
13362 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
13363 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
13364 format_warning_at_substring with just a shorter name instead of
13365 const function pointer.
13366 (fmtwarn_n): New function.
13367 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
13368 appropriate, get rid of all the fmtstr temporaries, move conditionals
13369 with G_() wrapped string literals directly into fmtwarn arguments,
13370 cast dir.len to (int), formatting fixes.
13372 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
13374 * doc/invoke.texi: Remove "Cilk Plus" references.
13376 2018-03-02 Jakub Jelinek <jakub@redhat.com>
13377 Richard Biener <rguenther@suse.de>
13380 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
13381 for error or warning attributes if CALL_FROM_THUNK_P is set.
13384 2018-03-02 Jakub Jelinek <jakub@redhat.com>
13387 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
13388 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
13391 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
13392 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
13394 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
13395 instead of -1U in last predictors element's probability member.
13397 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
13400 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
13401 arguments if they are comparable.
13403 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
13405 PR tree-optimization/84634
13406 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
13407 masks and masked_loop_p with a single loop_masks, making sure it's
13408 null for bb vectorization.
13410 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
13412 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
13413 (vect_analyze_data_ref_access): Use loop->safe_len rather than
13414 loop->force_vectorize to check whether there is no alias.
13416 2018-03-02 Jakub Jelinek <jakub@redhat.com>
13419 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
13421 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
13423 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
13424 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
13425 instead of a loop around prev_real_insn.
13426 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
13429 PR inline-asm/84625
13430 * config/i386/i386.c (ix86_print_operand): Use conditional
13431 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
13434 2018-03-02 Richard Biener <rguenther@suse.de>
13436 PR tree-optimization/84427
13437 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
13438 (bitmap_set_subtract_values): Rewrite to handle multiple
13441 (prune_clobbered_mems): Likewise.
13442 (phi_translate): Take edge instead of pred/phiblock.
13443 (phi_translate_1): Likewise.
13444 (phi_translate_set): Likewise. Insert all translated
13445 exprs for a value into the set, keeping possibly multiple
13446 expressions per value.
13447 (compute_antic_aux): Adjust for phi_translate changes.
13448 When intersecting union the expressions and prune those
13449 not in the final value set, keeping possibly multiple
13450 expressions per value. Do not use value-insertion
13451 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
13452 all expressions. Add verification that the value-sets
13453 only shrink during iteration.
13454 (compute_partial_antic_aux): Adjust for the phi_translate changes.
13455 (do_pre_regular_insertion): Likewise.
13456 (do_pre_partial_partial_insertion): Likewise.
13458 2018-03-02 Richard Biener <rguenther@suse.de>
13461 * config/darwin.c (saved_debug_info_level): New static global.
13462 (darwin_asm_lto_start): Disable debug info generation for LTO out.
13463 (darwin_asm_lto_end): Restore debug info generation settings.
13465 2018-03-01 Martin Liska <mliska@suse.cz>
13468 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
13469 volatile arguments.
13471 2018-03-01 Richard Biener <rguenther@suse.de>
13474 * dwarf2out.c (gen_variable_die): Properly handle late VLA
13475 type annotation with LTO when debug was disabled at compile-time.
13477 2018-03-01 Matthew Fortune <mfortune@gmail.com>
13479 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
13481 (mips_final_postscan_insn): Likewise.
13483 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
13485 PR rtl-optimization/84528
13486 * alias.c (init_alias_target): Add commentary.
13487 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
13488 a unique base value if the frame pointer is not eliminated
13489 to the stack pointer.
13491 2018-03-01 Tom de Vries <tom@codesourcery.com>
13493 PR rtl-optimization/83327
13494 * lra-int.h (hard_regs_spilled_into): Declare.
13495 * lra.c (hard_regs_spilled_into): Define.
13496 (init_reg_info): Init hard_regs_spilled_into.
13497 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
13498 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
13499 (process_bb_lives): Handle hard_regs_spilled_into.
13500 (lra_create_live_ranges_1): Before doing liveness propagation, clear
13501 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
13503 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
13505 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
13506 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
13507 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
13508 * config/rs6000/aix72.h: New file.
13510 2018-02-28 Jakub Jelinek <jakub@redhat.com>
13512 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
13513 instead of warning_at with conditional singular and plural messages
13517 * stor-layout.c (update_alignment_for_field): For
13518 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
13519 && !DECL_PACKED (field), do the alignment update, just use
13520 only desired_align instead of MAX (type_align, desired_align)
13522 (place_field): Don't do known_align < desired_align handling
13523 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
13524 is non-NULL, instead do it after rli->prev_field handling and
13525 only if not within a bitfield word. For DECL_PACKED (field)
13526 use type_align of BITS_PER_UNIT.
13528 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
13530 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
13531 superfluous parentheses and trailing spaces.
13533 2018-02-28 Richard Biener <rguenther@suse.de>
13535 PR tree-optimization/84584
13536 * graphite-scop-detection.c (scop_detection::add_scop): Discard
13537 SCoPs with fake exit edge.
13539 2018-02-28 Martin Liska <mliska@suse.cz>
13542 * timevar.c (timer::print): Fix format to properly print 100%
13545 2018-02-28 Richard Biener <rguenther@suse.de>
13547 PR middle-end/84607
13548 * genmatch.c (capture_info::walk_match): Do not mark
13549 captured expressions without operands as expr_p given
13550 they act more like predicates and should be subject to
13551 "lost tail" side-effect preserving.
13553 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
13555 PR rtl-optimization/81611
13556 * auto-inc-dec.c (attempt_change): Move dead note from
13557 mem_insn if it's the next use of regno
13558 (find_address): Take address use of reg holding
13559 non-incremented value. Add parm to limit search to the named
13561 (merge_in_block): Attempt to use a mem insn that is the next
13562 use of the original regno.
13564 2018-02-27 Martin Sebor <msebor@redhat.com>
13567 * doc/invoke.texi (-Wmissing-attributes): New option.
13568 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
13570 2018-02-27 Martin Sebor <msebor@redhat.com>
13572 PR translation/84207
13573 * diagnostic-core.h (warning_n, error_n, inform_n): Change
13574 n argument to unsigned HOST_WIDE_INT.
13575 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
13576 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
13577 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
13578 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
13580 2018-02-27 Richard Biener <rguenther@suse.de>
13582 PR tree-optimization/84512
13583 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
13584 Do not use the estimate returned from record_stmt_cost for
13585 the scalar iteration cost but sum properly using add_stmt_cost.
13587 2018-02-27 Richard Biener <rguenther@suse.de>
13589 PR tree-optimization/84466
13590 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
13591 Adjust last change to less strictly validate use operands.
13593 2018-02-27 Martin Liska <mliska@suse.cz>
13595 PR gcov-profile/84548
13596 * gcov.c (process_file): Allow partial overlap and consider it
13597 also as group functions.
13598 (output_lines): Properly calculate range of lines for a group.
13600 2018-02-27 Martin Liska <mliska@suse.cz>
13602 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
13603 'ggc' suffixes. Change first column width.
13604 (timer::print): Fix formatting of the column.
13606 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
13608 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
13609 preserve inline entry blocks for the sake of debug inline
13610 entry point markers alone.
13611 (remove_unused_locals): Suggest in comments a better place to
13612 force the preservation of inline entry blocks that are
13613 otherwise unused, but do not preserve them.
13615 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
13617 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
13619 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
13622 * config/i386/constraints.md (Bs): Replace
13623 ix86_indirect_branch_register with
13624 TARGET_INDIRECT_BRANCH_REGISTER.
13626 * config/i386/i386.md (indirect_jump): Likewise.
13627 (tablejump): Likewise.
13628 (*sibcall_memory): Likewise.
13629 (*sibcall_value_memory): Likewise.
13630 Peepholes of indirect call and jump via memory: Likewise.
13631 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
13632 (*sibcall_value_GOT_32): Likewise.
13633 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13634 (GOT_memory_operand): Likewise.
13635 (call_insn_operand): Likewise.
13636 (sibcall_insn_operand): Likewise.
13637 (GOT32_symbol_operand): Likewise.
13638 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
13640 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
13642 PR rtl-optimization/83496
13643 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
13644 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
13645 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
13646 redundant insn, if any.
13647 (relax_delay_slots): Likewise.
13648 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
13650 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
13652 PR tree-optimization/83965
13653 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
13654 that grouped statements are part of a reduction chain. Return
13655 true if the statement is not marked as a reduction itself but
13656 is part of a group.
13657 (vect_recog_dot_prod_pattern): Don't check whether the statement
13658 is part of a group here.
13659 (vect_recog_sad_pattern): Likewise.
13660 (vect_recog_widen_sum_pattern): Likewise.
13662 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
13665 * final.c (rest_of_clean_state): Also look for calls inside sequences.
13667 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
13670 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
13672 (ix86_output_indirect_function_return): New prototype.
13673 (ix86_split_simple_return_pop_internal): Likewise.
13674 * config/i386/i386.c (indirect_return_via_cx): New.
13675 (indirect_return_via_cx_bnd): Likewise.
13676 (indirect_thunk_name): Handle return va CX_REG.
13677 (output_indirect_thunk_function): Create alias for
13678 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
13679 (ix86_output_indirect_jmp): Remove the bool argument.
13680 (ix86_output_indirect_function_return): New function.
13681 (ix86_split_simple_return_pop_internal): Likewise.
13682 * config/i386/i386.md (*indirect_jump): Don't pass false
13683 to ix86_output_indirect_jmp.
13684 (*tablejump_1): Likewise.
13685 (simple_return_pop_internal): Change it to define_insn_and_split.
13686 Call ix86_split_simple_return_pop_internal to split it for
13687 -mfunction-return=.
13688 (simple_return_indirect_internal): Call
13689 ix86_output_indirect_function_return instead of
13690 ix86_output_indirect_jmp.
13692 2018-02-26 Jakub Jelinek <jakub@redhat.com>
13695 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
13696 memset and value initialization afterwards.
13698 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
13700 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
13702 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
13705 * common/config/aarch64/aarch64-common.c
13706 (aarch_option_optimization_table[]): Switch
13707 off fomit-frame-pointer
13709 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
13710 Chung-Ju Wu <jasonwucj@gmail.com>
13712 * config/nds32/nds32-multiple.md (load_multiple): Disallow
13714 (store_multiple): Ditto.
13716 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
13718 * config.gcc: Add --with-cpu support for nds32 target.
13719 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
13720 * config/nds32/nds32.opt: Add -mcpu= option.
13722 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
13724 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
13725 isel=yes): Warn for these deprecated options.
13727 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
13729 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
13730 ISA_2_5_MASKS_EMBEDDED.
13732 2018-02-23 Jakub Jelinek <jakub@redhat.com>
13734 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
13735 p->max as pointers rather than using iterative_hash_expr.
13737 2018-02-23 Carl Love <cel@us.ibm.com>
13739 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
13740 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
13742 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
13743 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
13744 P8V_BUILTIN_VEC_VUNSIGNED2.
13746 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
13749 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
13750 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
13752 * lra-constraints.c (curr_insn_transform): Set up
13753 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
13755 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
13756 * lra-eliminations.c (spill_pseudos): Ditto.
13757 (process_insn_for_elimination): Ditto.
13758 * lra-lives.c (reg_early_clobber_p): Use the new macros.
13759 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
13760 LRA_NON_CLOBBERED_ALT.
13762 2018-02-22 Martin Sebor <msebor@redhat.com>
13764 PR tree-optimization/84480
13765 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
13766 to maybe_diag_stxncpy_trunc. Call it.
13767 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
13768 from gimple_fold_builtin_strcpy. Print inlining stack.
13769 (handle_builtin_stxncpy): Print inlining stack.
13770 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
13772 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
13775 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
13776 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
13777 and -fcheck-pointer-bounds are used together.
13778 (indirect_thunk_prefix): New enum.
13779 (indirect_thunk_need_prefix): New function.
13780 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
13781 "_nt" instead of "_bnd" for NOTRACK prefix.
13782 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
13783 (output_indirect_thunk_function): Likewise.
13785 (ix86_code_end): Update output_indirect_thunk_function calls.
13786 (ix86_output_indirect_branch_via_reg): Replace
13787 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
13788 (ix86_output_indirect_branch_via_push): Likewise.
13789 (ix86_output_function_return): Likewise.
13790 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
13791 incompatible with -fcf-protection=branch and
13792 -fcheck-pointer-bounds.
13794 2018-02-22 Steve Ellcey <sellcey@cavium.com>
13797 * config/aarch64/aarch64.c (aarch64_print_address_internal):
13798 Change gcc_assert call to output_operand_lossage.
13800 2018-02-22 Steve Ellcey <sellcey@cavium.com>
13802 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
13804 2018-02-22 DJ Delorie <dj@redhat.com>
13805 Sebastian Perta <sebastian.perta@renesas.com>
13806 Oleg Endo <olegendo@gcc.gnu.org>
13808 * config/rx/rx.c (rx_rtx_costs): New function.
13809 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
13811 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
13813 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
13815 2018-02-22 Martin Liska <mliska@suse.cz>
13818 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
13819 Add "native" as a possible value.
13821 2018-02-22 Martin Liska <mliska@suse.cz>
13824 * config/i386/i386.c (ix86_option_override_internal):
13825 Add "native" as a possible value for -march and -mtune.
13827 2018-02-22 Jakub Jelinek <jakub@redhat.com>
13830 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
13831 to all type variants.
13833 PR tree-optimization/84503
13834 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
13835 width as info->bitpos + info->bitsize - start.
13836 (merged_store_group::merge_overlapping): Simplify width computation.
13837 (check_no_overlap): New function.
13838 (imm_store_chain_info::try_coalesce_bswap): Compute expected
13839 start + width and last_order of the group, fail if check_no_overlap
13841 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
13842 to group if check_no_overlap fails.
13844 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
13846 * config/rs6000/altivec.md: Delete contraint arguments to
13847 define_expand, define_split, and define_peephole2, and in
13848 define_insn_and_split if always unused.
13849 * config/rs6000/darwin.md: Ditto.
13850 * config/rs6000/dfp.md: Ditto.
13851 * config/rs6000/rs6000.md: Ditto.
13852 * config/rs6000/sync.md: Ditto.
13853 * config/rs6000/vector.md: Ditto.
13854 * config/rs6000/vsx.md: Ditto.
13856 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
13858 * config/rs6000/altivec.md: Write output control strings as braced
13859 blocks instead of double-quoted strings.
13860 * config/rs6000/darwin.md: Ditto.
13861 * config/rs6000/rs6000.md: Ditto.
13862 * config/rs6000/vector.md: Ditto.
13863 * config/rs6000/vsx.md: Ditto.
13865 2018-02-21 Jason Merrill <jason@redhat.com>
13867 PR c++/84314 - ICE with templates and fastcall attribute.
13868 * attribs.c (build_type_attribute_qual_variant): Remove assert.
13870 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
13872 * ipa-cp.c (determine_versionability): Fix comment typos.
13874 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
13877 * ipa-cp.c (determine_versionability): Do not version functions caling
13880 2018-02-21 Martin Liska <mliska@suse.cz>
13883 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
13884 Add "native" as a possible value.
13885 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
13886 the macro when native cpu detection is available.
13888 2018-02-21 Martin Liska <mliska@suse.cz>
13891 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
13892 Add "native" as a possible value.
13893 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
13894 when native cpu detection is available.
13896 2018-02-21 Jakub Jelinek <jakub@redhat.com>
13897 Martin Sebor <msebor@redhat.com>
13899 PR tree-optimization/84478
13900 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
13902 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
13903 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
13904 support which is conservatively correct, for 2 only stay conservative
13905 for maxlen. Formatting and comment capitalization fixes. Add STRICT
13906 argument to the 2 argument get_range_strlen, adjust 6 arg
13907 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
13909 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
13910 (gimple_fold_builtin_strlen): Pass true as last argument to
13913 2018-02-20 Martin Sebor <msebor@redhat.com>
13915 PR middle-end/84095
13916 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
13917 (builtin_memref::set_base_and_offset): Same. Handle inner references.
13918 (builtin_memref::builtin_memref): Factor out parts into
13919 set_base_and_offset and call it.
13921 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
13923 PR middle-end/84406
13924 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
13925 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
13926 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
13927 search at the associated MODE_INT.
13929 2018-02-20 Jeff Law <law@redhat.com>
13931 PR middle-end/82123
13932 PR tree-optimization/81592
13933 PR middle-end/79257
13934 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
13935 for range data rather than using global data.
13936 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
13937 range data rather than using global data.
13938 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
13939 pass it to children as needed.
13940 (struct directive::fmtresult): Similarly.
13941 (struct directive::set_width): Similarly.
13942 (struct directive::set_precision): Similarly.
13943 (format_integer, format_directive, parse_directive): Similarly.
13944 (format_none): Accept unnamed vr_values parameter.
13945 (format_percent, format_floating, format_character): Similarly.
13946 (format_string, format_plain): Similarly.
13947 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
13948 the EVRP range analyzer for range data rather than using global data.
13949 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
13950 gimple-ssa-evrp-analyze.h
13951 (class sprintf_dom_walker): Add after_dom_children member function.
13952 Add evrp_range_analyzer member.
13953 (sprintf_dom_walker::before_dom_children): Call into the EVRP
13954 range analyzer as needed.
13955 (sprintf_dom_walker::after_dom_children): New member function.
13956 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
13958 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
13959 (evrp_range_analyzer::pop_to_marker): Likewise.
13961 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
13963 PR tree-optimization/84419
13964 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
13965 with the required type if its current type is compatible but
13968 2018-02-20 Jakub Jelinek <jakub@redhat.com>
13970 PR middle-end/82004
13971 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
13972 after vectorization.
13974 2018-02-20 Martin Liska <mliska@suse.cz>
13977 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
13978 possible values if we don't have a hint.
13980 2018-02-20 Martin Liska <mliska@suse.cz>
13984 * final.c (shorten_branches): Build align_tab array with one
13986 * opts.c (finish_options): Add alignment option limit check.
13987 (MAX_CODE_ALIGN): Likewise.
13988 (MAX_CODE_ALIGN_VALUE): Likewise.
13989 * doc/invoke.texi: Document maximum allowed option value for
13990 all -falign-* options.
13992 2018-02-19 Jakub Jelinek <jakub@redhat.com>
13995 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
13996 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
13997 * var-tracking.c (emit_note_insn_var_location): Remove all references
13998 to NOTE_INSN_CALL_ARG_LOCATION.
13999 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
14000 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
14001 Use copy_rtx_if_shared.
14002 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
14003 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
14004 (dwarf2out_var_location): Remove handling of
14005 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
14007 * final.c (final_scan_insn): Remove all references to
14008 NOTE_INSN_CALL_ARG_LOCATION.
14009 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
14010 before dumping final insns.
14011 * except.c (emit_note_eh_region_end): Remove all references to
14012 NOTE_INSN_CALL_ARG_LOCATION.
14013 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
14014 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
14015 * config/arc/arc.c (hwloop_optimize): Likewise.
14016 * config/arm/arm.c (create_fix_barrier): Likewise.
14017 * config/s390/s390.c (s390_chunkify_start): Likewise.
14018 * config/sh/sh.c (find_barrier): Likewise.
14019 * config/i386/i386.c (rest_of_insert_endbranch,
14020 ix86_seh_fixup_eh_fallthru): Likewise.
14021 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
14022 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
14023 * config/frv/frv.c (frv_function_prologue): Likewise.
14024 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
14026 (note_outside_basic_block_p): Remove all references to
14027 NOTE_INSN_CALL_ARG_LOCATION.
14028 * gengtype.c (adjust_field_rtx_def): Likewise.
14029 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
14031 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
14032 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
14035 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
14038 PR tree-optimization/84452
14039 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
14040 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
14043 2018-02-19 Martin Liska <mliska@suse.cz>
14046 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
14048 2018-02-19 Martin Liska <mliska@suse.cz>
14049 Richard Sandiford <richard.sandiford@linaro.org>
14051 PR tree-optimization/82491
14052 * gimple-fold.c (get_base_constructor): Make earlier bail out
14055 2018-02-19 Carl Love <cel@us.ibm.com>
14057 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
14058 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
14060 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
14061 P8V_BUILTIN_VEC_NEG.
14063 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
14065 * config/rl78/rl78.md (movdf): New define expand.
14067 2018-02-19 Martin Liska <mliska@suse.cz>
14070 * doc/invoke.texi: Fix typo.
14071 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
14073 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
14075 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
14076 handle rs6000_single_float and rs6000_double_float specially for
14079 2018-02-16 Jeff Law <law@redhat.com>
14081 * config/rx/rx.c (add_pop_cfi_notes): New function.;
14082 (pop_regs): Use it.
14084 2018-02-16 Jakub Jelinek <jakub@redhat.com>
14087 * ipa-inline.c (inline_small_functions): Fix a typo.
14089 2018-02-16 Nathan Sidwell <nathan@acm.org>
14091 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
14093 2018-02-16 Carl Love <cel@us.ibm.com>
14095 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
14096 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
14097 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
14098 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
14099 expansion to P8V_BUILTIN_VEC_FLOAT2.
14101 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
14103 PR rtl-optimization/70023
14104 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
14105 src_regno into account.
14107 2018-02-16 Carl Love <cel@us.ibm.com>
14109 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
14110 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
14111 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
14112 * config/rs6000/rs6000.c: Remove case statements for
14113 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
14114 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
14115 and P9V_BUILTIN_VEC_VINSERT4B.
14116 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
14117 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
14118 * config/rs6000/vsx.md:
14119 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
14122 2018-02-16 Carl Love <cel@us.ibm.com>
14124 * config/rs6000/altivec.h: Add builtin names vec_extract4b
14126 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
14128 * config/rs6000/rs6000-c.c: Add the definitions for
14129 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
14130 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
14131 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
14132 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
14133 definition for insert4b and define insn *insert3b_internal.
14134 * doc/extend.texi: Add documentation for vec_extract4b.
14136 2018-02-16 Nathan Sidwell <nathan@acm.org>
14138 * doc/extend.texi (Backwards Compatibility): Mention friend
14139 injection. Note for-scope is deprecated.
14140 * doc/invoke.texi (-ffriend-injection): Deprecate.
14142 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
14144 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
14145 that moved to I2, also allow destinations that are a paradoxical
14146 subreg (instead of a normal reg).
14148 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
14151 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
14154 2018-02-16 Richard Biener <rguenther@suse.de>
14156 PR tree-optimization/84037
14157 PR tree-optimization/84016
14159 * config/i386/i386.c (ix86_builtin_vectorization_cost):
14160 Adjust vec_construct for the fact we need additional higher latency
14161 128bit inserts for AVX256 and AVX512 vector builds.
14162 (ix86_add_stmt_cost): Scale vector construction cost for
14165 2018-02-16 Richard Biener <rguenther@suse.de>
14167 PR tree-optimization/84417
14168 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
14169 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
14170 (non_rewritable_lvalue_p): Likewise, use poly-ints.
14172 2018-02-16 Martin Liska <mliska@suse.cz>
14175 * internal-fn.def (ASAN_CHECK): Set proper flags.
14176 (ASAN_MARK): Likewise.
14178 2018-02-16 Julia Koval <julia.koval@intel.com>
14180 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
14181 from PTA_CANNONLAKE.
14183 2018-02-16 Jakub Jelinek <jakub@redhat.com>
14186 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
14187 Use ++iter rather than iter++ for std::list iterators.
14188 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
14189 defer deleting them until all nodes in the forest are processed. Do
14190 free even leaf nodes. Change to_process into auto_vec.
14193 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
14194 * vec.h (vec_default_construct): Use memset instead of placement new
14195 if BROKEN_VALUE_INITIALIZATION is defined.
14196 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
14197 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
14200 PR rtl-optimization/83723
14201 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
14202 * lra.c (lra_substitute_pseudo): Likewise. If true, use
14203 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
14205 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
14207 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
14209 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
14211 PR rtl-optimization/81443
14212 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
14213 from inner REGs to paradoxical SUBREGs.
14215 2018-02-16 Richard Biener <rguenther@suse.de>
14217 PR tree-optimization/84399
14218 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
14219 For operands we can analyze at their definition make sure we can
14220 analyze them at each use as well.
14222 2018-02-16 Richard Biener <rguenther@suse.de>
14224 PR tree-optimization/84190
14225 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
14226 volatile accesses if the decl isn't volatile.
14228 2018-02-15 Jason Merrill <jason@redhat.com>
14230 PR c++/84314 - ICE with templates and fastcall attribute.
14231 * attribs.c (build_type_attribute_qual_variant): Don't clobber
14232 TYPE_CANONICAL on an existing type.
14234 2018-02-15 Jakub Jelinek <jakub@redhat.com>
14236 PR tree-optimization/84383
14237 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
14238 dstoff nor call operand_equal_p if dstbase is NULL.
14240 PR tree-optimization/84334
14241 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
14242 also a CONSTANT_CLASS_P, punt.
14244 2018-02-14 Jim Wilson <jimw@sifive.com>
14246 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
14247 first SMALL_OPERAND check. New local min_second_step. Move assert
14248 to where locals are set. Add TARGET_RVC support.
14249 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
14251 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
14253 * doc/invoke.texi: Correct -Wformat-overflow code sample.
14255 2018-02-14 Martin Sebor <msebor@redhat.com>
14257 PR tree-optimization/83698
14258 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
14259 arrays constrain the offset range to their bounds.
14260 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
14261 (builtin_access::overlap): Avoid setting the size of overlap if it's
14263 (maybe_diag_overlap): Also consider arrays when deciding what values
14264 of offsets to include in diagnostics.
14266 2018-02-14 Martin Sebor <msebor@redhat.com>
14269 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
14270 that correspond to the kind of a declaration.
14272 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
14275 * config/pa/pa.md: Load address of PIC label using the linkage table
14276 if the label is nonlocal.
14278 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
14280 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
14281 warning message if user requests -maltivec=be.
14282 * doc/invoke.texi: Document deprecation of -maltivec=be.
14284 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
14287 * config/rs6000/rs6000-c.c: Update definitions for
14288 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
14289 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
14291 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
14294 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
14295 add _get_ssp intrinsics. Remove argument from
14296 __builtin_ia32_rdssp[d|q].
14297 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
14298 * config/i386/i386-builtin.def: Remove argument from
14299 __builtin_ia32_rdssp[d|q].
14300 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
14301 ix86_expand_special_args_builtin for _rdssp[d|q].
14302 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
14303 Clear register before usage.
14304 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
14305 Add documentation for new _get_ssp and _inc_ssp intrinsics.
14307 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
14309 PR tree-optimization/84357
14310 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
14311 operand 1 of an ARRAY_REF too.
14313 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
14316 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
14317 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
14319 (set_of_reg): New struct.
14320 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
14321 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
14322 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
14324 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
14325 Split into bitclr, bitset, bitinvert patterns if appropriate.
14326 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
14327 use rx_fuse_in_memory_bitop.
14328 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
14329 to named insn, correct maximum insn length.
14331 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
14334 * machmode.def: Define a complex mode for PARTIAL_INT.
14335 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
14337 * doc/rtl.texi: Document CSPImode.
14338 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
14340 (msp430_hard_regno_nregs_with_padding): Likewise.
14342 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
14345 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
14347 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
14349 PR rtl-optimization/84169
14350 * combine.c (try_combine): New variable split_i2i3. Set it to true if
14351 we generated a parallel as new i3 and we split that to new i2 and i3
14352 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
14353 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
14354 those to i2, not i1. Partially rewrite this scan code.
14356 2018-02-13 Jakub Jelinek <jakub@redhat.com>
14359 * stor-layout.c (place_field): For variable length fields, adjust
14360 offset_align afterwards not just based on the field's alignment,
14361 but also on the size.
14363 PR middle-end/84309
14364 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
14365 of exps and logs in the use_exp2 case.
14367 2018-02-13 Jeff Law <law@redhat.com>
14369 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
14370 entry for "vector".
14372 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
14375 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
14379 * common.opt (gas-loc-support, gas-locview-support): New.
14380 (ginline-points, ginternal-reset-location-views): New.
14381 * doc/invoke.texi: Document them. Use @itemx where intended.
14382 (gvariable-location-views): Adjust.
14383 * target.def (reset_location_view): New.
14384 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
14385 (TARGET_RESET_LOCATION_VIEW): New.
14386 * doc/tm.texi: Rebuilt.
14387 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
14388 (dwarf2out_default_as_locview_support): New.
14389 (output_asm_line_debug_info): Use option variables.
14390 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
14391 (output_loc_list): Likewise.
14392 (add_high_low_attributes): Check option variables.
14393 Don't output entry view attribute in strict mode.
14394 (gen_inlined_subroutine_die): Check option variables.
14395 (dwarf2out_inline_entry): Likewise.
14396 (init_sections_and_labels): Likewise.
14397 (dwarf2out_early_finish): Likewise.
14398 (maybe_reset_location_view): New, from...
14399 (dwarf2out_var_location): ... here. Call it.
14400 * debug.h (dwarf2out_default_as_loc_support): Declare.
14401 (dwarf2out_default_as_locview_support): Declare.
14402 * hooks.c (hook_int_rtx_insn_0): New.
14403 * hooks.h (hook_int_rtx_insn_0): Declare.
14404 * toplev.c (process_options): Take -gas-loc-support and
14405 -gas-locview-support from dwarf2out. Enable
14406 -gvariable-location-views by default only with locview
14407 assembler support. Enable -ginternal-reset-location-views by
14408 default only if the target defines the corresponding hook.
14409 Enable -ginline-points by default if location views are
14410 enabled; force it disabled if statement frontiers are
14412 * tree-inline.c (expand_call_inline): Check option variables.
14413 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
14415 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
14417 PR tree-optimization/84321
14418 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
14419 handling. Also check whether the anti-range contains any values
14420 that satisfy the mask; switch to a VR_RANGE if not.
14422 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
14425 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
14427 2018-02-13 Martin Jambor <mjambor@suse.cz>
14430 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
14431 of call statements, also set location of a load to a temporary.
14433 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
14435 * config/rl78/rl78.c (add_vector_labels): New function.
14436 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
14437 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
14438 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
14439 which checks that no arguments are passed.
14440 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
14441 * doc/extend.texi: Documentation for the new attribute.
14443 2018-02-13 Andreas Schwab <schwab@suse.de>
14445 * config/riscv/linux.h (CPP_SPEC): Define.
14447 2018-02-13 Jakub Jelinek <jakub@redhat.com>
14450 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
14451 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
14452 OPTION_MASK_ISA_AES as first argument to def_builtin_const
14453 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
14454 instead of OPTION_MASK_ISA_PCLMUL as first argument to
14455 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
14456 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
14457 temporarily for AES and PCLMUL builtins.
14459 PR tree-optimization/84339
14460 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
14461 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
14464 PR middle-end/84309
14465 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
14466 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
14467 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
14469 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
14471 * omp-simd-clone.h: New file.
14472 * omp-simd-clone.c: Include omp-simd-clone.h.
14473 (expand_simd_clones): No longer static.
14474 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
14475 cgraph.h and omp-simd-clone.h.
14476 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
14477 (vect_recog_widen_shift_pattern): Formatting fix.
14478 (vect_pattern_recog_1): Don't check optab for calls.
14481 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
14482 operands[2] into a REG before using gen_lowpart on it.
14484 2018-02-12 Jeff Law <law@redhat.com>
14487 * config/sh/sh.c (find_barrier): Consider a sibling call
14490 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
14491 successfully back substituting a reg.
14493 2018-02-12 Richard Biener <rguenther@suse.de>
14495 PR tree-optimization/84037
14496 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
14497 parameter, move visited init to caller.
14498 (vect_slp_analyze_operations): Separate cost from validity
14499 check, initialize visited once for all instances.
14500 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
14502 * tree-vect-stmts.c (vect_model_simple_cost): Make early
14504 (vect_model_promotion_demotion_cost): Likewise.
14505 (vectorizable_bswap): Guard cost modeling with !slp_node
14506 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
14508 (vectorizable_call): Likewise.
14509 (vectorizable_conversion): Likewise.
14510 (vectorizable_assignment): Likewise.
14511 (vectorizable_shift): Likewise.
14512 (vectorizable_operation): Likewise.
14513 (vectorizable_store): Likewise.
14514 (vectorizable_load): Likewise.
14515 (vectorizable_condition): Likewise.
14516 (vectorizable_comparison): Likewise.
14518 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
14521 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
14522 (ASAN_MARK): Fix fnspec to account for return value, change pointer
14523 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
14525 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
14527 PR middle-end/83665
14528 * params.def (inline-min-speedup): Increase from 8 to 15.
14529 (max-inline-insns-auto): Decrease from 40 to 30.
14530 * ipa-split.c (consider_split): Add some buffer for function to
14531 be considered inlining candidate.
14532 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
14535 2018-02-12 Richard Biener <rguenther@suse.de>
14537 PR tree-optimization/84037
14538 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
14539 matched stmts if we cannot swap the non-matched ones.
14541 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
14543 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
14544 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
14545 _mm_maskz_scalef_round_ss): New intrinsics.
14546 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
14547 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
14548 __builtin_ia32_scalefss_round): Remove.
14549 (__builtin_ia32_scalefsd_mask_round,
14550 __builtin_ia32_scalefss_mask_round): New intrinsics.
14551 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
14552 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
14553 ((match_operand:VF_128 2 "<round_nimm_predicate>"
14554 "<round_constraint>")): Changed to ...
14555 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
14556 "<round_scalar_constraint>")): ... this.
14557 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
14558 %0, %1, %2<round_op3>}"): Changed to ...
14559 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
14560 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
14561 %2<round_scalar_mask_op3>}"): ... this.
14562 * config/i386/subst.md (round_scalar_nimm_predicate): New.
14564 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
14566 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
14567 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
14568 (_mm_maskz_sqrt_round_ss): New intrinsics.
14569 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
14570 (__builtin_ia32_sqrtsd_mask_round)
14571 (__builtin_ia32_sqrtss_mask_round): New builtins.
14572 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
14573 (__builtin_ia32_sqrtss_round): Remove.
14574 (__builtin_ia32_sqrtsd_mask_round)
14575 (__builtin_ia32_sqrtss_mask_round): New builtins.
14576 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
14577 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
14578 ((match_operand:VF_128 1 "vector_operand"
14579 "xBm,<round_constraint>")): Changed to ...
14580 ((match_operand:VF_128 1 "vector_operand"
14581 "xBm,<round_scalar_constraint>")): ... this.
14582 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
14583 %0, %2, %<iptr>1<round_op3>}): Changed to ...
14584 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
14585 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
14586 %<iptr>1<round_scalar_mask_op3>}): ... this.
14587 ((set_attr "prefix" "<round_prefix>")): Changed to ...
14588 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
14590 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
14593 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
14594 Cast vec_cmpeq result to correct type.
14595 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
14596 Cast vec_cmpgt result to correct type.
14598 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
14600 * final.c (final_scan_insn_1): Renamed from...
14601 (final_scan_insn): ... this. New wrapper, to recover
14602 seen from the outermost call in recursive ones.
14603 * config/sparc/sparc.c (output_return): Drop seen from call.
14604 (output_sibcall): Likewise.
14605 * config/visium/visium.c (output_branch): Likewise.
14607 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
14609 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
14612 2018-02-10 Alan Modra <amodra@gmail.com>
14615 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
14616 Specify LR as an input.
14618 2018-02-10 Jakub Jelinek <jakub@redhat.com>
14621 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
14622 remove_member_access_dummy_vars): New functions.
14623 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
14624 lower_omp_1, execute_lower_omp): Use them.
14626 PR rtl-optimization/84308
14627 * shrink-wrap.c (spread_components): Release todo vector.
14629 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
14631 PR rtl-optimization/57193
14632 * ira-color.c (struct allocno_color_data): Add member
14633 conflict_allocno_hard_prefs.
14634 (update_conflict_allocno_hard_prefs): New.
14635 (bucket_allocno_compare_func): Add a preference based on
14636 conflict_allocno_hard_prefs.
14637 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
14638 (color_allocnos): Remove a dead code. Initiate
14639 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
14641 2018-02-09 Jakub Jelinek <jakub@redhat.com>
14644 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
14645 constraint from =wa to wa. Avoid a subreg on the output operand,
14646 instead use a pseudo and subreg it in a move.
14647 (p9_xxbrd_<mode>): Changed to ...
14648 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
14649 (p9_xxbrd_v2df): New expander.
14650 (p9_xxbrw_<mode>): Changed to ...
14651 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
14652 (p9_xxbrw_v4sf): New expander.
14654 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
14656 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
14658 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
14661 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
14662 multiply in 32-bit mode.
14663 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
14664 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
14667 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
14669 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
14670 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
14671 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
14672 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
14674 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
14677 * dwarf2out.c (is_trivial_indirect_ref): New function.
14678 (dwarf2out_late_global_decl): Do not generate a location
14679 attribute for variables that have a non-trivial DECL_VALUE_EXPR
14680 and that are not defined in the current unit.
14682 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
14684 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
14685 instead of a libcall for UNORDERED.
14687 2018-02-09 Tamar Christina <tamar.christina@arm.com>
14690 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
14691 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
14693 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14696 * config/s390/s390.c (s390_set_current_function): Invoke
14697 s390_indirect_branch_settings also if fndecl didn't change.
14699 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
14701 * config/rs6000/rs6000.md (blockage): Set length to zero.
14703 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
14705 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
14707 2018-02-09 Jakub Jelinek <jakub@redhat.com>
14710 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
14711 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
14715 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
14716 PARALLEL incoming that failed vt_get_decl_and_offset check.
14718 PR middle-end/84237
14719 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
14720 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
14722 (get_variable_section): For decls in named .bss* sections pass true as
14723 second argument to bss_initializer_p.
14725 2018-02-09 Marek Polacek <polacek@redhat.com>
14726 Jakub Jelinek <jakub@redhat.com>
14729 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
14730 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
14731 Sync some changes from cxx_fold_indirect_ref.
14733 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
14735 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
14737 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
14738 (BLOCK_INLINE_ENTRY_LABEL): New.
14739 (dwarf2out_var_location): Disregard inline entry markers.
14740 (inline_entry_data): New struct.
14741 (inline_entry_data_hasher): New hashtable type.
14742 (inline_entry_data_hasher::hash): New.
14743 (inline_entry_data_hasher::equal): New.
14744 (inline_entry_data_table): New variable.
14745 (add_high_low_attributes): Add DW_AT_entry_pc and
14746 DW_AT_GNU_entry_view attributes if a pending entry is found
14747 in inline_entry_data_table. Add old entry_pc attribute only
14748 if debug nonbinding markers are disabled.
14749 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
14750 markers are enabled.
14751 (block_within_block_p, dwarf2out_inline_entry): New.
14752 (dwarf2out_finish): Check that no entries remained in
14753 inline_entry_data_table.
14754 * final.c (reemit_insn_block_notes): Handle inline entry notes.
14755 (final_scan_insn, notice_source_line): Likewise.
14756 (rest_of_clean_state): Skip inline entry markers.
14757 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
14759 * gimple.c (gimple_build_debug_inline_entry): New.
14760 * gimple.h (enum gimple_debug_subcode): Add
14761 GIMPLE_DEBUG_INLINE_ENTRY.
14762 (gimple_build_debug_inline_entry): Declare.
14763 (gimple_debug_inline_entry_p): New.
14764 (gimple_debug_nonbind_marker_p): Adjust.
14765 * insn-notes.def (INLINE_ENTRY): New.
14766 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
14767 inline entry marker notes.
14768 (print_insn): Likewise.
14769 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
14770 (INSN_DEBUG_MARKER_KIND): Likewise.
14771 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
14772 * tree-inline.c (expand_call_inline): Build and insert
14773 debug_inline_entry stmt.
14774 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
14775 inline entry blocks early, if nonbind markers are enabled.
14776 (dump_scope_block): Dump fragment info.
14777 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
14778 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
14779 (gimple_build_debug_inline_entry): New.
14780 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
14781 Enable/disable inline entry points too.
14782 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
14783 (DEBUG_INSN): Describe inline entry markers.
14785 * common.opt (gvariable-location-views): New.
14786 (gvariable-location-views=incompat5): New.
14787 * config.in: Rebuilt.
14788 * configure: Rebuilt.
14789 * configure.ac: Test assembler for view support.
14790 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
14791 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
14792 * dwarf2out.c (var_loc_view): New typedef.
14793 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
14794 (dwarf2out_locviews_in_attribute): New.
14795 (dwarf2out_locviews_in_loclist): New.
14796 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
14797 (enum dw_line_info_opcode): Add LI_adv_address.
14798 (struct dw_line_info_table): Add view.
14799 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
14800 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
14801 (zero_view_p): New variable.
14802 (ZERO_VIEW_P): New macro.
14803 (output_asm_line_debug_info): New.
14804 (struct var_loc_node): Add view.
14805 (add_AT_view_list, AT_loc_list): New.
14806 (add_var_loc_to_decl): Add view param. Test it against last.
14807 (new_loc_list): Add view params. Record them.
14808 (AT_loc_list_ptr): Handle loc and view lists.
14809 (view_list_to_loc_list_val_node): New.
14810 (print_dw_val): Handle dw_val_class_view_list.
14811 (size_of_die): Likewise.
14812 (value_format): Likewise.
14813 (loc_list_has_views): New.
14814 (gen_llsym): Set vl_symbol too.
14815 (maybe_gen_llsym, skip_loc_list_entry): New.
14816 (dwarf2out_maybe_output_loclist_view_pair): New.
14817 (output_loc_list): Output view list or entries too.
14818 (output_view_list_offset): New.
14819 (output_die): Handle dw_val_class_view_list.
14820 (output_dwarf_version): New.
14821 (output_compilation_unit_header): Use it.
14822 (output_skeleton_debug_sections): Likewise.
14823 (output_rnglists, output_line_info): Likewise.
14824 (output_pubnames, output_aranges): Update version comments.
14825 (output_one_line_info_table): Output view numbers in asm comments.
14826 (dw_loc_list): Determine current endview, pass it to new_loc_list.
14827 Call maybe_gen_llsym.
14828 (loc_list_from_tree_1): Adjust.
14829 (add_AT_location_description): Create view list attribute if
14830 needed, check it's absent otherwise.
14831 (convert_cfa_to_fb_loc_list): Adjust.
14832 (maybe_emit_file): Call output_asm_line_debug_info for test.
14833 (dwarf2out_var_location): Reset views as needed. Precompute
14834 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
14835 attribute. Set view.
14836 (new_line_info_table): Reset next view.
14837 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
14838 (dwarf2out_source_line): Likewise. Output view resets and labels to
14839 the assembler, or select appropriate line info opcodes.
14840 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
14841 (optimize_string_length): Catch it. Adjust.
14842 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
14843 dw_val_class_view_list, and remove it if no longer needed.
14844 (hash_loc_list): Hash view numbers.
14845 (loc_list_hasher::equal): Compare them.
14846 (optimize_location_lists): Check whether a view list symbol is
14847 needed, and whether the locview attribute is present, and
14848 whether they match. Remove the locview attribute if no longer
14850 (index_location_lists): Call skip_loc_list_entry for test.
14851 (dwarf2out_finish): Call output_asm_line_debug_info for test.
14852 Use output_dwarf_version.
14853 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
14854 (struct dw_val_node): Add val_view_list.
14855 * final.c (SEEN_NEXT_VIEW): New.
14856 (set_next_view_needed): New.
14857 (clear_next_view_needed): New.
14858 (maybe_output_next_view): New.
14859 (final_start_function): Rename to...
14860 (final_start_function_1): ... this. Take pointer to FIRST,
14861 add SEEN parameter. Emit param bindings in the initial view.
14862 (final_start_function): Reintroduce SEEN-less interface.
14863 (final): Rename to...
14864 (final_1): ... this. Take SEEN parameter. Output final pending
14865 next view at the end.
14866 (final): Reintroduce seen-less interface.
14867 (final_scan_insn): Output pending next view before switching
14868 sections or ending a block. Mark the next view as needed when
14869 outputting variable locations. Notify debug backend of section
14870 changes, and of location view changes.
14871 (rest_of_handle_final): Adjust.
14872 * toplev.c (process_options): Autodetect value for debug variable
14873 location views option. Warn on incompat5 without -gdwarf-5.
14874 * doc/invoke.texi (gvariable-location-views): New.
14875 (gvariable-location-views=incompat5): New.
14876 (gno-variable-location-views): New.
14878 2018-02-08 David Malcolm <dmalcolm@redhat.com>
14880 PR tree-optimization/84136
14881 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
14882 that the result of find_edge is non-NULL.
14884 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
14887 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
14888 storing integer register in SImode. Fix cost of 256 and 512
14889 byte aligned SSE register store.
14891 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
14893 * config/i386/i386.c (ix86_multiplication_cost): Fix
14894 multiplication cost for TARGET_AVX512DQ.
14896 2018-02-08 Marek Polacek <polacek@redhat.com>
14898 PR tree-optimization/84238
14899 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
14902 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
14904 PR tree-optimization/84265
14905 * tree-vect-stmts.c (vectorizable_store): Don't treat
14906 VMAT_CONTIGUOUS accesses as grouped.
14907 (vectorizable_load): Likewise.
14909 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
14911 PR tree-optimization/81635
14912 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
14913 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
14914 (test_round_for_mask): New functions.
14915 (wide_int_cc_tests): Call test_round_for_mask.
14916 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
14917 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
14918 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
14919 range returned by get_range_info.
14921 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
14924 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
14925 * symtab.c: Include builtins.h
14926 (symtab_node::output_to_lto_symbol_table_p): Move here
14927 from lto-streamer-out.c:output_symbol_p.
14928 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
14929 (output_symbol_p): Move all logic to symtab.c
14930 (produce_symtab): Update.
14932 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14934 * config/s390/s390-opts.h (enum indirect_branch): Define.
14935 * config/s390/s390-protos.h (s390_return_addr_from_memory)
14936 (s390_indirect_branch_via_thunk)
14937 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
14938 (enum s390_indirect_branch_type): Define.
14939 * config/s390/s390.c (struct s390_frame_layout, struct
14940 machine_function): Remove.
14941 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
14942 (indirect_branch_table_label_no, indirect_branch_table_name):
14944 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
14945 (enum s390_indirect_branch_option): Define.
14946 (s390_return_addr_from_memory): New function.
14947 (s390_handle_string_attribute): New function.
14948 (s390_attribute_table): Add new attribute handler.
14949 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
14950 (s390_indirect_branch_via_thunk): New function.
14951 (s390_indirect_branch_via_inline_thunk): New function.
14952 (s390_function_ok_for_sibcall): When jumping via thunk disallow
14953 sibling call optimization for non z10 compiles.
14954 (s390_emit_call): Force indirect branch target to be a single
14955 register. Add r1 clobber for non-z10 compiles.
14956 (s390_emit_epilogue): Emit return jump via return_use expander.
14957 (s390_reorg): Handle JUMP_INSNs as execute targets.
14958 (s390_option_override_internal): Perform validity checks for the
14959 new command line options.
14960 (s390_indirect_branch_attrvalue): New function.
14961 (s390_indirect_branch_settings): New function.
14962 (s390_set_current_function): Invoke s390_indirect_branch_settings.
14963 (s390_output_indirect_thunk_function): New function.
14964 (s390_code_end): Implement target hook.
14965 (s390_case_values_threshold): Implement target hook.
14966 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
14968 * config/s390/s390.h (struct s390_frame_layout)
14969 (struct machine_function): Move here from s390.c.
14970 (TARGET_INDIRECT_BRANCH_NOBP_RET)
14971 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
14972 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
14973 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
14974 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
14975 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
14976 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
14977 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
14978 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
14979 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
14980 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
14981 (mnemonic attribute): Add values which aren't recognized
14983 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
14984 pattern for branch conversion. Fix mnemonic attribute.
14985 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
14986 indirect branch via thunk if requested.
14987 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
14988 ("*indirect_jump"): Disable for branch conversion using out of
14990 ("indirect_jump_via_thunk<mode>_z10")
14991 ("indirect_jump_via_thunk<mode>")
14992 ("indirect_jump_via_inlinethunk<mode>_z10")
14993 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
14994 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
14995 ("casesi_jump_via_inlinethunk<mode>_z10")
14996 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
14997 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
14998 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
14999 ("*indirect2_jump"): Disable for branch conversion.
15000 ("casesi_jump"): Turn into expander and expand patterns for branch
15002 ("return_use"): New expander.
15003 ("*return"): Emit return via thunk and rename it to ...
15004 ("*return<mode>"): ... this one.
15005 * config/s390/s390.opt: Add new options and and enum for the
15008 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
15010 * lra-constraints.c (match_reload): Unconditionally use
15011 gen_lowpart_SUBREG, rather than selecting between that
15012 and equivalent gen_rtx_SUBREG code.
15014 2018-02-08 Richard Biener <rguenther@suse.de>
15016 PR tree-optimization/84233
15017 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
15018 changed flag instead of boguously re-using phi_inserted.
15020 2018-02-08 Martin Jambor <mjambor@suse.cz>
15022 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
15023 static local variables.
15025 2018-02-08 Richard Biener <rguenther@suse.de>
15027 PR tree-optimization/84278
15028 * tree-vect-stmts.c (vectorizable_store): When looking for
15029 smaller vector types to perform grouped strided loads/stores
15030 make sure the mode is supported by the target.
15031 (vectorizable_load): Likewise.
15033 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
15035 * config/aarch64/aarch64.c (aarch64_components_for_bb):
15036 Increase LDP/STP opportunities by adding adjacent callee-saves.
15038 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
15040 PR rtl-optimization/84068
15041 PR rtl-optimization/83459
15042 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
15044 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
15046 PR tree-optimization/84224
15047 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
15048 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
15049 non-zero arguments.
15051 2018-02-07 Iain Sandoe <iain@codesourcery.com>
15054 * config/rs6000/altivec.md (*restore_world): Remove LR use.
15055 * config/rs6000/predicates.md (restore_world_operation): Adjust op
15056 count, remove one USE.
15058 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
15060 * doc/install.texi (Configuration): Document the
15061 --with-long-double-format={ibm,ieee} PowerPC configuration
15065 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
15066 Convert from define_expand to be define_insn_and_split. Rework
15067 float/double/_Float128 conversions to QI/HI/SImode to work with
15068 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
15069 conversions to QI/HImode types did a store and then a load to
15070 truncate the value. For conversions to VSX registers, don't split
15071 the insn, instead emit the code directly. Use the code iterator
15072 any_fix to combine signed and unsigned conversions.
15073 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
15074 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
15075 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
15076 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
15077 (fix_<mode>di2_hw): Likewise.
15078 (fixuns_<mode>di2_hw): Likewise.
15079 (fix_<mode>si2_hw): Likewise.
15080 (fixuns_<mode>si2_hw): Likewise.
15081 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
15082 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
15083 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
15084 fix<uns>_trunc<SFDF:mode>si2_p8.
15085 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
15086 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
15087 (fix<uns>_<mode>_mem): Likewise.
15088 (fctiw<u>z_<mode>_mem): Likewise.
15089 (fix<uns>_<mode>_mem): Likewise.
15090 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
15091 the register allocator from doing a direct move to the GPRs to do
15092 a store, and instead use the ISA 3.0 store byte/half-word from
15093 vector register instruction. For IEEE 128-bit floating point,
15094 also optimize stores of 32-bit ints.
15095 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
15097 2018-02-07 Alan Hayward <alan.hayward@arm.com>
15099 * genextract.c (push_pathstr_operand): New function to support
15101 (walk_rtx): Call push_pathstr_operand.
15102 (print_path): Support [a-zA-Z].
15104 2018-02-07 Richard Biener <rguenther@suse.de>
15106 PR tree-optimization/84037
15107 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
15108 (cse_and_gimplify_to_preheader): Declare.
15109 (vect_get_place_in_interleaving_chain): Likewise.
15110 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15112 (_loop_vec_info::~_loop_vec_info): Delete it.
15113 (cse_and_gimplify_to_preheader): New function.
15114 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
15115 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
15116 (vectorizable_load): Likewise. For grouped stores always base
15117 the IV on the first element.
15118 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
15119 condition before gimplifying.
15121 2018-02-07 Jakub Jelinek <jakub@redhat.com>
15123 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
15124 *DIV_EXPR and *MOD_EXPR.
15126 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
15129 * config/i386/i386.c (ix86_option_override_internal): Mask out
15130 the CF_SET bit when checking -fcf-protection.
15132 2018-02-07 Tom de Vries <tom@codesourcery.com>
15135 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
15138 2018-02-07 Richard Biener <rguenther@suse.de>
15140 PR tree-optimization/84204
15141 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
15144 PR tree-optimization/84205
15145 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
15146 special-case isl_ast_op_zdiv_r.
15148 PR tree-optimization/84223
15149 * graphite-scop-detection.c (gather_bbs::before_dom_children):
15150 Only add conditions from within the region.
15151 (gather_bbs::after_dom_children): Adjust.
15153 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
15156 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
15157 * config/avr/avr.md: Only post-reload split REG-REG moves if
15158 either register is GENERAL_REG_P.
15160 2018-02-07 Jakub Jelinek <jakub@redhat.com>
15162 PR tree-optimization/84235
15163 * tree-ssa-scopedtables.c
15164 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
15165 if the subtraction is performed in floating point type where NaNs are
15166 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
15167 build 1. Formatting fix.
15169 2018-02-06 Jakub Jelinek <jakub@redhat.com>
15172 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
15173 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
15174 and skip it regardless of bb boundaries. Use CALL_P macro,
15175 don't test INSN_P (insn) together with CALL_P or JUMP_P check
15176 unnecessarily, formatting fix.
15178 2018-02-06 Michael Collison <michael.collison@arm.com>
15180 * config/arm/thumb2.md:
15181 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
15182 (*thumb_mov_notscc): Ditto.
15184 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
15187 * config/rs6000/rs6000.md (su code attribute): Use "u" for
15188 unsigned_fix, not "s".
15190 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15192 * configure.ac (gcc_fn_eh_frame_ro): New function.
15193 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
15194 correct .eh_frame permissions.
15195 * configure: Regenerate.
15197 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
15199 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
15200 irrelevant options.
15202 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15204 * config/rs6000/rs6000.c (rs6000_option_override_internal):
15205 Display warning message for -mno-speculate-indirect-jumps.
15207 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
15209 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
15211 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
15213 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
15215 PR tree-optimization/84225
15216 * tree-eh.c (find_trapping_overflow): Only call
15217 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
15219 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
15222 * config/i386/i386.c: Reimplement the check of possible options
15223 -mibt/-mshstk conbination. Change error messages.
15224 * doc/invoke.texi: Fix a typo: remove extra '='.
15226 2018-02-06 Marek Polacek <polacek@redhat.com>
15228 PR tree-optimization/84228
15229 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
15231 2018-02-06 Tamar Christina <tamar.christina@arm.com>
15234 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
15235 emitted arch directives.
15236 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
15237 __ARM_FEATURE_COPROC before changing architectures.
15239 2018-02-06 Richard Biener <rguenther@suse.de>
15241 * config/i386/i386.c (print_reg): Fix typo.
15242 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
15244 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
15246 * configure: Regenerate.
15248 2018-02-05 Martin Sebor <msebor@redhat.com>
15250 PR tree-optimization/83369
15251 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
15254 2018-02-05 Martin Liska <mliska@suse.cz>
15256 * doc/invoke.texi: Cherry-pick upstream r323995.
15258 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
15260 * ira.c (ira_init_register_move_cost): Adjust comment.
15262 2018-02-05 Martin Liska <mliska@suse.cz>
15264 PR gcov-profile/84137
15265 * doc/gcov.texi: Fix typo in documentation.
15267 2018-02-05 Martin Liska <mliska@suse.cz>
15269 PR gcov-profile/83879
15270 * doc/gcov.texi: Document necessity of --dynamic-list-data when
15271 using dlopen functionality.
15273 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
15275 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
15276 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
15277 _mm_maskz_range_ss, _mm_mask_range_round_ss,
15278 _mm_maskz_range_round_ss): New intrinsics.
15279 (__builtin_ia32_rangesd128_round)
15280 (__builtin_ia32_rangess128_round): Remove.
15281 (__builtin_ia32_rangesd128_mask_round,
15282 __builtin_ia32_rangess128_mask_round): New builtins.
15283 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
15284 __builtin_ia32_rangess128_round): Remove.
15285 (__builtin_ia32_rangesd128_mask_round,
15286 __builtin_ia32_rangess128_mask_round): New builtins.
15287 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
15288 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
15289 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
15290 "<round_saeonly_constraint>")): Changed to ...
15291 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
15292 "<round_saeonly_scalar_constraint>")): ... this.
15293 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
15294 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
15295 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
15296 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
15297 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
15299 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
15301 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
15303 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
15304 Remove all values except native, 8540 and 8548.
15306 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
15308 * config/i386/i386.c (ix86_output_function_return): Pass
15309 INVALID_REGNUM, instead of -1, as invalid register number to
15310 indirect_thunk_name and output_indirect_thunk.
15312 2018-02-02 Julia Koval <julia.koval@intel.com>
15314 * config.gcc: Add -march=icelake.
15315 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
15316 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
15317 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
15318 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
15319 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
15320 (processor_target_table): Add icelake.
15321 (ix86_option_override_internal): Handle new PTAs.
15322 (get_builtin_code_for_version): Handle icelake.
15323 (M_INTEL_COREI7_ICELAKE): New.
15324 (fold_builtin_cpu): Handle icelake.
15325 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
15326 * doc/invoke.texi: Add -march=icelake.
15328 2018-02-02 Julia Koval <julia.koval@intel.com>
15330 * config/i386/i386.c (ix86_option_override_internal): Change flags type
15331 to wide_int_bitmask.
15332 * wide-int-bitmask.h: New.
15334 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
15337 * config/i386/i386.md: Replace Pmode with word_mode in
15338 builtin_setjmp_setup and builtin_longjmp to support x32.
15340 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
15344 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
15346 (rs6000_supported_cpu_names): New static variable.
15347 (linux_cpu_translation_table): Likewise.
15348 (elf_platform) <cpu>: Define new static variable and use it.
15349 Translate kernel AT_PLATFORM name to canonical name if needed.
15350 Error if platform name is unknown.
15352 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
15355 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
15357 2018-02-01 Jeff Law <law@redhat.com>
15360 * config/i386/i386.c (release_scratch_register_on_entry): Add new
15361 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
15362 the scratch if RELEASE_VIA_POP is false.
15363 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
15364 If we have to save a temporary register, decrement SIZE appropriately.
15365 Pass new arguments to release_scratch_register_on_entry.
15366 (ix86_adjust_stack_and_probe): Likewise.
15367 (ix86_emit_probe_stack_range): Pass new arguments to
15368 release_scratch_register_on_entry.
15370 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
15372 PR rtl-optimization/84157
15373 * combine.c (change_zero_ext): Use REG_P predicate in
15374 front of HARD_REGISTER_P predicate.
15376 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
15378 * config/avr/avr.c (avr_option_override): Move disabling of
15379 -fdelete-null-pointer-checks to...
15380 * common/config/avr/avr-common.c (avr_option_optimization_table):
15383 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15385 PR tree-optimization/81635
15386 * tree-data-ref.c (split_constant_offset_1): For types that
15387 wrap on overflow, try to use range info to prove that wrapping
15390 2018-02-01 Renlin Li <renlin.li@arm.com>
15393 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
15394 TAILCALL_ADDR_REGS.
15395 (aarch64_register_move_cost): Likewise.
15396 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
15397 TAILCALL_ADDR_REGS.
15398 (REG_CLASS_NAMES): Likewise.
15399 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
15400 TAILCALL_ADDR_REGS. Remove IP registers.
15401 * config/aarch64/aarch64.md (Ucs): Update register constraint.
15403 2018-02-01 Richard Biener <rguenther@suse.de>
15405 * domwalk.h (dom_walker::dom_walker): Add additional constructor
15406 for specifying RPO order and allow NULL for that.
15407 * domwalk.c (dom_walker::dom_walker): Likewise.
15408 (dom_walker::walk): Handle NULL RPO order.
15409 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
15411 (rewrite_update_dom_walker): Likewise.
15412 (mark_def_dom_walker): Likewise.
15414 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15416 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
15417 (aarch64_maybe_expand_sve_subreg_move): Declare.
15418 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
15419 * config/aarch64/predicates.md (aarch64_any_register_operand): New
15421 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
15422 that are semantically a reverse operation.
15423 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
15424 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
15425 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
15427 (aarch64_can_change_mode_class): For big-endian, forbid changes
15428 between two SVE modes if they have different element sizes.
15430 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15432 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
15433 the TImode handling for big-endian targets.
15435 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15437 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
15438 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
15440 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
15441 Remove BSWAP handing for big-endian targets and use the form of
15442 LD1RQ appropariate for the mode.
15444 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15446 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
15447 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
15448 duplicated element.
15450 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15453 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
15454 check for operands that need to go through aarch64_sve_reload_be.
15456 2018-02-01 Jakub Jelinek <jakub@redhat.com>
15458 PR tree-optimization/81661
15459 PR tree-optimization/84117
15460 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
15461 * tree-eh.c: Include gimplify.h.
15462 (find_trapping_overflow, replace_trapping_overflow,
15463 rewrite_to_non_trapping_overflow): New functions.
15464 * tree-vect-loop.c: Include tree-eh.h.
15465 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
15466 * tree-data-ref.c: Include tree-eh.h.
15467 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
15469 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
15471 PR rtl-optimization/84123
15472 * combine.c (change_zero_ext): Check if hard register satisfies
15473 can_change_dest_mode before calling gen_lowpart_SUBREG.
15475 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
15478 * ira.c (ira_init_register_move_cost): Remove assert.
15480 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
15482 PR rtl-optimization/84071
15483 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
15484 * doc/tm.texi: Regenerate.
15486 2018-01-31 Richard Biener <rguenther@suse.de>
15488 PR tree-optimization/84132
15489 * tree-data-ref.c (analyze_miv_subscript): Properly
15490 check whether evolution_function_is_affine_multivariate_p
15491 before calling gcd_of_steps_may_divide_p.
15493 2018-01-31 Julia Koval <julia.koval@intel.com>
15496 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
15497 * config/i386/i386.md (rdpid_rex64) New.
15498 (rdpid): Make 32bit only.
15500 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
15503 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
15504 an IDENTIFIER_NODE for FUNCTION_TYPE's.
15506 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
15509 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
15511 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
15513 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
15515 PR rtl-optimization/84071
15516 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
15517 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
15519 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
15521 * config/arc/arc.c (arc_handle_aux_attribute): New function.
15522 (arc_attribute_table): Add 'aux' attribute.
15523 (arc_in_small_data_p): Consider aux like variables.
15524 (arc_is_aux_reg_p): New function.
15525 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
15526 (arc_get_aux_arg): New function.
15527 (prepare_move_operands): Handle aux-register access.
15528 (arc_handle_aux_attribute): New function.
15529 * doc/extend.texi (ARC Variable attributes): Add subsection.
15531 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
15533 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
15534 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
15535 (arc_attribute_table): Add 'uncached' attribute.
15536 (arc_print_operand): Print '.di' flag for uncached memory
15538 (arc_in_small_data_p): Do not consider for small data the uncached
15540 (arc_is_uncached_mem_p): New function.
15541 * config/arc/predicates.md (compact_store_memory_operand): Check
15542 for uncached memory accesses.
15543 (nonvol_nonimm_operand): Likewise.
15544 * doc/extend.texi (ARC Type Attribute): New subsection.
15546 2018-01-31 Jakub Jelinek <jakub@redhat.com>
15549 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
15550 falign-loops=): Add Optimization flag.
15552 2018-01-30 Jeff Law <law@redhat.com>
15555 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
15556 INT_REGISTERS_SAVED. Check it prior to calling
15557 get_scratch_register_on_entry.
15558 (ix86_adjust_stack_and_probe): Similarly.
15559 (ix86_emit_probe_stack_range): Similarly.
15560 (ix86_expand_prologue): Corresponding changes.
15562 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15565 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
15566 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
15568 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
15571 * lra-constraints.c (curr_insn_transform): Process AND in the
15574 2018-01-30 Jakub Jelinek <jakub@redhat.com>
15576 PR rtl-optimization/83986
15577 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
15578 dependence against last_pending_memory_flush in addition to
15579 pending_jump_insns.
15581 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
15583 PR tree-optimization/81611
15584 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
15587 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15590 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
15593 2018-01-30 Richard Biener <rguenther@suse.de>
15594 Jakub Jelinek <jakub@redhat.com>
15596 PR tree-optimization/84111
15597 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
15598 inner loops added during recursion, as they don't have up-to-date
15601 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
15604 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
15605 (can_inline_edge_by_limits_p): ... here.
15606 (can_early_inline_edge_p, check_callers,
15607 update_caller_keys, update_callee_keys, recursive_inlining,
15608 add_new_edges_to_heap, speculation_useful_p,
15609 inline_small_functions,
15610 inline_small_functions, flatten_function,
15611 inline_to_all_callers_1): Update.
15613 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
15615 * profile-count.c (profile_count::combine_with_ipa_count): Handle
15618 2018-01-30 Richard Biener <rguenther@suse.de>
15620 PR tree-optimization/83008
15621 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
15622 invariant and constant vector uses in stmts when they need
15623 more than one stmt.
15625 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15628 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
15629 * configure: Regenerate.
15631 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
15633 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
15635 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
15636 Use gen_rtx_REG rather than gen_lowpart.
15638 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
15640 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
15641 rather than 0 when creating partial subregs.
15643 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
15645 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
15648 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
15651 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
15652 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
15653 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
15654 flags. This restores the settings used before the 2017-07-24.
15655 Turning off pre increment/decrement/modify allows IVOPTS to
15656 optimize DF/SF loops where the index is an int.
15658 2018-01-29 Richard Biener <rguenther@suse.de>
15659 Kelvin Nilsen <kelvin@gcc.gnu.org>
15662 * tree-vect-stmts.c (vectorizable_call): Don't call
15663 targetm.vectorize_builtin_md_vectorized_function if callee is
15666 2018-01-22 Carl Love <cel@us.ibm.com>
15668 * doc/extend.tex: Fix typo in second arg in
15669 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
15671 2018-01-29 Richard Biener <rguenther@suse.de>
15673 PR tree-optimization/84086
15674 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
15675 (flush_ssaname_freelist): When SSA names were released reset
15676 the SCEV hash table.
15678 2018-01-29 Richard Biener <rguenther@suse.de>
15680 PR tree-optimization/84057
15681 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
15682 removed paths when removing edges.
15684 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
15686 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
15687 -mfunction-return=@var{choice}.
15689 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
15691 PR diagnostic/84034
15692 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
15693 Handle CR like TAB.
15694 (layout::print_source_line): Likewise.
15695 (test_get_line_width_without_trailing_whitespace): Add test cases.
15697 2018-01-27 Jakub Jelinek <jakub@redhat.com>
15699 PR middle-end/84040
15700 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
15703 2018-01-26 Jim Wilson <jimw@sifive.com>
15705 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
15707 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
15710 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15712 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
15713 and CMP + SUB-immediate -> SUBS.
15715 2018-01-26 Martin Sebor <msebor@redhat.com>
15717 PR tree-optimization/83896
15718 * tree-ssa-strlen.c (get_string_len): Rename...
15719 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
15720 Avoid assuming length is constant.
15721 (handle_char_store): Use HOST_WIDE_INT for string length.
15723 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
15726 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
15727 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
15729 2018-01-26 Richard Biener <rguenther@suse.de>
15731 PR rtl-optimization/84003
15732 * dse.c (record_store): Only record redundant stores when
15733 the earlier store aliases at least all accesses the later one does.
15735 2018-01-26 Jakub Jelinek <jakub@redhat.com>
15737 PR rtl-optimization/83985
15738 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
15739 REG_CFA_RESTORE insns.
15740 (delete_unmarked_insns): Don't ignore separate shrink wrapping
15741 REG_CFA_RESTORE insns here.
15744 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
15745 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
15747 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
15749 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
15750 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
15751 (arc_init): Likewise.
15752 (arc_override_options): Likewise.
15753 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
15755 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
15757 * config/arc/arc.h (TARGET_DBNZ): Define.
15758 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
15759 properly set the tune attribute.
15760 (dbnz): Use TARGET_DBNZ guard.
15761 * config/arc/arc.opt (mtune): Add core3 option.
15763 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
15765 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
15766 recognize new pic like addresses.
15767 (arc_delegitimize_address): Clean up.
15769 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
15771 * config/arc/arc-arches.def: Option mrf16 valid for all
15773 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
15774 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
15775 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
15776 * config/arc/arc-tables.opt: Regenerate.
15777 * config/arc/arc.c (arc_conditional_register_usage): Handle
15778 reduced register file case.
15779 (arc_file_start): Set must have build attributes.
15780 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
15781 mrf16 option value.
15782 * config/arc/arc.opt (mrf16): Add new option.
15783 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
15784 * config/arc/genmultilib.awk: Handle new mrf16 option.
15785 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
15786 * config/arc/t-multilib: Regenerate.
15787 * doc/invoke.texi (ARC Options): Document mrf16 option.
15789 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
15791 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
15792 * config/arc/arc.c (arc_handle_secure_attribute): New function.
15793 (arc_attribute_table): Add 'secure_call' attribute.
15794 (arc_print_operand): Print secure call operand.
15795 (arc_function_ok_for_sibcall): Don't optimize tail calls when
15797 (arc_is_secure_call_p): New function. * config/arc/arc.md
15798 (call_i): Add support for sjli instruction.
15799 (call_value_i): Likewise.
15800 * config/arc/constraints.md (Csc): New constraint.
15802 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
15803 John Eric Martin <John.Martin@emmicro-us.com>
15805 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
15806 * config/arc/arc.c (_arc_jli_section): New struct.
15807 (arc_jli_section): New type.
15808 (rc_jli_sections): New static variable.
15809 (arc_handle_jli_attribute): New function.
15810 (arc_attribute_table): Add jli_always and jli_fixed attribute.
15811 (arc_file_end): New function.
15812 (TARGET_ASM_FILE_END): Define.
15813 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
15814 (arc_add_jli_section): New function.
15815 (jli_call_scan): Likewise.
15816 (arc_reorg): Call jli_call_scan.
15817 (arc_output_addsi): Remove 'S' from printing asm operand.
15818 (arc_is_jli_call_p): New function.
15819 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
15821 (movhi_insn): Likewise.
15822 (movsi_insn): Likewise.
15823 (movsi_set_cc_insn): Likewise.
15824 (loadqi_update): Likewise.
15825 (load_zeroextendqisi_update): Likewise.
15826 (load_signextendqisi_update): Likewise.
15827 (loadhi_update): Likewise.
15828 (load_zeroextendhisi_update): Likewise.
15829 (load_signextendhisi_update): Likewise.
15830 (loadsi_update): Likewise.
15831 (loadsf_update): Likewise.
15832 (movsicc_insn): Likewise.
15833 (bset_insn): Likewise.
15834 (bxor_insn): Likewise.
15835 (bclr_insn): Likewise.
15836 (bmsk_insn): Likewise.
15837 (bicsi3_insn): Likewise.
15838 (cmpsi_cc_c_insn): Likewise.
15839 (movsi_ne): Likewise.
15840 (movsi_cond_exec): Likewise.
15841 (clrsbsi2): Likewise.
15842 (norm_f): Likewise.
15852 (call_i): Remove 'S' asm letter, add jli instruction.
15853 (call_value_i): Likewise.
15854 * config/arc/arc.op (mjli-always): New option.
15855 * config/arc/constraints.md (Cji): New constraint.
15856 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
15858 (subsf3_fpx): Likewise.
15859 (mulsf3_fpx): Likewise.
15860 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
15862 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
15863 function attrbutes.
15864 * doc/invoke.texi (ARC): Document mjli-always option.
15866 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
15868 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
15869 avoid addition with 0 and use incw and decw where possible.
15871 2018-01-26 Richard Biener <rguenther@suse.de>
15873 PR tree-optimization/81082
15874 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
15875 association if it requires casting to unsigned.
15876 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
15877 from fold_plusminus_mult_expr to catch important cases late when
15878 range info is available.
15880 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15882 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
15883 * configure.ac (hidden_linkonce): New test.
15884 * configure: Regenerate.
15885 * config.in: Regenerate.
15887 2018-01-26 Julia Koval <julia.koval@intel.com>
15889 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
15890 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
15891 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
15892 _mm_mask_bitshuffle_epi64_mask): Fix type.
15893 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
15894 USI_FTYPE_V4DI_V4DI_USI): Remove.
15895 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
15896 __builtin_ia32_vpshufbitqmb256_mask,
15897 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
15898 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
15899 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
15901 2018-01-26 Alan Modra <amodra@gmail.com>
15904 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
15905 UNSPEC_VBPERMQ. Sort other unspecs.
15907 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
15909 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
15911 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
15913 PR middle-end/83055
15914 * predict.c (drop_profile): Do not push/pop cfun; update also
15916 (handle_missing_profiles): Fix logic looking for zero profiles.
15918 2018-01-25 Jakub Jelinek <jakub@redhat.com>
15920 PR middle-end/83977
15921 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
15922 on functions with #pragma omp declare simd or functions with simd
15924 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
15925 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
15926 Remove trailing \n from warning_at calls.
15928 2018-01-25 Tom de Vries <tom@codesourcery.com>
15931 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
15932 for neutered workers.
15934 2018-01-24 Joseph Myers <joseph@codesourcery.com>
15937 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
15938 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
15940 2018-01-24 Jeff Law <law@redhat.com>
15943 * i386.c (get_probe_interval): Move to earlier point.
15944 (ix86_compute_frame_layout): If -fstack-clash-protection and
15945 the frame is larger than the probe interval, then use pushes
15946 to save registers rather than reg->mem moves.
15947 (ix86_expand_prologue): Remove conditional for int_registers_saved
15950 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
15953 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
15954 min/max for never referenced object.
15956 2018-01-24 Jakub Jelinek <jakub@redhat.com>
15958 PR middle-end/83977
15959 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
15961 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
15962 attributes from DECL_ATTRIBUTES (decl) without affecting
15963 DECL_ATTRIBUTES (current_function_decl).
15964 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
15965 functions with non-NULL DECL_ABSTRACT_ORIGIN.
15967 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
15969 PR tree-optimization/83979
15970 * fold-const.c (fold_comparison): Use constant_boolean_node
15971 instead of boolean_{true,false}_node.
15973 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
15975 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
15978 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15980 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
15981 Simplify the clause that sets the length attribute.
15982 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
15983 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
15984 clause that sets the length attribute.
15985 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
15987 2018-01-24 Tom de Vries <tom@codesourcery.com>
15990 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
15991 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
15992 Add strict parameter.
15993 (prevent_branch_around_nothing): Insert dummy insn between branch to
15994 label and label with no ptx insn inbetween.
15995 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
15997 2018-01-24 Tom de Vries <tom@codesourcery.com>
16000 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
16001 for neutered threads in warp.
16002 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
16004 2018-01-24 Richard Biener <rguenther@suse.de>
16006 PR tree-optimization/83176
16007 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
16010 2018-01-24 Richard Biener <rguenther@suse.de>
16012 PR tree-optimization/82819
16013 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
16014 code generating pluses that are no-ops in the target precision.
16016 2018-01-24 Richard Biener <rguenther@suse.de>
16018 PR middle-end/84000
16019 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
16021 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
16023 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
16024 to merge probabilities.
16025 * predict.c (probably_never_executed): Also mark as cold functions
16026 with global 0 profile and guessed local profile.
16027 * profile-count.c (profile_probability::combine_with_count): New
16029 * profile-count.h (profile_probability::operator*,
16030 profile_probability::operator*=, profile_probability::operator/,
16031 profile_probability::operator/=): Reduce precision to adjusted
16032 and set value to guessed on contradictory divisions.
16033 (profile_probability::combine_with_freq): Remove.
16034 (profile_probability::combine_wiht_count): Declare.
16035 (profile_count::force_nonzero):: Set to adjusted.
16036 (profile_count::probability_in):: Set quality to adjusted.
16037 * tree-ssa-tail-merge.c (replace_block_by): Use
16038 combine_with_count.
16040 2018-01-23 Andrew Waterman <andrew@sifive.com>
16041 Jim Wilson <jimw@sifive.com>
16043 * config/riscv/riscv.c (riscv_stack_boundary): New.
16044 (riscv_option_override): Set riscv_stack_boundary. Handle
16045 riscv_preferred_stack_boundary_arg.
16046 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
16047 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
16048 (STACK_BOUNDARY): Set to riscv_stack_boundary.
16049 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
16050 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
16051 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
16053 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
16056 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
16057 of struct ix86_frame.
16058 (ix86_expand_epilogue): Likewise. Add a local variable for
16059 the reg_save_offset field in struct ix86_frame.
16061 2018-01-23 Bin Cheng <bin.cheng@arm.com>
16063 PR tree-optimization/82604
16064 * tree-loop-distribution.c (enum partition_kind): New enum item
16065 PKIND_PARTIAL_MEMSET.
16066 (partition_builtin_p): Support above new enum item.
16067 (generate_code_for_partition): Ditto.
16068 (compute_access_range): Differentiate cases that equality can be
16069 proven at all loops, the innermost loops or no loops.
16070 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
16071 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
16072 (finalize_partitions, distribute_loop): Don't fuse partition of
16073 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
16074 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
16075 parloop is enabled.
16077 2018-01-23 Martin Liska <mliska@suse.cz>
16079 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
16080 order to ignore the predictor.
16081 (PRED_POLYMORPHIC_CALL): Likewise.
16082 (PRED_RECURSIVE_CALL): Likewise.
16084 2018-01-23 Martin Liska <mliska@suse.cz>
16086 * tree-profile.c (tree_profiling): Print function header to
16087 aware reader which function we are working on.
16088 * value-prof.c (gimple_find_values_to_profile): Do not print
16089 not interesting value histograms.
16091 2018-01-23 Martin Liska <mliska@suse.cz>
16093 * profile-count.h (enum profile_quality): Add
16094 profile_uninitialized as the first value. Do not number values
16095 as they are zero based.
16096 (profile_count::verify): Update sanity check.
16097 (profile_probability::verify): Likewise.
16099 2018-01-23 Nathan Sidwell <nathan@acm.org>
16101 * doc/invoke.texi (ffor-scope): Deprecate.
16103 2018-01-23 David Malcolm <dmalcolm@redhat.com>
16105 PR tree-optimization/83510
16106 * domwalk.c (set_all_edges_as_executable): New function.
16107 (dom_walker::dom_walker): Convert bool param
16108 "skip_unreachable_blocks" to enum reachability. Move setup of
16109 edge flags to set_all_edges_as_executable and only do it when
16110 reachability is REACHABLE_BLOCKS.
16111 * domwalk.h (enum dom_walker::reachability): New enum.
16112 (dom_walker::dom_walker): Convert bool param
16113 "skip_unreachable_blocks" to enum reachability.
16114 (set_all_edges_as_executable): New decl.
16115 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
16116 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
16118 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
16119 but converting true to REACHABLE_BLOCKS.
16120 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
16122 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
16123 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
16124 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
16126 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
16127 if check_all_array_refs will be called.
16129 2018-01-23 David Malcolm <dmalcolm@redhat.com>
16131 * tree.c (selftest::test_location_wrappers): Add more test
16134 2018-01-23 David Malcolm <dmalcolm@redhat.com>
16136 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
16137 (selftest::test_bit_in_range): Likewise.
16139 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
16142 * doc/sourcebuild.texi (vect_float): Say that the selector
16143 only describes the situation when -funsafe-math-optimizations is on.
16144 (vect_float_strict): Document.
16146 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
16148 PR tree-optimization/83965
16149 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
16150 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
16151 instead of checking only for a reduction.
16152 (vect_recog_widen_sum_pattern): Likewise.
16154 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
16156 * predict.c (probably_never_executed): Only use precise profile info.
16157 (compute_function_frequency): Skip after inlining hack since we now
16158 have quality checking.
16160 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
16162 * profile-count.h (profile_probability::very_unlikely,
16163 profile_probability::unlikely, profile_probability::even): Set
16164 precision to guessed.
16166 2018-01-23 Richard Biener <rguenther@suse.de>
16168 PR tree-optimization/83963
16169 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
16170 Properly terminate dominator walk when crossing the exit edge not
16171 when visiting its source block.
16173 2018-01-23 Jakub Jelinek <jakub@redhat.com>
16176 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
16177 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
16179 2018-01-22 Jakub Jelinek <jakub@redhat.com>
16181 PR tree-optimization/83957
16182 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
16183 semicolon after for body surrounded by braces.
16185 PR tree-optimization/83081
16186 * profile-count.h (profile_probability::split): New method.
16187 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
16188 Use profile_probability::split.
16189 (do_compare_rtx_and_jump): Fix adjustment of probabilities
16190 when splitting a single conditional jump into 2.
16192 2018-01-22 David Malcolm <dmalcolm@redhat.com>
16194 PR tree-optimization/69452
16195 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
16198 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16200 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
16201 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
16202 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
16204 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16206 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
16208 * config/rl78/rl78.md (movdi): New define_expand.
16209 * config/rl78/rl78.c (rl78_split_movdi): New function.
16211 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
16214 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
16216 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
16217 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
16218 128-bit to produce an UNSPEC move to get the double word with the
16219 signbit and then a shift directly to do signbit.
16220 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
16221 implementation with a new version that just does either a direct
16222 move or a regular move. Move memory interface to separate insns.
16223 Move insns so they are next to the expander.
16224 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
16225 with signbit move. Split big and little endian case.
16226 (signbit<mode>2_dm_mem_le): Likewise.
16227 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
16228 (signbit<mode>2_dm2): Likewise.
16230 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16232 * config/rl78/rl78.md (anddi3): New define_expand.
16234 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16236 * config/rl78/rl78.md (umindi3): New define_expand.
16238 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16240 * config/rl78/rl78.md (smindi3): New define_expand.
16242 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16244 * config/rl78/rl78.md (smaxdi3): New define_expand.
16246 2018-01-22 Carl Love <cel@us.ibm.com>
16248 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
16249 LVX_V1TI): Add macro expansion.
16250 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
16251 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
16252 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
16253 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
16254 Change check to determine if the instruction is a byte reversing
16255 entry. Fix typo in comment.
16256 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
16257 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
16258 Add def_builtin calls for new builtins.
16259 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
16260 Add define_insn expansion.
16262 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16264 * config/rl78/rl78.md (umaxdi3): New define_expand.
16266 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16268 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
16269 for non-QImode registers.
16271 2018-01-22 Richard Biener <rguenther@suse.de>
16273 PR tree-optimization/83963
16274 * graphite-scop-detection.c (scop_detection::get_sese): Delay
16275 including the loop exit block.
16276 (scop_detection::merge_sese): Likewise.
16277 (scop_detection::add_scop): Do it here instead.
16279 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16281 * doc/sourcebuild.texi (arm_softfloat): Document.
16283 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
16286 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
16287 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
16288 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
16290 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16291 David Edelsohn <dje.gcc@gmail.com>
16294 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
16295 Change "crset eq" to "crset 2".
16296 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
16297 (*call_indirect_aix<mode>_nospec): Likewise.
16298 (*call_value_indirect_aix<mode>_nospec): Likewise.
16299 (*call_indirect_elfv2<mode>_nospec): Likewise.
16300 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
16301 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
16302 change assembly output from . to $.
16303 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
16304 (indirect_jump<mode>_nospec): Change assembly output from . to $.
16305 (*tablejump<mode>_internal1_nospec): Likewise.
16307 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
16310 * config/sh/sh_optimize_sett_clrt.cc:
16311 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
16313 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
16315 PR tree-optimization/83940
16316 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
16317 offset_dt to vect_constant_def rather than vect_unknown_def_type.
16318 (vect_check_load_store_mask): Add a mask_dt_out parameter and
16319 use it to pass back the definition type.
16320 (vect_check_store_rhs): Likewise rhs_dt_out.
16321 (vect_build_gather_load_calls): Add a mask_dt argument and use
16322 it instead of a call to vect_is_simple_use.
16323 (vectorizable_store): Update calls to vect_check_load_store_mask
16324 and vect_check_store_rhs. Use the dt returned by the latter instead
16325 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
16326 instead of calls to vect_is_simple_use. Pass the scalar rather
16327 than the vector operand to vect_is_simple_use when handling
16328 second and subsequent copies of an rhs value.
16329 (vectorizable_load): Update calls to vect_check_load_store_mask
16330 and vect_build_gather_load_calls. Use the cached mask_dt and
16331 gs_info.offset_dt instead of calls to vect_is_simple_use.
16333 2018-01-20 Jakub Jelinek <jakub@redhat.com>
16335 PR middle-end/83945
16336 * tree-emutls.c: Include gimplify.h.
16337 (lower_emutls_2): New function.
16338 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
16339 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
16340 it before further processing.
16343 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
16344 UINTVAL (trueop1) instead of INTVAL (op1).
16346 2018-01-19 Jakub Jelinek <jakub@redhat.com>
16350 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
16351 INCOMING_FRAME_SP_OFFSET if not defined.
16352 (scan_trace): Add ENTRY argument. If true and
16353 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
16354 emit a note to adjust the CFA offset.
16355 (create_cfi_notes): Adjust scan_trace callers.
16356 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
16357 INCOMING_FRAME_SP_OFFSET in the CIE.
16358 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
16359 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
16361 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
16362 * doc/tm.texi: Regenerated.
16364 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
16366 PR rtl-optimization/83147
16367 * lra-constraints.c (remove_inheritance_pseudos): Use
16368 lra_substitute_pseudo_within_insn.
16370 2018-01-19 Tom de Vries <tom@codesourcery.com>
16371 Cesar Philippidis <cesar@codesourcery.com>
16374 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
16376 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
16379 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
16380 spaces for function labels.
16382 2018-01-19 Martin Liska <mliska@suse.cz>
16384 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
16385 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
16386 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
16387 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
16388 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
16389 (PRED_CONST_RETURN): Change from 69 to 65.
16390 (PRED_NULL_RETURN): Change from 91 to 71.
16391 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
16392 (PRED_LOOP_GUARD): Change from 66 to 73.
16394 2018-01-19 Martin Liska <mliska@suse.cz>
16396 * predict.c (predict_insn_def): Add new assert.
16397 (struct branch_predictor): Change type to signed integer.
16398 (test_prediction_value_range): Amend test to cover
16399 PROB_UNINITIALIZED.
16400 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
16401 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
16402 (PRED_LOOP_ITERATIONS_MAX): Likewise.
16403 (PRED_LOOP_IV_COMPARE): Likewise.
16404 * predict.h (PROB_UNINITIALIZED): Define new constant.
16406 2018-01-19 Martin Liska <mliska@suse.cz>
16408 * predict.c (dump_prediction): Add new format for
16409 analyze_brprob.py script which is enabled with -details
16411 * profile-count.h (precise_p): New function.
16413 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
16415 PR tree-optimization/83922
16416 * tree-vect-loop.c (vect_verify_full_masking): Return false if
16417 there are no statements that need masking.
16418 (vect_active_double_reduction_p): New function.
16419 (vect_analyze_loop_operations): Use it when handling phis that
16420 are not in the loop header.
16422 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
16424 PR tree-optimization/83914
16425 * tree-vect-loop.c (vectorizable_induction): Don't convert
16426 init_expr or apply the peeling adjustment for inductions
16427 that are nested within the vectorized loop.
16429 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16431 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
16434 2018-01-18 Jakub Jelinek <jakub@redhat.com>
16438 * function.h (gimplify_parameters): Add gimple_seq * argument.
16439 * function.c: Include gimple.h and options.h.
16440 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
16441 for the added local temporaries if needed.
16442 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
16443 if there are any parameter cleanups, wrap whole body into a
16444 try/finally with the cleanups.
16446 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
16449 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
16450 Use GET_MODE_CLASS for scalar floating point.
16452 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
16456 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
16457 Fix call of call_cgraph_insertion_hooks.
16459 2018-01-18 Martin Sebor <msebor@redhat.com>
16461 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
16463 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
16466 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
16469 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
16472 * common.opt: (-ffile-prefix-map): New option.
16473 * opts.c (common_handle_option): Defer it.
16474 * opts-global.c (handle_common_deferred_options): Handle it.
16475 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
16476 * file-prefix-map.h: New file.
16477 (remap_debug_filename, add_debug_prefix_map): ...here.
16478 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
16479 * final.c (debug_prefix_map, add_debug_prefix_map
16480 remap_debug_filename): Move to...
16481 * file-prefix-map.c: New file.
16482 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
16483 generalize, get rid of alloca(), use strrchr() instead of strchr().
16484 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
16485 Implement in terms of add_prefix_map().
16486 (remap_macro_filename, remap_debug_filename): Implement in term of
16488 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
16489 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
16490 * dbxout.c: Include file-prefix-map.h.
16491 * varasm.c: Likewise.
16492 * vmsdbgout.c: Likewise.
16493 * xcoffout.c: Likewise.
16494 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
16495 * doc/cppopts.texi (-fmacro-prefix-map): Document.
16496 * doc/invoke.texi (-ffile-prefix-map): Document.
16497 (-fdebug-prefix-map): Update description.
16499 2018-01-18 Martin Liska <mliska@suse.cz>
16501 * config/i386/i386.c (indirect_thunk_name): Document that also
16503 (output_indirect_thunk): Document why both instructions
16504 (pause and lfence) are generated.
16506 2018-01-18 Richard Biener <rguenther@suse.de>
16508 PR tree-optimization/83887
16509 * graphite-scop-detection.c
16510 (scop_detection::get_nearest_dom_with_single_entry): Remove.
16511 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
16512 (scop_detection::merge_sese): Re-implement with a flood-fill
16513 algorithm that properly finds a SESE region if it exists.
16515 2018-01-18 Jakub Jelinek <jakub@redhat.com>
16518 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
16519 pointer_diff optimizations use view_convert instead of convert.
16521 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16523 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
16524 Generate different code for -mno-speculate-indirect-jumps.
16525 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
16526 (*call_indirect_aix<mode>): Disable for
16527 -mno-speculate-indirect-jumps.
16528 (*call_indirect_aix<mode>_nospec): New define_insn.
16529 (*call_value_indirect_aix<mode>): Disable for
16530 -mno-speculate-indirect-jumps.
16531 (*call_value_indirect_aix<mode>_nospec): New define_insn.
16532 (*sibcall_nonlocal_sysv<mode>): Generate different code for
16533 -mno-speculate-indirect-jumps.
16534 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
16536 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
16538 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
16539 long double type, set the flags for noting the default long double
16540 type, even if we don't pass or return a long double type.
16542 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
16545 * ipa-inline.c (flatten_function): Do not overwrite final inlining
16548 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
16550 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
16551 support for merge[hl].
16552 (fold_mergehl_helper): New helper function.
16553 (tree-vector-builder.h): New #include for tree_vector_builder usage.
16554 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
16555 (altivec_vmrglw_direct): Add xxmrglw insn.
16557 2018-01-17 Andrew Waterman <andrew@sifive.com>
16559 * config/riscv/riscv.c (riscv_conditional_register_usage): If
16560 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
16562 2018-01-17 David Malcolm <dmalcolm@redhat.com>
16565 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
16566 call the lto_location_cache before reading the
16567 DECL_SOURCE_LOCATION of the types.
16569 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
16570 Richard Sandiford <richard.sandiford@linaro.org>
16572 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
16573 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
16574 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
16575 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
16576 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
16578 * config/aarch64/constraints.md (aarch64_movti_operand):
16580 * config/aarch64/predicates.md (Uti): Add new constraint.
16582 2018-01-17 Carl Love <cel@us.ibm.com>
16584 * config/rs6000/vsx.md (define_expand xl_len_r,
16585 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
16586 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
16588 (define_expand, define_insn): Move the shift left from the
16589 define_insn to the define_expand for lxvl and stxvl instructions.
16590 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
16591 and XL_LEN_R definitions to PURE.
16593 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
16595 * config/i386/i386.c (indirect_thunk_name): Declare regno
16596 as unsigned int. Compare regno with INVALID_REGNUM.
16597 (output_indirect_thunk): Ditto.
16598 (output_indirect_thunk_function): Ditto.
16599 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
16600 in the call to output_indirect_thunk_function.
16602 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
16604 PR middle-end/83884
16605 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
16606 rather than the size of inner_type to determine the stack slot size
16607 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
16609 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
16612 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
16615 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
16617 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
16618 endian Linux systems to optionally enable multilibs for selecting
16619 the long double type if the user configured an explicit type.
16620 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
16621 have no long double multilibs if not defined.
16622 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
16623 warn if the user used -mabi={ieee,ibm}longdouble and we built
16624 multilibs for long double.
16625 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
16626 appropriate multilib option.
16627 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
16629 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
16630 for building long double multilibs.
16631 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
16633 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
16635 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
16638 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
16640 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
16643 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
16646 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
16649 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
16651 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
16652 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
16654 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
16656 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
16657 different rtl trees depending on TARGET_64BIT.
16658 (rs6000_gen_lvx): Likewise.
16660 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
16662 * config/visium/visium.md (nop): Tweak comment.
16663 (hazard_nop): Likewise.
16665 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16667 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
16668 -mspeculate-indirect-jumps.
16669 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
16670 for -mno-speculate-indirect-jumps.
16671 (*call_indirect_elfv2<mode>_nospec): New define_insn.
16672 (*call_value_indirect_elfv2<mode>): Disable for
16673 -mno-speculate-indirect-jumps.
16674 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
16675 (indirect_jump): Emit different RTL for
16676 -mno-speculate-indirect-jumps.
16677 (*indirect_jump<mode>): Disable for
16678 -mno-speculate-indirect-jumps.
16679 (*indirect_jump<mode>_nospec): New define_insn.
16680 (tablejump): Emit different RTL for
16681 -mno-speculate-indirect-jumps.
16682 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
16683 (tablejumpsi_nospec): New define_expand.
16684 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
16685 (tablejumpdi_nospec): New define_expand.
16686 (*tablejump<mode>_internal1): Disable for
16687 -mno-speculate-indirect-jumps.
16688 (*tablejump<mode>_internal1_nospec): New define_insn.
16689 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
16692 2018-01-16 Artyom Skrobov tyomitch@gmail.com
16694 * caller-save.c (insert_save): Drop unnecessary parameter. All
16697 2018-01-16 Jakub Jelinek <jakub@redhat.com>
16698 Richard Biener <rguenth@suse.de>
16701 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
16702 return early, inline manually is_gimple_sizepos. Make sure if we
16703 call gimplify_expr we don't end up with a gimple constant.
16704 * tree.c (variably_modified_type_p): Don't return true for
16705 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
16706 * gimplify.h (is_gimple_sizepos): Remove.
16708 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
16710 PR tree-optimization/83857
16711 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
16712 vectorizable_live_operation for pure SLP statements.
16713 (vectorizable_live_operation): Handle PHIs.
16715 2018-01-16 Richard Biener <rguenther@suse.de>
16717 PR tree-optimization/83867
16718 * tree-vect-stmts.c (vect_transform_stmt): Precompute
16719 nested_in_vect_loop_p since the scalar stmt may get invalidated.
16721 2018-01-16 Jakub Jelinek <jakub@redhat.com>
16724 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
16725 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
16726 If off is not INTEGER_CST, issue a may not be aligned warning
16727 rather than isn't aligned. Use isn%'t rather than isn't.
16728 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
16730 <case MULT_EXPR>: Improve the case when bottom and one of the
16731 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
16732 operand, in that case check if the other operand is multiple of
16733 bottom divided by the INTEGER_CST operand.
16735 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
16738 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
16739 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
16740 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
16741 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
16742 * config/pa/pa.c (pa_function_arg_advance): Likewise.
16743 (pa_function_arg, pa_arg_partial_bytes): Likewise.
16744 (pa_function_arg_size): New function.
16746 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
16748 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
16749 in a separate statement.
16751 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
16753 PR tree-optimization/83847
16754 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
16755 group gathers and scatters.
16757 2018-01-16 Jakub Jelinek <jakub@redhat.com>
16759 PR rtl-optimization/86620
16760 * params.def (max-sched-ready-insns): Bump minimum value to 1.
16762 PR rtl-optimization/83213
16763 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
16764 to last if both are JUMP_INSNs.
16766 PR tree-optimization/83843
16767 * gimple-ssa-store-merging.c
16768 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
16769 store_immediate_info for bswap/nop orig_stores.
16771 2018-01-15 Andrew Waterman <andrew@sifive.com>
16773 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
16775 <UDIV>: Increase cost if !TARGET_DIV.
16777 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
16779 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
16780 (define_attr "cr_logical_3op"): New.
16781 (cceq_ior_compare): Adjust.
16782 (cceq_ior_compare_complement): Adjust.
16783 (*cceq_rev_compare): Adjust.
16784 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
16785 (is_cracked_insn): Adjust.
16786 (insn_must_be_first_in_group): Adjust.
16787 * config/rs6000/40x.md: Adjust.
16788 * config/rs6000/440.md: Adjust.
16789 * config/rs6000/476.md: Adjust.
16790 * config/rs6000/601.md: Adjust.
16791 * config/rs6000/603.md: Adjust.
16792 * config/rs6000/6xx.md: Adjust.
16793 * config/rs6000/7450.md: Adjust.
16794 * config/rs6000/7xx.md: Adjust.
16795 * config/rs6000/8540.md: Adjust.
16796 * config/rs6000/cell.md: Adjust.
16797 * config/rs6000/e300c2c3.md: Adjust.
16798 * config/rs6000/e500mc.md: Adjust.
16799 * config/rs6000/e500mc64.md: Adjust.
16800 * config/rs6000/e5500.md: Adjust.
16801 * config/rs6000/e6500.md: Adjust.
16802 * config/rs6000/mpc.md: Adjust.
16803 * config/rs6000/power4.md: Adjust.
16804 * config/rs6000/power5.md: Adjust.
16805 * config/rs6000/power6.md: Adjust.
16806 * config/rs6000/power7.md: Adjust.
16807 * config/rs6000/power8.md: Adjust.
16808 * config/rs6000/power9.md: Adjust.
16809 * config/rs6000/rs64.md: Adjust.
16810 * config/rs6000/titan.md: Adjust.
16812 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
16814 * config/i386/predicates.md (indirect_branch_operand): Rewrite
16815 ix86_indirect_branch_register logic.
16817 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
16819 * config/i386/constraints.md (Bs): Update
16820 ix86_indirect_branch_register check. Don't check
16821 ix86_indirect_branch_register with GOT_memory_operand.
16823 * config/i386/predicates.md (GOT_memory_operand): Don't check
16824 ix86_indirect_branch_register here.
16825 (GOT32_symbol_operand): Likewise.
16827 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
16829 * config/i386/predicates.md (constant_call_address_operand):
16830 Rewrite ix86_indirect_branch_register logic.
16831 (sibcall_insn_operand): Likewise.
16833 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
16835 * config/i386/constraints.md (Bs): Replace
16836 ix86_indirect_branch_thunk_register with
16837 ix86_indirect_branch_register.
16839 * config/i386/i386.md (indirect_jump): Likewise.
16840 (tablejump): Likewise.
16841 (*sibcall_memory): Likewise.
16842 (*sibcall_value_memory): Likewise.
16843 Peepholes of indirect call and jump via memory: Likewise.
16844 * config/i386/i386.opt: Likewise.
16845 * config/i386/predicates.md (indirect_branch_operand): Likewise.
16846 (GOT_memory_operand): Likewise.
16847 (call_insn_operand): Likewise.
16848 (sibcall_insn_operand): Likewise.
16849 (GOT32_symbol_operand): Likewise.
16851 2018-01-15 Jakub Jelinek <jakub@redhat.com>
16853 PR middle-end/83837
16854 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
16855 type rather than type addr's type points to.
16856 (expand_omp_atomic_mutex): Likewise.
16857 (expand_omp_atomic): Likewise.
16859 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
16862 * config/i386/i386.c (output_indirect_thunk_function): Use
16863 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
16864 for __x86_return_thunk.
16866 2018-01-15 Richard Biener <rguenther@suse.de>
16868 PR middle-end/83850
16869 * expmed.c (extract_bit_field_1): Fix typo.
16871 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16874 * config/arm/iterators.md (VF): New mode iterator.
16875 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
16876 Remove integer-related logic from pattern.
16877 (neon_vabd<mode>_3): Likewise.
16879 2018-01-15 Jakub Jelinek <jakub@redhat.com>
16881 PR middle-end/82694
16882 * common.opt (fstrict-overflow): No longer an alias.
16883 (fwrapv-pointer): New option.
16884 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
16885 also for pointer types based on flag_wrapv_pointer.
16886 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
16887 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
16888 opts->x_flag_wrapv got set.
16889 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
16890 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
16891 POINTER_TYPE_OVERFLOW_UNDEFINED.
16892 * match.pd: Likewise in address comparison pattern.
16893 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
16895 2018-01-15 Richard Biener <rguenther@suse.de>
16898 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
16899 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
16900 Reset type names to their identifier if their TYPE_DECL doesn't
16901 have linkage (and thus is used for ODR and devirt).
16902 (save_debug_info_for_decl): Remove.
16903 (save_debug_info_for_type): Likewise.
16904 (add_tree_to_fld_list): Adjust.
16905 * tree-pretty-print.c (dump_generic_node): Make dumping of
16906 type names more robust.
16908 2018-01-15 Richard Biener <rguenther@suse.de>
16910 * BASE-VER: Bump to 8.0.1.
16912 2018-01-14 Martin Sebor <msebor@redhat.com>
16915 * builtins.c (check_access): Avoid warning when the no-warning bit
16918 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
16920 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
16921 * ira-color (allocno_hard_regs_compare): Likewise.
16923 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
16926 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
16927 Use .pushsection/.popsection.
16929 2018-01-14 Martin Sebor <msebor@redhat.com>
16932 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
16934 2018-01-14 Jakub Jelinek <jakub@redhat.com>
16936 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
16937 entry from extra_headers.
16938 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
16939 extra_headers, make the list bitwise identical to the i?86-*-* one.
16941 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
16943 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
16944 -mcmodel=large with -mindirect-branch=thunk,
16945 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
16946 -mfunction-return=thunk-extern.
16947 * doc/invoke.texi: Document -mcmodel=large is incompatible with
16948 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
16949 -mfunction-return=thunk and -mfunction-return=thunk-extern.
16951 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
16953 * config/i386/i386.c (print_reg): Print the name of the full
16954 integer register without '%'.
16955 (ix86_print_operand): Handle 'V'.
16956 * doc/extend.texi: Document 'V' modifier.
16958 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
16960 * config/i386/constraints.md (Bs): Disallow memory operand for
16961 -mindirect-branch-register.
16963 * config/i386/predicates.md (indirect_branch_operand): Likewise.
16964 (GOT_memory_operand): Likewise.
16965 (call_insn_operand): Likewise.
16966 (sibcall_insn_operand): Likewise.
16967 (GOT32_symbol_operand): Likewise.
16968 * config/i386/i386.md (indirect_jump): Call convert_memory_address
16969 for -mindirect-branch-register.
16970 (tablejump): Likewise.
16971 (*sibcall_memory): Likewise.
16972 (*sibcall_value_memory): Likewise.
16973 Disallow peepholes of indirect call and jump via memory for
16974 -mindirect-branch-register.
16975 (*call_pop): Replace m with Bw.
16976 (*call_value_pop): Likewise.
16977 (*sibcall_pop_memory): Replace m with Bs.
16978 * config/i386/i386.opt (mindirect-branch-register): New option.
16979 * doc/invoke.texi: Document -mindirect-branch-register option.
16981 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
16983 * config/i386/i386-protos.h (ix86_output_function_return): New.
16984 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
16985 set function_return_type.
16986 (indirect_thunk_name): Add ret_p to indicate thunk for function
16988 (output_indirect_thunk_function): Pass false to
16989 indirect_thunk_name.
16990 (ix86_output_indirect_branch_via_reg): Likewise.
16991 (ix86_output_indirect_branch_via_push): Likewise.
16992 (output_indirect_thunk_function): Create alias for function
16993 return thunk if regno < 0.
16994 (ix86_output_function_return): New function.
16995 (ix86_handle_fndecl_attribute): Handle function_return.
16996 (ix86_attribute_table): Add function_return.
16997 * config/i386/i386.h (machine_function): Add
16998 function_return_type.
16999 * config/i386/i386.md (simple_return_internal): Use
17000 ix86_output_function_return.
17001 (simple_return_internal_long): Likewise.
17002 * config/i386/i386.opt (mfunction-return=): New option.
17003 (indirect_branch): Mention -mfunction-return=.
17004 * doc/extend.texi: Document function_return function attribute.
17005 * doc/invoke.texi: Document -mfunction-return= option.
17007 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
17009 * config/i386/i386-opts.h (indirect_branch): New.
17010 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
17011 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
17012 with local indirect jump when converting indirect call and jump.
17013 (ix86_set_indirect_branch_type): New.
17014 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
17015 (indirectlabelno): New.
17016 (indirect_thunk_needed): Likewise.
17017 (indirect_thunk_bnd_needed): Likewise.
17018 (indirect_thunks_used): Likewise.
17019 (indirect_thunks_bnd_used): Likewise.
17020 (INDIRECT_LABEL): Likewise.
17021 (indirect_thunk_name): Likewise.
17022 (output_indirect_thunk): Likewise.
17023 (output_indirect_thunk_function): Likewise.
17024 (ix86_output_indirect_branch_via_reg): Likewise.
17025 (ix86_output_indirect_branch_via_push): Likewise.
17026 (ix86_output_indirect_branch): Likewise.
17027 (ix86_output_indirect_jmp): Likewise.
17028 (ix86_code_end): Call output_indirect_thunk_function if needed.
17029 (ix86_output_call_insn): Call ix86_output_indirect_branch if
17031 (ix86_handle_fndecl_attribute): Handle indirect_branch.
17032 (ix86_attribute_table): Add indirect_branch.
17033 * config/i386/i386.h (machine_function): Add indirect_branch_type
17034 and has_local_indirect_jump.
17035 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
17037 (tablejump): Likewise.
17038 (*indirect_jump): Use ix86_output_indirect_jmp.
17039 (*tablejump_1): Likewise.
17040 (simple_return_indirect_internal): Likewise.
17041 * config/i386/i386.opt (mindirect-branch=): New option.
17042 (indirect_branch): New.
17045 (thunk-inline): Likewise.
17046 (thunk-extern): Likewise.
17047 * doc/extend.texi: Document indirect_branch function attribute.
17048 * doc/invoke.texi: Document -mindirect-branch= option.
17050 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
17053 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
17055 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
17057 * ipa-inline.c (want_inline_small_function_p): Return false if
17058 inlining has already failed with CIF_FINAL_ERROR.
17059 (update_caller_keys): Call want_inline_small_function_p before
17061 (update_callee_keys): Likewise.
17063 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
17065 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
17067 (rs6000_quadword_masked_address_p): Likewise.
17068 (quad_aligned_load_p): Likewise.
17069 (quad_aligned_store_p): Likewise.
17070 (const_load_sequence_p): Add comment to describe the outer-most loop.
17071 (mimic_memory_attributes_and_flags): New function.
17072 (rs6000_gen_stvx): Likewise.
17073 (replace_swapped_aligned_store): Likewise.
17074 (rs6000_gen_lvx): Likewise.
17075 (replace_swapped_aligned_load): Likewise.
17076 (replace_swapped_load_constant): Capitalize argument name in
17077 comment describing this function.
17078 (rs6000_analyze_swaps): Add a third pass to search for vector loads
17079 and stores that access quad-word aligned addresses and replace
17080 with stvx or lvx instructions when appropriate.
17081 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
17082 New function prototype.
17083 (rs6000_quadword_masked_address_p): Likewise.
17084 (rs6000_gen_lvx): Likewise.
17085 (rs6000_gen_stvx): Likewise.
17086 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
17087 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
17088 when memory address is aligned.
17089 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
17090 this split to select lvx instruction when memory address is aligned.
17091 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
17092 instruction when memory address is aligned.
17093 (*vsx_le_perm_load_v16qi): Likewise.
17094 (four unnamed splitters): Modify to select the stvx instruction
17095 when memory is aligned.
17097 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
17099 * predict.c (determine_unlikely_bbs): Handle correctly BBs
17100 which appears in the queue multiple times.
17102 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17103 Alan Hayward <alan.hayward@arm.com>
17104 David Sherwood <david.sherwood@arm.com>
17106 * tree-vectorizer.h (vec_lower_bound): New structure.
17107 (_loop_vec_info): Add check_nonzero and lower_bounds.
17108 (LOOP_VINFO_CHECK_NONZERO): New macro.
17109 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
17110 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
17111 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
17112 fields. Make seg_len the distance travelled, not including the
17114 (dr_direction_indicator): Declare.
17115 (dr_zero_step_indicator): Likewise.
17116 (dr_known_forward_stride_p): Likewise.
17117 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
17119 (runtime_alias_check_p): Allow runtime alias checks with
17121 (operator ==): Compare access_size and align.
17122 (prune_runtime_alias_test_list): Rework for new distinction between
17123 the access_size and seg_len.
17124 (create_intersect_range_checks_index): Likewise. Cope with polynomial
17126 (get_segment_min_max): New function.
17127 (create_intersect_range_checks): Use it.
17128 (dr_step_indicator): New function.
17129 (dr_direction_indicator): Likewise.
17130 (dr_zero_step_indicator): Likewise.
17131 (dr_known_forward_stride_p): Likewise.
17132 * tree-loop-distribution.c (data_ref_segment_size): Return
17133 DR_STEP * (niters - 1).
17134 (compute_alias_check_pairs): Update call to the dr_with_seg_len
17136 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
17137 (vect_preserves_scalar_order_p): New function, split out from...
17138 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
17139 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
17140 (vect_vfa_access_size): New function.
17141 (vect_vfa_align): Likewise.
17142 (vect_compile_time_alias): Take access_size_a and access_b arguments.
17143 (dump_lower_bound): New function.
17144 (vect_check_lower_bound): Likewise.
17145 (vect_small_gap_p): Likewise.
17146 (vectorizable_with_step_bound_p): Likewise.
17147 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
17148 depencies if the vectorization factor is 1. Convert the checks
17149 for nonzero steps into checks on the bounds of DR_STEP. Try using
17150 a bunds check for variable steps if the minimum required step is
17151 relatively small. Update calls to the dr_with_seg_len
17152 constructor and to vect_compile_time_alias.
17153 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
17155 (vect_loop_versioning): Call it.
17156 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
17158 (vect_estimate_min_profitable_iters): Account for any bounds checks.
17160 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17161 Alan Hayward <alan.hayward@arm.com>
17162 David Sherwood <david.sherwood@arm.com>
17164 * doc/sourcebuild.texi (vect_scatter_store): Document.
17165 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
17167 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
17169 * genopinit.c (main): Add supports_vec_scatter_store and
17170 supports_vec_scatter_store_cached to target_optabs.
17171 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
17172 IFN_MASK_SCATTER_STORE.
17173 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
17175 * internal-fn.h (internal_store_fn_p): Declare.
17176 (internal_fn_stored_value_index): Likewise.
17177 * internal-fn.c (scatter_store_direct): New macro.
17178 (expand_scatter_store_optab_fn): New function.
17179 (direct_scatter_store_optab_supported_p): New macro.
17180 (internal_store_fn_p): New function.
17181 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
17182 IFN_MASK_SCATTER_STORE.
17183 (internal_fn_mask_index): Likewise.
17184 (internal_fn_stored_value_index): New function.
17185 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
17186 for scatter stores.
17187 * optabs-query.h (supports_vec_scatter_store_p): Declare.
17188 * optabs-query.c (supports_vec_scatter_store_p): New function.
17189 * tree-vectorizer.h (vect_get_store_rhs): Declare.
17190 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
17191 true for scatter stores.
17192 (vect_gather_scatter_fn_p): Handle scatter stores too.
17193 (vect_check_gather_scatter): Consider using scatter stores if
17194 supports_vec_scatter_store_p.
17195 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
17196 scatter stores too.
17197 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
17198 internal_fn_stored_value_index.
17199 (check_load_store_masking): Handle scatter stores too.
17200 (vect_get_store_rhs): Make public.
17201 (vectorizable_call): Use internal_store_fn_p.
17202 (vectorizable_store): Handle scatter store internal functions.
17203 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
17204 when deciding whether the end of the group has been reached.
17205 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
17206 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
17207 (mask_scatter_store<mode>): New insns.
17209 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17210 Alan Hayward <alan.hayward@arm.com>
17211 David Sherwood <david.sherwood@arm.com>
17213 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
17214 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
17215 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
17217 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
17218 Use vect_truncate_gather_scatter_offset if we can't treat the
17219 operation as a normal gather load or scatter store.
17220 (get_group_load_store_type): Take the gather_scatter_info
17221 as argument. Try using a gather load or scatter store for
17222 single-element groups.
17223 (get_load_store_type): Update calls to get_group_load_store_type
17224 and vect_use_strided_gather_scatters_p.
17226 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17227 Alan Hayward <alan.hayward@arm.com>
17228 David Sherwood <david.sherwood@arm.com>
17230 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
17231 optional tree argument.
17232 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
17234 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
17235 but continue to use the current value as a fallback.
17236 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
17237 to compare the updates.
17238 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
17239 (get_load_store_type): Use it when handling a strided access.
17240 (vect_get_strided_load_store_ops): New function.
17241 (vect_get_data_ptr_increment): Likewise.
17242 (vectorizable_load): Handle strided gather loads. Always pass
17243 a step to vect_create_data_ref_ptr and bump_vector_ptr.
17245 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17246 Alan Hayward <alan.hayward@arm.com>
17247 David Sherwood <david.sherwood@arm.com>
17249 * doc/md.texi (gather_load@var{m}): Document.
17250 (mask_gather_load@var{m}): Likewise.
17251 * genopinit.c (main): Add supports_vec_gather_load and
17252 supports_vec_gather_load_cached to target_optabs.
17253 * optabs-tree.c (init_tree_optimization_optabs): Use
17254 ggc_cleared_alloc to allocate target_optabs.
17255 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
17256 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
17258 * internal-fn.h (internal_load_fn_p): Declare.
17259 (internal_gather_scatter_fn_p): Likewise.
17260 (internal_fn_mask_index): Likewise.
17261 (internal_gather_scatter_fn_supported_p): Likewise.
17262 * internal-fn.c (gather_load_direct): New macro.
17263 (expand_gather_load_optab_fn): New function.
17264 (direct_gather_load_optab_supported_p): New macro.
17265 (direct_internal_fn_optab): New function.
17266 (internal_load_fn_p): Likewise.
17267 (internal_gather_scatter_fn_p): Likewise.
17268 (internal_fn_mask_index): Likewise.
17269 (internal_gather_scatter_fn_supported_p): Likewise.
17270 * optabs-query.c (supports_at_least_one_mode_p): New function.
17271 (supports_vec_gather_load_p): Likewise.
17272 * optabs-query.h (supports_vec_gather_load_p): Declare.
17273 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
17274 and memory_type field.
17275 (NUM_PATTERNS): Bump to 15.
17276 * tree-vect-data-refs.c: Include internal-fn.h.
17277 (vect_gather_scatter_fn_p): New function.
17278 (vect_describe_gather_scatter_call): Likewise.
17279 (vect_check_gather_scatter): Try using internal functions for
17280 gather loads. Recognize existing calls to a gather load function.
17281 (vect_analyze_data_refs): Consider using gather loads if
17282 supports_vec_gather_load_p.
17283 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
17284 (vect_get_gather_scatter_offset_type): Likewise.
17285 (vect_convert_mask_for_vectype): Likewise.
17286 (vect_add_conversion_to_patterm): Likewise.
17287 (vect_try_gather_scatter_pattern): Likewise.
17288 (vect_recog_gather_scatter_pattern): New pattern recognizer.
17289 (vect_vect_recog_func_ptrs): Add it.
17290 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
17291 internal_fn_mask_index and internal_gather_scatter_fn_p.
17292 (check_load_store_masking): Take the gather_scatter_info as an
17293 argument and handle gather loads.
17294 (vect_get_gather_scatter_ops): New function.
17295 (vectorizable_call): Check internal_load_fn_p.
17296 (vectorizable_load): Likewise. Handle gather load internal
17298 (vectorizable_store): Update call to check_load_store_masking.
17299 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
17300 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
17301 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
17302 (aarch64_gather_scale_operand_d): New predicates.
17303 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
17304 (mask_gather_load<mode>): New insns.
17306 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17307 Alan Hayward <alan.hayward@arm.com>
17308 David Sherwood <david.sherwood@arm.com>
17310 * optabs.def (fold_left_plus_optab): New optab.
17311 * doc/md.texi (fold_left_plus_@var{m}): Document.
17312 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
17313 * internal-fn.c (fold_left_direct): Define.
17314 (expand_fold_left_optab_fn): Likewise.
17315 (direct_fold_left_optab_supported_p): Likewise.
17316 * fold-const-call.c (fold_const_fold_left): New function.
17317 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
17318 * tree-parloops.c (valid_reduction_p): New function.
17319 (gather_scalar_reductions): Use it.
17320 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
17321 (vect_finish_replace_stmt): Declare.
17322 * tree-vect-loop.c (fold_left_reduction_fn): New function.
17323 (needs_fold_left_reduction_p): New function, split out from...
17324 (vect_is_simple_reduction): ...here. Accept reductions that
17325 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
17326 (vect_force_simple_reduction): Also store the reduction type in
17327 the assignment's STMT_VINFO_REDUC_TYPE.
17328 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
17329 (merge_with_identity): New function.
17330 (vect_expand_fold_left): Likewise.
17331 (vectorize_fold_left_reduction): Likewise.
17332 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
17333 scalar phi in place for it. Check for target support and reject
17334 cases that would reassociate the operation. Defer the transform
17335 phase to vectorize_fold_left_reduction.
17336 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
17337 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
17338 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
17340 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17342 * tree-if-conv.c (predicate_mem_writes): Remove redundant
17343 call to ifc_temp_var.
17345 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17346 Alan Hayward <alan.hayward@arm.com>
17347 David Sherwood <david.sherwood@arm.com>
17349 * target.def (legitimize_address_displacement): Take the original
17350 offset as a poly_int.
17351 * targhooks.h (default_legitimize_address_displacement): Update
17353 * targhooks.c (default_legitimize_address_displacement): Likewise.
17354 * doc/tm.texi: Regenerate.
17355 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
17356 as an argument, moving assert of ad->disp == ad->disp_term to...
17357 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
17358 Try calling targetm.legitimize_address_displacement before expanding
17359 the address rather than afterwards, and adjust for the new interface.
17360 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
17361 Match the new hook interface. Handle SVE addresses.
17362 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
17363 new hook interface.
17365 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17367 * Makefile.in (OBJS): Add early-remat.o.
17368 * target.def (select_early_remat_modes): New hook.
17369 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
17370 * doc/tm.texi: Regenerate.
17371 * targhooks.h (default_select_early_remat_modes): Declare.
17372 * targhooks.c (default_select_early_remat_modes): New function.
17373 * timevar.def (TV_EARLY_REMAT): New timevar.
17374 * passes.def (pass_early_remat): New pass.
17375 * tree-pass.h (make_pass_early_remat): Declare.
17376 * early-remat.c: New file.
17377 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
17379 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
17381 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17382 Alan Hayward <alan.hayward@arm.com>
17383 David Sherwood <david.sherwood@arm.com>
17385 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
17386 vfm1 with a bound_epilog parameter.
17387 (vect_do_peeling): Update calls accordingly, and move the prologue
17388 call earlier in the function. Treat the base bound_epilog as 0 for
17389 fully-masked loops and retain vf - 1 for other loops. Add 1 to
17390 this base when peeling for gaps.
17391 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
17392 with fully-masked loops.
17393 (vect_estimate_min_profitable_iters): Handle the single peeled
17394 iteration in that case.
17396 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17397 Alan Hayward <alan.hayward@arm.com>
17398 David Sherwood <david.sherwood@arm.com>
17400 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
17401 single-element interleaving even if the size is not a power of 2.
17402 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
17403 accesses for single-element interleaving if the group size is
17406 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17407 Alan Hayward <alan.hayward@arm.com>
17408 David Sherwood <david.sherwood@arm.com>
17410 * doc/md.texi (fold_extract_last_@var{m}): Document.
17411 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
17412 * optabs.def (fold_extract_last_optab): New optab.
17413 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
17414 * internal-fn.c (fold_extract_direct): New macro.
17415 (expand_fold_extract_optab_fn): Likewise.
17416 (direct_fold_extract_optab_supported_p): Likewise.
17417 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
17418 * tree-vect-loop.c (vect_model_reduction_cost): Handle
17419 EXTRACT_LAST_REDUCTION.
17420 (get_initial_def_for_reduction): Do not create an initial vector
17421 for EXTRACT_LAST_REDUCTION reductions.
17422 (vectorizable_reduction): Leave the scalar phi in place for
17423 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
17424 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
17425 epilogue code for EXTRACT_LAST_REDUCTION and defer the
17426 transform phase to vectorizable_condition.
17427 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
17429 (vect_finish_stmt_generation): ...here.
17430 (vect_finish_replace_stmt): New function.
17431 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
17432 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
17434 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
17436 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17437 Alan Hayward <alan.hayward@arm.com>
17438 David Sherwood <david.sherwood@arm.com>
17440 * doc/md.texi (extract_last_@var{m}): Document.
17441 * optabs.def (extract_last_optab): New optab.
17442 * internal-fn.def (EXTRACT_LAST): New internal function.
17443 * internal-fn.c (cond_unary_direct): New macro.
17444 (expand_cond_unary_optab_fn): Likewise.
17445 (direct_cond_unary_optab_supported_p): Likewise.
17446 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
17447 loops using EXTRACT_LAST.
17448 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
17449 (extract_last_<mode>): ...this optab.
17450 (vec_extract<mode><Vel>): Update accordingly.
17452 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17453 Alan Hayward <alan.hayward@arm.com>
17454 David Sherwood <david.sherwood@arm.com>
17456 * target.def (empty_mask_is_expensive): New hook.
17457 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
17458 * doc/tm.texi: Regenerate.
17459 * targhooks.h (default_empty_mask_is_expensive): Declare.
17460 * targhooks.c (default_empty_mask_is_expensive): New function.
17461 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
17462 if the target says that empty masks are expensive.
17463 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
17465 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
17467 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17468 Alan Hayward <alan.hayward@arm.com>
17469 David Sherwood <david.sherwood@arm.com>
17471 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
17472 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
17473 (vect_use_loop_mask_for_alignment_p): New function.
17474 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
17475 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
17476 niters_skip argument. Make sure that the first niters_skip elements
17477 of the first iteration are inactive.
17478 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
17479 Update call to vect_set_loop_masks_directly.
17480 (get_misalign_in_elems): New function, split out from...
17481 (vect_gen_prolog_loop_niters): ...here.
17482 (vect_update_init_of_dr): Take a code argument that specifies whether
17483 the adjustment should be added or subtracted.
17484 (vect_update_init_of_drs): Likewise.
17485 (vect_prepare_for_masked_peels): New function.
17486 (vect_do_peeling): Skip prologue peeling if we're using a mask
17487 instead. Update call to vect_update_inits_of_drs.
17488 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
17490 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
17491 alignment. Do not include the number of peeled iterations in
17492 the minimum threshold in that case.
17493 (vectorizable_induction): Adjust the start value down by
17494 LOOP_VINFO_MASK_SKIP_NITERS iterations.
17495 (vect_transform_loop): Call vect_prepare_for_masked_peels.
17496 Take the number of skipped iterations into account when calculating
17498 * tree-vect-stmts.c (vect_gen_while_not): New function.
17500 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17501 Alan Hayward <alan.hayward@arm.com>
17502 David Sherwood <david.sherwood@arm.com>
17504 * doc/sourcebuild.texi (vect_fully_masked): Document.
17505 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
17506 default value to 0.
17507 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
17509 (vect_analyze_loop_2): ...here. Don't check the vectorization
17510 factor against the number of loop iterations if the loop is
17513 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17514 Alan Hayward <alan.hayward@arm.com>
17515 David Sherwood <david.sherwood@arm.com>
17517 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
17518 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
17519 (dump_groups): Update accordingly.
17520 (iv_use::mem_type): New member variable.
17521 (address_p): New function.
17522 (record_use): Add a mem_type argument and initialize the new
17524 (record_group_use): Add a mem_type argument. Use address_p.
17525 Remove obsolete null checks of base_object. Update call to record_use.
17526 (find_interesting_uses_op): Update call to record_group_use.
17527 (find_interesting_uses_cond): Likewise.
17528 (find_interesting_uses_address): Likewise.
17529 (get_mem_type_for_internal_fn): New function.
17530 (find_address_like_use): Likewise.
17531 (find_interesting_uses_stmt): Try find_address_like_use before
17532 calling find_interesting_uses_op.
17533 (addr_offset_valid_p): Use the iv mem_type field as the type
17534 of the addressed memory.
17535 (add_autoinc_candidates): Likewise.
17536 (get_address_cost): Likewise.
17537 (split_small_address_groups_p): Use address_p.
17538 (split_address_groups): Likewise.
17539 (add_iv_candidate_for_use): Likewise.
17540 (autoinc_possible_for_pair): Likewise.
17541 (rewrite_groups): Likewise.
17542 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
17543 (determine_group_iv_cost): Update after split of USE_ADDRESS.
17544 (get_alias_ptr_type_for_ptr_address): New function.
17545 (rewrite_use_address): Rewrite address uses in calls that were
17546 identified by find_address_like_use.
17548 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17549 Alan Hayward <alan.hayward@arm.com>
17550 David Sherwood <david.sherwood@arm.com>
17552 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
17554 * gimple-expr.h (is_gimple_addressable: Likewise.
17555 * gimple-expr.c (is_gimple_address): Likewise.
17556 * internal-fn.c (expand_call_mem_ref): New function.
17557 (expand_mask_load_optab_fn): Use it.
17558 (expand_mask_store_optab_fn): Likewise.
17560 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17561 Alan Hayward <alan.hayward@arm.com>
17562 David Sherwood <david.sherwood@arm.com>
17564 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
17565 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
17566 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
17567 (cond_umax@var{mode}): Document.
17568 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
17569 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
17570 (cond_umin_optab, cond_umax_optab): New optabs.
17571 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
17572 (COND_IOR, COND_XOR): New internal functions.
17573 * internal-fn.h (get_conditional_internal_fn): Declare.
17574 * internal-fn.c (cond_binary_direct): New macro.
17575 (expand_cond_binary_optab_fn): Likewise.
17576 (direct_cond_binary_optab_supported_p): Likewise.
17577 (get_conditional_internal_fn): New function.
17578 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
17579 Cope with reduction statements that are vectorized as calls rather
17581 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
17582 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
17583 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
17584 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
17585 (UNSPEC_COND_EOR): New unspecs.
17586 (optab): Add mappings for them.
17587 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
17588 (sve_int_op, sve_fp_op): New int attributes.
17590 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17591 Alan Hayward <alan.hayward@arm.com>
17592 David Sherwood <david.sherwood@arm.com>
17594 * optabs.def (while_ult_optab): New optab.
17595 * doc/md.texi (while_ult@var{m}@var{n}): Document.
17596 * internal-fn.def (WHILE_ULT): New internal function.
17597 * internal-fn.h (direct_internal_fn_supported_p): New override
17598 that takes two types as argument.
17599 * internal-fn.c (while_direct): New macro.
17600 (expand_while_optab_fn): New function.
17601 (convert_optab_supported_p): Likewise.
17602 (direct_while_optab_supported_p): New macro.
17603 * wide-int.h (wi::udiv_ceil): New function.
17604 * tree-vectorizer.h (rgroup_masks): New structure.
17605 (vec_loop_masks): New typedef.
17606 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
17607 and fully_masked_p.
17608 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
17609 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
17610 (vect_max_vf): New function.
17611 (slpeel_make_loop_iterate_ntimes): Delete.
17612 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
17613 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
17614 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
17615 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
17616 internal-fn.h, stor-layout.h and optabs-query.h.
17617 (vect_set_loop_mask): New function.
17618 (add_preheader_seq): Likewise.
17619 (add_header_seq): Likewise.
17620 (interleave_supported_p): Likewise.
17621 (vect_maybe_permute_loop_masks): Likewise.
17622 (vect_set_loop_masks_directly): Likewise.
17623 (vect_set_loop_condition_masked): Likewise.
17624 (vect_set_loop_condition_unmasked): New function, split out from
17625 slpeel_make_loop_iterate_ntimes.
17626 (slpeel_make_loop_iterate_ntimes): Rename to..
17627 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
17628 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
17629 (vect_do_peeling): Update call accordingly.
17630 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
17632 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
17633 mask_compare_type, can_fully_mask_p and fully_masked_p.
17634 (release_vec_loop_masks): New function.
17635 (_loop_vec_info): Use it to free the loop masks.
17636 (can_produce_all_loop_masks_p): New function.
17637 (vect_get_max_nscalars_per_iter): Likewise.
17638 (vect_verify_full_masking): Likewise.
17639 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
17640 retries, and free the mask rgroups before retrying. Check loop-wide
17641 reasons for disallowing fully-masked loops. Make the final decision
17642 about whether use a fully-masked loop or not.
17643 (vect_estimate_min_profitable_iters): Do not assume that peeling
17644 for the number of iterations will be needed for fully-masked loops.
17645 (vectorizable_reduction): Disable fully-masked loops.
17646 (vectorizable_live_operation): Likewise.
17647 (vect_halve_mask_nunits): New function.
17648 (vect_double_mask_nunits): Likewise.
17649 (vect_record_loop_mask): Likewise.
17650 (vect_get_loop_mask): Likewise.
17651 (vect_transform_loop): Handle the case in which the final loop
17652 iteration might handle a partial vector. Call vect_set_loop_condition
17653 instead of slpeel_make_loop_iterate_ntimes.
17654 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
17655 (check_load_store_masking): New function.
17656 (prepare_load_store_mask): Likewise.
17657 (vectorizable_store): Handle fully-masked loops.
17658 (vectorizable_load): Likewise.
17659 (supportable_widening_operation): Use vect_halve_mask_nunits for
17661 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
17662 (vect_gen_while): New function.
17663 * config/aarch64/aarch64.md (umax<mode>3): New expander.
17664 (aarch64_uqdec<mode>): New insn.
17666 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17667 Alan Hayward <alan.hayward@arm.com>
17668 David Sherwood <david.sherwood@arm.com>
17670 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
17671 (reduc_xor_scal_optab): New optabs.
17672 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
17673 (reduc_xor_scal_@var{m}): Document.
17674 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
17675 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
17676 internal functions.
17677 * fold-const-call.c (fold_const_call): Handle them.
17678 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
17679 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
17680 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
17681 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
17682 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
17683 (UNSPEC_XORV): New unspecs.
17684 (optab): Add entries for them.
17685 (BITWISEV): New int iterator.
17686 (bit_reduc_op): New int attributes.
17688 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17689 Alan Hayward <alan.hayward@arm.com>
17690 David Sherwood <david.sherwood@arm.com>
17692 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
17693 * internal-fn.def (VEC_SHL_INSERT): New internal function.
17694 * optabs.def (vec_shl_insert_optab): New optab.
17695 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
17696 (duplicate_and_interleave): Likewise.
17697 * tree-vect-loop.c: Include internal-fn.h.
17698 (neutral_op_for_slp_reduction): New function, split out from
17699 get_initial_defs_for_reduction.
17700 (get_initial_def_for_reduction): Handle option 2 for variable-length
17701 vectors by loading the neutral value into a vector and then shifting
17702 the initial value into element 0.
17703 (get_initial_defs_for_reduction): Replace the code argument with
17704 the neutral value calculated by neutral_op_for_slp_reduction.
17705 Use gimple_build_vector for constant-length vectors.
17706 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
17707 but the first group_size elements have a neutral value.
17708 Use duplicate_and_interleave otherwise.
17709 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
17710 Update call to get_initial_defs_for_reduction. Handle SLP
17711 reductions for variable-length vectors by creating one vector
17712 result for each scalar result, with the elements associated
17713 with other scalar results stubbed out with the neutral value.
17714 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
17715 Require IFN_VEC_SHL_INSERT for double reductions on
17716 variable-length vectors, or SLP reductions that have
17717 a neutral value. Require can_duplicate_and_interleave_p
17718 support for variable-length unchained SLP reductions if there
17719 is no neutral value, such as for MIN/MAX reductions. Also require
17720 the number of vector elements to be a multiple of the number of
17721 SLP statements when doing variable-length unchained SLP reductions.
17722 Update call to vect_create_epilog_for_reduction.
17723 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
17724 and remove initial values.
17725 (duplicate_and_interleave): Make public.
17726 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
17727 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
17729 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17730 Alan Hayward <alan.hayward@arm.com>
17731 David Sherwood <david.sherwood@arm.com>
17733 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
17734 (can_duplicate_and_interleave_p): New function.
17735 (vect_get_and_check_slp_defs): Take the vector of statements
17736 rather than just the current one. Remove excess parentheses.
17737 Restriction rejectinon of vect_constant_def and vect_external_def
17738 for variable-length vectors to boolean types, or types for which
17739 can_duplicate_and_interleave_p is false.
17740 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
17741 (duplicate_and_interleave): New function.
17742 (vect_get_constant_vectors): Use gimple_build_vector for
17743 constant-length vectors and suitable variable-length constant
17744 vectors. Use duplicate_and_interleave for other variable-length
17745 vectors. Don't defer the update when inserting new statements.
17747 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17748 Alan Hayward <alan.hayward@arm.com>
17749 David Sherwood <david.sherwood@arm.com>
17751 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
17752 min_profitable_iters doesn't go negative.
17754 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17755 Alan Hayward <alan.hayward@arm.com>
17756 David Sherwood <david.sherwood@arm.com>
17758 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
17759 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
17760 * optabs.def (vec_mask_load_lanes_optab): New optab.
17761 (vec_mask_store_lanes_optab): Likewise.
17762 * internal-fn.def (MASK_LOAD_LANES): New internal function.
17763 (MASK_STORE_LANES): Likewise.
17764 * internal-fn.c (mask_load_lanes_direct): New macro.
17765 (mask_store_lanes_direct): Likewise.
17766 (expand_mask_load_optab_fn): Handle masked operations.
17767 (expand_mask_load_lanes_optab_fn): New macro.
17768 (expand_mask_store_optab_fn): Handle masked operations.
17769 (expand_mask_store_lanes_optab_fn): New macro.
17770 (direct_mask_load_lanes_optab_supported_p): Likewise.
17771 (direct_mask_store_lanes_optab_supported_p): Likewise.
17772 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
17774 (vect_load_lanes_supported): Likewise.
17775 * tree-vect-data-refs.c (strip_conversion): New function.
17776 (can_group_stmts_p): Likewise.
17777 (vect_analyze_data_ref_accesses): Use it instead of checking
17778 for a pair of assignments.
17779 (vect_store_lanes_supported): Take a masked_p parameter.
17780 (vect_load_lanes_supported): Likewise.
17781 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
17782 vect_store_lanes_supported and vect_load_lanes_supported.
17783 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
17784 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
17785 parameter. Don't allow gaps for masked accesses.
17786 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
17787 and vect_load_lanes_supported.
17788 (get_load_store_type): Take a masked_p parameter and update
17789 call to get_group_load_store_type.
17790 (vectorizable_store): Update call to get_load_store_type.
17791 Handle IFN_MASK_STORE_LANES.
17792 (vectorizable_load): Update call to get_load_store_type.
17793 Handle IFN_MASK_LOAD_LANES.
17795 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17796 Alan Hayward <alan.hayward@arm.com>
17797 David Sherwood <david.sherwood@arm.com>
17799 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
17801 * config/aarch64/aarch64-protos.h
17802 (aarch64_sve_struct_memory_operand_p): Declare.
17803 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
17804 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
17805 (VPRED, vpred): Handle SVE structure modes.
17806 * config/aarch64/constraints.md (Utx): New constraint.
17807 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
17808 (aarch64_sve_struct_nonimmediate_operand): New predicates.
17809 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
17810 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
17811 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
17812 structure modes. Split into pieces after RA.
17813 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
17814 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
17816 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
17817 SVE structure modes.
17818 (aarch64_classify_address): Likewise.
17819 (sizetochar): Move earlier in file.
17820 (aarch64_print_operand): Handle SVE register lists.
17821 (aarch64_array_mode): New function.
17822 (aarch64_sve_struct_memory_operand_p): Likewise.
17823 (TARGET_ARRAY_MODE): Redefine.
17825 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17826 Alan Hayward <alan.hayward@arm.com>
17827 David Sherwood <david.sherwood@arm.com>
17829 * target.def (array_mode): New target hook.
17830 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
17831 * doc/tm.texi: Regenerate.
17832 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
17833 * hooks.c (hook_optmode_mode_uhwi_none): New function.
17834 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
17835 targetm.array_mode.
17836 * stor-layout.c (mode_for_array): Likewise. Support polynomial
17839 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17840 Alan Hayward <alan.hayward@arm.com>
17841 David Sherwood <david.sherwood@arm.com>
17843 * fold-const.c (fold_binary_loc): Check the argument types
17844 rather than the result type when testing for a vector operation.
17846 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17848 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
17849 * doc/tm.texi: Regenerate.
17851 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17852 Alan Hayward <alan.hayward@arm.com>
17853 David Sherwood <david.sherwood@arm.com>
17855 * doc/invoke.texi (-msve-vector-bits=): Document new option.
17856 (sve): Document new AArch64 extension.
17857 * doc/md.texi (w): Extend the description of the AArch64
17858 constraint to include SVE vectors.
17859 (Upl, Upa): Document new AArch64 predicate constraints.
17860 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
17862 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
17863 (msve-vector-bits=): New option.
17864 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
17865 SVE when these are disabled.
17866 (sve): New extension.
17867 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
17868 modes. Adjust their number of units based on aarch64_sve_vg.
17869 (MAX_BITSIZE_MODE_ANY_MODE): Define.
17870 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
17871 aarch64_addr_query_type.
17872 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
17873 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
17874 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
17875 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
17876 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
17877 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
17878 (aarch64_simd_imm_zero_p): Delete.
17879 (aarch64_check_zero_based_sve_index_immediate): Declare.
17880 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
17881 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
17882 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
17883 (aarch64_sve_float_mul_immediate_p): Likewise.
17884 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
17885 rather than an rtx.
17886 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
17887 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
17888 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
17889 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
17890 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
17891 (aarch64_regmode_natural_size): Likewise.
17892 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
17893 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
17895 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
17896 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
17897 for VG and the SVE predicate registers.
17898 (V_ALIASES): Add a "z"-prefixed alias.
17899 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
17900 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
17901 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
17902 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
17903 (REG_CLASS_NAMES): Add entries for them.
17904 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
17905 and the predicate registers.
17906 (aarch64_sve_vg): Declare.
17907 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
17908 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
17909 (REGMODE_NATURAL_SIZE): Define.
17910 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
17912 * config/aarch64/aarch64.c: Include cfgrtl.h.
17913 (simd_immediate_info): Add a constructor for series vectors,
17914 and an associated step field.
17915 (aarch64_sve_vg): New variable.
17916 (aarch64_dbx_register_number): Handle VG and the predicate registers.
17917 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
17918 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
17919 (VEC_ANY_DATA, VEC_STRUCT): New constants.
17920 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
17921 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
17922 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
17923 (aarch64_get_mask_mode): New functions.
17924 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
17925 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
17926 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
17927 predicate modes and predicate registers. Explicitly restrict
17928 GPRs to modes of 16 bytes or smaller. Only allow FP registers
17929 to store a vector mode if it is recognized by
17930 aarch64_classify_vector_mode.
17931 (aarch64_regmode_natural_size): New function.
17932 (aarch64_hard_regno_caller_save_mode): Return the original mode
17934 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
17935 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
17936 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
17937 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
17939 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
17940 does not overlap dest if the function is frame-related. Handle
17942 (aarch64_split_add_offset): New function.
17943 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
17944 them aarch64_add_offset.
17945 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
17946 and update call to aarch64_sub_sp.
17947 (aarch64_add_cfa_expression): New function.
17948 (aarch64_expand_prologue): Pass extra temporary registers to the
17949 functions above. Handle the case in which we need to emit new
17950 DW_CFA_expressions for registers that were originally saved
17951 relative to the stack pointer, but now have to be expressed
17952 relative to the frame pointer.
17953 (aarch64_output_mi_thunk): Pass extra temporary registers to the
17955 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
17956 IP0 and IP1 values for SVE frames.
17957 (aarch64_expand_vec_series): New function.
17958 (aarch64_expand_sve_widened_duplicate): Likewise.
17959 (aarch64_expand_sve_const_vector): Likewise.
17960 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
17961 Handle SVE constants. Use emit_move_insn to move a force_const_mem
17962 into the register, rather than emitting a SET directly.
17963 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
17964 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
17965 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
17966 (offset_9bit_signed_scaled_p): New functions.
17967 (aarch64_replicate_bitmask_imm): New function.
17968 (aarch64_bitmask_imm): Use it.
17969 (aarch64_cannot_force_const_mem): Reject expressions involving
17970 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
17971 (aarch64_classify_index): Handle SVE indices, by requiring
17972 a plain register index with a scale that matches the element size.
17973 (aarch64_classify_address): Handle SVE addresses. Assert that
17974 the mode of the address is VOIDmode or an integer mode.
17975 Update call to aarch64_classify_symbol.
17976 (aarch64_classify_symbolic_expression): Update call to
17977 aarch64_classify_symbol.
17978 (aarch64_const_vec_all_in_range_p): New function.
17979 (aarch64_print_vector_float_operand): Likewise.
17980 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
17981 "vN" for FP registers with SVE modes. Handle (const ...) vectors
17982 and the FP immediates 1.0 and 0.5.
17983 (aarch64_print_address_internal): Handle SVE addresses.
17984 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
17985 (aarch64_regno_regclass): Handle predicate registers.
17986 (aarch64_secondary_reload): Handle big-endian reloads of SVE
17988 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
17989 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
17990 (aarch64_convert_sve_vector_bits): New function.
17991 (aarch64_override_options): Use it to handle -msve-vector-bits=.
17992 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
17993 rather than an rtx.
17994 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
17995 Handle SVE vector and predicate modes. Accept VL-based constants
17996 that need only one temporary register, and VL offsets that require
17997 no temporary registers.
17998 (aarch64_conditional_register_usage): Mark the predicate registers
17999 as fixed if SVE isn't available.
18000 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
18001 Return true for SVE vector and predicate modes.
18002 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
18003 rather than an unsigned int. Handle SVE modes.
18004 (aarch64_preferred_simd_mode): Update call accordingly. Handle
18006 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
18008 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
18009 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
18010 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
18011 (aarch64_sve_float_mul_immediate_p): New functions.
18012 (aarch64_sve_valid_immediate): New function.
18013 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
18014 Explicitly reject structure modes. Check for INDEX constants.
18015 Handle PTRUE and PFALSE constants.
18016 (aarch64_check_zero_based_sve_index_immediate): New function.
18017 (aarch64_simd_imm_zero_p): Delete.
18018 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
18019 vector modes. Accept constants in the range of CNT[BHWD].
18020 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
18021 ask for an Advanced SIMD mode.
18022 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
18023 (aarch64_simd_vector_alignment): Handle SVE predicates.
18024 (aarch64_vectorize_preferred_vector_alignment): New function.
18025 (aarch64_simd_vector_alignment_reachable): Use it instead of
18027 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
18028 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
18030 (MAX_VECT_LEN): Delete.
18031 (expand_vec_perm_d): Add a vec_flags field.
18032 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
18033 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
18034 (aarch64_evpc_ext): Don't apply a big-endian lane correction
18036 (aarch64_evpc_rev): Rename to...
18037 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
18038 (aarch64_evpc_rev_global): New function.
18039 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
18040 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
18042 (aarch64_evpc_sve_tbl): New function.
18043 (aarch64_expand_vec_perm_const_1): Update after rename of
18044 aarch64_evpc_rev. Handle SVE permutes too, trying
18045 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
18046 than aarch64_evpc_tbl.
18047 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
18048 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
18049 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
18050 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
18051 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
18052 (aarch64_expand_sve_vcond): New functions.
18053 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
18054 of aarch64_vector_mode_p.
18055 (aarch64_dwarf_poly_indeterminate_value): New function.
18056 (aarch64_compute_pressure_classes): Likewise.
18057 (aarch64_can_change_mode_class): Likewise.
18058 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
18059 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
18060 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
18061 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
18062 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
18063 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
18064 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
18065 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
18067 (Dn, Dl, Dr): Accept const as well as const_vector.
18068 (Dz): Likewise. Compare against CONST0_RTX.
18069 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
18070 of "vector" where appropriate.
18071 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
18072 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
18073 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
18074 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
18075 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
18076 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
18077 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
18078 (v_int_equiv): Extend to SVE modes.
18079 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
18081 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
18082 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
18083 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
18084 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
18085 (SVE_COND_FP_CMP): New int iterators.
18086 (perm_hilo): Handle the new unpack unspecs.
18087 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
18089 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
18090 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
18091 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
18092 (aarch64_equality_operator, aarch64_constant_vector_operand)
18093 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
18094 (aarch64_sve_nonimmediate_operand): Likewise.
18095 (aarch64_sve_general_operand): Likewise.
18096 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
18097 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
18098 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
18099 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
18100 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
18101 (aarch64_sve_float_arith_immediate): Likewise.
18102 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
18103 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
18104 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
18105 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
18106 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
18107 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
18108 (aarch64_sve_float_arith_operand): Likewise.
18109 (aarch64_sve_float_arith_with_sub_operand): Likewise.
18110 (aarch64_sve_float_mul_operand): Likewise.
18111 (aarch64_sve_vec_perm_operand): Likewise.
18112 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
18113 (aarch64_mov_operand): Accept const_poly_int and const_vector.
18114 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
18115 as well as const_vector.
18116 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
18117 in file. Use CONST0_RTX and CONSTM1_RTX.
18118 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
18119 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
18120 Use aarch64_simd_imm_zero.
18121 * config/aarch64/aarch64-sve.md: New file.
18122 * config/aarch64/aarch64.md: Include it.
18123 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
18124 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
18125 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
18126 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
18127 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
18128 (sve): New attribute.
18129 (enabled): Disable instructions with the sve attribute unless
18131 (movqi, movhi): Pass CONST_POLY_INT operaneds through
18132 aarch64_expand_mov_immediate.
18133 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
18134 CNT[BHSD] immediates.
18135 (movti): Split CONST_POLY_INT moves into two halves.
18136 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
18137 Split additions that need a temporary here if the destination
18138 is the stack pointer.
18139 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
18140 (*add<mode>3_poly_1): New instruction.
18141 (set_clobber_cc): New expander.
18143 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18145 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
18146 parameter and use it instead of GET_MODE_SIZE (innermode). Use
18147 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
18148 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
18149 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
18150 Change innermode from fixed_mode_size to machine_mode.
18151 (simplify_subreg): Update call accordingly. Handle a constant-sized
18152 subreg of a variable-length CONST_VECTOR.
18154 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18155 Alan Hayward <alan.hayward@arm.com>
18156 David Sherwood <david.sherwood@arm.com>
18158 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
18159 (add_offset_to_base): New function, split out from...
18160 (create_mem_ref): ...here. When handling a scale other than 1,
18161 check first whether the address is valid without the offset.
18162 Add it into the base if so, leaving the index and scale as-is.
18164 2018-01-12 Jakub Jelinek <jakub@redhat.com>
18167 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
18168 fold_for_warn before checking if arg2 is INTEGER_CST.
18170 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
18172 * config/rs6000/predicates.md (load_multiple_operation): Delete.
18173 (store_multiple_operation): Delete.
18174 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
18175 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
18176 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
18177 guarded by TARGET_STRING.
18178 (rs6000_output_load_multiple): Delete.
18179 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
18180 OPTION_MASK_STRING / TARGET_STRING handling.
18181 (print_operand) <'N', 'O'>: Add comment that these are unused now.
18182 (const rs6000_opt_masks) <"string">: Change mask to 0.
18183 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
18184 (MASK_STRING): Delete.
18185 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
18187 (load_multiple): Delete.
18194 (store_multiple): Delete.
18201 (movmemsi_8reg): Delete.
18202 (corresponding unnamed define_insn): Delete.
18203 (movmemsi_6reg): Delete.
18204 (corresponding unnamed define_insn): Delete.
18205 (movmemsi_4reg): Delete.
18206 (corresponding unnamed define_insn): Delete.
18207 (movmemsi_2reg): Delete.
18208 (corresponding unnamed define_insn): Delete.
18209 (movmemsi_1reg): Delete.
18210 (corresponding unnamed define_insn): Delete.
18211 * config/rs6000/rs6000.opt (mno-string): New.
18212 (mstring): Replace by deprecation warning stub.
18213 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
18215 2018-01-12 Jakub Jelinek <jakub@redhat.com>
18217 * regrename.c (regrename_do_replace): If replacing the same
18218 reg multiple times, try to reuse last created gen_raw_REG.
18221 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
18222 main to workaround a bug in GDB.
18224 2018-01-12 Tom de Vries <tom@codesourcery.com>
18227 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
18229 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
18231 PR rtl-optimization/80481
18232 * ira-color.c (get_cap_member): New function.
18233 (allocnos_conflict_by_live_ranges_p): Use it.
18234 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
18235 (setup_slot_coalesced_allocno_live_ranges): Ditto.
18237 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
18240 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
18241 (*saddl_se_1): Ditto.
18242 (*ssubsi_1): Ditto.
18243 (*ssubl_se_1): Ditto.
18245 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
18247 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
18248 rather than wi::to_widest for DR_INITs.
18249 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
18250 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
18251 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
18253 (vect_analyze_group_access_1): Note that here.
18255 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
18257 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
18258 polynomial type sizes.
18260 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
18262 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
18263 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
18264 (gimple_add_tmp_var): Likewise.
18266 2018-01-12 Martin Liska <mliska@suse.cz>
18268 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
18269 (gimple_alloc_sizes): Likewise.
18270 (dump_gimple_statistics): Use PRIu64 in printf format.
18271 * gimple.h: Change uint64_t to int.
18273 2018-01-12 Martin Liska <mliska@suse.cz>
18275 * tree-core.h: Use uint64_t instead of int.
18276 * tree.c (tree_node_counts): Likewise.
18277 (tree_node_sizes): Likewise.
18278 (dump_tree_statistics): Use PRIu64 in printf format.
18280 2018-01-12 Martin Liska <mliska@suse.cz>
18282 * Makefile.in: As qsort_chk is implemented in vec.c, add
18283 vec.o to linkage of gencfn-macros.
18284 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
18285 passing the info to record_node_allocation_statistics.
18286 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
18288 * ggc-common.c (struct ggc_usage): Add operator== and use
18289 it in operator< and compare function.
18290 * mem-stats.h (struct mem_usage): Likewise.
18291 * vec.c (struct vec_usage): Remove operator< and compare
18292 function. Can be simply inherited.
18294 2018-01-12 Martin Jambor <mjambor@suse.cz>
18297 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
18298 * tree-ssa-math-opts.c: Include domwalk.h.
18299 (convert_mult_to_fma_1): New function.
18300 (fma_transformation_info): New type.
18301 (fma_deferring_state): Likewise.
18302 (cancel_fma_deferring): New function.
18303 (result_of_phi): Likewise.
18304 (last_fma_candidate_feeds_initial_phi): Likewise.
18305 (convert_mult_to_fma): Added deferring logic, split actual
18306 transformation to convert_mult_to_fma_1.
18307 (math_opts_dom_walker): New type.
18308 (math_opts_dom_walker::after_dom_children): New method, body moved
18309 here from pass_optimize_widening_mul::execute, added deferring logic
18311 (pass_optimize_widening_mul::execute): Moved most of code to
18312 math_opts_dom_walker::after_dom_children.
18313 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
18314 * config/i386/i386.c (ix86_option_override_internal): Added
18315 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
18317 2018-01-12 Richard Biener <rguenther@suse.de>
18320 * dwarf2out.c (gen_variable_die): Do not reset old_die for
18321 inline instance vars.
18323 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
18326 * config/rx/rx.c (rx_is_restricted_memory_address):
18327 Handle SUBREG case.
18329 2018-01-12 Richard Biener <rguenther@suse.de>
18331 PR tree-optimization/80846
18332 * target.def (split_reduction): New target hook.
18333 * targhooks.c (default_split_reduction): New function.
18334 * targhooks.h (default_split_reduction): Declare.
18335 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
18336 target requests first reduce vectors by combining low and high
18338 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
18339 (get_vectype_for_scalar_type_and_size): Export.
18340 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
18341 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
18342 * doc/tm.texi: Regenerate.
18343 * config/i386/i386.c (ix86_split_reduction): Implement
18344 TARGET_VECTORIZE_SPLIT_REDUCTION.
18346 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
18349 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
18350 in PIC mode except for TARGET_VXWORKS_RTP.
18351 * config/sparc/sparc.c: Include cfgrtl.h.
18352 (TARGET_INIT_PIC_REG): Define.
18353 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
18354 (sparc_pic_register_p): New predicate.
18355 (sparc_legitimate_address_p): Use it.
18356 (sparc_legitimize_pic_address): Likewise.
18357 (sparc_delegitimize_address): Likewise.
18358 (sparc_mode_dependent_address_p): Likewise.
18359 (gen_load_pcrel_sym): Remove 4th parameter.
18360 (load_got_register): Adjust call to above. Remove obsolete stuff.
18361 (sparc_expand_prologue): Do not call load_got_register here.
18362 (sparc_flat_expand_prologue): Likewise.
18363 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
18364 (sparc_use_pseudo_pic_reg): New function.
18365 (sparc_init_pic_reg): Likewise.
18366 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
18367 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
18369 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
18371 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
18372 Add item for branch_cost.
18374 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
18376 PR rtl-optimization/83565
18377 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
18378 not extend the result to a larger mode for rotate operations.
18379 (num_sign_bit_copies1): Likewise.
18381 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18384 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
18386 Use values-Xc.o for -pedantic.
18387 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
18389 2018-01-12 Martin Liska <mliska@suse.cz>
18392 * ipa-devirt.c (final_warning_record::grow_type_warnings):
18394 (possible_polymorphic_call_targets): Use it.
18395 (ipa_devirt): Likewise.
18397 2018-01-12 Martin Liska <mliska@suse.cz>
18399 * profile-count.h (enum profile_quality): Use 0 as invalid
18400 enum value of profile_quality.
18402 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
18404 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
18405 -mext-string options.
18407 2018-01-12 Richard Biener <rguenther@suse.de>
18409 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
18410 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
18411 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
18413 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
18415 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
18417 * configure.ac (--with-long-double-format): Add support for the
18418 configuration option to change the default long double format on
18420 * config.gcc (powerpc*-linux*-*): Likewise.
18421 * configure: Regenerate.
18422 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
18423 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
18424 used without modification.
18426 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18428 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
18429 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
18430 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
18431 MISC_BUILTIN_SPEC_BARRIER.
18432 (rs6000_init_builtins): Likewise.
18433 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
18435 (speculation_barrier): New define_insn.
18436 * doc/extend.texi: Document __builtin_speculation_barrier.
18438 2018-01-11 Jakub Jelinek <jakub@redhat.com>
18441 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
18442 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
18443 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
18445 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
18446 integral modes instead of "ss" and "sd".
18447 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
18448 vectors with 32-bit and 64-bit elements.
18449 (vecdupssescalarmodesuffix): New mode attribute.
18450 (vec_dup<mode>): Use it.
18452 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
18455 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
18456 frame if argument is passed on stack.
18458 2018-01-11 Jakub Jelinek <jakub@redhat.com>
18461 * ree.c (combine_reaching_defs): Optimize also
18462 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
18463 reg2=any_extend(exp); reg1=reg2;, formatting fix.
18465 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
18467 PR middle-end/83189
18468 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
18470 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
18472 PR middle-end/83718
18473 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
18474 after they are computed.
18476 2018-01-11 Bin Cheng <bin.cheng@arm.com>
18478 PR tree-optimization/83695
18479 * gimple-loop-linterchange.cc
18480 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
18481 reset cached scev information after interchange.
18482 (pass_linterchange::execute): Remove call to scev_reset_htab.
18484 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18486 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
18487 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
18488 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
18489 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
18490 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
18491 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
18492 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
18493 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
18494 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
18495 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
18496 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
18497 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
18498 (V_lane_reg): Likewise.
18499 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
18501 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
18502 (vfmal_lane_low<mode>_intrinsic,
18503 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
18504 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
18505 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
18506 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
18507 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
18508 vfmsl_lane_high<mode>_intrinsic): New define_insns.
18510 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18512 * config/arm/arm-cpus.in (fp16fml): New feature.
18513 (ALL_SIMD): Add fp16fml.
18514 (armv8.2-a): Add fp16fml as an option.
18515 (armv8.3-a): Likewise.
18516 (armv8.4-a): Add fp16fml as part of fp16.
18517 * config/arm/arm.h (TARGET_FP16FML): Define.
18518 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
18520 * config/arm/arm-modes.def (V2HF): Define.
18521 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
18522 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
18523 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
18524 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
18525 vfmsl_low, vfmsl_high): New set of builtins.
18526 * config/arm/iterators.md (PLUSMINUS): New code iterator.
18527 (vfml_op): New code attribute.
18528 (VFMLHALVES): New int iterator.
18529 (VFML, VFMLSEL): New mode attributes.
18530 (V_reg): Define mapping for V2HF.
18531 (V_hi, V_lo): New mode attributes.
18532 (VF_constraint): Likewise.
18533 (vfml_half, vfml_half_selector): New int attributes.
18534 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
18536 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
18537 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
18539 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
18540 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
18541 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
18542 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
18544 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
18545 Document new effective target and option set.
18547 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18549 * config/arm/arm-cpus.in (armv8_4): New feature.
18550 (ARMv8_4a): New fgroup.
18551 (armv8.4-a): New arch.
18552 * config/arm/arm-tables.opt: Regenerate.
18553 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
18554 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
18555 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
18556 Add matching rules for -march=armv8.4-a and extensions.
18557 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
18559 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
18562 * config/rx/rx.md (BW): New mode attribute.
18563 (sync_lock_test_and_setsi): Add mode suffix to insn output.
18565 2018-01-11 Richard Biener <rguenther@suse.de>
18567 PR tree-optimization/83435
18568 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
18569 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
18570 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
18572 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
18573 Alan Hayward <alan.hayward@arm.com>
18574 David Sherwood <david.sherwood@arm.com>
18576 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
18578 (aarch64_classify_address): Initialize it. Track polynomial offsets.
18579 (aarch64_print_address_internal): Use it to check for a zero offset.
18581 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
18582 Alan Hayward <alan.hayward@arm.com>
18583 David Sherwood <david.sherwood@arm.com>
18585 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
18586 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
18587 Return a poly_int64 rather than a HOST_WIDE_INT.
18588 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
18589 rather than a HOST_WIDE_INT.
18590 * config/aarch64/aarch64.h (aarch64_frame): Protect with
18591 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
18592 hard_fp_offset, frame_size, initial_adjust, callee_offset and
18593 final_offset from HOST_WIDE_INT to poly_int64.
18594 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
18595 to_constant when getting the number of units in an Advanced SIMD
18597 (aarch64_builtin_vectorized_function): Check for a constant number
18599 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
18601 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
18602 attribute instead of GET_MODE_NUNITS.
18603 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
18604 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
18605 GET_MODE_SIZE for fixed-size registers.
18606 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
18607 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
18608 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
18609 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
18610 (aarch64_print_operand, aarch64_print_address_internal)
18611 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
18612 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
18613 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
18614 Handle polynomial GET_MODE_SIZE.
18615 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
18616 wider than SImode without modification.
18617 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
18618 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
18619 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
18620 passing and returning SVE modes.
18621 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
18622 rather than GEN_INT.
18623 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
18624 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
18625 (aarch64_allocate_and_probe_stack_space): Likewise.
18626 (aarch64_layout_frame): Cope with polynomial offsets.
18627 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
18628 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
18629 polynomial offsets.
18630 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
18631 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
18632 poly_int64 rather than a HOST_WIDE_INT.
18633 (aarch64_get_separate_components, aarch64_process_components)
18634 (aarch64_expand_prologue, aarch64_expand_epilogue)
18635 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
18636 (aarch64_anchor_offset): New function, split out from...
18637 (aarch64_legitimize_address): ...here.
18638 (aarch64_builtin_vectorization_cost): Handle polynomial
18639 TYPE_VECTOR_SUBPARTS.
18640 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
18642 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
18643 number of elements from the PARALLEL rather than the mode.
18644 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
18645 rather than GET_MODE_BITSIZE.
18646 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
18647 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
18648 (aarch64_expand_vec_perm_const_1): Handle polynomial
18649 d->perm.length () and d->perm elements.
18650 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
18651 Apply to_constant to d->perm elements.
18652 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
18653 polynomial CONST_VECTOR_NUNITS.
18654 (aarch64_move_pointer): Take amount as a poly_int64 rather
18656 (aarch64_progress_pointer): Avoid temporary variable.
18657 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
18658 the mode attribute instead of GET_MODE.
18660 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
18661 Alan Hayward <alan.hayward@arm.com>
18662 David Sherwood <david.sherwood@arm.com>
18664 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
18665 x exists before using it.
18666 (aarch64_add_constant_internal): Rename to...
18667 (aarch64_add_offset_1): ...this. Replace regnum with separate
18668 src and dest rtxes. Handle the case in which they're different,
18669 including when the offset is zero. Replace scratchreg with an rtx.
18670 Use 2 additions if there is no spare register into which we can
18671 move a 16-bit constant.
18672 (aarch64_add_constant): Delete.
18673 (aarch64_add_offset): Replace reg with separate src and dest
18674 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
18675 Use aarch64_add_offset_1.
18676 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
18677 an rtx rather than an int. Take the delta as a poly_int64
18678 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
18679 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
18680 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
18681 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
18682 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
18683 and aarch64_add_sp.
18684 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
18685 aarch64_add_constant.
18687 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
18689 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
18690 Use scalar_float_mode.
18692 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
18694 * config/aarch64/aarch64-simd.md
18695 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
18696 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
18697 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
18698 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
18699 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
18700 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
18701 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
18702 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
18703 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
18704 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
18706 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18709 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
18710 targ_options->x_arm_arch_string is non NULL.
18712 2018-01-11 Tamar Christina <tamar.christina@arm.com>
18714 * config/aarch64/aarch64.h
18715 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
18717 2018-01-11 Sudakshina Das <sudi.das@arm.com>
18720 * expmed.c (emit_store_flag_force): Swap if const op0
18721 and change VOIDmode to mode of op0.
18723 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
18725 PR rtl-optimization/83761
18726 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
18727 than bytes to mode_for_size.
18729 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
18731 PR middle-end/83189
18732 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
18733 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
18736 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
18738 PR middle-end/83575
18739 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
18740 when in layout mode.
18741 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
18742 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
18745 2018-01-10 Michael Collison <michael.collison@arm.com>
18747 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
18748 * config/aarch64/aarch64-option-extension.def: Add
18749 AARCH64_OPT_EXTENSION of 'fp16fml'.
18750 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
18751 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
18752 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
18753 * config/aarch64/constraints.md (Ui7): New constraint.
18754 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
18755 (VFMLA_SEL_W): Ditto.
18758 (VFMLA16_LOW): New int iterator.
18759 (VFMLA16_HIGH): Ditto.
18760 (UNSPEC_FMLAL): New unspec.
18761 (UNSPEC_FMLSL): Ditto.
18762 (UNSPEC_FMLAL2): Ditto.
18763 (UNSPEC_FMLSL2): Ditto.
18764 (f16mac): New code attribute.
18765 * config/aarch64/aarch64-simd-builtins.def
18766 (aarch64_fmlal_lowv2sf): Ditto.
18767 (aarch64_fmlsl_lowv2sf): Ditto.
18768 (aarch64_fmlalq_lowv4sf): Ditto.
18769 (aarch64_fmlslq_lowv4sf): Ditto.
18770 (aarch64_fmlal_highv2sf): Ditto.
18771 (aarch64_fmlsl_highv2sf): Ditto.
18772 (aarch64_fmlalq_highv4sf): Ditto.
18773 (aarch64_fmlslq_highv4sf): Ditto.
18774 (aarch64_fmlal_lane_lowv2sf): Ditto.
18775 (aarch64_fmlsl_lane_lowv2sf): Ditto.
18776 (aarch64_fmlal_laneq_lowv2sf): Ditto.
18777 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
18778 (aarch64_fmlalq_lane_lowv4sf): Ditto.
18779 (aarch64_fmlsl_lane_lowv4sf): Ditto.
18780 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
18781 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
18782 (aarch64_fmlal_lane_highv2sf): Ditto.
18783 (aarch64_fmlsl_lane_highv2sf): Ditto.
18784 (aarch64_fmlal_laneq_highv2sf): Ditto.
18785 (aarch64_fmlsl_laneq_highv2sf): Ditto.
18786 (aarch64_fmlalq_lane_highv4sf): Ditto.
18787 (aarch64_fmlsl_lane_highv4sf): Ditto.
18788 (aarch64_fmlalq_laneq_highv4sf): Ditto.
18789 (aarch64_fmlsl_laneq_highv4sf): Ditto.
18790 * config/aarch64/aarch64-simd.md:
18791 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
18792 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
18793 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
18794 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
18795 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
18796 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
18797 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
18798 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
18799 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
18800 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
18801 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
18802 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
18803 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
18804 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
18805 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
18806 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
18807 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
18808 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
18809 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
18810 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
18811 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
18812 (vfmlsl_low_u32): Ditto.
18813 (vfmlalq_low_u32): Ditto.
18814 (vfmlslq_low_u32): Ditto.
18815 (vfmlal_high_u32): Ditto.
18816 (vfmlsl_high_u32): Ditto.
18817 (vfmlalq_high_u32): Ditto.
18818 (vfmlslq_high_u32): Ditto.
18819 (vfmlal_lane_low_u32): Ditto.
18820 (vfmlsl_lane_low_u32): Ditto.
18821 (vfmlal_laneq_low_u32): Ditto.
18822 (vfmlsl_laneq_low_u32): Ditto.
18823 (vfmlalq_lane_low_u32): Ditto.
18824 (vfmlslq_lane_low_u32): Ditto.
18825 (vfmlalq_laneq_low_u32): Ditto.
18826 (vfmlslq_laneq_low_u32): Ditto.
18827 (vfmlal_lane_high_u32): Ditto.
18828 (vfmlsl_lane_high_u32): Ditto.
18829 (vfmlal_laneq_high_u32): Ditto.
18830 (vfmlsl_laneq_high_u32): Ditto.
18831 (vfmlalq_lane_high_u32): Ditto.
18832 (vfmlslq_lane_high_u32): Ditto.
18833 (vfmlalq_laneq_high_u32): Ditto.
18834 (vfmlslq_laneq_high_u32): Ditto.
18835 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
18836 (AARCH64_FL_FOR_ARCH8_4): New.
18837 (AARCH64_ISA_F16FML): New ISA flag.
18838 (TARGET_F16FML): New feature flag for fp16fml.
18839 (doc/invoke.texi): Document new fp16fml option.
18841 2018-01-10 Michael Collison <michael.collison@arm.com>
18843 * config/aarch64/aarch64-builtins.c:
18844 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
18845 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
18846 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
18847 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
18848 (AARCH64_ISA_SHA3): New ISA flag.
18849 (TARGET_SHA3): New feature flag for sha3.
18850 * config/aarch64/iterators.md (sha512_op): New int attribute.
18851 (CRYPTO_SHA512): New int iterator.
18852 (UNSPEC_SHA512H): New unspec.
18853 (UNSPEC_SHA512H2): Ditto.
18854 (UNSPEC_SHA512SU0): Ditto.
18855 (UNSPEC_SHA512SU1): Ditto.
18856 * config/aarch64/aarch64-simd-builtins.def
18857 (aarch64_crypto_sha512hqv2di): New builtin.
18858 (aarch64_crypto_sha512h2qv2di): Ditto.
18859 (aarch64_crypto_sha512su0qv2di): Ditto.
18860 (aarch64_crypto_sha512su1qv2di): Ditto.
18861 (aarch64_eor3qv8hi): Ditto.
18862 (aarch64_rax1qv2di): Ditto.
18863 (aarch64_xarqv2di): Ditto.
18864 (aarch64_bcaxqv8hi): Ditto.
18865 * config/aarch64/aarch64-simd.md:
18866 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
18867 (aarch64_crypto_sha512su0qv2di): Ditto.
18868 (aarch64_crypto_sha512su1qv2di): Ditto.
18869 (aarch64_eor3qv8hi): Ditto.
18870 (aarch64_rax1qv2di): Ditto.
18871 (aarch64_xarqv2di): Ditto.
18872 (aarch64_bcaxqv8hi): Ditto.
18873 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
18874 (vsha512h2q_u64): Ditto.
18875 (vsha512su0q_u64): Ditto.
18876 (vsha512su1q_u64): Ditto.
18877 (veor3q_u16): Ditto.
18878 (vrax1q_u64): Ditto.
18879 (vxarq_u64): Ditto.
18880 (vbcaxq_u16): Ditto.
18881 * config/arm/types.md (crypto_sha512): New type attribute.
18882 (crypto_sha3): Ditto.
18883 (doc/invoke.texi): Document new sha3 option.
18885 2018-01-10 Michael Collison <michael.collison@arm.com>
18887 * config/aarch64/aarch64-builtins.c:
18888 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
18889 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
18890 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
18891 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
18892 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
18893 (AARCH64_ISA_SM4): New ISA flag.
18894 (TARGET_SM4): New feature flag for sm4.
18895 * config/aarch64/aarch64-simd-builtins.def
18896 (aarch64_sm3ss1qv4si): Ditto.
18897 (aarch64_sm3tt1aq4si): Ditto.
18898 (aarch64_sm3tt1bq4si): Ditto.
18899 (aarch64_sm3tt2aq4si): Ditto.
18900 (aarch64_sm3tt2bq4si): Ditto.
18901 (aarch64_sm3partw1qv4si): Ditto.
18902 (aarch64_sm3partw2qv4si): Ditto.
18903 (aarch64_sm4eqv4si): Ditto.
18904 (aarch64_sm4ekeyqv4si): Ditto.
18905 * config/aarch64/aarch64-simd.md:
18906 (aarch64_sm3ss1qv4si): Ditto.
18907 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
18908 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
18909 (aarch64_sm4eqv4si): Ditto.
18910 (aarch64_sm4ekeyqv4si): Ditto.
18911 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
18912 (sm3part_op): Ditto.
18913 (CRYPTO_SM3TT): Ditto.
18914 (CRYPTO_SM3PART): Ditto.
18915 (UNSPEC_SM3SS1): New unspec.
18916 (UNSPEC_SM3TT1A): Ditto.
18917 (UNSPEC_SM3TT1B): Ditto.
18918 (UNSPEC_SM3TT2A): Ditto.
18919 (UNSPEC_SM3TT2B): Ditto.
18920 (UNSPEC_SM3PARTW1): Ditto.
18921 (UNSPEC_SM3PARTW2): Ditto.
18922 (UNSPEC_SM4E): Ditto.
18923 (UNSPEC_SM4EKEY): Ditto.
18924 * config/aarch64/constraints.md (Ui2): New constraint.
18925 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
18926 * config/arm/types.md (crypto_sm3): New type attribute.
18927 (crypto_sm4): Ditto.
18928 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
18929 (vsm3tt1aq_u32): Ditto.
18930 (vsm3tt1bq_u32): Ditto.
18931 (vsm3tt2aq_u32): Ditto.
18932 (vsm3tt2bq_u32): Ditto.
18933 (vsm3partw1q_u32): Ditto.
18934 (vsm3partw2q_u32): Ditto.
18935 (vsm4eq_u32): Ditto.
18936 (vsm4ekeyq_u32): Ditto.
18937 (doc/invoke.texi): Document new sm4 option.
18939 2018-01-10 Michael Collison <michael.collison@arm.com>
18941 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
18942 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
18943 (AARCH64_FL_FOR_ARCH8_4): New.
18944 (AARCH64_FL_V8_4): New flag.
18945 (doc/invoke.texi): Document new armv8.4-a option.
18947 2018-01-10 Michael Collison <michael.collison@arm.com>
18949 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
18950 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
18951 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
18952 * config/aarch64/aarch64-option-extension.def: Add
18953 AARCH64_OPT_EXTENSION of 'sha2'.
18954 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
18955 (crypto): Disable sha2 and aes if crypto disabled.
18956 (crypto): Enable aes and sha2 if enabled.
18957 (simd): Disable sha2 and aes if simd disabled.
18958 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
18960 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
18961 (TARGET_SHA2): New feature flag for sha2.
18962 (TARGET_AES): New feature flag for aes.
18963 * config/aarch64/aarch64-simd.md:
18964 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
18965 conditional on TARGET_AES.
18966 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
18967 (aarch64_crypto_sha1hsi): Make pattern conditional
18969 (aarch64_crypto_sha1hv4si): Ditto.
18970 (aarch64_be_crypto_sha1hv4si): Ditto.
18971 (aarch64_crypto_sha1su1v4si): Ditto.
18972 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
18973 (aarch64_crypto_sha1su0v4si): Ditto.
18974 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
18975 (aarch64_crypto_sha256su0v4si): Ditto.
18976 (aarch64_crypto_sha256su1v4si): Ditto.
18977 (doc/invoke.texi): Document new aes and sha2 options.
18979 2018-01-10 Martin Sebor <msebor@redhat.com>
18981 PR tree-optimization/83781
18982 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
18985 2018-01-11 Martin Sebor <msebor@gmail.com>
18986 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18988 PR tree-optimization/83501
18989 PR tree-optimization/81703
18991 * tree-ssa-strlen.c (get_string_cst): Rename...
18992 (get_string_len): ...to this. Handle global constants.
18993 (handle_char_store): Adjust.
18995 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
18996 Jim Wilson <jimw@sifive.com>
18998 * config/riscv/riscv-protos.h (riscv_output_return): New.
18999 * config/riscv/riscv.c (struct machine_function): New naked_p field.
19000 (riscv_attribute_table, riscv_output_return),
19001 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
19002 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
19003 (riscv_compute_frame_info): Only compute frame->mask if not a naked
19005 (riscv_expand_prologue): Add early return for naked function.
19006 (riscv_expand_epilogue): Likewise.
19007 (riscv_function_ok_for_sibcall): Return false for naked function.
19008 (riscv_set_current_function): New.
19009 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
19010 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
19011 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
19012 * doc/extend.texi (RISC-V Function Attributes): New.
19014 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
19016 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
19017 check for 128-bit long double before checking TCmode.
19018 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
19019 128-bit long doubles before checking TFmode or TCmode.
19020 (FLOAT128_IBM_P): Likewise.
19022 2018-01-10 Martin Sebor <msebor@redhat.com>
19024 PR tree-optimization/83671
19025 * builtins.c (c_strlen): Unconditionally return zero for the empty
19027 Use -Warray-bounds for warnings.
19028 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
19029 for non-constant array indices with COMPONENT_REF, arrays of
19030 arrays, and pointers to arrays.
19031 (gimple_fold_builtin_strlen): Determine and set length range for
19032 non-constant character arrays.
19034 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
19036 PR middle-end/81897
19037 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
19040 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
19042 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
19044 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
19047 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
19048 VECTOR_MEM_ALTIVEC_OR_VSX_P.
19049 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
19050 indexed_or_indirect_operand predicate.
19051 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
19052 (*vsx_le_perm_load_v8hi): Likewise.
19053 (*vsx_le_perm_load_v16qi): Likewise.
19054 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
19055 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
19056 (*vsx_le_perm_store_v8hi): Likewise.
19057 (*vsx_le_perm_store_v16qi): Likewise.
19058 (eight unnamed splitters): Likewise.
19060 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
19062 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
19063 * config/rs6000/emmintrin.h: Likewise.
19064 * config/rs6000/mmintrin.h: Likewise.
19065 * config/rs6000/xmmintrin.h: Likewise.
19067 2018-01-10 David Malcolm <dmalcolm@redhat.com>
19070 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
19072 * tree.c (tree_nop_conversion): Return true for location wrapper
19074 (maybe_wrap_with_location): New function.
19075 (selftest::check_strip_nops): New function.
19076 (selftest::test_location_wrappers): New function.
19077 (selftest::tree_c_tests): Call it.
19078 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
19079 (maybe_wrap_with_location): New decl.
19080 (EXPR_LOCATION_WRAPPER_P): New macro.
19081 (location_wrapper_p): New inline function.
19082 (tree_strip_any_location_wrapper): New inline function.
19084 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
19087 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
19088 stack_realign_offset for the largest alignment of stack slot
19090 (ix86_find_max_used_stack_alignment): New function.
19091 (ix86_finalize_stack_frame_flags): Use it. Set
19092 max_used_stack_alignment if we don't realign stack.
19093 * config/i386/i386.h (machine_function): Add
19094 max_used_stack_alignment.
19096 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
19098 * config/arm/arm.opt (-mbranch-cost): New option.
19099 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
19102 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
19105 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
19106 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
19108 2018-01-10 Richard Biener <rguenther@suse.de>
19111 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
19112 early out so it also covers the case where we have a non-NULL
19115 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
19117 PR tree-optimization/83753
19118 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
19119 for non-strided grouped accesses if the number of elements is 1.
19121 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
19124 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
19125 * i386.h (TARGET_USE_GATHER): Define.
19126 * x86-tune.def (X86_TUNE_USE_GATHER): New.
19128 2018-01-10 Martin Liska <mliska@suse.cz>
19131 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
19132 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
19134 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
19135 CLEANUP_NO_PARTITIONING is not set.
19137 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
19139 * doc/rtl.texi: Remove documentation of (const ...) wrappers
19140 for vectors, as a partial revert of r254296.
19141 * rtl.h (const_vec_p): Delete.
19142 (const_vec_duplicate_p): Don't test for vector CONSTs.
19143 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
19144 * expmed.c (make_tree): Likewise.
19147 * common.md (E, F): Use CONSTANT_P instead of checking for
19149 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
19150 checking for CONST_VECTOR.
19152 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
19154 PR middle-end/83575
19155 * predict.c (force_edge_cold): Handle in more sane way edges
19156 with no prediction.
19158 2018-01-09 Carl Love <cel@us.ibm.com>
19160 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
19162 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
19163 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
19164 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
19165 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
19166 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
19167 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
19168 * config/rs6000/rs6000-protos.h: Add extern defition for
19169 rs6000_generate_float2_double_code.
19170 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
19172 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
19173 (float2_v2df): Add define_expand.
19175 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
19178 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
19179 op_mode in the force_to_mode call.
19181 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
19183 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
19184 instead of checking each element individually.
19185 (aarch64_evpc_uzp): Likewise.
19186 (aarch64_evpc_zip): Likewise.
19187 (aarch64_evpc_ext): Likewise.
19188 (aarch64_evpc_rev): Likewise.
19189 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
19190 instead of checking each element individually. Return true without
19192 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
19193 whether all selected elements come from the same input, instead of
19194 checking each element individually. Remove calls to gen_rtx_REG,
19195 start_sequence and end_sequence and instead assert that no rtl is
19198 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
19200 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
19201 order of HIGH and CONST checks.
19203 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
19205 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
19206 if the destination isn't an SSA_NAME.
19208 2018-01-09 Richard Biener <rguenther@suse.de>
19210 PR tree-optimization/83668
19211 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
19213 (canonicalize_loop_form): ... here, renamed from ...
19214 (canonicalize_loop_closed_ssa_form): ... this and amended to
19215 swap successor edges for loop exit blocks to make us use
19216 the RPO order we need for initial schedule generation.
19218 2018-01-09 Joseph Myers <joseph@codesourcery.com>
19220 PR tree-optimization/64811
19221 * match.pd: When optimizing comparisons with Inf, avoid
19222 introducing or losing exceptions from comparisons with NaN.
19224 2018-01-09 Martin Liska <mliska@suse.cz>
19227 * asan.c (shadow_mem_size): Add gcc_assert.
19229 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
19231 Don't save registers in main().
19234 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
19235 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
19236 * config/avr/avr.c (avr_set_current_function): Don't error if
19237 naked, OS_task or OS_main are specified at the same time.
19238 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
19240 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
19242 * common/config/avr/avr-common.c (avr_option_optimization_table):
19243 Switch on -mmain-is-OS_task for optimizing compilations.
19245 2018-01-09 Richard Biener <rguenther@suse.de>
19247 PR tree-optimization/83572
19248 * graphite.c: Include cfganal.h.
19249 (graphite_transform_loops): Connect infinite loops to exit
19250 and remove fake edges at the end.
19252 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
19254 * ipa-inline.c (edge_badness): Revert accidental checkin.
19256 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
19259 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
19260 symbols; not inline clones.
19262 2018-01-09 Jakub Jelinek <jakub@redhat.com>
19265 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
19266 hard registers. Formatting fixes.
19268 PR preprocessor/83722
19269 * gcc.c (try_generate_repro): Pass
19270 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
19271 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
19274 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
19275 Kito Cheng <kito.cheng@gmail.com>
19277 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
19278 (riscv_leaf_function_p): Delete.
19279 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
19281 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19283 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
19285 (do_ifelse): New function.
19286 (do_isel): New function.
19287 (do_sub3): New function.
19288 (do_add3): New function.
19289 (do_load_mask_compare): New function.
19290 (do_overlap_load_compare): New function.
19291 (expand_compare_loop): New function.
19292 (expand_block_compare): Call expand_compare_loop() when appropriate.
19293 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
19294 option description.
19295 (-mblock-compare-inline-loop-limit): New option.
19297 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19300 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
19301 Reverse order of second and third operands in first alternative.
19302 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
19303 of first and second elements in UNSPEC_VPERMR vector.
19304 (altivec_expand_vec_perm_le): Likewise.
19306 2018-01-08 Jeff Law <law@redhat.com>
19308 PR rtl-optimizatin/81308
19309 * tree-switch-conversion.c (cfg_altered): New file scoped static.
19310 (process_switch): If group_case_labels makes a change, then set
19312 (pass_convert_switch::execute): If a switch is converted, then
19313 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
19315 PR rtl-optimization/81308
19316 * recog.c (split_all_insns): Conditionally cleanup the CFG after
19319 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
19321 PR target/83663 - Revert r255946
19322 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
19323 generation for cases where splatting a value is not useful.
19324 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
19325 across a vec_duplicate and a paradoxical subreg forming a vector
19326 mode to a vec_concat.
19328 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19330 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
19331 -march=armv8.3-a variants.
19332 * config/arm/t-multilib: Likewise.
19333 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
19335 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19337 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
19339 (cceq_ior_compare_complement): Give it a name so I can use it, and
19340 change boolean_or_operator predicate to boolean_operator so it can
19341 be used to generate a crand.
19342 (eqne): New code iterator.
19343 (bd/bd_neg): New code_attrs.
19344 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
19345 a single define_insn.
19346 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
19347 decrement (bdnzt/bdnzf/bdzt/bdzf).
19348 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
19349 with the new names of the branch decrement patterns, and added the
19350 names of the branch decrement conditional patterns.
19352 2018-01-08 Richard Biener <rguenther@suse.de>
19354 PR tree-optimization/83563
19355 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
19358 2018-01-08 Richard Biener <rguenther@suse.de>
19360 PR middle-end/83713
19361 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
19363 2018-01-08 Richard Biener <rguenther@suse.de>
19365 PR tree-optimization/83685
19366 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
19367 references to abnormals.
19369 2018-01-08 Richard Biener <rguenther@suse.de>
19372 * dwarf2out.c (output_indirect_strings): Handle empty
19373 skeleton_debug_str_hash.
19374 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
19376 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
19378 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
19379 (emit_store_direct): Likewise.
19380 (arc_trampoline_adjust_address): Likewise.
19381 (arc_asm_trampoline_template): New function.
19382 (arc_initialize_trampoline): Use asm_trampoline_template.
19383 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
19384 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
19385 * config/arc/arc.md (flush_icache): Delete pattern.
19387 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
19389 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
19390 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
19393 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
19396 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
19397 by not USED_FOR_TARGET.
19398 (make_pass_resolve_sw_modes): Likewise.
19400 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
19402 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
19405 2018-01-08 Richard Biener <rguenther@suse.de>
19407 PR middle-end/83580
19408 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
19410 2018-01-08 Richard Biener <rguenther@suse.de>
19412 PR middle-end/83517
19413 * match.pd ((t * 2) / 2) -> t): Add missing :c.
19415 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
19417 PR middle-end/81897
19418 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
19419 basic blocks with a small number of successors.
19420 (convert_control_dep_chain_into_preds): Improve handling of
19422 (dump_predicates): Split apart into...
19423 (dump_pred_chain): ...here...
19424 (dump_pred_info): ...and here.
19425 (can_one_predicate_be_invalidated_p): Add debugging printfs.
19426 (can_chain_union_be_invalidated_p): Improve check for invalidation
19428 (uninit_uses_cannot_happen): Avoid unnecessary if
19429 convert_control_dep_chain_into_preds yielded nothing.
19431 2018-01-06 Martin Sebor <msebor@redhat.com>
19433 PR tree-optimization/83640
19434 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
19435 subtracting negative offset from size.
19436 (builtin_access::overlap): Adjust offset bounds of the access to fall
19437 within the size of the object if possible.
19439 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
19441 PR rtl-optimization/83699
19442 * expmed.c (extract_bit_field_1): Restrict the vector usage of
19443 extract_bit_field_as_subreg to cases in which the extracted
19444 value is also a vector.
19446 * lra-constraints.c (process_alt_operands): Test for the equivalence
19447 substitutions when detecting a possible reload cycle.
19449 2018-01-06 Jakub Jelinek <jakub@redhat.com>
19452 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
19453 by default if flag_selective_schedling{,2}. Formatting fixes.
19455 PR rtl-optimization/83682
19456 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
19457 if it has non-VECTOR_MODE element mode.
19458 (vec_duplicate_p): Likewise.
19460 PR middle-end/83694
19461 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
19462 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
19464 2018-01-05 Jakub Jelinek <jakub@redhat.com>
19467 * config/i386/i386-builtin.def
19468 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
19469 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
19470 Require also OPTION_MASK_ISA_AVX512F in addition to
19471 OPTION_MASK_ISA_GFNI.
19472 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
19473 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
19474 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
19475 to OPTION_MASK_ISA_GFNI.
19476 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
19477 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
19478 OPTION_MASK_ISA_AVX512BW.
19479 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
19480 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
19481 addition to OPTION_MASK_ISA_GFNI.
19482 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
19483 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
19484 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
19485 to OPTION_MASK_ISA_GFNI.
19486 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
19487 a requirement for all ISAs rather than any of them with a few
19489 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
19491 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
19492 bitmasks to be enabled with 3 exceptions, instead of requiring any
19493 enabled ISA with lots of exceptions.
19494 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
19495 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
19496 Change avx512bw in isa attribute to avx512f.
19497 * config/i386/sgxintrin.h: Add license boilerplate.
19498 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
19499 to __AVX512F__ and __AVX512VL to __AVX512VL__.
19500 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
19501 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
19503 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
19504 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
19505 temporarily sse2 rather than sse if not enabled already.
19508 * config/i386/sse.md (VI248_VLBW): Rename to ...
19509 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
19510 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
19511 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
19512 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
19513 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
19514 mode iterator instead of VI248_VLBW.
19516 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
19518 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
19519 (record_modified): Skip clobbers; add debug output.
19520 (param_change_prob): Use sreal frequencies.
19522 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
19524 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
19525 punt for user-aligned variables.
19527 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
19529 * tree-chrec.c (chrec_contains_symbols): Return true for
19532 2018-01-05 Sudakshina Das <sudi.das@arm.com>
19535 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
19536 of (x|y) == x for BICS pattern.
19538 2018-01-05 Jakub Jelinek <jakub@redhat.com>
19540 PR tree-optimization/83605
19541 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
19542 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
19545 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
19547 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
19548 * config/epiphany/rtems.h: New file.
19550 2018-01-04 Jakub Jelinek <jakub@redhat.com>
19551 Uros Bizjak <ubizjak@gmail.com>
19554 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
19555 QIreg_operand instead of register_operand predicate.
19556 * config/i386/i386.c (ix86_rop_should_change_byte_p,
19557 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
19558 comments instead of -fmitigate[-_]rop.
19560 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
19563 * cgraphunit.c (symbol_table::compile): Switch to text_section
19564 before calling assembly_start debug hook.
19565 * run-rtl-passes.c (run_rtl_passes): Likewise.
19568 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19570 * tree-vrp.c (extract_range_from_binary_expr_1): Check
19571 range_int_cst_p rather than !symbolic_range_p before calling
19572 extract_range_from_multiplicative_op_1.
19574 2018-01-04 Jeff Law <law@redhat.com>
19576 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
19577 redundant test in assertion.
19579 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19581 * doc/rtl.texi: Document machine_mode wrapper classes.
19583 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19585 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
19586 using tree_to_uhwi.
19588 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19590 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
19591 the VEC_PERM_EXPR fold to fail.
19593 2018-01-04 Jakub Jelinek <jakub@redhat.com>
19596 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
19597 to switched_sections.
19599 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19602 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
19603 test for d.testing.
19605 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
19608 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
19609 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
19611 2018-01-04 Jakub Jelinek <jakub@redhat.com>
19614 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
19615 is BLKmode and bitpos not zero or mode change is needed.
19617 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19620 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
19623 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
19626 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
19627 instead of MULT rtx. Update all corresponding splitters.
19628 (*saddl_se): Ditto.
19629 (*ssub<modesuffix>): Ditto.
19630 (*ssubl_se): Ditto.
19631 (*cmp_sadd_di): Update split patterns.
19632 (*cmp_sadd_si): Ditto.
19633 (*cmp_sadd_sidi): Ditto.
19634 (*cmp_ssub_di): Ditto.
19635 (*cmp_ssub_si): Ditto.
19636 (*cmp_ssub_sidi): Ditto.
19637 * config/alpha/predicates.md (const23_operand): New predicate.
19638 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
19639 Look for ASHIFT, not MULT inner operand.
19640 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
19642 2018-01-04 Martin Liska <mliska@suse.cz>
19644 PR gcov-profile/83669
19645 * gcov.c (output_intermediate_file): Add version to intermediate
19647 * doc/gcov.texi: Document new field 'version' in intermediate
19648 file format. Fix location of '-k' option of gcov command.
19650 2018-01-04 Martin Liska <mliska@suse.cz>
19653 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
19655 2018-01-04 Jakub Jelinek <jakub@redhat.com>
19657 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
19659 2018-01-03 Martin Sebor <msebor@redhat.com>
19661 PR tree-optimization/83655
19662 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
19663 checking calls with invalid arguments.
19665 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19667 * tree-vect-stmts.c (vect_get_store_rhs): New function.
19668 (vectorizable_mask_load_store): Delete.
19669 (vectorizable_call): Return false for masked loads and stores.
19670 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
19671 instead of gimple_assign_rhs1.
19672 (vectorizable_load): Handle IFN_MASK_LOAD.
19673 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
19675 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19677 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
19679 (vectorizable_mask_load_store): ...here.
19680 (vectorizable_load): ...and here.
19682 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19684 * tree-vect-stmts.c (vect_build_all_ones_mask)
19685 (vect_build_zero_merge_argument): New functions, split out from...
19686 (vectorizable_load): ...here.
19688 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19690 * tree-vect-stmts.c (vect_check_store_rhs): New function,
19692 (vectorizable_mask_load_store): ...here.
19693 (vectorizable_store): ...and here.
19695 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19697 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
19699 (vectorizable_mask_load_store): ...here.
19701 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19703 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
19704 (vect_model_store_cost): Take a vec_load_store_type instead of a
19706 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
19707 (vect_model_store_cost): Take a vec_load_store_type instead of a
19709 (vectorizable_mask_load_store): Update accordingly.
19710 (vectorizable_store): Likewise.
19711 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
19713 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19715 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
19716 IFN_MASK_LOAD calls here rather than...
19717 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
19719 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19720 Alan Hayward <alan.hayward@arm.com>
19721 David Sherwood <david.sherwood@arm.com>
19723 * expmed.c (extract_bit_field_1): For vector extracts,
19724 fall back to extract_bit_field_as_subreg if vec_extract
19727 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19728 Alan Hayward <alan.hayward@arm.com>
19729 David Sherwood <david.sherwood@arm.com>
19731 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
19732 they are variable or constant sized.
19733 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
19734 slots for constant-sized data.
19736 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19737 Alan Hayward <alan.hayward@arm.com>
19738 David Sherwood <david.sherwood@arm.com>
19740 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
19741 handling COND_EXPRs with boolean comparisons, try to find a better
19742 basis for the mask type than the boolean itself.
19744 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19746 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
19747 is calculated and how it can be overridden.
19748 * genmodes.c (max_bitsize_mode_any_mode): New variable.
19749 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
19751 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
19754 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19755 Alan Hayward <alan.hayward@arm.com>
19756 David Sherwood <david.sherwood@arm.com>
19758 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
19759 Remove the mode argument.
19760 (aarch64_simd_valid_immediate): Remove the mode and inverse
19762 * config/aarch64/iterators.md (bitsize): New iterator.
19763 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
19764 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
19765 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
19766 aarch64_simd_valid_immediate.
19767 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
19768 (aarch64_reg_or_bic_imm): Likewise.
19769 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
19770 with an insn_type enum and msl with a modifier_type enum.
19771 Replace element_width with a scalar_mode. Change the shift
19772 to unsigned int. Add constructors for scalar_float_mode and
19773 scalar_int_mode elements.
19774 (aarch64_vect_float_const_representable_p): Delete.
19775 (aarch64_can_const_movi_rtx_p)
19776 (aarch64_simd_scalar_immediate_valid_for_move)
19777 (aarch64_simd_make_constant): Update call to
19778 aarch64_simd_valid_immediate.
19779 (aarch64_advsimd_valid_immediate_hs): New function.
19780 (aarch64_advsimd_valid_immediate): Likewise.
19781 (aarch64_simd_valid_immediate): Remove mode and inverse
19782 arguments. Rewrite to use the above. Use const_vec_duplicate_p
19783 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
19784 and aarch64_float_const_representable_p on the result.
19785 (aarch64_output_simd_mov_immediate): Remove mode argument.
19786 Update call to aarch64_simd_valid_immediate and use of
19787 simd_immediate_info.
19788 (aarch64_output_scalar_simd_mov_immediate): Update call
19791 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19792 Alan Hayward <alan.hayward@arm.com>
19793 David Sherwood <david.sherwood@arm.com>
19795 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
19796 (mode_nunits): Likewise CONST_MODE_NUNITS.
19797 * machmode.def (ADJUST_NUNITS): Document.
19798 * genmodes.c (mode_data::need_nunits_adj): New field.
19799 (blank_mode): Update accordingly.
19800 (adj_nunits): New variable.
19801 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
19803 (emit_mode_size_inline): Set need_bytesize_adj for all modes
19804 listed in adj_nunits.
19805 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
19806 listed in adj_nunits. Don't emit case statements for such modes.
19807 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
19808 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
19809 nothing if adj_nunits is nonnull.
19810 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
19811 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
19812 (emit_mode_fbit): Update use of print_maybe_const_decl.
19813 (emit_move_size): Likewise. Treat the array as non-const
19815 (emit_mode_adjustments): Handle adj_nunits.
19817 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19819 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
19820 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
19821 (VECTOR_MODES): Use it.
19822 (make_vector_modes): Take the prefix as an argument.
19824 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19825 Alan Hayward <alan.hayward@arm.com>
19826 David Sherwood <david.sherwood@arm.com>
19828 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
19829 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
19830 for MODE_VECTOR_BOOL.
19831 * machmode.def (VECTOR_BOOL_MODE): Document.
19832 * genmodes.c (VECTOR_BOOL_MODE): New macro.
19833 (make_vector_bool_mode): New function.
19834 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
19836 * lto-streamer-in.c (lto_input_mode_table): Likewise.
19837 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
19839 * stor-layout.c (int_mode_for_mode): Likewise.
19840 * tree.c (build_vector_type_for_mode): Likewise.
19841 * varasm.c (output_constant_pool_2): Likewise.
19842 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
19843 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
19844 for MODE_VECTOR_BOOL.
19845 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
19846 of mode class checks.
19847 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
19848 instead of a list of mode class checks.
19849 (expand_vector_scalar_condition): Likewise.
19850 (type_for_widest_vector_mode): Handle BImode as an inner mode.
19852 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19853 Alan Hayward <alan.hayward@arm.com>
19854 David Sherwood <david.sherwood@arm.com>
19856 * machmode.h (mode_size): Change from unsigned short to
19858 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
19859 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
19860 or if measurement_type is not polynomial.
19861 (fixed_size_mode::includes_p): Check for constant-sized modes.
19862 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
19863 return a poly_uint16 rather than an unsigned short.
19864 (emit_mode_size): Change the type of mode_size from unsigned short
19865 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
19866 (emit_mode_adjustments): Cope with polynomial vector sizes.
19867 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
19869 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
19871 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
19872 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
19873 * caller-save.c (setup_save_areas): Likewise.
19874 (replace_reg_with_saved_mem): Likewise.
19875 * calls.c (emit_library_call_value_1): Likewise.
19876 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
19877 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
19878 (gen_lowpart_for_combine): Likewise.
19879 * convert.c (convert_to_integer_1): Likewise.
19880 * cse.c (equiv_constant, cse_insn): Likewise.
19881 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
19882 (cselib_subst_to_values): Likewise.
19883 * dce.c (word_dce_process_block): Likewise.
19884 * df-problems.c (df_word_lr_mark_ref): Likewise.
19885 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
19886 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
19887 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
19888 (rtl_for_decl_location): Likewise.
19889 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
19890 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
19891 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
19892 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
19893 (expand_expr_real_1): Likewise.
19894 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
19895 (pad_below): Likewise.
19896 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
19897 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
19898 * ira.c (get_subreg_tracking_sizes): Likewise.
19899 * ira-build.c (ira_create_allocno_objects): Likewise.
19900 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
19901 (ira_sort_regnos_for_alter_reg): Likewise.
19902 * ira-costs.c (record_operand_costs): Likewise.
19903 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
19904 (resolve_simple_move): Likewise.
19905 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
19906 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
19907 (lra_constraints): Likewise.
19908 (CONST_POOL_OK_P): Reject variable-sized modes.
19909 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
19910 (add_pseudo_to_slot, lra_spill): Likewise.
19911 * omp-low.c (omp_clause_aligned_alignment): Likewise.
19912 * optabs-query.c (get_best_extraction_insn): Likewise.
19913 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
19914 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
19915 (expand_mult_highpart, valid_multiword_target_p): Likewise.
19916 * recog.c (offsettable_address_addr_space_p): Likewise.
19917 * regcprop.c (maybe_mode_change): Likewise.
19918 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
19919 * regrename.c (build_def_use): Likewise.
19920 * regstat.c (dump_reg_info): Likewise.
19921 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
19922 (find_reloads, find_reloads_subreg_address): Likewise.
19923 * reload1.c (eliminate_regs_1): Likewise.
19924 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
19925 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
19926 (simplify_binary_operation_1, simplify_subreg): Likewise.
19927 * targhooks.c (default_function_arg_padding): Likewise.
19928 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
19929 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
19930 (verify_gimple_assign_ternary): Likewise.
19931 * tree-inline.c (estimate_move_cost): Likewise.
19932 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19933 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
19934 (get_address_cost_ainc): Likewise.
19935 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
19936 (vect_supportable_dr_alignment): Likewise.
19937 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
19938 (vectorizable_reduction): Likewise.
19939 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
19940 (vectorizable_operation, vectorizable_load): Likewise.
19941 * tree.c (build_same_sized_truth_vector_type): Likewise.
19942 * valtrack.c (cleanup_auto_inc_dec): Likewise.
19943 * var-tracking.c (emit_note_insn_var_location): Likewise.
19944 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
19945 (ADDR_VEC_ALIGN): Likewise.
19947 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19948 Alan Hayward <alan.hayward@arm.com>
19949 David Sherwood <david.sherwood@arm.com>
19951 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
19953 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
19954 or if measurement_type is polynomial.
19955 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
19956 * combine.c (make_extraction): Likewise.
19957 * dse.c (find_shift_sequence): Likewise.
19958 * dwarf2out.c (mem_loc_descriptor): Likewise.
19959 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
19960 (extract_bit_field, extract_low_bits): Likewise.
19961 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
19962 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
19963 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
19964 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
19965 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
19966 * reload.c (find_reloads): Likewise.
19967 * reload1.c (alter_reg): Likewise.
19968 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
19969 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
19970 * tree-if-conv.c (predicate_mem_writes): Likewise.
19971 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
19972 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
19973 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
19974 * valtrack.c (dead_debug_insert_temp): Likewise.
19975 * varasm.c (mergeable_constant_section): Likewise.
19976 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
19978 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19979 Alan Hayward <alan.hayward@arm.com>
19980 David Sherwood <david.sherwood@arm.com>
19982 * expr.c (expand_assignment): Cope with polynomial mode sizes
19983 when assigning to a CONCAT.
19985 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19986 Alan Hayward <alan.hayward@arm.com>
19987 David Sherwood <david.sherwood@arm.com>
19989 * machmode.h (mode_precision): Change from unsigned short to
19991 (mode_to_precision): Return a poly_uint16 rather than an unsigned
19993 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
19994 or if measurement_type is not polynomial.
19995 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
19996 in which the mode is already known to be a scalar_int_mode.
19997 * genmodes.c (emit_mode_precision): Change the type of mode_precision
19998 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
20000 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
20001 for GET_MODE_PRECISION.
20002 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
20003 for GET_MODE_PRECISION.
20004 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
20006 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
20007 (expand_field_assignment, make_extraction): Likewise.
20008 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
20009 (get_last_value): Likewise.
20010 * convert.c (convert_to_integer_1): Likewise.
20011 * cse.c (cse_insn): Likewise.
20012 * expr.c (expand_expr_real_1): Likewise.
20013 * lra-constraints.c (simplify_operand_subreg): Likewise.
20014 * optabs-query.c (can_atomic_load_p): Likewise.
20015 * optabs.c (expand_atomic_load): Likewise.
20016 (expand_atomic_store): Likewise.
20017 * ree.c (combine_reaching_defs): Likewise.
20018 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
20019 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
20020 * tree.h (type_has_mode_precision_p): Likewise.
20021 * ubsan.c (instrument_si_overflow): Likewise.
20023 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20024 Alan Hayward <alan.hayward@arm.com>
20025 David Sherwood <david.sherwood@arm.com>
20027 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
20028 polynomial numbers of units.
20029 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
20030 (valid_vector_subparts_p): New function.
20031 (build_vector_type): Remove temporary shim and take the number
20032 of units as a poly_uint64 rather than an int.
20033 (build_opaque_vector_type): Take the number of units as a
20034 poly_uint64 rather than an int.
20035 * tree.c (build_vector_from_ctor): Handle polynomial
20036 TYPE_VECTOR_SUBPARTS.
20037 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
20038 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
20039 (build_vector_from_val): If the number of units is variable,
20040 use build_vec_duplicate_cst for constant operands and
20041 VEC_DUPLICATE_EXPR otherwise.
20042 (make_vector_type): Remove temporary is_constant ().
20043 (build_vector_type, build_opaque_vector_type): Take the number of
20044 units as a poly_uint64 rather than an int.
20045 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
20047 * cfgexpand.c (expand_debug_expr): Likewise.
20048 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
20049 (store_constructor, expand_expr_real_1): Likewise.
20050 (const_scalar_mask_from_tree): Likewise.
20051 * fold-const-call.c (fold_const_reduction): Likewise.
20052 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
20053 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
20054 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
20055 (fold_relational_const): Likewise.
20056 (native_interpret_vector): Likewise. Change the size from an
20057 int to an unsigned int.
20058 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
20059 TYPE_VECTOR_SUBPARTS.
20060 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
20061 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
20062 duplicating a non-constant operand into a variable-length vector.
20063 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
20064 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
20065 * ipa-icf.c (sem_variable::equals): Likewise.
20066 * match.pd: Likewise.
20067 * omp-simd-clone.c (simd_clone_subparts): Likewise.
20068 * print-tree.c (print_node): Likewise.
20069 * stor-layout.c (layout_type): Likewise.
20070 * targhooks.c (default_builtin_vectorization_cost): Likewise.
20071 * tree-cfg.c (verify_gimple_comparison): Likewise.
20072 (verify_gimple_assign_binary): Likewise.
20073 (verify_gimple_assign_ternary): Likewise.
20074 (verify_gimple_assign_single): Likewise.
20075 * tree-pretty-print.c (dump_generic_node): Likewise.
20076 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
20077 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
20078 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
20079 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
20080 (vect_shift_permute_load_chain): Likewise.
20081 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
20082 (expand_vector_condition, optimize_vector_constructor): Likewise.
20083 (lower_vec_perm, get_compute_type): Likewise.
20084 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
20085 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
20086 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
20087 (vect_recog_mask_conversion_pattern): Likewise.
20088 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
20089 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
20090 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
20091 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
20092 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
20093 (vectorizable_shift, vectorizable_operation, vectorizable_store)
20094 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
20095 (supportable_widening_operation): Likewise.
20096 (supportable_narrowing_operation): Likewise.
20097 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
20099 * varasm.c (output_constant): Likewise.
20101 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20102 Alan Hayward <alan.hayward@arm.com>
20103 David Sherwood <david.sherwood@arm.com>
20105 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
20106 so that both the length == 3 and length != 3 cases set up their
20107 own permute vectors. Add comments explaining why we know the
20108 number of elements is constant.
20109 (vect_permute_load_chain): Likewise.
20111 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20112 Alan Hayward <alan.hayward@arm.com>
20113 David Sherwood <david.sherwood@arm.com>
20115 * machmode.h (mode_nunits): Change from unsigned char to
20117 (ONLY_FIXED_SIZE_MODES): New macro.
20118 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
20119 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
20120 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
20122 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
20123 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
20124 or if measurement_type is not polynomial.
20125 * genmodes.c (ZERO_COEFFS): New macro.
20126 (emit_mode_nunits_inline): Make mode_nunits_inline return a
20128 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
20129 Use ZERO_COEFFS when emitting initializers.
20130 * data-streamer.h (bp_pack_poly_value): New function.
20131 (bp_unpack_poly_value): Likewise.
20132 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
20133 for GET_MODE_NUNITS.
20134 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
20135 for GET_MODE_NUNITS.
20136 * tree.c (make_vector_type): Remove temporary shim and make
20137 the real function take the number of units as a poly_uint64
20138 rather than an int.
20139 (build_vector_type_for_mode): Handle polynomial nunits.
20140 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
20141 * emit-rtl.c (const_vec_series_p_1): Likewise.
20142 (gen_rtx_CONST_VECTOR): Likewise.
20143 * fold-const.c (test_vec_duplicate_folding): Likewise.
20144 * genrecog.c (validate_pattern): Likewise.
20145 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
20146 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
20147 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
20148 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
20149 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
20150 * rtlanal.c (subreg_get_info): Likewise.
20151 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
20152 (vect_grouped_load_supported): Likewise.
20153 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
20154 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
20155 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
20156 (simplify_const_unary_operation, simplify_binary_operation_1)
20157 (simplify_const_binary_operation, simplify_ternary_operation)
20158 (test_vector_ops_duplicate, test_vector_ops): Likewise.
20159 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
20160 instead of CONST_VECTOR_NUNITS.
20161 * varasm.c (output_constant_pool_2): Likewise.
20162 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
20163 explicit-encoded elements in the XVEC for variable-length vectors.
20165 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20167 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
20169 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20170 Alan Hayward <alan.hayward@arm.com>
20171 David Sherwood <david.sherwood@arm.com>
20173 * coretypes.h (fixed_size_mode): Declare.
20174 (fixed_size_mode_pod): New typedef.
20175 * builtins.h (target_builtins::x_apply_args_mode)
20176 (target_builtins::x_apply_result_mode): Change type to
20177 fixed_size_mode_pod.
20178 * builtins.c (apply_args_size, apply_result_size, result_vector)
20179 (expand_builtin_apply_args_1, expand_builtin_apply)
20180 (expand_builtin_return): Update accordingly.
20182 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20184 * cse.c (hash_rtx_cb): Hash only the encoded elements.
20185 * cselib.c (cselib_hash_rtx): Likewise.
20186 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
20187 CONST_VECTOR encoding.
20189 2018-01-03 Jakub Jelinek <jakub@redhat.com>
20190 Jeff Law <law@redhat.com>
20193 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
20194 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
20195 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
20196 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
20199 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
20200 explicitly probe *sp in a noreturn function if there were any callee
20201 register saves or frame pointer is needed.
20203 2018-01-03 Jakub Jelinek <jakub@redhat.com>
20206 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
20207 BLKmode for ternary, binary or unary expressions.
20210 * var-tracking.c (delete_vta_debug_insn): New inline function.
20211 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
20212 insns from get_insns () to NULL instead of each bb separately.
20213 Use delete_vta_debug_insn. No longer static.
20214 (vt_debug_insns_local, variable_tracking_main_1): Adjust
20215 delete_vta_debug_insns callers.
20216 * rtl.h (delete_vta_debug_insns): Declare.
20217 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
20218 instead of variable_tracking_main.
20220 2018-01-03 Martin Sebor <msebor@redhat.com>
20222 PR tree-optimization/83603
20223 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
20224 arguments past the endof the argument list in functions declared
20225 without a prototype.
20226 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
20227 Avoid checking when arguments are null.
20229 2018-01-03 Martin Sebor <msebor@redhat.com>
20232 * doc/extend.texi (attribute const): Fix a typo.
20233 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
20234 issuing -Wsuggest-attribute for void functions.
20236 2018-01-03 Martin Sebor <msebor@redhat.com>
20238 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
20239 offset_int::from instead of wide_int::to_shwi.
20240 (maybe_diag_overlap): Remove assertion.
20241 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
20242 * gimple-ssa-sprintf.c (format_directive): Same.
20243 (parse_directive): Same.
20244 (sprintf_dom_walker::compute_format_length): Same.
20245 (try_substitute_return_value): Same.
20247 2018-01-03 Jeff Law <law@redhat.com>
20249 PR middle-end/83654
20250 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
20251 non-constant residual for zero at runtime and avoid probing in
20252 that case. Reorganize code for trailing problem to mirror handling
20255 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
20257 PR tree-optimization/83501
20258 * tree-ssa-strlen.c (get_string_cst): New.
20259 (handle_char_store): Call get_string_cst.
20261 2018-01-03 Martin Liska <mliska@suse.cz>
20263 PR tree-optimization/83593
20264 * tree-ssa-strlen.c: Include tree-cfg.h.
20265 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
20266 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
20267 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
20269 (strlen_dom_walker::before_dom_children): Call
20270 gimple_purge_dead_eh_edges. Dump tranformation with details
20272 (strlen_dom_walker::before_dom_children): Update call by adding
20273 new argument cleanup_eh.
20274 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
20276 2018-01-03 Martin Liska <mliska@suse.cz>
20279 * cif-code.def (VARIADIC_THUNK): New enum value.
20280 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
20283 2018-01-03 Jan Beulich <jbeulich@suse.com>
20285 * sse.md (mov<mode>_internal): Tighten condition for when to use
20286 vmovdqu<ssescalarsize> for TI and OI modes.
20288 2018-01-03 Jakub Jelinek <jakub@redhat.com>
20290 Update copyright years.
20292 2018-01-03 Martin Liska <mliska@suse.cz>
20295 * ipa-visibility.c (function_and_variable_visibility): Skip
20296 functions with noipa attribure.
20298 2018-01-03 Jakub Jelinek <jakub@redhat.com>
20300 * gcc.c (process_command): Update copyright notice dates.
20301 * gcov-dump.c (print_version): Ditto.
20302 * gcov.c (print_version): Ditto.
20303 * gcov-tool.c (print_version): Ditto.
20304 * gengtype.c (create_file): Ditto.
20305 * doc/cpp.texi: Bump @copying's copyright year.
20306 * doc/cppinternals.texi: Ditto.
20307 * doc/gcc.texi: Ditto.
20308 * doc/gccint.texi: Ditto.
20309 * doc/gcov.texi: Ditto.
20310 * doc/install.texi: Ditto.
20311 * doc/invoke.texi: Ditto.
20313 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20315 * vector-builder.h (vector_builder::m_full_nelts): Change from
20316 unsigned int to poly_uint64.
20317 (vector_builder::full_nelts): Update prototype accordingly.
20318 (vector_builder::new_vector): Likewise.
20319 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
20320 (vector_builder::operator ==): Likewise.
20321 (vector_builder::finalize): Likewise.
20322 * int-vector-builder.h (int_vector_builder::int_vector_builder):
20323 Take the number of elements as a poly_uint64 rather than an
20325 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
20326 from unsigned int to poly_uint64.
20327 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
20328 (vec_perm_indices::new_vector): Likewise.
20329 (vec_perm_indices::length): Likewise.
20330 (vec_perm_indices::nelts_per_input): Likewise.
20331 (vec_perm_indices::input_nelts): Likewise.
20332 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
20333 number of elements per input as a poly_uint64 rather than an
20334 unsigned int. Use the original encoding for variable-length
20335 vectors, rather than clamping each individual element.
20336 For the second and subsequent elements in each pattern,
20337 clamp the step and base before clamping their sum.
20338 (vec_perm_indices::series_p): Handle polynomial element counts.
20339 (vec_perm_indices::all_in_range_p): Likewise.
20340 (vec_perm_indices_to_tree): Likewise.
20341 (vec_perm_indices_to_rtx): Likewise.
20342 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
20343 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
20344 (tree_vector_builder::new_binary_operation): Handle polynomial
20345 element counts. Return false if we need to know the number
20346 of elements at compile time.
20347 * fold-const.c (fold_vec_perm): Punt if the number of elements
20348 isn't known at compile time.
20350 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20352 * vec-perm-indices.h (vec_perm_builder): Change element type
20353 from HOST_WIDE_INT to poly_int64.
20354 (vec_perm_indices::element_type): Update accordingly.
20355 (vec_perm_indices::clamp): Handle polynomial element_types.
20356 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
20357 (vec_perm_indices::all_in_range_p): Likewise.
20358 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
20360 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
20361 polynomial vec_perm_indices element types.
20362 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
20363 * fold-const.c (fold_vec_perm): Likewise.
20364 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
20365 * tree-vect-generic.c (lower_vec_perm): Likewise.
20366 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
20367 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
20368 element type to HOST_WIDE_INT.
20370 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20371 Alan Hayward <alan.hayward@arm.com>
20372 David Sherwood <david.sherwood@arm.com>
20374 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
20375 rather than an int. Use plus_constant.
20376 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
20377 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
20379 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20380 Alan Hayward <alan.hayward@arm.com>
20381 David Sherwood <david.sherwood@arm.com>
20383 * calls.c (emit_call_1, expand_call): Change struct_value_size from
20384 a HOST_WIDE_INT to a poly_int64.
20386 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20387 Alan Hayward <alan.hayward@arm.com>
20388 David Sherwood <david.sherwood@arm.com>
20390 * calls.c (load_register_parameters): Cope with polynomial
20391 mode sizes. Require a constant size for BLKmode parameters
20392 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
20393 forces a parameter to be padded at the lsb end in order to
20394 fill a complete number of words, require the parameter size
20395 to be ordered wrt UNITS_PER_WORD.
20397 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20398 Alan Hayward <alan.hayward@arm.com>
20399 David Sherwood <david.sherwood@arm.com>
20401 * reload1.c (spill_stack_slot_width): Change element type
20402 from unsigned int to poly_uint64_pod.
20403 (alter_reg): Treat mode sizes as polynomial.
20405 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20406 Alan Hayward <alan.hayward@arm.com>
20407 David Sherwood <david.sherwood@arm.com>
20409 * reload.c (complex_word_subreg_p): New function.
20410 (reload_inner_reg_of_subreg, push_reload): Use it.
20412 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20413 Alan Hayward <alan.hayward@arm.com>
20414 David Sherwood <david.sherwood@arm.com>
20416 * lra-constraints.c (process_alt_operands): Reject matched
20417 operands whose sizes aren't ordered.
20418 (match_reload): Refer to this check here.
20420 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20421 Alan Hayward <alan.hayward@arm.com>
20422 David Sherwood <david.sherwood@arm.com>
20424 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
20425 that the mode size is in the set {1, 2, 4, 8, 16}.
20427 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20428 Alan Hayward <alan.hayward@arm.com>
20429 David Sherwood <david.sherwood@arm.com>
20431 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
20432 Use plus_constant instead of gen_rtx_PLUS.
20434 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20435 Alan Hayward <alan.hayward@arm.com>
20436 David Sherwood <david.sherwood@arm.com>
20438 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
20439 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
20440 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
20441 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
20442 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
20443 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
20444 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
20445 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
20446 * config/i386/i386.c (ix86_push_rounding): ...this new function.
20447 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
20449 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
20450 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
20451 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
20452 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
20453 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
20454 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
20455 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
20456 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
20457 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
20458 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
20460 * expr.c (emit_move_resolve_push): Treat the input and result
20461 of PUSH_ROUNDING as a poly_int64.
20462 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
20463 (emit_push_insn): Likewise.
20464 * lra-eliminations.c (mark_not_eliminable): Likewise.
20465 * recog.c (push_operand): Likewise.
20466 * reload1.c (elimination_effects): Likewise.
20467 * rtlanal.c (nonzero_bits1): Likewise.
20468 * calls.c (store_one_arg): Likewise. Require the padding to be
20469 known at compile time.
20471 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20472 Alan Hayward <alan.hayward@arm.com>
20473 David Sherwood <david.sherwood@arm.com>
20475 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
20476 Use plus_constant instead of gen_rtx_PLUS.
20478 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20479 Alan Hayward <alan.hayward@arm.com>
20480 David Sherwood <david.sherwood@arm.com>
20482 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
20483 rather than an int.
20485 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20486 Alan Hayward <alan.hayward@arm.com>
20487 David Sherwood <david.sherwood@arm.com>
20489 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
20490 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
20491 via stack temporaries. Treat the mode size as polynomial too.
20493 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20494 Alan Hayward <alan.hayward@arm.com>
20495 David Sherwood <david.sherwood@arm.com>
20497 * expr.c (expand_expr_real_2): When handling conversions involving
20498 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
20499 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
20500 as a poly_uint64 too.
20502 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20503 Alan Hayward <alan.hayward@arm.com>
20504 David Sherwood <david.sherwood@arm.com>
20506 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
20508 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20509 Alan Hayward <alan.hayward@arm.com>
20510 David Sherwood <david.sherwood@arm.com>
20512 * combine.c (can_change_dest_mode): Handle polynomial
20513 REGMODE_NATURAL_SIZE.
20514 * expmed.c (store_bit_field_1): Likewise.
20515 * expr.c (store_constructor): Likewise.
20516 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
20517 and polynomial REGMODE_NATURAL_SIZE.
20518 (gen_lowpart_common): Likewise.
20519 * reginfo.c (record_subregs_of_mode): Likewise.
20520 * rtlanal.c (read_modify_subreg_p): Likewise.
20522 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20523 Alan Hayward <alan.hayward@arm.com>
20524 David Sherwood <david.sherwood@arm.com>
20526 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
20527 numbers of elements.
20529 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20530 Alan Hayward <alan.hayward@arm.com>
20531 David Sherwood <david.sherwood@arm.com>
20533 * match.pd: Cope with polynomial numbers of vector elements.
20535 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20536 Alan Hayward <alan.hayward@arm.com>
20537 David Sherwood <david.sherwood@arm.com>
20539 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
20540 in a POINTER_PLUS_EXPR.
20542 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20543 Alan Hayward <alan.hayward@arm.com>
20544 David Sherwood <david.sherwood@arm.com>
20546 * omp-simd-clone.c (simd_clone_subparts): New function.
20547 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
20548 (ipa_simd_modify_function_body): Likewise.
20550 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20551 Alan Hayward <alan.hayward@arm.com>
20552 David Sherwood <david.sherwood@arm.com>
20554 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
20555 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
20556 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
20557 (expand_vector_condition, vector_element): Likewise.
20558 (subparts_gt): New function.
20559 (get_compute_type): Use subparts_gt.
20560 (count_type_subparts): Delete.
20561 (expand_vector_operations_1): Use subparts_gt instead of
20562 count_type_subparts.
20564 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20565 Alan Hayward <alan.hayward@arm.com>
20566 David Sherwood <david.sherwood@arm.com>
20568 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
20569 (vect_compile_time_alias): ...this new function. Do the calculation
20570 on poly_ints rather than trees.
20571 (vect_prune_runtime_alias_test_list): Update call accordingly.
20573 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20574 Alan Hayward <alan.hayward@arm.com>
20575 David Sherwood <david.sherwood@arm.com>
20577 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
20579 (vect_schedule_slp_instance): Likewise.
20581 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20582 Alan Hayward <alan.hayward@arm.com>
20583 David Sherwood <david.sherwood@arm.com>
20585 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
20586 constant and extern definitions for variable-length vectors.
20587 (vect_get_constant_vectors): Note that the number of units
20588 is known to be constant.
20590 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20591 Alan Hayward <alan.hayward@arm.com>
20592 David Sherwood <david.sherwood@arm.com>
20594 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
20595 of units as polynomial. Choose between WIDE and NARROW based
20598 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20599 Alan Hayward <alan.hayward@arm.com>
20600 David Sherwood <david.sherwood@arm.com>
20602 * tree-vect-stmts.c (simd_clone_subparts): New function.
20603 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
20605 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20606 Alan Hayward <alan.hayward@arm.com>
20607 David Sherwood <david.sherwood@arm.com>
20609 * tree-vect-stmts.c (vectorizable_call): Treat the number of
20610 vectors as polynomial. Use build_index_vector for
20611 IFN_GOMP_SIMD_LANE.
20613 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20614 Alan Hayward <alan.hayward@arm.com>
20615 David Sherwood <david.sherwood@arm.com>
20617 * tree-vect-stmts.c (get_load_store_type): Treat the number of
20618 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
20619 for variable-length vectors.
20620 (vectorizable_mask_load_store): Treat the number of units as
20621 polynomial, asserting that it is constant if the condition has
20622 already been enforced.
20623 (vectorizable_store, vectorizable_load): Likewise.
20625 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20626 Alan Hayward <alan.hayward@arm.com>
20627 David Sherwood <david.sherwood@arm.com>
20629 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
20630 of units as polynomial. Punt if we can't tell at compile time
20631 which vector contains the final result.
20633 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20634 Alan Hayward <alan.hayward@arm.com>
20635 David Sherwood <david.sherwood@arm.com>
20637 * tree-vect-loop.c (vectorizable_induction): Treat the number
20638 of units as polynomial. Punt on SLP inductions. Use an integer
20639 VEC_SERIES_EXPR for variable-length integer reductions. Use a
20640 cast of such a series for variable-length floating-point
20643 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20644 Alan Hayward <alan.hayward@arm.com>
20645 David Sherwood <david.sherwood@arm.com>
20647 * tree.h (build_index_vector): Declare.
20648 * tree.c (build_index_vector): New function.
20649 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
20650 of units as polynomial, forcibly converting it to a constant if
20651 vectorizable_reduction has already enforced the condition.
20652 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
20653 to create a {1,2,3,...} vector.
20654 (vectorizable_reduction): Treat the number of units as polynomial.
20655 Choose vectype_in based on the largest scalar element size rather
20656 than the smallest number of units. Enforce the restrictions
20659 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20660 Alan Hayward <alan.hayward@arm.com>
20661 David Sherwood <david.sherwood@arm.com>
20663 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
20664 number of units as polynomial.
20666 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20667 Alan Hayward <alan.hayward@arm.com>
20668 David Sherwood <david.sherwood@arm.com>
20670 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
20671 * target.def (autovectorize_vector_sizes): Return the vector sizes
20672 by pointer, using vector_sizes rather than a bitmask.
20673 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
20674 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
20675 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
20677 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
20678 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
20679 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
20680 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
20681 * omp-general.c (omp_max_vf): Likewise.
20682 * omp-low.c (omp_clause_aligned_alignment): Likewise.
20683 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
20684 * tree-vect-loop.c (vect_analyze_loop): Likewise.
20685 * tree-vect-slp.c (vect_slp_bb): Likewise.
20686 * doc/tm.texi: Regenerate.
20687 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
20689 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
20690 the vector size as a poly_uint64 rather than an unsigned int.
20691 (current_vector_size): Change from an unsigned int to a poly_uint64.
20692 (get_vectype_for_scalar_type): Update accordingly.
20693 * tree.h (build_truth_vector_type): Take the size and number of
20694 units as a poly_uint64 rather than an unsigned int.
20695 (build_vector_type): Add a temporary overload that takes
20696 the number of units as a poly_uint64 rather than an unsigned int.
20697 * tree.c (make_vector_type): Likewise.
20698 (build_truth_vector_type): Take the number of units as a poly_uint64
20699 rather than an unsigned int.
20701 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20702 Alan Hayward <alan.hayward@arm.com>
20703 David Sherwood <david.sherwood@arm.com>
20705 * target.def (get_mask_mode): Take the number of units and length
20706 as poly_uint64s rather than unsigned ints.
20707 * targhooks.h (default_get_mask_mode): Update accordingly.
20708 * targhooks.c (default_get_mask_mode): Likewise.
20709 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
20710 * doc/tm.texi: Regenerate.
20712 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20713 Alan Hayward <alan.hayward@arm.com>
20714 David Sherwood <david.sherwood@arm.com>
20716 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
20717 * omp-general.c (omp_max_vf): Likewise.
20718 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
20719 (expand_omp_simd): Handle polynomial safelen.
20720 * omp-low.c (omplow_simd_context): Add a default constructor.
20721 (omplow_simd_context::max_vf): Change from int to poly_uint64.
20722 (lower_rec_simd_input_clauses): Update accordingly.
20723 (lower_rec_input_clauses): Likewise.
20725 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20726 Alan Hayward <alan.hayward@arm.com>
20727 David Sherwood <david.sherwood@arm.com>
20729 * tree-vectorizer.h (vect_nunits_for_cost): New function.
20730 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
20731 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
20732 (vect_analyze_slp_cost): Likewise.
20733 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
20734 (vect_model_load_cost): Likewise.
20736 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20737 Alan Hayward <alan.hayward@arm.com>
20738 David Sherwood <david.sherwood@arm.com>
20740 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
20741 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
20742 from an unsigned int * to a poly_uint64_pod *.
20743 (calculate_unrolling_factor): New function.
20744 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
20746 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20747 Alan Hayward <alan.hayward@arm.com>
20748 David Sherwood <david.sherwood@arm.com>
20750 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
20751 from an unsigned int to a poly_uint64.
20752 (_loop_vec_info::slp_unrolling_factor): Likewise.
20753 (_loop_vec_info::vectorization_factor): Change from an int
20755 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
20756 (vect_get_num_vectors): New function.
20757 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
20758 (vect_get_num_copies): Use vect_get_num_vectors.
20759 (vect_analyze_data_ref_dependences): Change max_vf from an int *
20760 to an unsigned int *.
20761 (vect_analyze_data_refs): Change min_vf from an int * to a
20763 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
20764 than an unsigned HOST_WIDE_INT.
20765 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
20766 (vect_analyze_data_ref_dependence): Change max_vf from an int *
20767 to an unsigned int *.
20768 (vect_analyze_data_ref_dependences): Likewise.
20769 (vect_compute_data_ref_alignment): Handle polynomial vf.
20770 (vect_enhance_data_refs_alignment): Likewise.
20771 (vect_prune_runtime_alias_test_list): Likewise.
20772 (vect_shift_permute_load_chain): Likewise.
20773 (vect_supportable_dr_alignment): Likewise.
20774 (dependence_distance_ge_vf): Take the vectorization factor as a
20775 poly_uint64 rather than an unsigned HOST_WIDE_INT.
20776 (vect_analyze_data_refs): Change min_vf from an int * to a
20778 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
20779 vfm1 as a poly_uint64 rather than an int. Make the same change
20780 for the returned bound_scalar.
20781 (vect_gen_vector_loop_niters): Handle polynomial vf.
20782 (vect_do_peeling): Likewise. Update call to
20783 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
20784 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
20786 * tree-vect-loop.c (vect_determine_vectorization_factor)
20787 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
20788 (vect_get_known_peeling_cost): Likewise.
20789 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
20790 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
20791 (vect_transform_loop): Likewise. Use the lowest possible VF when
20792 updating the upper bounds of the loop.
20793 (vect_min_worthwhile_factor): Make static. Return an unsigned int
20794 rather than an int.
20795 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
20796 polynomial unroll factors.
20797 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
20798 (vect_make_slp_decision): Likewise.
20799 (vect_supported_load_permutation_p): Likewise, and polynomial
20801 (vect_analyze_slp_cost): Handle polynomial vf.
20802 (vect_slp_analyze_node_operations): Likewise.
20803 (vect_slp_analyze_bb_1): Likewise.
20804 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
20805 than an unsigned HOST_WIDE_INT.
20806 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
20807 (vectorizable_load): Handle polynomial vf.
20808 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
20810 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
20812 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20813 Alan Hayward <alan.hayward@arm.com>
20814 David Sherwood <david.sherwood@arm.com>
20816 * match.pd: Handle bit operations involving three constants
20817 and try to fold one pair.
20819 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20821 * tree-vect-loop-manip.c: Include gimple-fold.h.
20822 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
20823 niters_maybe_zero parameters. Handle other cases besides a step of 1.
20824 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
20825 Add a path that uses a step of VF instead of 1, but disable it
20827 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
20828 and niters_no_overflow parameters. Update calls to
20829 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
20830 Create a new SSA name if the latter choses to use a ste other
20831 than zero, and return it via niters_vector_mult_vf_var.
20832 * tree-vect-loop.c (vect_transform_loop): Update calls to
20833 vect_do_peeling, vect_gen_vector_loop_niters and
20834 slpeel_make_loop_iterate_ntimes.
20835 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
20836 (vect_gen_vector_loop_niters): Update declarations after above changes.
20838 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
20840 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
20841 128-bit round to integer instructions.
20842 (ceil<mode>2): Likewise.
20843 (btrunc<mode>2): Likewise.
20844 (round<mode>2): Likewise.
20846 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
20848 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
20849 unaligned VSX load/store on P8/P9.
20850 (expand_block_clear): Allow the use of unaligned VSX
20851 load/store on P8/P9.
20853 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20855 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
20857 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
20858 swap associated with both a load and a store.
20860 2018-01-02 Andrew Waterman <andrew@sifive.com>
20862 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
20863 * config/riscv/riscv.md (clear_cache): Use it.
20865 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
20867 * web.c: Remove out-of-date comment.
20869 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20871 * expr.c (fixup_args_size_notes): Check that any existing
20872 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
20873 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
20874 (emit_single_push_insn): ...here.
20876 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20878 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
20879 (const_vector_encoded_nelts): New function.
20880 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
20881 (const_vector_int_elt, const_vector_elt): Declare.
20882 * emit-rtl.c (const_vector_int_elt_1): New function.
20883 (const_vector_elt): Likewise.
20884 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
20885 of CONST_VECTOR_ELT.
20887 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20889 * expr.c: Include rtx-vector-builder.h.
20890 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
20891 directly on the tree encoding.
20892 (const_vector_from_tree): Likewise.
20893 * optabs.c: Include rtx-vector-builder.h.
20894 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
20895 sequence of "u" values.
20896 * vec-perm-indices.c: Include rtx-vector-builder.h.
20897 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
20898 directly on the vec_perm_indices encoding.
20900 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20902 * doc/rtl.texi (const_vector): Describe new encoding scheme.
20903 * Makefile.in (OBJS): Add rtx-vector-builder.o.
20904 * rtx-vector-builder.h: New file.
20905 * rtx-vector-builder.c: Likewise.
20906 * rtl.h (rtx_def::u2): Add a const_vector field.
20907 (CONST_VECTOR_NPATTERNS): New macro.
20908 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
20909 (CONST_VECTOR_DUPLICATE_P): Likewise.
20910 (CONST_VECTOR_STEPPED_P): Likewise.
20911 (CONST_VECTOR_ENCODED_ELT): Likewise.
20912 (const_vec_duplicate_p): Check for a duplicated vector encoding.
20913 (unwrap_const_vec_duplicate): Likewise.
20914 (const_vec_series_p): Check for a non-duplicated vector encoding.
20915 Say that the function only returns true for integer vectors.
20916 * emit-rtl.c: Include rtx-vector-builder.h.
20917 (gen_const_vec_duplicate_1): Delete.
20918 (gen_const_vector): Call gen_const_vec_duplicate instead of
20919 gen_const_vec_duplicate_1.
20920 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
20921 (gen_const_vec_duplicate): Use rtx_vector_builder.
20922 (gen_const_vec_series): Likewise.
20923 (gen_rtx_CONST_VECTOR): Likewise.
20924 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
20925 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
20926 Build a new vector rather than modifying a CONST_VECTOR in-place.
20927 (handle_special_swappables): Update call accordingly.
20928 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
20929 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
20930 Build a new vector rather than modifying a CONST_VECTOR in-place.
20931 (handle_special_swappables): Update call accordingly.
20933 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20935 * simplify-rtx.c (simplify_const_binary_operation): Use
20936 CONST_VECTOR_ELT instead of XVECEXP.
20938 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20940 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
20941 the selector elements to be different from the data elements
20942 if the selector is a VECTOR_CST.
20943 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
20944 ssizetype for the selector.
20946 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20948 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
20949 before testing each element individually.
20950 * tree-vect-generic.c (lower_vec_perm): Likewise.
20952 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20954 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
20955 * selftest-run-tests.c (selftest::run_tests): Call it.
20956 * vector-builder.h (vector_builder::operator ==): New function.
20957 (vector_builder::operator !=): Likewise.
20958 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
20959 (vec_perm_indices::all_from_input_p): New function.
20960 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
20961 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
20962 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
20963 instead of reading the VECTOR_CST directly. Detect whether both
20964 vector inputs are the same before constructing the vec_perm_indices,
20965 and update the number of inputs argument accordingly. Use the
20966 utility functions added above. Only construct sel2 if we need to.
20968 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20970 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
20971 the broadcast of the low byte.
20972 (expand_mult_highpart): Use an explicit encoding for the permutes.
20973 * optabs-query.c (can_mult_highpart_p): Likewise.
20974 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
20975 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
20976 (vectorizable_bswap): Likewise.
20977 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
20978 explicit encoding for the power-of-2 permutes.
20979 (vect_permute_store_chain): Likewise.
20980 (vect_grouped_load_supported): Likewise.
20981 (vect_permute_load_chain): Likewise.
20983 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20985 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
20986 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
20987 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
20988 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
20989 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
20990 (vect_gen_perm_mask_any): Likewise.
20992 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20994 * int-vector-builder.h: New file.
20995 * vec-perm-indices.h: Include int-vector-builder.h.
20996 (vec_perm_indices): Redefine as an int_vector_builder.
20997 (auto_vec_perm_indices): Delete.
20998 (vec_perm_builder): Redefine as a stand-alone class.
20999 (vec_perm_indices::vec_perm_indices): New function.
21000 (vec_perm_indices::clamp): Likewise.
21001 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
21002 (vec_perm_indices::new_vector): New function.
21003 (vec_perm_indices::new_expanded_vector): Update for new
21004 vec_perm_indices class.
21005 (vec_perm_indices::rotate_inputs): New function.
21006 (vec_perm_indices::all_in_range_p): Operate directly on the
21007 encoded form, without computing elided elements.
21008 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
21009 encoding. Update for new vec_perm_indices class.
21010 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
21011 the given vec_perm_builder.
21012 (expand_vec_perm_var): Update vec_perm_builder constructor.
21013 (expand_mult_highpart): Use vec_perm_builder instead of
21014 auto_vec_perm_indices.
21015 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
21016 vec_perm_indices instead of auto_vec_perm_indices. Use a single
21017 or double series encoding as appropriate.
21018 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
21019 vec_perm_indices instead of auto_vec_perm_indices.
21020 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
21021 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
21022 (vect_permute_store_chain): Likewise.
21023 (vect_grouped_load_supported): Likewise.
21024 (vect_permute_load_chain): Likewise.
21025 (vect_shift_permute_load_chain): Likewise.
21026 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
21027 (vect_transform_slp_perm_load): Likewise.
21028 (vect_schedule_slp_instance): Likewise.
21029 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
21030 (vectorizable_mask_load_store): Likewise.
21031 (vectorizable_bswap): Likewise.
21032 (vectorizable_store): Likewise.
21033 (vectorizable_load): Likewise.
21034 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
21035 vec_perm_indices instead of auto_vec_perm_indices. Use
21036 tree_to_vec_perm_builder to read the vector from a tree.
21037 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
21038 vec_perm_builder instead of a vec_perm_indices.
21039 (have_whole_vector_shift): Use vec_perm_builder and
21040 vec_perm_indices instead of auto_vec_perm_indices. Leave the
21041 truncation to calc_vec_perm_mask_for_shift.
21042 (vect_create_epilog_for_reduction): Likewise.
21043 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
21044 from auto_vec_perm_indices to vec_perm_indices.
21045 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
21046 instead of changing individual elements.
21047 (aarch64_vectorize_vec_perm_const): Use new_vector to install
21048 the vector in d.perm.
21049 * config/arm/arm.c (expand_vec_perm_d::perm): Change
21050 from auto_vec_perm_indices to vec_perm_indices.
21051 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
21052 instead of changing individual elements.
21053 (arm_vectorize_vec_perm_const): Use new_vector to install
21054 the vector in d.perm.
21055 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
21056 Update vec_perm_builder constructor.
21057 (rs6000_expand_interleave): Likewise.
21058 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
21059 (rs6000_expand_interleave): Likewise.
21061 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21063 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
21064 to qimode could truncate the indices.
21065 * optabs.c (expand_vec_perm_var): Likewise.
21067 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21069 * Makefile.in (OBJS): Add vec-perm-indices.o.
21070 * vec-perm-indices.h: New file.
21071 * vec-perm-indices.c: Likewise.
21072 * target.h (vec_perm_indices): Replace with a forward class
21074 (auto_vec_perm_indices): Move to vec-perm-indices.h.
21075 * optabs.h: Include vec-perm-indices.h.
21076 (expand_vec_perm): Delete.
21077 (selector_fits_mode_p, expand_vec_perm_var): Declare.
21078 (expand_vec_perm_const): Declare.
21079 * target.def (vec_perm_const_ok): Replace with...
21080 (vec_perm_const): ...this new hook.
21081 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
21082 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
21083 * doc/tm.texi: Regenerate.
21084 * optabs.def (vec_perm_const): Delete.
21085 * doc/md.texi (vec_perm_const): Likewise.
21086 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
21087 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
21088 expand_vec_perm for constant permutation vectors. Assert that
21089 the mode of variable permutation vectors is the integer equivalent
21090 of the mode that is being permuted.
21091 * optabs-query.h (selector_fits_mode_p): Declare.
21092 * optabs-query.c: Include vec-perm-indices.h.
21093 (selector_fits_mode_p): New function.
21094 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
21095 is defined, instead of checking whether the vec_perm_const_optab
21096 exists. Use targetm.vectorize.vec_perm_const instead of
21097 targetm.vectorize.vec_perm_const_ok. Check whether the indices
21098 fit in the vector mode before using a variable permute.
21099 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
21100 vec_perm_indices instead of an rtx.
21101 (expand_vec_perm): Replace with...
21102 (expand_vec_perm_const): ...this new function. Take the selector
21103 as a vec_perm_indices rather than an rtx. Also take the mode of
21104 the selector. Update call to shift_amt_for_vec_perm_mask.
21105 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
21106 Use vec_perm_indices::new_expanded_vector to expand the original
21107 selector into bytes. Check whether the indices fit in the vector
21108 mode before using a variable permute.
21109 (expand_vec_perm_var): Make global.
21110 (expand_mult_highpart): Use expand_vec_perm_const.
21111 * fold-const.c: Includes vec-perm-indices.h.
21112 * tree-ssa-forwprop.c: Likewise.
21113 * tree-vect-data-refs.c: Likewise.
21114 * tree-vect-generic.c: Likewise.
21115 * tree-vect-loop.c: Likewise.
21116 * tree-vect-slp.c: Likewise.
21117 * tree-vect-stmts.c: Likewise.
21118 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
21120 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
21121 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
21122 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
21123 (aarch64_vectorize_vec_perm_const): ...this new function.
21124 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
21125 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
21126 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
21127 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
21128 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
21129 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
21130 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
21132 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
21133 check for NEON modes.
21134 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
21135 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
21136 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
21137 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
21139 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
21140 the old VEC_PERM_CONST conditions.
21141 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
21142 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
21143 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
21144 (ia64_vectorize_vec_perm_const_ok): Merge into...
21145 (ia64_vectorize_vec_perm_const): ...this new function.
21146 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
21147 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
21148 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
21149 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
21150 * config/mips/mips.c (mips_expand_vec_perm_const)
21151 (mips_vectorize_vec_perm_const_ok): Merge into...
21152 (mips_vectorize_vec_perm_const): ...this new function.
21153 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
21154 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
21155 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
21156 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
21157 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
21158 (rs6000_expand_vec_perm_const): Delete.
21159 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
21161 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
21162 (altivec_expand_vec_perm_const_le): Take each operand individually.
21163 Operate on constant selectors rather than rtxes.
21164 (altivec_expand_vec_perm_const): Likewise. Update call to
21165 altivec_expand_vec_perm_const_le.
21166 (rs6000_expand_vec_perm_const): Delete.
21167 (rs6000_vectorize_vec_perm_const_ok): Delete.
21168 (rs6000_vectorize_vec_perm_const): New function.
21169 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
21170 an element count and rtx array.
21171 (rs6000_expand_extract_even): Update call accordingly.
21172 (rs6000_expand_interleave): Likewise.
21173 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
21174 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
21175 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
21176 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
21177 (rs6000_expand_vec_perm_const): Delete.
21178 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
21179 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
21180 (altivec_expand_vec_perm_const_le): Take each operand individually.
21181 Operate on constant selectors rather than rtxes.
21182 (altivec_expand_vec_perm_const): Likewise. Update call to
21183 altivec_expand_vec_perm_const_le.
21184 (rs6000_expand_vec_perm_const): Delete.
21185 (rs6000_vectorize_vec_perm_const_ok): Delete.
21186 (rs6000_vectorize_vec_perm_const): New function. Remove stray
21187 reference to the SPE evmerge intructions.
21188 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
21189 an element count and rtx array.
21190 (rs6000_expand_extract_even): Update call accordingly.
21191 (rs6000_expand_interleave): Likewise.
21192 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
21193 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
21195 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
21197 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21199 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
21200 vector mode and that that mode matches the mode of the data
21202 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
21203 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
21204 directly using expand_vec_perm_1 when forcing selectors into
21206 (expand_vec_perm_var): New function, split out from expand_vec_perm.
21208 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21210 * optabs-query.h (can_vec_perm_p): Delete.
21211 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
21212 * optabs-query.c (can_vec_perm_p): Split into...
21213 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
21214 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
21215 particular selector is valid.
21216 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
21217 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
21218 (vect_grouped_load_supported): Likewise.
21219 (vect_shift_permute_load_chain): Likewise.
21220 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
21221 (vect_transform_slp_perm_load): Likewise.
21222 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
21223 (vectorizable_bswap): Likewise.
21224 (vect_gen_perm_mask_checked): Likewise.
21225 * fold-const.c (fold_ternary_loc): Likewise. Don't take
21226 implementations of variable permutation vectors into account
21227 when deciding which selector to use.
21228 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
21229 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
21230 with a false third argument.
21231 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
21232 to test whether the constant selector is valid and can_vec_perm_var_p
21233 to test whether a variable selector is valid.
21235 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21237 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
21238 * optabs-query.c (can_vec_perm_p): Likewise.
21239 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
21240 instead of vec_perm_indices.
21241 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
21242 (vect_gen_perm_mask_checked): Likewise,
21243 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
21244 (vect_gen_perm_mask_checked): Likewise,
21246 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21248 * optabs-query.h (qimode_for_vec_perm): Declare.
21249 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
21250 (qimode_for_vec_perm): ...this new function.
21251 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
21253 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
21255 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
21256 does not have a conditional at the top.
21258 2018-01-02 Richard Biener <rguenther@suse.de>
21260 * ipa-inline.c (big_speedup_p): Fix expression.
21262 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
21265 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
21268 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
21272 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
21273 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
21274 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
21275 cond_taken_branch_cost 3->4.
21277 2018-01-01 Jakub Jelinek <jakub@redhat.com>
21279 PR tree-optimization/83581
21280 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
21281 TODO_cleanup_cfg if any changes have been made.
21283 PR middle-end/83608
21284 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
21285 convert_modes if target mode has the right side, but different mode
21288 PR middle-end/83609
21289 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
21290 last argument when extracting from CONCAT. If either from_real or
21291 from_imag is NULL, use expansion through memory. If result is not
21292 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
21293 the parts directly to inner mode, if even that fails, use expansion
21296 PR middle-end/83623
21297 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
21298 check for bswap in mode rather than HImode and use that in expand_unop
21301 Copyright (C) 2018 Free Software Foundation, Inc.
21303 Copying and distribution of this file, with or without modification,
21304 are permitted in any medium without royalty provided the copyright
21305 notice and this notice are preserved.